<%BANNER%>

Essays on Latin American Infrastructure

Permanent Link: http://ufdc.ufl.edu/UFE0022423/00001

Material Information

Title: Essays on Latin American Infrastructure Empirical Studies of Sector Performance
Physical Description: 1 online resource (121 p.)
Language: english
Creator: Corton, Maria
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Economics -- Dissertations, Academic -- UF
Genre: Economics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This study examines the performance of utilities in the water sector of Central American countries and Peru. Performance is assessed as technical efficiency and it is measured utilizing different methodologies. Chapter 1 explains the methodologies utilized in the different performance analyses presented in Chapters 2, 3 and 4. Chapter 2 provides a comprehensive efficiency analysis of water service providers in six countries in the Central American region. Pressures for sector reform have stimulated interest in identifying and understanding the factors that can contribute to network expansion, improved service quality, and cost containment. The aim of the analysis is to provide policymakers and investment funds institutions with quantitative evidence on the effectiveness of the regional water sectors and utilities under different institutional arrangements. In addition to key sector performance indicators, the analysis considers total factor productivity indexes, a non-deterministic frontier and a stochastic cost frontier. Differences in results from these methodologies are due to the assumptions imposed to each model specification and data employed. The Central American performance indicators are compared to similar indicators for service providers in Latin America. From the group of countries analyzed in Central America, the water service provider from Panama is the best performer according to results from all the methodologies. Chapter 3 investigates the presence of economies of scale and technical efficiencies in the water and sanitation sector of Peru. Chapter 3 analyses the sector structure considering size and location of service providers. The aim is to address the merits of introducing additional suppliers in some regions and possible merging of small utilities in other regions. The analysis employs a stochastic cost frontier model. Overall, economies of scale are present in all firms in the forest region and in the small firms located on the coast. These findings support the aggregation process: consolidating some utilities could lower costs. Chapter 4 examines the impact on efficiency of water utilities providing service to multiple jurisdictions. The issue analyzed is whether the expansion of service across multiple jurisdictions is more efficient than expanding service within a single jurisdiction. In this context, a jurisdiction is a unit of government designed to carry out public functions within a specific territory. The hypothesis is that utilities answering to a heterogeneous group of jurisdictional authorities are less efficient than those reporting to a single authority. Political issues and bargaining for resources may contribute to production inefficiencies affecting firms costs. The resulting inefficiencies may offset any scale economies associated with serving a larger area. Results from the analysis support the stated hypothesis by finding less efficiency associated to utilities providing service to more than one province when they are serving in a department where there is more than one provider.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Maria Corton.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Berg, Sanford V.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022423:00001

Permanent Link: http://ufdc.ufl.edu/UFE0022423/00001

Material Information

Title: Essays on Latin American Infrastructure Empirical Studies of Sector Performance
Physical Description: 1 online resource (121 p.)
Language: english
Creator: Corton, Maria
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Economics -- Dissertations, Academic -- UF
Genre: Economics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This study examines the performance of utilities in the water sector of Central American countries and Peru. Performance is assessed as technical efficiency and it is measured utilizing different methodologies. Chapter 1 explains the methodologies utilized in the different performance analyses presented in Chapters 2, 3 and 4. Chapter 2 provides a comprehensive efficiency analysis of water service providers in six countries in the Central American region. Pressures for sector reform have stimulated interest in identifying and understanding the factors that can contribute to network expansion, improved service quality, and cost containment. The aim of the analysis is to provide policymakers and investment funds institutions with quantitative evidence on the effectiveness of the regional water sectors and utilities under different institutional arrangements. In addition to key sector performance indicators, the analysis considers total factor productivity indexes, a non-deterministic frontier and a stochastic cost frontier. Differences in results from these methodologies are due to the assumptions imposed to each model specification and data employed. The Central American performance indicators are compared to similar indicators for service providers in Latin America. From the group of countries analyzed in Central America, the water service provider from Panama is the best performer according to results from all the methodologies. Chapter 3 investigates the presence of economies of scale and technical efficiencies in the water and sanitation sector of Peru. Chapter 3 analyses the sector structure considering size and location of service providers. The aim is to address the merits of introducing additional suppliers in some regions and possible merging of small utilities in other regions. The analysis employs a stochastic cost frontier model. Overall, economies of scale are present in all firms in the forest region and in the small firms located on the coast. These findings support the aggregation process: consolidating some utilities could lower costs. Chapter 4 examines the impact on efficiency of water utilities providing service to multiple jurisdictions. The issue analyzed is whether the expansion of service across multiple jurisdictions is more efficient than expanding service within a single jurisdiction. In this context, a jurisdiction is a unit of government designed to carry out public functions within a specific territory. The hypothesis is that utilities answering to a heterogeneous group of jurisdictional authorities are less efficient than those reporting to a single authority. Political issues and bargaining for resources may contribute to production inefficiencies affecting firms costs. The resulting inefficiencies may offset any scale economies associated with serving a larger area. Results from the analysis support the stated hypothesis by finding less efficiency associated to utilities providing service to more than one province when they are serving in a department where there is more than one provider.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Maria Corton.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Berg, Sanford V.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022423:00001


This item has the following downloads:


Full Text





ESSAYS ON LATIN AMERICAN INFRASTRUCTURE:
EMPIRICAL STUDIES OF SECTOR PERFORMANCE





















By

MARIA LUISA CORTON


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2008




































2008 Maria Luisa Corton


































To Aldo Alejandro









ACKNOWLEDGMENTS

This work would not have been possible without the support of Sanford Berg. His

support, insight and direction were invaluable to the completion of my dissertation. I am also

thankful for my involvement with the PURC team. They have supported me as a family

throughout this process. In particular, I would like to thank Cynthia for reminding me that there

was an end to the tunnel all along the way. Also, I am very appreciative for Steven Slutsky's

thoughtful comments and for letting me in to the grad group every other year. I am very grateful

to Jonathan Hamilton, for his help and insights, particularly in the final stage of this project. It

was an honor to work with my committee. Each member has introduced me to a different facet

of economics. Finally, I am grateful to the Department of Economics and PURC for their

flexibility and support during this long journey.









TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ..............................................................................................................4

LIST OF TABLES ......... ..... .... ............. .......................................... 7

LIST O F FIG U RE S ................................................................. 8

LIST O F FIG U RE S ................................................................. 8

ABSTRAC T ...........................................................................................

CHAPTER

1 IN T R O D U C T IO N ......................................................................................................1 1

Performance Indicators and Total Factor Productivity ..................................................13
F ro n tiers ................ ............. ..................................................................................................... 15
N on-param etric Frontiers .............................................................17
P aram etric F rontiers ................................................................20
E explaining Inefficiency ................................................................23

2 BENCHMARKING CENTRAL AMERICAN WATER UTILITIES ...............................27

D ata C o llectio n P ro ce ss .................................................................................................... 2 8
Sector Performance Indicators ..................................... ........... ........ ................. 31
O operational Indicators .................................................................................. ..33
F in an cial In dictators .............................................................................................. .. .... 3 5
Q u ality In dictators ........................................................................................3 6
Total Factor Productivity A nalysis................................................................................... 37
D ata Envelopm ent Analysis (DEA) .................. .............................. .............. .. ........ .... 38
Stochastic Frontier .................. .................................... ........................... 42
Concluding O bservations...................................................44

3 SECTOR FRAGMENTATION AND AGGREGATION OF SERVICE PROVISION
IN THE WATER SECTOR...................... .............. ........ .......56

In du story F ram ew ork ....................................................................................................58
Model Specification and Data .................................. .........................64
E m pirical R results ............................................................................... ................. 70
R egularity of the C ost Function ......................................................................... ......70
E fficien cy .........................................................................7 2
Environm mental and Output Variables ........................................................................ 73
E con om ies of S cale ............................................................................. ..................... 76
Input Factors and Price E lasticity ........................................................................ ...... 78
C including O b servations............................................................................................79









4 INFRASTRUCTURE SERVICE PROVISION TO MULTIPLE JURISDICTIONS:
A N E FFIC IEN C Y A N A L Y SIS ................................................................. .....................88

P political Su b -div ision s .............................................................................. .. ................ .. 9 1
O w nership Structure of U tilities.................................................. ... .... 93
Contraction and Expansion of Service Provision......................................... ...... ............. 93
C change in Service C overage.............................................. .................... ............... 95
Model Specification........ ............ ....... ... .. ..... ...... ............ ....96
Results from Estimation .................................. .. ...... ... ..............100
C including O b servations............................................................................ ....................104

APPENDIX

A DESCRIPTION OF VARIABLES FOR DATA COLLECTION ....................................113

B SUMMARY OF PERFORMANCE ANALISYS METHODOLOGIES.............................115

L IST O F R E F E R E N C E S ......... ................. ............................................................................. 116

B IO G R A PH ICA L SK ETCH ........................................................................................ 121









LIST OF TABLES


Table page

2-1 Share of water coverage within country by 2005 ................................... ...... ...............49

2-2 Operational performance indicators by 2005...... ...................... ...............52

2-3 Summary of finance indicators-average values from 2002 to 2005 ..............................53

2-4 Percentage of change values for 2002 to 2005 ...................................... ............... 53

2-5 Summary of quality indicators- average values from 2002 to 2005..............................54

2-6 Total factor productivity indexes from 2002 to 2005 ................................ ............... 54

2-7 DEA technical efficiency and scale impact on efficiency for 2005......................... 54

2-8 Efficiency rank and operating cost reductions............................................................. 55

3-1 Change in network length between 1996 and 2003 ............................ ................82

3-2 Summary statistics for outputs and price of inputs ...................................... .................82

3-3 Summary statistics for utilities' specific characteristics.................................................83

3-4 Estimation results for the translog cost function..................................... ............... 84

3-5 Technical inefficiency statistics for 2003 ............................................... ............... 85

3-6 Cost reduction statistics for 2003 ....................................................................... 86

3-7 Economies of scale ............................... ... ..... .... ................. 86

3-8 Distribution of firm s by regions and size....................... ........................ ............... 86

4-1 Number of provinces, districts and utilities providing service by department ...............106

4-2 Service coverage increase from 1996 to 2003 and population growth by region............07

4-3 Mean values for population and coverage increases............................................108

4-4 Estimation results for variables comprising the intercept of the translog model............. 109

4-5 Inefficiency statistics by region for 2003+ .................. ............ ...............111

4-6 Results from efficiency regression (Dependent variable average efficiency) ...............12









LIST OF FIGURES


Figure pe

2-1 Average volume of water delivered, billed and lost from 2002 to 2005 ...........................49

2-2 Average number of connections from 2002 to 2005 ............................... ............... .49

2-3 Average population with water service and total population in the area from 2002 to
2 0 0 5 .............. ....................... ....................................... ......... ..... 50

2-4 Average changes in number of water connections and network length from 2002 to
2 0 0 5 .............. ....................... ....................................... ......... ..... 50

2-5 Average coverage and network density from 2002 to 2005 ........................................... 51

2-6 Average volume of water billed and water consumption from 2002 to 2005 ..................51

2-7 Average operating cost per connection and network density from 2002 to 2005 .............52

2-8 Average operating cost per cubic meter from 2002 to 2005......................................53

2-9 Number of connections and workers by 2005 ....................................... ............... 55

3-1 Technical frontier for 1996 and 2003 ............. ... ........ ................ ..... 85

3-2 Fitted average total cost per cubic meter of water billed for 2003 ................ ........ 87

4-1 Schematic structure of jurisdictions in Peru and examples of service provision.............107

4-2 Departments of Peru showing those with variation in the number of districts..............108

4-3 Values of predicted inefficiency for 2003 .......................................... 108

4-4 Movement towards efficient frontier from 1996 to 2003 .............. ...................110

4-5 Actual total costs (black line) and prediction (dashed gray line) for 2003 ..................111









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

ESSAYS ON LATIN AMERICAN INFRASTRUCTURE:
EMPIRICAL STUDIES OF SECTOR PERFORMANCE

By

Maria Luisa Corton

August 2008

Chair: Sanford Berg
Major: Economics

This study examines the performance of utilities in the water sector of Central American

countries and Peru. Performance is assessed as technical efficiency and it is measured utilizing

different methodologies. Chapter 1 explains the methodologies utilized in the different

performance analyses presented in Chapters 2, 3 and 4. Chapter 2 provides a comprehensive

efficiency analysis of water service providers in six countries in the Central American region.

Pressures for sector reform have stimulated interest in identifying and understanding the factors

that can contribute to network expansion, improved service quality, and cost containment. The

aim of the analysis is to provide policymakers and investment funds institutions with quantitative

evidence on the effectiveness of the regional water sectors and utilities under different

institutional arrangements. In addition to key sector performance indicators, the analysis

considers total factor productivity indexes, a non-deterministic frontier and a stochastic cost

frontier. Differences in results from these methodologies are due to the assumptions imposed to

each model specification and data employed. The Central American performance indicators are

compared to similar indicators for service providers in Latin America. From the group of

countries analyzed in Central America, the water service provider from Panama is the best

performer according to results from all the methodologies.









Chapter 3 investigates the presence of economies of scale and technical efficiencies in the

water and sanitation sector of Peru. Chapter 3 analyses the sector structure considering size and

location of service providers. The aim is to address the merits of introducing additional suppliers

in some regions and possible merging of small utilities in other regions. The analysis employs a

stochastic cost frontier model. Overall, economies of scale are present in all firms in the forest

region and in the small firms located on the coast. These findings support the aggregation

process: consolidating some utilities could lower costs.

Chapter 4 examines the impact on efficiency of water utilities providing service to multiple

jurisdictions. The issue analyzed is whether the expansion of service across multiple jurisdictions

is more efficient than expanding service within a single jurisdiction. In this context, a jurisdiction

is a unit of government designed to carry out public functions within a specific territory. The

hypothesis is that utilities answering to a heterogeneous group of jurisdictional authorities are

less efficient than those reporting to a single authority. Political issues and bargaining for

resources may contribute to production inefficiencies affecting firms' costs. The resulting

inefficiencies may offset any scale economies associated with serving a larger area. Results from

the analysis support the stated hypothesis by finding less efficiency associated to utilities

providing service to more than one province when they are serving in a department where there

is more than one provider.









CHAPTER 1
INTRODUCTION

Developments over the past decade in quantitative techniques and pressures for

infrastructure reform have stimulated interest in identifying and understanding the factors that

can contribute to improving infrastructure performance. Utility managers, infrastructure sectors'

associations, regulators, and other groups have begun to undertake statistical analyses to measure

utilities' performance over time, across countries and geographic regions. Estache, Perelman and

Trujillo (2005) present an analysis of reforms in the electricity, telecommunications,

transportation and water sectors in the last decade. The authors provide a comprehensive list of

empirical studies in each sector highlighting the factors each study considered.

This study's contribution to the infrastructure literature is twofold. First it adds a regional

analysis and a country case analysis to the water sector literature. Secondly, it provides an

analysis of unique factors that affect infrastructure performance such as sector fragmentation and

the provision of service to multiple jurisdictions. These governance aspects are analyzed to

determine the impact of infrastructure decentralization on utilities' performance a unique

approach not yet covered in the empirical literature.

Empirical studies in infrastructure addresses privatization, the type of regulation in place,

or utilities' ownership as factors affecting performance. At a regional level, empirical studies are

limited, mainly focusing on the electricity sector. For example, Zhang, Parker and Kirkpatrick

(2008) investigate the effect of privatization, competition and regulation on the performance of

electricity utilities across 36 developing countries from 1985 to 2003. Estache and Rossi (2008)

analyze the effect on the electricity sector's performance of having a regulatory agency in place

for 51 developing countries from 1985 to 2005. Similarly, Jamasb and Pollitt (2003) review the

effect of regulation on the performance of a group of European electricity distributors.









Regarding the water sector, the empirical literature related to more than one country is also

limited. Sabbioni (2007) examines the relative efficiency of private and public utilities in the

water and sanitation sector in Brazil considering different regions within the country. Clarke,

Kosec and Wallsten (2004) examined the impact of private participation on the performance of

the water sectors of Brazil, Argentina and Bolivia measured as the increase on service coverage.

In Chapter 2, several methodologies for assessing water utility performance are applied to

six countries in Central America: Costa Rica, El Salvador, Guatemala, Honduras, Nicaragua, and

Panama. In Chapter 3, the performance of the water and sanitation sector in Peru is investigated

to explore the impact of firm specific production characteristics on the performance of service

providers. The aim is to address the merits of introducing additional suppliers in some regions

and merging of small utilities in others. In Chapter 4, the attention focuses on the role of service

provision to multiple jurisdictions and its impact on efficiency. The goal is to investigate the

effect on performance of the decentralization of service provision to municipalities.

Decentralization is examined by Pearce-Oroz (2006). The author utilized a set of single

dimensional indicators to examine the performance of decentralized versus centralized utilities in

Honduras. This analysis is partial since it is based on performance measures that focus on

specific areas of the production process of service providers. This analysis does not consider

factors specific to the environmental circumstances of the production process or those particular

to the operation practices of the utilities. In addition, this type of analysis does not include the

interrelationships among the factors of production.

In contrast, in Chapter 4 the impact of decentralization is considered by estimating a

stochastic cost frontier where specific characteristics of the production process and their









interrelationship are taken on account. The following sections provide a set of performance

measurement methodologies which are utilized in this study.

Performance Indicators and Total Factor Productivity

A substantial body of technical literature exists regarding how to measure performance.

Coelli, Estache, Perelman and Trujillo (2003) present a survey of different methodologies on this

topic. The methodologies considered in this study are performance indicators, total factor

productivity indexes, and frontiers.

The simplest types of performance measures are single dimensional indicators such as

labor productivity, service coverage, and non-revenue water. These indicators focus on a specific

area of performance within the production process. These types of indicators fail to account for

the relationships among all the elements of production and generally do not consider the impact

of environmental or country specific factors.

Productivity indexes represent a way of measuring performance over time since they

identify production differences between two time periods. A comprehensive exposition and

analysis of such measures is found in Balk (2003). The basic idea behind a Total Factor

Productivity (TFP) index is to determine how much output is produced due to each unit of input,

which is expressed by Equation 1-1.

TFP = YX (1-1)

Here, Y is the level of output and X the level of input utilized in the production process. In

practice, there may be more than one output produced from a combination of inputs. In such a

case, a TFP index is constructed as the ratio of an output index to an input index. The input index

should reflect the relative importance of each input in producing the output(s) and the output

index should reflect the relative importance of each output. These relationships are represented









by weights. Assume r is the weight given to k outputs and s the weight to n inputs. Equation 1-2

is the general formula to calculate a TFP index measuring productivity change between two

periods of time, say from year 0 to year 1.

k r1yk / S X
TFP1 k
TFP, ZrkYk /Z SnX
k (1-2)

Three aspects are relevant when setting the above mentioned weights: the selection of the

elements that represent the weights, the mathematical or functional form that combine them and

whether the weights are the same for the two analyzed periods or not. The weights are generally

prices for input factors and costs for outputs. In Equation 1-2, the weights are assumed the same

for both periods but they may not be which yields different alternatives for calculating TFP

indexes. When considering the initial set of prices, Equation 1-2 represents a Laspeyres index

while using the final period prices produces a Paasche index. The Fisher index utilizes the

geometric mean of the two periods.

Regarding the functional form to relate the weights, the indexes described so far imply a

linear functional form. The Tornqvist index, represented in Equation 1-3, uses a logarithm form

which is more flexible in reflecting a production technology.

In = P (r + r) ln (s, + o) ln X
TFPo 2 k Yo 2 X
.TFP= 1( .) 2 s ) K (1-3)

By understanding the sources of productivity change, managers can focus attention on

areas that seem weak. At the same time, by understanding the sources of these changes, policy

makers, investors and other stakeholders can point to the most productive firms as examples of

strong performance promoting the diffusion of best practice to all firms.









In Chapter 2, a comprehensive analysis of performance indicators and Total Factor

Productivity measures is performed for the water sector in the Central America region. Chapter 2

examines performance patterns across six countries, focusing on three performance indicators:

operational, financial and quality. In addition, trends for 2002 to 2005 are examined by

computing total factor productivity indexes for water utilities in each country. A central point in

performing these calculations was the availability of data. The data collection procedure is given

attention and is described in the Chapter 2.

Frontiers

A frontier approach has been widely used to analyze efficiency as a measure of a firm's

performance. The measure of efficiency is linked to the functional representation of production

technology structure (Aigner and Chu 1968 and Lovell 1995). From production theory, an

industry production function represents the relationship between input factors and output which

underlines a specific production process technology such that maximum product is obtained from

a combination of input factors within the industry environment.

Conceptually, the industry production function is a frontier determined by the production

process of those firms attaining maximum output with a set of inputs. Other firms in the industry

fall short of the frontier due to the presence of production inefficiencies. Suppose a firm uses a

set of inputs x= (xl, x2, .., Xn)E 9", available at prices w= (wi, w2, ....., Wn ) e 9++ to produce

output y e 9+ in an industry environment characterized by exogeneous variables

z= (zi, z2,.., zj)e 9V. Exogeneous or environmental variables are not controlled by the firm

decisions. These environmental variables are characteristics of the firm, such as ownership, or

features of the operating environment such as the country economic characteristics, competitive

conditions or geographical characteristics where the firm is located.









A production frontier is represented by a function f : 9x 9W -> 9+ with a parameter

vector f such thatf(x, z; fl) denotes the maximum output obtainable from input vector x in

environment z. If the firm minimizes costs, the dual of a production frontier is a cost frontier

represented by a function c 93+x 91" x 91 -> 91 with the same parameter f such that

c (w, y, z; fl) denotes the minimum expenditure on inputs required to produce output y with input

price vector w in environment z. In this context, the cost frontier represents the best that can be

achieved in environment z so observed cost cannot be less than minimum possible cost, in other

words w'x > c(y, w, z; F). The inequality leads to measures of inefficiency. In particular, cost

inefficiency is measured by the ratio defined in Equation 1-4.

c(y, w, z; /)
( l<1 (1-4)
(w x)

Aigner, Lovell and Schmidt (1977), pointed out that the possibility of estimating a frontier

as opposed to an average production function to examine firms' performance came after Farrell's

(1957) pioneering work on efficiency. Farrell constructed an envelope isoquant for the industry

production function. Murillo-Zamorano (2004) presents a comprehensive set of techniques to

calculate or estimate frontiers. Previous work summarizing the methodologies related to frontiers

and efficiency analysis include that by Kalirajan and Shand (1999) who review and compare

methodologies for measuring technical efficiency.

Following Murillo-Zamorano (2004), frontiers are classified as non-parametric, those

where the frontier is not determined by a functional form, and parametric, those that have a

specific functional form to specify the optimal frontier. Among the parametric frontiers, there are

deterministic or stochastic frontiers, which considers the residual term including the presence of

noise and inefficiency, and non-deterministic those that view the residual term comprised only









by inefficiency. Naturally, frontier techniques may yield different results due to the different

underlying assumptions in determining the optimal frontier.

Different frontier techniques are employed for examining the performance of water utilities

in Central America in Chapter 2. In Chapter 3 a stochastic cost frontier is utilized to address the

performance of water utilities in Peru. Chapter 4 utilizes the same frontier technique but now it

includes different environmental variables. In addition, in Chapter 4 efficiency is calculated by

means of a non-parametric frontier. Results from the efficiency calculation are utilized to

examine the presence of efficiency drivers in the sector.

Non-parametric Frontiers

Data Envelopment Analysis (DEA) is the most commonly used non-parametric frontier

methodology. Charnes, Cooper and Rhodes (1978) were first to present the concept of the

relative ranking of decision making units according to their efficiency. A DEA calculation

determines simple relationships among variables. For example, utilities that produce far less

output than other utilities, which are using the same input levels, are deemed to be relatively

inefficient. This methodology is viewed as an "extreme point" method because it compares

production of each firm with the "best" producers. A recent empirical study by Mathur (2007)

utilizing DEA for the telecommunications sector of India provides a detailed and rigorous

illustration of this methodology which is commonly applied in infrastructure to measure relative

performance. Efficiency results from a DEA frontier are contingent to three main factors:

* The composition (homogeneity) of the sample set of firms to be analyzed which is critical
in determining the set of best producers to be compared with each firm

* The set of selected inputs and outputs which establishes the comparison terms

* The quality of the data since this methodology assumes that there are no errors









A DEA analysis consists of measuring the efficiency of any firm as obtained by the ratio of

weighted outputs to weighted inputs subject to the condition that similar ratios for every firm are

less than or equal to unity. This relationship is expressed mathematically in Equation 1-5.



Max... = subject to '1 <1
Jx,, Y EJX


a,.,p 0; j = ...n; i=1...m; r=l...s (1-5)

Here, Yo, Xo are observed output and input variable vectors of the firm under evaluation; a

and P are the weights to be applied to all units; i represents an input within a set of m, r an output

within a set of s, andj one of the n firms. From Equation 1-5, possible values for the measure of

efficiency, Ao, ranges from zero to unit. This means that each firm can weight inputs and outputs

differently as long as the ratio of their linear combination is less or equal to one. The efficiency

of firm zero is rated relative to all firms. It is general practice to utilize the dual expression of

Equation 1-5 under a linear programming framework instead as it is shown in Equation 1-6. In

this case, rho (p) is a vector of intensity parameters which allows for the convex combination of

the observed inputs and outputs. In other words, for an inefficient firm the efficient level of each

production factor gets expressed as a linear combination of these intensity parameters as weights

of the production factors of the peer group.


Min A
A
subject to

Y, < pY
pX < AX ,
P e RN (1-6)









The output from a DEA exercise is the proportion by which the observed inputs could be

contracted if the firm were to operate efficiently. Intuitively this means that the same level of

output can be produced with fewer inputs, so it is referred as an input efficiency approach. The

implicit assumption is that managers minimize input usage given output level. In the economic

literature, this is referred as the Farrell's measure of input efficiency (Farrell 1957).

Alternatively, the output efficiency approach considers the maximal proportional output

expansion with the input vector held fixed, so managers maximize output given a set of inputs.

Unless constant returns to scale are assumed, each of these approaches yield a different scalar

value of efficiency. A comparison of results from input and output oriented models is found in

Orea, Roibas and Wall (2004). A study by Banker, Cooper, Seiford, Thrall and Zhu (2004)

addresses details and applications of input and output efficiency approaches under variable and

constant returns to scale.

For regulated industries, such as the water sector in Central America and Peru, an input

approach is the natural option given that utilities most generally have service obligations to all

customers under a fixed tariff. This approach implies that firms are fully capable of reallocating

resources when improving efficiency. In Chapter 2, the DEA methodology is utilized to assess

the utilities' performance in the Central America water sectors in 2005. A set of two inputs and

two outputs was considered: labor and network length, and volume of water and number of

connections respectively. Robust performance comparisons require analysts to obtain

comparable data across firms, select appropriate empirical methodologies, and check for

consistency across different methodologies. In Chapter 4, a DEA approach is utilized for

robustness check of the stochastic frontier model utilized in the performance analysis of the

water sector in Peru. Technical efficiency is calculated for each year of the panel data set and









then regressed on the set of covariates utilized in the stochastic model to check for their

explanatory power on efficiency.

A Malmquist index measures the Total Factor Productivity change between two time

periods utilizing the ratio of the distances of each data point relative to a common production

technology. When calculating this index it is common practice to utilize DEA to calculate these

distances. Following Fare, Grosskopf, Norris and Zhang (1994), and considering the input

perspective already selected, a Malmquist technology change component based on the geometric

mean of the considered periods is defined in Equation 1-7.

-1/2
TFP, Do(Yo,Xo) D,(Y,,XDX,)DI(YX) (1-7)
TFP, D,(Y,,X,) Do (Y,,X,)D,(Y,,X,)

The first ratio term to the right of Equation 1-7 indicates a measure of input-oriented

technical efficiency change for the analyzed period (the catching up effect or movement towards

the frontier). Negative values indicate efficiency has declined over the period (the initial

efficiency value is higher than the final value). Positive values indicate increased efficiency. A

value equal to zero indicates no efficiency change. The term within brackets represents technical

change calculated as the geometric mean of the shift in technology between two periods. It is

important to notice that while a TFP index is calculated only on reference to a particular firm a

firm change of productivity over time the efficiency change component of the Malmquist index

is calculated with respect to the movement of a firm towards the optimal frontier which is

determined by a group of firms. This efficiency change component is calculated for the water

utilities in Central America.

Parametric Frontiers

Finding key determinants of a firm's performance requires knowledge of its production

technology, or the particular way the firm selects inputs to produce a set of outputs. This









knowledge translates into a functional mathematical form for the production technology which

recognizes the relationship among factors entering the production process. Aigner et al. (1977)

were first in introducing a stochastic frontier approach in parallel with Meeusen and Van den

Broeck (1977).

According to the frontier classification presented by Murillo-Zamorano (2004), depending

on the specification of the error term a frontier may be deterministic or non-deterministic. A

deterministic frontier views a firm's deviations from the frontier as pure inefficiency whereas a

stochastic frontier recognizes the presence of noise and measurement errors. The problem with

the deterministic frontier is that the estimated inefficiency may be confounded with noise. In

addition, any misspecification or missing variables are also considered as part of the inefficiency

term. This can lead to biased estimates of the position and shape of the frontier surface. Jensen

(2005) compares both approaches for cross section models and draws conclusions depending

upon the size of the data set, technology functional form and whether the objective is to estimate

accurate inefficiency levels or efficiency ranking.

A stochastic frontier framework considers a two component disturbance term: the typical

noise (v), a symmetric component, and inefficiency (u), which represents those factors under the

firm's control such as technical and economic inefficiency, the will and effort of the producer

and his employees, and defective and damaged products. For the case of a cost function

specification, u is added to the frontier, showing the cost inefficiency coming from actual cost

being far from the minimum possible cost frontier. A stochastic cost frontier is estimated for the

water utilities in Central America and in Peru.









Following Atkinson and Cornwell (1994), the general form for a cost frontier capturing

input technical efficiency, which measures the potential each firm has to reduce cost (holding

output constant), is specified in Equation 1-8.

C,(Y, /u,)= min[(Wf/u,)(uX,) f(uX,)= Y]= (1/u,)C,(Y,W) (1-8)

In Equation 1-8 the time subscript is been omitted for simplicity, Y, is observed output of

firm i ,fis a production function, common to all firms, We, = (w1, W21 ...,wji) is a vector of

input prices, X, = (xl,, x2, ..., x,,) is a vector of inputs, and u, 0 < u, < 1, is a parameter which

measures the extent to which minimal input usage differs from actual input usage. The last

equality in Equation 1-8 follows from the fact that a cost function is linearly homogeneous in

input prices. To reach the frontier each firm must lower cost by the amount specified in

Equation 1-9.

[(C, /u) C] (1-9)

Applying natural logs to Equation 1-8, yields Equation 1-10.

In C, (Y, W,/u,) = ln(1/u,) + In C, (Y, W) > -ln u, + In C, (Y, W) (1-10)

Regarding the inclusion of time in the model specification, Cornwell, Schmidt and Sickles

(1990) utilizes a quadratic functional form of a time trend variable included in the intercept; Lee

and Schmidt (1993) includes in the intercept a time trend variable interacted with environmental

variables. Some authors such as Atkinson and Halabi (2005), Battese and Coelli (1995) and

Atkinson, Cornwell and Honerkamp (1999) introduce the time variable twice: in the production

function accounting for technical change and in the explanation of the inefficiency effect.

Time is included in the intercept of the functional form of technology to account for

technology changes in the analysis of the water sector of Peru. Although the water industry is not

characterized by substantial or rapid technical changes as in the case of the telecommunications









industry, the eight year period under analysis is long enough to attempt to capture such an effect.

In addition, to address the possibility of inefficiency changing over the eight year period, time is

included in the econometric model following the functional form specified by Battese and Coelli

(1992) for the inefficiency component, (uit), which is defined in Equation 1-11.

ut = exp[-q(t -T)]; u, iid N(p, 2) (1-11)

In Equation 1-11, the u, are assumed to be independent and identically distributed non-

negative truncations of the N((/, r2) distribution. Eta (rq) is a scalar parameter to be estimated, t

represents each time period within the T total number of periods. With this specification, as t

increases, u,t decreases if r is higher than zero, remains constant if r equals zero and increases if

q is less than zero. The first case suggests that firms have improved their level of efficiency over

the time period under consideration. This specification assumes all firms have followed the same

trend. This is restrictive but a reasonable assumption for the water sector in Peru since the group

of utilities is delivering service under the same economic, social and political circumstances.

Results from estimation support the hypothesis of inefficiency varying over time and following

the specification in Equation 1-11.

Explaining Inefficiency

When the objective is to explain inefficiency in a stochastic model, the introduction of

environmental variables requires elaboration. As it is been previously explained, in addition to

output and input factors, environmental variables represent specific characteristics related to the

environment where the utilities deliver service usually explaining production differences either

across countries, regions or among firms. Firm specific characteristics are also referred as firm

heterogeneity. In a panel data context unobserved firm heterogeneity is referred as the panel

effect where a panel is comprised by observations for each time period for each firm. In a panel









data frontier analysis, unobserved inefficiency is usually treated as the panel effect. The work of

Aigner et al. (1977) on frontiers was extended to the panel data case by Pitt and Lee (1981) and

Schmidt and Sickles (1984). The estimation of a frontier panel data proceeds by first finding the

best (minimum) cost-performer-year for each firm, in other words the best performer within each

panel. The frontier consists of all identified best (minimum) cost performers.

Empirical researchers have attempted to disentangle firm heterogeneity from inefficiency

given that unobserved inefficiency may be confounded by unobserved firm heterogeneity

(Greene 2005a, 2005b and 2004; Farsi, Filippini and Greene 2005; Hattori 2002).

Coelli, Perelman and Romano (1999) point out that, when environmental variables are

included in the technology functional form they affect the shape of the technology and the

estimated residual includes a net measure of inefficiency. Otherwise, the estimated residual

represents a gross measure of inefficiency. The issue is all about how to include environmental

variables in a model provided that they may explain observed inefficiency.

To illustrate this point, consider a cost function, C, f(.) a technology function, Wt a vector

of input prices, Yt a vector of outputs, and Zt a vector of environmental variables. The previous

classification of inefficiency translates into Equation 1-12 and Equation 1-13. The parameters a,

y, p are to be estimated and u and v represent inefficiency and noise respectively.

Ct = f(Wt, ItZ,; ca, 7, ()+ ,, and E,, = u, + v, (1-12)

C,t = f(Wt, t,; a, y) + E,, and E,, = u, + v,, where u = Zt +, (1-13)

According to Coelli et al. (1999), a net measure of inefficiency is obtained when the model

is specified as in Equation 1-12 where the vector of environmental variables (Z) is included in

the production technology functional form. A gross measure of inefficiency is obtained when the

model is specified as in Equation 1-13 where the set of explanatory variables are directly









explaining the inefficiency term u. If the environmental variables do explain inefficiency then

they are correlated to the residual term in Equation 1-12. In this case the estimated coefficients

are biased if using an Ordinary Least Squares estimator. A fixed effects estimator is more

appropriate in this case given that it allows explanatory variables to be correlated to the residual

term.

In the case of specification defined in Equation 1-13 several aspects need to be considered.

First, if the set of environmental variables do explain inefficiency using the right term of

Equation 1-13 to estimate u then leaving these variables out of the left term specification of

Equation 1-13 will produce a residual E that includes missing observed inefficiency as well as

unobserved inefficiency.

Secondly, the use of a two step estimation procedure, first predicting C, under the

assumption that the residual E is identically distributed and then using the predicted inefficiency

to regress the environmental variables contradicts the identical and independent distribution

assumption previously made. Battese and Coelli (1993, and 1995) have reviewed the evolution

of the literature regarding how inefficiency is estimated and they pointed out that a simultaneous

estimation procedure is more appropriate given the issues previously discussed. Either using the

model in Equation 1-12 or 1-13, a robustness test is appropriate. This test can be performed by

estimating or calculating the inefficiency term by a different methodology and then checking the

explanatory power of the set of environmental variables.

Ultimately, whether environmental variables directly explain firm's possible inefficiency

or if they have an effect on the technology shape is a matter of interpretation related to the

specific objective of the analysis. In this study, a stochastic cost frontier is estimated for the set

of water utilities in Central America and the municipal water utilities in Peru. Environmental









variables are introduced in the intercept of the technology functional form assuming that

unobserved inefficiency is not correlated to these variables. A robustness check is performed in

Chapter 4 by calculating a DEA frontier to obtain inefficiency values and investigate the

explanatory power of the environmental variables.









CHAPTER 2
BENCHMARKING CENTRAL AMERICAN WATER UTILITIES

The purpose of Chapter 2 is to analyze the relative performance of water utilities in the

Central American region to identify best performers and areas of weakness in the sector. The

results can help decision makers' better direct investment funds into projects that will enhance

performance in the water sector in this region. This study is part of a Public Utility Research

Center project funded by the Inter-American Development Bank as part of a new investment

program which considers the possibility of providing loans to water utilities without requiring

sovereign guarantees.

A limitation associated with studies of Central America is the scarcity of data related to the

water sector. The first steps of the study involved examining existing data and defining a set of

variables to be collected. Appendix A presents the description of the identified variables. This set

of variables was kept simple and limited to reduce possible road blocks during the collection

process. During the data collection process, some factors were found to be limiting and others

were critical for the success of the process. These factors are described in the data collection

process section.

The subsequent steps of the study related to performance measurement. The basic concepts

and methodologies were described in Chapter 1. With key input, output, and quality

information, basic performance comparisons can be made. A set of performance indicators

commonly used among practitioners in the water sector was calculated to provide a very simple

picture of the sector's performance characteristics in the region. Some of these performance

indicators were compared to those presented by the Association of Water and Sanitation

Regulatory Entities from Latin America (Asociaci6n de Entes Reguladores de Agua Potable y

Saneamiento de las Americas, ADERASA) benchmarking task force in its most recent annual









report for Latin American countries (Informe Annual Benchmarking 2006;

http://www.aderasa.org/es/documentos3.htm?x=753, last visit: 5/25/2008) For information on

the countries and regulatory authorities comprising this association see Corton and Molinary

(2008). The availability of data from 2002 to 2005 allowed assessing performance in the region

through the calculation of total factor productivity indexes.

In addition, to provide a more specific picture of the efficiencies associated to production

practices in the region, a production frontier using data envelopment analysis and a stochastic

cost frontier were included in the analysis. Differences on some of the values obtained from this

variety of performance measures are based on the different assumptions underlying each

methodology. Appendix B explains the purpose of utilizing each methodology. A best performer

was found consistently through the calculation of the performance methodologies. The following

sections describe each major step conducted on the analysis of the water sector performance in

this region.

Data Collection Process

The starting point of this study was to build a verifiable data base taking into account the

data already available. To this end, the author requested and collected data from ADERASA and

the International Benchmarking Network for Water and Sanitation Utilities (IBNET). The

ADERASA data base is comprised of data that comes from the regulatory agencies in each Latin

American country. This information is reported by some of the utilities in each sector but not all.

Appropriate contact was established with these utilities to verify the existing data and to obtain

missing values. In addition, Guatemala and El Salvador are not members of ADERASA, so data

for these countries were collected for a first time. The adopted strategy for the collection process

was incremental in the sense that data were sent to the source several times for verification. A

new and refined data set for the water sector in Central America emerged from this process.









Nevertheless, only a subset of variables was used for the analysis because not all countries

reported all variables in every year. Consequently, the number of observations was reduced to

allow the data set to be comparable for all utilities and to include all countries.

The total time spent for data collection was six months. Data owners' response was slow

for some utilities given ongoing political and institutional changes at the time. In addition,

several factors were identified as affecting data availability within the region: the ongoing water

sector restructuring, the low level of water infrastructure in place, and the low presence of

information technology among the service providers.

From an institutional point of view, Costa Rica, Panama and Honduras have independent

regulatory agencies but El Salvador, Guatemala and Nicaragua still have central government

bodies overseeing the water sector. Some of these central government institutions are

undergoing restructuring such as the transfer of sector responsibilities to municipalities. Some

countries have just finished a major restructuring of this type. Because these changes generally

imply changes within the company staff, the flow and registering of data may get interrupted,

affecting data collection procedures.

El Salvador, Honduras and Nicaragua show a low level of infrastructure in place. This

promotes the presence of a large number of local independent water providers which complicates

the data collection process. Solo (1998) provides a detailed description on the role of these

independent providers within the water sector of Latin American countries.

Overall, the water sector in this region is fragmented given the decentralization of service

provision into municipalities. For instance, for Honduras, with 271 municipalities, and for

Guatemala with more than 300 municipalities the difficulty in assembling a comprehensive data

set is evident. Consequently, available data came from the municipalities serving the largest









cities. This fragmentation greatly hampers performance analysis for the water sector in this

region.

Finally, the development of information technology is central to any data collection

initiative. To frame the status of the region on this matter, the information and communication

technology (ICT) diffusion index utilized by the United Nations Conference on Trade and

Development (UNCTAD) is utilized (UNCTAD 2006). More specifically, the ICT index

includes a connectivity index to measure technology infrastructure development. This

connectivity index includes per capital number of Internet hosts, number of PCs, number of

telephone mainlines and the number of mobile subscribers. The connectivity index for 2005 was

0.75 for the United States. In the same year this index was 0.20 for Costa Rica; 0.10 for El

Salvador and Panama; 0.08 for Guatemala and 0.04 for Honduras and Nicaragua. Information

technology is the core to any structured data collection procedure.

The availability of an information system specific for the sector is crucial for any data

collection process. Initiatives in this respect are only incipient. In 2004, a workshop hosted by

Peru with participation of several Latin American countries representatives from the water sector

gathered initial ideas and directed some efforts into the development of a water sector

information system common to the region. A similar initiative hosted by El Salvador developed

in late 2006.

The presence of technology is necessary but not sufficient for improved information on

water utility performance. When designing rules for the sector, the government needs to

consider not only the utilities, the main entity responsible for collecting appropriate data, but also

the role that each stakeholder plays in the flow of data within the sector. For instance, the

reporting of data to the regulator from the utilities needs to be stated by law and not taken as an









informal relationship between the parties. In the same way it is important to establish formal

communication channels among all the sector stakeholders, such as environmental or municipal

development agencies, in a way that data collection programs and possible data repositories are

well identified and efforts are not duplicated.

Because data from all service providers are not available for this study, knowing the share

of the population served with respect to total country population, known as service coverage,

permits identification of the comprehensiveness of this study. Table 2-1 summarizes the share of

water coverage for each country. Costa Rica, Guatemala and Honduras are represented in the

analysis by two service providers of different sizes, which are going to be referred as the small

and large providers for each of these countries. The rest of countries are represented only by one

operator.

Sector Performance Indicators

An initial step in analyzing performance is to calculate the most commonly used

performance indicators in this sector. These have been classified in this study as operational,

financial and quality indicators. Since they are benchmarked against those indicators calculated

by the ADERASA Benchmarking group, their definitions are kept the same to maintain

consistency across the region.

In general, the extent of water service provision can be measured by volume of water,

number of connections and population served. These factors provide a good picture of the size of

the company providing service. Overall, in the Central American countries the water utilities can

be broadly classified as small, medium and large. Figure 2-1 shows average volume of water

delivered, water billed and lost for 2002 to 2005. From a volume of water point of view, Panama

has the largest provider. Note that for this group of countries and in general in Latin American









countries, volume of water lost usually referred as non-revenue water, represents a large portion

of water produced.

Figure 2-2 depicts average number of water and sewerage connections and water

connections with meters. El Salvador has the larger system from a number of connections point

of view. On average, the small proportion of sewerage connections relative to water connections

may suggest cost restrictions to expand the sewerage systems in this region. In addition, the level

of metering varies among countries suggesting not only the possibility of cost restrictions but

also the different social acceptance of this practice.

The population dimension is shown in Figure 2-3 which depicts the variation of the mean

coverage values among service providers for these countries. Although Panama and Costa Rica

provide the largest volume of water delivered to customers, their number of connections and

amount of population served is smaller than that of the water provider in El Salvador the largest

provider from a water volume perspective. This illustrates some of the differences among

countries' water systems.

The availability of several years of data allows us to analyze the changes occurring in the

number of connections and network length which imply system expansion at different stages of

the investment cycle. Clearly, national priorities and funding sources affect the pace and pattern

of system expansion. Figure 2-4 show changes in the number of water connections and network

length occurred from 2002 to 2005 to illustrate system expansion as shares of total system

expansion for each country. The service provider from Guatemala shows a higher increase in

network length with respect to that of number of connections. This may suggest a system

expansion of the transportation segment where pipes but not connections are added as opposed to

an expansion of the distribution segment where pipes and connections are both added.









Alternatively it may indicate earlier stages of the distribution network expansion where

customers have not been connected yet. The service provider from Nicaragua presents an

opposite situation. Here the increase in the number of connections is higher than the increase of

length of network. This may be explained by connections added to satisfy commercial and

industrial customers who generally do not add to population served.

Operational Indicators

Water lost or non-revenue water reflects deficiencies in either operational or commercial

practices. The extent of water losses may reflect a cost tradeoff between increasing water

production and repairing network leaks to keep up with water demand. In other words, to satisfy

demand, managers may find it more costly to repair leaks and to control water losses than

increase water production. Pipe leaks on the transmission segment require costly maintenance

outlays, particularly on long or dispersed networks. Operational water losses arise in transit

while in the transport or distribution network, and are calculated as volume of water produced

less water delivered to the distribution network, expressed as a fraction of the volume of water

produced.

Referring to the distribution system, water losses may be either due to water theft,

representing commercial losses, or to leakage from pipes. Given the characteristics of this sector

it may be difficult for firms to control commercial losses if doing so entails denying the service

to the poorest segments of the population. For the distribution network, water losses are

measured as the difference between water delivered and water billed. Another way of viewing

this indicator is to calculate the ratio of water billed to water delivered to the distribution network

which is referred by the ADERASA benchmarking group as an indicator for commercial

efficiency. For utilities in the Central America sample, this indicator is equal to 55% which is

higher than the ADERASA value of 40%.









The metering indicator is calculated as the ratio of the number of connections with an

installed meter to the number of total connections. Meter acquisition and installation costs are

high. In some countries there is a direct allocation of metering costs to the consumer, which

generally translates into higher tariffs and/or connection fees. The higher is the level of metering,

the higher the possibility of identifying water losses, the more accurate will collection

information be and as a consequence, revenues may be higher. Overall, metering median value is

56%, which is lower than the 75% median value for ADERASA members.

Service coverage is calculated as the ratio of population with water service to total

population in the area of service of the utility. The median value for water service coverage in

this region is 90% which is close to the ADERASA value of 89%. There is a noticeable coverage

gap between large and medium-small utilities. In Central America, coverage is equal to 92% for

large firms, 66% for medium firms and 85% for the small utilities group.

Water companies with a similar scale, measured by number of connected properties, may

have different costs due to differences in network characteristics, such as length. Larger firms

could have lower costs due to a large amount of customers per kilometer of pipe, rather than due

to scale economies originated from total output. To explore this issue, network density,

measured as the ratio of the number of connections to network length is the performance

indicator considered in this analysis. The median value for network density equals 95 for utilities

in Central America. Larger firms have denser networks than medium and small firms. Figure 2-5

shows average values for coverage and network density from 2002 to 2005.

The low coverage and high network density values found for the Guatemala utility suggest

that the system may be expanded by increasing the length of network to reach out under-served

populated areas. The low coverage and low network density found in Honduras may indicate that









the system can be expanded by adding more connections to satisfy underserved population in the

area.

The ADERASA benchmarking group utilizes the ratio of volume of water billed to

population with water service as an indicator for water consumption. The median consumption

value for the region equals 219 liters per person per day, which is slightly higher than the

ADERASA value of 172. The group of smaller companies is characterized by a higher

consumption level of 323 liters per person per day as opposed to a lower value of 222 satisfied

by larger firms. Figure 2-6 depicts this indicator.

The number of workers per one thousand connections is used in the water sector literature

as signaling labor efficiencies or inefficiencies. A large value suggests the company is using a

higher than efficient number of workers on its production process. The median value for this

indicator equals 6.6, which is twice the value found for ADERASA members suggesting high

labor inefficiencies. Note that the number of workers considered for this indicator is a total figure

which includes the number of workers under a contract agreement. These workers do not have

any of the salary benefits provided by the company. Table 2-2 summarizes average values for the

operational performance indicators from 2002 to 2005.

Financial Indicators

Financial indicators in this sector are expressed in currency amounts per connection or

cubic meters of water. Operating costs for this region include labor and energy costs, chemicals,

administrative and sales expenses. Depreciation and finance expenses are considered to be part

of total costs. On average, operating costs are $91/connection. Figure 2-7 shows average

operating costs per connection and its relationship to average network density for the analyzed

period. Higher values of network density are associated with lower values for operating costs

per connection. Average operating cost per cubic meter of water delivered per utility is depicted









in Figure 2-8. The median operating cost per cubic meter is $0.10, half the cost of ADERASA

member countries.

For the large group of the utilities in the region, the median administrative expense per

connection equals $27, whereas it is $34 for the small group. Both values are lower than the

similar indicator for ADERASA members which equal $47. Table 2-3 summarizes average

values for the financial indicators between 2002 and 2005.

Table 2-4 presents a summary of percentage changes for some of these indicators from

2002 to 2005. Changes in operating costs for 2002 to 2005 are of small magnitude. Costa Rica

displays a significant increase in cost of workers (54%) and administrative expenses (51%)

which may explain the increase in its operating costs with respect to that of other utilities. On the

other hand, Panama's increase in operating costs (18%) may be explained by an increase in

energy costs (54%). El Salvador displays a decrease in administrative expenses (18%) which

may explain the decrease of its operating costs (10%). Service providers which presented an

increase in energy costs may be reflecting a combination of increases in input prices and greater

utilization of energy inputs to service larger systems. Significant increases in cost of workers

could be due to an increased focus on hiring professionals with managerial skills.

Quality Indicators

Compliance with water quality standards has a median value of 95.96% for utilities in

Central American countries. Continuity measured as the number of hours with water service

ranges from 20 to 24 hours on average per year. Number of complaints per connection (median

value per year) is similar for both ADERASA and Central American utilities (0.21). The median

annual number of leaks per km of pipe is 2.53 for ADERASA members, almost half the value

found on Central America countries, 5.19. This suggests a lower degree of pipes service

maintenance for Central America water networks compared with the Latin American set of water









networks. This also explains the higher value of water lost in the Central American water

utilities compared to this value for utilities in Latin America. Table 2-5 shows average annual

values for the quality indicators discussed during the period 2002 to 2005.

Total Factor Productivity Analysis

The basic idea behind a Total Factor Productivity (TFP) index is to know how much output

is produced for a given level of inputs. Four alternatives for calculating TFP indexes are

considered for the Central American water sector, which have been already defined in Chapter 1:

1) the Laspeyres index which considers the initial set of prices as the base period; 2) the Paasche

index which utilizes the final period prices; 3) the Fisher index which uses the geometric mean

of the two periods; and 4) the Tornqvist index which uses a logarithm functional form to express

the inputs and outputs relationships.

Defining the weight of outputs for calculating these indexes requires detailed information

on the production technology of these companies as well as more specific data. In order to

simplify the analysis such that weights for the outputs are not needed, two sets of TFP measures

are calculated: one considering volume of water billed as the output and another considering the

number of connections. As in most empirical studies related to the water sector, labor and energy

are the input factors under consideration. However, not all service providers reported energy

volume which limited the calculation of these indexes only to a subset of utilities. The weights

for the input factors are calculated as the ratio of their respective costs_relative to operating costs.

See Table 2-6 for obtained results.

The Laspeyeres, Paasche and Fisher TFP indexes yield similar results. This may be

explained by the fact that the length of period is short which produces only a small variation

when calculating the weights for the different indexes. All companies are more productive from

the view point of number of connections as opposed to volume of water billed. Panama is the









only country displaying increased productivity over the period when considering both number of

connections and volume of water. The productivity increase for Panama ranges from 31% to

53%. Nicaragua displays a very small decrease in productivity ranging from 2% to 5%. Finally,

El Salvador depicts a small productivity decrease ranging from 6% to 17%. Presumably,

increases in TFP should track decreases in average cost if all the other factors of production

besides labor and energy remain constant. The service provider for Panama is expected to

perform better with respect to other providers in the frontier performance assessment.

Data Envelopment Analysis (DEA)

This section examines each firm's relative technical efficiency in 2005 using Data

Envelopment Analysis (DEA). As in the case of TFP indexes, DEA assumes that the data

contains no measurement errors. A difference with respect to the TFP analysis is that the DEA

methodology allows us to consider a linear combination of outputs and inputs for the production

process without specifying their weights. Rather, these weights are calculated with respect to the

combination of these factors found on best producers.

A DEA model collapses into the selection of an appropriate set of inputs and outputs

involved in the production process. For the Central American water sectors, labor and capital are

selected as input factors and volume of water billed and number of connections are the outputs.

The amount of energy utilized in the production process was used in the calculation of the TFP

indexes and is generally used as an input factor in the production process of water utilities.

However data availability is limited for this set of companies. To include all firms in the

calculation of the frontier, length of network instead of volume of energy is considered as input

factor. Network length is utilized as a proxy to represent capital in the infrastructure empirical

literature. The rationale for doing this is the high amount of capital necessary to lay down pipes









compared to capital needs for pumps and treatment facilities. Labor is represented in the model

by the amount of total workers.

Even when the inefficiency or efficiency of a service provider may be due to its production

process per se, a firm can be favored or hindered by country specific circumstances. Indeed,

when considering several countries a major challenge is to appropriately account for each

country's political, social and economic differences. The empirical literature on cross country

studies for the water sector is limited. Clarke et al. (2004) utilize GNP per capital when analyzing

the impact of private participation in the water sectors of Brazil, Argentina and Bolivia. In the

electricity sector, Estache, Rossi and Ruzzier (2004) utilize GNP per capital on a DEA and

stochastic frontiers to account for countries' differences when assessing efficiency for South

America's main electricity distribution companies. Zhang et al. (2007) utilizes GDP per capital in

a stochastic model to assess the impact of privatization, regulation and competition on the

performance of utilities in the electricity sector of 36 developing countries over a period of 18

years. A study by Estache and Rossi (2008) considers GDP per capital among a set of covariates

to capture country particularities on a difference in differences analysis of the electricity sector

across 51 developing countries.

The World Bank utilizes the level of gross national income calculated by the Atlas method

to classify economies and set the lending eligibilities. The GNI adds to the GDP the income

received from other countries, less similar payments made to other countries. The Atlas method

uses a three year average of exchange rates to smooth effects of transitory exchange rate

fluctuations. The assumption in this study is that variations of country's GNI may impact the

performance of water utilities since this variable captures country specific economic









circumstances affecting production practices of these firms. Under a DEA framework, a way to

examine this influence is to include GNI as an "additional resource" to the utility.

Utilities in this region have different sizes so it is appropriate to account for firms' scale

when measuring efficiency. A variable returns to scale approach allows increasing or decreasing

efficiency based on the size of the firms. Alternatively, a constant return to scale approach means

that firms are able to linearly scale the inputs and outputs without increasing or decreasing

efficiency. The ratio of the efficiency value produced by a constant returns approach to the

efficiency value calculated from a variable returns to scale approach produces a scale impact

value. Table 2-7 shows results for the DEA technical efficiency values under the variable returns

to scale approach considering three options. Each option is based on utilizing a different set of

input factors. The first column corresponds to efficiency results when utilizing only labor. The

second column shows the efficiency result when using labor and network length and in the third

column efficiency results consider GNI. The last three columns of the table present the scale

effect for each of these model alternatives.

Overall, higher efficiency values are obtained when including network length and

subsequently adding GNI. The interpretation of results is based on the third column which

includes the three inputs discussed. Including all the input factors implies considering more

production characteristics which improves the quality of results.

Efficiency results on the third column of Table 2-7indicate that the service providers in

Panama, El Salvador, Nicaragua, and the small providers in Costa Rica, Guatemala and

Honduras are all 100% efficient. Efficiency is equal to 63% for the large provider of Costa Rica.

This means that this utility could produce the same output with approximately 63% of its

resources (labor, network length and GNI). This result also means that this provider is only 37%









inefficient. This value is obtained by subtracting 63% from 100%. According to values obtained

from scale impact calculations, 44% of the efficiency value for the utility in Honduras comes

from considering its size.

Figure 2-9 shows labor as the input factor on the x axis and number of connections as the

output on the Y axis. This graph shows only one dimension of the frontier since the other input

and output factors are not shown. Nevertheless, it gives an idea of each country relative position.

For instance, El Salvador, and Nicaragua produce a very similar amount of connections.

However the amount of labor differs widely. While El Salvador is found 100% efficient,

Nicaragua can work with 60% of its labor amount if considering labor the only input factor.

Costa Rica and Nicaragua levels of efficiency are very close (62% and 59% respectively).

Obviously, these efficiency results exclude a number of factors affecting production

conditions such as hydrology, total population, geography, topology, service quality levels, and

other elements affecting the production process. Nevertheless, these results provide a first cut at

evaluating relative performance when considering number of workers and size of the network to

produce water and water connections.

Further understanding a DEA analysis requires considering the efficient input and output

target levels for inefficient firms, which correspond to decreases in inputs and increases in

outputs with respect to the level of inputs and outputs of the respective peer group. The rationale

for this is that in some cases, the decrease in inputs is not enough to bring a company to the

frontier, so an increase in output is also necessary. According to DEA results, the large provider

from Costa Rica and Honduras need to reduce labor and increase actual volume of water.

In this DEA analysis, only data from 2005 were utilized. The Malmquist technical

efficiency change component for the period 2002 to 2005 is calculated for all firms except for the









large service provider from Honduras. The utilities from El Salvador, Nicaragua, and the small

utilities from Costa Rica and Honduras show no change in technical efficiency over the period.

The utilities from Panama and Guatemala show an increase in efficiency of 8% and 3%

respectively. The large service provider from Costa Rica shows a decrease in efficiency of 2%.

Stochastic Frontier

This section examines cost efficiency by statistically estimating cost relationships

according to input prices, given a level of output produced. The ideal framework would be to

completely specify a cost function including outputs, input prices and those specific factors

capturing possible cost differences among firms and countries. The limitation associated with

data availability restricts the analysis to the inclusion of only four explanatory variables in the

economic model.

The merit of performing this econometric exercise is to provide a general approximation

for the ranking of firms, not the level. In addition, this methodology recognizes the presence of

data errors which is an important element for the Central American analysis. The DEA approach

considers efficiency with respect to the best performers, given the variables selected whereas an

estimated cost frontier is a measure of central tendency considering all firms not just those on the

frontier.

The economic model is a cost function specified by volume of water billed (VolBil), price

of labor, price of energy, and GNI. Operational costs plus administrative expenses divided by the

number of connections comprise operating costs per connection which is the dependent variable.

All variables are normalized by the number of connections, to control for firm size. The price of

labor (LabPrice) is equal to total cost of workers divided by total number of workers. The price

of energy (EnegPrice) is total energy expenses divided by length of the network. The economic

model translates into Equation 2-1 (time and firm specific subscripts are omitted for simplicity).









I OliipCost = a + GNI + yVolBil + fLLab Pr ice + P/Eneg Pr ice (2-1)

In Equation 2-1, a, Vp, 7, FL, and fE are parameters to be estimated. All variables are in

natural logarithm form. The data set is an unbalanced panel for the period 2002 to 2005.

Empirical researchers often introduce a time trend in the model to capture possible technology

shifts due to technological changes. Given the short period of time plus the limitation on the

number of explanatory variables, GNI instead of a time trend is included to capture possible

economic shocks occurred over the period. Nevertheless, time is included in the estimation of the

cost frontier to explain possible changes of efficiency over time. The maintained assumption is

that these countries have changed efficiency behavior over time, independently. A stochastic cost

panel data frontier specified by the model in Equation 2-1 plus the idiosyncratic error term (v)

which is independently and identical distributed as N (0, o) and independently from

regressors is estimated. The only panel-specific effect is the random inefficiency term (u). The

frontier estimator assumes the effects (error term) are not correlated to the explanatory variables.

Inefficiency was allow to vary with time according to the Battese and Coelli (1992)'

specification for inefficiency. However, the coefficient for the parameter rl was not statistically

significant which implies that inefficiency has not changed over time and it does not follow an

exponential path. According to estimation results, the operating costs for the region are

represented by the Equation 2-2 (standard errors are in parenthesis).

UnitOpCost = (-5.3) 0.1* GNI +0.6 VolBil + 0.7 *LabPrice (0.1) EnegPrice (2-2)
(1.81) (0.08) (0.12) (0.12) (0.07)

The signs of coefficients are as expected and the statistical significance is high for the

output variable and price of labor indicating a high economic impact on costs. Energy price and

GNI are not statistically significant.









An increase of 1% in volume of water per connection produces an increase in unit costs of

approximately 0.6%. An increase in price per worker per connection of 1% produces an increase

in unit costs of approximately 0.7%. Price of labor provides the highest impact on costs from

relative to the effect of other variables. Increasing volume has a less than proportional increase

effect on costs. On average it indicates the presence of economies of scale in the region. Table

2-8 shows country ranks according to how far each firm is from the cost frontier and possible

reduction of operating costs by 2005. This reduction is calculated as the ratio of the estimated

inefficiency to actual operating costs.

The differences between results from the DEA frontier and the estimated cost frontier can

be explained by the fact that when calculating the DEA frontier we were looking at a contraction

of inputs for a given level of output. A cost frontier looks at minimizing costs given input prices

and output. When assuming the minimum set of inputs for a given level of output, the DEA

abstracts from other factors influencing the production process, such as the price of inputs

included in the cost frontier. However, Panama is been consistently the best performer for all the

methodologies. Performance differences for the other countries should be considered on a case

by case basis.

Concluding Observations

A major contribution from this study is the creation of a unique data base and a

comprehensive data collection process. Considering total water service coverage in the region

the data collected is very representative from a sector wide perspective. The quality of the data

set is good in the sense that it came from and it was reviewed and certified by data owners.

A major conclusion from this analysis points towards additional efforts for improving data

collection procedures in the region. Besides the scarce presence of information technology

limiting record keeping within these utilities, difficulties also may be due to the fragmented









service provision in some countries. A higher level of coordination is needed if data are to be

collected and trends analyzed. Such an initiative may require an analysis of stakeholders'

responsibilities regarding monitoring and storage of data. Coordination is needed among

stakeholders regarding what and how to collect data. Information technology is central to any

structured data collection procedure: the availability of an information system specific for the

sector is crucial for any data collection process within this region. Government, policy makers

and fund providers need to consider the role of technology improvement in the region.

In the process of identifying segments of the industry with no data, policy-makers,

regulators and managerial staff have been encouraged to expand efforts to seek disaggregated

data. Such data are necessary for further quantitative analysis, providing more complete

information regarding sector performance. This study is comprehensive, ranging from the

examination of the water sector structure characteristics of each country to individual firm

performance behavior under very specific scenarios.

A set of methodologies was utilized to assess sector performance, such as performance

indicators, total factor productivity analysis, production and cost frontiers. The role of

benchmarking is fundamental for assess the better allocation of resources. Making comparisons

based on individual indices is fraught with difficulty. Nevertheless, benchmarking techniques

that draw upon performance indicators is an important step towards improved understanding of

the efficiency behavior of water utilities.

Focusing now, on the results for countries in Central America, the relative small number of

sewerage connections compared with water connections may reflect costs restrictions in the area.

The study did identified different stages of the investment cycles for the group of countries.

Overall, system expansion seems to be in balance in terms of adding more connections or more









pipes to the network. Guatemala and Nicaragua are characterized by different behavior.

Guatemala seems to be adding more pipes and Nicaragua more connections. While the former is

trying to reach under-served areas, the latter is trying to satisfy local customers demand.

Investment for these two types of systems expansion is different. Adding more pipes require

higher levels of capital than adding only connections.

The amount of non revenue water is higher in Central American countries (55%) with

respect to that on Latin American countries (40%). This may be a consequence of poor metering

systems as it is reflected in a 56% metering value for these countries with respect to a 75% for

the sample of ADERASA Latin American countries. It may also be explained by the amount of

pipe leaks which is very high when compared to the same indicator for Latin American

countries.

Large firms have denser networks than smaller firms, reflecting the low investment

capacity of small providers. Guatemala shows a low level of coverage compared to its high

network density which is consistent with the extension of its network system through adding

more pipes as it was previously mentioned. In Honduras, the low coverage ratio and network

density implies that the expansion of the system should be through adding more connections to

satisfy under-served population in the area. Water consumption is higher in Central American

countries when compared to Latin American countries which reflect abundant sources of water

in the region. Labor inefficiencies are twice those found in Latin American Countries.

On the financial side operating costs per cubic meter of water are half the value found in

Latin American countries. Administrative expenses are found to be higher in small firms.

Changes in operating costs over the period are reflecting changes in energy costs, as in the case

of Panama, or labor costs as the case of Costa Rica.









When trying to assess changes over the period, calculating different total factor

productivity indices yield similar results given the short period analyzed. Panama is the only

country displaying an increase in productivity when considering labor and energy as input

factors over the period 2002 to 2005. The service provider for this country coincides in showing

an increase in efficiency as a result of calculating the Malmquist efficiency change component.

Nicaragua is the country within the group with a small decrease on efficiency change (2%).

When calculating a technical frontier for 2005 by means of the DEA methodology, higher

efficiency values are obtained from models that include GNI as an additional input factor. The

sensitivity of results to the inclusion of the GNI variable illustrates the importance of country

characteristics in explaining efficiency within the region. It was also important to consider the

scale of the service providers as it was reflected by the large scale impact values.

Results indicate that the service providers in Panama, El Salvador, Nicaragua, and the

small providers in Costa Rica, Guatemala and Honduras are all 100% efficient. Part of this

efficiency for El Salvador, Nicaragua and the small providers is due to their scale. The fact that

such a large number of firms are found to be 100% efficient indicates that the group of firms is

not heterogeneous enough in the sense that the set of input and output factors considered are not

sufficient to explain possible production differences among these countries.

Results from the stochastic cost frontier estimation indicate that when considering output

and input prices, the highest impact on operating costs per connection comes from labor prices.

Estimated firm's ranks according to how far is each firm from the cost frontier shows Nicaragua

as the most efficient firm. The service provider from Panama is ranking number 2 which

coincides with the fact that its service provider is positioned as the best performer from other

methodologies. The ranking of other providers need to be assessed on a case by case basis since









some results are contradictory. Data limitations affect model specification: not all elements

affecting cost are in the model.

For regulatory agencies, related government institutions, and funding agencies, this study

may contain additional information for their strategic planning and decision making processes.

This study should be viewed as a first step in the analysis of water utilities in Central America.

As data from additional years become available and more utilities provide information, analysts

will be able to conduct much more thorough analyses of sector performance. Hopefully, the

results presented here will serve as a catalyst for more comprehensive data collection/verification

initiatives in the region and for additional quantitative studies.










Table 2-1. Share of water coverage within country by 2005
Country Share of Coverage
Costa Rica 51%
El Salvador 94%
Guatemala 10%
Honduras 20%
Nicaragua 52%
Panama 66%


E Delivered

O Billed

o Lost


0 ILI I IPL joiELL J p^
Panama CostaRica ElSalvd. NicaraguaGuatemala Hond. CostaRica Hond. Guatemala

Figure 2-1. Average volume of water delivered, billed and lost from 2002 to 2005


== Water

-------Sew

----.- Wat w/Meters


PanamaCostaRicaElSalvd.NicaraguSuatemala Hond. CostaRica Hond. Guatemala


Figure 2-2. Average number of connections from 2002 to 2005


600

500


400
-o
d 300

S200

100

n\
















3,500 Total

c3,000 l \\ ith \\ater Ser\

S2,500

2,000

1,500

1,000

500

Or
Panama CostaRica ElSalvd. NicaraguzGuatemala Hond. CostaRica Hond. Guatemala

Figure 2-3. Average population with water service and total population in the area from 2002 to
2005






0 WaterConx 0 NetLeng


100% -


75%


50%


25%


0%
Panama CostaRica ElSalvador Nicaragua Guatemala Honduras

Figure 2-4. Average changes in number of water connections and network length from 2002 to
2005
















Coverage Network
Network
Coverage Density
100% .----Network
100%0
Density 140

80% *- -- 120

S- ,100
60%- 80


40% 60

S40
20%-
20

0% 0
Panama CostaRica ElSalvd. NicaraguaGuatemala Hond. CostaRica Hond. Guatemala


Figure 2-5. Average coverage and network density from 2002 to 2005



$/
Connection 200

250 OpCost/Conx
I
--- Consumption ,
200 150


150 ,
100

100

50
50


0 0
PanamaCostaRicaElSalvd.Nicaragufiuatemala Hond. CostaRica Hond. Guatemala


Figure 2-6. Average volume of water billed and water consumption from 2002 to 2005














Table 2-2. Operational performance indicators by 2005
Vol Vol Num Vol Pop
Met
Del Lost Conn Del Serv
Mill Per
Country/Units m3 % Miles % Person Miles
Panama 452 58 448 41 126 2,004
Costa Rica-L 305 49 457 94 76 1,978
El Salvador 259 619 55 84 3,093
Nicaragua 257 44 457 48 39 2,870
Guatemala-L 122 55 195 84 81 1,045
Honduras-L 75 63 105 35 67 707
Costa Rica-S 28 55 48 97 66 228
Honduras-S 10 44 11 77 157 53
Guatemala-S 7.6 10 56 183 42


Net Net
Cov
Length Dens
Conn/
% Km km
92 4,727 95
99 6,437 71
90 4,391 141
91 4,604 99
93 5,013 128
69 2,800 38
100 678 71
71 77 144
72 232 42


$/
Connection

200



150



100



50



0 -


-- OpCost/Conx
- *- NetwDen


Connection
/km

150




100




50




0


Panama CostaRica ElSalvd. NicaraguaGuatemala Hond. CostaRica Hond. Guatemala


Figure 2-7. Average operating cost per connection and network density from 2002 to 2005


work!


work/
1000
Conn










$/m3
0.30-
0.25--
0.20-
0.15

0.10
0.05
0.00 r


IIaI


PanamaCostaRicaE1Salvd.Nicaragu6uatemala Hond. CostaRica Hond. Guatemala


Figure 2-8. Average operating cost per cubic meter from 2002 to 2005


Table 2-3. Summary of finance indicators-average values from 2002 to 2005
Country OpCost LaborCost EnergyCost AdmExp FinExp OpCost/Connection
Units $/m3 $/m3 $/m3 $/m3 $/m3 $
Panama 0.10 0.04 0.04 0.03 0.004 103
Costa Rica-L 0.14 0.07 0.02 0.07 0.041 91
El Salvador 0.05 0.01 0.01 0.03 0.006 21
Nicaragua 0.11 0.07 0.06 0.04 0.027 61
Guatemala-L 0.22 0.07 0.12 0.03 0.034 138
Honduras-L 0.28 0.11 0.03 0.00 0.000 201
Costa Rica-S 0.09 0.02 0.02 0.06 0.022 51
Honduras-S 0.10 0.03 0.02 0.01 0.000 89
Guatemala-S 0.08 0.05 0.06 0.03 0.000 66


Table 2-4. Percentage of change values from 2002 to 2005
Country VolW Pop Number Net Cost of Energy Admin Op
Deliv Served Connect Length Work Costs Exp Costs
Panama 17% 6% 24% 26% 2% 54% -22% 18%
Costa Rica-L 2% 27% 17% 13% 54% 15% 251% 25%
El Salvador -11% 7% 7% 0% 8% 2% -18% -10%
Nicaragua 11% -4% 14% 5% -1% 14% 17% 4%
Guatemala-L 1% 24% 4% 10% 1% 16% 95% 18%
Honduras-L 0% 0% 0% 0% 0% 0% 0% 0%
Costa Rica-S 12% 10% 10% 7% 16% 109% 45% 29%
Honduras-S 14% -4% 30% 38% 57% 18% 297% 315%
Guatemala-S 0% 0% 0% 0% 0% 0% 0% 0%









Table 2-5. Summary of quality indicators- average values from 2002 to 2005
Quality Continuity Complaints Leaks
Country/ Units /Connect /km
Panama 69 21 0.03 6
Costa Rica-L 98 24 0.70 3
El Salvador 90 0 0.20 10
Nicaragua 100 20 0.24 3
Guatemala -L 100 9 0.01 5
Honduras-L 100 7 0.13 0
Costa Rica-S 99 24 0.30 22
Honduras-S 98 24 0.13 12
Guatemala-S 0 22 0.11 5


Table 2-6. Total factor productivity indexes from 2002 to 2005
Index Laspeyres, Paasche &Fisher TFP Tornqvist

Output Variable Volume #Connect Volume #Connect
Panama 51% 53% 31% 32%
El Salvador -17% -6% -9% 0
Nicaragua -5% 1% -5% -2%



Table 2-7. DEA technical efficiency and scale impact on efficiency for 2005
Variable Scale Impact on
Returns Efficiency
Inputs Labor Labor Labor Labor Labor Labor
Netlength Netlength Netlength Netlength
GNI2005 GNI2005 GNI2005
Panama 1.00 1.00 1.00 1.00 1.00 1.00
El Salvador 1.00 1.00 1.00 0.73 1.00 1.00
CostaRica-L 0.62 0.62 0.63 0.60 0.92 1.00
Nicaragua 0.59 0.70 1.00 0.59 0.99 1.00
Guatemala-L 0.52 0.85 0.85 0.72 0.98 1.00
Honduras-L 0.34 0.38 0.99 0.99 0.97 0.44
CostaRica-S 1.00 1.00 1.00 1.00 1.00 1.00
Guatemala-S 1.00 1.00 1.00 0.62 0.62 0.62
Honduras-S 1.00 1.00 1.00 0.39 1.00 1.00












Number
Of
Connections
(Thousands)


700

600

500
El S
400 -

300 -

200 Guatel

100 Costa Honduras
Costa Rica

0 1000 2
0 1000 20(


Panama
a


Costa Rica
Nicaragua
alvador


mala


)0 3000


4000


Number of Workers

Figure 2-9. Number of connections and workers by 2005


Table 2-8. Efficiency rank and operating cost reductions

Country Frontier Rank Cost Reduction in 2005

Nicaragua 1 2%
Panama 2 2%
El Salvador 3 3%
Costa Rica-S 4 3%
Guatemala-S 5 4%
Costa Rica-L 6 3%
Guatemala-L 7 12%
Honduras-L 8 14%









CHAPTER 3
SECTOR FRAGMENTATION AND AGGREGATION OF SERVICE PROVISION IN THE
WATER SECTOR

The policy decisions issued in 1995 by the government of Peru pointed towards

restructuring the water sector by means of service decentralization to municipalities. The aim

was to give more decision-making responsibilities to those with more information about costs

and production conditions (Tamayo, Barrentes, Conterno and Bustamente 1999; Corton2003). In

its 2002-2006 management report, the Peru water regulatory agency (SUNASS) identifies key

elements of the current situation of the water and sanitation sector: political interference in water

service providers' decisions, sector fragmentation, the lack of sector investment, and tariff

stagnation (SUNASS 2006). The sector fragmentation issue is addressed in this study.

Sector fragmentation refers to the presence of a large number of very small service

providers. Of the 49 service companies comprising the sector, 10% provide service to less than

1,000 connections and 39% have between 1,000 and 10,000 connections. Large utilities serving

more than 40,000 connections represent 16% of the sector and those utilities serving between

10,000 and 25,000 connections represent 25% of the sector. The average number of

municipalities served by large companies is 17. Medium size companies serve on average 5

municipalities and very small firms provide service to one municipality. Beyond the financial

sustainability of these companies, a key aspect is that these smaller entities will not always have

the fundraising capabilities to invest in the water networks systems. Consequently, their

attraction to private investors may be limited.

SUNASS argued that aggregation of small service providers may enable the achievement

of economies of scale, providing service to a larger customer base at a lower cost (and/or with

higher quality). Moreover, professional capacity in a large scale operation may be enhanced









together with other potential production and service efficiencies. These efficiencies result from

joint administration and operation, which renders the sector more attractive to private investors.

The objective of this study is to analyze the structure of the water industry in Peru, to

investigate the presence of economies of scale and cost efficiencies considering the region where

the firm is located and specific production characteristics. The aim is to provide policy makers

with information about whether aggregation of service providers is optimal based on their

location and specific characteristics. The degree of scale economies will help to determine the

merits of aggregation of small service providers. If there are no economies of scale, splitting

large service providers by localities of service may be an optimal decision from a sector structure

point of view. Otherwise, aggregation is recommended.

All companies for this sector provide sewerage service so cost efficiency is examined

considering the volume of water billed and the number of sewerage connections as outputs of the

production process. Volume of water loss is included as an additional output. Water losses

originate from pipe leaks and non authorized connections. In the water sector of Peru, volume of

water loss represents 47% of volume of water produced. The intuition behind the inclusion of

this variable as an output comes from assuming that it is hard for firms to control commercial

losses if that entails denying the service to the poorest segments of the population. In this sense

water lost is interpreted as a valued output by consumers. In addition, producing water has no

cost for the firm other than the marginal cost of treatment. Managers consider a tradeoff between

repairing pipe leaks and increasing water produced when satisfying water demand. Under this

framework, the hypothesis is that water service providers jointly produce water lost and billed.

The input factors considered are labor and capital.









The study utilizes a data set of 44 water service companies, for the period 1996 to 2003.

Particular attention is given to the modeling process to ensure accuracy on capturing the behavior

and production conditions of these utilities. Considering a long run scenario for the analysis,

where all factor inputs are variable, the basis for the economic model is a total cost

translogarithm functional specification. This functional form gives flexibility to the production

technology not imposing constant returns to scale, allowing efficiency to vary with scale.

Production practices differ according to the region where the utility is located. Technology

shifts may occur as a result of variations in national investment and GDP, as well as time which

represent possible technological changes in the sector. A set of specific characteristics to the

utility production process are also included such as network density, the number of active

connections, continuity, and the number of municipalities served. The econometric approach is

based on the estimation of a stochastic frontier where inefficiency is allowed to vary over time.

The intuition is that the time period under analysis is long enough to expect that firms may have

modified their cost efficiency behavior.

Results from the analysis found the production technology not homothetic implying that

the input mix varies with the scale of the firm. In addition, the production technology was found

to be no homogeneous which means that returns to scale varies with the scale of the firm. These

findings implies that any merging or aggregation of these firms require detailed analysis of the

particular production characteristics of each firm. Utilities in the forest region and small firms

on the coast exhibit economies of scale which support the aggregation process. The introduction

of competition is suggested in the mountains and the coast.

Industry Framework

Following Garcia and Thomas (2001)'s analysis of the French water industry the water

sector production process is characterized by five main functions. The first function, production









and treatment, covers operations for water extraction, either from underground or surface, and

preliminary treatment (disinfection, filtering, softening). In the case of Peru, water as a

commodity has no cost for the company and the possibility of water purchases is not considered

in this analysis. Thus, costs incurred while performing this function comprise only water

extraction and preliminary treatment.

The second step is to transport water from production facilities through transmission

pipelines. This can be done by means of pumping or by gravity. Costs from performing this

function are proportional to the amount of pumping required. Storing water in facilities such as

water tanks or towers is the third function. The fourth production process is comprised of

pressurization of water from the stocking facilities into pipelines, which is done by installing

pumping stations. Finally, distribution of water tofinal customers through distribution mains and

customer service lines completes the production process. These five functions are all performed

by each service provider in Peru. Thus, these utilities are vertically integrated.

Each service company is located within one of Peru's geographical regions: mountains (19

companies), forest (7 companies) and coast (18 companies). The water source found in these

regions, surface or underground, requires a different production technology. In general, surface

water does not need pumping but may need a more intense water treatment to meet quality

standards required before distribution. In this study, the region in which each company is

located identifies production technology differences and they are included in the economic

model.

Volume of water sold to different types of customers, population served and the number of

connections are among the most generally identified outputs in the water industry literature.

Recent empirical studies indicate that authors view service companies as producing other goods









besides the classical output, volume of water delivered. Renzetti (1999) identifies water sold to

residential and non-residential customers in the analysis of cost supply and pricing practices of

the water and sewerage utilities in Ontario. Saal and Parker (2000) utilize the resident water

supply population served in their study of the England and Wales water sector. Garcia and

Thomas (2001) define volume of water loss as a product, in addition to volume of water

delivered in the French water sector. Aubert and Reynaud (2005) utilize volume of water

delivered and the number of customers served in a variable cost function for the Wisconsin water

sector. Saal and Parker (2004) consider water volume delivered and the number of connections

in a study of the England and Welsh water industry.

Neuberg (1977) introduced the "separate marketability of components" as a necessary (but

not sufficient) condition to define the vector of outputs. This issue, followed by several authors

in the electricity sector, has not been explored in the water sector (Hattori, Jamasb, and Pollit

2005; Lowry, Getachew and Hovde 2005; Jamasb and Pollit 2003; Hattori 2002 and Rossi 2001).

Neuberg's argument is that identification of outputs needs to take into account what the company

is actually selling.

Following this line of thought, and according to the water production process previously

described, volume of water billed and lost as well as sewerage connections are the outputs

identified in this analysis. Between 1996 and 2003, water customers in Peru were charged

according to volume of water usage. Water billed is the final output to these customers and it is a

main driver of costs for service providers. Water received by end customers have gone through

the entire production process of these companies, from production to distribution so the final

product, "water billed" has costs associated to each segment of this industry. Volume of water

produced is not a cost driver in this industry given that costs associated to production are only









marginal. Water produced is not the final product to be sold but rather it represents an input to

the provision of water service.

Regarding water lost, two important elements are addressed to support its consideration as

an output in this sector. First, in this particular industry where water utilities operate all the

segments of the service production and producing water implies only the marginal cost of water

treatment, the extent of water losses partially reflects a cost tradeoff between increasing water

production and repairing network leaks to keep up with water demand. Managers may find it

more costly to repair leaks than increase water production to satisfy demand. Pipe leaks require

costly maintenance outlays, particularly on long or dispersed networks.

Alternatively, when considering the distribution network, water losses are related to non

authorized connections. Some authors refer to this issue as water theft or commercial losses,

hence the "non revenue" water terminology. Given the characteristics of this country where a

large portion of the population lives in poverty conditions1 it may be hard for water utilities to

control water lost, if that entails denying the service to the poorest segments of the population. In

this sense, water lost may be valued as an output by a significant segment of the population. In

addition, managers, and in particular the members of the company board, may find it more

advantageous to keep water needs of this population segment satisfied if that secures wining their

votes during municipal elections.

Regarding the measurement of the water lost share due to costs tradeoffs and which

corresponds to non authorized connections, data available on pipe leaks is not reliable. Counting

the number of leaks require either costly electronic equipment to measure the flow of water or a

maintenance routine to manually do the counting which requires staff and vehicles. None of

1 A level of 48.7% of poverty by 2005 is reported by the World Bank: http://go.worldbank.org/AHUP42HWRO (last
visit: 05/25/2008)









these mechanisms are reported as used by these water utilities. Data on non authorized

connections is not available. Still, some water may actually be wasted which from a welfare

perspective the waste of a natural resource is a loss. However, from the firm's perspective it

seems that benefits from the costs savings originated from not repairing the leaks as well as the

benefits from securing the votes from a particular segment of the population are high enough to

keep them producing water losses. In the Peru water sector the median volume of water lost is

50% of volume produced. To put this number in perspective, water utilities from developed

countries have 15% of water volume lost. For Latin American countries this value equals 60%

and 35% for developing countries (Kingdom, Liemberger and Marin 2006). Two leading studies

on this issue are Garcia and Thomas (2001) who found a tradeoff between water produced and

network leaks in the French water industry and Antoniolli and Filippini (2001) that included

percentage of water loss as a firm specific characteristic in the analysis of the Italian water

industry.

Under this framework, the joint production of water loss and water billed is considered in

this analysis. Water lost is calculated as the difference of volume of water produced and billed.

In addition, all service companies in Peru offer sewerage services. Sewerage production

accounts for a high portion of a service's costs, but it is not charged separately from water. For

each 100 water connections a median firm serves 85 sewerage connections. Thus the number of

sewerage connections is identified as an output for this production process.

With respect to water quality, it could be viewed as an output of the water service

provision process depending upon the country and the value consumers place on it compared to

the provision of the usual outputs in this sector: water and sewerage. Without doubt, in

developed countries where water and sewerage coverage rates are almost 100%, water quality is









highly valued by final customers and its production implies high costs given the high quality

standards in these countries. In the case of Peru, by 2004 the population was 27.5 million people

of which 74% were located in urban areas. Of this urban population, 75% is within the scope of

service of water service companies. On average, service companies deliver water to 84% of the

total population in their area of service. From the perspective of customers in these countries, it

is reasonably to assume that quality of water is of less value comparing to water or sewerage

service. In addition, water quality standards are still basic when compared to those of developed

countries. The costs associated to meeting these standards are marginal compared to other costs

in the production process such as lying pipes or installing water pumps.

A variable that represents a measure of quality of service is continuity. This variable is

defined as the average number of hours a day customers receive water service over a year. This

characteristic is related to water pressure within pipes, which is determined by the placement of

compressors along the network, the volume of water, and its flow within pipes according to

differences in gravity. When the source of water is located in the mountains or hills, gravity

promotes water flow. When serving customers in a flatter geography, the amount of required

pumping is higher. Costs will be higher for firms located in the coast, compared to costs for

firms in the mountains or the forest when maintaining the same level of continuity. This variable

is considered in this analysis not as an output but as a characteristic of each utility service

provision.

Finally, SUNASS classifies service companies according to the number of water

connections, as large, medium and small. Large companies are those with more than 40,000

connections (8 companies); small companies are those with less than 10,000 and more than 1,000

connections (19 companies); medium size companies are those with number of connections









between 10,000 and 40,000 connections (17 companies). In this study, each firm is classified as

belonging to one of these groups according to the company's number of connections in 1996.

The ratio of active connections to total connections is interpreted in this analysis as an

indicator for service/maintenance efficiency. Its median value was 84% in 2003. A connection is

a piece of equipment subject to functional problems and wearing out. Connections are classified

as inactive when they are not able to deliver water to the customer. This occurs when there is a

functional problem, which once fixed allows the connection to be active again. Another possible

reason for a connection to be inactive is suspension of service due to non payment, but this is not

the case in Peru. If adequate maintenance is provided on a regular basis, all connections are

supposed to be active during their operating life. If keeping all connections active is costly, firms

may prefer to maintain a certain amount of inactive connections to reduce costs. Thus, including

this variable allows one to examine the impact on costs of keeping connections inactive

reflecting maintenance inefficiencies.

Model Specification and Data

This analysis assumes that the firm acts to minimize its long run costs where all input

factors are free to adjust and output is exogenously determined by the obligation to supply

customers. The assumption that managers are cost minimizers may seem strong in the context of

this sector. However, the length of the period under consideration might help to smooth out any

differences from managerial objectives. Municipal leaders stay in power usually for one year.

Network length is commonly used as a proxy for capital in empirical work. The rationale

for doing this is the high amount of capital necessary to lay down pipes compared to the capital

needs for other types of network system developments such as pumps or treatment facilities.

Table 3-1 depicts change in network length for each different group of firms illustrating a

considerable expansion of the network systems in this sector.









Peru is not an exception regarding the difficulties Latin American companies face when

calculating price of capital. Finance expenses plus depreciation divided by length of network is

considered as a proxy for price of capital in this study. An increase in a firm's debt is assumed

to occur when the company expands its network length. Further, finance expenses are assumed

to reflect this increase and subsequently the network expansion. This variable will capture

characteristics related to managerial debt financing abilities and the possible value of investing in

network developments. Nevertheless, the proxy for capital utilized is under-representing the

relative risk of investing in this sector as opposed to in other markets.

Labor costs represent approximately 60% of firm's total cost. For the purpose of the cost

analysis, labor is classified into two types according to the contractual obligations acquired by

the utility at the time of hiring. The direct labor type is comprised of workers who have

permanent positions in the company and are entitled to the company's labor benefits. The

indirect labor group is comprised of workers under explicitly limited terms of employment,

regarding time and salary. This group has no firm working benefits other than the monetary

amount agreed at the initial time of work contract or agreement. Prices for direct and indirect

labor are calculated as annual labor costs divided by the number of workers for each case

assuming that the number of hours employed by indirect and direct workers on average is

similar. Table 3-2 shows summary statistics for the identified outputs and price of inputs.

Total cost is comprised of sales cost, sales expenses, administrative expenses, finance

expenses and depreciation representing total operating expenditures. It is assumed that

accounting definitions adopted by all firms in the sample are the same and that the depreciation

value is based on a similar estimation procedure for the assets in place. The median value for









total annual costs in current US Dollars2 for large utilities is $16 whereas it is equal to $2 and $

0.4 for medium and small firms respectively.

A transcendental logarithmic specification provides a second-order approximation for the

cost functional form. Its advantages are that it places no a priori restriction on substitution among

factors of production and that it allows scale economies to vary with level of output, not

imposing homotheticity or unitary elasticity of substitution. These conditions are tested at

estimation time. Three outputs and three input factors are considered in this cost analysis. The

use of a translog functional form is appropriate because it readily allows for the identification of

scope economies among the set of outputs by observing the sign of the coefficients for the

outputs interacted terms. The coefficients of these terms measure the increase in marginal cost of

one of the outputs when the other is increased. It is of interest in this study to examine possible

scope economies between water delivered and lost and between sewerage and water. A finding

of diseconomies of scope between any of these outputs suggests that the production process need

to be re-assessed possibly separating the production of the output involved. However, a

weakness of this specification is that it assumes symmetry between any pair of outputs, which

may not necessarily apply here.

In addition, in a translog functional form specification the coefficients for the interacted

price terms identify the marginal cost of increasing one of the inputs as the price of the other

increases. If the coefficient is positive, it indicates possible substitutability. Otherwise, it signals

complementarities among the input factors. In this study it is interesting to examine possible

substitutability or complementarities between capital and labor. Findings in one direction or



2 Exchange rates obtained from Peru's Central Bank web site; http://www.bcrp.gob.pe (last visited 05/24/2008)









another may suggest over-usage of the input factor involved. Examples of a translog functional

specification in the water sector are Aubert and Reynaud (2005) in analyzing the Wisconsin

water utilities; Saal and Parker (2004) in the English and Welsch water industry; Garcia and

Thomas (2001) for the French water sector; Fabbri and Fraquelli (2000) in the Italian water

industry and Asthon (1999) in the English and Welsch water industry. Omitting firm and time

subscripts for clarity, the economic model is specified in Equation 3-1.


lnTC=a+a,, lnYm ,+ lnI+ 7,,, lnYm ln4
m j m n
1 1(3-1)



In Equation 3-1, TC is total costs defined before; Ym is the vector of the m outputs already

identified (volume of water billed-Y1, the number of sewerage connections-Y2, and volume of

water loss- Y3); Pi is the vector of i input prices previously identified (P1, P2 are prices for direct

and indirect labor respectively and P3 is the price of capital); and the a's and y's are parameters

to be estimated.

According to Equation 3-1, a firm's costs are represented by a single functional form

common to all firms, so there are no systematic differences in technology among them. To

account for possible technology differences, dummy variables representing the region where the

firm is located are included in the first order price inputs and output coefficients in Equation 3-2.

aa = a+ + afR1+ af3R3; a, = bo + b R1 + bR3 (3-2)


R1 takes the value one if the firm is located in region 1 (Mountains), and zero otherwise.

R3 takes the value one if the firm is located in region 3(Coast), zero otherwise. The region forest

(R2) is captured in the intercept, which in this case corresponds to the coefficient of the first

order outputs and price of inputs. This selection does not affect the results. The a's and b 's are









parameters to be estimated and represent deviations from region 2 taken as a base. This

functional specification focuses on how outputs and price of inputs on the coast and the

mountains differ from those in the forest region.

In characterizing the production process for the Peru water and sewerage sector, several

variables specific to each service company have been identified (environmental variables):

continuity of service (CONTINUITY) and maintenance efficiency represented by the ratio of

active to total connections (SMAINTENANCE). The number of districts served by each

company (NDISTRICTS) was assumed to represent possible political factors reducing costs

according to findings from Corton (2003). Disagreements or agreements among province and

district municipalities' leaders were assumed to have an impact on the number of localities

served by a company. Alternatively, the number of localities may capture the effect of having to

serve a larger area. Additional localities may need additional local offices, staff and maintenance

service, which may represent increased costs. This variable is also included in the present

analysis.

Network density (NDENSITY) equals the number of total connections divided by the

length of the network. The length of the network includes transmission and distribution pipes up

to the customer connection. Water companies with a similar size, as measured by number of

connected properties may have different costs because of network differences, such as its length

and type of customers. Larger firms could have lower costs due to greater network density rather

than a scale economy benefit per se. The bigger is the value of this variable, the denser the

network (more connections per km of pipe). As network density increases, average costs are

expected to decline considering that fixed costs are spread over a larger amount of customers

(connections). This variable distinguishes possible firm's gains due to scale economies from









those related to the network characteristics. It is important to notice that in this analysis all

customers (and connections) are assumed to be residential.

All these environmental variables vary with the region in which the utility is providing

service. They are introduced in the economic model interacted with the region dummies, in the

same way inputs and outputs are included. Table 3-3 depicts descriptive statistics for these

variables.

Finally, besides considering a time trend to capture possible technological changes over

time, gross domestic product (GDPv) and national investment (NIv) variations are included in

the model to capture specific country characteristics related to annual economic circumstances.

The assumption is that technological changes occurring between 1996 and 2003, which are

captured by a time trend variable, may have a different impact on sector's total costs than the

possible effect of the annual economic changes captured by these variables. After 2001, GDP

annual percentage of change increased, yet national investment level for 2000-2005 was only one

third of that in the earlier five year period. The government no longer had the capacity to fund

infrastructure projects so it is plausible to believe that firm's costs were affected. For instance,

firms had to raise capital to continue system expansion which may have had measurable effects

on utilities' total costs.

In this analysis, the environmental variables identified are included in the intercept so they

explain cost shifts, rather than technology shape or inefficiency. This choice translates into the

following specification for the intercept of the economic model defined in Equation 3-1.

a = a +QoTP +olnGDPv, +( InNIv +E 2 InZ, + n I nZ,,Rl+ lnnZR3 (3-3)
I I I









In Equation 3-3, ao is the intercept common to all firms; Tis the time trend; GDPv, NIv are

time varying common for each firm; Z is the vector of identified environmental variables; R1 and

R3 are the dummies for the mountains and coast regions, respectively.

Empirical Results

The model specified in Equations 3-1, 3-2 and 3-3 is estimated as a stochastic cost panel

frontier. To address the possibility of inefficiency changing over the analyzed period, time is

included in the estimation process by following Battese and Coelli's (1992) functional form

specification for the inefficiency component, (u, ) as it was explained in Chapter 1. For purposes

of estimation, the usual idiosyncratic error term (v) is added to the economic model. This error

term is independently and identical distributed following a normal distribution with zero mean,

and it is independent from the regressors. Estimation proceeds using a balanced data set

consisting of 42 companies comprising 336 observations.

Regularity of the Cost Function

Table 3-4 shows results from estimation. A well behaved cost function is concave in input

prices and non-decreasing in outputs. Assuming the cost function is twice continuously

differentiable, a necessary and sufficient condition for it to be concave in prices is that the matrix

of second order partial derivatives of the cost function with respect to prices be negative semi-

definite. In the case of the translog flexible functional form, for this to hold it is necessary to

impose symmetry on the parameters of interacted price terms. This is accomplished by applying

7YrY = 7rY for y Yn YmYm and y, = 7,, for iJj. In addition, following Diewert and Wales

(1987), the price shares need to be positive over the price domain. These conditions hold for all

regions.









A cost function must be homogeneous of degree one in prices to correspond to a well-

behaved production function. This implies that for a fixed level of output, total cost must

increase proportionally when all prices increase proportionally. This translates into applying to

the parameters of the model the conditions specified in Equation 3-4.


-'= 1, -y =0, 2", 1= ", = 2"", = 0 (3-4)


Imposing these restrictions is equivalent to normalizing prices and total cost by one of the

prices. In particular this is done by dividing total costs, price of direct labor and price of capital

by the price of indirect workers (P2). The selection of the price for normalization does not alter

the results.

A cost function corresponds to a homothetic production technology if and only if the cost

function can be written as a separable function in output and factor prices. This implies assuming

that the input mix is constant with scale. A homogeneous technology is a special case of a

homothetic technology when the elasticity of cost with respect to output is constant.

Homogeneity implies that returns to scale are invariant to the production mix and scale of the

firm. For the case of the translog model, the homotheticity and homogeneity conditions are tested

using the Likelihood Ratio test after imposing the restrictions specified in Equations 3-5 and 3-6

(Diewert 1974).

Homotheticity requires: y, = 0 (3-5)

Homogeneity in outputs requires: y = 0; y, = 0 (3-6)


The sequence for testing these restrictions follows Christensen and Greene (1976). A chi-

squared one-sided upper tail test at significance level of 0.001% rejects the null hypothesis of









homotheticity and homogeneity. The highly statistically significant coefficients for the

interacted price terms evidences that unitary elasticity does not hold in this data set.

Efficiency

For the frontier estimator the likelihood function is expressed in terms of the variance

parameters, a2 = v 2 + u, 2 and gamma, y = a, 2 2. The closer the value of gamma is to one,

the more inefficiency as opposed to noise accounts for explaining the model disturbance term

variance. A gamma value of 0.41 indicates that the presence of noise (or unobserved firm

specific characteristics) is still important in this data set.

The coefficient for q, the parameter included in the inefficiency specification is positive

and highly statistically significant. This indicates that the exponential specification imposed to

the inefficiency term is appropriate for explaining its behavior for the service providers in this

sector. The mean value for inefficiency, u,, is statistically significant at a 95% level, which

indicates that inefficiency does have an important explanatory role in determining costs in this

sector. It is worth noting that this term may contain unobserved time variant firm specific effects

not captured by the variables specified in the economic model.

Figure 3-1 illustrates the movement of firms from 1996(dashed line) to 2003 (solid line)

with respect to the optimal frontier represented by the horizontal line y=l (the closer the value to

1 the more efficient is the firm). On the horizontal axis firms are sorted by size and the vertical

axis represent technical efficiency values. Table 3-5 shows statistics for the extent to which

minimal input usage differs from actual input usage in 2003, represented by the estimate of

minus the natural log of the technical efficiency via E (uit eit)). Small companies are slightly

more efficient than large size companies. The most efficient small firms are located in the coast.

Firms located in the mountains are similarly efficient regardless of their size.









Table 3-6 shows cost reduction statistics calculated as the ratio of actual cost less predicted

cost divided by actual cost. If firms were to behave efficiently with respect to their predicted

costs for 2003, the reduction of costs found for small firms is on average of 4%; for large firms

this reduction is 16% and for medium size firms it is 17%. There is a noticeable difference in

reducing usage of resources for large firms depending upon their location. The only large firm

located in the forest should reduce by double the amount of resources with respect to the

reduction amount for large firms on the coast. There is also a noticeable difference in the amount

of cost reduction among firms located on the coast. Small firms are able to reduce their costs by

50% less than the amount large firms would reduce theirs. At the same time, large firms on the

coast are able to reduce their cost by 50% less than medium size firms would reduce theirs.

Environmental and Output Variables

Turning to the interpretation of coefficients, neither national investment nor GDP

variations are statistically significant. Time is not statistically significant either. A possible

explanation for this may be the fact that monetary variables converted to current US Dollars may

have captured country economic fluctuations. An alternative model using soles, the currency of

Peru, instead of US dollars yields the time coefficient statistically significant at a 90% level. The

coefficient is positive but very small in magnitude indicating a small economic impact. Still in

this model, neither GDP nor national investment variations are statistically significant

With respect to the number of districts (NDISTRICTS), Corton (2003) previously reported

cost savings for firms serving more than 5 districts. In the present analysis, this variable is not

statistically significant. An alternate model was estimated, replacing NDISTRICTS by a dummy

which takes value equal to one if the firm serves more than 5 localities. Estimation results were

similar. Further analysis with respect to the provision of service according to the number of

districts and provinces is included in Chapter 4.









The coefficient for the ratio of active to total connections (SMAINTENANCE) for firms at

the coast in comparison to those in the forest or the mountains is statistically significant. A 1%

increase in this variable (which implies an increase in the number of active connections)

produces a decrease in costs of approximately 0.5% only for firms located at the coast. This

variable was included as an indicator of maintenance efficiency assuming that keeping

connections active has an impact on total costs. An explanation for firms on the coast to have

lower costs as they keep a higher level of connections active implies that these firms are more

efficient compared to firms in other regions from the view point of maintenance. It may be the

case that these firms have maintenance schedules under control such that replacing costly

connection equipment are kept to a minimum. It is feasible to believe that the smaller territory

and a not challenging topography compared to the mountains and the forest provide a

comparative advantage to these firms when it comes to maintaining the operational level of the

network.

The network density (NDENSITY) coefficient is negative and statistically significant at a

99% level. Differences among regions do not have an impact on network density. An increase of

1% for network density produces a decrease in costs of approximately 0.4%. The possible effect

of a denser network on average costs is clear. On this total cost model this variable is capturing

the scale of the network. In this sense the economies of scale results will be net of this network

dimension. The negative coefficient for this variable implies that firms are still in the position of

adding more connections to their networks, a reasonably result considering the under-service

issue in this sector.

Differences in continuity (CONTINUITY) among regions are statistically significant. The

negative signs on coefficients for the coast and mountains mean costs savings when continuity is









increased independently of the topographic conditions of these two distinct regions. A possible

interpretation for this is higher continuity implies higher efficiency and consequently lower

costs. If efficient firms are those delivering higher continuity levels, then efficient firms exhibit

lower costs. Overall, there may be other factors interacting with the possible effects of this

variable in addition to the topographic aspects considered here. An alternative model excluding

continuity was estimated. The differences in the magnitude of the coefficients were small. The

signs of the coefficients and statistical significance of the variables did not change.

The first order coefficients for volume of water billed (Y1) are not statistically significant.

The coefficient for the interacted term with volume of water lost (Y1Y3) is statistically

significant at the 99% level. This finding indicates that the volume of water billed does not

explain costs by itself, but rather it does so by being jointly produced with volume of water lost.

The coefficient is positive indicating diseconomies of scope when producing these two outputs

together. A 1% increase in delivering water billed and lost jointly implies an increase in costs of

0.1% for the firm. This is the cost may be interpreted as the cost the firm pays for allowing non

authorized connections to exist.

On the other hand, the coefficient for volume of water lost (Y3) is negative and statistically

different from zero. The negative coefficient and its magnitude explain why firms allow water to

be lost. Producing water lost reduce costs by approximately 1.3% for each 1% its volume

production is increased. This result can be interpreted as the cost savings from not repairing the

leaks. In this way, the cost tradeoff hypothesis is supported. The net effect from the cost tradeoff

and the diseconomies of scope yields costs savings around 1% for each 1% increase in water

lost.









Two alternate models were estimated to check the sensitivity of this model. Volume of

water produced instead of billed not including water lost yields a much larger residual. A large

residual indicates the presence of missing explanatory variables. In this case it suggests the

explanatory importance of water lost in the model. In addition, none of the coefficients for water

produced is statistically significant indicating that this variable has not explanatory power on

costs. Another model was estimated including water lost as an environmental variable instead of

an output. A chi-squared one-sided upper tail test at significance level of 0.001% rejects the null

hypothesis of not including Y3 interacted with the other outputs and input factors.

Regarding the number of sewerage connections (Y2) is positive and statistically different

from zero indicating that total costs increase as number of sewerage connections increase. The

coefficient for the squared term (Y2Y2) is statistically significant and positive indicating that

marginal cost increases as the number of sewerage connections increases and that this effect has

explanatory power on costs. The coefficient for the interacted term with water lost (Y2Y3),

although small in magnitude is negative and statistically different from zero, indicating

economies of scope when producing these two outputs. The interpretation of this result is

complicated by the imposed symmetry. When adding the interacted term Y1Y2Y3 to the model,

only Y1Y3 is statically significant and still positive.

Economies of Scale

The assessment of economies of scale is fundamental for the characterization of the most

economical efficient structure organization. Evidence of economies of scale supports a smaller

number of firms to supply industry output, whereas diseconomies of scale indicates introduction

of competition as more efficient for the market structure. Following Christensen & Greene

(1976) and Hanoch (1975), economies of scale in this study are defined in terms of the

relationship between total cost and output along the expansion path, where input prices are









constant and costs are minimized at every level of output. This is considered the appropriate

scenario for this analysis given the cost minimization framework. See Hanoch (1975) for details

on the alternative approach implying the relative increase in output due to a proportional increase

in all input quantities.

The measure utilized to calculate economies of scale is the elasticity of cost (E) with

respect to output, which is defined as the proportional increase in cost resulting from a

proportional increase in the level of output. Because of the multi-output scenario, economies of

scale are referred to as ray economies of scale meaning that what leads the less than proportional

increase in cost is strictly the same proportional increase in the level of "all" outputs.

Conversely, "ray" diseconomies of scale are present when a higher than proportional increase in

cost occurs after an equal proportion on the level of "all" outputs is increased. Following Baumol

(1976) and Panzar and Willig (1977), a local measure of overall scale economies for a multi-

product firm is defined in Equation 3-7.

SlnTC
ES = 1/ fori =1..3 ; e ln (3-7)
1 ln

Equation 3-7 produces numbers that are higher than one for positive economies of scale;

less than one for diseconomies of scale and unity for constant returns to scale. The calculation is

performed from estimated values for each firm according to region and then firms are classified

by size. The standard deviation from calculating cost elasticities for each region is very high

suggesting a high degree of variation within each region. Table 3-7 shows economies of scale for

each region.

To provide some intuition on the interpretation of the scale economies findings, Table 3-8

shows distribution of firms in each region according to their size. There are no large firms

located in the mountains. Medium and small firms in this region present small diseconomies of









scale. The coast is the most populated region and firms located there are predominantly larger

than firms in other regions. Those large firms located in the coast are experiencing large

diseconomies of scale, yet the only three small companies in this region show large economies of

scale. Medium size firms located in the coast experience constant returns to scale. Economies of

scale are present for all firms in the forest.

Additional intuition is provided by Figure 3-2 which depicts fitted average total cost per

cubic meter of water billed for 2003. Firms are on the x-axis sorted by size from the smallest to

the largest. The shape of the average total cost curve shows average total cost per unit declining

as size of company increases. The slope of the curve is steeper for small firms meaning that they

are able to enjoy the benefits of larger economies of scale compared to larger firms. The curve

gets flatter as it reaches the largest firm. This suggests that scale economies have already been

exhausted by large firms. Overall, for the Peru water sector, the optimal firm size is medium if

the firm is located on the coast. If the firm is located in the mountains small may be considered

the optimum size (these firms exhibit close to constant returns to scale).

Input Factors and Price Elasticity

Finally, regarding the price of inputs, the coefficients are all positive as expected. The

price of direct labor is not statistically different from zero, yet the price of capital is highly

statistically significant when considering price differences for firms located in the mountains

compared with those in the forest.

The coefficients for the squared terms on both prices, labor (P1P1) and capital (P3P3) are

statistically significant at a 99% level, meaning that these prices do have an important role in

explaining costs. The sign of the squared labor coefficient is positive indicating that marginal

cost increases as labor is added. The sign for the squared capital coefficient is negative meaning

that marginal cost decreases as capital (debt) increases. These findings may signal, on average,









over-usage of labor (higher than optimal number of workers) and low levels of debt with the

subsequent indication of under-investment in the sector.

In a translog functional specification the coefficients for the interacted price terms identify

the marginal cost of increasing one of the inputs as the price of the other increases. If the

coefficient is positive, it indicates possible substitutability. Otherwise, it signals

complementarities among the input factors. The coefficient for the interacted prices term (P1P3)

is negative and statistically significant at a 99% level, indicating complementariness between

capital and labor. A negative yet small value for this partial elasticity is a consequence of having

more than two inputs in the model. Therefore, the degree of complementarities becomes less

restricted with respect to the two inputs case.

To examine this issue, Morishima partial elasticities of substitution, which do not impose

symmetry among factors, are calculated. The Morishima partial elasticity of substitution is a

measure of elasticity of substitution utilized in the multi input case. The elasticity value obtained

for the inputs direct labor and capital is equal to one and the value for the mirror combination is

equal to 0.8. These close to unity values indicate that these inputs are needed in relatively fixed

proportions within the production process.

Concluding Observations

The fragmentation of the water and sewerage sector was among the concerns of Peru's

water and sewerage regulatory agency at the end of 2006. Findings from this study reveal

important aspects about the performance of utilities in this sector. Results from the analysis

indicate that when considering the optimal usage of resources, in particular labor and capital,

small firms are on average more efficient than firms of large or medium size. This conclusion

comes after finding that small firms are closer to the efficient frontier than other companies.









A reduction in resource usage translates into a reduction of costs. If small firms were to

behave technically efficiently, they would need to decrease their mean cost by approximately 4%

over the analyzed period. This cost reduction is equal to 16% for large firms and 17% for

medium size firms.

The shape of the average total cost curve for this sector shows that there are scale

economies to be enjoyed by small and medium size firms, indicating possibilities for

aggregation. For firms located in the mountains, introduction of competition is indicated given

the presence of diseconomies of scale in this region. The optimal size for a firm located in the

mountains ranges on providing service to 1,000 to 10,000 connections. These companies are

classified in this sector as small. On the coast the optimal firm' size is serving between 10,000

and 40,000 connections.

Political influences affecting large water service providers have been a concern for

SUNASS for a long time. Finally, in 2007 rules were changed to include customer

representatives on the board. Nevertheless, the proposal of institutional aggregation by SUNASS

comes as a more aggressive approach to address this issue. Findings from this study indicate that

large firms on the coast have already exhausted scale economies, so aggregation in terms of

joining assets is not recommended. Introduction of competition in the mountains is also advised.

With respect to medium size companies, aggregation is indicated by the presence of large scale

economies for these firms if located in the forest. However, results from the analysis found the

production technology not homothetic implying that the input mix varies with the scale of the

firm. In addition, the production technology was found to be no homogeneous which means that

returns to scale varies with the scale of the firm. These findings imply that any merging or









aggregation among these firms would require a detailed assessment of the particular firm

production characteristics, including topological, hydrological, and geographical constraints.

In a previous study by the author, the number of localities served by a firm is found to be

statistically significant and negative. The present analysis does not contribute with additional

information about this issue. Chapter 4 presents a more detailed assessment on this matter.

Findings about the set of outputs indicate that the cost paid by firms for allowing water

billed and lost to be produced together is 0.1% per each 1% increase of joint production. This

cost represent the cost paid by firms by allowing non authorized connections to be present in the

network and secure the votes of the population being served with water that is not billed. When it

comes to the production of water lost alone, an increase of 1% of its volume produces a decrease

on costs of 1.3%. This cost reduction represents the cost savings from the tradeoff of pumping

more water in the network instead of repairing pipe leaks. In conclusion, firms have costs

savings by producing water lost which generate incentives for them to keep the joint production

of water lost and billed.

With respect to the number of sewerage connections, results indicate that marginal cost

increases as the number of sewerage connections increases. This result and the possibility of

finding diseconomies of scope for the joint production of water and sewerage may set the stage

to a more detailed analysis into delivering water and sewerage in a separate way.

Regarding price of inputs, findings might signal, on average, an over-usage of labor

(higher than optimal number of workers) and low levels of debt with the subsequent presence of

under investment in the sector. Overall, companies need these inputs in fixed proportions.











Table 3-1. Change in network length between 1996 and 2003
Mean StDv Median
Large 25% 16% 23%
Medium 60% 53% 51%
Small 56% 70% 34%
All 52% 57% 42%


Table 3-2. Summary statistics for outputs and price of inputs
Description Var Mean StDv Median Min Max
Volbill Millions m3 Y1 Large 67.6 133.3 26.2 9.2 396.8
Medium 4.7 2.2 4.2 2.0 10.3
Small 1.3 0.5 1.2 0.2 2.4
All 15.1 59.5 3.5 0.1 432.3
Sewcox- Thousands Y2 Large 198.5 321.6 99.5 36.4 989
Medium 15.9 5.9 14.7 6.6 30.1
Small 4.5 2.2 3.8 1.6 9.0
All 45.7 151.8 12.0 0.9 1,397
Volwaterloss- Millions m3 Y3 Large 49.2 90.8 17.1 3.7 272.9
Medium 5.0 2.6 4.9 1.5 9.1
Small 1.4 1.1 1.1 0.05 3.3
All 11.9 41.2 3.2 0 317.4
PriceDWork $/worker P1 Large 12.7 4.0 13.2 6.8 20.3
Medium 10.2 6.1 9.1 6.2 34.2
Small 7.7 2.5 7.1 3.1 12.1
All 9.7 5.4 8.8 1.7 40.4
PricelWork $/worker P2 Large 16.2 12.2 17.7 0.4 30.3
Medium 15.9 24.7 4.3 1.0 106.6
Small 8.6 7.8 5.0 1.6 24.3
All 13.2 45.6 4.5 0.1 768.0
Financecost $/km P3 Large 4.3 2.1 4.0 1.6 7.1
Medium 3.1 2.3 2.7 1.0 11.7
Small 2.5 2.2 1.8 0.3 9.0
All 3.1 2.7 2.4 0.1 14.5


I










Table 3-3. Summary statistics for utilities' specific characteristics


Variable
Number of Localities


Actconx/totconx


Network density


Continuity


Group
(Z1) Large
Medium
Small
(Z2) Large
Medium
Small
(Z4) Large
Medium
Small
(Z5) Large
Medium
Small


Mean
17
5
1
0.82
0.80
0.84
124
125
131
14
15
16


StDv
13.8
3.3
0.5
0.08
0.10
0.08
33
28
52
5
5
6


Median
17.5
3
1
0.83
0.83
0.86
111
118
118
14
14
18


Min
3
1
1
0.65
0.57
0.65
106
85
61
7
6
2


Max
41
14
3
0.91
0.92
0.94
204
196
226
22
22
23










Table 3-4. Estimation results for the translog cost function


Var

T

NIv

GDPv

DISTRICTS

DISTRICTS x R1

DISTRICTS x R3

MAINTENANCE

MAINTENANCE x R1

MAINTENANCE x R3

NDENSITY

NDENSITY x R1

NDENSITY x R3

CONTINUITY

CONTINUITY x R1

CONTINUITY x R3

Intercept

Gamma+


Coeff
-0.001
(0.0143)
0.036
(0.0356)
0.005
(0.004)
0.047
(0.0777)
-0.039
(0.0861)
0.046
(0.094)
0.363
(0.2265)
-0.482
(0.3036)
-0.882***
(0.280)
-0.456***
(0.088)
0.045
(0.1137)
-0.029
(0.1193)
0.172
(0.1136)
-0.304**
(0.1413)
-0.216*
(0.1281)
-6.380
(5.424)
0.45
(0.170)


Confidence levels: *** 99%; ** 95%;


0.1149*** 0.167*
Eta 0.0362) Mu (u)++ 0.090)
(0.0362) (0.090)
* 90% Data set: 336 observations; Standard Errors in


parenthesis. Loglikelihood=130.84732; Dependent variable: Ln (TotalCost); R1 = Mountains;
R3= Coast; Y1= Vol. Water Billed; Y2= Sewerage connections; Y3=Vol. Water Lost
Pl=Price of direct labor; P3=Price of capital
+ Gamma is defined as 7 =0, /o 2 where 0 2 is the sum of the variance for the term parameters
noise and inefficiency (a 2 = o 2 + 2 2)
++ Predicted Mu (u) after estimation comes from the estimate of minus the natural log of the
technical efficiency via E(uit I eit) where eit is the residual term.


Var Coeff
0.749
Y1
(1.186)
0.043
YlxR 0043
(0.1038)
-0.028
Y1xR3
(0.107)
2.113*
Y2
(1.212)
-0.032
Y2xR1
(0.1296)
-0.141
Y2xR3
(0.1496)
-1.347***
Y3
(0.4409)
0.072
Y3xR1
(0.0517)
0.152**
Y3xR3
(0.0642)
0.475
P1
(0.4096)
-0.086
PlxR1
(0.0559)
-0.033
PlxR3
(0.053)
0.068
P3
(0.400)
0.210***
P3xR1
(0.0484)
0.047
P3xR3 00
(0.0517)


Var Coeff
-0.022
Y1P1
(0.0484)
0.117**
Y1P3
(0.0455)
0.044
Y2P1
(0.0486)
-0.054
Y2P3
(0.044)
0.015
Y3P1
(0.0177)
-0.030*
Y3P3
(0.0184)
0.034
Y1Y1
(0.081)
0.189**
Y2Y2
(0.0966)
0.003
Y3Y3
0.0037)
-0.258
Y1Y2
(0.1694)
0.129***
Y1Y3
(0.0455)
-0.095**
Y2Y3
(0.0442)
0.087***
P1P1
0.0131)
-0.080***
P3P3
(0.0125)
-0.107***
P1P3
(0.020)














II Ii 11



1.0 ,- ,
I i I, 4. II I \ I i I I 5 I






-- ,TE96
TEO3
0.5 -


0.0
Utilities sorted by size


Figure 3-1. Technical frontier for 1996 and 2003


Table 3-5. Technical inefficiency statistics for 2003
Mean StDv Median Min Max
Large Mountains -
Forest 0.31 0.31 0.31 0.31
Coast 0.16 0.06 0.16 0.09 0.26
All 0.18 0.08 0.18 0.09 0.31
Medium Mountains 0.22 0.11 0.23 0.03 0.36
Forest 0.17 0.16 0.17 0.06 0.28
Coast 0.22 0.11 0.24 0.07 0.32
All 0.21 0.11 0.23 0.03 0.36
Small Mountains 0.2 0.12 0.21 0.06 0.34
Forest 0.14 0.13 0.12 0.04 0.37
Coast 0.13 0.09 0.15 0.03 0.2
All 0.16 0.11 0.15 0.03 0.37










Table 3-6. Cost reduction statistics for 2003
Mean StDv Median Min Max
Large Mountains -
Forest 0.76 0.76 0.76 0.76
Coast 0.07 0.26 0.09 0.36 0.39
All 0.16 0.35 0.13 0.36 0.76
Medium Mountains 0.18 0.20 0.12 0.04 0.68
Forest 0.18 0.10 0.18 0.10 0.25
Coast 0.15 0.09 0.13 0.01 0.28
All 0.17 0.15 0.13 0.01 0.69
Small Mountains 0.02 0.03 0.02 0.03 0.07
Forest 0.08 0.12 0.09 0.25 0.05
Coast 0.03 0.06 0.05 0.09 0.04
All 0.04 0.08 0.03 0.05 0.26



Table 3-7. Economies of scale
Large Medium Small All
Mountains 0.87 0.93 1.09

Forest 3.33 4.16 1.32 2.27

Coast 0.46 1 2.17 0.97





Table 3-8. Distribution of firms by regions and size
Large Medium Small All
Mountains 0 9 7 16
Forest 1 2 6 9
Coast 7 8 3 18
All Regions 8 19 16 43



















Cn













Water Volume

Figure 3-2. Fitted average total cost per cubic meter of water billed for 2003









CHAPTER 4
INFRASTRUCTURE SERVICE PROVISION TO MULTIPLE JURISDICTIONS: AN
EFFICIENCY ANALYSIS

When the objective is to fully expand infrastructure service across a country,

governments face the challenge of considering not only the technical characteristics of the sector

but also the country's economic, social and political structure. During the last decade,

infrastructure reforms undertaken by Latin American governments have often involved

decentralization, transferring service provision responsibilities to municipalities. The water

sectors of Peru, Nicaragua, and Honduras are a few examples. The objective was to give more

decision-making responsibilities to those with more information on costs and production

conditions to improve sector performance. However, outcomes do not always match the formal

aims of such restructuring initiatives.

This study examines the impact on efficiency of providing service to multiple

jurisdictions over the period 1996 to 2003. The issue analyzed is whether a utility expanding

service across multiple political subdivisions is more efficient than a utility expanding service

within a single political subdivision. In this context, a political subdivision is a unit of

government designed to carry out public functions within a specific territory. Utilities are

municipal-owned service providers. The hypothesis is that utilities answering to a heterogeneous

group of jurisdictional authorities are less efficient than those reporting to a single authority.

Political issues and bargaining for resources may contribute to production inefficiencies affecting

utilities' costs. The resulting inefficiencies may offset any scale economies associated with

serving a larger area.

This study is unique in that there is no empirical study in the infrastructure literature

examining the impact on utilities performance of this type of reform. Results from this analysis

provide insights to policy makers on an efficient governance structure for service provision









under a multiple political subdivisions framework, where the aim is to expand infrastructure

service within a country.

To illustrate the issue, the water sector of Peru is analyzed. The political process existing

in the country during the 1990s and the beginning of the current decade illustrates the political

bargaining process in a multiple jurisdiction organization. After some initial steps during the

early 1990s towards expanding service to under served areas of the country, the second half of

this decade represented a slowdown in this initiative, partially due to the country's economic

conditions and also because of the political climate present at that time (Corton 2003; Tamayo et

al. 1999).

During the authoritarian government of the 1990s, the country witnessed a breakdown of

political parties. To some extent, the resulting political vacuum was filled by a surge of

independent political leaders coming from the group of province municipal authorities. Provinces

are the largest organized unit of local government within Peru's departments. A department is

the main jurisdictional subdivision. These provincial municipal leaders tried to appeal to social

interests aimed at opposing the authoritarian government regime. They gained importance as

political representatives of social discontent locally and at the departmental level. These province

municipal leaders looked for regional blocks of power by attracting the interests of rural and

small towns and small district authorities, who were initially neglected by government as they

were seen as marginal lobbyists to its central favoritism.

However, by overestimating their political capabilities, this contingent of leaders ended up

pursuing individual interests rather than reinforcing each other's political clout in opposing the

regime. In a sense, the political pattern reflected individualism and dispersion rather than a

consensual force of opposition.









In this political system, governors lobbied the central government to obtain resources for

their departments. Presumably, these resources were directed to provinces where their

authorities had a common political ground with the respective governor. Evidence of

infrastructure development was present in those districts where the capital city of the province

was located. Consequently, the presence of political rivalry among province leaders affected

policy implementation, particularly with respect to allocation of resources in infrastructure.

As would be expected, conflicts of power permeated decisions coming from the board of

directors of water service companies due to divergent and individual political interests. For

instance, in the earlier stages of introducing private participation into the sector, the

organizational structure of one of the largest utilities (providing service to twenty eight districts)

was challenged by a lack of agreement among the directors of the board regarding whether to

allow this private participation process to go on in some of the districts served by the company.

Some local authorities might have viewed their political power threatened by this initiative.

Indeed, the water regulatory agency in Peru is currently assessing the issue of whether

having fewer rather than more political authorities for governance benefits water providers and

sector performance. The central issue from the regulatory perspective is that the decentralized

framework facilitates political interference. Requiring these utilities to report directly to a higher

authority level, such as the governor, would reduce the number of jurisdictions overseeing the

company, yielding efficiency improvements for the sector.

This study examines the effect on utilities' costs of providing service to multiple

jurisdictions. The analysis involves estimating a stochastic cost frontier where the number of

provinces served, service coverage and whether the water utility is the only provider within a

department are variables shifting the cost frontier. Results from estimating the econometric









model support the hypothesis that providing service within a province is more efficient than

extending service to nearby provinces. In addition, those utilities that have expanded service to

nearby districts show lower costs in the short run. However, expanding or contracting the service

to nearby districts is costly in the long run. This result may be interpreted by a lack of adjustment

of inputs to the changed production structure. The result is also consistent with the cost of

political interference in the utility's service provision decisions. Finally, results suggest that the

presence of more than one service provider within a department translates into lower costs for

these utilities when compared to the costs of a utility serving a department alone.

Political Sub-divisions

Peru is a unitary decentralized republic divided into twenty four main political

subdivisions called departments. The largest organized unit of local government within

departments is called a province. Within a province there is an additional level of unit of

government which is called a district. There are 195 provinces and 1639 districts in the country.

A department has on average 8 provinces and a province has on average 65 districts. Broadly

speaking, a department parallels a state in the United States and a province has similarities to a

county. Provinces and districts are municipally governed. As a decentralized republic, the

governor of each department coordinates with province authorities on the implementation of

development plans and allocation of resources.

A water company may offer services to more than one district, which is the minimum

service unit, and to more than one province. The average number of districts served by a utility is

five and the maximum number is twenty eight. Available data indicate that each district and

province is served by only one utility.

Departments may be served by more than one service provider. No utility provides service

to more than one department. Table 4-1 displays for each department: population and population









increase, the number of provinces and districts contained in the department, the number of

companies providing service and the number of provinces and districts served by such providers.

Note the variation in the number of provinces within departments and the number of districts

within each province. A department may have a minimum of one and a maximum of twenty

provinces. A province may have a minimum of five and a maximum of 161 districts. Note also

how evident is the under provision of service in some departments.

For a better understanding of this structure, see Figure 4-1 which represents these political

subdivisions in a very simplified way. The large squares depict two possible types of

departments within the country: type A, which has several water service providers and type B

which has only one provider. The smaller squares represent the provinces inside any department.

The rectangles inside each province are the districts being served by a utility. Groups of

rectangles are examples of utilities providing service to several districts. Utility 1 provides

service to several districts inside one province. Utility 2 provides service to several districts in

two provinces. Utility 3 is providing service to three provinces. In each province, this utility

delivers service to several districts. Note that although departments and provinces seem of the

same size they may not be. Districts served by each utility are displayed in different sizes.

This framework allows one to examine the presence of a competitive effect within

departments where there is more than one operator. Departmental authorities have discretion in

allocating resources to provinces. If there is more than one utility serving a department,

competitive pressures in requesting additional resources may affect the production process'

efficiency. Competition for funding might encourage the utility board of directors to demonstrate

that they are using resources wisely.









Ownership Structure of Utilities

Water service companies are owned by the districts to which they deliver service. The

number of shares owned by each district is determined by the proportion of district to province

inhabitants. This number is computed independently of actual population served by the utility.

More populated districts have a larger ownership stake on a utility that provides service to

multiple districts within a province.

A company's board of directors is comprised of a maximum of six members, each with a

one year term and the opportunity for reelection. Directors are elected by a stakeholder

committee, which is comprised of province and district representatives. Each committee

representative proposes a candidate to hold a board of director's position. There should be no

more than two directors representing one district.

Assuming that each district representative proposes a candidate from his own district,

province representatives are able to favor one district over others when proposing candidates to

the board. When there are homogeneous interests within the board of directors, this ownership

structure provides an incentive scheme which favors the expansion of service to highly populated

districts.

Providing service to more than one province may introduce a level of heterogeneity within

the board of directors that may impact a firm's efficiency. Possible conflicts of interest among

board members representing different provinces may get in the way of important economic or

strategic decisions. The case of a large utility facing a private participation decision mentioned

earlier illustrates how conflicts arise.

Contraction and Expansion of Service Provision

To further analyze the way these firms provide service, it is useful to examine how firms

contract and expand service provision across political subdivisions. Providing service to









neighbor districts could be viewed as a natural way for a company to expand its service,

assuming it has the resources to undertake such expansion and there is no other service provider

in the area. Service expansion by means of adding more connections to an existing network may

occur when local demand has not been fully satisfied or the local population is increasing.

Additionally, nearby districts may offer an attractive opportunity for a firm to decrease average

costs in the long run. From an engineering point of view, anticipating network capacity and

expansion is done before pipes are laid down. Adding pipes to extend the network may involve

costs that could have been avoided with appropriate planning.

Alternatively, several motives might account for a company's reduction of service. The

pipeline infrastructure in a location may be very old or in poor condition, so maintenance or

replacement could be too costly. It may be the case that the district in question comprises a

poor/rural area where the rate of uncollected bills is high or volume of nonrevenue water is high.

Therefore, by removing those districts that do not contribute to revenue, the company may be

able to increase profits or funds available for service improvement.

Another cause for removing districts from service may be stagnation or slow district

population growth, as people move to more populated areas where economic and infrastructure

development is more likely. The assumption when a district is removed from service is that

either a neighbor company takes over the service (appearing as an increase of districts served by

that company), or the district stops receiving service altogether. This may occur in very small

districts where service provision by independent providers is an alternative.

Data available for these for these utilities show six companies with a reduction of number

of districts served, and eleven show an increase. Approximately 50% of the increases happened









among companies serving more than six districts. Most decreases took place in companies

serving less than six districts and all of them are located in type A departments.

Change in Service Coverage

A change in the number of districts served does not necessarily parallel a change in

population served. The relationship between population served and the total population in the

area of a firm's service is known as service coverage. Available data on population served is a

total amount representing population served for all districts. Thus, absent more detailed data,

there is no way to identify population served by district.

Table 4-2 shows statistics for changes in service coverage and population growth by

region. On average, there has been an increase in coverage of approximately 20% over the

period. Coverage has increased twice as much for companies located in the mountains (27%), the

region with the lowest population growth, compared to the coverage increase of those on the

coast (13%), the region with the highest population growth. This outcome is consistent with the

government's aim of expanding water service to less populated areas in the initial

decentralization stage.

To relate the change of service across districts to change in total service coverage, Figure

4-2 shows a map of Peru with the departmental division displaying solid circles to indicate those

departments where companies have increased the number of districts served and dashed circles

to indicate those that have reduced their number. Note that the changes occurred in the number

of districts served for utilities in departments located on the coast or nearby, which are the most

populated departments. However, there are no changes in the number of districts for utilities in

departments where the increase in coverage is highest, such as those in the forest and the

mountains.









A closer look at these changes and their relationship with coverage and population

increases is given in Table 4-3. Firms that have added districts to their service are located in

departments with the highest increase in population growth and the lowest increase in coverage

(18%). Firms that have reduced the number of districts served are the ones located in

departments with the lowest rate of population increase and with the highest increase in service

coverage (28%). A plausible explanation is that firms increased local service coverage after a

reduction of service scope.

On average, utilities serving in type A departments show an increase in coverage of 14%

whereas those in type B departments have an increase of 9%. Tracking the increase in the

number of districts with an increase in coverage may be interpreted as a consequence of firm's

informed/planned decisions aimed at expanding service. Laying down more pipes implies higher

costs than merely adding more connections to an existing network. In this sense, the increase in

coverage is expected to be more costly if it includes an expansion of network compared with

only adding connections. On the other hand, increasing coverage over the same network may

decrease average costs if fixed costs are spread over a larger set of customers.

Model Specification

The competitive pressure effect among utilities within the same department is captured by

a dummy (MORETHANONE) which takes the value of one if the company is serving a

department type A, zero otherwise. It is expected that this competitive pressure has an effect on

utility's costs such that they are lower with respect to costs of utilities serving departments alone.

NPROVJ is a variable defined equal to the number of provinces served for firms located in

type A departments. It is expected that the larger the number of jurisdictions served by a utility

the higher its costs. This means that while serving in departments where there are several

utilities, utilities serving more rather than less provinces have higher costs.









CHCOVERAGE is defined as the percentage of change in firms' coverage from 1996 to

2003. The increase in coverage over the period could increase total costs if it has implied laying

down pipes to nearby districts which could have required additional capital and/or increasing

debt.

NEGCHIDX is a dummy variable which takes a value equal to one when there has been a

decrease in the number of districts served, and zero otherwise. A decrease in the number of

districts served should translate into a total cost decrease not only in the short run but also in the

long run as the company adjusts for a shorter network, fewer workers, lower local office

expenses and lower maintenance costs over the time period. This dummy is interacted with a

time trend to capture this long term effect (NEGCHIDXT).

Alternatively, POSCHIDX is a dummy defined with a value of one when a utility has

increased the number of districts served, and zero otherwise. This dummy is also interacted with

a time trend to capture the effect of having added more districts over the period. The coefficient

for this variable represents the cost difference for a utility that have extended its service toward

nearby districts with respect to utilities that have not done so. In the short run, when controlling

for coverage, network density and the number of provinces this coefficient is expected to be

negative as expanding service to additional districts translates into a larger network and

associated scale economies.

The dummies R1 for the mountains and R3 for the coast previously utilized for capturing

regional differences in the model presented in Chapter 3 are included in this model in the same

way: interacted with outputs and input prices. Network density, continuity of service and a time

trend variable are also included in the model. The variables that were not found to be statistically

significant in the previous model are not included. The homogeneity in outputs and









homotheticity hypothesis for the technology was rejected in the model presented in Chapter 3.

Thus, the same translog functional form is kept for the specification of this model. Including the

identified variables, the economic model is specified in Equation 4-1.


lnTC = a + n Y, + 1 ln n + I Y,, lnYm ln
m I m n
1 (4-1)
+ I y lnp lnP + Ym .,lnY lnP
2 J m J

Here TC is total costs as defined before; Y, is the vector of the m outputs already identified

(volume of water billed- Y1, the number of sewerage connections- Y2, and volume of water loss-

Y3); Pi is the vector of i input prices previously identified (PI, P2 are prices for direct and

indirect labor respectively, and P3 is the price of capital); and the a's and 7 's are parameters to be

estimated. The intercept a is specified in Equation 4-2.

a = a, + + a2MORETHANONE + a3NPROVJ + a4CHCOVERAGE (4-2)

A firm is input technically inefficient if it fails to produce maximal output from a given

quantity of inputs. Following Atkinson and Cornell (1994), the general form for a cost frontier

capturing input technical efficiency, which measures the potential each firm has to reduce costs

(holding output constant), is depicted in Equation 4-3.

C,(Y,,P /u,) = min[(P '/u,)(u,X,) I f(u,X,)= Y,]= (1/u,)C,(Y,,P,) (4-3)

Where the time subscript and other variables already identified in the model have been

omitted for simplicity, Yi is observed output of firm i ,fis a production function, common to all

firms, P ', = (pli, P2i ..., pji ) is a vector of input prices, andX, = (xli x2i ..., xji ) is a vector of

inputs. The parameter ui, 0 < ui < 1, measures the extent to which minimal input usage differs

from actual input usage and provides a measure of technical inefficiency. The last equality in

Equation 4-3 follows from the fact that a cost function is linearly homogeneous in input prices.









Another way to express Equation 4-3 is to think of the difference between minimum and actual

costs as the amount needed to reach the frontier, so each firm must lower cost by the amount

specified in Equation 4-4.

[(C,/u)-C,] (4-4)

Applying natural logs to Equation 4-3 yields Equation 4-5.

In C,(Y,,P, /u,) = In(/u,) + In C,(Y,,P,) = -Inu, + InC, (Y,,P,) (4-5)

If the variable C represents total costs as defined for this model, the second term of

Equation 4-5 is specified by Equations 4-1 and 4-2. In a panel data frontier analysis, inefficiency

is usually treated as the panel effect. In this framework, the parameter u encompasses those

unobserved factors out of a firm's control which represent the firm's specific level of technical

inefficiency. As such, this term is assumed not to be correlated with the explanatory variables.

Given the length of the period in consideration and the general characteristics of the

country and sector described so far, it is possible to infer that the firm's specific inefficiency

behavior has changed over time. Moreover, this change may be the same fraction for all firms in

this sector. To assess this possibility, time is included in the econometric model following

Battese and Coelli's (1993) functional form specification for the inefficiency component, (ui )

provided in Chapter 1.

For purposes of estimation, the usual idiosyncratic error term (v) is added to the

econometric model. This error term is independently and identically distributed following a

normal distribution with zero mean, and it is independent from explanatory variables.

Estimation proceeds using a balanced data set of 41 companies comprising 328 observations.

Lima's water utility, Sedapal, was removed from the data set given its high values for the

variables utilized in this model. It represents an outlier and is likely to distort the results.









Results from Estimation

For the frontier estimator, the likelihood function is expressed in terms of the variance

parameters, a 2 = V 2 + a, 2 and gamma, 7 = u 2 / 2. The closer the value of gamma is to one,

the more inefficiency as opposed to noise explains the variance of the disturbance term. The

variances for the normal noise and inefficiency terms are 0.02 and 0.026 respectively. The value

obtained for gamma indicates that unobserved inefficiency represents 60% of the residual term

whereas noise comprises 40%. Figure 4-3 displays the prediction of the inefficient parameter p

for 2003 showing full dispersion. Utilities are represented on the X axis sorted by their number

of connections which indicates their size and the inefficiency values are on the Y axis. Possible

values for p range from zero to one. The closer is the value to zero, the higher the efficiency.

Table 4-4 presents results from the estimation. Time is not statistically significant

indicating that technological changes intended to be captured by this variable have had no impact

in this sector costs over the period analyzed. However, the coefficient for q, the parameter

explaining the inefficiency behavior for these utilities over time, is positive and statistically

significant at 1% level, indicating that the exponential specification imposed is appropriate. It

indicates that utilities' efficiency has change over time following an exponential distribution.

Figure 4-4 illustrates the predicted technical efficiency for 1996 and 2003. The horizontal

line y=l represents the efficient frontier. Predicted technical efficiency after estimation comes

from calculating the technical efficiency term via E{exp(uit |eit)}. The closer the value to one the

more efficient is the firm. The model fit is illustrated in Figure 4-5. Actual costs for 2003 are in

natural logarithm form on the Y axis and are represented by a black line. Predicted costs are

represented by a gray dashed line and utilities are sorted by size on the X axis. The small gap

between actual and predicted costs indicates that the selected variables and functional form for

the economic model closely represent total costs for this sector.









Table 4-5 shows utilities' inefficiency statistics in 2003 by region. These values represent

(u) the extent to which minimal input usage differs from actual input usage. Predicted u after

estimation comes from the estimate of minus the natural log of the technical efficiency via E(uit

eit) where eit is the residual term. In other words, these values represent the factor by which input

usage needs to be reduced if firms are to behave efficiently. On average, water utilities in this

sector could reduce input usage by 27%. The differences in inefficiency among regions are not

large.

Turning to the interpretation of the coefficients for the explanatory variables, those utilities

providing service in departments type A have lower costs than those serving in departments' type

B. This is explained by the negative and statistically significant coefficient for the

MORETHANONE dummy which represents a 2.5% decrease in costs. This value explains the

cost difference between firms located in departments where there are additional service providers

with respect to utilities serving alone within a department. This cost reduction is paralleled in this

model by an increased efficiency for these utilities which is shown by smaller u values. Indeed,

utilities in departments' type B show a mean value for inefficiency equal to 31% whereas those

serving with other providers have a mean inefficiency value of 26%.

However, for utilities providing service along with other providers, a one percentage

increase in number of provinces served increases costs by 0.21%. This is reflected by the

positive and statistically significant coefficient for the number of provinces variable (NPROVJ).

Given that the maximum number of provinces served by a utility is eight, the impact on costs for

a utility increasing service to one additional province is 2.5%. This result offsets the cost

decrease obtained from the presence of competition.









The coefficient for the increased coverage variable is not statistically significant indicating

that this variable has no explanatory power on total costs. Coefficient magnitudes and signs for

network density, continuity and the set of outputs and input prices are very similar to those

obtained on the estimation of the cost model in Chapter 3. Overall, utilities with denser networks

have lower total costs. Continuity of service is not statistically significant indicating that this

variable has no explanatory power on total costs for this group of utilities.

A sensitivity analysis is performed including a few variables to test variation of results. In

Table 4-4, model (a) includes the number of districts served by each utility. The obtained

coefficients are not statistically significant and very small. In model (b), coverage increase and

the number of provinces served are interacted with the regional dummies. In this case, the

magnitude of coefficients for the variables MORETHANONE, NPROVJ and CHCOVERAGE

are slightly larger. The larger magnitude among coefficients of NPROVJ is the one related to the

mountains. This indicates that when expanding service to nearby provinces, utilities located in

departments type A that are in the mountains have higher costs than firms in other departments

and located in other regions.

The next two model variations address the possible effect on total costs of expansions or

contractions of service to nearby districts. In model (C), the coefficient for NEGCHIDX is not

statistically significant. However when considering the effect over time, utilities reducing the

number of districts have 0.2% higher costs than that of utilities not changing the number of

districts or increasing the number of districts. The variation of other coefficients magnitude is

very small.

In model (d) the coefficient for the dummy POSCHIDX is statistically significant and has

a negative impact on costs. This means that utilities expanding service to nearby districts have









total costs lower than those not doing so. When considering the effect of adding districts to

service over time, these utilities have 0.4% higher costs. This is shown by a positive and

statistically significant coefficient for the variable POSCHIDXT.

Finally, it is important to investigate what drives inefficiency in this sector directly,

utilizing another technique. The difficulty in doing this is that inefficiency is not observed. A

central assumption when defining the econometric model was that the unobserved inefficiency

term was not correlated with the set of explanatory variables. If it is, estimated coefficients are

biased. Inefficiency can be calculated by utilizing a DEA technique, as it was explained in

Chapter 1, where the same set of inputs and outputs considered in the cost model are included.

Once these efficiency values are obtained, they are regressed using an Ordinary Least Square

estimator. The idea is to investigate whether this inefficiency value can be explained by the

explanatory variables used as the shifters of the cost frontier. If these variables are drivers of

inefficiency, they may as well be correlated with the inefficiency term from the cost model,

which would produce biased coefficient estimates.

This robustness test proceeds by calculating a DEA frontier for each year, imposing

variable returns to scale in the calculation to allow efficiency to vary with the size of the utilities

and considering the inputs and outputs included in the cost model. The idea is to include as many

characteristics of the production process as possible. One aspect that is missed in this DEA

calculation is the translog functional form and the fact that input quantities instead of prices are

utilized. Thus, these calculated efficiency values represent a production frontier rather than a cost

frontier.

These new inefficiency values are regressed on the set of variables, shifting the stochastic

cost frontier using an Ordinary Least Squares estimator. In addition, the dummies for each region









are included in this model. Results from estimation are presented in Table 4-6. None of the

coefficients are statistically significant indicating that these variables do not explain technical

efficiency.

Concluding Observations

Results from this study help explain the performance of water utilities with respect to the

provision of service to multiple jurisdictions. It examines the impacts of policies promoting the

decentralization of water provision to municipalities compared to providing service to more

aggregated jurisdictions such as departments. The issue turns on the governance structure, which

affects the potential scope for political interference and conflict

Decentralizing water provision to expand service to less populated areas has increased

local coverage in less populated departments. The expansion of service implies lower costs if

utilities do so locally by increasing provision to districts within the same province. However, the

expansion of service to nearby districts has occurred within the most populated departments.

This pattern can be interpreted as a direct consequence of the ownership structure of these

utilities. Having data on population served by district for each utility could provide additional

insights on the issues analyzed in this study. Additional research is warranted.

The analysis also shows that changing the number of districts served implies some costs in

the long run. This finding may be a consequence of the lack of adjustment on the amount of

resources utilized in production as the size of the network changes. Alternatively, this result

might be interpreted as the cost of political interference on strategic decisions of these utilities.

The changes in the number of districts served occur only for utilities in the highest populated

departments and the increase in coverage for these departments is small relative to increases in

coverage in other departments. This observation suggests that changes in the number of districts









served may be a consequence of a quest for power, since expanding service to highly populated

districts increases the utility's ownership shares for those districts.

Results from the analysis also show that utilities located in departments where there are

other service providers have 2.5% lower total costs than those providing service in departments

where there is no other service supplier. Firms providing service alone show lower levels of

coverage, which could be due to the absence of competition. From a policy making perspective,

this result points towards developing an incentive scheme to introduce additional suppliers at the

department level. Additionally, utilities providing service in departments where there are other

suppliers have 2.7% higher costs for each additional province they serve. This result offsets the

lower costs due to competitive pressure present when there are additional suppliers in one

department. This finding supports the hypothesis that having more as opposed to less political

authorities to respond to introduces inefficiencies on the utilities production practices. As a

conclusion, within a department, an efficient way of expanding service is by introducing

additional service providers such that they serve only one province.










Table 4-1. Number of provinces, districts and utilities providing service by department


Region
Mountain













Forest








Coast


Department
Ancash
Apurimac
Ayacucho
Cajamarca
Cusco
Huancavelica
Huanuco
Junin
Puno
Amazonas
Loreto
Madre de
Dios
San Martin
Ucayali
Arequipa
Ica
La Libertad
Lambayeque
Lima +
Moquegua
Piura
Tacna
Tumbes


+ Lima city population 6,954,517


Population
census 2005
1,039,415
418,882
619,338
1,359,023
1,171,503
447,054
730,871
1,091,619
1,245,508
389,700
884,144

92,024
669,973
402,445
1,140,810
665,592
1,539,774
1,091,535
7,819,436
159,306
1,630,772
274,496
191.713


1993-
2005
Pop
Growth
6%
6%
21%
5%
10%
12%
8%
0%
13%
10%
20%

32%
17%
21%
21%
15%
20%
15%
20%
22%
16%
23%
21%


Num of
District
146
73
100
114
95
87
65
114
96
77
44


Serv.
Prov
7
1
2
5
6
2
2
7
7
3
3


Serv.
Dist.
7
1
2
6
7
4
3
10
8
6
3


Num of
Province
20
7
11
13
13
7
11
9
13
7
7


Num.of
Utilities
2
1
1
2
3
1
1
3
4
3
1

1
2
1
1
4
1
1
4
2
1
1
1










Utility 1
K


Department A
Province Province


ILcrn^


Province


Province


Department B


Province


Province


/II I__


Utility
Utility2


Province


LL~


Province



ZJYII


Figure 4-1. Schematic structure of jurisdictions in Peru and examples of service provision


Table 4-2. Service coverage increase from 1996 to 2003 and population growth by region
Service Coverage Pop. Growth
Region Mean Median St.Dev Min Max
Mountains 27% 31% 14% 0% 54% 9%
Forest 22% 21% 19% 2% 65% 17%
Coast 13% 10% 15% -7% 56% 18%
All 20% 17% 17% -7% 65% 15%


Utility3 -------














~~ -. -.4, /

-'Ad


O # Districts
Increased

# Districts
J Decreased


*I
N.


LlIN.I A


Figure 4-2. Departments of Peru showing those with variation in the number of districts

Table 4-3. Mean values for population and coverage increases
Number 0%Change %Change
Subset of Utilities Number Total %Change %Change
Utilities in Pop. in Coverag
Not changing number of districts served 24 14% 19%
Adding number of districts to service 11 16% 18%
Reducing number of districts 6 11% 28%
Total with a change in districts served 17 14% 22%
Total 41 14% 20%
Reducing number of districts + those not changing 6+24 30 14% 21%
Adding districts to service + those not changing 11+24 35 15% 19%


Inefficiency 0.6
Mean


0.4 -1


0.2


0-


0 A


* *


U
*


4 9*)
ScR


Utilities sorted by size


Figure 4-3. Values of predicted inefficiency for 2003


to


9 9


i










Table 4-4. Estimation results for variables comprising the intercept of the translog model
Model Main (a) (b) (c) (d)
T -0.001 0.05 -0.004 -0.009 -0.011
(0.012) (0.05) (0.009) (0.011) (0.008)
-0.25** -0.15* -0.28*** -0.28*** -0.22***
MORETHANONE
(0.102) (0.091) (0.099) (0.106) (0.084)
NPROVN S 0.21** 0.12 0.12 0.21** 0.196**
NPROVINCES
(0.086) (0.08) (0.09) (0.09) (0.084)
0.21
NPROVINCES x R1 -21
(0.134)
0.17
NPROVINCES x R3- -0.
(0.154)
-0.087 0.32 -0.82 -0.03 -0.229
COVERAGE
(0.274) (0.28) (0.54) (0.286) (0.251)
0.55
COVERAGE x R1 -55
(0.744)
0.99
COVERAGE xR3 0
(0.728)
0.06
DISTRICTS -0
(0.085)
-0.06
DISTRICTS x R1 0
(0.085)
0.102
DISTRICTS x R3 0
(0.101)
-0.12
NEGCHIDX -2
(0.139)
0.027**
NEGCHIDX xT 0
(0.012)
-0.28*
POSCHIDX -
(0.157)
0.045***
POSCHIDX xT 0
(0.015)
-0.32*** -0.42*** -0.279** -0.315*** -0.29**
NETDENSITY
(0.108) (0.112) (0.128) (0.111) (0.106)
-0.015 -0.051 -0.03 0.011 -0.085
NETDENSITY x R1
(0.136) (0.136) (0.155) (0.139) (0.130)
-0.189 -0.037 -0.23 -0.186 -0.201
NETDENSITY x R3
(0.129) (0.136) (0.159) (0.129) (0.129)
NTINUTY 0.059 0.161 0.047 0.002 0.129
CONTINUITY
(0.118) (0.113) (0.116) (0.125) (0.107)
-0.22 -0.32* -0.248 -0.15 -0.29*
CONTINUITY x R1
(0.145) (0.140) (0.148) (0.152) (0.132)
ONTINU-Y 0.07 -0.19 -0.05 0.017 -0.20
CONTINUITY(0.136) (0.128) (0.137) (0.148) (0.130)x R3
(0.136) (0.128) (0.137) (0.148) (0.130)










Table 4-4. Continued
Model Main (a) (b) (c) (d)
Number of Variables 40 43 44 42 42
LogLikelihood 114.82414 120.22713 117.5489 117.61988 119.51025
0.257* 0.261** 0.119
Mu (u) + 0.141 0.129
(0.102) (0.103) (0.142)
0.086** 0.088*** 0.074*** 0.095***
(0.102) (0.023) (0.024) (0.024)
Gamma +++ 0.567 0.63 0.698 0.62 0.668
Dependent variable: Ln (TotalCost); Number of observations 328
Confidence levels: *** 99%; ** 95%; 90%; Standard Errors in parenthesis.
+ Predicted Mu (u) after estimation comes from the estimate of minus the natural log of the
technical efficiency via E(uit eit) where eit is the residual term.
++Eta is defined as
+++ Gamma is defined as 7 = o. 21/ 2 where o 2 is the sum of the variance for the term
parameters noise and inefficiency (a 2 = o 2 + o, 2)


Technical
Efficiency
2.5


2


1.5


1


0.5


0


mI

S I
,
*//. \-\ : '


I' I
5I
I ,


. 1996


I, i I
'I' ,r
I'


Utilities sorted by size


Figure 4-4. Movement towards efficient frontier from 1996 to 2003












Ln (costs)
2

1

0

-1

-2


Utilities sorted by size

Figure 4-5. Actual total costs (black line) and prediction (dashed gray line) for 2003


Table 4-5. Inefficiency statistics by region for 2003
Mean SD p50 min
Mountains 28% 14% 28% 5%
Forest 22% 13% 19% 5%
Coast 28% 14% 26% 3%
All 27% 14% 27% 3%
(Inefficiency is estimated as minus the natural log of the TE via E


max
46%
42%
57%
57%
(uit I eit))










Table 4-6. Results from efficiency regression-dependent variable average efficiency
Model Coefficient
-0.017
(0.077)
0.034
(0.086)
-0.036
MORETHANONE036
(0.076)
0.015
NPROV015
(0.015)
0.105
COVERAGE
(0.293)
0.019
NEGCHIDX
(0.065)
0.020
PDXJ (0.071)
0.0003
NETDENSITY
(0.0007)
0.002
CONTINUITY
(0.005)
R-squared 0.123
Standard Errors are in parenthesis.









APPENDIX A
DESCRIPTION OF VARIABLES FOR DATA COLLECTION

General Characteristics for Each Service Provider in Central America

1. Service provider name and year beginning operations

2. Public (government) or private operation

3. Scope of the company according to country jurisdictional organization: state, municipal,
district

4. Type of service: water, sewerage or both.

5. Source of water: surface or underground

6. Localities: refers to the number of localities where the company provides service according
to the country jurisdictional organization (districts, municipal or status)

7. Region: refers to the region the company is located in case the country has specific regions
regarding some type of geographic or topological characteristics, eg mountains or coast. In
case the company serves different locations indicate region for each location.

Outputs

1. Volume of water (cubic meters or litters annual):
a) Produced: total annual gross volume extracted from its origin point, whether surface or
underground.

b) Delivered: total annual volume entering into the distribution network.

c) Billed: total annual volume that is billed to customers.

d) Collected: total annual volume for those bills that have been collected.

e) Treated: volume of water treated alter used by customers.

f) Total volume of water lost

g) Commercial lost: difference between volume delivered to the distribution network and
volume billed.

h) Operational lost: difference between volume produced and volume delivered to the
distribution network.

2. Number of water connections and sewerage:
a) Total
b) Residential, Commercial and Industrial
c) With meter









d) Active
e) Number of consumers water and sewerage:
f) Total population (usually from country census)
g) Population served with water and sewerage
h) Number of inhabitants per connection (or number of inhabitants per household)
3. Network length water and sewerage (km): total network length including the distribution
network.

Inputs

1. Number of workers and its costs (or expense):
a) Total
b) By contract(limited time or money amount)
c) Fixed (they receive company benefits)
d) Types of workers administrative, operational, managerial
2. Volume of energy and its cost (Kva or another unit).
3. Capital stock:
a) Non-current assets
b) Accumulated depreciation
c) Annual depreciation
4. Administrative Expenses
5. Financial Expenses
6. Operating costs
7. Total Cost

Quality

1. Water quality: any variable defining water quality (e.g. percentage of residual chlorine)
2. Continuity: number of hours a day customers receive water service
3. Quality of service:
a) Number of complaints received or served
b) Number of failures or problems received or fixed
4. Number of network leaks









APPENDIX B
SUMMARY OF PERFORMANCE ANALISYS METHODOLOGIES


Technology
Methodology Outputs Inputs functional Period Purpose
form


Performance
Indicators



TFP-
Laspeyres,
Paasche,
Fisher


TFP-
Tomqvist


DEA
frontier


Malmquist
catch-up
effect



Stochastic
Cost frontier


Water lost,
metering,
coverage,
network
density, water
consumption

1) VolWaterBil

2)#Connections


1) VolWaterBil

2)#Connections


1) VolWaterBil
+ number of
connections

2) VolWaterBil
+ net.density


1) VolWaterBil
+ number of
connect

2) VolWaterBil
+ net.density


VolWaterBil


Number of workers/
1000connect





Labor + Energy





Labor + Energy


1)GNI + Labor +
netleng

2) GNI + Labor




1)GNI + Labor +
netlength

2) GNI + Labor


GNI
Labor
Energy


Ratio of
single
variable



Ratio of
linear
combination


Ratio of log
form
combination


Ratio of
linear
combination






Ratio of
Distances


Productivity
2002 change of one
2005 factor for a
particular firm


2002
2005



2002
2005


Productivity
change of a
few factors for
a particular
firm
Productivity
change of a
few factors for
a particular
firm


Tech
efficiency of a
firm
considering
2005 multiple inputs


2002
2005


Linear log 2002
combination 2005


and outputs,
with respect to
best practice
within a group
Movement of a
firm towards
tech efficient
frontier set by
a group
Central
tendency of
cost efficiency
for a group of
firms









LIST OF REFERENCES


Aigner, D., & Chu, S. (1968). On estimating the industry production function. The American
Economic Review, 58, 826-839.

Aigner, D., Lovell, C., & Schmidt, P. (1977). Formulation and estimation of stochastic frontier
production function models. Journal ofEconometrics, 6, 21-37.

Antonioli B. & Filippini M. (2001). The use of variable cost function in the regulation of the
Italian water industry. Utilities Policy, 10, 181-187.

Ashton, John (1999). Economies of scale, economies of capital utilization and capital utilization
in the English and Welsh water industry. Bournemouth University, School of Accounting
and Finance, Working Papers Series No17.

Atkinson, S. & Cornwell, C. (1994). Estimation of output and input technical efficiency using a
flexible functional form and panel data. International Economic Review, 35, 245-255.

Atkinson, S., Cornwell, C. & Honerkamp, O. (1999). Measuring and decomposing productivity
change: stochastic distance function estimation vs DEA. Terry College of Business, Dept
of Economics, University of Georgia, Working Paper Series 99-478.

Atkinson, S. & Halabi, C. (2005). Economic efficiency and productivity growth in the post-
privatization Chilean hydroelectric industry. Journal ofProductivity Analysis, 2, 245-
273.

Aubert, C. & Reynaud, A. (2005). The impact of regulation on cost efficiency, an empirical
analysis of Wisconsin Water Utilities. Journal ofProductivity Analysis, 23, 383-409.

Baumol, W. (1976). Scale Economies, average cost, and the profitability of marginal cost
pricing. Essays in Urban Economics and Public Finance, R.E. Grieson, Lexington, Mass.
43-57.

Balk, B. (2003). The Residual: On monitoring and benchmarking firms, industries and
economies with respect to productivity. Journal ofProductivity Analysis, 20, 5-47.

Banker, R., Cooper, W., Seiford, L. Thrall, R. & Zhu, J. (2004). Returns to scale in different
DEA models. European Journal of Operational Research, 154, 354-362.

Battese, G. & Coelli, T. (1992). Frontier production functions, technical efficiency and panel
data: with application to paddy farmers in India. Journal ofProductivity Analysis, 3, 153-
169.

Battese, G. & Coelli, T. (1993). A stochastic frontier production function incorporating a model
for technical efficiency effects. Working Papers in Econometrics and Applied Statistics,
No 69, Department of Econometrics, University of New England Armidale.









Battese, G. & Coelli T. (1995). A model for technical inefficiency effects in a stochastic frontier
production functions for panel data. Empirical Economics, 20,325-332.

Clarke, G., Kosec, K., & Wallsten, S. (2004). Has private participation in water and sewerage
improved coverage? Empirical evidence from Latin America. World Bank Policy
Research Working Paper No. 3445.

Coelli, T, Perelman, S., & Romano, E. (1999). Accounting for environmental influences in
stochastic frontier models: with application to international airlines. Journal of
Productivity Analysis, 11,251-273.

Coelli, T., Estache, A., Perelman, S., & Trujillo, L. (2003). A primer on efficiency measurement
for utilities and transport regulators. World Bank Development Studies, 26062, the World
Bank, Washington D.C.

Cornwell, C., P. Schmidt and R. Sickles (1990). Production frontiers with cross-sectional and
time-series variation in efficiency levels. Journal ofEconometrics, 46, 185-200.

Corton, M., & Molinari, A. (2008). ADERASA's role in regulatory collaboration in the
Americas. Water 21, February 2008, 23-26.

Corton, M. (2003). Benchmarking in the Latin American water sector: the case of Peru. Utilities
Policy, 11,133-142.

Charnes, A., Cooper, W., & Rhodes, E. (1978). Measuring the efficiency of decision making
units. European Journal of Operations Research, 2, 429-444.

Christensen L. and W. Greene (1976). Economies of scale in U.S. electric power generation. The
Journal of Political Economy, 84(4), 655-676.

Diewert, W.E. (1974). Applications of duality theory, in Frontiers of quantitative economics,
edited by M.D. Intriligator and D.A. Kendrick. Amsterdam: North-Holland.

Diewert, W.E., & Wales, T.J. (1987). Flexible functional forms and global curvature conditions.
Econometrica, 55(1), 43-68.

Estache, A., Perelman, S., & Trujillo, L. (2005). Infrastructure performance and reform in
developing and transition economies: Evidence from a survey of productivity measures.
World Bank Policy Research Working Paper 3514.

Estache, A., Rossi, M. & Russier. C. (2004). The case for international coordination of electricity
regulation: Evidence from the measurement of efficiency in South America. Journal of
Regulatory Economics, 25(3), 271-295.

Estache, A., & Rossi, M. (2008). Regulatory agencies: Impact on firm performance and social
welfare. The World Bank, Policy Research Working Paper 4509.









Fabbri, P., & Fraquelli, G. (2000). Costs and structure of technology in the Italian water industry.
Empirica. 27, 65-82.

Fare, R., Grosskopf, S., Norris, M. & Zhang Z. (1994). Productivity growth, technical progress,
and efficiency change in industrialized countries. The American Economic Review, 84,
66-83.

Farrell, M. (1957). The measurement of productive efficiency. Journal of the Royal Statistical
Society, Series A, 120(3), 253-282.

Farsi M., Filippini, M., & Greene, W. (2005). Efficiency measurement in network industries:
application to the Swiss railway companies. Journal ofRegulatory Economics, 28(1), 69-
90.

Garcia, S. &Thomas, A. (2001). The structure of municipal water supply costs: application to a
panel of French local communities. Journal ofProductivity Analysis, 16, 5-29.

Greene, W. (2005a). Reconsidering heterogeneity in panel data estimators of the stochastic
frontier model. Journal ofEconometrics, 126, 269-303.

Greene, W. (2005b). Fixed and random effects in stochastic frontier models. Journal of
Productivity Analysis, 23, 7-32.

Greene, W. (2004). Distinguishing between heterogeneity and inefficiency: stochastic frontier
analysis of the World Health Organization's panel data on national health care systems.
Econometrics and Health Economics, 13, 959-980.

Hanoch, G. (1975). The elasticity of scale and the shape of average costs. American Economic
Review, 65(3), 492-497.

Hattori, T. (2002). Relative performance of US and Japanese electricity distribution: an
application of stochastic frontier analysis. Journal ofProductivity Analysis, 18, 269-284.

Hattori, T., Jamasb T., & Pollitt, M. (2005). Electricity distribution in the UK and Japan: A
comparative efficiency analysis 1985-1998. The Energy Journal, 26(2), 23-47.

Jamasb, T., & Pollitt, M. (2003). International benchmarking and regulation, an application to
European electricity distribution utilities. Energy Policy, 31(15), 1609-1623.

Jensen, U. (2005). Misspecification preferred: the sensitivity of inefficiency rankings. Journal of
Productivity Analysis, 23, 223-244.

Kalirajan, K., & Shand, R. (1999). Frontier production functions and technical efficiency
measures. Journal of Economic Surveys, 13(2), 149-172.









Kingdom, B., Liemberger, R. & Marin, P. (2006). The challenge of reducing non-revenue water
(NRW) in developing countries, how the private sector can help: a look at performance
based serving contracting. World Bank Water Supply and Sanitation Sector Board
Discussion Paper Series, 8, December 2006.

Lee, Y. & Schmidt, P. (1993). A Production frontier model with flexible temporal variation in
technical inefficiency. In The Measurement of Productive Efficiency: Techniques and
Applications, Harold F., Lovell, C., Schmidt, S. (Eds.), Oxford: Oxford University Press:
237-255.

Lovell, Knox C. (1995). Econometric efficiency analysis: a policy-oriented review. European
Journal of Operational Research, 80,452-461.

Lowry, M., Getachew, L., & Hovde, D. (2005). Econometric benchmarking of cost performance:
The case of U.S. power distributors. The Energy Journal, 26(3), 75-92.

Mathur, S. (2007). Indian IT and ICT Industry: A performance Analysis using Data Envelopment
Analysis and Malmquist Index. Global Economy Journal, 7, 1-40.

Meeusen, W., & Van den Broeck, J. (1977). Efficiency estimation from Cobb-Douglas
production functions with composed error. International Economic Review, 18, 435-444.

Murillo-Zamorano, L. (2004). Economic efficiency and frontier techniques. Journal of Economic
Surveys, 18(1), 33-45.

Neuberg, Leland (1977). Two issues in the municipal ownership of electric power distribution.
The Bell Journal ofEconomics, 8(1), 303-323.

Orea, L., Roibas, D., & Wall, A. (2004). Choosing the technical efficiency orientation to analyze
firm's technology: a model selection test approach. Journal ofProductivity Analysis, 22,
51-71.

Panzar, J.C., & Willig, R. (1977). Two Issues in the municipal ownership of electric power
distribution systems. The Bell Journal of Economics, 8, 303-323.

Pearce-Oroz, G. (2006). The viability of decentralized water and sanitation provision in
developing countries: the case of Honduras. Water Policy, 8, 31-50.

Pitt, M., & Lee, L. (1981). The measurement sources of technical inefficiency in the Indonesian
weaving industry. Journal of Development Economics, 9, 43-64.

Renzetti, S. (1999). Municipal Water Supply and Sewage Treatment Costs, Prices and
Distortions. Canadian Journal of Economics 32(3), 688-704.

Rossi, Martin (2001). Technical change and efficiency measures: the post privatization in the gas
distribution sector in Argentina. Energy Economics, 23,295-304.

Sabbioni, G. (2007). Efficiency in the Brazilian sanitation sector. Utilities Policy, 16, 11-20.









Saal, D., & Parker, D. (2000). The impact of Privatization and regulation on the Water and
Sewerage Industry in England and Wales: A translog Cost Function Model. Managerial
and Decision Economics, 21(6), 253-268.

Saal, D. & Parker, D. (2004). Productivity and price performance in the privatized water and
sewerage companies of England and Wales. Journal ofRegulatory Economics, 20(1), 61-
90.

Schmidt, P., & Stickles, R. (1984). Production frontiers and panel data. Journal of Business and
Economic Statistics, 4, 367-374.

Solo, T.M. (1998). Competition in water and sanitation- the role of small scale entrepreneurs.
World Bank Viewpoint, Dec.: Note 165.

SUNASS (2006). Infraestructura de agua potable y alcantarillado urban en el Peru: un reto
pendiente. Copyright SUNASS, Mayo 2008. http://www.sunass.gob.pe/svipu.jsp (last
visit: 05/25/2008).

Tamayo, G., Barrentes, R., Conterno, E., & Bustamante, E. (1999). Reform efforts and low-level
equilibrium in the Peruvian water sector In: Spilled Water: Institutional Commitment in
the Provision of water Services. Savedoff, W., P. Spiller (Eds.), Inter-American
Development Bank, Washington D.C., pp. 89-133.

UNCTAD United Nations Conference on Trade and Development. The digital divide report:
ICT diffusion index 2005. UNCTAD/ITE/IPC/2006/5 ; United Nations Publication;
Copyright United Nations, 2006. http://www.unctad.org/en/docs/iteipc20065_en.pdf (last
visit: 05/24/2008).

Zhang, Y., Parker, D., & Kirkpatrick, C. (2008). Electricity sector reform in developing
countries: an econometric assessment of the effects of privatization, competition and
regulation. Journal ofRegulatory Economics, 33, 159-178.









BIOGRAPHICAL SKETCH

Maria Luisa has a degree on computer science engineering. Her career developed as a

computer engineer for a long period of time while holding a variety of positions in different

public and private organizations. Maria Luisa also has experience in academia, teaching

computer science courses. In 1994, she was accepted at Jacksonville University to pursue a MBA

program in international business. She was holding a Fulbright/Mariscal de Ayacucho

scholarship from her country, Venezuela. In 1996, she finished the MBA with an award for high

achievement. In 1997, she applied to the University of Florida for a PhD degree in economics.

She was accepted and started the program, but she decided to end it as a master's degree by year

2000. In 2004, she resumed the PhD program and graduated in August 2008.





PAGE 1

1 ESSAYS ON LATIN AMERIC AN INFRASTRUCTURE: EMPIRICAL STUDIES OF SECTOR PERFORMANCE By MARIA LUISA CORTON A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008

PAGE 2

2 2008 Maria Luisa Corton

PAGE 3

3 To Aldo Alejandro

PAGE 4

4 ACKNOWLEDGMENTS This work would not have been possible without the support of Sanford Berg. His support, insight and direction were invaluable to the com pletion of my dissertation. I am also thankful for my involvement with the PURC t eam. They have supported me as a family throughout this process. In part icular, I would like to thank Cynt hia for reminding me that there was an end to the tunnel all along the way. Als o, I am very appreciative for Steven Slutskys thoughtful comments and for letting me in to the grad group every other year I am very grateful to Jonathan Hamilton, for his help an d insights, particularly in the fi nal stage of this project. It was an honor to work with my committee. Each member has introduced me to a different facet of economics. Finally, I am grateful to the Department of Economi cs and PURC for their flexibility and support during this long journey.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........7 LIST OF FIGURES.........................................................................................................................8 LIST OF FIGURES.........................................................................................................................8 ABSTRACT.....................................................................................................................................9 CHAPTER 1 INTRODUCTION ..................................................................................................................11 Performance Indicators and Total Factor Productivity........................................................... 13 Frontiers ..................................................................................................................................15 Non-parametric Frontiers................................................................................................17 Parametric Frontiers........................................................................................................20 Explaining Inefficiency........................................................................................................ ..23 2 BENCHMARKING CENTRAL AMERICAN WATER UTILITIES............................... 27 Data Collection Process..........................................................................................................28 Sector Performance Indicators................................................................................................31 Operational Indicators..................................................................................................... 33 Financial Indicators......................................................................................................... 35 Quality Indicators............................................................................................................ 36 Total Factor Productivity Analysis......................................................................................... 37 Data Envelopment Analysis (DEA)........................................................................................ 38 Stochastic Frontier..................................................................................................................42 Concluding Observations........................................................................................................ 44 3 SECTOR FRAGMENTATION AND AGGREGATION OF SERVICE PROVISION IN THE WATER SECTOR.................................................................................................... 56 Industry F ramework................................................................................................................58 Model Specification and Data................................................................................................ 64 Empirical Results.............................................................................................................. ......70 Regularity of the Cost Function ...................................................................................... 70 Efficiency........................................................................................................................72 Environmental and Output Variables.............................................................................. 73 Economies of Scale......................................................................................................... 76 Input Factors and Price Elasticity.................................................................................... 78 Concluding Observations ........................................................................................................ 79

PAGE 6

6 4 INFRASTRUCTURE SERVICE PROVIS ION TO MU LTIPLE JURISDICTIONS: AN EFFICIENCY ANALYSIS............................................................................................ 88 Political Sub-divisions............................................................................................................91 Ownership Structure of Utilities............................................................................................. 93 Contraction and Expansi on of Service Provision................................................................... 93 Change in Service Coverage................................................................................................... 95 Model Specification................................................................................................................96 Results from Estimation....................................................................................................... 100 Concluding Observations...................................................................................................... 104 APPENDIX A DESCRIPTION OF VARIABLES FOR DATA COLLECTION........................................ 113 B SUMMARY OF PERFORMANCE ANALISYS METHODOLOGIES............................. 115 LIST OF REFERENCES.............................................................................................................116 BIOGRAPHICAL SKETCH.......................................................................................................121

PAGE 7

7 LIST OF TABLES Table page 2-1 Share of water coverage within country by 2005.............................................................. 49 2-2 Operational performance indicators by 2005..................................................................... 52 2-3 Summary of finance indicator s-average values from 2002 to 2005 ..................................53 2-4 Percentage of change values for 2002 to 2005.................................................................. 53 2-5 Summary of quality indicatorsaverage values from 2002 to 2005 .................................. 54 2-6 Total factor productivity indexes from 2002 to 2005........................................................ 54 2-7 DEA technical efficiency and scale impact on efficiency for 2005................................... 54 2-8 Efficiency rank and operating cost reductions................................................................... 55 3-1 Change in network length between 1996 and 2003...........................................................82 3-2 Summary statistics for out puts and price of inputs............................................................ 82 3-3 Summary statistics for utilit ies specific characteristics.................................................... 83 3-4 Estimation results for the translog cost function................................................................ 84 3-5 Technical inefficien cy statistics for 2003..........................................................................85 3-6 Cost reduction statistics for 2003....................................................................................... 86 3-7 Economies of scale......................................................................................................... ...86 3-8 Distribution of fi rm s by regions and size.......................................................................... 86 4-1 Number of provinces, districts and utilities prov iding service by departme nt............... 106 4-2 Service coverage increase from 199 6 to 2003 and population growth by region............ 107 4-3 Mean values for populati on and coverage increases........................................................ 108 4-4 Estimation results for variables compri sing the intercept of the translog model............. 109 4-5 Inefficiency statis tics by region for 2003+...................................................................... 111 4-6 Results from efficiency regression (Dependent variable average efficiency)................. 112

PAGE 8

8 LIST OF FIGURES Figure page 2-1 Average volume of water delivered, billed and lost from 2002 to 2005 ........................... 49 2-2 Average number of connections from 2002 to 2005.........................................................49 2-3 Average population with water service and total population in the area from 2002 to 2005........................................................................................................................... .........50 2-4 Average changes in number of water c onnections and network length from 2002 to 2005....................................................................................................................................50 2-5 Average coverage and network density from 2002 to 2005.............................................. 51 2-6 Average volume of water billed a nd water consumption from 2002 to 2005................... 51 2-7 Average operating cost per connection and network density from 2002 to 2005............. 52 2-8 Average operating cost per cubic meter from 2002 to 2005.............................................. 53 2-9 Number of connections and workers by 2005................................................................... 55 3-1 Technical frontier for 1996 and 2003................................................................................ 85 3-2 Fitted average total cost per c ubic meter of water billed for 2003.................................... 87 4-1 Schematic structure of jurisdictions in Peru and examples of s ervice provision............. 107 4-2 Departments of Peru showing those with variation in the number of districts ................ 108 4-3 Values of predicted inefficiency for 2003....................................................................... 108 4-4 Movement towards efficient frontier from 1996 to 2003................................................ 110 4-5 Actual total costs (black line) a nd prediction (dashed gray line) for 2003...................... 111

PAGE 9

9 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ESSAYS ON LATIN AMERICAN INFRASTRUCTURE: EMPIRICAL STUDIES OF SECTOR PERFORMANCE By Maria Luisa Corton August 2008 Chair: Sanford Berg Major: Economics This study examines the performance of utilitie s in the water sector of Central American countries and Peru. Performance is assessed as te chnical efficiency and it is measured utilizing different methodologies. Chapter 1 explains the methodologies utilized in the different performance analyses presented in Chapters 2, 3 and 4. Chapter 2 provides a comprehensive efficiency analysis of water service providers in six countries in the Central American region. Pressures for sector reform have stimulated inte rest in identifying and understanding the factors that can contribute to network expansion, improved service quality, and cost containment. The aim of the analysis is to provide policymakers and investment funds institutions with quantitative evidence on the effectiveness of the regional water sectors and utilities under different institutional arrangements. In addition to key sector performance indicators, the analysis considers total factor productiv ity indexes, a non-deterministic frontier and a stochastic cost frontier. Differences in results from these met hodologies are due to the assumptions imposed to each model specification and data employed. The Central American performance indicators are compared to similar indicators for service provi ders in Latin America. From the group of countries analyzed in Central America, the wate r service provider from Panama is the best performer according to results from all the methodologies.

PAGE 10

10 Chapter 3 investigates the presence of economies of scale and technical efficiencies in the water and sanitation sector of Peru. Chapter 3 an alyses the sector struct ure considering size and location of service providers. The aim is to addr ess the merits of introducing additional suppliers in some regions and possible merging of small u tilities in other regions. The analysis employs a stochastic cost frontier model. Ov erall, economies of scale are pres ent in all firms in the forest region and in the small firms located on the coast. These findings support the aggregation process: consolidating some utilities could lower costs. Chapter 4 examines the impact on efficiency of water utilities providing service to multiple jurisdictions. The issue analyzed is whether the expansion of service across multiple jurisdictions is more efficient than expanding se rvice within a single jurisdiction. In this context, a jurisdiction is a unit of government designed to carry out public functions within a sp ecific territory. The hypothesis is that utilities answer ing to a heterogeneous group of jurisdictional authorities are less efficient than those reporting to a single authority. Political issu es and bargaining for resources may contribute to production inefficienci es affecting firms costs. The resulting inefficiencies may offset any scale economies asso ciated with serving a larger area. Results from the analysis support the stated hypothesis by fi nding less efficiency associated to utilities providing service to more than one province when they are serving in a department where there is more than one provider.

PAGE 11

11 CHAPTER 1 INTRODUCTION Developments over the past decade in quantitative techniques and pressures for infrastructure reform have stimulated interest in identifying and understa nding the factors that can contribute to improving infrastructure performance. Utility managers, infrastructure sectors associations, regulators, and othe r groups have begun to undertake st atistical analyses to measure utilities performance over time, across countries and geographic regions. Estache, Perelman and Trujillo (2005) present an analysis of reforms in the electricity, telecommunications, transportation and water sectors in the last decad e. The authors provide a comprehensive list of empirical studies in each sector highlight ing the factors each study considered. This studys contribution to the infrastructure literature is twofold. First it adds a regional analysis and a country case anal ysis to the water sector litera ture. Secondly, it provides an analysis of unique factors that a ffect infrastructure performance such as sector fragmentation and the provision of service to multiple jurisdic tions. These governance aspects are analyzed to determine the impact of infrastructure decentr alization on utilities performance a unique approach not yet covered in the empirical literature. Empirical studies in infrastruct ure addresses privatization, the type of regulation in place, or utilities ownership as factor s affecting performance. At a regi onal level, empirical studies are limited, mainly focusing on the el ectricity sector. For example, Zhang, Parker and Kirkpatrick (2008) investigate the effect of privatization, competition and re gulation on the performance of electricity utilities across 36 developing countri es from 1985 to 2003. Estache and Rossi (2008) analyze the effect on the electric ity sectors performance of having a regulatory agency in place for 51 developing countries from 1985 to 2005. Si milarly, Jamasb and Pollitt (2003) review the effect of regulation on the performance of a group of Europ ean electricity distributors.

PAGE 12

12 Regarding the water sector, the empirical literatu re related to more than one country is also limited. Sabbioni (2007) examines the relative effici ency of private and pu blic utilities in the water and sanitation sector in Brazil considering different regi ons within the country. Clarke, Kosec and Wallsten (2004) examined the impact of private participation on the performance of the water sectors of Brazil, Argentina and Bolivia measured as the increase on service coverage. In Chapter 2, several methodologies for assessing water utility perfor mance are applied to six countries in Central America: Costa Rica, El Salvador, Guatemala, Honduras, Nicaragua, and Panama. In Chapter 3, the performance of the water and sanitation sector in Peru is investigated to explore the impact of firm specific producti on characteristics on the performance of service providers. The aim is to address the merits of introducing additional suppliers in some regions and merging of small utilities in others. In Chap ter 4, the attention focuses on the role of service provision to multiple jurisdictions and its impact on efficiency. The goal is to investigate the effect on performance of the decentralization of service provision to municipalities. Decentralization is examined by Pearce-Oroz (2006). The author utilized a set of single dimensional indicators to examine the performance of decentralized versus centralized utilities in Honduras. This analysis is par tial since it is base d on performance measures that focus on specific areas of the production process of serv ice providers. This analysis does not consider factors specific to the environmental circumstances of the production proce ss or those particular to the operation practices of the u tilities. In addition, this type of analysis does not include the interrelationships among th e factors of production. In contrast, in Chapter 4 the impact of d ecentralization is considered by estimating a stochastic cost frontier wher e specific characteristics of the production process and their

PAGE 13

13 interrelationship are ta ken on account. The following sections provide a set of performance measurement methodologies which are utilized in this study. Performance Indicators and Total Factor Productivity A substantial body of technical literature exis ts regarding how to m easure performance. Coelli, Estache, Perelman and Trujillo (2003) present a survey of different methodologies on this topic. The methodologies considered in this study are performance indi cators, total factor productivity indexes, and frontiers. The simplest types of performance measures are single dimensional indicators such as labor productivity, service covera ge, and non-revenue water. These indicators focus on a specific area of performance within the production process. These types of indicators fail to account for the relationships among all the el ements of production and generally do not consider the impact of environmental or country specific factors. Productivity indexes represent a way of m easuring performance over time since they identify production differences between two ti me periods. A comprehensive exposition and analysis of such measures is found in Balk (2003). The basic idea behind a Total Factor Productivity (TFP) index is to determine how much output is produced due to each unit of input, which is expressed by Equation 1-1. XYTFP / (1-1) Here, Y is the level of output and X the level of input utilized in the production process. In practice, there may be more than one output prod uced from a combination of inputs. In such a case, a TFP index is constructed as the ratio of an output index to an input index. The input index should reflect the relative importance of each input in producing the output(s) and the output index should reflect the relative importance of each output. These relationships are represented

PAGE 14

14 by weights. Assume r is the wei ght given to k outputs and s the weight to n inputs. Equation 1-2 is the general formula to calculate a TFP i ndex measuring productivity change between two periods of time, say from year 0 to year 1. 0 0 1 1 0 1/ /nn k k k nn k k kXsYr XsYr TFP TFP (1-2) Three aspects are relevant when setting the a bove mentioned weights: the selection of the elements that represent the weights, the mathema tical or functional form that combine them and whether the weights are the same for the two anal yzed periods or not. Th e weights are generally prices for input factors and cost s for outputs. In Equation 1-2, th e weights are assumed the same for both periods but they may not be which yiel ds different alternatives for calculating TFP indexes. When considering the initial set of prices, Equation 1-2 repres ents a Laspeyres index while using the final period prices produces a Pa asche index. The Fisher index utilizes the geometric mean of the two periods. Regarding the functional form to relate the we ights, the indexes described so far imply a linear functional form. The Tornqvist index, repres ented in Equation 1-3, us es a logarithm form which is more flexible in reflecting a production technology. n kX X ss Y Y rr TFP TFP0 1 01 0 1 01 0 1ln)( 2 1 ln)( 2 1 ln (1-3) By understanding the sources of productivity change, manage rs can focus attention on areas that seem weak. At the same time, by unde rstanding the sources of these changes, policy makers, investors and other stakeholders can point to the most productive firms as examples of strong performance promoting the diffusion of best practice to all firms.

PAGE 15

15 In Chapter 2, a comprehensive analysis of performance indicators and Total Factor Productivity measures is pe rformed for the water sector in th e Central America region. Chapter 2 examines performance patterns across six count ries, focusing on three performance indicators: operational, financial and quality. In addi tion, trends for 2002 to 2005 are examined by computing total factor productivity indexes for wa ter utilities in each country. A central point in performing these calculations was the availability of data. The data collection procedure is given attention and is described in the Chapter 2. Frontiers A frontier approach has been widely used to analyze efficiency as a m easure of a firms performance. The measure of efficiency is li nked to the functional representation of production technology structure (Aigner and Chu 1968 and Lovell 1995). From production theory, an industry production function repres ents the relationship between i nput factors and output which underlines a specific production pr ocess technology such that ma ximum product is obtained from a combination of input factors within the industry environment. Conceptually, the industry production functi on is a frontier determined by the production process of those firms attaining maximum output w ith a set of inputs. Other firms in the industry fall short of the frontier due to the presence of production inefficiencies. Suppose a firm uses a set of inputs x = (x1, x2, .., xn)n available at prices w = (w1, w2,..., wn ) to produce output y in an industry environment characterized by exogeneous variables z = (z1, z2,.., zj)j. Exogeneous or environmental variab les are not controlled by the firm decisions. These environmental variables are characteristics of the firm, such as ownership, or features of the operating environment such as th e country economic characteristics, competitive conditions or geographical characteris tics where the firm is located.

PAGE 16

16 A production frontier is re presented by a function jnxf : with a parameter vector such that f (x, z; ) denotes the maximum output obtainable from input vector x in environment z. If the firm minimizes costs, th e dual of a production frontier is a cost frontier represented by a function j nxxc :with the same parameter such that c ( w, y, z; ) denotes the minimum expenditure on inputs required to produce output y with input price vector w in environment z In this context, the cost frontier represents the best that can be achieved in environment z so observed cost cannot be less th an minimum possible cost, in other words(,,;)Twxcywz The inequality leads to measures of inefficiency. In particular, cost inefficiency is measured by the ratio defined in Equation 1-4. (,,;) 1 ()Tcywz wx (1-4) Aigner, Lovell and Schmidt (1977), pointed out th at the possibility of estimating a frontier as opposed to an average production function to ex amine firms performance came after Farrells (1957) pioneering work on efficien cy. Farrell constructed an enve lope isoquant for the industry production function. Murillo-Zamorano (2004) presen ts a comprehensive set of techniques to calculate or estimate frontiers. Previous work summarizing the methodologies related to frontiers and efficiency analysis include that by Kalir ajan and Shand (1999) who review and compare methodologies for measuring technical efficiency. Following Murillo-Zamorano (2004), frontiers are classified as non-parametric, those where the frontier is not determined by a functio nal form, and parametric, those that have a specific functional form to specify the optimal fr ontier. Among the parametric frontiers, there are deterministic or stochastic frontiers, which cons iders the residual term in cluding the presence of noise and inefficiency, and non-deterministic those that view th e residual term comprised only

PAGE 17

17 by inefficiency. Naturally, frontie r techniques may yield different results due to the different underlying assumptions in determining the optimal frontier. Different frontier techniques are employed for examining the performance of water utilities in Central America in Chapter 2. In Chapter 3 a stoc hastic cost frontier is utilized to address the performance of water utilities in Peru. Chapter 4 utilizes the sa me frontier technique but now it includes different environmental va riables. In addition, in Chapter 4 efficiency is calculated by means of a non-parametric frontier. Results from the efficiency calculation are utilized to examine the presence of efficien cy drivers in the sector. Non-parametric Frontiers Data Envelopment Analysis (DEA) is the m ost commonly used non-parametric frontier methodology. Charnes, Cooper and Rhodes (1978) we re first to present the concept of the relative ranking of d ecision making units acco rding to their efficiency. A DEA calculation determines simple relationships among variable s. For example, utilities that produce far less output than other utilitie s, which are using the same input le vels, are deemed to be relatively inefficient. This methodology is viewed as an extreme point method because it compares production of each firm with the best producer s. A recent empirical study by Mathur (2007) utilizing DEA for the telecommunications sector of India provides a detailed and rigorous illustration of this methodology which is commonly applied in infrastructure to measure relative performance. Efficiency results from a DEA fr ontier are contingent to three main factors: The composition (homogeneity) of the sample set of firms to be analyz ed which is critical in determining the set of best producer s to be compared with each firm The set of selected inputs and outputs whic h establishes the comparison terms The quality of the data since this me thodology assumes that there are no errors

PAGE 18

18 A DEA analysis consists of measuring the effici ency of any firm as obtained by the ratio of weighted outputs to weighted inpu ts subject to the condition that similar ratios for every firm are less than or equal to unity. Th is relationship is expressed ma thematically in Equation 1-5. srminj X Y tosubject X Y Maxir m i iji s r rjr m i ii s r rr 1;1;1;0, 11 1 1 0 1 0 0 (1-5) Here, Y0, X0 are observed output and input variable vectors of the firm under evaluation; and are the weights to be a pplied to all units; i repr esents an input within a set of m, r an output within a set of s, and j one of the n firms. From Equation 1-5, possible va lues for the measure of efficiency, 0, ranges from zero to unit. This means that each firm can weight inputs and outputs differently as long as the ratio of their linear co mbination is less or equal to one. The efficiency of firm zero is rated relative to all firms. It is general practice to uti lize the dual expression of Equation 1-5 under a linear progra mming framework instead as it is shown in Equation 1-6. In this case, rho ( ) is a vector of intensity parameters which allows for the convex combination of the observed inputs and outputs. In other words, for an inefficient firm the efficient level of each production factor gets expressed as a linear combin ation of these intensity parameters as weights of the production factors of the peer group. NR XX YY to subject Min 0 0 (1-6)

PAGE 19

19 The output from a DEA exercise is the propor tion by which the obser ved inputs could be contracted if the firm were to operate efficiently. Intuitively this means that the same level of output can be produced with fewer inputs, so it is referred as an input efficiency approach. The implicit assumption is that managers minimize i nput usage given output level. In the economic literature, this is referred as the Farrells measure of i nput efficiency (Farrell 1957). Alternatively, the output efficiency appro ach considers the maximal proportional output expansion with the input vector held fixed, so managers maximize output given a set of inputs. Unless constant returns to scale are assumed, each of these approaches yield a different scalar value of efficiency. A comparison of results from input and output oriented models is found in Orea, Roibas and Wall (2004). A study by Banker, Cooper, Seiford, Thrall and Zhu (2004) addresses details and applications of input and output efficiency approaches under variable and constant returns to scale. For regulated industries, such as the water se ctor in Central America and Peru, an input approach is the natural option given that utilities most generally have se rvice obligations to all customers under a fixed tariff. This approach imp lies that firms are fully capable of reallocating resources when improving efficiency. In Chapter 2, the DEA methodology is utilized to assess the utilities performan ce in the Central America water sect ors in 2005. A set of two inputs and two outputs was considered: labor and networ k length, and volume of water and number of connections respectively. Robus t performance comparisons require analysts to obtain comparable data across firms, select appropr iate empirical methodologies, and check for consistency across different me thodologies. In Chapter 4, a DEA approach is utilized for robustness check of the stochastic frontier model utilized in the perfor mance analysis of the water sector in Peru. Technical e fficiency is calculated for each year of the panel data set and

PAGE 20

20 then regressed on the set of covariates utilized in the stochastic model to check for their explanatory power on efficiency. A Malmquist index measures the Total F actor Productivity change between two time periods utilizing the ratio of the distances of each data point relativ e to a common production technology. When calculating this in dex it is common practice to ut ilize DEA to calculate these distances. Following Fare, Grosskopf, Norris and Zhang (1994), and considering the input perspective already selected, a Malmquist technology change component based on the geometric mean of the considered periods is defined in Equation 1-7. 2/1 110000 111001 111 000 0 1),(),( ),(),( ),( ),( XYDXYD XYDXYD XYD XYD TFP TFP (1-7) The first ratio term to the right of Equati on 1-7 indicates a measure of input-oriented technical efficiency change for the analyzed peri od (the catching up effect or movement towards the frontier). Negative values indicate efficien cy has declined over th e period (the initial efficiency value is higher than the final value). Positive values indicate increased efficiency. A value equal to zero indicates no efficiency change. The term within brackets represents technical change calculated as the geometric mean of th e shift in technology between two periods. It is important to notice that while a TFP index is calcu lated only on reference to a particular firm a firm change of productivity over time the effici ency change component of the Malmquist index is calculated with respect to the movement of a firm towards the optimal frontier which is determined by a group of firms. This efficiency change component is calculated for the water utilities in Central America. Parametric Frontiers Finding key determinants of a fir ms perf ormance requires knowle dge of its production technology, or the particular way the firm selects inputs to pr oduce a set of outputs. This

PAGE 21

21 knowledge translates into a functional mathematical form for the production technology which recognizes the relationship among factors entering the production process. Aigner et al. (1977) were first in introducing a stoc hastic frontier approach in pa rallel with Meeusen and Van den Broeck (1977). According to the frontier classification pr esented by Murillo-Zamorano (2004), depending on the specification of the error term a frontier may be deterministic or non-deterministic. A deterministic frontier views a firms deviations from the frontier as pure inefficiency whereas a stochastic frontier recognizes the presence of noise and measurement errors. The problem with the deterministic frontier is that the estimated inefficiency may be confounded with noise. In addition, any misspecification or missi ng variables are also considered as part of the inefficiency term. This can lead to biased estimates of the position and shape of the fr ontier surface. Jensen (2005) compares both approaches for cross s ection models and draws conclusions depending upon the size of the data set, technology functional form and whether the objective is to estimate accurate inefficiency levels or efficiency ranking. A stochastic frontier framework considers a two component disturbance term: the typical noise ( v ), a symmetric component, and inefficiency ( u ) which represents those factors under the firms control such as technical and economic in efficiency, the will and effort of the producer and his employees, and defective and damage d products. For the case of a cost function specification, u is added to the frontier, showing the co st inefficiency coming from actual cost being far from the minimum possibl e cost frontier. A stochastic co st frontier is estimated for the water utilities in Central America and in Peru.

PAGE 22

22 Following Atkinson and Cornwell (1994), the ge neral form for a cost frontier capturing input technical efficiency, which measures the potential each fi rm has to reduce cost (holding output constant), is specified in Equation 1-8. (,/)min[(/)()|()](1/)(,)T iiii iiiiiiiiiiiCYWuWuuXfuXYuCYW (1-8) In Equation 1-8 the time subscript is been omitted for simplicity, Yi is observed output of firm i f is a production function, common to all firms, WT i = (w1i w2i ,wji ) is a vector of input prices, Xi = (x1i x2i xji ) is a vector of inputs, and ui 0 ui 1, is a parameter which measures the extent to which minimal input usage differs from actual input usage. The last equality in Equation 1-8 follows from the fact that a cost function is linearly homogeneous in input prices. To reach the frontier each firm must lower cost by th e amount specified in Equation 1-9. ])/[(iiiCuC (1-9) Applying natural logs to Equa tion 1-8, yields Equation 1-10. ln(,/)ln(1/)ln(,)lnln(,)iiiiiiiiiiiiCYWuuCYWuCYW (1-10) Regarding the inclusion of time in the mode l specification, Cornwell, Schmidt and Sickles (1990) utilizes a quadratic functi onal form of a time trend variable included in the intercept; Lee and Schmidt (1993) includes in the intercept a time trend variable interacted with environmental variables. Some authors such as Atkinson a nd Halabi (2005), Battese and Coelli (1995) and Atkinson, Cornwell and Honerkamp (1999) introdu ce the time variable twice: in the production function accounting for technical change and in th e explanation of the inefficiency effect. Time is included in the intercept of the functional form of technology to account for technology changes in the analysis of the water sector of Peru. Although the water industry is not characterized by substantial or ra pid technical changes as in th e case of the telecommunications

PAGE 23

23 industry, the eight year period under analysis is long enough to attempt to capture such an effect. In addition, to address the possibi lity of inefficiency changing ove r the eight year period, time is included in the econometric model following the functional form specified by Battese and Coelli (1992) for the inefficiency component, ( uit ), which is defined in Equation 1-11. 2exp[()];~(,)iti iuutTuiidN (1-11) In Equation 1-11, the ui are assumed to be independent and identically distributed nonnegative truncations of the 2(,)N distribution. Eta () is a scalar parameter to be estimated, t represents each time period within the T total number of periods. With th is specification, as t increases, uit decreases if is higher than zero, remains constant if equals zero and increases if is less than zero. The first case suggests that fi rms have improved their level of efficiency over the time period under considerati on. This specification assumes a ll firms have followed the same trend. This is restrictive but a reasonable assumption for the wate r sector in Peru since the group of utilities is delivering service under the same economic, social and political circumstances. Results from estimation support the hypothesis of inefficiency varying over time and following the specification in Equation 1-11. Explaining Inefficiency When the objective is to explain inefficiency in a stochastic mode l, the introduction of environmental variables requires elaboration. As it is been previously explained, in addition to output and input factors, environm ental variables repres ent specific characteri stics related to the environment where the utilities deliver service usually explaining production differences either across countries, regions or among firms. Firm specific characteristics are also referred as firm heterogeneity. In a panel data context unobserved firm heterogeneity is referred as the panel effect where a panel is comprised by observati ons for each time period for each firm. In a panel

PAGE 24

24 data frontier analysis, unobserved in efficiency is usually treated as the panel effect. The work of Aigner et al. (1977) on frontiers was extended to the panel data case by Pitt and Lee (1981) and Schmidt and Sickles (1984). The estimation of a frontier panel data proceeds by first finding the best (minimum) cost-performer-year for each firm, in other words the best performer within each panel. The frontier consists of all iden tified best (minimum) cost performers. Empirical researchers have attempted to dise ntangle firm heterogene ity from inefficiency given that unobserved inefficiency may be confounded by unobserved firm heterogeneity (Greene 2005a, 2005b and 2004; Farsi, Fili ppini and Greene 2005; Hattori 2002). Coelli, Perelman and Romano (1999) point ou t that, when environmental variables are included in the technology func tional form they affect the shape of the technology and the estimated residual includes a net measure of in efficiency. Otherwise, the estimated residual represents a gross measure of inefficiency. The issue is all about how to include environmental variables in a model provided that they may explain observed inefficiency. To illustrate this point, c onsider a cost function, C, f(.) a technology function, Wit a vector of input prices, Yit a vector of outputs, and Zit a vector of environmental variables. The previous classification of inefficiency translates into Equation 1-12 and Equation 1-13. The parameters are to be estimated and u and v represent inefficiency and noise respectively. (,,;,,)it ititit itCfWYZ and itititvu (1-12) (,,;,)it itit itCfWY and itititvu where ititituZ (1-13) According to Coelli et al. (1999), a net measure of inefficiency is obtained when the model is specified as in Equation 1-12 where th e vector of environmental variables (Z) is included in the production technology functional form. A gross measure of inefficiency is obtained when the model is specified as in Equation 1-13 where th e set of explanatory variables are directly

PAGE 25

25 explaining the inefficiency term u. If the environmental variables do explain inefficiency then they are correlated to the residual term in Equa tion 1-12. In this case the estimated coefficients are biased if using an Ordinary Least Squares estimator. A fixed effects estimator is more appropriate in this case given that it allows explanatory variables to be correlated to the residual term. In the case of specification defined in Equation 1-13 several aspects need to be considered. First, if the set of environmental variables do explain inefficiency using the righ t term of Equation 1-13 to estimate u then leaving these variables out of the left term specification of Equation 1-13 will produce a residual that includes missing observed inefficiency as well as unobserved inefficiency. Secondly, the use of a two step estim ation procedure, first predicting Cit under the assumption that the residual is identically distributed and then using the predicted inefficiency to regress the environmental va riables contradicts the identical and independent distribution assumption previously made. Battese and Coelli (1993, and 1995) have re viewed the evolution of the literature regarding how inefficiency is estimated and they pointed out that a simultaneous estimation procedure is more appr opriate given the issues previous ly discussed. Either using the model in Equation 1-12 or 1-13, a ro bustness test is appropriate. This test can be performed by estimating or calculating the inefficiency term by a different methodology and then checking the explanatory power of the set of environmental variables. Ultimately, whether environmental variables directly explain firms possible inefficiency or if they have an effect on the technology sh ape is a matter of interp retation related to the specific objective of the analysis. In this study, a stochastic co st frontier is estimated for the set of water utilities in Central America and the mu nicipal water utilities in Peru. Environmental

PAGE 26

26 variables are introduced in the intercept of the technology functional form assuming that unobserved inefficiency is not corre lated to these variables. A r obustness check is performed in Chapter 4 by calculating a DEA frontier to obt ain inefficiency values and investigate the explanatory power of the environmental variables.

PAGE 27

27 CHAPTER 2 BENCHMARKING CENTRAL AMERICAN WATER UTILITIES The purpose of Chapter 2 is to analyze the rela tive performance of water utilities in the Central American region to identify best perfor mers and areas of weakness in the sector. The results can help decision makers be tter direct investment funds in to projects that will enhance performance in the water sector in this region. Th is study is part of a Public Utility Research Center project funded by the Inter-American De velopment Bank as part of a new investment program which considers the possibility of provid ing loans to water utilities without requiring sovereign guarantees. A limitation associated with studies of Central Am erica is the scarcity of data related to the water sector. The first steps of the study involve d examining existing data and defining a set of variables to be collected. Appendix A presents the description of the identified variables. This set of variables was kept simple and limited to reduce possible road blocks during the collection process. During the data collection process, so me factors were found to be limiting and others were critical for the success of the process. These factors are de scribed in the data collection process section. The subsequent steps of the study related to performance measurement. The basic concepts and methodologies were described in Chap ter 1. With key input, output, and quality information, basic performance comparisons can be made. A set of performance indicators commonly used among practitioners in the water sector was calculat ed to provide a very simple picture of the sectors perfor mance characteristics in the re gion. Some of th ese performance indicators were compared to those presente d by the Association of Water and Sanitation Regulatory Entities from Latin America (Asociacin de Entes Reguladores de Agua Potable y Saneamiento de las Americas, ADERASA) benchmar king task force in its most recent annual

PAGE 28

28 report for Latin American countries (Informe Annual Benchmarking 2006; http://www.aderasa.org/es/documentos3.htm?x=753 last visit: 5/25/2008) For information on the countries and regulatory authorities com pri sing this association see Corton and Molinary (2008). The availability of data from 2002 to 2005 allowed assessing performance in the region through the calculation of total factor productivity indexes. In addition, to provide a more specific picture of the efficien cies associated to production practices in the region, a producti on frontier using data envelopment analysis and a stochastic cost frontier were included in the analysis. Differences on some of the values obtained from this variety of performance measures are based on the different assump tions underlying each methodology. Appendix B explains the purpose of utilizing each methodology. A best performer was found consistently through the calculation of the performance methodologies. The following sections describe each major step conducted on th e analysis of the water sector performance in this region. Data Collection Process The starting point of this study was to build a verifiable data base taking into account the data already available. To this end, the author requested and collected data from ADERASA and the International Benchmarking Network for Water and Sanitation Utilities (IBNET). The ADERASA data base is comprised of data that comes from the regulatory agencies in each Latin American country. This information is reported by some of the utilit ies in each sector but not all. Appropriate contact was establishe d with these utilities to verify the existing data and to obtain missing values. In addition, Guatemala and El Salv ador are not members of ADERASA, so data for these countries were collected for a first time. The adopted st rategy for the collection process was incremental in the sense that data were sent to the source several times for verification. A new and refined data set for the water sector in Central America emerged from this process.

PAGE 29

29 Nevertheless, only a subset of variables was used for the analysis because not all countries reported all variables in every year. Consequently, the number of observations was reduced to allow the data set to be comparable for al l utilities and to include all countries. The total time spent for data collection was six months. Data owners response was slow for some utilities given ongoing political and in stitutional changes at the time. In addition, several factors were iden tified as affecting data availability within the region: the ongoing water sector restructuring, the low level of water infrastructure in place, and the low presence of information technology among the service providers. From an institutional point of view, Costa Rica, Panama and Honduras have independent regulatory agencies but El Salv ador, Guatemala and Nicaragua still have central government bodies overseeing the water sector. Some of these central government institutions are undergoing restructuring such as th e transfer of sector responsibilities to municipalities. Some countries have just finished a major restructuring of this type. Because these changes generally imply changes within the company staff, the flow and registering of data may get interrupted, affecting data collection procedures. El Salvador, Honduras and Nicaragua show a lo w level of infrastructure in place. This promotes the presence of a large number of local independent water providers which complicates the data collection process. Solo (1998) provide s a detailed description on the role of these independent providers within the water sector of Latin American countries. Overall, the water sector in this region is fragmented given the decentralization of service provision into municipalities. For instance, for Honduras, with 271 m unicipalities, and for Guatemala with more than 300 municipalities th e difficulty in assembling a comprehensive data set is evident. Consequently, available data came from the m unicipalities serving the largest

PAGE 30

30 cities. This fragmentation greatly hampers perfor mance analysis for the water sector in this region. Finally, the development of information tec hnology is central to any data collection initiative. To frame the status of the region on this matter, the information and communication technology (ICT) diffusion index utilized by th e United Nations Conference on Trade and Development (UNCTAD) is util ized (UNCTAD 2006). More specifically, the ICT index includes a connectivity index to measure technology infrastructure development. This connectivity index includes per capita number of Internet hosts, number of PCs, number of telephone mainlines and the number of mobile s ubscribers. The connectivity index for 2005 was 0.75 for the United States. In the same year this index was 0.20 for Costa Rica; 0.10 for El Salvador and Panama; 0.08 for Guatemala and 0.04 for Honduras and Nicaragua. Information technology is the core to any structur ed data collection procedure. The availability of an information system specific for the sector is crucial for any data collection process. Initiatives in this respect are only inci pient. In 2004, a workshop hosted by Peru with participation of several Latin American countries representatives from the water sector gathered initial ideas and directed some efforts into the development of a water sector information system common to the region. A simila r initiative hosted by El Salvador developed in late 2006. The presence of technology is necessary but not sufficient for improved information on water utility performance. When designing ru les for the sector, the government needs to consider not only the utilities, the main entity responsible for colle cting appropriate data, but also the role that each stakeholder plays in the flow of data within the sector. For instance, the reporting of data to the regulator from the utilitie s needs to be stated by law and not taken as an

PAGE 31

31 informal relationship between the parties. In th e same way it is important to establish formal communication channels among all the sector stake holders, such as environmental or municipal development agencies, in a way that data collection programs and possible data repositories are well identified and efforts are not duplicated. Because data from all service providers are not available for this study, knowing the share of the population served with respect to tota l country population, known as service coverage, permits identification of the co mprehensiveness of this study. Table 2-1 summarizes the share of water coverage for each country. Costa Rica, Guatemala and Honduras are represented in the analysis by two service providers of different si zes, which are going to be referred as the small and large providers for each of these countries. The rest of countries are represented only by one operator. Sector Performance Indicators An initial step in analyzing performance is to calculate the most commonly used performance indicators in this sector. These ha ve been classified in this study as operational, financial and quality indicators. Since they are benchmarked against thos e indicators calculated by the ADERASA Benchmarking gr oup, their definitions are kept the same to maintain consistency across the region. In general, the extent of water service pr ovision can be measured by volume of water, number of connections and populatio n served. These factors provide a good picture of the size of the company providing service. Overall, in the Ce ntral American countries the water utilities can be broadly classified as sma ll, medium and large. Figure 21 shows average volume of water delivered, water billed and lost for 2002 to 2005. From a volume of water point of view, Panama has the largest provider. Note that for this group of countries and in general in Latin American

PAGE 32

32 countries, volume of water lost usually referred as non-revenue wa ter, represents a large portion of water produced. Figure 2-2 depicts average number of wate r and sewerage connections and water connections with meters. El Salvador has the larg er system from a number of connections point of view. On average, the small proportion of sewe rage connections relative to water connections may suggest cost restrictions to expand the sewera ge systems in this region. In addition, the level of metering varies among countries suggesting not only the possibility of cost restrictions but also the different social accep tance of this practice. The population dimension is shown in Figure 23 which depicts the variation of the mean coverage values among service providers for th ese countries. Although Panama and Costa Rica provide the largest volume of water delivered to customers, their number of connections and amount of population served is smaller than that of the water provider in El Salvador the largest provider from a water volume perspective. This illustrates some of the differences among countries water systems. The availability of several years of data allows us to analyze the changes occurring in the number of connections and network length which imply system expansion at different stages of the investment cycle. Clearly, national prioritie s and funding sources affe ct the pace and pattern of system expansion. Figure 2-4 show changes in the number of water connections and network length occurred from 2002 to 2005 to illustrate system expansion as shares of total system expansion for each country. The service provider from Guatemala shows a higher increase in network length with respect to that of number of connectio ns. This may suggest a system expansion of the transportation segment where pi pes but not connections are added as opposed to an expansion of the distribution segment wh ere pipes and connections are both added.

PAGE 33

33 Alternatively it may indicate earlier stages of the distribution network expansion where customers have not been connected yet. The service provider from Nicaragua presents an opposite situation. Here the increase in the number of connections is higher than the increase of length of network. This may be explained by connections added to satisfy commercial and industrial customers who generally do not add to population served. Operational Indicators Water lost or non-revenue water reflects deficiencies in either operational or commercial practices. The extent of water losses may reflect a cost tradeoff between increasing water production and repairing netw ork leaks to keep up with water demand. In other words, to satisfy demand, managers may find it more costly to re pair leaks and to control water losses than increase water production. Pipe leaks on the tran smission segment require costly maintenance outlays, particularly on long or dispersed networ ks. Operational water lo sses arise in transit while in the transport or distri bution network, and are calculated as volume of water produced less water delivered to the dist ribution network, expre ssed as a fraction of the volume of water produced. Referring to the distribution system, water lo sses may be either due to water theft, representing commercial losses, or to leakage from pipes. Given the characteristics of this sector it may be difficult for firms to control commerci al losses if doing so entails denying the service to the poorest segments of the population. For the distributio n network, water losses are measured as the difference between water deliver ed and water billed. Another way of viewing this indicator is to calculate the ratio of water billed to water delivered to the distribution network which is referred by the ADERASA benchmarki ng group as an indica tor for commercial efficiency. For utilities in the Central America sample, this indicator is equal to 55% which is higher than the ADERASA value of 40%.

PAGE 34

34 The metering indicator is calculated as the ratio of the number of connections with an installed meter to the number of total connections. Meter acquisition and in stallation costs are high. In some countries there is a direct allo cation of metering costs to the consumer, which generally translates into higher tariffs and/or connection fees. Th e higher is the level of metering, the higher the possibility of identifying water losses, the more accurate will collection information be and as a consequence, revenues ma y be higher. Overall, metering median value is 56%, which is lower than the 75% me dian value for ADERASA members. Service coverage is calculated as the ratio of popul ation with water service to total population in the area of service of the utility. The median value for water service coverage in this region is 90% which is close to the ADERASA value of 89%. There is a noticeable coverage gap between large and medium-small utilities. In Central America, coverage is equal to 92% for large firms, 66% for medium firms a nd 85% for the small utilities group. Water companies with a similar scale, measur ed by number of connected properties, may have different costs due to differences in networ k characteristics, such as length. Larger firms could have lower costs due to a large amount of cu stomers per kilometer of pipe, rather than due to scale economies originated from to tal output. To explore this issue, network density, measured as the ratio of the number of connections to network lengt h is the performance indicator considered in this analysis. The median value for network density equals 95 for utilities in Central America. Larger firms have denser ne tworks than medium and small firms. Figure 2-5 shows average values for coverage and network density from 2002 to 2005. The low coverage and high network density va lues found for the Guatemala utility suggest that the system may be expanded by increasing th e length of network to reach out under-served populated areas. The low coverage and low networ k density found in Honduras may indicate that

PAGE 35

35 the system can be expanded by adding more conn ections to satisfy underserved population in the area. The ADERASA benchmarking group utilizes th e ratio of volume of water billed to population with water servic e as an indicator for water consumption. The median consumption value for the region equals 219 liters per person per day, which is slightly higher than the ADERASA value of 172. The group of smaller companies is characterized by a higher consumption level of 323 liters per person per da y as opposed to a lower value of 222 satisfied by larger firms. Figure 2-6 depicts this indicator. The number of workers per one thousand connections is used in the water sector literature as signaling labor efficiencies or inefficiencies. A large value suggests the company is using a higher than efficient number of workers on its production process. The median value for this indicator equals 6.6, which is twice the valu e found for ADERASA members suggesting high labor inefficiencies. Note that the number of work ers considered for this indicator is a total figure which includes the number of workers under a cont ract agreement. These workers do not have any of the salary benefits provided by the comp any. Table 2-2 summarizes average values for the operational performance indicators from 2002 to 2005. Financial Indicators Financial indicators in this sector are expressed in curr ency amounts per connection or cubic m eters of water. Operating costs for this region include labor and energy costs, chemicals, administrative and sales expenses. Depreciation a nd finance expenses are considered to be part of total costs. On average, operating cost s are $91/connection. Figure 2-7 shows average operating costs per connection and its relationship to average netw ork density for the analyzed period. Higher values of network density are as sociated with lower values for operating costs per connection. Average operating cost per cubic meter of water deliv ered per utility is depicted

PAGE 36

36 in Figure 2-8. The median operating cost per cu bic meter is $0.10, half the cost of ADERASA member countries. For the large group of the utilities in the region, the median administrative expense per connection equals $27, whereas it is $34 for the small group. Both values are lower than the similar indicator for ADERASA members which equal $47. Table 2-3 summarizes average values for the financial indicators between 2002 and 2005. Table 2-4 presents a summary of percentage changes for some of these indicators from 2002 to 2005. Changes in operating costs for 2002 to 2005 are of small magnitude. Costa Rica displays a significant increase in cost of wo rkers (54%) and administrative expenses (51%) which may explain the increase in it s operating costs with respect to th at of other utilities. On the other hand, Panamas increase in operating costs (18%) may be explained by an increase in energy costs (54%). El Salvador displays a decr ease in administrative expenses (18%) which may explain the decrease of its operating costs ( 10%). Service providers which presented an increase in energy costs may be reflecting a combin ation of increases in i nput prices and greater utilization of energy inputs to service larger systems. Significant increases in cost of workers could be due to an increased focus on hiring professionals with managerial skills. Quality Indicators Comp liance with water quality standards has a median value of 95.96% for utilities in Central American countries. Continuity measured as the number of hours with water service ranges from 20 to 24 hours on average per year. Number of complaints per connection (median value per year) is similar for both ADERASA a nd Central American utili ties (0.21). The median annual number of leaks per km of pipe is 2.53 for ADERASA members, almost half the value found on Central America countries, 5.19. This suggests a lower degr ee of pipes service maintenance for Central America water networks co mpared with the Latin American set of water

PAGE 37

37 networks. This also explains the higher value of water lost in the Central American water utilities compared to this valu e for utilities in Latin America. Table 2-5 shows average annual values for the quality indicators di scussed during the period 2002 to 2005. Total Factor Productivity Analysis The basic idea behind a Total Fa ctor Productivity (TFP) index is to know how m uch output is produced for a given level of inputs. Four alternatives for calculating TFP indexes are considered for the Central American water sector, which have been already defined in Chapter 1: 1) the Laspeyres index which considers the initial set of prices as the base period; 2) the Paasche index which utilizes the final period prices; 3) the Fisher index which uses the geometric mean of the two periods; and 4) the Tornqvist index which uses a logarith m functional form to express the inputs and outputs relationships. Defining the weight of outputs for calculating these indexes requires detailed information on the production technology of these companies as well as more specific data. In order to simplify the analysis such that weights for the ou tputs are not needed, two sets of TFP measures are calculated: one considering volume of water billed as the output and another considering the number of connections. As in mo st empirical studies related to the water sector, labor and energy are the input factors under cons ideration. However, not all serv ice providers reported energy volume which limited the calculation of these indexes only to a subs et of utilities. The weights for the input factors are calculated as the ratio of thei r respective costs relative to operating costs. See Table 2-6 for obtained results. The Laspeyeres, Paasche and Fisher TFP i ndexes yield similar results. This may be explained by the fact that the length of period is short which produces only a small variation when calculating the weights for the different indexes. All comp anies are more productive from the view point of number of connections as opposed to volume of water billed. Panama is the

PAGE 38

38 only country displaying increased pr oductivity over the period when considering both number of connections and volume of water. The productiv ity increase for Panama ranges from 31% to 53%. Nicaragua displays a very small decrease in productivity ranging from 2% to 5%. Finally, El Salvador depicts a small productivity de crease ranging from 6% to 17%. Presumably, increases in TFP should track decr eases in average cost if all the other factors of production besides labor and energy remain constant. Th e service provider for Panama is expected to perform better with respect to other providers in the frontier performance assessment. Data Envelopment Analysis (DEA) This section examines each firm s relativ e technical efficiency in 2005 using Data Envelopment Analysis (DEA). As in the case of TFP indexes, DEA assumes that the data contains no measurement errors. A difference with respect to the TFP analysis is that the DEA methodology allows us to consider a linear comb ination of outputs and inputs for the production process without specifying their weights. Rather, these weights are calculate d with respect to the combination of these factors found on best producers. A DEA model collapses into the selection of an appropriate set of inputs and outputs involved in the production process. For the Centra l American water sectors, labor and capital are selected as input factors and volume of water billed and number of connections are the outputs. The amount of energy utilized in the production process was used in the calculation of the TFP indexes and is generally used as an input fact or in the production pro cess of water utilities. However data availability is limited for this set of companies. To include all firms in the calculation of the frontier, length of network instead of volume of energy is considered as input factor. Network length is utilized as a proxy to represent capital in the infrastructure empirical literature. The rationale for doing this is the hi gh amount of capital necess ary to lay down pipes

PAGE 39

39 compared to capital needs for pumps and treatmen t facilities. Labor is represented in the model by the amount of total workers. Even when the inefficiency or efficiency of a service provider may be due to its production process per se, a firm can be favored or hi ndered by country specific circumstances. Indeed, when considering several countries a major cha llenge is to appropriate ly account for each countrys political, social and economic differenc es. The empirical lite rature on cross country studies for the water sector is limited. Clarke et al. (2004) utilize GNP per capita when analyzing the impact of private participation in the water sectors of Brazil, Argentina and Bolivia. In the electricity sector, Estache, Rossi and Ruzzier (2004) utilize GNP per capita on a DEA and stochastic frontiers to account for countries di fferences when assessing efficiency for South Americas main electricity distribution companies. Zhang et al. (2007) utili zes GDP per capita in a stochastic model to assess the impact of privatization, regulation and competition on the performance of utilities in the electricity sector of 36 devel oping countries over a period of 18 years. A study by Estache and Rossi (2008) cons iders GDP per capita among a set of covariates to capture country particularities on a difference in differences analysis of the electricity sector across 51 developing countries. The World Bank utilizes the level of gross national income calculated by the Atlas method to classify economies and set the lending eligib ilities. The GNI adds to the GDP the income received from other countries, less similar payments made to other countries. The Atlas method uses a three year average of exchange rates to smooth effects of transitory exchange rate fluctuations. The assumption in this study is that variations of countrys GNI may impact the performance of water utilities since this variable captures country specific economic

PAGE 40

40 circumstances affecting production practices of these firms. Under a DEA framework, a way to examine this influence is to include GNI as an additional resource to the utility. Utilities in this region have different sizes so it is appropriate to account for firms scale when measuring efficiency. A variable returns to scale approa ch allows increasing or decreasing efficiency based on the size of the firms. Alterna tively, a constant return to scale approach means that firms are able to linearly scale the inputs and outputs wi thout increasing or decreasing efficiency. The ratio of the efficiency value pr oduced by a constant re turns approach to the efficiency value calculated from a variable retu rns to scale approach produces a scale impact value. Table 2-7 shows results for the DEA technical efficiency values under the variable returns to scale approach considering three options. Each option is based on utiliz ing a different set of input factors. The first column corresponds to e fficiency results when utilizing only labor. The second column shows the efficiency result when using labor and network length and in the third column efficiency results consider GNI. The last three columns of the table present the scale effect for each of thes e model alternatives. Overall, higher efficiency values are obt ained when including network length and subsequently adding GNI. The interpretation of results is based on the third column which includes the three inputs discussed. Including all the input factors imp lies considering more production characteristics which impr oves the quality of results. Efficiency results on the third column of Ta ble 2-7indicate that the service providers in Panama, El Salvador, Nicaragua, and the sma ll providers in Costa Rica, Guatemala and Honduras are all 100% efficient. E fficiency is equal to 63% for th e large provider of Costa Rica. This means that this utility could produce th e same output with approximately 63% of its resources (labor, network length a nd GNI). This result also means that this provider is only 37%

PAGE 41

41 inefficient. This value is obtai ned by subtracting 63% from 100% According to values obtained from scale impact calculations, 44% of the effi ciency value for the utility in Honduras comes from considering its size. Figure 2-9 shows labor as the i nput factor on the x axis and nu mber of connections as the output on the Y axis. This graph shows only one di mension of the frontier since the other input and output factors are not shown. Nevertheless, it gives an idea of each country relative position. For instance, El Salvador, and Nicaragua pr oduce a very similar amount of connections. However the amount of labor differs widely. While El Salvador is found 100% efficient, Nicaragua can work with 60% of its labor amount if consideri ng labor the only input factor. Costa Rica and Nicaragua levels of efficiency are very close (62% and 59% respectively). Obviously, these efficiency results exclude a number of factors affecting production conditions such as hydrology, total population, geography, topology, service quality levels, and other elements affecting the production process. Ne vertheless, these results provide a first cut at evaluating relative performance when considering number of workers and size of the network to produce water and water connections. Further understanding a DEA analysis requires considering the efficient input and output target levels for inefficient firms, which correspond to decreases in inputs and increases in outputs with respect to th e level of inputs and outputs of th e respective peer group. The rationale for this is that in some cases, the decrease in inputs is not enough to bring a company to the frontier, so an increase in out put is also necessary. According to DEA results, the large provider from Costa Rica and Honduras need to reduce la bor and increase actual volume of water. In this DEA analysis, only data from 2005 were utilized. The Malmquist technical efficiency change component for the period 2002 to 2005 is calculated for all firms except for the

PAGE 42

42 large service provider from Honduras. The utilitie s from El Salvador, Nicaragua, and the small utilities from Costa Rica and Honduras show no ch ange in technical efficiency over the period. The utilities from Panama and Guatemala show an increase in efficiency of 8% and 3% respectively. The large service provider from Costa Rica shows a decrease in efficiency of 2%. Stochastic Frontier This section examines cost efficiency by statistically estimating cost relationships according to input prices, given a level of out put produced. The ideal framework would be to completely specify a cost function including out puts, input prices and those specific factors capturing possible cost differences among firms and countries. The limitation associated with data availability restricts the analysis to the in clusion of only four explanatory variables in the economic model. The merit of performing this econometric exercise is to provide a general approximation for the ranking of firms, not the level. In a ddition, this methodology recognizes the presence of data errors which is an importa nt element for the Central American analysis. The DEA approach considers efficiency with respect to the best pe rformers, given the variables selected whereas an estimated cost frontier is a measure of central te ndency considering all firms not just those on the frontier. The economic model is a cost function specified by volume of water billed (VolBil), price of labor, price of energy, and GNI. Operational co sts plus administrative expenses divided by the number of connections comprise operating costs per connection which is the dependent variable. All variables are normalized by the number of connections, to control for firm size. The price of labor (LabPrice) is equal to total cost of work ers divided by total number of workers. The price of energy (EnegPrice) is total energy expenses divided by length of the network. The economic model translates into Equation 2-1 (time and firm specific subscripts are omitted for simplicity).

PAGE 43

43 Pr PrLEUnitOpCostGNIVolBilLabiceEnegice (2-1) In Equation 2-1, L, and E are parameters to be estimated. All variables are in natural logarithm form. The data set is an unbalanced panel for the period 2002 to 2005. Empirical researchers often introduce a time tren d in the model to capture possible technology shifts due to technological ch anges. Given the short period of time plus the limitation on the number of explanatory variables, GNI instead of a time trend is included to capture possible economic shocks occurred over the period. Neverthele ss, time is included in the estimation of the cost frontier to explain possible changes of efficiency over time The maintained assumption is that these countries have changed efficiency beha vior over time, independently. A stochastic cost panel data frontier specified by the model in Equation 2-1 plus the idiosyncratic error term ( v) which is independently and identical distributed as ),0(~2 vN and independently from regressors is estimated. The onl y panel-specific effect is the random inefficiency term ( u) The frontier estimator assumes the effects (error term) are not correlated to the explanatory variables. Inefficiency was allow to vary with time according to the Battese and Coelli (1992) specification for inefficiency. However, the coefficient for the parameter was not statistically significant which implies that inefficiency has not changed over time and it does not follow an exponential path. According to estimation resu lts, the operating costs for the region are represented by the Equation 2-2 (stand ard errors are in parenthesis). (5.3)0.1*0.6*0.7*Pr(0.1)*Pr UnitOpCostGNIVolBilLabiceEnegice (2-2) (1.81) (0.08) (0.12) (0.12) (0.07) The signs of coefficients are as expected a nd the statistical signifi cance is high for the output variable and price of la bor indicating a high economic impact on costs. Energy price and GNI are not statistically significant.

PAGE 44

44 An increase of 1% in volume of water per connection produces an increase in unit costs of approximately 0.6%. An increase in price per work er per connection of 1% produces an increase in unit costs of approximately 0.7%. Price of labor provides the highest impact on costs from relative to the effect of other variables. Increasing volum e has a less than proportional increase effect on costs. On average it indicates the pres ence of economies of scale in the region. Table 2-8 shows country ranks according to how far each firm is from the cost frontier and possible reduction of operating costs by 2005. This reduction is calculated as the ratio of the estimated inefficiency to actual operating costs. The differences between results from the DEA frontier and the estimated cost frontier can be explained by the fact that when calculating th e DEA frontier we were looking at a contraction of inputs for a given level of out put. A cost frontier looks at minimizing costs given input prices and output. When assuming the minimum set of inputs for a given level of output, the DEA abstracts from other factors influencing the pr oduction process, such as the price of inputs included in the cost frontier. However, Panama is been consistently the be st performer for all the methodologies. Performance differences for the ot her countries should be considered on a case by case basis. Concluding Observations A ma jor contribution from this study is the creation of a unique data base and a comprehensive data collection process. Consider ing total water service coverage in the region the data collected is very representative from a sector wide perspective. The quality of the data set is good in the sense that it came from and it was reviewed and certified by data owners. A major conclusion from this analysis points towards additional efforts for improving data collection procedures in the region. Besides the scarce presence of information technology limiting record keeping within thes e utilities, difficulties also may be due to the fragmented

PAGE 45

45 service provision in some countri es. A higher level of coordination is needed if data are to be collected and trends analyzed. Such an initia tive may require an analysis of stakeholders responsibilities regardin g monitoring and storage of data Coordination is needed among stakeholders regarding what and how to collect data. Information technology is central to any structured data collection procedure: the availability of an information system specific for the sector is crucial for any data collection process with in this region. Government, policy makers and fund providers need to consider the role of technology improvement in the region. In the process of identifying segments of the industry with no data, policy-makers, regulators and managerial staff have been encouraged to expand efforts to seek disaggregated data. Such data are necessary for further qua ntitative analysis, providing more complete information regarding sector performance. This study is comprehensive, ranging from the examination of the water sector structure charact eristics of each country to individual firm performance behavior under ve ry specific scenarios. A set of methodologies was utilized to assess sector perfor mance, such as performance indicators, total factor productivity analysis, production and co st frontiers. The role of benchmarking is fundamental for assess the better allocation of resources. Making comparisons based on individual indices is fraught with difficulty. Nevertheless, benchmarking techniques that draw upon performance indicators is an im portant step towards improved understanding of the efficiency behavior of water utilities. Focusing now, on the results for countries in Ce ntral America, the rela tive small number of sewerage connections compared w ith water connections may reflect costs restrictions in the area. The study did identified different stages of th e investment cycles fo r the group of countries. Overall, system expansion seems to be in balan ce in terms of adding more connections or more

PAGE 46

46 pipes to the network. Guatemala and Nicaragua are characterized by different behavior. Guatemala seems to be adding more pipes and Nicaragua more connections. While the former is trying to reach under-served areas, the latter is trying to sa tisfy local customers demand. Investment for these two types of systems expa nsion is different. Adding more pipes require higher levels of capital than adding only connections. The amount of non revenue water is higher in Central American countries (55%) with respect to that on Latin American countries (40%). This may be a consequence of poor metering systems as it is reflected in a 56% metering valu e for these countries with respect to a 75% for the sample of ADERASA Latin American countries It may also be explained by the amount of pipe leaks which is very high when compared to the same indicator for Latin American countries. Large firms have denser networks than sm aller firms, reflecting the low investment capacity of small providers. Guatemala shows a low level of coverage compared to its high network density which is consistent with the extension of its network system through adding more pipes as it was previously mentioned. In Honduras, the low coverage ratio and network density implies that the expansi on of the system should be thr ough adding more connections to satisfy under-served population in the area. Wate r consumption is higher in Central American countries when compared to Latin American c ountries which reflect abundant sources of water in the region. Labor ineffici encies are twice those found in Latin American Countries. On the financial side operating costs per cubi c meter of water are ha lf the value found in Latin American countries. Admini strative expenses are found to be higher in small firms. Changes in operating costs over th e period are reflecting changes in energy costs, as in the case of Panama, or labor costs as the case of Costa Rica.

PAGE 47

47 When trying to assess changes over the period, calculating diffe rent total factor productivity indices yield similar results given the short period analyzed. Panama is the only country displaying an increase in productivity when consider ing labor and energy as input factors over the period 2002 to 2005. The service pr ovider for this country coincides in showing an increase in efficiency as a result of calculating th e Malmquist efficiency change component. Nicaragua is the country within the group with a small decrease on efficiency change (2%). When calculating a technical frontier for 2005 by means of the DEA methodology, higher efficiency values are obtained from models that include GNI as an additional input factor. The sensitivity of results to the inclusion of the GNI variable illustrates th e importance of country characteristics in explaining efficiency within th e region. It was also important to consider the scale of the service providers as it was re flected by the large s cale impact values. Results indicate that the service providers in Panama, El Salvador, Nicaragua, and the small providers in Costa Rica, Guatemala and Ho nduras are all 100% efficient. Part of this efficiency for El Salvador, Nicaragua and the small providers is due to thei r scale. The fact that such a large number of firms are found to be 100 % efficient indicates that the group of firms is not heterogeneous enough in the se nse that the set of input and out put factors considered are not sufficient to explain possible production differences among these countries. Results from the stochastic cost frontier esti mation indicate that when considering output and input prices, the highest impact on operati ng costs per connection co mes from labor prices. Estimated firms ranks according to how far is each firm from the cost frontier shows Nicaragua as the most efficient firm. The service provi der from Panama is ranking number 2 which coincides with the fact that its service provider is positioned as the best performer from other methodologies. The ranking of other providers need to be assessed on a case by case basis since

PAGE 48

48 some results are contradictory. Data limitations affect model specification: not all elements affecting cost are in the model. For regulatory agencies, related government institutions, and funding agencies, this study may contain additional information for their stra tegic planning and decision making processes. This study should be viewed as a first step in the analysis of water utilities in Central America. As data from additional years become available and more utilities provide information, analysts will be able to conduct much more thorough analys es of sector performance. Hopefully, the results presented here will serve as a catalyst for more comprehensive data collection/verification initiatives in the region and for additional quantit ative studies.

PAGE 49

49 Table 2-1. Share of water coverage within country by 2005 Country Share of Coverage Costa Rica 51% El Salvador 94% Guatemala 10% Honduras 20% Nicaragua 52% Panama 66% 0 50 100 150 200 250 300 350 400 450Millm 3 PanamaCostaRicaElSalvd.NicaraguaGuatemalaHond.CostaRicaHond.Guatemala Delivered Billed Lost Figure 2-1. Average volume of water delive red, billed and lost from 2002 to 2005 0 100 200 300 400 500 600PanamaCostaRicaElSalvd.Nicaragua GuatemalaHond.CostaRicaHond.GuatemalaThousand s Water Sew Wat w/Meters Figure 2-2. Average number of connections from 2002 to 2005

PAGE 50

50 0 500 1,000 1,500 2,000 2,500 3,000 3,500Thous inhab i PanamaCostaRicaElSalvd.NicaraguaGuatemalaHond.CostaRicaHond.Guatemala Total With Water Serv Figure 2-3. Average population with water serv ice and total population in the area from 2002 to 2005 0% 25% 50% 75% 100%PanamaCostaRicaElSalvadorNicaraguaGuatemalaHonduras WaterConx NetLeng Figure 2-4. Average changes in number of wate r connections and netw ork length from 2002 to 2005

PAGE 51

51 0% 20% 40% 60% 80% 100%PanamaCostaRicaElSalvd.NicaraguaGuatemalaHond.CostaRicaHond.Guatemala0 20 40 60 80 100 120 140 Coverage Network Density Figure 2-5. Average coverage and network density from 2002 to 2005 0 50 100 150 200 250PanamaCostaRicaElSalvd.Nicaragua GuatemalaHond.CostaRicaHond.Guatemala0 50 100 150 200m3/hab OpCost/Conx Consumption Figure 2-6. Average volume of water billed and water consumption from 2002 to 2005 Network Density Coverage $/ Connection

PAGE 52

52 Table 2-2. Operational performance indicators by 2005 Vol Del Vol Lost Num Conn Met Vol Del Pop Serv Cov Net Length Net Dens work/ 1000 Country/Units Mill m3 % Miles % Per Person Miles % Km Conn/ km Conn Panama 452 58 448 41 126 2,004 92 4,727 95 5.6 Costa Rica-L 305 49 457 94 76 1,978 99 6,437 71 6.7 El Salvador 259 619 55 84 3,093 90 4,391 141 4.2 Nicaragua 257 44 457 48 39 2,870 91 4,604 99 6.7 Guatemala-L 122 55 195 84 81 1,045 93 5,013 128 7.5 Honduras-L 75 63 105 35 67 707 69 2,800 38 10.5 Costa Rica-S 28 55 48 97 66 228 100 678 71 2.5 Honduras-S 10 44 11 77 157 53 71 77 144 6.6 Guatemala-S 7.6 10 56 183 42 72 232 42 9.6 0 50 100 150 200PanamaCostaRicaElSalvd.NicaraguaGuatemalaHond.CostaRicaHond.Guatemala0 50 100 150 OpCost/Conx NetwDen Figure 2-7. Average operating cost per conne ction and network density from 2002 to 2005 $/ Connection Connection /km

PAGE 53

53 0.00 0.05 0.10 0.15 0.20 0.25 0.30 PanamaCostaRicaElSalvd.Nicaragua GuatemalaHond.CostaRicaHond.Guatemala Figure 2-8. Average operating cost per cubic meter from 2002 to 2005 Table 2-3. Summary of finance indica tors-average values from 2002 to 2005 Country OpCost LaborCost EnergyCost AdmExp FinExp OpCost/Connection Units $/m3 $/m3 $/m3 $/m3 $/m3 $ Panama 0.10 0.04 0.04 0.03 0.004 103 Costa Rica-L 0.14 0.07 0.02 0.07 0.041 91 El Salvador 0.05 0.01 0.01 0.03 0.006 21 N icaragua 0.11 0.07 0.06 0.04 0.027 61 GuatemalaL 0.22 0.07 0.12 0.03 0.034 138 Honduras-L 0.28 0.11 0.03 0.00 0.000 201 Costa Rica-S 0.09 0.02 0.02 0.06 0.022 51 Honduras-S 0.10 0.03 0.02 0.01 0.000 89 Guatemala-S 0.08 0.05 0.06 0.03 0.000 66 Table 2-4. Percentage of change values from 2002 to 2005 Country VolW Pop Number Net Cost of Energy Admin Op Deliv Served Connect Length Work Costs Exp Costs Panama 17% 6% 24% 26% 2% 54% -22% 18% Costa Rica-L 2% 27% 17% 13% 54% 15% 251% 25% El Salvador -11% 7% 7% 0% 8% 2% -18% -10% Nicaragua 11% -4% 14% 5% -1% 14% 17% 4% Guatemala-L 1% 24% 4% 10% 1% 16% 95% 18% Honduras-L 0% 0% 0% 0% 0% 0% 0% 0% Costa Rica-S 12% 10% 10% 7% 16% 109% 45% 29% Honduras-S 14% -4% 30% 38% 57% 18% 297% 315% Guatemala-S 0% 0% 0% 0% 0% 0% 0% 0% $/m3

PAGE 54

54 Table 2-5. Summary of quality indicatorsaverage values from 2002 to 2005 Quality Continuity Complaints Leaks Country/ Units /Connect /km Panama 69 21 0.03 6 Costa Rica-L 98 24 0.70 3 El Salvador 90 0 0.20 10 Nicaragua 100 20 0.24 3 Guatemala -L 100 9 0.01 5 Honduras-L 100 7 0.13 0 Costa Rica-S 99 24 0.30 22 Honduras-S 98 24 0.13 12 Guatemala-S 0 22 0.11 5 Table 2-6. Total factor productivity indexes from 2002 to 2005 Index Laspeyres, Paasche &Fisher TFP Tornqvist Output Variable Volume #Connect Volume #Connect Panama 51% 53% 31% 32% El Salvador -17% -6% -9% 0 Nicaragua -5% 1% -5% -2% Table 2-7. DEA technical efficiency a nd scale impact on efficiency for 2005 Variable Returns Scale Impact on Efficiency Inputs Labor Labor Labor Labor Labor Labor Netlength Netlength Netlength Netlength GNI2005 GNI2005 GNI2005 Panama 1.00 1.00 1.00 1.00 1.00 1.00 El Salvador 1.00 1.00 1.00 0.73 1.00 1.00 CostaRica-L 0.62 0.62 0.63 0.60 0.92 1.00 Nicaragua 0.59 0.70 1.00 0.59 0.99 1.00 Guatemala-L 0.52 0.85 0.85 0.72 0.98 1.00 Honduras-L 0.34 0.38 0.99 0.99 0.97 0.44 CostaRica-S 1.00 1.00 1.00 1.00 1.00 1.00 Guatemala-S 1.00 1.00 1.00 0.62 0.62 0.62 Honduras-S 1.00 1.00 1.00 0.39 1.00 1.00

PAGE 55

55 0 100 200 300 400 500 600 700 01000200030004000 Number of Workers Figure 2-9. Number of conn ections and workers by 2005 Table 2-8. Efficiency rank a nd operating cost reductions Country Frontier Rank Cost Reduction in 2005 Nicaragua 1 2% Panama 2 2% El Salvador 3 3% Costa Rica-S 4 3% Guatemala-S 5 4% Costa Rica-L 6 3% Guatemala-L 7 12% Honduras-L 8 14% Panama Costa Rica Nicaragua Guatemala Honduras El Salvador Costa Rica Number Of Connections (Thousands)

PAGE 56

56 CHAPTER 3 SECTOR FRAGMENTATION AND AGGREGATION OF SERVICE PROVISION IN THE WATE R SECTOR The policy decisions issued in 1995 by the government of Peru pointed towards restructuring the water sector by means of servic e decentralization to municipalities. The aim was to give more decision-maki ng responsibilities to those with more information about costs and production conditions (Tamayo, Barrentes, C onterno and Bustamente 1999; Corton2003). In its 2002-2006 management report, the Peru wate r regulatory agency (SUNASS) identifies key elements of the current situation of the water an d sanitation sector: political interference in water service providers decisions, sector fragmenta tion, the lack of sector investment, and tariff stagnation (SUNASS 2006). The sector fragmentation issue is addressed in this study. Sector fragmentation refers to the presen ce of a large number of very small service providers. Of the 49 service companies comprising the sector, 10% provide service to less than 1,000 connections and 39% have between 1,000 a nd 10,000 connections. Large utilities serving more than 40,000 connections represent 16% of th e sector and those util ities serving between 10,000 and 25,000 connections represent 25% of the sector. The average number of municipalities served by large companies is 17. Medium size companies serve on average 5 municipalities and very small firms provide serv ice to one municipalit y. Beyond the financial sustainability of these companies, a key aspect is that these smaller entities will not always have the fundraising capabilities to invest in the water networks systems. Consequently, their attraction to private investors may be limited. SUNASS argued that aggregation of small service providers may enable the achievement of economies of scale, providing se rvice to a larger customer base at a lower cost (and/or with higher quality). Moreover, professional capacity in a large scale operation may be enhanced

PAGE 57

57 together with other potential pr oduction and service efficiencies. These efficiencies result from joint administration and operation, which renders the sector more attractive to private investors. The objective of this study is to analyze the structure of the water industry in Peru, to investigate the presence of economies of scale an d cost efficiencies considering the region where the firm is located and specific production characte ristics. The aim is to provide policy makers with information about whether aggregation of service providers is optimal based on their location and specific characteristics. The degree of scale economies will help to determine the merits of aggregation of small service provide rs. If there are no economies of scale, splitting large service providers by localities of service may be an optimal decision from a sector structure point of view. Otherwise, aggregation is recommended. All companies for this sector provide sewerage service so cost efficiency is examined considering the volume of water billed and the number of sewerage connections as outputs of the production process. Volume of water loss is included as an additional output. Water losses originate from pipe leaks and non authorized connections. In the wa ter sector of Peru, volume of water loss represents 47% of volume of water produced. The intuition behind the inclusion of this variable as an output comes from assuming that it is hard for firms to control commercial losses if that entails denying the service to the poorest segments of the population. In this sense water lost is interpreted as a valued output by consumers. In addition, producing water has no cost for the firm other than the marginal cost of treatment. Managers consider a tradeoff between repairing pipe leaks and incr easing water produced when satisfying water demand. Under this framework, the hypothesis is that water service providers jointly produce water lost and billed. The input factors considered are labor and capital.

PAGE 58

58 The study utilizes a data set of 44 water se rvice companies, for the period 1996 to 2003. Particular attention is given to the modeling process to ensure accuracy on capturing the behavior and production conditions of these utilities. Cons idering a long run scenario for the analysis, where all factor inputs are variable, the ba sis for the economic model is a total cost translogarithm functional specifica tion. This functional form gives flexibility to the production technology not imposing constant returns to scale, allowing efficiency to vary with scale. Production practices differ according to the regi on where the utility is located. Technology shifts may occur as a result of variations in na tional investment and GDP, as well as time which represent possible technological changes in the s ector. A set of specific characteristics to the utility production process are also included su ch as network density, the number of active connections, continuity, and the number of municipaliti es served. The econometric approach is based on the estimation of a stochastic frontier wh ere inefficiency is allowed to vary over time. The intuition is that the time pe riod under analysis is long enough to expect that firms may have modified their cost efficiency behavior. Results from the analysis found the produc tion technology not homothetic implying that the input mix varies with the scale of the fi rm. In addition, the produc tion technology was found to be no homogeneous which means that returns to scale varies wi th the scale of the firm. These findings implies that any merging or aggregation of these firms re quire detailed analysis of the particular production characteristic s of each firm. Utili ties in the forest region and small firms on the coast exhibit economies of scale which su pport the aggregation pr ocess. The introduction of competition is suggested in the mountains and the coast. Industry Framework Following Garcia and Thomas (2001)s analysis of the French water industry the water sector production process is characterized by five m ain functions The first function, production

PAGE 59

59 and treatment, covers operations for water extracti on, either from underground or surface, and preliminary treatment (disinfect ion, filtering, softening). In the case of Peru, water as a commodity has no cost for the company and the possi bility of water purchas es is not considered in this analysis. Thus, costs incurred while performing this function comprise only water extraction and preliminary treatment. The second step is to transport water from production facilities through transmission pipelines. This can be done by means of pu mping or by gravity. Costs from performing this function are proportional to the amount of pumping required. Storing water in facilities such as water tanks or towers is the third function. The fourth produc tion process is comprised of pressurization of water from the stocking facilities into pipelines, which is done by installing pumping stations. Finally, distribution of water to final customers through distribution mains and customer service lines complete s the production process. These fi ve functions are all performed by each service provider in Peru. Thus, th ese utilities are ver tically integrated. Each service company is located within one of Perus geographical regions: mountains (19 companies), forest (7 companies) and coast (18 companies). The water source found in these regions, surface or underground, requires a different production technology. In general, surface water does not need pumping but may need a more intense water treatment to meet quality standards required before distribution. In th is study, the region in which each company is located identifies production tec hnology differences and they are included in the economic model. Volume of water sold to different types of customers, population serv ed and the number of connections are among the most generally identi fied outputs in the water industry literature. Recent empirical studies indicate that authors view service companies as producing other goods

PAGE 60

60 besides the classical output, vol ume of water delivered. Renzetti (1999) identifies water sold to residential and non-residential customers in the an alysis of cost supply and pricing practices of the water and sewerage utilities in Ontario. Saal and Parker ( 2000) utilize the resident water supply population served in their study of the England and Wales water sector. Garcia and Thomas (2001) define volume of water loss as a product, in addition to volume of water delivered in the French water sector. Aube rt and Reynaud (2005) ut ilize volume of water delivered and the number of customers served in a variable cost function for the Wisconsin water sector. Saal and Parker (2004) c onsider water volume delivered and the number of connections in a study of the England and Welsh water industry. Neuberg (1977) introduced the separate market ability of components as a necessary (but not sufficient) condition to define the vector of outputs. This issue, followed by several authors in the electricity sector, has not been explored in the water se ctor (Hattori, Jamasb, and Pollit 2005; Lowry, Getachew and Hovde 2005; Jamasb and Pollit 2003; Hattori 2002 and Rossi 2001). Neubergs argument is that identification of outputs needs to take into account what the company is actually selling. Following this line of thought, a nd according to the water pr oduction process previously described, volume of water bill ed and lost as well as sewerage connections are the outputs identified in this analysis. Between 1996 a nd 2003, water customers in Peru were charged according to volume of water usage. Water billed is the final output to these customers and it is a main driver of costs for servic e providers. Water received by end customers have gone through the entire production process of these companies, from production to distribution so the final product, water billed has costs associated to each segment of this industry. Volume of water produced is not a cost driver in this industry gi ven that costs associated to production are only

PAGE 61

61 marginal. Water produced is not the final product to be sold but rather it re presents an input to the provision of water service. Regarding water lost, two important elements are addressed to support its consideration as an output in this sector. First, in this particular industry where water utilities operate all the segments of the service production and producing wa ter implies only the marginal cost of water treatment, the extent of water losses partially reflects a cost tradeoff between increasing water production and repairing network leaks to keep up with water demand. Managers may find it more costly to repair leaks than increase wa ter production to satisfy demand. Pipe leaks require costly maintenance outlays, particular ly on long or dispersed networks. Alternatively, when considering the distribut ion network, water losses are related to non authorized connections. Some aut hors refer to this issue as water theft or commercial losses, hence the non revenue water terminology. Given the characteristics of this country where a large portion of the population lives in poverty conditions1 it may be hard for water utilities to control water lost, if that entails denying the service to the poorest segments of the population. In this sense, water lost may be valued as an output by a significant segm ent of the population. In addition, managers, and in particular the members of the company board, may find it more advantageous to keep water needs of this populati on segment satisfied if that secures wining their votes during municipal elections. Regarding the measurement of the water lost share due to costs tradeoffs and which corresponds to non authorized connections, data av ailable on pipe leaks is not reliable. Counting the number of leaks require either costly electron ic equipment to measure the flow of water or a maintenance routine to manually do the counting which requires staff and vehicles. None of 1 A level of 48.7% of poverty by 2005 is reported by the World Bank: http://go.worldbank.org/AHUP42HWR0 (last visit: 05/25/2008)

PAGE 62

62 these mechanisms are reported as used by th ese water utilities. Da ta on non authorized connections is not available. St ill, some water may actually be wasted which from a welfare perspective the waste of a natural resource is a loss. However, from the firms perspective it seems that benefits from the costs savings originat ed from not repairing the leaks as well as the benefits from securing the votes from a partic ular segment of the population are high enough to keep them producing water losses. In the Peru wa ter sector the median volume of water lost is 50% of volume produced. To put this number in perspective, water utilities from developed countries have 15% of water volume lost. For La tin American countries this value equals 60% and 35% for developing countries (Kingdom, Liem berger and Marin 2006). Two leading studies on this issue are Garcia and Thomas (2001) who found a tradeoff between water produced and network leaks in the French water industry and Antoniolli and Filippini (2001) that included percentage of water loss as a firm specific charac teristic in the analysis of the Italian water industry. Under this framework, the joint production of water loss and water billed is considered in this analysis. Water lost is calculated as the difference of volume of water produced and billed. In addition, all service companies in Peru o ffer sewerage services. Sewerage production accounts for a high portion of a services costs, but it is not charged separately from water. For each 100 water connections a median firm serves 85 sewerage connections. Thus the number of sewerage connections is identified as an output for this production process. With respect to water quality, it could be viewed as an output of the water service provision process depending upon the country and the value consumers place on it compared to the provision of the usual outpu ts in this sector: water a nd sewerage. Without doubt, in developed countries where water and sewerage coverage rates ar e almost 100%, water quality is

PAGE 63

63 highly valued by final customers and its product ion implies high costs given the high quality standards in these countries. In the case of Peru, by 2004 the population was 27.5 million people of which 74% were located in urban areas. Of th is urban population, 75% is within the scope of service of water service companies. On average, service companies deliver water to 84% of the total population in their ar ea of service. From the perspective of customers in these countries, it is reasonably to assume that quality of water is of less value comparing to water or sewerage service. In addition, water quality standards are still basic when compared to those of developed countries. The costs associated to meeting these standards are marginal compared to other costs in the production process such as ly ing pipes or installing water pumps. A variable that represents a measure of quality of service is continuity. This variable is defined as the average number of hours a day cu stomers receive water service over a year. This characteristic is related to water pressure within pipes, which is determined by the placement of compressors along the network, the volume of wate r, and its flow within pipes according to differences in gravity. When the source of water is located in the mountains or hills, gravity promotes water flow. When serving customers in a flatter geography, the amount of required pumping is higher. Costs will be higher for firms located in the coast, compared to costs for firms in the mountains or the forest when maintain ing the same level of c ontinuity. This variable is considered in this analysis not as an output but as a characteristic of each utility service provision. Finally, SUNASS classifies service comp anies according to the number of water connections, as large, medium and small. La rge companies are those with more than 40,000 connections (8 companies); small companies are those with less than 10,000 and more than 1,000 connections (19 companies); medium size comp anies are those with number of connections

PAGE 64

64 between 10,000 and 40,000 connections (17 companies). In this study, each firm is classified as belonging to one of these groups according to the companys number of connections in 1996. The ratio of active connections to total connections is interpreted in this analysis as an indicator for service/maintenan ce efficiency. Its median value was 84% in 2003. A connection is a piece of equipment subject to functional proble ms and wearing out. Connections are classified as inactive when they are not able to deliver wate r to the customer. This occurs when there is a functional problem, which once fixed allows the c onnection to be active again. Another possible reason for a connection to be inactive is suspension of service due to non payment, but this is not the case in Peru. If adequate maintenance is provided on a regular basis, all connections are supposed to be active during their operating life. If keeping all c onnections active is costly, firms may prefer to maintain a certain amount of inac tive connections to reduce costs. Thus, including this variable allows one to examine the impact on costs of keeping connections inactive reflecting maintenance inefficiencies. Model Specification and Data This analysis assume s that the firm acts to minimize its long run costs where all input factors are free to adjust and output is exogenously determined by the obligation to supply customers. The assumption that managers are cost minimizers may seem strong in the context of this sector. However, the length of the period un der consideration might help to smooth out any differences from managerial objec tives. Municipal leaders stay in power usually for one year. Network length is commonly used as a proxy for capital in empirica l work. The rationale for doing this is the high amount of capital necess ary to lay down pipes co mpared to the capital needs for other types of network system developm ents such as pumps or treatment facilities. Table 3-1 depicts change in network length for each differe nt group of firms illustrating a considerable expansion of the ne twork systems in this sector.

PAGE 65

65 Peru is not an exception regarding the difficulties Latin American companies face when calculating price of capital. Finan ce expenses plus depreciation di vided by length of network is considered as a proxy for price of capital in this study. An increas e in a firms debt is assumed to occur when the company expands its network length. Further, finance expenses are assumed to reflect this increase and s ubsequently the network expansion. This variable will capture characteristics related to manageri al debt financing abilities and th e possible value of investing in network developments. Nevertheless, the proxy for capital utilized is under-representing the relative risk of investing in this sect or as opposed to in other markets. Labor costs represent approximate ly 60% of firms total cost. For the purpose of the cost analysis, labor is classified into two types acco rding to the contractual obligations acquired by the utility at the time of hiring. The direct labor type is comprised of workers who have permanent positions in the company and are ent itled to the companys labor benefits. The indirect labor group is comprise d of workers under explicitly limited terms of employment, regarding time and salary. This group has no firm working benefits other than the monetary amount agreed at the initial time of work contract or agreement. Prices for direct and indirect labor are calculated as annual labor costs divided by the numb er of workers for each case assuming that the number of hours employed by indirect and direct workers on average is similar. Table 3-2 shows summary statistics for the identi fied outputs and price of inputs. Total cost is comprised of sales cost, sale s expenses, administrative expenses, finance expenses and depreciation representing total operating expenditures. It is assumed that accounting definitions adopted by all firms in the sample are the same and that the depreciation value is based on a similar estimation procedure fo r the assets in place. The median value for

PAGE 66

66 total annual costs in current US Dollars2 for large utilities is $16 whereas it is equal to $2 and $ 0.4 for medium and small firms respectively. A transcendental logarithmic specification provides a second-order approximation for the cost functional form. Its advantages are that it places no a priori restriction on substitution among factors of production and that it allows scale economies to vary with level of output, not imposing homotheticity or unitary elasticity of substitution. These conditions are tested at estimation time. Three outputs and three input fact ors are considered in this cost analysis. The use of a translog functional form is appropriate b ecause it readily allows fo r the identification of scope economies among the set of outputs by observing the sign of the coefficients for the outputs interacted terms. The coefficients of these terms measure the increase in marginal cost of one of the outputs when the other is increased. It is of interest in this study to examine possible scope economies between water delivered and lo st and between sewerage and water. A finding of diseconomies of scope between any of these outputs suggests that th e production process need to be re-assessed possibly se parating the production of the output involved. However, a weakness of this specification is that it assumes symmetry between any pair of outputs, which may not necessarily apply here. In addition, in a translog functional form speci fication the coefficients for the interacted price terms identify the marginal cost of increa sing one of the inputs as the price of the other increases. If the coefficient is positive, it indica tes possible substitutability. Otherwise, it signals complementarities among the input factors. In this study it is interesting to examine possible substitutability or complementar ities between capital and labor. Findings in one direction or 2 Exchange rates obtained from Perus Central Bank web site; http://www.bcrp.gob.pe (last visited 05/24/2008)

PAGE 67

67 another may suggest over-usage of the input fact or involved. Examples of a translog functional specification in the water sector are Aubert and Reynaud (2005) in analyzing the Wisconsin water utilities; Saal and Parker (2004) in the English and Welsch water industry; Garcia and Thomas (2001) for the French water sector; Fabb ri and Fraquelli (2000) in the Italian water industry and Asthon (1999) in the English and Welsch water industry. Omitting firm and time subscripts for clarity, the economic m odel is specified in Equation 3-1. 1 ln lnln lnln 2 1 lnln lnln 2mm n mYmiiYYmn mim n ijij Yjmj ij mjTC YP YY PPYP (3-1) In Equation 3-1, TC is tota l costs defined before; Ym is the vector of the m outputs already identified (volume of water billed-Y1, the number of sewerage connections-Y2, and volume of water lossY3); Pi is the vector of i input pr ices previously identified (P1, P2 are prices for direct and indirect labor respectively and P3 is the price of capital); and the s and s are parameters to be estimated. According to Equation 3-1, a firms costs are represented by a single functional form common to all firms, so there are no system atic differences in technology among them. To account for possible technology diffe rences, dummy variables representing the region where the firm is located are included in the first order price inputs and output coe fficients in Equation 3-2. 31 ;313 10 3 10RbRbbRaRaaiii i M MM YM (3-2) R1 takes the value one if the firm is located in region 1 (Mountains), and zero otherwise. R3 takes the value one if the firm is located in re gion 3(Coast), zero otherwise. The region forest (R2) is captured in the intercept, which in this case corresponds to the coefficient of the first order outputs and price of inputs. This se lection does not affect the results. The as and bs are

PAGE 68

68 parameters to be estimated and represent devi ations from region 2 taken as a base. This functional specification focuses on how outputs and price of inputs on the coast and the mountains differ from thos e in the forest region. In characterizing the production process for th e Peru water and sewerage sector, several variables specific to each service company have been identified (environmental variables): continuity of service (CONTINUITY) and maintenance efficiency repres ented by the ratio of active to total connections (SMAINTENANCE). The number of districts served by each company (NDISTRICTS) was assumed to repres ent possible political factors reducing costs according to findings from Corton (2003). Disagreements or agreements among province and district municipalities leaders were assumed to have an imp act on the number of localities served by a company. Alternatively, the number of localities may capture the effect of having to serve a larger area. Additional localities may need additional local offices staff and maintenance service, which may represent increased costs. Th is variable is also included in the present analysis. Network density (NDENSITY) equals the number of total connections divided by the length of the network. The length of the networ k includes transmission and distribution pipes up to the customer connection. Water companies with a similar size, as measured by number of connected properties may have different costs because of network differences, such as its length and type of customers. Larger firms could have lower costs due to greater network density rather than a scale economy benefit per se. The bigger is the value of this variable, the denser the network (more connections per km of pipe). As network density increases, average costs are expected to decline considering that fixed costs are spread over a larger amount of customers (connections). This variable distinguishes possi ble firms gains due to scale economies from

PAGE 69

69 those related to the network characteristics. It is important to notice that in this analysis all customers (and connections) are assumed to be residential. All these environmental variables vary with the region in which the utility is providing service. They are introduced in the economic mo del interacted with the region dummies, in the same way inputs and outputs are included. Table 3-3 depicts desc riptive statistics for these variables. Finally, besides considering a time trend to capture possible tec hnological changes over time, gross domestic product (GDPv) and national investment (NIv) vari ations are included in the model to capture specific country characte ristics related to annual economic circumstances. The assumption is that technological changes occurring between 1996 and 2003, which are captured by a time trend variable, may have a diffe rent impact on sectors total costs than the possible effect of the annual economic cha nges captured by these variables. After 2001, GDP annual percentage of change increased, yet nati onal investment level for 2000-2005 was only one third of that in the earlier five year period. The government no longer had the capacity to fund infrastructure projects so it is plausible to believe that firms costs were affected. For instance, firms had to raise capital to continue system expansion which may have had measurable effects on utilities total costs. In this analysis, the environmental variables identified are included in the intercept so they explain cost shifts, rather than technology shape or inefficiency. This choice translates into the following specification for the intercept of the economic model defined in Equation 3-1. 02 1 3lnln lnln1ln3TtGtNItZRitZRit ZRit jjjTGDPvNIvZZRZR (3-3)

PAGE 70

70 In Equation 3-3, 0 is the intercept common to all firms; T is the time trend; GDPv, NIv are time varying common for each firm; Z is the vector of identified environmental variables; R1 and R3 are the dummies for the mountains and coast regions, respectively. Empirical Results The model s pecified in Equations 3-1, 3-2 and 3-3 is estimated as a stochastic cost panel frontier. To address the possibility of ineffici ency changing over the an alyzed period, time is included in the estimation process by following Battese and Coellis (1992) functional form specification for the inefficiency component, (uit ) as it was explained in Chapter 1. For purposes of estimation, the usual id iosyncratic error term (v) is added to the econo mic model. This error term is independently and identical distributed following a norm al distribution with zero mean, and it is independent from the regressors. Es timation proceeds using a balanced data set consisting of 42 companies comprising 336 observations. Regularity of the Cost Function Table 3-4 shows results from estim ation. A well behaved cost function is concave in input prices and non-decreasing in outputs. Assuming the cost f unction is twice continuously differentiable, a necessary and sufficient condition for it to be concave in pri ces is that the matrix of second order partial derivatives of the cost f unction with respect to pr ices be negative semidefinite. In the case of the translog flexible f unctional form, for this to hold it is necessary to impose symmetry on the parameters of interacted price terms. This is accomplished by applying mn nmYYYY for ym yn ym ym and jiij for i j. In addition, following Diewert and Wales (1987), the price shares need to be positive over the price domain. These conditions hold for all regions.

PAGE 71

71 A cost function must be homogeneous of degr ee one in prices to correspond to a wellbehaved production function. This implies that for a fixed leve l of output, total cost must increase proportionally when all prices increase proportionally. This translates into applying to the parameters of the model the co nditions specified in Equation 3-4. 1,0,0iY ii ji ji j iiiji j (3-4) Imposing these restrictions is equivalent to normalizing prices and to tal cost by one of the prices. In particular this is done by dividing total costs, price of direct labo r and price of capital by the price of indirect workers (P2). The selection of the price for normalization does not alter the results. A cost function corresponds to a homothetic production technology if and only if the cost function can be written as a separable function in output and factor prices. This implies assuming that the input mix is constant with scale. A homogeneous technology is a special case of a homothetic technology when the elasticity of cost with respect to output is constant. Homogeneity implies that return s to scale are invariant to th e production mix and scale of the firm. For the case of the translog model, the homot heticity and homogeneit y conditions are tested using the Likelihood Ratio test after imposing the re strictions specified in Equations 3-5 and 3-6 (Diewert 1974). Homotheticity requires: 0 Yi (3-5) Homogeneity in outputs requires:0;0 YY Yi (3-6) The sequence for testing these restrictions follows Christensen and Greene (1976). A chisquared one-sided upper tail test at significance level of 0.001% rejects the null hypothesis of

PAGE 72

72 homotheticity and homogeneity. The highly stat istically significant coefficients for the interacted price terms evidences that unitary elasticity does no t hold in this data set. Efficiency For the fr ontier estimator the likelihood function is expressed in terms of the variance parameters, 2 = V 2 + u 2 and gamma, = u 2 / 2. The closer the value of gamma is to one, the more inefficiency as opposed to noise acc ounts for explaining the model disturbance term variance. A gamma value of 0.41 indicates that the presence of noise (or unobserved firm specific characteristics) is still important in this data set. The coefficient for the parameter included in the ine fficiency specification is positive and highly statistically significant. This indicates that the exponential specification imposed to the inefficiency term is appropria te for explaining its behavior fo r the service providers in this sector. The mean value for inefficiency, ui, is statistically significant at a 95% level, which indicates that inefficiency does ha ve an important explanatory role in determining costs in this sector. It is worth noting that this term may contain unobserved time variant firm specific effects not captured by the variables specified in the economic model. Figure 3-1 illustrates the move ment of firms from 1996(dashed line) to 2003 (solid line) with respect to the optimal frontie r represented by the horizontal line y=1 (the closer the value to 1 the more efficient is the firm). On the horizonta l axis firms are sorted by size and the vertical axis represent technical efficiency values. Tabl e 3-5 shows statistics for the extent to which minimal input usage differs from actual input usage in 2003, represented by the estimate of minus the natural log of the technical efficiency via E (uit | eit)). Small companies are slightly more efficient than large size companies. The most efficient small firms are located in the coast. Firms located in the mountains are simila rly efficient regardle ss of their size.

PAGE 73

73 Table 3-6 shows cost reduction statistics calculat ed as the ratio of actual cost less predicted cost divided by actual cost. If firms were to beha ve efficiently with respect to their predicted costs for 2003, the reduction of costs found for sm all firms is on average of 4%; for large firms this reduction is 16% and for medium size firms it is 17%. There is a noticeable difference in reducing usage of resources for large firms de pending upon their locati on. The only large firm located in the forest should reduce by double the amount of re sources with respect to the reduction amount for large firms on the coast. Ther e is also a noticeable difference in the amount of cost reduction among firms located on the coast. Small firms are able to reduce their costs by 50% less than the amount large firms would reduce theirs. At the same time, large firms on the coast are able to reduce their cost by 50% less than medium size firms would reduce theirs. Environmental and Output Variables Turning to the interpre tation of coefficients, neither national investment nor GDP variations are statistically signi ficant. Time is not statistica lly significant either. A possible explanation for this may be the fact that moneta ry variables converted to current US Dollars may have captured country economic fluctuations. An alternative model using soles, the currency of Peru, instead of US dollars yields the time coefficient statistically significant at a 90% level. The coefficient is positive but very small in magnit ude indicating a small economic impact. Still in this model, neither GDP nor national investment variations are statis tically significant With respect to the number of districts (NDI STRICTS), Corton (2003) previously reported cost savings for firms serving more than 5 district s. In the present analysis, this variable is not statistically significant. An alternate mode l was estimated, replacing NDISTRICTS by a dummy which takes value equal to one if the firm serves more than 5 localities. Estimation results were similar. Further analysis with respect to th e provision of service according to the number of districts and provinces is included in Chapter 4.

PAGE 74

74 The coefficient for the ratio of active to to tal connections (SMAINTENANCE) for firms at the coast in comparison to those in the forest or the mountains is statistically significant. A 1% increase in this variable (which implies an increase in the number of active connections) produces a decrease in costs of approximately 0.5% only for firms located at the coast. This variable was included as an indicator of maintenance effi ciency assuming that keeping connections active has an impact on total costs. An explanation for firms on the coast to have lower costs as they keep a highe r level of connections active implies that these firms are more efficient compared to firms in other regions from the view point of maintenance. It may be the case that these firms have maintenance schedu les under control such that replacing costly connection equipments are kept to a minimum. It is feasible to believe that the smaller territory and a not challenging topography compared to the mountains and the forest provide a comparative advantage to these firms when it come s to maintaining the operational level of the network. The network density (NDENSITY) coefficient is negative and statistically significant at a 99% level. Differences among regions do not have an impact on network density. An increase of 1% for network density produces a decrease in costs of approximately 0.4%. The possible effect of a denser network on average cost s is clear. On this total cost model this variable is capturing the scale of the network. In this sense the economie s of scale results will be net of this network dimension. The negative coefficient for this variable implies that firms are still in the position of adding more connections to their networks, a reasonably result consid ering the under-service issue in this sector. Differences in continuity (CONTINUITY) among regions are statistically significant. The negative signs on coefficients for the coast and m ountains mean costs savings when continuity is

PAGE 75

75 increased independently of the topographic conditions of these two distinct regions. A possible interpretation for this is higher continuity im plies higher efficiency and consequently lower costs. If efficient firms are those delivering highe r continuity levels, then efficient firms exhibit lower costs. Overall, there may be other factors interacting with the possible effects of this variable in addition to the topographic aspects considered here. An alternative model excluding continuity was estimated. The differences in the ma gnitude of the coefficients were small. The signs of the coefficients a nd statistical significance of the variables did not change. The first order coefficients for volume of wate r billed (Y1) are not st atistically significant. The coefficient for the interacted term with volume of water lost (Y 1Y3) is statistically significant at the 99% level. This finding indicat es that the volume of water billed does not explain costs by itself, but rather it does so by being jointly produced with volume of water lost. The coefficient is positive indicating diseconom ies of scope when producing these two outputs together. A 1% increase in deliveri ng water billed and lost jointly implies an increase in costs of 0.1% for the firm. This is the cost may be interpreted as the cost the firm pays for allowing non authorized connections to exist. On the other hand, the coefficient for volume of water lost (Y3) is ne gative and statistically different from zero. The negative coefficient and its magnitude explain why firms allow water to be lost. Producing water lost reduce costs by approximately 1.3% for each 1% its volume production is increased. This result can be interpre ted as the cost savings from not repairing the leaks. In this way, the cost tradeoff hypothesis is supported. The net effect from the cost tradeoff and the diseconomies of scope yields costs sa vings around 1% for each 1% increase in water lost.

PAGE 76

76 Two alternate models were estimated to check the sensitivity of this model. Volume of water produced instead of billed not including wate r lost yields a much larger residual. A large residual indicates the presence of missing explanatory variables. In this case it suggests the explanatory importance of water lost in the model. In addition, none of the coefficients for water produced is statistically signifi cant indicating that th is variable has not explanatory power on costs. Another model was estimated including water lost as an environmenta l variable instead of an output. A chi-squared one-sided upper tail test at significance level of 0.001% rejects the null hypothesis of not including Y3 interacted w ith the other outputs and input factors. Regarding the number of sewerage connections (Y2) is positive and statistically different from zero indicating that total costs increase as number of sewerage connections increase. The coefficient for the squared term (Y2Y2) is sta tistically significant and positive indicating that marginal cost increases as the number of sewerage connections increases a nd that this effect has explanatory power on costs. The coefficient for the interacted term w ith water lost (Y2Y3), although small in magnitude is negative and statistically different from zero, indicating economies of scope when producing these two output s. The interpretation of this result is complicated by the imposed symmetry. When adding the interacted term Y1Y2Y3 to the model, only Y1Y3 is statically si gnificant and s till positive. Economies of Scale The assessment of econom ies of scale is funda mental for the characterization of the most economical efficient structure organization. Evidence of economies of scale supports a smaller number of firms to supply industry output, wher eas diseconomies of s cale indicates introduction of competition as more efficient for the mark et structure. Following Christensen & Greene (1976) and Hanoch (1975), economies of scale in this study are defined in terms of the relationship between total cost and output along the expansion path, where input prices are

PAGE 77

77 constant and costs are minimized at every level of output. This is considered the appropriate scenario for this anal ysis given the cost minimization framework. See Hanoch (1975) for details on the alternative approach implyi ng the relative increase in output due to a proportional increase in all input quantities. The measure utilized to calculate economies of scale is the el asticity of cost ( ) with respect to output, which is defined as the proportional increase in cost resulting from a proportional increase in the level of output. Becau se of the multi-output scenario, economies of scale are referred to as ray econom ies of scale meaning that what leads the less than proportional increase in cost is strictly the same propor tional increase in the le vel of all outputs. Conversely, ray diseconomies of scale are present when a higher than proportional increase in cost occurs after an equal propor tion on the level of all outputs is increased. Following Baumol (1976) and Panzar and Willig (1977), a local m easure of overall scale economies for a multiproduct firm is defined in Equation 3-7. i ifori ES3..1 /1 ; i iY TC ln ln (3-7) Equation 3-7 produces numbers that are higher than one for positive economies of scale; less than one for diseconomies of scale and unity fo r constant returns to scale. The calculation is performed from estimated values for each firm acco rding to region and then firms are classified by size. The standard deviation from calculating cost elasticities for each region is very high suggesting a high degree of variation within each region. Table 3-7 shows economies of scale for each region. To provide some intuition on the interpretation of the scale economies findings, Table 3-8 shows distribution of firms in each region according to their size. There are no large firms located in the mountains. Medium and small firms in this region present small diseconomies of

PAGE 78

78 scale. The coast is the most populated region and firms located there are predominantly larger than firms in other regions. Those large firms located in the coast are experiencing large diseconomies of scale, yet the only three small companies in this region show large economies of scale. Medium size firms located in the coast expe rience constant returns to scale. Economies of scale are present for all firms in the forest. Additional intuition is provided by Figure 3-2 which depicts fitted average total cost per cubic meter of water billed for 2003. Firms are on th e x-axis sorted by size from the smallest to the largest. The shape of the av erage total cost curve shows aver age total cost per unit declining as size of company increases. The slope of the cu rve is steeper for small firms meaning that they are able to enjoy the benefits of larger economie s of scale compared to larger firms. The curve gets flatter as it reaches the largest firm. This suggests that scale economies have already been exhausted by large firms. Overall, for the Peru water sector, the optimal firm size is medium if the firm is located on the coast. If the firm is located in the mountains small may be considered the optimum size (these firms exhibit cl ose to constant returns to scale). Input Factors and Price Elasticity Finally, regarding the price of inputs, the co efficients are all positive as expected. The price of direct labor is not statistically different from zero, yet the price of capital is highly statistically significant when considering price differences for firms located in the mountains compared with those in the forest. The coefficients for the squared terms on both prices, labor (P1P1) a nd capital (P3P3) are statistically significant at a 99% level, meaning th at these prices do have an important role in explaining costs. The sign of th e squared labor coefficient is pos itive indicating that marginal cost increases as labor is added. The sign for the squared capital coefficient is negative meaning that marginal cost decreases as capital (debt) increases. These findings may signal, on average,

PAGE 79

79 over-usage of labor (higher than optimal number of workers) and low levels of debt with the subsequent indication of unde r-investment in the sector. In a translog functional specification the coefficients for the interacted price terms identify the marginal cost of increasing one of the input s as the price of the ot her increases. If the coefficient is positive, it indicates possi ble substitutability. Otherwise, it signals complementarities among the input factors. The coefficient for the interacted prices term (P1P3) is negative and statistically si gnificant at a 99% level, indica ting complementariness between capital and labor. A negative yet small value for this partial elasticity is a consequence of having more than two inputs in the model. Therefore, the degree of complementarities becomes less restricted with respect to the two inputs case. To examine this issue, Morishima partial elas ticities of substitution, which do not impose symmetry among factors, are calculated. The Mori shima partial elasticity of substitution is a measure of elasticity of substitution utilized in the multi input case. The elasticity value obtained for the inputs direct labor and cap ital is equal to one and the valu e for the mirror combination is equal to 0.8. These close to unity values indicate that these inputs are needed in relatively fixed proportions within the production process. Concluding Observations The fragmentation of the water and sewerage sector was among the concerns of Perus water and sewerage regulatory agency at th e end of 2006. Findings fr om this study reveal important aspects about the perfor mance of utilities in this sector. Results from the analysis indicate that when considering the optimal usage of resources, in particular labor and capital, small firms are on average more efficient than firms of large or medium size. This conclusion comes after finding that small firms are closer to the efficient frontier than other companies.

PAGE 80

80 A reduction in resource usage translates into a reduction of costs. If small firms were to behave technically efficiently, th ey would need to decrease their mean cost by approximately 4% over the analyzed period. This cost reduction is equal to 16% for large firms and 17% for medium size firms. The shape of the average total cost curve fo r this sector shows that there are scale economies to be enjoyed by small and medium size firms, indicating possibilities for aggregation. For firms located in the mountains introduction of competiti on is indicated given the presence of diseconomies of scale in this region. The optimal size for a firm located in the mountains ranges on providing service to 1,000 to 10,000 connections. These companies are classified in this sector as small. On the co ast the optimal firm size is serving between 10,000 and 40,000 connections. Political influences affecting large water service providers have been a concern for SUNASS for a long time. Finally, in 2007 rule s were changed to include customer representatives on the board. Nevertheless, the proposal of institutional aggregation by SUNASS comes as a more aggressive approach to address this issue. Findings from this study indicate that large firms on the coast have already exhauste d scale economies, so aggregation in terms of joining assets is not recommended. Introduction of competition in the mountains is also advised. With respect to medium size companies, aggregat ion is indicated by the presence of large scale economies for these firms if located in the forest However, results from the analysis found the production technology not homothetic implying that th e input mix varies w ith the scale of the firm. In addition, the produc tion technology was found to be no homogeneous which means that returns to scale varies with th e scale of the firm. These findings imply that any merging or

PAGE 81

81 aggregation among these firms would require a de tailed assessment of the particular firm production characteristics, incl uding topological, hydrological, and geographical constraints. In a previous study by the author, the number of localities served by a firm is found to be statistically significant and negative. The presen t analysis does not contribute with additional information about this issue. Chapter 4 presen ts a more detailed assessment on this matter. Findings about the set of outputs indicate that the cost paid by firms for allowing water billed and lost to be produced together is 0.1% per each 1% increase of joint production. This cost represent the cost paid by firms by allowing non authorized connections to be present in the network and secure the votes of th e population being served with wa ter that is not billed. When it comes to the production of water lost alone, an increase of 1% of its volume produces a decrease on costs of 1.3%. This cost re duction represents the cost savi ngs from the tradeoff of pumping more water in the network instead of repairing pipe leaks. In conclu sion, firms have costs savings by producing water lost which generate in centives for them to k eep the joint production of water lost and billed. With respect to the number of sewerage conn ections, results indicate that marginal cost increases as the number of sewerage connections increases. This result and the possibility of finding diseconomies of scope for the joint production of water a nd sewerage may set the stage to a more detailed analysis into delivering water and sewerage in a separate way. Regarding price of inputs, fi ndings might signal, on average, an over-usage of labor (higher than optimal number of wo rkers) and low levels of debt with the subsequent presence of under investment in the sector. Overall, comp anies need these inputs in fixed proportions.

PAGE 82

82 Table 3-1. Change in network length between 1996 and 2003 Mean StDv Median Large 25% 16% 23% Medium 60% 53% 51% Small 56% 70% 34% All 52% 57% 42% Table 3-2. Summary statistics for outputs and price of inputs Description Var Mean StDv Median Min Max Volbill Millions m3 Y1 Large 67.6 133.3 26.2 9.2 396.8 Medium 4.7 2.2 4.2 2.0 10.3 Small 1.3 0.5 1.2 0.2 2.4 All 15.1 59.5 3.5 0.1 432.3 SewcoxThousands Y2 Large 198.5 321.6 99.5 36.4 989 Medium 15.9 5.9 14.7 6.6 30.1 Small 4.5 2.2 3.8 1.6 9.0 All 45.7 151.8 12.0 0.9 1,397 VolwaterlossMillions m3 Y3 Large 49.2 90.8 17.1 3.7 272.9 Medium 5.0 2.6 4.9 1.5 9.1 Small 1.4 1.1 1.1 0.05 3.3 All 11.9 41.2 3.2 0 317.4 PriceDWork $/worker P1 Large 12.7 4.0 13.2 6.8 20.3 Medium 10.2 6.1 9.1 6.2 34.2 Small 7.7 2.5 7.1 3.1 12.1 All 9.7 5.4 8.8 1.7 40.4 PriceIWork $/worker P2 Large 16.2 12.2 17.7 0.4 30.3 Medium 15.9 24.7 4.3 1.0 106.6 Small 8.6 7.8 5.0 1.6 24.3 All 13.2 45.6 4.5 0.1 768.0 Financecost $/km P3 Large 4.3 2.1 4.0 1.6 7.1 Medium 3.1 2.3 2.7 1.0 11.7 Small 2.5 2.2 1.8 0.3 9.0 All 3.1 2.7 2.4 0.1 14.5

PAGE 83

83 Table 3-3. Summary statistics for utilities specific characteristics Variable Group Mean StDv Median Min Max Number of Localities (Z1) Large 17 13.8 17.5 3 41 Medium 5 3.3 3 1 14 Small 1 0.5 1 1 3 Actconx/totconx (Z2) Large 0.82 0.08 0.83 0.65 0.91 Medium 0.80 0.10 0.83 0.57 0.92 Small 0.84 0.08 0.86 0.65 0.94 Network density (Z4) Large 124 33 111 106 204 Medium 125 28 118 85 196 Small 131 52 118 61 226 Continuity (Z5) Large 14 5 14 7 22 Medium 15 5 14 6 22 Small 16 6 18 2 23

PAGE 84

84 Table 3-4. Estimation results fo r the translog cost function Var Coeff Var Coeff Var Coeff T -0.001 (0.0143) Y1 0.749 (1.186) Y1P1 -0.022 (0.0484) NIv 0.036 (0.0356) Y1xR1 0.043 (0.1038) Y1P3 0.117** (0.0455) GDPv 0.005 (0.004) Y1xR3 -0.028 (0.107) Y2P1 0.044 (0.0486) NDISTRICTS 0.047 (0.0777) Y2 2.113* (1.212) Y2P3 -0.054 (0.044) NDISTRICTS x R1 -0.039 (0.0861) Y2xR1 -0.032 (0.1296) Y3P1 0.015 (0.0177) NDISTRICTS x R3 0.046 (0.094) Y2xR3 -0.141 (0.1496) Y3P3 -0.030* (0.0184) SMAINTENANCE 0.363 (0.2265) Y3 -1.347*** (0.4409) Y1Y1 0.034 (0.081) SMAINTENANCE x R1 -0.482 (0.3036) Y3xR1 0.072 (0.0517) Y2Y2 0.189** (0.0966) SMAINTENANCE x R3 -0.882*** (0.280) Y3xR3 0.152** (0.0642) Y3Y3 0.003 0.0037) NDENSITY -0.456*** (0.088) P1 0.475 (0.4096) Y1Y2 -0.258 (0.1694) NDENSITY x R1 0.045 (0.1137) P1xR1 -0.086 (0.0559) Y1Y3 0.129*** (0.0455) NDENSITY x R3 -0.029 (0.1193) P1xR3 -0.033 (0.053) Y2Y3 -0.095** (0.0442) CONTINUITY 0.172 (0.1136) P3 0.06 8 (0.400) P1P1 0.08 7*** 0.0131) CONTINUITY x R1 -0.304** (0.1413) P3xR1 0.210*** (0.0484) P3P3 -0.080*** (0.0125) CONTINUITY x R3 -0.216* (0.1281) P3xR3 0.047 (0.0517) P1P3 -0.107*** (0.020) Intercept -6.380 (5.424) Gamma+ 0.45 (0.170) Eta 0.1149*** (0.0362) Mu (u)++ 0.167* (0.090) Confidence levels: *** 99%; ** 95%; 90% Data set: 336 observations; Standard Errors in parenthesis. Loglikelihood=130.84732; Dependent va riable: Ln (TotalCo st); R1 = Mountains; R3= Coast; Y1= Vol. Water Billed; Y2= Se werage connections; Y3=Vol. Water Lost P1=Price of direct labor; P3=Price of capital + Gamma is defined as = u 2 / 2 where 2 is the sum of the variance for the term parameters noise and inefficiency ( 2 = V 2 + u 2) ++ Predicted Mu (u) after estimation comes from the estim ate of minus the natural log of the technical efficiency via E( uit | eit) where eit is the residual term.

PAGE 85

85 0.0 0.5 1.0 1.5 2.0 2.5 TE96 TE03 Figure 3-1. Technical frontier for 1996 and 2003 Table 3-5. Technical inefficiency statistics for 2003 Mean StDv Median Min Max Large Mountains Forest 0.31 0.31 0.31 0.31 Coast 0.16 0.06 0.16 0.09 0.26 All 0.18 0.08 0.18 0.09 0.31 Medium Mountains 0.22 0.11 0.23 0.03 0.36 Forest 0.17 0.16 0.17 0.06 0.28 Coast 0.22 0.11 0.24 0.07 0.32 All 0.21 0.11 0.23 0.03 0.36 Small Mountains 0.2 0.12 0.21 0.06 0.34 Forest 0.14 0.13 0.12 0.04 0.37 Coast 0.13 0.09 0.15 0.03 0.2 All 0.16 0.11 0.15 0.03 0.37 Utilities sorted by size

PAGE 86

86 Table 3-6. Cost reduction statistics for 2003 Mean StDv Median Min Max Large Mountains Forest 0.76 0.76 0.76 0.76 Coast 0.07 0.26 0.09 0.36 0.39 All 0.16 0.35 0.13 0.36 0.76 Medium Mountains 0.18 0.20 0.12 0.04 0.68 Forest 0.18 0.10 0.18 0.10 0.25 Coast 0.15 0.09 0.13 0.01 0.28 All 0.17 0.15 0.13 0.01 0.69 Small Mountains 0.02 0.03 0.02 0.03 0.07 Forest 0.08 0.12 0.09 0.25 0.05 Coast 0.03 0.06 0.05 0.09 0.04 All 0.04 0.08 0.03 0.05 0.26 Table 3-7. Economies of scale Large MediumSmall All Mountains 0.87 0.93 1.09 Forest 3.33 4.16 1.32 2.27 Coast 0.46 1 2.17 0.97 Table 3-8. Distribution of firms by regions and size Large Medium Small All Mountains 0 9 7 16 Forest 1 2 6 9 Coast 7 8 3 18 All Regions 8 19 16 43

PAGE 87

87 AverageCost Water Volume Figure 3-2. Fitted average total cost pe r cubic meter of water billed for 2003

PAGE 88

88 CHAPTER 4 INFRASTRUCTURE SERVICE PROVISION TO MULTIP LE JURISDICTIONS: AN EFFICIENCY ANALYSIS When the objective is to fully expand infrastructure service across a country, governments face the challenge of considering not onl y the technical characteristics of the sector but also the countrys economic, social and political structure. During the last decade, infrastructure reforms undertaken by Latin American governments have often involved decentralization, transferring se rvice provision responsibilities to municipalities. The water sectors of Peru, Nicaragua, and Honduras are a few examples. The objective was to give more decision-making responsibilities to those w ith more information on costs and production conditions to improve sector performance. However, outcomes do not always match the formal aims of such restructuring initiatives. This study examines the impact on efficiency of providing service to multiple jurisdictions over the period 1996 to 2003. The i ssue analyzed is whether a utility expanding service across multiple political subdivisions is more efficient than a utility expanding service within a single political subdivi sion. In this context, a politi cal subdivision is a unit of government designed to carry out public functions within a sp ecific territory. Utilities are municipal-owned service providers. The hypothesis is that utilitie s answering to a heterogeneous group of jurisdictional authoritie s are less efficient than those reporting to a single authority. Political issues and bargaining for resources may contribute to production in efficiencies affecting utilities costs. The resulting inefficiencies may offset any scale economies associated with serving a larger area. This study is unique in that there is no empi rical study in the infras tructure literature examining the impact on utilities performance of th is type of reform. Results from this analysis provide insights to policy makers on an effici ent governance structure for service provision

PAGE 89

89 under a multiple political subdivisions framework, where the aim is to expand infrastructure service within a country. To illustrate the issue, the water sector of Peru is analyzed. The political process existing in the country during the 1990s and the beginning of the current decade illu strates the political bargaining process in a multiple jurisdiction or ganization. After some initial steps during the early 1990s towards expanding service to under served areas of the country, the second half of this decade represented a slowdown in this initiative, partiall y due to the countrys economic conditions and also because of the political clim ate present at that time (Corton 2003; Tamayo et al. 1999). During the authoritarian government of the 1990s, the country witnessed a breakdown of political parties. To some extent, the resulting political vacuum was filled by a surge of independent political leaders coming from the gr oup of province municipal authorities. Provinces are the largest organized unit of local government within Perus departments. A department is the main jurisdictional subdivision. These provincial municipal leader s tried to appeal to social interests aimed at opposing the authoritarian gove rnment regime. They gained importance as political representatives of social discontent loca lly and at the department al level. These province municipal leaders looked for regional blocks of power by attracting the interests of rural and small towns and small district au thorities, who were initially neglected by government as they were seen as marginal lobbyist s to its central favoritism. However, by overestimating their political capab ilities, this contingent of leaders ended up pursuing individual interests rather than reinforc ing each others political clout in opposing the regime. In a sense, the political pattern refl ected individualism and dispersion rather than a consensual force of opposition.

PAGE 90

90 In this political system, gove rnors lobbied the ce ntral government to obtain resources for their departments. Presumably, these resour ces were directed to provinces where their authorities had a common political ground w ith the respective gove rnor. Evidence of infrastructure development was present in those districts where the capital city of the province was located. Consequently, the presence of political rivalry among province leaders affected policy implementation, particularly with respect to allocation of resources in infrastructure. As would be expected, conflicts of power pe rmeated decisions coming from the board of directors of water service companies due to di vergent and individual political interests. For instance, in the earlier stages of introducing private participation into the sector, the organizational structure of one of the largest utilities (providing se rvice to twenty eight districts) was challenged by a lack of agreement among the directors of the board regarding whether to allow this private participation process to go on in some of the districts served by the company. Some local authorities might have viewed their political power threatened by this initiative. Indeed, the water regulatory agency in Peru is currently assessing the issue of whether having fewer rather than more political authorities for governan ce benefits water providers and sector performance. The central issue from the regulatory perspect ive is that the decentralized framework facilitates political in terference. Requiring these utilities to report directly to a higher authority level, such as the governor, would re duce the number of jurisd ictions overseeing the company, yielding efficiency improvements for the sector. This study examines the effect on utilities costs of providing service to multiple jurisdictions. The analysis involves estimating a stochastic cost frontie r where the number of provinces served, service coverage and whether the water utility is the only provider within a department are variables shifting the cost frontier. Results from estimating the econometric

PAGE 91

91 model support the hypothesis that providing service with in a province is more efficient than extending service to nearby provinces In addition, those utilities that have expanded service to nearby districts show lower costs in the short run. However, expa nding or contracting the service to nearby districts is costly in the long run. This result may be interpreted by a lack of adjustment of inputs to the changed production structure. The result is also consistent with the cost of political interference in the utilitys service provision decisions. Finally, results suggest that the presence of more than one service provider within a department translates into lower costs for these utilities when compared to the costs of a utility serving a department alone. Political Sub-divisions Peru is a unitary decentralized republic divided into twenty four main political subdivisions called departments The largest organized unit of local government within departments is called a province Within a province there is an additional level of unit of government which is called a district There are 195 provinces and 1639 districts in the country. A department has on average 8 provinces and a province has on average 65 districts. Broadly speaking, a department parallels a state in the United States and a provin ce has similarities to a county. Provinces and districts are municipally governed. As a decentralized republic, the governor of each department coordinates with province authorities on the implementation of development plans and allocation of resources. A water company may offer services to more than one district, which is the minimum service unit, and to more than one province. The av erage number of districts served by a utility is five and the maximum number is twenty eight. Available data indicate that each district and province is served by only one utility. Departments may be served by more than one service provider. No utility provides service to more than one department. Table 4-1 displays for each department: population and population

PAGE 92

92 increase, the number of provin ces and districts contained in the department, the number of companies providing service and the number of pr ovinces and districts served by such providers. Note the variation in the number of provinces within departments and the number of districts within each province. A department may have a minimum of one and a maximum of twenty provinces. A province may have a minimum of five and a maximum of 161 districts. Note also how evident is the under provision of service in some departments. For a better understanding of this structure, see Figure 4-1 which repr esents these political subdivisions in a very simplified way. The large squares depict two possible types of departments within the country: type A, which has several water service providers and type B which has only one provider. The smaller squares represent the provinces inside any department. The rectangles inside each province are the di stricts being served by a utility. Groups of rectangles are examples of utilities providing serv ice to several districts. Utility 1 provides service to several districts inside one province. Utility 2 provides service to several districts in two provinces. Utility 3 is providi ng service to three provinces. In each province, this utility delivers service to severa l districts. Note that although depa rtments and provinces seem of the same size they may not be. Districts served by eac h utility are displayed in different sizes. This framework allows one to examine the presence of a competitive effect within departments where there is more than one operator. Departmental authori ties have discretion in allocating resources to provinces. If there is more than one utility serving a department, competitive pressures in requesting additional resources may affect the production process efficiency. Competition for funding might encourage the utility board of di rectors to demonstrate that they are using resources wisely.

PAGE 93

93 Ownership Structure of Utilities Water service companies are owned by the distri cts to which they deliver service. The number of shares owned by each district is determined by the proportion of district to province inhabitants. This number is computed independe ntly of actual population served by the utility. More populated districts have a larger ownership stake on a utility that provides service to multiple districts within a province. A companys board of directors is comprised of a maximum of six members, each with a one year term and the opportunity for reelection. Directors are elected by a stakeholder committee, which is comprised of province a nd district representatives. Each committee representative proposes a candidate to hold a board of directors position. There should be no more than two directors repr esenting one district. Assuming that each district representative proposes a candidate from his own district, province representatives are able to favor one district over others when proposing candidates to the board. When there are homogeneous interests w ithin the board of dir ectors, this ownership structure provides an incentive scheme which favors the expansi on of service to highly populated districts. Providing service to more than one province may introduce a level of heterogeneity within the board of directors that may impact a firms efficiency. Possible c onflicts of interest among board members representing different provinces may get in the way of important economic or strategic decisions. The case of a large utility facing a private participation decision mentioned earlier illustrates how conflicts arise. Contraction and Expansion of Service Provision To further analyze the way these firms provide service, it is useful to examine how firms contract and expand service provision across political subdivisions. Providing service to

PAGE 94

94 neighbor districts could be view ed as a natural way for a company to expand its service, assuming it has the resources to undertake such expansion and there is no other service provider in the area. Service expansion by means of addi ng more connections to an existing network may occur when local demand has not been fully satisfied or the local population is increasing. Additionally, nearby districts may offer an attractive opportunity fo r a firm to decrease average costs in the long run. From an engineering point of view, anticipa ting network capacity and expansion is done before pipes are laid down. Adding pipes to extend the network may involve costs that could have been avoi ded with appropriate planning. Alternatively, several motives might account for a companys reduction of service. The pipeline infrastructure in a location may be very old or in poor condition, so maintenance or replacement could be too costly. It may be th e case that the district in question comprises a poor/rural area where the rate of uncollected bills is high or volu me of nonrevenue water is high. Therefore, by removing those districts that do not contribute to revenue, the company may be able to increase profits or funds available for service improvement. Another cause for removing districts from service may be st agnation or slow district population growth, as people move to more populated areas where economic and infrastructure development is more likely. The assumption when a district is removed from service is that either a neighbor company takes over the service (a ppearing as an increase of districts served by that company), or the district stops receiving service altogether. This may occur in very small districts where service pr ovision by independent provide rs is an alternative. Data available for these for these utilities s how six companies with a reduction of number of districts served, and eleven show an increas e. Approximately 50% of the increases happened

PAGE 95

95 among companies serving more than six distri cts. Most decreases took place in companies serving less than six district s and all of them are located in type A departments. Change in Service Coverage A change in the number of districts served does not necessarily parallel a change in population served. The relationship between popul ation served and the total population in the area of a firms service is known as service coverage. Available data on population served is a total amount representing population served for all districts. Thus, absent more detailed data, there is no way to identify population served by district. Table 4-2 shows statistics for changes in service coverage an d population growth by region. On average, there has been an increa se in coverage of approximately 20% over the period. Coverage has increased twice as much fo r companies located in the mountains (27%), the region with the lowest population growth, compar ed to the coverage in crease of those on the coast (13%), the region with the highest population growth. This outcome is consistent with the governments aim of expanding water servi ce to less populated ar eas in the initial decentralization stage. To relate the change of service across district s to change in total service coverage, Figure 4-2 shows a map of Peru with the departmental division displaying solid ci rcles to indicate those departments where companies have increased the number of districts serv ed and dashed circles to indicate those that have reduced their number. Note that the changes occurred in the number of districts served for utilities in departments located on the coas t or nearby, which are the most populated departments. However, th ere are no changes in the number of districts for utilities in departments where the increase in coverage is highest, such as those in the forest and the mountains.

PAGE 96

96 A closer look at these changes and their relationship with coverage and population increases is given in Table 4-3. Firms that have added districts to their service are located in departments with the highest increase in populati on growth and the lowest increase in coverage (18%). Firms that have reduced the number of districts served are the ones located in departments with the lowest rate of population increase and with th e highest increase in service coverage (28%). A plausible expl anation is that firms increased local service coverage after a reduction of service scope. On average, utilities serving in type A depart ments show an increase in coverage of 14% whereas those in type B departments have an increase of 9%. Tracking the increase in the number of districts with an incr ease in coverage may be interpre ted as a consequence of firms informed/planned decisions aimed at expanding se rvice. Laying down more pipes implies higher costs than merely adding more connections to an existing network. In this sense, the increase in coverage is expected to be more costly if it includes an expansion of network compared with only adding connections. On the other hand, in creasing coverage over the same network may decrease average costs if fixed costs are sp read over a larger set of customers. Model Specification The competitive pressure effect among utilities within the same department is captured by a dummy (MORETHANONE) which takes the valu e of one if the company is serving a department type A, zero otherwise. It is expect ed that this competitive pressure has an effect on utilitys costs such that they are lower with respec t to costs of utilities se rving departments alone. NPROVJ is a variable defined equal to the number of provinces served for firms located in type A departments. It is expected that the larg er the number of jurisdictions served by a utility the higher its costs. This means that while se rving in departments where there are several utilities, utilities serving more rather than less provinces have higher costs.

PAGE 97

97 CHCOVERAGE is defined as the percentage of change in firms coverage from 1996 to 2003. The increase in coverage ove r the period could increase total costs if it has implied laying down pipes to nearby districts which could have required additional capit al and/or increasing debt. NEGCHIDX is a dummy variable which takes a value equal to one when there has been a decrease in the number of districts served, a nd zero otherwise. A decrease in the number of districts served should translate in to a total cost decrease not only in the short run but also in the long run as the company adjusts for a shorter network, fewer workers, lower local office expenses and lower maintenance co sts over the time period. This dummy is interacted with a time trend to capture this l ong term effect (NEGCHIDXT). Alternatively, POSCHIDX is a dummy defined with a value of one when a utility has increased the number of districts served, and zero otherwise. This dummy is also interacted with a time trend to capture the effect of having adde d more districts over the period. The coefficient for this variable represents the cost difference fo r a utility that have extended its service toward nearby districts with respect to ut ilities that have not done so. In the short run, when controlling for coverage, network density and the number of provinces this coefficient is expected to be negative as expanding service to additional dist ricts translates into a larger network and associated scale economies. The dummies R1 for the mountains and R3 for the coast previously utilized for capturing regional differences in the model presented in Ch apter 3 are included in this model in the same way: interacted with outputs and input prices. Network density, continuity of service and a time trend variable are also included in the model. The variables that we re not found to be statistically significant in the previous model are not included. The homogeneity in outputs and

PAGE 98

98 homotheticity hypothesis for the technology was reje cted in the model presented in Chapter 3. Thus, the same translog functional form is kept fo r the specification of this model. Including the identified variables, the economic model is specified in Equation 4-1. 1 ln lnln lnln 2 1 lnln lnln 2mm n mYmiiYYmn mim n ijij Yjmj ij mjTC YP YY PPYP (4-1) Here TC is total costs as defined before; Ym is the vector of the m outputs already identified (volume of water billed-Y1, the number of sewerage connections-Y2, and volume of water lossY3); Pi is the vector of i input prices previously identified (P1, P2 are prices for direct and indirect labor respectively, and P3 is the price of capital); and the s and s are parameters to be estimated. The intercept is specified in Equation 4-2. CHCOVERAGE NPROVJ E MORETHANON T4 3 210 (4-2) A firm is input technically inefficient if it fails to produce maximal output from a given quantity of inputs. Following Atkinson and Corne ll (1994), the general form for a cost frontier capturing input technical efficiency, which measur es the potential each firm has to reduce costs (holding output constant), is depicted in Equation 4-3. ),()/1(])(|))(/min[()/,(' iiii iii iiii iiiiPYCuYXufXuuP uPYC (4-3) Where the time subscript and other variables al ready identified in the model have been omitted for simplicity, Yi is observed output of firm i f is a production function, common to all firms, Pi = (p1i p2i pji ) is a vector of input prices, and Xi = (x1i x2i xji ) is a vector of inputs. The parameter ui, 0 ui 1, measures the extent to which minimal input usage differs from actual input usage and provides a measure of technical inefficiency. The last equality in Equation 4-3 follows from the fact that a cost f unction is linearly homogeneous in input prices.

PAGE 99

99 Another way to express Equation 4-3 is to think of the differe nce between minimum and actual costs as the amount needed to reach the frontie r, so each firm must lower cost by the amount specified in Equation 4-4. ])/[(iiiCuC (4-4) Applying natural logs to Equa tion 4-3 yields Equation 4-5. ),(lnln),(ln)/1ln()/,(lniii i iii i iiiiPYCuPYCuuPYC (4-5) If the variable C represents total costs as defined for this model, the second term of Equation 4-5 is specified by Equatio ns 4-1 and 4-2. In a panel data frontier analysis, inefficiency is usually treated as the panel effect In this framework, the parameter u encompasses those unobserved factors out of a firms control which represent the firm s specific level of technical inefficiency. As such, this term is assumed not to be correlated with the explanatory variables. Given the length of the period in considera tion and the general ch aracteristics of the country and sector described so far, it is possibl e to infer that the firms specific inefficiency behavior has changed over time. Moreover, this ch ange may be the same fraction for all firms in this sector. To assess this possibility, time is included in the econometric model following Battese and Coellis (1993) functional form sp ecification for the inefficiency component, (uit ) provided in Chapter 1. For purposes of estimation, the us ual idiosyncratic error term (v) is added to the econometric model. This error term is indepe ndently and identically distributed following a normal distribution with zero mean, and it is independent from explanatory variables. Estimation proceeds using a balanced data se t of 41 companies comprising 328 observations. Limas water utility, Sedapal, was removed from the data set given its high values for the variables utilized in this model. It represents an outlier and is likely to distort the results.

PAGE 100

100 Results from Estimation For the frontier estimator, the likelihood func tion is expressed in terms of the variance parameters, 2 = V 2 + u 2 and gamma, = u 2 / 2. The closer the value of gamma is to one, the more inefficiency as opposed to noise explai ns the variance of the disturbance term. The variances for the normal noise and inefficiency terms are 0.02 and 0.026 respectively. The value obtained for gamma indicates that unobserved inefficiency represen ts 60% of the residual term whereas noise comprises 40%. Figure 4-3 displays the prediction of the inefficient parameter for 2003 showing full dispersion. Utilities are repr esented on the X axis sorted by their number of connections which indicates their size and the inefficiency values are on the Y axis. Possible values for range from zero to one. The closer is th e value to zero, the higher the efficiency. Table 4-4 presents results from the estimation. Time is not statistically significant indicating that tec hnological changes intended to be capture d by this variable have had no impact in this sector costs over the period an alyzed. However, the coefficient for the parameter explaining the inefficiency behavior for these ut ilities over time, is positive and statistically significant at 1% level, indicating that the exponential specification imposed is appropriate. It indicates that utilities effi ciency has change over time follo wing an exponential distribution. Figure 4-4 illustrates the pred icted technical efficiency for 1996 and 2003. The horizontal line y=1 represents the efficient frontier. Predic ted technical efficiency after estimation comes from calculating the technical efficiency term via E{exp(uit |eit)}. The closer the value to one the more efficient is the firm. The mo del fit is illustrated in Figure 4-5. Actual costs for 2003 are in natural logarithm form on the Y axis and are re presented by a black line. Predicted costs are represented by a gray dashed line and utilities are sorted by size on the X axis. The small gap between actual and predicted cost s indicates that the selected va riables and functional form for the economic model closely represent total costs for this sector.

PAGE 101

101 Table 4-5 shows utilities inefficiency statis tics in 2003 by region. These values represent ( u) the extent to which minimal input usage differs from actual input usage. Predicted u after estimation comes from the estimate of minus the natural log of the technical efficiency via E( uit | eit) where eit is the residual term. In other words, thes e values represent the factor by which input usage needs to be reduced if firms are to behave efficiently. On average, water utilities in this sector could reduce input usage by 27%. The differences in inefficiency among regions are not large. Turning to the interpretation of the coefficients for the explanatory variables, those utilities providing service in departments ty pe A have lower costs than thos e serving in departments type B. This is explained by the negative and st atistically significant coefficient for the MORETHANONE dummy which represents a 2.5% d ecrease in costs. This value explains the cost difference between firms located in departme nts where there are additional service providers with respect to utilities serving alone within a department. This cost reduction is paralleled in this model by an increased efficiency for th ese utilities which is shown by smaller u values Indeed, utilities in departments type B show a mean va lue for inefficiency equal to 31% whereas those serving with other providers have a mean inefficiency value of 26%. However, for utilities providi ng service along with other pr oviders, a one percentage increase in number of provinces served incr eases costs by 0.21%. This is reflected by the positive and statistically significant coefficient fo r the number of provinces variable (NPROVJ). Given that the maximum number of provinces served by a utility is eight, the impact on costs for a utility increasing service to one additional province is 2.5%. This result offsets the cost decrease obtained from the presence of competition.

PAGE 102

102 The coefficient for the increased coverage variab le is not statistically significant indicating that this variable has no explan atory power on total costs. Coe fficient magnitudes and signs for network density, continuity and the set of outputs and input prices are very similar to those obtained on the estimation of the cost model in Chap ter 3. Overall, utilities with denser networks have lower total costs. Continuity of service is not statistically significant indicating that this variable has no explanatory power on total costs for this group of utilities. A sensitivity analysis is performed including a few variables to test variation of results. In Table 4-4, model (a) includes the number of districts served by each utility. The obtained coefficients are not statistically significant and ve ry small. In model (b), coverage increase and the number of provinces served are interacted with the regional dummies. In this case, the magnitude of coefficients for the variab les MORETHANONE, NPROVJ and CHCOVERAGE are slightly larger. The larger magnitude among coe fficients of NPROVJ is the one related to the mountains. This indicates that when expanding se rvice to nearby provinces utilities located in departments type A that are in the mountains have higher costs th an firms in other departments and located in other regions. The next two model variations address the possi ble effect on total cost s of expansions or contractions of service to near by districts. In model (C), the coefficient for NEGCHIDX is not statistically significant. However when consider ing the effect over time, utilities reducing the number of districts have 0.2% higher costs than that of utilities not changing the number of districts or increasing the number of districts. The variation of other coefficients magnitude is very small. In model (d) the coefficient for the dummy POSCHIDX is statistically significant and has a negative impact on costs. This means that ut ilities expanding service to nearby districts have

PAGE 103

103 total costs lower than those not doing so. When considering the effect of adding districts to service over time, these utilities have 0.4% higher costs. This is shown by a positive and statistically significant coefficient for the variable POSCHIDXT. Finally, it is important to investigate what dr ives inefficiency in this sector directly, utilizing another technique. The di fficulty in doing this is that inefficiency is not observed. A central assumption when defining the econometric model was that the unobserved inefficiency term was not correlated with the set of explanatory vari ables. If it is, estimated coefficients are biased. Inefficiency can be calculated by util izing a DEA technique, as it was explained in Chapter 1, where the same set of inputs and outputs considered in the cost model are included. Once these efficiency values are obtained, they are regressed using an Ordinary Least Square estimator. The idea is to invest igate whether this inefficiency value can be explained by the explanatory variables used as the shifters of the cost frontier. If these variables are drivers of inefficiency, they may as well be correlated with the inefficiency term from the cost model, which would produce biased coefficient estimates. This robustness test proceeds by calculat ing a DEA frontier for each year, imposing variable returns to scale in the calculation to allo w efficiency to vary with the size of the utilities and considering the inputs and output s included in the cost model. The idea is to include as many characteristics of the production process as possible. One aspect that is missed in this DEA calculation is the translog functional form and the fact that input quantities instead of prices are utilized. Thus, these calculated efficiency values represent a production frontier rather than a cost frontier. These new inefficiency values are regressed on the set of variables, shifting the stochastic cost frontier using an Ordinary Least Squares es timator. In addition, the dummies for each region

PAGE 104

104 are included in this model. Results from estimati on are presented in Table 4-6. None of the coefficients are statistically si gnificant indicating that these variables do not explain technical efficiency. Concluding Observations Results from this study help explain the perfor mance of water utilitie s with respect to the provision of service to multiple jurisdictions. It examines the impacts of policies promoting the decentralization of water provisi on to municipalities compared to providing service to more aggregated jurisdictions such as departments. The issue turns on the governance structure, which affects the potential scope for political interference and conflict Decentralizing water provision to expand se rvice to less populated areas has increased local coverage in less populated departments. The expansion of service implies lower costs if utilities do so locally by increasing provision to di stricts within the same province. However, the expansion of service to nearby districts has occurred within th e most populated departments. This pattern can be interpreted as a direct c onsequence of the ownership structure of these utilities. Having data on populati on served by district for each utility could provide additional insights on the issues analyzed in this study. Additional research is warranted. The analysis also shows that changing the number of districts served implies some costs in the long run. This finding may be a consequen ce of the lack of adjustment on the amount of resources utilized in production as the size of the network changes. Alte rnatively, this result might be interpreted as the cost of political interference on strategic decisions of these utilities. The changes in the number of districts served o ccur only for utilities in the highest populated departments and the increase in c overage for these departments is small relative to increases in coverage in other departments. This observation suggests that changes in the number of districts

PAGE 105

105 served may be a consequence of a quest for pow er, since expanding servi ce to highly populated districts increases the utilitys ownership shares for those districts. Results from the analysis also show that u tilities located in departments where there are other service providers have 2.5% lower total costs than those providing service in departments where there is no other service supplier. Firms providing service alone show lower levels of coverage, which could be due to the absence of competition. From a policy making perspective, this result points towards developing an incentiv e scheme to introduce additional suppliers at the department level. Additionally, utilities providing service in departments where there are other suppliers have 2.7% higher costs for each additiona l province they serve. This result offsets the lower costs due to competitive pressure presen t when there are additi onal suppliers in one department. This finding supports the hypothesis that having more as opposed to less political authorities to respond to introduces inefficien cies on the utilities production practices. As a conclusion, within a department, an effici ent way of expanding se rvice is by introducing additional service providers such that they serve only one province.

PAGE 106

106 Table 4-1. Number of provin ces, districts and utilities providing service by department Region Department Population census 2005 19932005 Pop Growth Num of Province Num of District Serv. Prov Serv. Dist. Num.of Utilities Mountain Ancash 1,039,415 6% 20 146 7 7 2 Apurimac 418,882 6% 7 73 1 1 1 Ayacucho 619,338 21% 11 100 2 2 1 Cajamarca 1,359,023 5% 13 114 5 6 2 Cusco 1,171,503 10% 13 95 6 7 3 Huancavelica 447,054 12% 7 87 2 4 1 Hunuco 730,871 8% 11 65 2 3 1 Junn 1,091,619 0% 9 114 7 10 3 Puno 1,245,508 13% 13 96 7 8 4 Forest Amazonas 389,700 10% 7 77 3 6 3 Loreto 884,144 20% 7 44 3 3 1 Madre de Dios 92,024 32% 3 8 1 1 1 San Martin 669,973 17% 10 67 9 10 2 Ucayali 402,445 21% 4 11 1 1 1 Coast Arequipa 1,140,810 21% 8 101 8 20 1 Ica 665,592 15% 5 36 5 15 4 La Libertad 1,539,774 20% 12 71 5 15 1 Lambayeque 1,091,535 15% 3 35 3 26 1 Lima + 7,819,436 20% 10 161 4 18 4 Moquegua 159,306 22% 3 17 2 2 2 Piura 1,630,772 16% 8 56 5 15 1 Tacna 274,496 23% 4 23 2 3 1 Tumbes 191,713 21% 3 10 3 15 1 + Lima city population 6,954,517

PAGE 107

107 Figure 4-1. Schematic structure of jurisdictions in Peru and examples of service provision Table 4-2. Service coverage increase fr om 1996 to 2003 and population growth by region Service Coverage Pop. Growth Region Mean Median St.Dev Min Max Mountains 27% 31% 14% 0% 54% 9% Forest 22% 21% 19% 2% 65% 17% Coast 13% 10% 15% -7% 56% 18% All 20% 17% 17% -7% 65% 15% Department A Department BProvince Province Province Province Utility1 Utility2 Utility3 Province Province Province Province

PAGE 108

108 Figure 4-2. Departments of Peru showing those with variatio n in the number of districts Table 4-3. Mean values for popul ation and coverage increases Subset of Utilities Number Utilities Total %Change in Pop. %Change in Coverag Not changing number of districts served 24 14% 19% Adding number of districts to service 11 16% 18% Reducing number of districts 6 11% 28% Total with a change in districts served 17 14% 22% Total 41 14% 20% Reducing number of districts + those not changing 6+24 30 14% 21% Adding districts to service + those not changing 11+24 35 15% 19% 0 0.2 0.4 0.6 Figure 4-3. Values of pred icted inefficiency for 2003 # Districts Increased # Districts Decreased LIMA Utilities sorted by size Inefficiency Mean

PAGE 109

109 Table 4-4. Estimation results for variables comprising the intercept of the translog model Model Main (a) (b) (c) (d) T -0.001 (0.012) 0.05 (0.05) -0.004 (0.009) -0.009 (0.011) -0.011 (0.008) MORETHANONE -0.25** (0.102) -0.15* (0.091) -0.28*** (0.099) -0.28*** (0.106) -0.22*** (0.084) NPROVINCES 0.21** (0.086) 0.12 (0.08) 0.12 ( 0.09) 0.21** (0.09) 0.196** (0.084) NPROVINCES x R1 0.21 (0.134) NPROVINCES x R3 0.17 (0.154) COVERAGE -0.087 (0.274) 0.32 (0.28) -0.82 (0.54) -0.03 (0.286) -0.229 (0.251) COVERAGE x R1 0.55 (0.744) COVERAGE x R3 0.99 (0.728) NDISTRICTS 0.06 (0.085) NDISTRICTS x R1 -0.06 (0.085) NDISTRICTS x R3 0.102 (0.101) NEGCHIDX -0.12 (0.139) NEGCHIDX x T 0.027** (0.012) POSCHIDX -0.28* (0.157) POSCHIDX x T 0.045*** (0.015) NETDENSITY -0.32*** (0.108) -0.42*** (0.112) -0.279** (0.128) -0.315*** (0.111) -0.29** (0.106) NETDENSITY x R1 -0.015 (0.136) -0.051 (0.136) -0.03 (0.155) 0.011 (0.139) -0.085 (0.130) NETDENSITY x R3 -0.189 (0.129) -0.037 (0.136) -0.23 (0.159) -0.186 (0.129) -0.201 (0.129) CONTINUITY 0.059 (0.118) 0.161 (0.113) 0.047 (0.116) 0.002 (0.125) 0.129 (0.107) CONTINUITY x R1 -0.22 (0.145) -0.32* (0.140) -0.248 (0.148) -0.15 (0.152) -0.29* (0.132) CONTINUITY x R3 -0.07 (0.136) -0.19 (0.128) -0.05 (0.137) 0.017 (0.148) -0.20 (0.130)

PAGE 110

110 Table 4-4. Continued Model Main (a) (b) (c) (d) Number of Variables 40 43 44 42 42 LogLikelihood 114.82414 120.22713 117.5489 117.61988 119.51025 Mu (u ) + 0.257* (0.102) 0.141 0.129 0.261** (0.103) 0.119 (0.142) Eta++ 0.086** (0.102) 0.088*** (0.023) 0.074*** (0.024) 0.095*** (0.024) Gamma +++ 0.567 0.63 0.698 0.62 0.668 Dependent variable: Ln (TotalCo st); Number of observations 328 Confidence levels: *** 99%; ** 95%; 90%; Standard Errors in parenthesis. + Predicted Mu (u) after estimation comes from the estimate of minus the natural log of the technical efficiency via E( uit | eit) where eit is the residual term. ++Eta is defined as +++ Gamma is defined as = u 2 / 2 where 2 is the sum of the variance for the term parameters noise and inefficiency ( 2 = V 2 + u 2) 0 0.5 1 1.5 2 2.5 3 11 12 13 14 1 Figure 4-4. Movement towards e fficient frontier from 1996 to 2003 Technical Efficiency Utilities sorted by size 1996 2003

PAGE 111

111 -6 -5 -4 -3 -2 -1 0 1 2 Figure 4-5. Actual total cost s (black line) and prediction (dashed gray line) for 2003 Table 4-5. Inefficiency st atistics by region for 2003 Mean SD p50 min max Mountains 28% 14% 28% 5% 46% Forest 22% 13% 19% 5% 42% Coast 28% 14% 26% 3% 57% All 27% 14% 27% 3% 57% (Inefficiency is estimated as minus the natural log of the TE via E (uit | eit)) Utilities sorted b y size Ln (costs)

PAGE 112

112 Table 4-6. Results from efficiency regressi on-dependent variable average efficiency Model Coefficient R1 -0.017 (0.077) R3 0.034 (0.086) MORETHANONE -0.036 (0.076) NPROV 0.015 (0.015) COVERAGE 0.105 (0.293) NEGCHIDX 0.019 (0.065) PIDXJ 0.020 (0.071) NETDENSITY 0.0003 (0.0007) CONTINUITY 0.002 (0.005) R-squared 0.123 Standard Errors are in parenthesis.

PAGE 113

113 APPENDIX A DESCRIPTION OF VARIABLES F O R DATA COLLECTION General Characteristics for Each S ervice Provider in Central America 1. Service provider name and year beginning operations 2. Public (government) or private operation 3. Scope of the company according to country jurisd ictional organization: stat e, municipal, district 4. Type of service: water, sewerage or both. 5. Source of water: surface or underground 6. Localities: refers to the number of locali ties where the company provides service according to the country jurisdictional organization (districts, municipal or status) 7. Region: refers to the region the company is lo cated in case the country has specific regions regarding some type of geographi c or topological characteristics, eg mountains or coast. In case the company serves different locations indicate region for each location. Outputs 1. Volume of water (cubic mete rs or litters annual): a) Produced: total annual gross volume extracted from its origin point, whether surface or underground. b) Delivered: total annual volume enteri ng into the distribution network. c) Billed: total annual volume that is billed to customers. d) Collected: total annual volume for thos e bills that have been collected. e) Treated: volume of water treated alter used by customers. f) Total volume of water lost g) Commercial lost: difference between volume delivered to the distribution network and volume billed. h) Operational lost: difference between volum e produced and volume delivered to the distribution network. 2. Number of water connections and sewerage: a) Total b) Residential, Commerci al and Industrial c) With meter

PAGE 114

114 d) Active e) Number of consumers water and sewerage: f) Total population (usually from country census) g) Population served with water and sewerage h) Number of inhabitants per connection (o r number of inhabitants per household) 3. Network length water and sewerage (km): to tal network length including the distribution network. Inputs 1. Number of workers and its costs (or expense): a) Total b) By contract(limited time or money amount) c) Fixed (they receive company benefits) d) Types of workers administrati ve, operational, managerial 2. Volume of energy and its co st (Kva or another unit). 3. Capital stock: a) Non-current assets b) Accumulated depreciation c) Annual depreciation 4. Administrative Expenses 5. Financial Expenses 6. Operating costs 7. Total Cost Quality 1. Water quality: any variable defining water qua lity (e.g. percentage of residual chlorine) 2. Continuity: number of hours a da y customers receive water service 3. Quality of service: a) Number of complaints received or served b) Number of failures or problems received or fixed 4. Number of network leaks

PAGE 115

115 APPENDIX B SUMMARY OF PERFORMANCE ANALISYS METHODOLOGIES Methodology Outputs Inputs Technology functional for m Period Purpose Performance Indicators Water lost, metering, coverage, network density, water consumption Number of workers/ 1000connect Ratio of single variable 2002 2005 Productivity change of one factor for a particular firm TFPLaspeyres, Paasche, Fisher 1) VolWaterBil 2)#Connections Labor + Energy Ratio of linear combination 2002 2005 Productivity change of a few factors for a particular firm TFPTornqvist 1) VolWaterBil 2)#Connections Labor + Energy Ratio of log form combination 2002 2005 Productivity change of a few factors for a particular firm DEA frontier 1) VolWaterBil + number of connections 2) VolWaterBil + net.density 1)GNI + Labor + netleng 2) GNI + Labor Ratio of linear combination 2005 Tech efficiency of a firm considering multiple inputs and outputs, with respect to best practice within a group Malmquist catch-up effect 1) VolWaterBil + number of connect 2) VolWaterBil + net.density 1)GNI + Labor + netlength 2) GNI + Labor Ratio of Distances 2002 2005 Movement of a firm towards tech efficient frontier set by a group Stochastic Cost frontier VolWaterBil GNI Labor Energy Linear log combination 2002 2005 Central tendency of cost efficiency for a group of firms

PAGE 116

116 LIST OF REFERENCES Aigner, D., & Chu, S. (1968). On esti m a ting the industry production function. The American Economic Review 58, 826. Aigner, D., Lovell, C., & Schmidt, P. (1977). Formulation and estimation of stochastic frontier production function models. Journal of Econometrics 6, 21. Antonioli B. & Filippini M. (2001) The use of variable cost function in the regulation of the Italian water industry. Utilities Policy 10, 181. Ashton, John (1999). Economies of scale, economies of capital util ization and capital utilization in the English and Welsh water industry. B ournemouth University, School of Accounting and Finance, Working Papers Series No17. Atkinson, S. & Cornwell, C. (1994). Estimation of output and input techni cal efficiency using a flexible functional fo rm and panel data. International Economic Review 35, 245. Atkinson, S., Cornwell, C. & Honerkamp, O. (1999). Measuring and decomposing productivity change: stochastic distance function estimation vs DEA. Terry College of Business, Dept of Economics, University of Geor gia, Working Paper Series 99. Atkinson, S. & Halabi, C. (2005). Economic e fficiency and productivity growth in the postprivatization Chilean hydroelectric industry. Journal of Productivity Analysis 2, 245 273. Aubert, C. & Reynaud, A. (2005). The impact of regulation on cost efficiency, an empirical analysis of Wisconsin Water Utilities. Journal of Productivity Analysis, 23, 383. Baumol, W. (1976). Scale Economies, average cost and the profitability of marginal cost pricing. Essays in Urban Economics and Public Finance R.E. Grieson, Lexington, Mass. 43. Balk, B. (2003). The Residual: On monitori ng and benchmarking firms, industries and economies with respect to productivity. Journal of Productivity Analysis 20, 5. Banker, R., Cooper, W., Seiford, L. Thrall, R. & Zhu, J. (2004). Returns to scale in different DEA models. European Journal of Operational Research 154, 354. Battese, G. & Coelli, T. (1992). Frontier produc tion functions, technical efficiency and panel data: with application to paddy farmers in India. Journal of Productivity Analysis 3, 153 169. Battese, G. & Coelli, T. (1993). A stochastic frontier production function incorporating a model for technical efficiency effects. Working Papers in Econometrics and Applied Statistics, No 69, Department of Econometrics, Un iversity of New England Armidale.

PAGE 117

117 Battese, G. & Coelli T. (1995). A model for technical inefficiency effects in a stochastic frontier production functions for panel data. Empirical Economics 20,325. Clarke, G., Kosec, K., & Wallsten, S. (2004). Ha s private participation in water and sewerage improved coverage? Empirical evidence from Latin America. World Bank Policy Research Working Paper No. 3445. Coelli, T, Perelman, S., & Romano, E. (1999). Accounting for environmental influences in stochastic frontier models: with appl ication to international airlines. Journal of Productivity Analysis 11,251. Coelli, T., Estache, A., Perelman, S., & Trujillo L. (2003). A primer on efficiency measurement for utilities and transport regulators. World Bank Development Studies, 26062, the World Bank, Washington D.C. Cornwell, C., P. Schmidt and R. Sickles (1990). Production frontiers with cross-sectional and time-series variation in efficiency levels. Journal of Econometrics 46, 185. Corton, M., & Molinari, A. (2008). ADERASAs role in regulatory collaboration in the Americas. Water 21, February 2008, 23. Corton, M. (2003). Benchmarking in the Latin American water sector: the case of Peru. Utilities Policy, 11,133. Charnes, A., Cooper, W., & Rhodes, E. (1978). Measuring the efficiency of decision making units. European Journal of Operations Research 2, 429. Christensen L. and W. Greene (1976). Economies of scale in U.S. elect ric power generation. The Journal of Political Economy, 84(4), 655. Diewert, W.E. (1974). Applications of duality th eory, in Frontiers of quantitative economics, edited by M.D. Intriligator and D.A. Kendrick. Amsterdam: North-Holland. Diewert, W.E., & Wales, T.J. (1987). Flexible functional forms and glob al curvature conditions. Econometrica 55(1), 43. Estache, A., Perelman, S., & Trujillo, L. (2005) Infrastructure performance and reform in developing and transition economies: Eviden ce from a survey of productivity measures. World Bank Policy Research Working Paper 3514. Estache, A., Rossi, M. & Russier. C. (2004). The case for internati onal coordination of electricity regulation: Evidence from the measuremen t of efficiency in South America. Journal of Regulatory Economics 25(3), 271. Estache, A., & Rossi, M. (2008). Regulatory ag encies: Impact on firm performance and social welfare. The World Bank, Policy Research Working Paper 4509.

PAGE 118

118 Fabbri, P., & Fraquelli, G. (2000). Costs and struct ure of technology in the Italian water industry. Empirica 27, 65. Fare, R., Grosskopf, S., Norris, M. & Zhang Z. (1994). Productivity growth, technical progress, and efficiency change in industrialized countries. The American Economic Review 84, 66. Farrell, M. (1957). The measurement of productive efficiency. Journal of the Royal Statistical Society Series A, 120(3), 253. Farsi M., Filippini, M., & Greene, W. (2005). Efficiency measur ement in network industries: application to the Swiss railway companies. Journal of Regulatory Economics 28(1), 69 90. Garcia, S. &Thomas, A. (2001). The structure of municipal water s upply costs: application to a panel of French local communities. Journal of Productivity Analysis 16, 5. Greene, W. (2005a). Reconsidering heterogeneity in panel data estimators of the stochastic frontier model. Journal of Econometrics 126, 269. Greene, W. (2005b). Fixed and random ef fects in stochastic frontier models. Journal of Productivity Analysis 23, 7. Greene, W. (2004). Distinguishing between heterogeneity and ine fficiency: stochastic frontier analysis of the World Health Organizations panel data on national health care systems. Econometrics and Health Economics 13, 959. Hanoch, G. (1975). The elasticity of scale and the shape of average costs. American Economic Review 65(3), 492. Hattori, T. (2002). Relative performance of US and Japanese electric ity distribution: an application of stochast ic frontier analysis. Journal of Productivity Analysis 18, 269. Hattori, T., Jamasb T., & Pollitt, M. (2005). Electricity distribution in the UK and Japan: A comparative efficiency analysis 1985. The Energy Journal 26(2), 23. Jam asb, T., & Pollitt, M. (2003). International benchmarking and re gulation, an application to European electricity distribution utilities. Energy Policy 31(15), 1609. Jensen, U. (2005). Misspecification preferred: the sensitivit y of inefficiency rankings. Journal of Productivity Analysis 23, 223. Kalirajan, K ., & Shand, R. (1999). Frontier production functions and technical efficiency m easures. Journal of Economic Surveys 13(2), 149.

PAGE 119

119 Kingdom, B., Liemberger, R. & Marin, P. (2006) The challenge of reducing non-revenue water (NRW) in developing countries, how the private sector can help: a look at performance based serving contracting. World Bank Water Supply and Sanitation Sector Board Discussion Paper Series, 8, December 2006. Lee, Y. & Schmidt, P. (1993). A Production frontie r model with flexible temporal variation in technical inefficiency. In The Measuremen t of Productive Efficiency: Techniques and Applications, Harold F., Lovell, C., Schmidt, S. (Eds.), Oxford: Oxford University Press: 237. Lovell, Knox C. (1995). Econometric efficien cy analysis: a policy-oriented review. European Journal of Operational Research 80,452. Lowry, M., Getachew, L., & Hovde, D. (2005). Ec onometric benchmarking of cost performance: The case of U.S. power distributors. The Energy Journal 26(3), 75. Mathur, S. (2007). Indian IT and ICT Industry: A performance Analysis using Data Envelopment Analysis and Malmquist Index. Global Economy Journal, 7, 1. Meeusen, W., & Van den Broeck, J. (1977). Efficiency estimation from Cobb-Douglas production functions with composed error. International Economic Review 18, 435. Murillo-Zamorano, L. (2004). Economic efficiency and frontier techniques. Journal of Economic Surveys, 18(1), 33. Neuberg, Leland (1977). Two issues in the munici pal ownership of electric power distribution. The Bell Journal of Economics 8(1), 303. Orea, L., Roibas, D., & Wall, A. (2004). Choosing th e technical efficiency orientation to analyze firms technology: a model selection test approach. Journal of Productivity Analysis 22, 51. Panzar, J.C., & Willig, R. (1977). Two Issues in the municipal ownership of electric power distribution systems. The Bell Journal of Economics 8, 303. Pearce-Oroz, G. (2006). The viability of decen tralized water and sanitation provision in developing countries: the case of Honduras. Water Policy 8, 31. Pitt, M., & Lee, L. (1981). The measurement sour ces of technical inefficiency in the Indonesian weaving industry. Journal of Development Economics 9, 43. Renzetti, S. (1999). Municipal Water Supply and Sewage Treatment Costs, Prices and Distortions. Canadian Journal of Economics 32(3), 688. Rossi, Martin (2001). Technical change and efficien cy measures: the post pr ivatization in the gas distribution sector in Argentina. Energy Economics, 23,295. Sabbioni, G. (2007). Efficiency in the Brazilian sanitation sector. Utilities Policy 16, 11.

PAGE 120

120 Saal, D., & Parker, D. (2000). The impact of Privatization and regul ation on the Water and Sewerage Industry in England and Wales: A translog Cost Function Model. Managerial and Decision Economics, 21(6), 253. Saal, D. & Parker, D. (2004). Productivity and price performance in the privatized water and sewerage companies of England and Wales. Journal of Regulatory Economics 20(1), 61 90. Schmidt, P., & Stickles, R. (1984) Production frontiers and panel data. Journal of Business and Economic Statistics 4, 367. Solo, T.M. (1998). Competition in water and sanita tion the role of sma ll scale entrepreneurs. World Bank Viewpoint, Dec.: Note 165. SUNASS (2006). Infraestructura de agua potable y alcantarillado urbano en el Peru: un reto pendiente. Copyright SUNASS, Mayo 2008. http://www.sunass.gob.pe/svipu.jsp (l ast visit: 05/25/2008). Tamayo, G., Barrentes, R., Conterno, E., & Bust amante, E. (1999). Reform efforts and low-level equilibrium in the Peruvian water sector In : Spilled Water: Institutional Commitment in the Provision of water Services. Savedoff, W., P. Spiller (Eds.), Inter-American Development Bank, Washington D.C., pp. 89. UNCTAD United Nations Conference on Trade a nd Development. The digital divide report: ICT diffusion index 2005. UNCTAD/ITE/IPC/ 2006/5 ; United Nations Publication; Copyright United Nations, 2006. http://www.unctad.org/en /docs/iteipc20065_en.pdf (last visit: 05/24/2008). Zhang, Y., Parker, D., & Kirkpatrick, C. ( 2008). Electricity sector reform in developing countries: an econom etric assessment of the effects of privatization, competition and regulation. Journal of Regulatory Economics 33, 159.

PAGE 121

121 BIOGRAPHICAL SKETCH Maria Luisa has a degree on com pute r scien ce engineering. Her career developed as a computer engineer for a long period of time wh ile holding a variety of positions in different public and private organizations. Maria Luisa also has experience in academia, teaching computer science courses. In 1994, she was accepted at Jacksonville University to pursue a MBA program in international business. She wa s holding a Fulbright/Mariscal de Ayacucho scholarship from her country, Vene zuela. In 1996, she finished the MBA with an award for high achievement. In 1997, she applied to the Universi ty of Florida for a PhD degree in economics. She was accepted and started the program, but she decided to end it as a ma sters degree by year 2000. In 2004, she resumed the PhD program and graduated in August 2008.


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E20101202_AAAABF INGEST_TIME 2010-12-02T12:47:23Z PACKAGE UFE0022423_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 51398 DFID F20101202_AAAUHH ORIGIN DEPOSITOR PATH corton_m_Page_030.pro GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
728e427f2b307dc3dc1309659cab0d61
SHA-1
9dfea85c6f1ed3da86d0ff549c33bdd5a8450955
25271604 F20101202_AAAUGT corton_m_Page_113.tif
0ffb1e4d4042772f6a17e2dcd584ecaa
c8dc1ad058c038cca6aee5a8aa63c246179b0feb
51942 F20101202_AAAUHI corton_m_Page_032.pro
8002e76a47eac36cf60e94640d6fe9f2
ab1df6066e242a67cc87e4a066e1687732d0bc2e
F20101202_AAAUGU corton_m_Page_114.tif
17117cfde7781df1271ecb08b5d78ddf
a2b3a69e67b5a1080f5b37471061cfca2be13f36
54240 F20101202_AAAUHJ corton_m_Page_038.pro
b43873841586124da83950ff707bcfe9
b2b1733c910400ea5d13a237f84c8223d904c3e7
F20101202_AAAUGV corton_m_Page_117.tif
496509896914834290beb5c0afbb3dce
eda7277856a938374b4b8bb39e9482b3b7f143a5
51753 F20101202_AAAUHK corton_m_Page_039.pro
8cff1c3f8dc10634daef38f4d5a4c5a2
5fa9213b7d22e3fbd2ace27522d67bda91918559
F20101202_AAAUGW corton_m_Page_119.tif
8b55f8fb4e469d4ecd1bb4680d464526
d73c462c4d83810a3ef71c43d52761762b50123b
55048 F20101202_AAAUHL corton_m_Page_040.pro
07b8f7ece5000add7c9db6ab3b7fb287
44bdc1a6393770cce645f39fbeec1c96154ac1f1
F20101202_AAAUGX corton_m_Page_120.tif
722591db3938c28d8ef7b003b140886b
1a9a239e711e7786fcb6061dde9be48f97f0cf5d
53670 F20101202_AAAUHM corton_m_Page_041.pro
b8dfc569e3d6c05a3771d998ed4f295d
aa3b483c29cd68819a564964b193439c0e6c5ab7
728 F20101202_AAAUGY corton_m_Page_003.pro
a3b6206ba4f56bbb9bf3c3ebf1959a62
163f0e92c345d97d03734b4ae52fe6b3e371192f
47495 F20101202_AAAUIA corton_m_Page_070.pro
0af68e640ae998ba7843459f89ae1986
dde9d723348e537d810b5870e6e24af36f2c3c66
54341 F20101202_AAAUHN corton_m_Page_042.pro
dfdf8e8e4b9461db2de3c320d2c94aea
83169aca7a005756060eafbff06e982d2359c733
45356 F20101202_AAAUIB corton_m_Page_071.pro
06843539d24ec5288faa9eea993bce15
d0b74a2eacf1b4f960d38f82472e46c318e9de0a
53266 F20101202_AAAUHO corton_m_Page_043.pro
661021504893bcfd10e6b3830766dd9e
03ba8101c7cd254c23ecf68ce63e075637c899de
23603 F20101202_AAAUGZ corton_m_Page_004.pro
81e12995fa6c65035100f1e21c525443
4ffea7a955c4d16a2bfe7d2b984bc338e2e4f0a4
53098 F20101202_AAAUIC corton_m_Page_074.pro
80c4498c80794e77a0c1f74c85fcbf3e
fecb9eea6b2270a14c49fc849b576cf1edc0739b
54002 F20101202_AAAUHP corton_m_Page_045.pro
4736f7f94dac6c3d8af9c93f6ce743d7
2087e896e0d6b9a6462374752f6988c4d3efc9f7
52361 F20101202_AAAUID corton_m_Page_077.pro
6d04db89cf3061ae9d72358538d1cd04
af35b077df471770a7472c03f5956d6543324011
14120 F20101202_AAAUHQ corton_m_Page_050.pro
af0ac46853b1ec6c9718595bf3fb2079
ad24d5e1ed79310484ba1c9316f6eb082a8b1667
56493 F20101202_AAAUIE corton_m_Page_078.pro
8f7cc3005dd8972d6c683193fc316779
b39a64073097036b301aa7e9a109c3217cd0c21d
53297 F20101202_AAAUHR corton_m_Page_057.pro
decebf336ed379aa4b7c0cf07c863e01
2076c30558baf48d9b7a354413624672adbbbdb0
52425 F20101202_AAAUIF corton_m_Page_079.pro
66f8ad10c95e0a8809bd4a763911b918
1e3c139f7f5c2b471825693608bd296f8757c308
52668 F20101202_AAAUHS corton_m_Page_059.pro
ff86a83f0c11d242282a2eca5a30d7fc
2c0b4b298c923bc075cd299530e68ec4306f14c1
49837 F20101202_AAAUIG corton_m_Page_081.pro
85d0b146f9f143247aaf22935f56252c
4971df2f79ed9bc03dcc997b7afba96a69bdebf1
55156 F20101202_AAAUHT corton_m_Page_061.pro
efdfaacd74b7efd4a34b01c7d37c85c2
b4a21100918dd75f3bcdfca2e8f0772d4bab7315
38587 F20101202_AAAUIH corton_m_Page_082.pro
496a85b14fee07d6fb48c61563139e8f
7ed42bd64d2174be78e4ea15155886d42a288520
54074 F20101202_AAAUHU corton_m_Page_062.pro
5b5e431493f1c2f535efb534cd7d3ed5
4f6e91e82e13328fac2b73a14643ba0ce8bd97a5
53554 F20101202_AAAUII corton_m_Page_084.pro
e072f6e1371afbccc8ac97bf6f5d4fd3
f84c7ef0a05cdfbc7ad711715538682e88aeff8d
54593 F20101202_AAAUHV corton_m_Page_063.pro
c6d25488b9ae6671677c1873bc45e6b3
321cfffae4c91f46412ed852f607ebd0764dcda4
56059 F20101202_AAAUIJ corton_m_Page_090.pro
6abd58d529629e64cec7468a8179285f
0fadaaf8797f755796ed05458edd3b422867bce9
53260 F20101202_AAAUHW corton_m_Page_066.pro
3ebba67c44c899e9f3d336ed1e4a1ad4
2fbb8552e0339af9443851e5a742e6acd2a2188f
52029 F20101202_AAAUIK corton_m_Page_092.pro
fdfe1f885c3e4b0ad96df41d3a97f45e
91e9aab21ebd826642f9831d316e0e0e7fce35e3
49619 F20101202_AAAUHX corton_m_Page_067.pro
8bee4ad59d3c08d42048cd1ec9569786
db13cf6d964a6116add76becf0590eeed57ec097
50162 F20101202_AAAUIL corton_m_Page_093.pro
bd16ee6b9afd9f4c5c36a422c4fdfa79
4507a8bc164b4821c5c97c875c6719da4d7a40ca
52846 F20101202_AAAUHY corton_m_Page_068.pro
cc0e69ecf07d676dcc1c256ff93e72c8
f1ee6f0eea4862cda89efd301e1b3cfa0a4a64fc
977 F20101202_AAAUJA corton_m_Page_004.txt
05d56a78afed59f27c23c51d029c50e7
116afc5a6da05548f8d2e280359a44d61f354eb3
52515 F20101202_AAAUIM corton_m_Page_094.pro
46e24e5f387c39ae1212bf7392fbb72d
5b3890997f751ce17bb618499211b1f16d7764fa
47398 F20101202_AAAUHZ corton_m_Page_069.pro
23c06d7f0fc92cdfd1131c3ef823462b
84dfa758c24fac8c24640911ea9446cdbecf7933
3922 F20101202_AAAUJB corton_m_Page_005.txt
8610947dc019279e186bf7dc1cfa1313
1303c50e51324126829b1e2fb29092d9d52bf73d
58527 F20101202_AAAUIN corton_m_Page_100.pro
d7dc85da4cd30ea3316c7055314ebc6a
9e126b71b34a698096dcf5f9537e89e6794b115b
2078 F20101202_AAAUJC corton_m_Page_009.txt
734c2c9924236332432ac5bb9d2217ff
a9caecda1b4b713da16927318bc3698baeb701f4
53367 F20101202_AAAUIO corton_m_Page_102.pro
418d3d7b45df3843a46d7153f389c27b
bbdeb0922b10c76bcb5d277d23dfdb6647a45ad0
1703 F20101202_AAAUJD corton_m_Page_010.txt
ae14f87f726d03aae118409a2115f97f
c517e5d25dd786f06d0ae5655ce0466d75457bea
53957 F20101202_AAAUIP corton_m_Page_103.pro
83a4ec0ddfaf68d7f6deaa574d0d8521
00df473cec270f3c169c0c0f2a99f9e28e24fef0
2029 F20101202_AAAUJE corton_m_Page_013.txt
1ae6cf730d490b354515edc253c13c21
f9bc1b56e0c8e35059658ea82cfc0832e725a63a
53075 F20101202_AAAUIQ corton_m_Page_104.pro
655ec618c007dd163559471f75a514df
fce5186397b875a31c33ad92ceb50cc1f87a736e
2104 F20101202_AAAUJF corton_m_Page_015.txt
257a582c55d7a5bf84a43f7da321a33e
7b796180414dbf95733bd98f291d9a0218dc19a6
33554 F20101202_AAAUIR corton_m_Page_105.pro
206f2c0d8bf379285dc2cee62e0d80f3
4ead623cf62ed2f6aa06d222d26d1ef82b3826cf
2123 F20101202_AAAUJG corton_m_Page_016.txt
04272af92521c89c5e0a1e710caf438e
b6b157e02aec2010b62aadce4931a165aa1c66b8
27159 F20101202_AAAUIS corton_m_Page_106.pro
d075f234f87449b433228ccd041857cb
b5337e3afd25c9449b3db9f9c81ce4f2ebdab242
2065 F20101202_AAAUJH corton_m_Page_017.txt
388778dbbd4ddbad05d6a8a672ff439a
4d32b4232494b3363bc678ad9f897ef6b8572e2a
18668 F20101202_AAAUIT corton_m_Page_107.pro
99dc68400d71ae40a6691775e6c8cec5
d777bef120672337a317dad4126270c53c79fe91
1755 F20101202_AAAUJI corton_m_Page_018.txt
5fc373cbe1d4a8281e7f4196530d3621
a4a7769e928a51b002fd86b5a0a8d0aa2b91c57f
27501 F20101202_AAAUIU corton_m_Page_108.pro
4fe10b1a7b31eaf00e22d43ab1c94fe8
405fc3f4540bcb0bb131f9d40c5dde94b268dd0d
2173 F20101202_AAAUJJ corton_m_Page_023.txt
1d7d47e4e86d1466f145850a1008b88f
351de3ada082ef9f92b929f78fae2fec78bb6217
31527 F20101202_AAAUIV corton_m_Page_110.pro
b06389830f9744ae1e2e860dae90cc9a
197096013edf06c0732dc6585d376f5ca78ae687
352 F20101202_AAAUJK corton_m_Page_026.txt
671ceba112e2f1c0d90f07a6ef7d8b61
5ca05d4b6e9c98a8638031a7fb0a8229750a5689
13161 F20101202_AAAUIW corton_m_Page_111.pro
52c95fd5fcb842db90dab904101e4184
ed8b2f94946dc025a7752fa0983f4fd4d10f7f7f
2186 F20101202_AAAUJL corton_m_Page_028.txt
ed28b63b442752e478e9508af2fee238
740471789f665ec834b271755053186702051300
43762 F20101202_AAAUIX corton_m_Page_113.pro
a8393aba76afe9d2bc7776b4d159d5a2
0ae97d1e06abade822edecdb96d24db3a069f4f6
2249 F20101202_AAAUKA corton_m_Page_058.txt
e096aece54ee2ea9527d3df4727235a2
73c742daa874cd813e9683b8ac8c91f78dd85072
2025 F20101202_AAAUJM corton_m_Page_030.txt
f18e3807b2fd5cf9822922cd94aaa7c9
864405b7d78902f52c4e17afcd29ca9b00f981b0
64427 F20101202_AAAUIY corton_m_Page_116.pro
5fe88815c936aed81c0fe2ab7009969e
7c86526fe81e395ca39488d43cbcb7bf6b9f2214
2081 F20101202_AAAUKB corton_m_Page_059.txt
13afedbd3c6b16fc8f1351cd49dc7a5c
21388cc91ffe9d9da5ffcdcfff29c6ddfedea9fe
F20101202_AAAUJN corton_m_Page_031.txt
84d071aa63db5206d4acd9598939ea81
192f100e79cf9f67cad8481ebe134bd0f1582589
67200 F20101202_AAAUIZ corton_m_Page_119.pro
81034de4785db080e8a308fd375f3b44
8469672e74978b8c0410103718fc48847aca0109
2141 F20101202_AAAUKC corton_m_Page_060.txt
df7dbb82f2d5f4e45a0fd3f1b06cd21d
6c0fe09b7c89f5d88a25f15cafb83129e87ac1b4
2048 F20101202_AAAUJO corton_m_Page_032.txt
064bb60589e7fb72362585f079132842
4b3ff9af13e8f31b1cc7205bfc12efffe8f8a7fd
2124 F20101202_AAAUKD corton_m_Page_062.txt
18ca1448b0d643c2a21185218e524b4b
3fd06797633b34247b2a702111a71f8e3da7f21c
2094 F20101202_AAAUJP corton_m_Page_033.txt
f368eca916d0d664b15574960a418810
45cc829115ae6135a2180d66652ee96ced641639
2241 F20101202_AAAUKE corton_m_Page_064.txt
a6eba6d1b331a834821dcc2eb0e74d8b
a8ce1872541abc9ba9d32ecb3aca686dddda402a
2089 F20101202_AAAUJQ corton_m_Page_035.txt
e96ceaf4a76ee79746795f308ab6a492
f56316cfa4fdc67ee2a41e6ac9ec3c8d50ed8ff5
1940 F20101202_AAAUKF corton_m_Page_069.txt
5ee664b21414e0de2cd3fd3efcdeca59
563ca9283c0b3e276eb0622cc3d2792f41ddacd0
2205 F20101202_AAAUJR corton_m_Page_037.txt
42c6bbb1a32e3f72d30a70ac7a1d39ea
0182fdaa544acb1fb482562856ab14518db192ad
1928 F20101202_AAAUKG corton_m_Page_070.txt
747fd48316ba984fd0e641624544e941
f63c8c0f12a2714dfcf84cdd5276b478a2cd81bd
2157 F20101202_AAAUJS corton_m_Page_038.txt
f6fbe098adca125a88947dcc9e90a2ca
96312018ec5602d4523b878c6b1bca34450609cb
1899 F20101202_AAAUKH corton_m_Page_071.txt
e9f5e1fa807094fbdf627c5f2879fec2
3c7399758d2af6424e217b9741dcacdbddea29af
2120 F20101202_AAAUJT corton_m_Page_041.txt
da411a2a59826eda616ece31b0f8099a
45d9b4328999141e0f768a3b392141fae1100dff
2117 F20101202_AAAUKI corton_m_Page_072.txt
a17feeba55130ef720cf7d8df65c3586
5876be4ad946a3b063dd2529a6d58f88cde28f3d
2170 F20101202_AAAUJU corton_m_Page_042.txt
01362e3e6301d50808272ba0e34459d7
42a9388aa4a372188c80b2f0f04acab43e4aa8c4
2223 F20101202_AAAUKJ corton_m_Page_076.txt
0963c0e16787a9dcde1cc9bcbbe5815e
2c3c76d1b790cb70ffe95ed27f90c3ea774c55a8
2145 F20101202_AAAUJV corton_m_Page_043.txt
71a90ba8f8e7b89540d174c08dc90ec9
876bcaf6486f3154621df1f8d4d82e217d54133d
F20101202_AAAUKK corton_m_Page_077.txt
d15717189a77aab11968c5306f52ce48
9a24b15eaaff8bf54e172e3bb0d2edf538c4a0d5
2167 F20101202_AAAUJW corton_m_Page_051.txt
f2418239d994ef706ec226cd67cd5542
ab2f4e8822346e0fc549b8f6a2b88193568d4662
2217 F20101202_AAAUKL corton_m_Page_078.txt
cfd3e1f9f946f596856467dfa7e1e0f1
132decd12ae5c33fa31f665a4228393e06fe94b4
1497 F20101202_AAAUJX corton_m_Page_052.txt
0b40ba56bf899023bdc2d22afb35d1d0
1043a67d739c4eca2453c2d1556578bbd5b4c6eb
2100 F20101202_AAAUKM corton_m_Page_079.txt
8d4fef0d75ff46f686c8d32d1ecc159c
dcb271673293d398feb7acd672b128727d32386f
2356 F20101202_AAAUJY corton_m_Page_054.txt
59beda975c2918a2b93a6ae79be41ccb
839bc73abc58b200cb7a8df51fb37ad65019763d
2028 F20101202_AAAULA corton_m_Page_101.txt
22d818e99bfab6f42ad2381d4fe02da6
e235c34d4da41361507e8d5daf9a04b17817bcf9
2791 F20101202_AAAUKN corton_m_Page_084.txt
e84c178db0fcca3f20771d10e89a1123
67eeabf271235b2e99b2404abd7317bcb52ad989
934 F20101202_AAAUJZ corton_m_Page_055.txt
e39728c49ae690a05911d237b3e41952
02cbac01b93d8fd5720bb571981cb1cb89c63f49
2130 F20101202_AAAULB corton_m_Page_103.txt
ae9bd523f87970308334139e5ce03e7e
11035bef7154f0712bff55f1c0d88a96ffeb95a2
1355 F20101202_AAAUKO corton_m_Page_085.txt
7c66710bd70f11891d0132cc39032a5e
e89e461c97f83fdb1411ba1fb586e699c8752d39
F20101202_AAAULC corton_m_Page_104.txt
da1c003a84bb74ff08bae06a3da9b216
517b74345e37068c03f2727d8abe5d885f880c63
1511 F20101202_AAAUKP corton_m_Page_086.txt
e796075c3bc45ad8a1160e05b89cace7
bf1b55b9322fa2293ac2226f261115cd809b30fd
1334 F20101202_AAAULD corton_m_Page_105.txt
18fca84ad3bd2b63e2cd9c17c668e425
89b45321c9da25ed827a911a2cd56818083cc690
2180 F20101202_AAAUKQ corton_m_Page_088.txt
369835d81286f1ef5a9d7e42cf035e89
6e3039489c6ea8dc6c36b906638ef41109bcd63c
1182 F20101202_AAAULE corton_m_Page_106.txt
0b3222bfe577da59932bdfb63c57e728
b09f5adc8989e44ce7703feead7240200c361cf4
2035 F20101202_AAAUKR corton_m_Page_089.txt
3b54290aca5fd9890b76c4965bb4aca1
1aa9c72fdec91dbdeb4a9349fa3af5c663c7c4e2
1436 F20101202_AAAULF corton_m_Page_108.txt
05babc19f82ba71304e2a861f13064c2
5632b45a32a9c9bfc9bbe45394a12fe046c8434e
2212 F20101202_AAAUKS corton_m_Page_090.txt
bfee37598a93a65f1c8b6319eb7b0dc0
06c0fef76c3e57bff24cdd8b72a3f9092bffc922
3863 F20101202_AAAULG corton_m_Page_109.txt
7e2e829c58924719947fb66e24c65a41
af4764c57fb27562128751dca4020ad57b0c13f0
750 F20101202_AAAULH corton_m_Page_111.txt
d71ce69a5f556aac86de4654a92cdd00
705c8167e949fe589e424075fb1c3513977fe897
2046 F20101202_AAAUKT corton_m_Page_093.txt
62aafdf2c1edc6d2109463e48fdd9d1c
9c7de6bb4bae43d9d403d64977754eca4cf8176c
1895 F20101202_AAAULI corton_m_Page_113.txt
0632c014ffb1653a81467dae7bfee481
d85b03400b46daec105e9d9e0bed73e6a996b9f2
2075 F20101202_AAAUKU corton_m_Page_094.txt
7618dd08d38b816668816eebfddae6a9
e2414094be0484eafc81faabab0b0ae78d9a09f3
2595 F20101202_AAAULJ corton_m_Page_117.txt
357aed44495aea8fce03490c28be6f76
c41d783e6d84c0e4bbbe4e9988e09b231ef86091
2001 F20101202_AAAUKV corton_m_Page_095.txt
63026498eef63026195a574a9221e6be
8c4b7663964e1417e31a2de70279ba2eb9569634
2406 F20101202_AAAULK corton_m_Page_118.txt
e4c0017e30886b36c4c0c77717ec8fbe
8512962f74a92a79f9c7a45911c12c573569f118
2201 F20101202_AAAUKW corton_m_Page_096.txt
20251c2a512cc5b44188a6d151d68bee
9e6b5f10262af779af49f685de048f44f4a5ae93
2732 F20101202_AAAULL corton_m_Page_119.txt
869cfc41c92832c57835207c6ff1501d
1bcc4388f676bc01dbc42584e1d1e700f0d00c69
2060 F20101202_AAAUKX corton_m_Page_097.txt
07dbcdbcccc1084bb3a4c88d45ab311f
13c8f84db832474bf47769770b7c784466f7e7a4
6294 F20101202_AAAUMA corton_m_Page_015thm.jpg
c65eb4f1d016361db6cae0a3098ed37d
14184a5577d062762677082744205dc58e48a773
1830 F20101202_AAAULM corton_m_Page_120.txt
479a374f62359e986f0f5ee2a020a7ef
e2c41fefed2e9ee2c24d257d4a9306fbdfa95700
2193 F20101202_AAAUKY corton_m_Page_098.txt
f98990e78fb896721ae305b08f0cf4eb
6404460c2eab3ba9f21144dd460ba282236c4c55
26574 F20101202_AAAUMB corton_m_Page_016.QC.jpg
1f33059f0d58db98ed19acdaf116b058
f67e12e463b18733a7609a797e898290f65ffb80
902 F20101202_AAAULN corton_m_Page_121.txt
2a750180466695326fa6e0df9e79919b
0a72316759509bf4a0f635cd478ce6debad904f7
2333 F20101202_AAAUKZ corton_m_Page_100.txt
1f49dbfa361c2441ec467cecba1d5a22
3b70c378836eaa914f66a8a6e7289b20aaf15df8
6327 F20101202_AAAUMC corton_m_Page_016thm.jpg
83895e1ac05bf5a8b5e50dc019a98882
ed25b1b5b3a521defa4029fb7ec02f8478be6067
467 F20101202_AAAULO corton_m_Page_002thm.jpg
22498a3b7615b0bdbd53debf9fc2dbff
6ea1ed68b60fee671f29cbbbfcf93008b293c49d
6269 F20101202_AAAUMD corton_m_Page_017thm.jpg
2e5201b83755af69b99eeccf223f959b
a61218bc1710a8a55d93f5b3561b6e7ae6d73f21
410 F20101202_AAAULP corton_m_Page_003thm.jpg
30fd540552155170db449dbadc0de192
871f3e78075ba420e10844b77a705a1f80a5ec79
5017 F20101202_AAAUME corton_m_Page_018thm.jpg
d1433cd11aef38dbfc3ed7bca2ef81a8
29e014b1491208f04bdd1e27afdea9a0978d65f4
5081 F20101202_AAAULQ corton_m_Page_005thm.jpg
641afdc9109f4a56d8dc640c5ae103d1
87075b12f6859d3f0d1195ad9beb3a03a1df693b
25416 F20101202_AAAUMF corton_m_Page_020.QC.jpg
3196ab748139496b990db44b00273422
2a1938ceee5886596768e05c56dd8b16fbb3317c
2971 F20101202_AAAULR corton_m_Page_006thm.jpg
8b3e0865329a0935d5e2a6d6998794c4
124a8d54692ca5363b2c828d1d44170d5c821054
25334 F20101202_AAAUMG corton_m_Page_021.QC.jpg
de387c76847678da0c3b56f2de7c9e31
c209427f945daeaa7303a1d700870cf7ffa8be07
5771 F20101202_AAAULS corton_m_Page_007thm.jpg
83372143900bdab08c9a0766aae8fe9b
09eabbb05251fb4dfb12cef895634add7d526188
5906 F20101202_AAAUMH corton_m_Page_021thm.jpg
22c821b7d494606fc2e946d0f9529c5f
8d4b26dcb5b1876022e3d51c7165402d7c6b56d1
5692 F20101202_AAAULT corton_m_Page_009thm.jpg
ca0332598c5ae5fe026de0bf26e72957
ec724d2c661bac748bd6f8c32d155fa1bf8ac170
6711 F20101202_AAAUMI corton_m_Page_024thm.jpg
783f9ede14b1b548f5328fb8d7db81cc
d79977e6b8de9ac5e49c8ff043684c224aa25b86
21981 F20101202_AAAULU corton_m_Page_010.QC.jpg
c4a2c3b1e8664461d3c373cdc9160965
b40c8046c70e7f6f4c1234394c580ca275e85b18
27323 F20101202_AAAUMJ corton_m_Page_025.QC.jpg
2c3ba972b85fac0cc5ec99271d15bcf4
0b9e9e09f4221761471ac5136f58090ae03bea24
5178 F20101202_AAAULV corton_m_Page_010thm.jpg
b2233c2a9f75f37647bec0ce6486ff2b
35fbc6a56a35d58ab1c1d3af5a7dec629eeb32b9
6670 F20101202_AAAUMK corton_m_Page_025thm.jpg
7224a135326f91061734e992ef49dea9
087022a528675bbfcd2f0d17a643e3f889b20899
27407 F20101202_AAAULW corton_m_Page_011.QC.jpg
a10afd469cafd816259046c15622d557
84463b3a27b348aefd5282b55d5b4ec129d48ebe
1539 F20101202_AAAUML corton_m_Page_026thm.jpg
e8d5d4f46daa2dee156dedcefc8e514c
0e1aa32858ce07e05974548d4a5f93191f9097d3
27005 F20101202_AAAULX corton_m_Page_012.QC.jpg
b663a71bbd5858562d85e7aa22431615
31484faaad6d2eb88914a772788ab50e843e4b29
6287 F20101202_AAAUMM corton_m_Page_027thm.jpg
7a07cc676bdc47832fe51ff1e8ed72ab
e0587fc13c97690123d4e1372b2a88b8ffbd9131
6053 F20101202_AAAULY corton_m_Page_013thm.jpg
4c74563f7cd31d9c6e9ff2adbc67920e
6a6db0e210a279cb03b5d9cb805b88d78a2baefc
27176 F20101202_AAAUNA corton_m_Page_045.QC.jpg
75ba9fec703b7fe601660b969b5e85f5
61a5af12aa6d0dca7695724ce77cd37f06e87fdb
6724 F20101202_AAAUMN corton_m_Page_028thm.jpg
7c5c53385b6e0614c24cbb87ad884aef
33b9ffec8cf88ad8d5ba6a6c0077acbc4de370a4
24076 F20101202_AAAULZ corton_m_Page_014.QC.jpg
964cf46eb0d3eac5dc1b78c52801d40b
207eac075390d7d3be8b33fcfb2bb5228ceb8753
6390 F20101202_AAAUNB corton_m_Page_045thm.jpg
20122d4b5572a9610840d99a78331c5e
45200a71c28137b54a1274eca3a4bad4daad55f3
26459 F20101202_AAAUMO corton_m_Page_030.QC.jpg
85024c85b7b56e91dc74d915c724386f
b36f2d75ab0072ef88fa7ec6011b4ab11418723f
1002 F20101202_AAATKA corton_m_Page_003.QC.jpg
80d932924762565906cc37c2bb72840c
229b162d13f6a946796c0dcd13cf20e466e9b905
28680 F20101202_AAAUNC corton_m_Page_047.QC.jpg
8b2816cdcc0632c168f3042ccb3f0441
5bc3e8e048327b596369bc7b066940793c3912b2
26548 F20101202_AAAUMP corton_m_Page_031.QC.jpg
0693b1272783a642258c31d0d42c9551
66e69bfc12264c4c0944978a17a8f07077e75664
2101 F20101202_AAATKB corton_m_Page_027.txt
c5329726c97055441931c6b6716480c6
d4f79b3dfba68211fb9ff54893ac941630233aaa
6885 F20101202_AAAUND corton_m_Page_047thm.jpg
0160308f43b44df5d0e09523d3c68e9a
5b453f5b334b2772f432f7a5bf03636123579c1e
26703 F20101202_AAAUMQ corton_m_Page_033.QC.jpg
4b64fdce0cb8b9e862cf9b1e88292f72
610c5a8871c7999858ad45dd486493d378e20d82
6412 F20101202_AAATKC corton_m_Page_030thm.jpg
ea4d354982295914c98384122e605249
23115610954492bd1e0b5516f2a825357cbea3ee
11027 F20101202_AAAUNE corton_m_Page_048.QC.jpg
9253d1293543ea74c5daf10e1df476f5
33b57c429cb0b10e9a036c85cb56ea76d6f9784c
28333 F20101202_AAAUMR corton_m_Page_034.QC.jpg
8e90615a7465a5555d92e49ad9dbe1fc
d74bc02eeb79cbc23265d71c70c7fd6318b41da3
753 F20101202_AAATKD corton_m_Page_083.txt
fb157319a87c2bc2027483aff5918142
b12ac0f3ea6775c0244b9422d6e9984e59f6524b
16425 F20101202_AAAUNF corton_m_Page_049.QC.jpg
94870b577011e296b411cc4a55aa1629
7d884f5654c151eaa9df7cf3e3e9130b311af3a7
6784 F20101202_AAAUMS corton_m_Page_034thm.jpg
8c79fe35fb522b958a8e01546067fc9d
e6a882f3da5329572a5dda62e075124919e944bd
28324 F20101202_AAATKE corton_m_Page_096.QC.jpg
09cd502d2042bf0f54cb59d344b0c3dd
0bf0a7952e8cf8129dc35ab9195a01e724ccb931
4699 F20101202_AAAUNG corton_m_Page_049thm.jpg
2059014e097fd973ec4e24ef664e6975
48aff7d145ed09604edb7d5677d01fefe51beff0
27956 F20101202_AAAUMT corton_m_Page_035.QC.jpg
beeff487b1f147b52f71897433378c8d
c17b0d3c25b256b08f8427c191241deeabab459a
4962 F20101202_AAATKF corton_m_Page_109thm.jpg
bbe16adb4f0cb0d81358514685a58bab
3dd6933bf1eaa6ff89856a37acf1f05ae5220143
13097 F20101202_AAAUNH corton_m_Page_050.QC.jpg
7a4147ae72a7bc3c1ed939eaeadcc68f
3e099ec501fd90c85d64ac70fd20ba69630b84a1
2031 F20101202_AAATJS corton_m_Page_008.txt
7944983eb7cfa3f1f9e7efb2b8e31eac
c1cc411ddc6d41a5a23d584f12a171ee7bc762f7
29148 F20101202_AAAUMU corton_m_Page_037.QC.jpg
21d47970757490b2e29dba202f985090
be0d457cd5257c993116e42978bbe292ff866a74
203923 F20101202_AAATKG corton_m_Page_026.jp2
f63120c9649c99475cecbe9828ef26bf
15bbf369f186dfdf1e7d2ec3ad9e15aee9c12796
4363 F20101202_AAAUNI corton_m_Page_050thm.jpg
589fb1bc56bc97d2b5013dacded40dde
e06f171a92a28047778b3b7372c0f5e5fd6e6a4b
3333 F20101202_AAATJT corton_m_Page_114thm.jpg
8383c2b13b568e17dad142286d32c075
41dfb7c1d9c9c206e23ab2e46905ab971ad19784
6665 F20101202_AAAUMV corton_m_Page_038thm.jpg
06424f7bdb003f213a5dc2e9b8316b5d
f3d184c07caefa569fb2240067e71f638b73c29a
26397 F20101202_AAATKH corton_m_Page_086.pro
cc96b431cbd6d3c166853b702d13eb98
b236bdc8b314b2af22b6fe18fbcbd013e3a0211e
17998 F20101202_AAAUNJ corton_m_Page_052.QC.jpg
76fd296351b60ea527b8072e5659a474
bc58b6faad18853c063779e84aaac02309d738b4
2115 F20101202_AAATJU corton_m_Page_065.txt
4cb7432637805ef3e8f03c17435f980c
85df84754f9297e792f1eba283cdd2d16bae6833
26063 F20101202_AAAUMW corton_m_Page_039.QC.jpg
d6c9f3ba8d7ebb426618b6d1de9e30d7
fb4e9e78465a2a02427ed5623cfa46669a15f124
89948 F20101202_AAATKI corton_m_Page_090.jpg
da70215abb25f633913d9a795dd41398
9fe5548f7ad461c4bc8423586de6c0ca735ef96e
21962 F20101202_AAAUNK corton_m_Page_053.QC.jpg
b7b30a530d69822f5c9ef32d3a493838
88cd543933e78fd6c76152a5156fef223fa2f789
6180 F20101202_AAAUOA corton_m_Page_067thm.jpg
d62a07eaf248145f0b3e2e169a7e3c8c
85becf3eac2baf1490bea7e262e757add5b10466
F20101202_AAATJV corton_m_Page_082.tif
e3cbb105873dc6720faaa1c06c17e471
ca394987c87a918620883ebae4423315ff3b9c61
6362 F20101202_AAAUMX corton_m_Page_039thm.jpg
c6a89ae589bc9039016d263968f30f32
0bb23b47c4b58bed1f3421bac2e6ffb5bfa71345
753078 F20101202_AAATKJ corton_m_Page_105.jp2
f5d40f9d89126f38d6640b95fb4b6d86
a689139809f79a6b735b8da2ca0a22d13eec56ff
5972 F20101202_AAAUNL corton_m_Page_053thm.jpg
0e091cbe7e84756a3a599d5b61470b54
46aa170a815cab9b92adec5f67f8bcb47b161720
5104 F20101202_AAATJW corton_m_Page_054thm.jpg
53455a1d99f7a6f27d91a7508bfd3e3f
4c8677118e83a1e37aa2c01952f7c1f1a0e85c9a
28152 F20101202_AAAUMY corton_m_Page_042.QC.jpg
f356b8da4765656754169282f85b53e6
2685807af3e55555d0b79610186023e409d15c27
27260 F20101202_AAATKK corton_m_Page_104.QC.jpg
fbb86580333c5b4da2f19936a2320dce
495844768845d492648106c5715ce4389f662bd5
9198 F20101202_AAAUNM corton_m_Page_055.QC.jpg
8b5aad8fcd0845bf2b26788678e5621c
927208b4d408d5448954b6b5148f2d34d2f8fea8
24967 F20101202_AAAUOB corton_m_Page_069.QC.jpg
c8b5e1de015046af52ad250b41fc68c0
4b56322695b8e4b380f58b0e6121ea324eee3665
664128 F20101202_AAATJX corton_m_Page_050.jp2
690b67f6aa3732cdb46feb639b50116f
19bf75c92320a720c72cf17100627b8627abdcac
28780 F20101202_AAAUMZ corton_m_Page_044.QC.jpg
80e19984539af4ba8a0b4dc58d1181b0
b978ff3c6f1de2d68000b2b151cdf0d30f697c4f
2068 F20101202_AAATKL corton_m_Page_099.txt
991051d1a5b6634dd5a24ec0b2fa1518
26db528b846d98a966cbc8b55b93a75dcbc6477f
2781 F20101202_AAAUNN corton_m_Page_055thm.jpg
793ac7983bc6d8fa66adbf865ff0b9f6
bcb6b747eb70879a26c3487ebfd6731057bf8136
5970 F20101202_AAAUOC corton_m_Page_070thm.jpg
a7cc8027ad66b0cdc6b26d682e30e0e9
29a25ad9f6c4092d0475cedf75b102c50a7d653d
25645 F20101202_AAATJY corton_m_Page_013.QC.jpg
f2776dcd1cdaa57110ba81da3b699494
daab34e34ed3d869d87bb3e4673c3c2deb5642dd
1051965 F20101202_AAATLA corton_m_Page_118.jp2
cdf23c69f1ad43b1b2d9d4de271dd68f
440d163c8a53ee5f397b31392abd3e5db71ca145
6299 F20101202_AAATKM corton_m_Page_089thm.jpg
bdf54b856535b00c67e128fb3567263d
dbed294c778bf16cd4f724dd3454b1abe9751d81
27312 F20101202_AAAUNO corton_m_Page_056.QC.jpg
dc04a4b168eb03525f1b81e4dee298e4
267220ae7bda473db2fdb0b248b94f33585b758d
22779 F20101202_AAAUOD corton_m_Page_071.QC.jpg
695b6b783dc53772c7241e24fd901a8e
38ab824868613f454cc6a4adeb3d351e6f708337
26604 F20101202_AAATJZ corton_m_Page_017.QC.jpg
b162a077ddc5a71b9daf6be168cd3f88
55f596e7c57cbc9377f5bc073c4fadc4791a3652
6483 F20101202_AAATLB corton_m_Page_102thm.jpg
d2b9f2e8e24f3d88ee152ad051bfc23e
72db5ea2c2913318b208edd2f24c0454245e8a29
F20101202_AAATKN corton_m_Page_020.tif
ebd15caaa3b38454c5bcd54d8d23e85c
96306a4e527b6f3c55f6ef8232e1bad1e98dbf1c
26985 F20101202_AAAUNP corton_m_Page_057.QC.jpg
f8d824a2b6daa2427f49b0f7f2ee07f5
0f782d605096cbb2e9d1ff679fba93711ce3135e
27120 F20101202_AAAUOE corton_m_Page_072.QC.jpg
46ea4fe9d112c6971fa689d352e98fa8
24d10d0ed30afb7135d7305f671901194d2eaa81
6256 F20101202_AAATLC corton_m_Page_020thm.jpg
c4cab9e85ca372776a04c429857558f7
6eb2a051fc419072abf5264e9fed50b8d32878a4
4197 F20101202_AAATKO corton_m_Page_105thm.jpg
816c7fa5273b5ec52c781f31961f3e0b
899ef00b75a0ab7b0c85a7265ad2a22afc390e42
6512 F20101202_AAAUNQ corton_m_Page_057thm.jpg
16aa3d008ed2ae9b8a4a68f701029de1
f9f4c526b331179212954ff96e1f5f7562a0cd98
6539 F20101202_AAAUOF corton_m_Page_072thm.jpg
8da3333967f77aef5feeb6e186e0329e
936f83f0f5f9e3f8848389848df9b4e6f6cfc1ea
27659 F20101202_AAATLD corton_m_Page_027.QC.jpg
f1eb439efbacba4962817015ecb22887
5fd4143acfaad61d05d0db0fed07dcf85d4e310c
6304 F20101202_AAATKP corton_m_Page_032thm.jpg
860f6131682bfb2db9feebc17793f77c
04c53d88af803a9884cb32747d60e07cb0ceacc1
6651 F20101202_AAAUNR corton_m_Page_058thm.jpg
a1427de3a19ba3f8e29d94e59122c4f1
f1031e36f613e5f94d1e7ffcb9bc3216117a8529
6900 F20101202_AAAUOG corton_m_Page_073thm.jpg
f7ceda6a7dbab4a7ccb87d4d9d4ad42d
564d04004fe7adadc19499903fa6854587a1f2e0
2097 F20101202_AAATLE corton_m_Page_020.txt
861a9d8fb0e1f8d1fafea4a573fd658b
5a8ef76241fef3c56711fc6d898a339cbf14f02d
88506 F20101202_AAATKQ corton_m_Page_038.jpg
afa7022d465e00b01fde962b8f37e09b
e2ceb353755de377812e7a8cbf92f6ef41120930
6468 F20101202_AAAUNS corton_m_Page_059thm.jpg
9ea7c3ae260ada6ffba34f58eae2d4af
d8062bf1f2724b53d0ee2cffe4d4ffdbf17bb865
6349 F20101202_AAAUOH corton_m_Page_074thm.jpg
0b0ebe77086bbad987bd20ff96cbc1c3
7404e6484078ad38974a36622ddb2cbf49311d8f
53357 F20101202_AAATLF corton_m_Page_088.pro
4070fa2588adf682aa7d10f944a68ce6
249433563899a19d230a2d636b1b633fdc74436e
56134 F20101202_AAATKR corton_m_Page_049.jpg
70bb2e88812e957d00852ec12c5be086
86aba585657a42d07c211aff20102b4d68a4a58f
6706 F20101202_AAAUNT corton_m_Page_060thm.jpg
7d08f59aa8c6eedff6c62c2203155179
3ea2bad19a1201234cf2135d49baf15ac620abf9
6237 F20101202_AAAUOI corton_m_Page_075thm.jpg
9a0dd69d7b65fe2695c546fbb9beef75
148409a5567dd1b1004816efe560c4b4bfe9e893
803724 F20101202_AAATLG corton_m_Page_108.jp2
562660cb33df21b914897b5b8e515891
90a59a97fa49ec7cf048569a5225656830768ec5
27411 F20101202_AAATKS corton_m_Page_043.QC.jpg
5943826102aac0afabdb82551a40b5ea
efa2d7ac1ec48e74772234d75bd27cc5feecd349
6668 F20101202_AAAUNU corton_m_Page_061thm.jpg
4db8fa3d6d323bf891ed584cb688a272
0d5711d15a36ca70f7e14de2a01229e481492a29
26340 F20101202_AAAUOJ corton_m_Page_077.QC.jpg
91b6c482dda9bc9489b9a98397a49166
29ee162afdb4f01c142862f4bbfc6dc711f1b56f
1051960 F20101202_AAATLH corton_m_Page_033.jp2
8e3fc876fa5ebb0099a3b19539cb11a4
550961ad6ae4d8a9a7130dee09aeca445b44717d
51248 F20101202_AAATKT corton_m_Page_101.pro
7afce672d9175c1a73d65d8830794155
f57f0da1cbecbc8a17398080de2c12baf95c80c0
27814 F20101202_AAAUNV corton_m_Page_062.QC.jpg
a3b68b85aa6502c265f19efbe2ac83df
cf59f4d658981319e85033b6e8eea12b5300ccb3
6470 F20101202_AAAUOK corton_m_Page_077thm.jpg
ba8842891147726f0b5318b13c18d852
310d1a43e65a7d9a4584a4e8c09df04ffc80f0d3
81985 F20101202_AAATLI corton_m_Page_081.jpg
a1bc99cb8b3eb993110818bf06d93175
4090621d315bcd46d049a017f15585bf257fe808
F20101202_AAATKU corton_m_Page_057.txt
af05d31691de9d7cf7dc055fc36853dd
5a8cbd16b5162c5a32add5eeab07aac204c21042
27491 F20101202_AAAUNW corton_m_Page_063.QC.jpg
933c852c50d21d135f7bff05352be879
808359ced9093bda9a421d7ae26035962e2e8e78
6566 F20101202_AAAUPA corton_m_Page_090thm.jpg
7d0b12b30787898e24c0f4ccd11ac208
587c51b6c0ba9b661fa005f463609c5868373658
28176 F20101202_AAAUOL corton_m_Page_078.QC.jpg
b25f3ecd7455f8e32991901784202a2d
c2dfbb909cdc7291fb7db5f0ce44b30174b62ea4
6552 F20101202_AAATLJ corton_m_Page_031thm.jpg
cd184d560804a8294dfe2c00eaf6cb96
b5f7919b0b2383f05632996b0981f6bf9443ac7f
F20101202_AAATKV corton_m_Page_112.tif
2baddd09063f2eef13151033bd87a2a4
4ead306314885e7d2301ffc36cb13e04f1206fea
27209 F20101202_AAAUNX corton_m_Page_065.QC.jpg
fd42d850fceeb3aaf93dac8057940b66
8374e088d3c043f6e7bb81e9dd146c120c253f32
6171 F20101202_AAAUPB corton_m_Page_093thm.jpg
0ee41f7f44e16a6d99c65c62ad92d13d
06b43add892a61390db347dd71ac1cc5b0ee1a15
6576 F20101202_AAAUOM corton_m_Page_079thm.jpg
14ebec7754f9a0210b3d093a6bd76fda
e86dcb57a592a6ae1b2b924fda8be7fb21b33095
6383 F20101202_AAAUNY corton_m_Page_065thm.jpg
346750197eeabdf1917739777c8fc548
b6afea39402f6f10954c0ea24fd73746afdd0959
2142 F20101202_AAATLK corton_m_Page_012.txt
7fc018cd7cef1e0a42b96b9f820db4b4
04661d03367f737fb7b2376111c3dfcd454890dc
87108 F20101202_AAATKW corton_m_Page_045.jpg
0445516473869e855aaa28193ec3ca14
af7e360eaa5749bc2880977b31b1ab9cd36f1358
25747 F20101202_AAAUON corton_m_Page_080.QC.jpg
7a6c8d19b5bc37b2cafd9c68678c73a1
27a1de82234bb39e1102e72eb701f114b4190678
26275 F20101202_AAAUNZ corton_m_Page_067.QC.jpg
a47288e7baaf2f16f301d42a64b9dffa
9a0dc900d5301b98a99915a72893109f772c706a
13270 F20101202_AAATLL corton_m_Page_107.QC.jpg
73343c4e4cbfd1634b125f26d2e5e34f
1e3e14be30660c1dfddb7677b761092ed8da86d8
F20101202_AAATKX corton_m_Page_067.tif
36015cd59ac02817710a99c86ce106f0
f3a6a331ca106c53c53110eb389c0951097f01f4
26584 F20101202_AAAUPC corton_m_Page_094.QC.jpg
60ac95a59a81f7d1318a47321c5c4140
381a68229662c8268767184ab0b6fe34de2f1e7c
6448 F20101202_AAAUOO corton_m_Page_080thm.jpg
2b9a72238b1d2a941bec6ff79f3cc806
d208cc0cd36b468768c7683190fd35b812cbea28
25793 F20101202_AAATMA corton_m_Page_093.QC.jpg
8b89437bfffe6b93a1bc63757da8cb38
fc4122bed39e298588ae1dfa54e9aaaf8acb6550
1988 F20101202_AAATLM corton_m_Page_080.txt
f5254bab8c70825abc93bc67abcad973
800c71244d5693109880330d918fa9487a64303d
52104 F20101202_AAATKY corton_m_Page_056.pro
e2d05fdfb51da7cf45a256ddc8a6620a
46a2e69cf9caee62fab29e3044429349e3fd8128
6245 F20101202_AAAUPD corton_m_Page_094thm.jpg
1823b1c335f97b6cff601075214ea73f
78c2d88fee4e67e3ba76bb45ba6a788494702d55
25629 F20101202_AAAUOP corton_m_Page_081.QC.jpg
4fd694048eea53cac6a56c29ab7d6354
3fd3370078e05c256110d0704ddc42cf642a92a7
1051979 F20101202_AAATMB corton_m_Page_015.jp2
7956c6c20080d0bd7a9bfa540d3d3bd4
35ac81b18c2d484b8aadf9d1390fe2e6ebca87ef
56223 F20101202_AAATLN corton_m_Page_064.pro
9a325f3eb9473c6176afc0b5c1b473b1
9a18ea0d61771b499e6fb698c042431d047a0088
2232 F20101202_AAATKZ corton_m_Page_047.txt
4761c501085c5384c7aaab077f42403b
0bd9e7b2ba4f39d68b76128bed0a1b84dae26280
25369 F20101202_AAAUPE corton_m_Page_095.QC.jpg
d48a1589f8a000e403fc05a501dd8088
33c3a757ea2bec39d21a81319df6442fe3ec3f17
18311 F20101202_AAAUOQ corton_m_Page_082.QC.jpg
f9b1c7b34cfe3fd3daa9c24f2fb06a44
8c2c273d684e3970b96e860cc1dfbfcc9df6ed84
1051985 F20101202_AAATMC corton_m_Page_029.jp2
f9b897b4f277dcc266e7c81c717082c1
4379436ded3d11d96b60fec2a8fcd45e31e447e4
25458 F20101202_AAATLO corton_m_Page_022.QC.jpg
24fcfb7b817770f1c71e25bc5101847c
bb736855e5aa78d05c2743dac0ee220366590a28
27212 F20101202_AAAUPF corton_m_Page_097.QC.jpg
390d607198b85eef119c5543378033b9
353f7d2ca73f37216a3c0bc6b264ce6ce6f2b05b
4754 F20101202_AAAUOR corton_m_Page_082thm.jpg
3c68016d0f3c2023aa9d70f5fd9d5b39
6e0ad8528a73de2db6045811fa6211c72030e80f
F20101202_AAATMD corton_m_Page_073.tif
4185fb1101668b9b0b031339fbffebc9
bc57a3ad198ba437ce283d43a790dd216f71bdee
83383 F20101202_AAATLP corton_m_Page_089.jpg
f72d30e6c451b48394b14ae95dee408b
c1ec3454675d47640fdd733d2e6ed9118e95323b
6510 F20101202_AAAUPG corton_m_Page_098thm.jpg
4dbd15a412a9dc452af3424b9c2a952a
8e9b9b693a816aaf7523a10dd591937b9c783782
6340 F20101202_AAAUOS corton_m_Page_084thm.jpg
5d4a5f9d054400cb7ba78f9dc11f4039
8c4ada485315d676929ebb0423ae952a16c18ef5
F20101202_AAATME corton_m_Page_090.tif
29632015ca93b562696b9216659c4d6e
c0f530214d87f8c4a35580cb439d197f7050a073
57995 F20101202_AAATLQ corton_m_Page_073.pro
63dd8c334cfde9ef84253c12d6652c57
c3eff77270dc385f41329881181d56e02bc4af71
25506 F20101202_AAAUPH corton_m_Page_099.QC.jpg
16ac80880ee430a50f62ced614df82d1
a1f8a114bef1b34978bada1e4b025c079a9661c9
13888 F20101202_AAAUOT corton_m_Page_085.QC.jpg
7607a62a5d4cda2635d218271f613a67
4cf2bb01f46a15cffe20212f2caa665870d7aafc
27589 F20101202_AAATLR corton_m_Page_066.QC.jpg
f2dc67df17b1a8084fbc5d81540ae4e0
690b1f04f2b59044621e62ef673b3b75ee9c2c5f
52427 F20101202_AAATMF corton_m_Page_016.pro
7b9857d12c0085f2821ca061663afcd3
1537d41b1653f434abc42a4e5e29bcdcdf0b5293
6182 F20101202_AAAUPI corton_m_Page_099thm.jpg
9324c1a8e0735514f9ed334bc574d143
8dd48ece5531421d12c3978e69f08e9978422af4
3858 F20101202_AAAUOU corton_m_Page_086thm.jpg
4330ad47ccd00db30db7a11c200581c6
6b3d5612b61b8b4517cc0f19a7171a1142253b1f
56215 F20101202_AAATLS corton_m_Page_105.jpg
335c856b7d951430aba54b0e6954b2e7
16ea0781986d2253c05de283fc5aa45f058933c7
52337 F20101202_AAATMG corton_m_Page_015.pro
56de9ec6590954ce2ee8604104b79215
a8ccff6b62095f56c01d68684625a9b0aca9a442
6804 F20101202_AAAUPJ corton_m_Page_100thm.jpg
e00800a786d97fb2c1225382481b55bd
78d204612dc565c4bbdb72074db02dd2d649b11d
5174 F20101202_AAAUOV corton_m_Page_087.QC.jpg
4d1518653d50302b702bdf7baf023c95
b0d23bc17d5034efa5a87e593bfa35740224403f
F20101202_AAATLT corton_m_Page_064.tif
8c0e8a775506662005fbc8894cb41041
b4a6b31a3742375b2d64c89fcc5040e7e0b7706d
F20101202_AAATMH corton_m_Page_013.tif
0809583a564b130a243882f1e43a8a31
8797c95faafab36597eca4186a9b180358788042
6262 F20101202_AAAUPK corton_m_Page_101thm.jpg
7ab6a7ae50d5477422367481848f75e3
c2c3d1969ca695dbe91669b5e948612a7d0a99a8
27585 F20101202_AAAUOW corton_m_Page_088.QC.jpg
f13a350341a5f917cb01382e90f27eca
e54fcb81e8c35a9a37ec1504693e255bd4e3d750
2133 F20101202_AAATLU corton_m_Page_056.txt
5741a017bcb02499e117f3d7d500d3ba
637eaa86eb895d2f18474907f2685744cbe2b131
145990 F20101202_AAATMI corton_m_Page_087.jp2
339c45af0bc9e50ff42713b732eccf0d
d7e880333c1105ee977a9e2783f57ef839e7f530
20598 F20101202_AAAUQA corton_m_Page_115.QC.jpg
3628d597233a7a4ecd2cb8c6435e1fb3
2d487e145c6e6490f7a4b488e933d16c9042ebfb
26991 F20101202_AAAUPL corton_m_Page_102.QC.jpg
7a7eb2f60fc9dd1a3599137badf3a4a1
9f2f96580ec2efc33853ab5f1d78fe57ebbc7535
6416 F20101202_AAAUOX corton_m_Page_088thm.jpg
279cd8e2f28d2c7a3453878e413cb023
d514eac56c4965df757a95786ba1440cea141c82
1051972 F20101202_AAATLV corton_m_Page_067.jp2
e717ba9c7aa1d3f91394d6bb51f025d3
440493b072c7daf43defe3761eb27fcb5ffbd69c
37636 F20101202_AAATMJ corton_m_Page_115.pro
8daaa49d4cf36cfa3e2d443cd0ab6dec
7cdc607b0093f3a02791b18561c14a72aaf4491c
27775 F20101202_AAAUQB corton_m_Page_116.QC.jpg
f61afefbe2c1292f2804bcade9fa23fe
785db592048738524bc3b5475454359e405bf7ea
27387 F20101202_AAAUPM corton_m_Page_103.QC.jpg
05e57c7ab5aa40e8a854b3d9de11a62e
0a6c64b11bddbb18ad3efe867a0593abade5e341
25769 F20101202_AAAUOY corton_m_Page_089.QC.jpg
785956a60fc1caaf1bbb6113ae22ce9e
4853149a8bed0875c70803ff1629ae0498ded32d
6359 F20101202_AAATLW corton_m_Page_011thm.jpg
3fce920093d3b6f1477e6246f0f2742f
4724bba23199d8eb753a0c3e63b02bd4b6944a0b
2093 F20101202_AAATMK corton_m_Page_025.txt
6566a78c5ec4bb1a69264063332f43e9
a04399541e9a1e20474de533b221d962c35dd3d6
6765 F20101202_AAAUQC corton_m_Page_116thm.jpg
e794e3ca9573917e829ae0332933eebe
b57953ff5198eb292f3b64f00cdb3f40b98ccd77
6451 F20101202_AAAUPN corton_m_Page_103thm.jpg
773a21278b7233c87e94e75a69602119
83c52a98a4a577842663303490f955dff4d94a60
27870 F20101202_AAAUOZ corton_m_Page_090.QC.jpg
5ddf27c330aedee711cd10fa9dad5285
98acb0897f0e3c95454e1765a3ecb43fa379027f
29428 F20101202_AAATLX corton_m_Page_073.QC.jpg
1177c8500be844895fdf702ff8d0ee3d
ed29cc9355e930fd720bc5fc40303c77aced3b2e
31901 F20101202_AAATNA corton_m_Page_111.jpg
ff1682858fa76884cb858be48176cd2a
1c05daca0ce315b45c7c6ad4cf03c7194aacdcef
1051982 F20101202_AAATML corton_m_Page_005.jp2
f3d934e37734cff660afc20193ad2d28
dfbda6394daa23674a223c01c1737c0465fd80dd
6564 F20101202_AAAUPO corton_m_Page_104thm.jpg
35716a4044b9d50d6873cf4e69d51c10
f92abcb044f5ad175e9b1fc505f169a4cb29ac1c
1051978 F20101202_AAATLY corton_m_Page_065.jp2
803f1f521f55f75dea249ca391428af1
93aa603b6710709a99e109f8b5eba458c476aefe
664974 F20101202_AAATMM corton_m_Page_110.jp2
77d22a5d52e936d502badaa1fa1dc650
dae92dfb40173da1ceaed5bcd03d949dc166dcb5
28870 F20101202_AAAUQD corton_m_Page_117.QC.jpg
b254b3a907386a73c15512ed2c443657
47e89359ac3cec38340f33f60e169acc09af6141
17495 F20101202_AAAUPP corton_m_Page_105.QC.jpg
4edbc9f0cd5c138e7c834655d537ac59
cf7d7cf9863d862e87c98d18094bdd385bb37979
6532 F20101202_AAATLZ corton_m_Page_112.QC.jpg
872d4b38b86e8a8332e105c45f998b04
9d5c9bc284787ff02d25fb566a66332c484a6a21
6191 F20101202_AAATNB corton_m_Page_092thm.jpg
4a6334977d6b76f266601b88aaa3be4f
beaec7f44307f1a218dfe6abe87f148bfe99c226
F20101202_AAATMN corton_m_Page_069.tif
bbf29ed4e7710d99da4cc5717c1e3e2b
e502c6d49bd268d15cb6c29a23754c24787c0cff
6920 F20101202_AAAUQE corton_m_Page_117thm.jpg
94dcc06f9efe0310df6c24020b29d469
0b5ebc7233a2e6341e9a8c58c2bce206cab3fa78
16410 F20101202_AAAUPQ corton_m_Page_106.QC.jpg
1aa8ef6059b63a46eda0b0d751a1c7c3
420f018f7bdc664dad218af18d746c26fc540008
10572 F20101202_AAATNC corton_m_Page_112.pro
029091c565a05372856a0ea6c9987d82
e236d45011da117e9257a77724ddfc32c91d7103
6603 F20101202_AAATMO corton_m_Page_044thm.jpg
99c4f770c5f511f6c06e90e9e91ab05b
cd9bcc273e08d6e07ad8565d06b7867302424f3c
26883 F20101202_AAAUQF corton_m_Page_118.QC.jpg
8a47d207a1ad2ab93540fb7f235681b1
1d99552b723c89076fa38c65f2e2e77895c4c94a
4660 F20101202_AAAUPR corton_m_Page_106thm.jpg
13f764951198b9831d1b999af1f87e3c
a4ab7ff7be71acc2667fb7d3b0427ceb1f8b6853
1807 F20101202_AAATND corton_m_Page_115.txt
ac24bab10a39150820fe7b8cae543c54
de382a66a06ff4b8ff477f9419d0337f6a673a12
27168 F20101202_AAATMP corton_m_Page_061.QC.jpg
840d73d9cc426fe90b6b146a60f6ba35
4b2ebf7943f978c53b223a5262fa7d00320ea5f9
6442 F20101202_AAAUQG corton_m_Page_118thm.jpg
31b63efffab59ab63eafec62b8d6cb9a
df66e26e0d8f61b53d101cf965e627c4f8a4e89d
5434 F20101202_AAAUPS corton_m_Page_108thm.jpg
69f5b0b01c59df538e0d908b82b48054
548179740c0788d8fc44a839218b2e7317550558
1051933 F20101202_AAATNE corton_m_Page_032.jp2
97a5004bd5045a7fc4b9086db54c8cd2
abfeac1c931e626e29430fc0841a3e64c8801e8c
26168 F20101202_AAATMQ corton_m_Page_101.QC.jpg
5e9ce1a91dd9a7de35366d9056397b6a
9138208228bbf51fd9fd45f921ca98ad5b2fb27a
30527 F20101202_AAAUQH corton_m_Page_119.QC.jpg
04065631abf28cbe94bb8bdc3621ee53
09ed34c9c1e559d0ff267f67215a594374e05c0c
16678 F20101202_AAAUPT corton_m_Page_110.QC.jpg
0beef7d659a9edb3d27a2cadee22a907
496d4055123153f585770181f39b14afddc194b0
24801 F20101202_AAATNF corton_m_Page_084.QC.jpg
efcf0058e56c402c4c0e964e0c094614
63384262404988dea2954ca440734bafeb4d7b3e
91267 F20101202_AAATMR corton_m_Page_076.jpg
95576a485041e24c4f3c17759b64364d
c12a7286b1f083978fdf7f9d88975b1cd719013a
7236 F20101202_AAAUQI corton_m_Page_119thm.jpg
cfcc92420fde5f1a4d2ab141f8ab0c21
ffa5b270751b0ff2c9c1242c8d0f6ad6d2b13efc
4160 F20101202_AAAUPU corton_m_Page_110thm.jpg
5d67f655fcba9d74f0e693c390e4001f
518d93101b678a01f3055675750ce62e37a02378
47546 F20101202_AAATNG corton_m_Page_009.pro
963f9e84f016ba1f77b0f34df2cb7945
1c292649865c92e4d0259f2624bc27c2f0860d1a
51765 F20101202_AAATMS corton_m_Page_098.pro
a0f006bd6f50f45b3d21870b4c73b8e9
3bb7471d759a6e488a953c5044e1f4658510afd8
5098 F20101202_AAAUQJ corton_m_Page_120thm.jpg
425f8a16e1d270a71eae36087e728353
0cadfd7e7fa39f6894245677ea425fcfc34734e4
10341 F20101202_AAAUPV corton_m_Page_111.QC.jpg
ab2d94b15cece073caf14324e1bfc6b8
de2e6d5ff98a1d574704ec7bbafffccff582b0a1
1051963 F20101202_AAATNH corton_m_Page_020.jp2
091217a17ea289cca69df12e7627ab7c
5fb3791c92be0b2ec441d006221136f762788509
F20101202_AAATMT corton_m_Page_037.tif
eb1d03501dafa7aaad5501418f788334
aa4ad2384cb3fe797ac5b4e00722e64f03139ba0
3086 F20101202_AAAUQK corton_m_Page_121thm.jpg
c3ac7bbed92b390876bb15b2ddc43fb6
0d4d55f53c1db0a71e23e1d5fefe5c610868bc38
2918 F20101202_AAAUPW corton_m_Page_111thm.jpg
b5edd0ebdb5536201c34fa1517174efc
251efee92f1319800f78afec5d9e7fff840c5c50
F20101202_AAATNI corton_m_Page_040.txt
55f562d7c860a20e36a135708378cddc
80cbaced2cf92f4c2a69d37a7b3ef9e91c31e8b1
1051924 F20101202_AAATMU corton_m_Page_090.jp2
a9b936c572be61b865451018b9fa1eba
1dcb61da30942bf5b6db239b7b05c4124f034bef
140452 F20101202_AAAUQL UFE0022423_00001.mets FULL
c2f8a883fc8f123391ef0ef26134989f
c99a781ba1c1da9d69f27f187270a43f92f539a6
22111 F20101202_AAAUPX corton_m_Page_113.QC.jpg
f6d74d2e334c0b55b71d816dc7594268
833c61b83af0d25b022c29749eba15e54a18a306
37236 F20101202_AAATNJ corton_m_Page_006.pro
c4ce052286155e5cdd7034bb13e485fd
3bf71f558d45a5bc6ba107bde617b55b067c89db
F20101202_AAATMV corton_m_Page_035.jp2
d045053180f73f239c46d058d5452598
b30d39f2cbf4f7ee8748a97ce65855286592d1c8
5585 F20101202_AAAUPY corton_m_Page_113thm.jpg
816884e00873cdb58534e0db1d356cce
84887470ca6ebfd315d1c21a93da4b38fbb424c6
35566 F20101202_AAATNK corton_m_Page_048.jpg
10f8e703a3b0fc07808310c469ca9e8a
35a3cefb8ad702d462cfac51df1809a799dfddf9
84838 F20101202_AAATMW corton_m_Page_016.jpg
a74fec28cdde528b17a4b53876f428b0
f3be80ee520d21ae7040b9f0316909033a077fa3
13397 F20101202_AAAUPZ corton_m_Page_114.QC.jpg
ae9a420db57245bf9a5e32cdf7d86714
105979b6d9de3e3eb5fd5bb0172f94c7aa1da9cd
F20101202_AAATOA corton_m_Page_079.tif
80b01759d1b5376bdb549e3b358213f2
6f55b7e6557af90cf983f55a554f5c8d7a4bf1e8
1051958 F20101202_AAATNL corton_m_Page_064.jp2
af04bda2b32aab902d6571569eb503a9
134db2c19da3cc31c221689959aea86634e6443b
43946 F20101202_AAATMX corton_m_Page_054.pro
14eb309d7ba005a60936261313321d55
96ed5891423ebe5d5051d0b5d4d1a09b5849e3f4
1982 F20101202_AAATOB corton_m_Page_081.txt
db16b34773192597426cb7f0659b20d5
3f186be5f144fa24386e5d6e0e295cea72c68fbf
83061 F20101202_AAATNM corton_m_Page_031.jpg
015cc3568ab47a01361360ae04881aaf
297f14fb9610779894e38331b3b2d87fb925a2e1
F20101202_AAATMY corton_m_Page_040.tif
d5da7089270feece28dae1bb099db789
5f7e6e9f4d042b180f11cdced71b62d0bd3353e5
F20101202_AAATNN corton_m_Page_076.tif
6c2469dab22f20e6fbb1766893fa0cff
a972c7dfd7af6d17168b34a15c8a2699b0ae9559
2757 F20101202_AAATMZ corton_m_Page_087.pro
656a528d2822acdcdc9b31b5cf4241f0
7a44fadb2c1147429c0f098b742a3b73767f2e55
18418 F20101202_AAATOC corton_m_Page_049.pro
1911ae9a47880c17dad666b24902c638
9f68d6ebe810a096a6aa9fdb2ea9aa656220458a
4465 F20101202_AAATNO corton_m_Page_051thm.jpg
b3d36c7bdd6e1d2f48a2a4e988695e99
73b0462a5c88f64ed831b76f04e010f60183bbec
28519 F20101202_AAATOD corton_m_Page_036.QC.jpg
ad2ecd54ac47f6ab966c35086f16257c
0bbd212b39b15b997f30e067a86e6e4c7a0ae785
79973 F20101202_AAATNP corton_m_Page_013.jpg
caaa9f507ee69105b2fae68c839aed36
c16da940722749b9c3fdac7957946c9d9296b4de
F20101202_AAATOE corton_m_Page_062.tif
d597aa25ed6181641f255c6dd671194e
f8a733baabd0c13088684ce7cc640a8785bd70b3
F20101202_AAATNQ corton_m_Page_116.tif
1cd8859048677523e4362aaaa1126904
74ce459a96cd14ba3a570b472660bbd3f3c5b2f8
88583 F20101202_AAATOF corton_m_Page_060.jpg
252946cafa7678187861b1818b562701
4979a866627d815f542a475950bafd372a7b8a69
1051983 F20101202_AAATNR corton_m_Page_007.jp2
e12a61c50ee7784b1f2d55140f5adb42
3e043a91ff50f6046d369223205bbe7a9adb7d92
86507 F20101202_AAATOG corton_m_Page_104.jpg
6d9f9d178225ace14c25d00984405970
df944963a89da436fb2c97a0458e72182fb539ea
6709 F20101202_AAATNS corton_m_Page_037thm.jpg
27abfb53b5abdf7c27bfe096f322749e
116a699f646aacee619f1b48d295b67445e48a79
5863 F20101202_AAATOH corton_m_Page_069thm.jpg
bcfd07b52fa321d241baaa463f7e2a04
e64b320f6b21bbc6721895306e65bc598b1a162c
811 F20101202_AAATNT corton_m_Page_048.txt
d90805d3776d4b1cf54a8e3ddad20fcf
b275d8e1bb094e8bb963a1c84d8bb41af60b37f6
86355 F20101202_AAATOI corton_m_Page_102.jpg
323f641196390b1227ac7af988ec65b8
85f5c2670bb21ebdb1dfc7231c25b8c2f9e9db84
F20101202_AAATNU corton_m_Page_010.tif
e2ea2d3f274f86385e16358ea2f57b57
42de0f04280c1b6566a794a13904190a0fbfa4cc
F20101202_AAATOJ corton_m_Page_059.tif
83d21246c929e8cdf714646682d6a856
e15aaac277c6310a4f50ee242a83cf9bd9ca0e6c
F20101202_AAATNV corton_m_Page_096.jp2
b811791acc27729569e24933a13b4785
2b194b22793fe2c081f80ff25a7e2cc79cd4e335
83912 F20101202_AAATOK corton_m_Page_075.jpg
adb2e94dd2ec7cb45935a4fd09de33c4
ec3fc6a6ac969b4cc5058df05d40c56bb296d1cc
6572 F20101202_AAATNW corton_m_Page_036thm.jpg
df8084ab1d20bcb728f7fe1c88646ad9
0e5ac10af7843d4e5377ced0f63db6a7c92ffef4
F20101202_AAATOL corton_m_Page_088.tif
dea55904a6c55da3d2eae68547a943b5
73789e635a7548a94eceedf63053162531a7a790
28657 F20101202_AAATNX corton_m_Page_052.pro
33b8b6921383afc8efb18cec2269fc7b
75bda6a80862cf0736ef071a247293b86c15c61a
F20101202_AAATPA corton_m_Page_044.tif
9de240a3a92e09798fe9fb68368bd056
19e3c48b4b8375380b8d1db9386cb38615fbf391
F20101202_AAATOM corton_m_Page_053.txt
10d4f95bbaca4a77b9f3a8598877af98
8cd6facb941bf99264e64b4f3db728e4101e15b1
24458 F20101202_AAATNY corton_m_Page_007.QC.jpg
a41ab3f428fb9ccbeebcc8d635e9f4a3
77a76a171d13be426fbb1ab69e2c5d0a8394e195
1051959 F20101202_AAATPB corton_m_Page_079.jp2
c6ef57b290efcafd80b4cdfbca798ae6
1914638e255197dfafbbfa7b7e8acb6ab1e131ca
52162 F20101202_AAATON corton_m_Page_017.pro
14528f9daa6abaabcc84ebc213a55034
37dcd2750d0ccf646f538a36770ac7eb71120dd4
1051986 F20101202_AAATNZ corton_m_Page_095.jp2
e6764a0bf8f12f8505ff18cb2eea7a78
f242e9f82ee461aa06869f78a96f2945072ef952
27626 F20101202_AAATPC corton_m_Page_060.QC.jpg
784f339dd4b65b4f22e9d0c72c6e83fd
1e5dbc654dff6fe06766c0aad7087cc552d9913e
49800 F20101202_AAATOO corton_m_Page_013.pro
91183c6b7bd18d4b4bd9aff50ab5b86f
b8087d0115af18a2822ed788955ac4642ace0196
F20101202_AAATOP corton_m_Page_068.tif
b38094723b064cdff825a0947c0d80b7
a067a43fa81646c54054c76aad89b9594d66dce6
6700 F20101202_AAATPD corton_m_Page_097thm.jpg
266c20241c641a056fada946937f9241
7d0cb8ff41374098946b18e8fe62419ad31ec4ce
89429 F20101202_AAATOQ corton_m_Page_088.jpg
e2eaa0faac3a49d3d3a68c8cda1a1728
40379435509022864e4265c231294465216a469c
6663 F20101202_AAATPE corton_m_Page_040thm.jpg
f7f89427a6741f77b69d6d9c4f59a7a8
b2a5e6a34cdd76637a5c30ad68ba72deb1356c6d
F20101202_AAATOR corton_m_Page_081.tif
ad2a6480016f6ed565eb4b23d2485302
f0ae3f117c5653c8ffe4a720f5b40d96ef36b6d4
251424 F20101202_AAATPF corton_m_Page_001.jp2
247fc7d654e504edb3b9aa40869e0e03
9bf32fe6c23fca3643c900a5e143f486d00e3575
6435 F20101202_AAATOS corton_m_Page_068thm.jpg
a4557912dc5905b226522c2037123755
077d16c14283e6381e7e118a5a0c3d1a8e62e08b
2030 F20101202_AAATPG corton_m_Page_067.txt
3cc7bd73b409c0a358607bc87520b5bd
3c5ed73882f61fbd88f7565da18ca342fa702b91
50280 F20101202_AAATOT corton_m_Page_080.pro
9e4da57427fb40bfdf3acd83555062b5
ee1715cf600ea057c4f6a10de0242427faa86fb1
F20101202_AAATPH corton_m_Page_048.tif
611a607bcf547e60456273ab72b70ac8
712a73f9c4082ae4e78096c401198ae7627ce1ba
6312 F20101202_AAATOU corton_m_Page_012thm.jpg
b28aaa2a33ef5ad08a5c8fb64eb893c8
96c93f84d70ebcf509f4c4824abcc261009d838d
50416 F20101202_AAATPI corton_m_Page_109.pro
83ba4dad7b256927c8e6426e63afcc4d
c32aa7fe8b040a2748ae0e2827149463ea0629b0
1051940 F20101202_AAATOV corton_m_Page_017.jp2
85d48c58d18770ba88e10d487cbfcf49
356bdd02319f26100f89cefa84826cbef2a73725
447 F20101202_AAATPJ corton_m_Page_001.txt
543d03be09cd62862abda6acf607b92a
c402889f9474cf811e401122b86b22ca57882b48
2628 F20101202_AAATOW corton_m_Page_116.txt
e0f793c305797b3996004919a71e4ae6
0e30eeec2428a4c41580ed42c936f2646f562c82
58680 F20101202_AAATPK corton_m_Page_118.pro
7f7a8cf33092b4a9d130467d7eb74804
be3f749ce0cba07a938c881b974d0fcbc551e388
F20101202_AAATOX corton_m_Page_011.tif
8b3892d63cf1fa7d7643c950ce1c12cf
3786b130910ea771e5a86b700325bf3034774a95
2099 F20101202_AAATQA corton_m_Page_066.txt
a4cb908360a1defec5086b46fb3aa0bb
6e6b43cbd2db302cbd798a5c7bd66b5f40ed438a
F20101202_AAATPL corton_m_Page_072.tif
7b4d37e097b802aedf619e577b6cd3dd
b7135ca3ff663a1d43267e6b4f89537bc42d388e
2039 F20101202_AAATOY corton_m_Page_039.txt
f291a2d602bb40b88852ae703cdd3da8
45d9ae285d73c22d9c6112d100362350fb7b0a57
1051969 F20101202_AAATQB corton_m_Page_088.jp2
c9a2417bee53a8a13c160b0d387ca34a
aa459015a0aaa53cc64e299ad9b79e5e08abbb7b
1967 F20101202_AAATPM corton_m_Page_014.txt
72768cc0fee0099863142722637d453a
9449a0c9edb20ee5be5301c6e361bb5662e2d55f
8914 F20101202_AAATOZ corton_m_Page_083.QC.jpg
bac61eff78d9b3e9f2a9bb201a3efa78
83f888648e52a102399afe0ec88dd465249b41f3
84709 F20101202_AAATQC corton_m_Page_015.jpg
8127348e1f4fe975fd596967bda1abff
67fad5ef6e7bd02282eee4f1178ed5ac98c6addb
84367 F20101202_AAATPN corton_m_Page_101.jpg
4507205b8229ddb26850810e77637ff8
e39316d0a4e605823cf69e0fd9b5f549728ae2d6
26213 F20101202_AAATQD corton_m_Page_046.QC.jpg
3d72de1d39d76acb502dffb3b7798a07
6467949d5328dec809b152b891777782a4878804
215 F20101202_AAATPO corton_m_Page_087.txt
58329b9b1ba2534fcbb94be0f6ac9213
1be2254f09cbc908c11a6716e1df3e00c5c58987
6251 F20101202_AAATPP corton_m_Page_095thm.jpg
26cf8861c57291ae53421e2a722d6385
4f5c5b5a73cdceda8cc4f072e4601f6815a444c7
938 F20101202_AAATQE corton_m_Page_002.pro
06d91499ba57ecf37f7390069de7fc35
50cf4c8d1e87a5ff5be1415c6209f556b9d3bd16
26514 F20101202_AAATPQ corton_m_Page_059.QC.jpg
9ca470525e700a22a0a9b83dd463a9ad
e4f12713ad61f3a85fef14f7d6a013673751321a
55179 F20101202_AAATQF corton_m_Page_091.pro
cabd10b517eb4c9fb3adc8db8de4b391
ea6d5c7059d798095f883b017340d45db6aed16f
6445 F20101202_AAATPR corton_m_Page_035thm.jpg
6973e4d3d6965ba3666cb4ab59c5a3a9
8e235c4631cc250092686060fb65171c46d9fdfe
967449 F20101202_AAATQG corton_m_Page_071.jp2
e08e3a8edd126d846da66f5f07a39d4d
4bb5452aff46b42543aa3151105f9ecea4e3368e
53161 F20101202_AAATPS corton_m_Page_035.pro
f4bbf27bb08f7a501a238385895585c7
2c6f377e88dd5c35056681cc7508cdb2a7fde431
F20101202_AAATQH corton_m_Page_009.tif
3c782cacbfcbb31b5cedad54965256f8
8aa261eed88ae1b59bb3a3a47d6020ad2a38b8a1
51597 F20101202_AAATPT corton_m_Page_075.pro
f6a1afbfd49eeca1a27ab118f36c45b6
5bb8a8ac191250073529e75d4f4def09205cafbd
53315 F20101202_AAATQI corton_m_Page_033.pro
759a1d812598198ef592352c67dc46a1
517610acf1861f0f53deeabaec79ee07ea49d571
611396 F20101202_AAATPU corton_m_Page_106.jp2
32c3495783fcadf72f0dc517e246a24b
828fdeb737704fcc396b138b61fda1a0d8bffa9d
F20101202_AAATQJ corton_m_Page_047.tif
11b1481a81a0ca2647c35d6f9e1acb0c
d6e9d4b600d5c70e8d201e9ca907ad18d48fdfae
89458 F20101202_AAATPV corton_m_Page_011.jpg
4cdb730cb0137bd42689639cf16b07de
9e9fdf92863b6dad16782c71a0fc5c168086fff4
51436 F20101202_AAATQK corton_m_Page_046.pro
3d228390113c2ab9089b0ae72e123286
c2b4b6d665da510a35d645d28e1ea1167527e789
55301 F20101202_AAATPW corton_m_Page_034.pro
8df991ff932a02ac87faef56785b437a
d070abd58aa5800cec4077e667e7ea025ff5d47f
F20101202_AAATRA corton_m_Page_009.jp2
74359b392140ce543f55951f35937794
ba48f15a26478d6b0effed095b630f542d52d4c1
85495 F20101202_AAATQL corton_m_Page_039.jpg
d7dfa6ea13cafc7b7342ad6e645591f5
c7c7ed30e1d64dc6b82cdc31bf01484e32eeb780
28969 F20101202_AAATPX corton_m_Page_114.pro
cce1cbfd48a3be12ac70f4e2d8900d76
ccbc953ebcd495c6bae7be838942037ea71af5d1
73156 F20101202_AAATRB corton_m_Page_071.jpg
937004326ee5943e0c12bc3a9f84e3a8
8463c14c7ad638ba37608fe7290c7889c5601855
43896 F20101202_AAATQM corton_m_Page_120.pro
d37595d6a636650b96bc9202dd547163
defd9b2eaf3985ff1788827f351c028fb613b104
F20101202_AAATPY corton_m_Page_031.jp2
4acf68a43ba2ba63c55698896e21260c
be2f1e53ad5e9cf3afeef3b4411a4b09738440f1
21334 F20101202_AAATRC corton_m_Page_003.jp2
3e6c55a9d8cb5f9124900693612e3756
045a35e8da1de020a5de0442a0bc8968a1198a97
5458 F20101202_AAATQN corton_m_Page_115thm.jpg
ae6cc9c0a9167af9aa059f9b4e3ce7b6
b2c981b6bc3f5b8c5d06d040dd38cf19d842c478
F20101202_AAATPZ corton_m_Page_019.jp2
a87d2b3115e0f720c9449284ffb1e175
11204298520bcb4307431dd1c6413b4ba7b2216f
2191 F20101202_AAATQO corton_m_Page_044.txt
3eb904d8ac75cea77039155893a1090f
81c6db1a518a1e665c3a50994da23b678225d322
87565 F20101202_AAATRD corton_m_Page_065.jpg
105f1053be88485935f981e4687dc3f6
88bc39464974f023690d59d680c7c88a17ea52f6
27293 F20101202_AAATQP corton_m_Page_068.QC.jpg
9e5db8720e6e4f203349d9a70320267d
06affc909eaf17a540529fad38bee164fb7179aa
2571 F20101202_AAATRE corton_m_Page_007.txt
de81e5df9e9a03b13ac7cf3f7ad5d6c7
3d57187728b05fc96c94b0ab5fdacf0d7ed83e36
1051976 F20101202_AAATQQ corton_m_Page_117.jp2
1f40fbe0a633a0bca3251e5a18452700
167a50ae2d817daa1dd58ad71c11ec6af97a5582
57944 F20101202_AAATQR corton_m_Page_108.jpg
d9bd90981f01b4479d7f4043f7a1c668
536e2dc0aaa74c567f2b93651494026a4d42f5b6
F20101202_AAATRF corton_m_Page_104.tif
1b475da897555c24127f88f8c5461b28
ae433bbf0048b5e9a78c86fb731a01dddc13405e
F20101202_AAATQS corton_m_Page_012.tif
b5b3841113200ca2a0267c77ef45f2de
84b453e7261f4c3d6ad93eab615d191588f08dce
1270 F20101202_AAATRG corton_m_Page_114.txt
7d7bbb389a47fd1b0c9d886be41ddfa2
935a5edec11c3f4581a55bf0a884e5374716142f
21494 F20101202_AAATQT corton_m_Page_112.jpg
9ce90966a205ec46fb7f190109517b74
71b198235515790706acbc901f9d6a51717d62e1
17686 F20101202_AAATRH corton_m_Page_055.pro
f4f283d49713de327c638f2231837e29
d51231d9b785a4ea82f91a04db202dfefa2d7d5b
6625 F20101202_AAATQU corton_m_Page_063thm.jpg
312496629e7ec02352fee646f9763857
8ac84c2edcdddb02a34c781b18e9ad0c1a1a7e08
26432 F20101202_AAATRI corton_m_Page_015.QC.jpg
a654609b12341b5f3d28e0003c278eb9
f8fea4e6bbb9c96ec9823050f43fb56e9c0aa014
819 F20101202_AAATQV corton_m_Page_112.txt
dabb3c97940ac7e435fd3d2d07609d4d
b0e96bb44a1e70704b4dab72754854289712644a
F20101202_AAATRJ corton_m_Page_115.tif
40d837f8f945fbb61379ac796935ea83
59792a1d5c26388024445e81d18db2549955396a
110831 F20101202_AAATQW corton_m_Page_119.jpg
1bf9f2825b2681e8bb4e3ea79de8b7e7
9d4c0330823c491dcf84c6345ed082e10f9ad68b
89257 F20101202_AAATRK corton_m_Page_056.jpg
8349a01716bbc9b48b2989b4b6e6ad99
00c7cf530e6b4b85e3bd1f0b1e74766c9693c5bb
14422 F20101202_AAATQX corton_m_Page_086.QC.jpg
d5e83a827b2981d6e956ecc2dc22a5f0
be34c759bc216b584e50c0bb561eb5d611be528a
53732 F20101202_AAATSA corton_m_Page_036.pro
5c390ec94bf5dfa4db158f96bfe149fd
b274098b54fa29a3a1568186fd9a30fb9669ba2e
1051944 F20101202_AAATRL corton_m_Page_092.jp2
e8bccefc8bba82b7454d0deea9918ae0
ba2962fa654cd471de06df732e623ce663989cd8
83400 F20101202_AAATQY corton_m_Page_084.jpg
2175852007e908178d2027e319bffd73
7f5ecc878f5be47c61d447ca6ac2995685b3f8ff
6656 F20101202_AAATSB corton_m_Page_078thm.jpg
0f0557bdfc0e5f0f9ef16a5e165a293b
86d4ee5c8061b256a8058e66f5b687a82e5295a9
24751 F20101202_AAATRM corton_m_Page_009.QC.jpg
b03fa00239d81503670c05d44aed4f05
839a95e4745d452a2c834b0de1afba769b1ef72e
F20101202_AAATQZ corton_m_Page_046.tif
4f8add9d45a279d96acc3c86b69f5cee
539c2fd848f144d82676ebe6dbf9b7ae7bd84d79
1051962 F20101202_AAATSC corton_m_Page_066.jp2
2d65cc47b91809982d73fa3f9b698b6d
fcf4e2e118d08793c2a5da5cb96a2e41f393f741
5640 F20101202_AAATRN corton_m_Page_026.QC.jpg
d8c3306f363b539f9f9e38601054c6e1
9c5ddf532deb4b7b8a11bb7236d208cd7e4f7ef2
26973 F20101202_AAATSD corton_m_Page_098.QC.jpg
a671c2eabe84433cce20b1bd4a814116
eabb45c02d2e148007656edf76bacf96dbfd1336
20270 F20101202_AAATRO corton_m_Page_048.pro
bc1fa23974e5e4c7b2db19e8ebbdf7cd
f82532bfe41acbb4f669ba1152d31c71b3a56f98
47040 F20101202_AAATSE corton_m_Page_051.jpg
47b78c0be194dc4c7fa6f14523572624
4e2ad66eae49676ff9400be8f20fbca69d15d6f3
329018 F20101202_AAATRP corton_m_Page_055.jp2
c96c8ac8c55aba507665c751af3db402
15e9888c7928d032b5c066bd90648d086fe2d26a
3777 F20101202_AAATSF corton_m_Page_002.jpg
68688781128f557e16b391a5b32122ba
b2d5c51d4be94f240ea7f4575ccb28768fa5f515
21794 F20101202_AAATRQ corton_m_Page_121.pro
1fa6f482dc463960632106a4ff107cac
871b129fa371f1ddb8edaff03da5d0429bb267cb
6508 F20101202_AAATRR corton_m_Page_033thm.jpg
a565df70f762410c558128a18017a3df
ebe7a93dd40497bf4d17e928ba0f098bc6bd6dbb
82504 F20101202_AAATSG corton_m_Page_080.jpg
2ecbef08c4aeaa371ab65688bc6d94e4
7078c5215b066861cb8f2f345bc4d469f0b7793c
410739 F20101202_AAATRS corton_m_Page_107.jp2
ec5c93bb97ff069caff607b76c991a81
ee376cd6e15860678e6cff1c38e6143061c18c1f
2168 F20101202_AAATSH corton_m_Page_061.txt
085075ca5e49bc62b805fea5c1ccf9b8
f39ab619ba14a58ea7bf9d389f16e92de092f0a1
5153 F20101202_AAATRT corton_m_Page_052thm.jpg
40abed74b2c9621c79d1406b5a4cd12e
0a1a3c5ed5b6d8e9fdcaff2d5a2c97028bc29772
462553 F20101202_AAATSI corton_m_Page_085.jp2
3f01e79626fe052f53e73819f68ea865
3a9bfa7348c07b2a4805c6bbd213f1cdffe8c249
F20101202_AAATRU corton_m_Page_038.tif
7c92c90b0740d1b77d96b844cd883ede
deed979075dbd5c18a91c955cd22cacfcb5243f9
70327 F20101202_AAATSJ corton_m_Page_010.jpg
9903a441fd2e35c81ab8679dfe5b32c5
7f1aa86494ec6c6bde8d861c4fcc20d7f47e48b7
51687 F20101202_AAATRV corton_m_Page_106.jpg
3bd36835e3ab49a77bb02e542a8a0251
ac3e31ba824c95ca1a59d695f094f9dadbeffab4
1051931 F20101202_AAATSK corton_m_Page_063.jp2
568557167446e39212284d98a8fdfcd4
702342908fa480b9fa25c3193b91be9d773d5d93
F20101202_AAATRW corton_m_Page_093.tif
a5d0ac0ea6203c4c15e812ac6e9e9931
3f2ba8aec9bb54af3d772ae4db2046f5c734d23c
91578 F20101202_AAATTA corton_m_Page_037.jpg
72482d5796b739925d6c83ba07b9eeb3
8c80121107407464a1c948534207555d35c49b9e
27455 F20101202_AAATSL corton_m_Page_074.QC.jpg
a3736ce8718d1838c1457672945c8bd0
2aef071bacfda961fcb502fce2c7f9d87151abea
318008 F20101202_AAATRX corton_m_Page_083.jp2
b89ef78bf5175ca7a7aa0de039df8bca
7efc0e69d74d536f0dadfe99f5888501528a0f28
86215 F20101202_AAATTB corton_m_Page_023.jpg
8d1013fb951cbc2ea406203befa799ed
e740c04463afff4c0ef559477aa87d50566cc9f5
F20101202_AAATSM corton_m_Page_030.jp2
e809ad94b6c1e4cffcad58ff29d715a8
157a2bb6373f6ea0746d60f1c6e445c3223a6df4
F20101202_AAATRY corton_m_Page_054.tif
ac3aa56583061994a01037d37ecef26c
89dcac69d242209f34fd92e3c3a9427533fad4aa
87642 F20101202_AAATTC corton_m_Page_103.jpg
7faff116145b1482fe806d76140394c1
56c63861ac4cfa1fd3781a03632bc5ac5b501d57
27861 F20101202_AAATSN corton_m_Page_040.QC.jpg
3bdd2777b2eb74510bcfab78b3586b03
00544b0c6c933e3aae54fa18a3cd2679619c529a
F20101202_AAATRZ corton_m_Page_035.tif
b3cd37c010419f0553ab14e8ecbfe5cc
1a4e66e8ae0f35465b8988a80d7b4e023eda9415
26199 F20101202_AAATTD corton_m_Page_075.QC.jpg
e8c66b79bfe634c29802494031c8558c
a3392e915e8ffd28ca158c992a915a0216905b59
19012 F20101202_AAATSO corton_m_Page_108.QC.jpg
35c68b96b200278a8dccdfda8b4afdf5
a41afece4fc67aa6de575a1503278cabc7129384
F20101202_AAATTE corton_m_Page_042.tif
0fe96b906ef92643a592334ec4fc56f2
f5bf5edb84ce855ac1e611f8cf373c0ef9063f6b
26431 F20101202_AAATSP corton_m_Page_032.QC.jpg
69b95bcaa726569aef849b78ed507cee
d9dfd0bd4ea328fd16a87e101f61ef683e07023e
54729 F20101202_AAATTF corton_m_Page_023.pro
5245962378f8d1041a20eb7efa9e789c
1b9ae054578a94eea28719131ca69966ab88d9a0
6421 F20101202_AAATSQ corton_m_Page_043thm.jpg
d957f13c6015e75df0e65845509a5520
7e824a22ed6681099b71164b066aad712f066d5f
F20101202_AAATTG corton_m_Page_039.tif
4c110e94421576502180dc805e1aff86
f9eb3d6633a57a992ccd838c92dbd25a0ee08e68
27398 F20101202_AAATSR corton_m_Page_023.QC.jpg
807eed1a98fcdfc8566d753e3e3a7965
0f7e7c6b6af8ff6d7b481ec90ff62667a6ae994c
F20101202_AAATSS corton_m_Page_001.QC.jpg
28f1ea421969f0bc7ebd4e9f73f71d73
9e8901a7b8665a4e9c3c747e3351216c163acb01
F20101202_AAATTH corton_m_Page_036.tif
b531dd0e31ec28619ed56f2192626592
99c06089f6f19d765cfe7e11adfc09c7e632990c
3547 F20101202_AAATST corton_m_Page_107thm.jpg
d0dacc45d0ed824dbba49555520df66f
3287dadb89ce7c6d434656c5c01577f44897c36c
672831 F20101202_AAATTI corton_m_Page_052.jp2
707697be4016b2ee4ca2b990aa84c4d6
a32ed239e03b10e95621dbe3e45538c86c54335f
22595 F20101202_AAATSU corton_m_Page_001.jpg
827966e45de774a85a338f2d6175824f
585001aaecd8b8e2593a6544ff793af6ae11fafd
74820 F20101202_AAATTJ corton_m_Page_008.jpg
71a723ee654b6498ee74e233bc2cb01d
85c2c69a53cc3b170dc4c35665e8af43e1bf61d6
64333 F20101202_AAATSV corton_m_Page_007.pro
8f46088baefb32ea243a85990e751595
857c0dab719f98471332e3dc06816399aa4824b9
F20101202_AAATTK corton_m_Page_121.tif
c2fbc84b1759b8c6f49fd40a30f50577
46d72bdc6aafbb74e6cf51112d7738ea5e9f9990
41728 F20101202_AAATSW corton_m_Page_004.jpg
6b29c3111ec47c5a68e988a730f46ead
3078a2388276037beb1e29ac052985019becc32a
1051980 F20101202_AAATTL corton_m_Page_013.jp2
b5b42635bc19e91e765d44a1582308c9
1c9423cc2910ec8a9bb515388e223dd02d707791
F20101202_AAATSX corton_m_Page_110.tif
f3a15ae5e24b88862320d9871957461a
6b3aa62672b482286e0a63023d5ce3ab88b8c395
F20101202_AAATUA corton_m_Page_066.tif
f6ddbb7308de54b20ee3a8c6c8b2c18b
30de88071a7e00e2840b47eebc7598ae518e1cb4
11750 F20101202_AAATTM corton_m_Page_006.QC.jpg
0b3c0224183bdc67a88825ee5df67411
354c6309db05f51af0e1b7f0b7aceeda827d9557
1051967 F20101202_AAATSY corton_m_Page_037.jp2
56422b0b2032491fed50ba093e8b6c50
60b83af78796e8cde108f678dd0f8971b20a6d53
852 F20101202_AAATUB corton_m_Page_050.txt
5c957b252838db9686677a766ec32e57
002b1a09b93f97ed24b9ec37532e68b7ac18fb68
F20101202_AAATTN corton_m_Page_019.txt
8835751a052f328327735848ad9d380b
13b431ea221020ab8b262c7b5bdee8d46cad8bef
2111 F20101202_AAATSZ corton_m_Page_036.txt
71b71b1d30897dc9206b9271a5840ad8
af6f1b1a6705c8c9f64304b9bfbab766345273fa
25518 F20101202_AAATUC corton_m_Page_051.pro
c322a01d3ac6e69021646962b14ada2c
b9cb828bc7137dc4025a69b58a589821b12fe9e7
26857 F20101202_AAATTO corton_m_Page_029.QC.jpg
cdf23ac9164309d56af89fb9e8c42c09
6a8550573ba0eb62960aa60ea86fab1c2c2b1adc
52118 F20101202_AAATUD corton_m_Page_097.pro
1a4b59cafccec829e73396cf99fe5935
b629686091267eb4d23527e035e17ecb6c576d87
64741 F20101202_AAATTP corton_m_Page_018.jpg
c2781e9477de952d0b6ecf4b6b36cc67
47da4b3e673e9a5a172aedfc0ae99d81fac410b8
510262 F20101202_AAATUE corton_m_Page_121.jp2
0553f97be20140f22a4315a0c41e65c7
1965701f72f9510ffe40b3767907b00e570ba75c
2487 F20101202_AAATTQ corton_m_Page_083thm.jpg
11c3b9b9d2acf36b7fdf19b8627e42ef
dd963037afca9b644ac75a0db52314f817cbd5ed
47241 F20101202_AAATUF corton_m_Page_050.jpg
3bd666ecd0744d1cf1ecf9a188411066
c3a19e41f54b60b9ba52b2b88ea39570ce4bb316
93083 F20101202_AAATTR corton_m_Page_028.jpg
98400722fd18910c17b4df07f63ae6ba
fb30d544f931ea4ed4e8c922d17556c9778ae010
F20101202_AAAUAA corton_m_Page_006.jp2
015e2c1f8f4c426f3d5bcba62fae20a8
84be1fe01c8cdce5cbcfe95c6420cf85fc147bc9
1051964 F20101202_AAATUG corton_m_Page_047.jp2
adff58a949dfee88c1e6bb00a23da593
f53a6d159311c13fdf2b8b499711592de1bd0d79
26326 F20101202_AAATTS corton_m_Page_092.QC.jpg
a99936cbf76a55da4957e0ee69ff9c97
ca62c04d29fd22ff54b51c06498ad1fd0e935305
27384 F20101202_AAAUAB corton_m_Page_002.jp2
03978d69fda89abb4308f065a99cefc1
c0085f73e8ba2173fa08fd55948ede2cf26c6508
51055 F20101202_AAATUH corton_m_Page_020.pro
b710d694b19de4761a29e3aadd249a1f
978eda25ee6c61183b8de52a4f0d6731f79d1358
56609 F20101202_AAATTT corton_m_Page_076.pro
bc6530681815dd14485bd7e9b7c9f4e9
2289ab29195f2fc376b38e4664140d4d29773b81
181752 F20101202_AAAUAC UFE0022423_00001.xml
66bc039bb2584cc4ec2eb771b5bede70
12a5f34377e4fbb39d5d82fe3e28da63412a6af7
56596 F20101202_AAATTU corton_m_Page_047.pro
97fe3902bd68577ef7466b51b024c7ae
a594148d869c765bc70a1d35517b6ab01d8f9654
2278 F20101202_AAATUI corton_m_Page_073.txt
fe00e04aa914b36e02efd479e88b05b8
e91c25e9d85363739b4ab5442624e772521d9a3e
84 F20101202_AAATTV corton_m_Page_003.txt
dc98758db7360bf0eb65dcfc4777b4a7
0acb8abf2536a55b78713f24300b07a16bf2ad82
96385 F20101202_AAATUJ corton_m_Page_005.pro
e1f631e55b0336ffaf7c040bf01bbe6a
9294b3f7e384dd6a6a27127f96233b37ba1e517b
1078 F20101202_AAATTW corton_m_Page_002.QC.jpg
ad0a4ccd2bc795058d6217a19f42fbbb
4229f023a900cb268d0841cd12c94d394d26d29c
3364 F20101202_AAAUAF corton_m_Page_003.jpg
978dd605db6d8c1069f65df148448899
c0e6fdcdd7f9cc28812820dd0f1f5c2e0ead6f5f
12570 F20101202_AAATUK corton_m_Page_121.QC.jpg
26b73cf9240be17a05ea34be63d5b8cc
5679b93667e7f2e8aca2867eb61ea6165f2894fc
1831 F20101202_AAATTX corton_m_Page_087thm.jpg
00a8e4806193a5afaf30dbee1ca38f83
59d01a3ef3e25d068c1ebfd18e4996869608959f
88848 F20101202_AAAUAG corton_m_Page_005.jpg
06863fb2c09ee1382f04c2f4ecb1ce67
837c85bfc411712a3e956fe95d2f31945eb8c838
91098 F20101202_AAATVA corton_m_Page_034.jpg
985ddb5a859214ab6766c25e9871c003
2cfac6d3f29be3a6c015a0f1952e24cbbd3ccb41
F20101202_AAATUL corton_m_Page_023.tif
b95e7878a61d53bff42dbb6bff1f7a6c
93eb6ae1121d0790d70ff06db6befe23c34d83e6
F20101202_AAATTY corton_m_Page_077.jp2
cc4157e6598873d6d9521a9eb65647d4
f8a823f052696c856216a26c6a9c6714e8d021b3
47159 F20101202_AAAUAH corton_m_Page_006.jpg
6d21fb9b26c0248e193816d7d8090b32
6912389d1dd9018d8622c6a5e90f9cdbc7589198
20187 F20101202_AAATVB corton_m_Page_018.QC.jpg
39e4868f2ee024a841db94e26f73f699
5c9a5fbd35078d2890574db1ec149e68acccf973
1051916 F20101202_AAATUM corton_m_Page_016.jp2
cd04c97b4d36f972e94b11d13b91e5e4
d5771cac27dfec2bb6cf460063ce6163bbac0ec7
F20101202_AAATTZ corton_m_Page_024.tif
e628a2f532a2035a8df150451d828a8b
6833ca7f821c5c09ef2e03d02e321e43ddedbcf7
87631 F20101202_AAAUAI corton_m_Page_007.jpg
ed900b159550a563b64827dd8f9536ab
8036977e2a09e997ed974a60a7f711082cdbddcd
F20101202_AAATVC corton_m_Page_092.tif
3763ba3f850e007a48760c4b2711b1a0
95d857a8b3e94b2e9591b52891a454191f1861e1
F20101202_AAATUN corton_m_Page_043.tif
2a4cfb28b3600efb56cd323c8117edc6
770bb919fbea8614af72c1a05f6229fd2cb526e1
81348 F20101202_AAAUAJ corton_m_Page_009.jpg
77bd766b2bbb9637c26a51df339a3282
74c6ad3aa39af470e898bcbbbd513cac5dba6358
54951 F20101202_AAATVD corton_m_Page_096.pro
dceb31201d3bb91b787e9915ea325831
4edb9d68e4896e08eca743134904b40b8c7f10d7
44707 F20101202_AAATUO corton_m_Page_053.pro
1abba4af366175575030a8992d184123
6938cbaad4fb44dbd605c7affd7d03ed133e03a2
88362 F20101202_AAAUAK corton_m_Page_012.jpg
d7f09baf06a1d152627d513bb891bc09
db0a27235393790cbe50b0b7829d6a977efc6fd8
1051957 F20101202_AAATVE corton_m_Page_091.jp2
b2c23a36678700ef9bed6bf949c584c2
ec506d7c423553ed6006120e84dcd84170d77174
F20101202_AAATUP corton_m_Page_086.tif
34b9c0091c6940f55f9d242e2f9ec51b
ae7a82d56fc8793674d8f2a835b02895bf526fb0
90466 F20101202_AAAUAL corton_m_Page_019.jpg
f6a67e6b3b0aacd150d1e1259e04e24c
dc8d839c26571e8d22430a96297cda84854a0d26
13523 F20101202_AAATVF corton_m_Page_004.QC.jpg
fce690f3bcbf45d085841fe9584218ea
96372c727d7a6519df5ad584be2e64b3ecff6970
2053 F20101202_AAATUQ corton_m_Page_092.txt
7cab48b4d0dfd838e2ece4a2ce734c63
852ba87005e281c4b00b6fc2042e3805c605ca9c
83248 F20101202_AAAUAM corton_m_Page_020.jpg
f4c2b620608425219b7b14c000285667
7a4bbd7620aa8e0c98e1342bec6c9827f1230a17
6565 F20101202_AAATVG corton_m_Page_046thm.jpg
7b8ea3e10d9d40e3ae9c842734bd5300
44f690c52e3b5a96fb52bae8e4bae844bd95c6f1
F20101202_AAATUR corton_m_Page_116.jp2
f905c7cac90db2e4c186205b8e3caabb
d166b4b6406dc7b3982e7ab7307d62ff6e1f533a
89146 F20101202_AAAUBA corton_m_Page_042.jpg
ecebcc3a678ab7eb6fccc0d510b5b6ec
354efdd9cbc941abd767074c5631a0df4caf38dc
79818 F20101202_AAAUAN corton_m_Page_021.jpg
a645e74727ee8be7121313df4da45128
77aab70124c261d9d9549b45be2542d7d5c7229c
28173 F20101202_AAATVH corton_m_Page_019.QC.jpg
ca85d35af68087e4c468631b4dcc269d
cc68a0ee4e6e124fca5237b482d196aaad8b009c
F20101202_AAATUS corton_m_Page_102.jp2
18f28757864d518060af73ae6bcc6f07
3ec89959282b4024c5beb1293953a45d5294f12f
86366 F20101202_AAAUBB corton_m_Page_043.jpg
885324c66e377afa5d5d38a8de032c38
0668672d1cbfe6109a085baffce17ca558269eb1
81176 F20101202_AAAUAO corton_m_Page_022.jpg
027eab9de30e639c14ce00223a61091a
00f8b0aef627d606a0c3fa675ccee65b0a563cfe
F20101202_AAATVI corton_m_Page_003.tif
71ea0c59efea79a90c242a68d382d336
5c1ace7ef9bb0f7e749453f3f0c18504808c2fc1
6402 F20101202_AAATUT corton_m_Page_022thm.jpg
06b8e0872ed36c82f61cd183508ef97e
0268e0c8e7be758d68bf8fffd645715d48a99d28
88458 F20101202_AAAUBC corton_m_Page_044.jpg
81b71534ad6d6f5a52e0d5ea467147ed
7314401ba40e1d6ecd28d3ac64ac4fbc05e3a7aa
88256 F20101202_AAAUAP corton_m_Page_024.jpg
888eeb208cf83e4930889ad9c4dc1cba
7f0cf8c150a41a15f99cb1316fdfbb2216d9610e
1051947 F20101202_AAATUU corton_m_Page_072.jp2
2193f0f453d4c12c3c743bff69ea0ecb
56aadef07438745875f74525cd3a9e45bbcc76db
85023 F20101202_AAAUBD corton_m_Page_046.jpg
b35578471625a0855755a73ccb579b16
74bb0408b2a74ea1fe5e135c4446bf434614526d
86487 F20101202_AAAUAQ corton_m_Page_025.jpg
55cd6bd792f31677e5d6ac2b3fdfdd34
3ffcd9e86d8aedf14927fae0acb5b7b1073c2e3f
5646 F20101202_AAATVJ corton_m_Page_014thm.jpg
ec44d88a8ec55d7f3750c55767baf67f
c3b6314c8f1bf07a970b8e14d0b9b9b6e6d856df
53176 F20101202_AAATUV corton_m_Page_025.pro
f729438ad72321342d26e98b56de11e3
57356a8ad6484c37fc3328b697796857b15c7f48
54418 F20101202_AAAUBE corton_m_Page_052.jpg
eed55462b5b8452e7a854a05b31a06d5
7326f548d8e9e97aa0089a75451e8fbb23ada7cd
17716 F20101202_AAAUAR corton_m_Page_026.jpg
e4809433e270e5d7b39e608d575da78f
afef354a353a4b26256c67e2259a8594bbd8bb5c
15926 F20101202_AAATVK corton_m_Page_087.jpg
8c149d32f94ade04c022ed70086d6f43
6852957a4273a34eed772190a266405f66ede8ae
940 F20101202_AAATUW corton_m_Page_107.txt
f3dc5a30fa5e520c8fd7aed9b5e7358c
6644097ea71dc0915b4ed634811186b661624f99
71097 F20101202_AAAUBF corton_m_Page_053.jpg
df5c71181950d2258b94adebaed3ac1e
c70c56ea7a3344b3e2ccc14b4c15dc7365331b32
88124 F20101202_AAAUAS corton_m_Page_027.jpg
1d1226273557f072e2745565b25be54e
5c5a0c38d9d5d08362b4ab88153439de95b6ed43
54042 F20101202_AAATVL corton_m_Page_012.pro
8ab15cb02394b38fb630690bcbde1ca7
81dab879234117ab50725974e07a52762c8e6ba7
5587 F20101202_AAATUX corton_m_Page_071thm.jpg
a271c7db537b734d9e5b5303cd1d6694
a87adffc646ade9b051258e8a66e92d2c3e5e11e
63617 F20101202_AAAUBG corton_m_Page_054.jpg
f6578f2384ac82e94d5e3afdfc1ff5c9
903ba5d0a0b65c3e275acac83ecb810fc095d835
83452 F20101202_AAATWA corton_m_Page_077.jpg
7035f43c67b531fe046f77295da3e7f5
f122c110ec312154122b5e2df2d8bc75ac86f0f3
2086 F20101202_AAATVM corton_m_Page_022.txt
efe53617193ebaddc4913b4621c4d694
85fa084ee805272d5ac2a372046bf39cbd4767ab
2206 F20101202_AAATUY corton_m_Page_091.txt
4a14ee81c798c94f3878b7274bd2547a
cfa9c77d621487e3d8071302b3dd5f1c04655464
27163 F20101202_AAAUBH corton_m_Page_055.jpg
ff6d42471e0e958662f818d1e53df26d
dd100b18a5d5602d5e8cdaa6fffb7f2583dd0556
1051911 F20101202_AAATWB corton_m_Page_042.jp2
282fe657b5a76fbdc038365bcc1467d6
b982431a1cc96b6e54e91b74ca23a9d133aadba2
87155 F20101202_AAAUAT corton_m_Page_029.jpg
e87680e43a2f46ed79aa85e2a4caa01a
94d9571342a3b70020516d078a3f36aa804ca506
1051951 F20101202_AAATVN corton_m_Page_098.jp2
ea2f9961dc2c1a00f4be38fde1dba974
b83e35bb92a14453e14ff911490a762f09263896
3221 F20101202_AAATUZ corton_m_Page_004thm.jpg
6bc46c0475fbb17c243cba1fa3ac1d08
45ca30f8b0c7ee0b548bf647a5de0e8fb5ce2d41
87116 F20101202_AAAUBI corton_m_Page_057.jpg
a92eda2c85207cc0aed436e7d6d9156d
84413bb7d4096110c9de56254f34a69a3b169faa
1581 F20101202_AAATWC corton_m_Page_001thm.jpg
4eb02a3439d79998747f3b720cd4770d
64849fa4f36970f4e8c9f73829d5a2ec5a1bd76b
83927 F20101202_AAAUAU corton_m_Page_030.jpg
d715918ea3bd2946232386880cf7f772
f42bf9cbe8fe47fbe656d4ea8504f20ac1807815
21498 F20101202_AAATVO corton_m_Page_120.QC.jpg
eb2bfa8d32d9895aea19284943a56da6
68a0e8877d8654ed088a9f31da35695ca1e8e7c9
85939 F20101202_AAAUBJ corton_m_Page_059.jpg
4f8c12f574d009978c9f7778279d4fed
dc2b0e07f00443f31a3cc627d5621dbd0868c5fc
92122 F20101202_AAATWD corton_m_Page_058.jpg
4399fc2f3f8784812625c109f1974d16
296a67564cad7f8268dc3f2b7e482a5663b22084
86049 F20101202_AAAUAV corton_m_Page_033.jpg
6ce9c9c8822e09df33c15cd33b6ff62a
7f371a56d3bf5a2fda9d9d62ba0b23608e2f3323
1051955 F20101202_AAATVP corton_m_Page_100.jp2
2da49e774870532de63a16f32078a108
9d8f092bf372a25684f4028229ad59cae7a1d25e
88470 F20101202_AAAUBK corton_m_Page_061.jpg
f776dd0e58bfcd298b88be5239014aba
d2ef821e3aaf44a43355cb0ee755833a38b09dd6
28298 F20101202_AAATWE corton_m_Page_091.QC.jpg
5189560b6c53da4d10ae1fb77ca3104e
69201df4256ffa710508ead3a5b6477f08b83a63
86579 F20101202_AAAUAW corton_m_Page_035.jpg
22b349d42c2658c569653a489e3d5bc4
e60ceed9cdf3e6a5e00da56e987161eb3c144036
6427 F20101202_AAATVQ corton_m_Page_056thm.jpg
c6f7a90f5f896f73987c189043243a31
604492475e299120d32fa1674e9b78d93098c129
87719 F20101202_AAAUBL corton_m_Page_062.jpg
7ed00b38ef6b7c792f616f2564df136d
a5a93429d3d4dfceb6bfb191ab2c30d4deb44073
49746 F20101202_AAATWF corton_m_Page_095.pro
0e7d69fda2827258719a869ded20564c
7e006b9f4f0ae92915243c58731ee0e65ddb275d
88246 F20101202_AAAUAX corton_m_Page_036.jpg
31c3aa3d7c54c8616342482c8fef3f68
8aedaa640d0e310411c0b34e5bce9b0489fd7162
F20101202_AAATVR corton_m_Page_034.txt
5ae90702a0984db204fbd5e1ec02b909
53b4185946e244e2f78bb63af6c5582dee27a515
81697 F20101202_AAAUCA corton_m_Page_093.jpg
0f60a53378485624fb8febc819607232
786b310a7400dc2fe364e7420de0613fec953ebe
88684 F20101202_AAAUBM corton_m_Page_063.jpg
06cb635e7db60285bebe9ab179e299a0
6249eacfcb78630b8d7ee9fd411354a66b82f887
1852 F20101202_AAATWG corton_m_Page_112thm.jpg
76084ea4a902da817ea358e87c753957
66000d6114e923e5e842af11488c4ad26c7fc3b6
56033 F20101202_AAATVS corton_m_Page_058.pro
cfabe85be480717bfbb421559c18ec62
0e6a9a72d7062f05039c68f454300fd4a8d267b6
86024 F20101202_AAAUCB corton_m_Page_094.jpg
b3315f9f1e442977fa68fd8db3f7a5f7
893ade1514eddbe60a101f45a1fea1d8c1118393
91222 F20101202_AAAUBN corton_m_Page_064.jpg
e3fae665e431a5ee8bb862c68ee3bb90
2f5d347671b31e40eab51f942ab34f92ed6a077c
2126 F20101202_AAATWH corton_m_Page_045.txt
d081f8b7745c19414fc625e1b0086cf5
a9a32c239dfb3bccca6f102b1a0330d6783867d1
89377 F20101202_AAAUAY corton_m_Page_040.jpg
ba910713de8bd37875ebaf48fe33208e
002f3a484ad37f2b2bbfe50bf04070fd35b986e1
546209 F20101202_AAATVT corton_m.pdf
735902aca86269c266ca500a113aaef7
3e4ce63f7f4f1e308c07552772fb38d53801d542
84546 F20101202_AAAUCC corton_m_Page_097.jpg
bd656c53d79656ce9f0cf117a9476217
8ccab96f94093cab8c6e8b6f79f29b8bf33a9281
82550 F20101202_AAAUBO corton_m_Page_067.jpg
1ae5e8145258fd878577785e85b98442
2d19d7f2fbe6103b98bd852499133d64bbe02e52
97035 F20101202_AAATWI corton_m_Page_116.jpg
2d2c252c313c2c6723fc6120124db9bc
24f63541bb6223c8852501af716912b3a68e41c4
87884 F20101202_AAAUAZ corton_m_Page_041.jpg
f4d91bd2b3eca225f452d73e12ac3cfc
1f9e2462b092f0f53897227c349d08ef30b59c9b
F20101202_AAATVU corton_m_Page_063.txt
ca934a3d7b0ac643ef02662abcd2af9e
d8c933b2303518ccaa4a1b15b3262e2891f46520
82741 F20101202_AAAUCD corton_m_Page_099.jpg
9e7f57a4ebbff912b636cafbfe7e9ce0
6ab5110ac249ceeb176ce4b7ff9cfb0f7c40b628
88208 F20101202_AAAUBP corton_m_Page_068.jpg
a22ddd42dd94f945434bb1859752e4e1
cc765685219354c6955163b53a199112dd38dd28
1051936 F20101202_AAATWJ corton_m_Page_059.jp2
208a66fbde367de7e8b3d30bfdf697e1
0dd0f60306d8a74411ee304193b8a6a3da412ebd
6032 F20101202_AAATVV corton_m_Page_081thm.jpg
7889b382fe94e5d8d8b40cea85c0b070
0435855ab9303fbadc5c84caacf5e16cff0b91f8
94037 F20101202_AAAUCE corton_m_Page_100.jpg
3a72374801fd044decee00d4d40e37c6
15a39e0e9c81421d04a293ed22dbbe52d6d468d6
78203 F20101202_AAAUBQ corton_m_Page_069.jpg
6093eb86613dd67ef28e32bed24c2cc4
24fade99ba56f3374facc54939aa6ae97593956b
48126 F20101202_AAATVW corton_m_Page_021.pro
edc102b2e11b79254a6f52b46938fbd9
905017fcce75133413cc5bacb33ab82cf65e813f
38737 F20101202_AAAUCF corton_m_Page_107.jpg
84bad4d4b4faf3762e85e7410d0026c5
f7ff8ae0c4dc55bcf21954bd396ea7f283a25dab
77654 F20101202_AAAUBR corton_m_Page_070.jpg
26b68af253c646e04e8cbfe44185de08
6a6b7453793b60a5033ccbe77af7091c359e860e
54846 F20101202_AAATWK corton_m_Page_024.pro
8e18371bb5755653020ccd3022a0d03c
1dff80e3e79cec6d37067454e6e9460788f89ceb
F20101202_AAATVX corton_m_Page_006.txt
137e7e92fde6bc02f399b08dbd84d91a
03b56b6aeaf5157fac90cf0aa9f0aa8a58960885
61198 F20101202_AAAUCG corton_m_Page_109.jpg
8c634211d31568831712b6d0ad35a99c
bff34676c32a6228d0e335ec9c0a28028c49b3db
F20101202_AAATXA corton_m_Page_097.tif
e8c1b108089eeb4101ffc2cfcbf692f2
fb6d359554dfe8f2d97cc10323f264e3a8f4e759
94704 F20101202_AAAUBS corton_m_Page_073.jpg
05913bdd49b9fa67068f068c3041a29d
07844bd765b648e94b76bed7aa58ce69e84051c0
1051984 F20101202_AAATWL corton_m_Page_062.jp2
5c15239e6280ee578ab4545f5d9bb873
134a5e4817d11ef37b8cf1c2cffba9be40dd169f
81469 F20101202_AAATVY corton_m_Page_095.jpg
a5b4443b29cc7b86867779b151b1f5c7
619791261c75427d4bbc6959f53ae73f8bfc106f
54898 F20101202_AAAUCH corton_m_Page_110.jpg
b5448a09bc80f2239cffff6189ef142a
4fab67b4f2e6d45e3de997de8c986248cc37b256
F20101202_AAATXB corton_m_Page_046.txt
a41cec7421098c12a8a0d03c62783c89
ed4488d3b8f0aaa52b9a47189facbed63ac9d708
86325 F20101202_AAAUBT corton_m_Page_074.jpg
b635d65d5088b6b603e91b81efc5dbb7
eb128691e3360d9812d4d4f08d3115a996aa35df
663910 F20101202_AAATWM corton_m_Page_082.jp2
1d28e8729f2598cf7e51c569b7bfd008
17cd95f7bb662b1c6ea9371cf09021a15517e735
85306 F20101202_AAATVZ corton_m_Page_032.jpg
dc92cf7dcbda7460ea12a451233e2f9d
5041b826a47dbe5e8ac9512cb058e7a37bc40ce7
72816 F20101202_AAAUCI corton_m_Page_113.jpg
1f7373a6adb307daaf7d745de454533b
25234bf11f35c95ba2c3805ccf4135444e74cdbc
F20101202_AAATXC corton_m_Page_031.tif
c13fe858babd1ecc3599174d7cf714f2
28a2fcadd856d0753f31dc7a6d0e86720ca74e46
54942 F20101202_AAATWN corton_m_Page_028.pro
d54b7b71fba2d83e5ffe47f859c77fa4
a7a8d3da52dc01225506ba4582c01c0f79dc9e97
49671 F20101202_AAAUCJ corton_m_Page_114.jpg
3ef3e88829a8e74d5536d239de4f576c
0cbe3307568a524ca560db8d226fbf071328636a
91437 F20101202_AAATXD corton_m_Page_047.jpg
d6db86db7dc1c3ccd9be6a3da08f78b8
8c67025b1825866deb47bea09ed7222dc8da52c0
83617 F20101202_AAAUBU corton_m_Page_079.jpg
ac2ef4f640b05177fbf638585fd9fd2b
84386b8dc92aa70e42f89bbab3e8773301feeb96
19540 F20101202_AAATWO corton_m_Page_054.QC.jpg
35575f03c6608aa9f81a7b665fa636b3
5e392b4e9c485d8602c97b9390ecd419ed971453
66171 F20101202_AAAUCK corton_m_Page_115.jpg
20380583dfec0687717e1c3bc1898b38
a45f0ec70791a553809371bc7e05e586d33f0680
29500 F20101202_AAATXE corton_m_Page_028.QC.jpg
648c88679c7668c9e1b9f8e67fa1b561
06166d3d421b2ff2815960a997647e5a6597dc5d
57483 F20101202_AAAUBV corton_m_Page_082.jpg
41dfad099dc6a423178d834685591665
9580365ab325f90df04cf9389739d463dad0ca45
F20101202_AAATWP corton_m_Page_077.tif
ecbdeed4a52344ba031c10662f2dcc3e
8cbcce7882699d2db7e1bd2040f11d3c7ab73fe1
93591 F20101202_AAAUCL corton_m_Page_118.jpg
d9001a8820073db9d0d69da32aaa0548
da753b320151857161ee8c0bdece7e155e89ae9f
F20101202_AAATXF corton_m_Page_033.tif
adafbf2dd6e62640b0be3c2b39319aed
b086feca52839b46768443d19ef904b9114bf109
28196 F20101202_AAAUBW corton_m_Page_083.jpg
c170e09291851eb321a7f6905892491d
542c1c29566e675a2da3c474c320f1b788416d6c
F20101202_AAATWQ corton_m_Page_034.tif
4eb60982e83690b71237dd56bc24cd2e
74534541dec618c3a6f4620cef6b6ff11fb06a25
79704 F20101202_AAAUCM corton_m_Page_120.jpg
f36930e34123e6c337a5834be0d52ab4
b7144fee6c1c336db37925aceafdaea158f494f8
5022 F20101202_AAATXG corton_m_Page_008thm.jpg
502a4e1486b6f55a6ab1871277dc3387
fe642296c007a4ad86dc736235e32c10ec48e841
41691 F20101202_AAAUBX corton_m_Page_085.jpg
1e321f615072b280f3b9131d08bcb5ff
929fa89ae98af7480d08a896c7d668dd3762aa27
6725 F20101202_AAATWR corton_m_Page_066thm.jpg
f2e704e6e49d3f1f28848c703b09aa69
6a1d677fc12e61f0d8a680625653eae2d342358b
F20101202_AAAUDA corton_m_Page_034.jp2
3993cc389cd713810bef5a19433180f7
b273d345e0526c878be99a3ca51799df2d78064a
40236 F20101202_AAAUCN corton_m_Page_121.jpg
087f7cbc02465b7f012d794f4980a4b2
68d2bbcc7f844c12fd51f0134c17897607f17cc7
6776 F20101202_AAATXH corton_m_Page_076thm.jpg
a0737a7a73c25edb5ab6e256487cbe7e
01a6328283d9328c41d0bd6aafb672e3b6b3cb86
42566 F20101202_AAAUBY corton_m_Page_086.jpg
eba128d3a0a4a95bf86e12a170144e86
2a6c7bfe093695fa2c3df2ca72f6676191bacf61
3722 F20101202_AAATWS corton_m_Page_085thm.jpg
fe3d84a4f20aad1afa379581cdb9056e
4e45cc41550b80e4087a09c37e911ffc4882dc5a
F20101202_AAAUDB corton_m_Page_036.jp2
ae80041314cc3a0ba8fd783817872190
e8b72fcc618b1f3f1314cce7943b0d96992a20c1
534326 F20101202_AAAUCO corton_m_Page_004.jp2
22d7924f03fdf53952436d8546bb392f
8d85ea9532f5f706caecaf91c5417b6b286e73d1
F20101202_AAATXI corton_m_Page_074.txt
fd176243ff400cd7375f6cfcc226010a
6b34e972604dbc6a434248309b44482f08d399d9
85011 F20101202_AAAUBZ corton_m_Page_092.jpg
43372bf0ea5c0ea3b1007202066b9ae9
b83d346351edd26c07ec72efdb44378cd316b2e2
89404 F20101202_AAATWT corton_m_Page_091.jpg
1123579345b24bfc7b07298fb3b6d910
471498241e4a60fdd3e3c5ce3a1db2d151db2cae
F20101202_AAAUDC corton_m_Page_039.jp2
bc03d76991b480fff105eb8d672e5d5c
1a1e5701a4dfee02ffbd000e1ad898eea7d48f65
F20101202_AAAUCP corton_m_Page_008.jp2
c8957a1e5f681c1387adc11b54da6d04
d9d74a91d6f74b70a4d278687fed1c5f07432abf
27586 F20101202_AAATXJ corton_m_Page_041.QC.jpg
766d95b64a876475b3d6161019547923
73fa50b5183c7cac2db087ab6b074c1e108f3310
19500 F20101202_AAATWU corton_m_Page_005.QC.jpg
f50f984473fabd6dda6d132a4a01e741
800260340df9aa17de1244efd516300f3367c006
1051945 F20101202_AAAUDD corton_m_Page_040.jp2
5b1176abf12852130c8ac2f2733712a3
4b360b515fcc887ce38aedc3e94dc8033aa52ca2
943917 F20101202_AAAUCQ corton_m_Page_010.jp2
3158947ee809c6541e8f52de903205a5
51b4bdf5d2fc59dbc016bbfe79dfab086fe432b2
42740 F20101202_AAATXK corton_m_Page_010.pro
54faab2b6e46079fa29c0c8b2de92ba1
eca3bb86379f92083004ad1bc3393220ae30ff5d
6626 F20101202_AAATWV corton_m_Page_062thm.jpg
e15427b8548254c60385fbcaf081b8a5
36e7702fa9bac99b61cb755e354dde565b1bb0d2
F20101202_AAAUDE corton_m_Page_041.jp2
325f9866b10c13b05f2c067eede3840b
94fe0e71025bf9bb06e4a5ec92bddddaebf4a1ba
F20101202_AAAUCR corton_m_Page_011.jp2
87ca41cccd00808483c69399b312e7dc
06d6b43d5cfbe7492bb6562d57783874921877f7
53589 F20101202_AAATWW corton_m_Page_065.pro
9ff22da064bf1b0d395efef234473210
4fe54f49af458a2d9da46deb4ecaf53ee46b36cd
1051906 F20101202_AAAUDF corton_m_Page_043.jp2
6636016f7bae3b21aebe08c85cf152a5
29d1eaa8b5a0dc267b4a7f94d8fe7bcf3ece784d
F20101202_AAATYA corton_m_Page_017.tif
3b86e895b0ce0719c8132a4559abf34a
00561f7e0c8767a4c3852c50fe69ea457f0aaff6
F20101202_AAAUCS corton_m_Page_012.jp2
0b2d87c54a249c2cdbe44c6a718fd04d
81894b39b5b2697971ec2e6b6c1fd223eb2faba8
29310 F20101202_AAATXL corton_m_Page_100.QC.jpg
a643167708affb8286093e97de41f314
7e6ff73988b21eb7a06c29dc4d659bf97e621194
54563 F20101202_AAATWX corton_m_Page_060.pro
fc5a7263003a8c6d11d3ba5c24dbf72b
50a6b4805891970cd99ff88931760155088eac2a
F20101202_AAAUDG corton_m_Page_044.jp2
667411f8e194ad1ef7bbd084ffe26ea7
81b46e50bbc10fbd14d4ed23888bf893d47c9ba8
88402 F20101202_AAATYB corton_m_Page_066.jpg
fd75e39c0c17ec6df500001aeccd71a3
be20398c43760646cc2965689233b89166f0ca71
841326 F20101202_AAAUCT corton_m_Page_018.jp2
93a62cbfd64202d995fb112d74787972
d21f037e18f4a2e1c11a07ca66712071b4af93bb
1907 F20101202_AAATXM corton_m_Page_021.txt
7579604ffd49125149f73b45c88ee060
07ec27f75f2412a5b087a0dc1fe1fa2786a19b21
1051974 F20101202_AAATWY corton_m_Page_060.jp2
4f520b266fab6de49c63a6f4f45f74b3
dc9fb05674ad758c0e38501fbb68e9769cbbccf3
F20101202_AAAUDH corton_m_Page_045.jp2
f7606aef67d8bdf28505d602dc53d5b0
b727260a479bb2ccc8d3e875c254cb51a9a07a20
1051943 F20101202_AAATYC corton_m_Page_038.jp2
2f7a98d9979127cc44c7eda526014aa4
f40bcc08f64e6415c00fd496f5ba75c651f94613
1051941 F20101202_AAAUCU corton_m_Page_021.jp2
afcd6a2cf8f6ecfe2b06adf9f5f9d09e
6f84338cd5240bdd111e6e19274759739faf059f
2195 F20101202_AAATXN corton_m_Page_024.txt
8081f9ccb7e262dc088aa45d09abb4f0
0c8af20ef7554e1bd97adcd86bf979947efa009e
F20101202_AAATWZ corton_m_Page_071.tif
72e8dea62e6c8a2c932df6c5b3b8998d
dd2762920ff006c66b4038244042e54c1be9a315
1051902 F20101202_AAAUDI corton_m_Page_046.jp2
d857e836ec3b4cbb82d1d9527be03370
f757b53a6c6e17a8584e72584a3ab1d323854642
29276 F20101202_AAATYD corton_m_Page_076.QC.jpg
bc8d981b866e3a56d22beef4279a6a2b
a5d26215168392229e734efda76235353df4af68
54743 F20101202_AAATXO corton_m_Page_011.pro
8a155ab0209114ffc760fa2da1cbc572
8e7b8b6e16db551026ed8bd6a4e681a3e10adf52
670562 F20101202_AAAUDJ corton_m_Page_049.jp2
cfbce188a118f26206553da6d183e296
9a826b32700ac5cea877874532a06bd4d14971c9
2114 F20101202_AAATYE corton_m_Page_102.txt
d5d5444e17e2e655bd3277ef66202123
496aa24af102a2f6ab5097b949b521a3103b56ab
1051968 F20101202_AAAUCV corton_m_Page_022.jp2
826093e90482a57f7c8e24c6a46f20de
40a2aad09ee64224491a7955c7c85be8c07e8d0a
104564 F20101202_AAATXP corton_m_Page_117.jpg
24e20f27434b553e09b03b45e183d1c3
c88259000666c5d6adbc27a6a21c13e85df33e85
1051961 F20101202_AAAUDK corton_m_Page_053.jp2
fa161c5cab610fd8e4e7afabf6f696a3
823421f6a62acd7734198be452bf702a0eebc34b
15503 F20101202_AAATYF corton_m_Page_051.QC.jpg
b7df19cf13fe6e427e99e136b8c54a79
7c6c736bc1d573a67851bf10d0bf32ae0eb5ef1c
F20101202_AAAUCW corton_m_Page_023.jp2
0d720d1e7f83f28f30064560242dfc60
fa7a8396ba3c509eef047f1e3a2f880918d4f0a9
456220 F20101202_AAATXQ corton_m_Page_048.jp2
16af7e3a803af8f0860d1a03782af21f
f9c25e18b3ae0dc5b538099967013a1dd2d54b51
F20101202_AAAUDL corton_m_Page_056.jp2
cbc1ccfb6e6551d444bed20aa568191f
dd0616b57c8b3eceb864cfda53ebedc3554bae75
6634 F20101202_AAATYG corton_m_Page_091thm.jpg
749567a5b1e56b2cb5189fd2e27d616f
1630c6213da28777dfb3dcb052a1f5743107968d
1051981 F20101202_AAAUCX corton_m_Page_024.jp2
ea297ab6bce198ab0c3ff6159cfa987d
ba68543434a38fa398d03434007c7b2bec995244
1051970 F20101202_AAATXR corton_m_Page_094.jp2
bd97f50d103002342e51444ba6a1e86e
d409414e14fe587d4159e5e24a5577dded1d848f
F20101202_AAAUEA corton_m_Page_097.jp2
e6cc3eca11f1ead0f576067529796122
0f593e6770d55651a44dafec0860f672cf76e775
F20101202_AAAUDM corton_m_Page_057.jp2
e2f40b0d58e2e2c30952653ce9b46ba4
483eb35bc0cf3e8b6f9681d4e7f917ded7c7edc2
560951 F20101202_AAATYH corton_m_Page_051.jp2
84ac838666f95b293d95165b9139126a
d1469214e73d6eb62cbd5c9839f703c05945e95f
1051966 F20101202_AAAUCY corton_m_Page_027.jp2
46469e6306e949796e645dcc662e5c98
1aa13f05cf3cf6989148d4b36c9b7824afcaa3b3
15264 F20101202_AAATXS corton_m_Page_083.pro
b784f8ccb9a70024ee1bc98197fa156c
e52b0412e2dec0a2c68f49100c5ef0a46fe34241
F20101202_AAAUEB corton_m_Page_099.jp2
cca7e24d2f0c9ccec8b7ac21e21d326b
bb9b7cf0feed33214ce7e0a46d4998f2b8d7ecdf
F20101202_AAAUDN corton_m_Page_058.jp2
b4ebbc3c0ef7e5f4323fd63d150ba747
c714aa4859c7ac2b95f516b9d86308b67a2c5dcd
54787 F20101202_AAATYI corton_m_Page_044.pro
b42bf4cc3feca9c7c7eefe976c5f1f63
9fb087a162fe6fe7f046ec8d10da9dfa77640f6e
F20101202_AAAUCZ corton_m_Page_028.jp2
c93946c5e7f6729a49c5c5c94233334c
3adea0d1efb12e46a2e9b0f54976667c3d84a58e
85498 F20101202_AAATXT corton_m_Page_017.jpg
892c7d546b2a6ce5b3111b7d341ca109
65734ff525c7213a11ef31076c2313087b563cd2
F20101202_AAAUEC corton_m_Page_101.jp2
5b4790c4ac8cbaec1dbb90cc8fe7e0c0
0591f9c9689390ec4d93dbf8edadf7a455599900
F20101202_AAAUDO corton_m_Page_061.jp2
eb765da1fad63cef1738268f6fd1216d
79bfe982be10484f3f3778cbccaf249443798785
2765 F20101202_AAATYJ corton_m_Page_048thm.jpg
627702248588fd60a2f962e438ffd135
27658c12d83e5ac2df036b0490cf06e936375ad4
764744 F20101202_AAATXU corton_m_Page_054.jp2
0bf14760c52c6e4c6ed322f13678c82b
b7028b45f2eaa889e655cb4fe2af8ea9f1fea773
F20101202_AAAUED corton_m_Page_103.jp2
c7031493eea11fa29d4a3cfb9ddf9378
2a8ad9f4672cfd3597b38dbcba53160f2949aec3
F20101202_AAAUDP corton_m_Page_068.jp2
337ae29a4cdfb8957ab68c8af6d413d2
cc8e38da877abe372354560aa7b7a0c317b11edd
26647 F20101202_AAATYK corton_m_Page_079.QC.jpg
cb554c18452ca260bfff856808d77dde
85d2b911fe2f8895ef39004e34b32cecf92b5634
98 F20101202_AAATXV corton_m_Page_002.txt
a859e9e50243bd72f50e2879e006836b
b714aa3021c9c028c33c218ab3c607267ebaa9c3
1051953 F20101202_AAAUEE corton_m_Page_104.jp2
dc2e26031bfd02178285e88efaecf3a5
0a8f7b073ccec20f1147cc5d1b0d0a91cdb73389
1047478 F20101202_AAAUDQ corton_m_Page_070.jp2
8fb0c210686105d8f5df07f1d27dbf80
1056cfb2778ab27002b2c27c843c8a6101178499
1051942 F20101202_AAATYL corton_m_Page_084.jp2
aba29a0bacd71e6fc3de623bbac4bfc1
220d4911218d04cad5ac1c167b023a143499ce1b
864 F20101202_AAATXW corton_m_Page_049.txt
b6b0b3a0e4283b4260c54b41a563c474
ec5149f1424c6da88dd77ab8c93cfc5176e5f51c
772203 F20101202_AAAUEF corton_m_Page_109.jp2
3a0665de92cfafb58d2125e2fc659fc1
08f7adf8b2cdf362851e4da9c77e9703dfea79bd
F20101202_AAAUDR corton_m_Page_073.jp2
7a3b9041366472b5bbc6cddd5b8a3ea0
d5eb24c5c964eb1b067f3c855351d0abc35e01d2
6486 F20101202_AAATXX corton_m_Page_064thm.jpg
4950fc22756e926bdc43a90c9d7dd591
d4a8dc2e7800a7abafb91697ea56a24a22699888
369915 F20101202_AAAUEG corton_m_Page_111.jp2
c7c9fbc48c8bf1276d25443c800b331c
bd5cc7c386645544cc17c0ee6883635dfd1ed500
6803 F20101202_AAATZA corton_m_Page_041thm.jpg
bee83e3b5e7cc932750efd5b714fb297
a5e7bf31628531f4714d13969fd41685cc48c739
F20101202_AAAUDS corton_m_Page_074.jp2
7489d0f77de0aae99b510f80fd602c87
3c290de57ab16ac060ebfacaa405b0e257553a24
F20101202_AAATYM corton_m_Page_083.tif
ea56bbc4225617e816fcbd78bc3cbe61
0120c0050dcff7d91398c2feac73efa7c6ec7ad3
89549 F20101202_AAATXY corton_m_Page_096.jpg
b6b78a5eb7968d7025cdde7c9b1a4ba5
ca528dc0cbacdf7eca1c48744a325995ff52d191
242402 F20101202_AAAUEH corton_m_Page_112.jp2
d1a584476349ee0f9ec3f755021b88eb
de32c49924ea67684d9056f6b933fc5d7e568f09
86054 F20101202_AAATZB corton_m_Page_072.jpg
b63e9103cac1ddf6a6ccce495dcfd62a
291dc92a6b60b834089d89f6c4f6b10e4832432f
F20101202_AAAUDT corton_m_Page_075.jp2
920777b4e25daa73262e6725fae35e6d
7d6d1c2f62717ce3a28c9605f9201fc86a7d201e
8346 F20101202_AAATYN corton_m_Page_001.pro
5ecfcbb565d77437c63db0ad7197d7fd
65deaf95ac2da01528d91cb1c8366794ee1a50ed
6808 F20101202_AAATXZ corton_m_Page_042thm.jpg
49b61b6b273a5623cd9539f438a09e5d
741e048bdd7b532632f380f21a5946452b0cd52b
974101 F20101202_AAAUEI corton_m_Page_113.jp2
508b4eb05380a6db717c1bd7305fbb28
9c279079ad7fdab6393b7b4de2fe1adb29c795e5
27885 F20101202_AAATZC corton_m_Page_024.QC.jpg
7f6606b252ec9f9d07d92d8007b8a0b3
b1d32413a664d0f9f26b331998cbb1a228b1f87e
F20101202_AAAUDU corton_m_Page_076.jp2
aecce4c19765f6a69babcab7f430ba6b
b7f7abcfc4225477ac7ba062af5284a93262dc30
F20101202_AAATYO corton_m_Page_005.tif
2316e71e07eed416d20cb0c47c91e151
60f12dde7e76c586204dd4038f897f50b37dafcb
640860 F20101202_AAAUEJ corton_m_Page_114.jp2
370b54cb2013b6ba519b388dfdb11304
44dd943f1220180451410e4381b267ff3139edfd
73945 F20101202_AAATZD corton_m_Page_014.jpg
c836603ce3015404aaef122c7e8bf2a9
1c28bf7b8a1cd0ab9a1a5ea1a93cefe7498d4d9e
F20101202_AAAUDV corton_m_Page_078.jp2
9089eae579d87ed49b5b6f3087cae0da
28220d800f0d1f46107dda80ccccff3bdbb66fda
19541 F20101202_AAATYP corton_m_Page_109.QC.jpg
e359dc25f16ded2b11e7719096e0644d
d2fc80ee858eb1281c8721f07ffe05d2ea1bbe2e
844572 F20101202_AAAUEK corton_m_Page_115.jp2
8803deef88a5be71936246f859f1fd9b
b8732d9783151ba44e964742ac513d0c681e901b
29147 F20101202_AAATZE corton_m_Page_058.QC.jpg
884aed1de7c86d1e7a429e60cef80a28
461d5d3391b65e795366a7b0d0ec272cf58e49cf
F20101202_AAATYQ corton_m_Page_075.txt
b5aca3f80d232e97201a3ded9ed7fed0
3c87def0b39ff10833ef40cefc1398b41600013a
F20101202_AAAUEL corton_m_Page_119.jp2
c84db3cec8159b9ab71f96f00d90067a
b7a504424a264bec8a3f015bef0ae3245f1ce2bc
1051977 F20101202_AAATZF corton_m_Page_089.jp2
2a34d9a9621a73397a4404a643eab74a
668ac27cb7f2796a29632ab923070796bfff0de8
F20101202_AAAUDW corton_m_Page_080.jp2
5a2b819dd8018c46b2aa7249045b6e97
21e8100ee81f4195a7f2189d170be0de148ff683
2118 F20101202_AAATYR corton_m_Page_029.txt
3f04cf0e5bc8799d8d64961899eee9ab
fa81bce21ac24836c4f3d648559f4b5b36124116
F20101202_AAAUFA corton_m_Page_025.tif
db028c511437f0df1f4a33c9856bce03
9bc90ff058a5e19eeee1b0f96413522d8663e477
F20101202_AAAUEM corton_m_Page_120.jp2
0ccfdda91df6ad0c0c1eb9310db59c1b
7b3e4253884f9eaae8d32af52ce736dbf6d7aeac
63582 F20101202_AAATZG corton_m_Page_117.pro
1b8d86475d422485123266059656187a
8325e6cd8f52e1b2bc16ba79bff9d8215396f293
F20101202_AAAUDX corton_m_Page_081.jp2
2797fb013aa5a1a19aaef82c3f4462df
bf89e9ea09a6f919397ba6766bbd1ff004773655
25548 F20101202_AAATYS corton_m_Page_070.QC.jpg
1a5bd8fa04f5fbf8a8521f7808dd9470
5cb29c1b5bb64b6b0a452dd6af2469520fdadc39
F20101202_AAAUFB corton_m_Page_026.tif
ab778fff0d6252a1fc07c70fcc574448
4b9877003f1b3feb2343b99b428c1f7293032644
F20101202_AAAUEN corton_m_Page_001.tif
27c916647b8cbb47c664bfcb2e068be2
f74206e0f6985a8d8187fd7bf6260402fdbb1ba1
6561 F20101202_AAATZH corton_m_Page_023thm.jpg
2cc8b32074b780aed89f0f5d2bad8705
19b8fd888ab2e65a7301d1ef127294c23e131b11
488780 F20101202_AAAUDY corton_m_Page_086.jp2
050dbd2ac55dd9b5b4a57ef071bea3f2
57b9394289470f49e17490fc4cff9328c33b3eae
982949 F20101202_AAATYT corton_m_Page_014.jp2
bfa800593e32df573c2615ff48264f08
9f7d8a39d81373967926ccdc1772d59e1481d5de
F20101202_AAAUFC corton_m_Page_027.tif
ff0938a3304117057a11b4ad4805022a
c99aa9a3951425d91f4194a4d02a48f91fc11c13
F20101202_AAAUEO corton_m_Page_002.tif
c395139e82af7bf56463951cb0b8a1fa
7990c5add2cae3dda394ee7df80573a34224a520
50597 F20101202_AAATZI corton_m_Page_031.pro
69864a27823d4a23dc128ee83c6c7498
63248219ae91f15157fae958080aabbeae09d3c0
F20101202_AAAUDZ corton_m_Page_093.jp2
763441e4a4af08f17d30dcf02b914225
f7ef236ef37c27470356a9a53b404a25b7d4ba1f
6696 F20101202_AAATYU corton_m_Page_019thm.jpg
3d21db4e860c50c47b557957aa9f0ba0
8779f24177f793fb7e9c584ef1546ab34482bcb4
F20101202_AAAUFD corton_m_Page_028.tif
0ee825045bfa476401c73aca1cfcb571
5b0642621f4fe02321dbcc63a488eef8b15b6366
F20101202_AAAUEP corton_m_Page_004.tif
b9b5d4b83209f6dacd0246df59bee50a
47d89f9de089fa6f453ffc31e8e8699a385a9cd7
50944 F20101202_AAATZJ corton_m_Page_099.pro
feee1463c0bc98b7ebb5e21053c14caa
93b8382ca6e73bd19bac19b983375ad8b03c3889
6581 F20101202_AAATYV corton_m_Page_096thm.jpg
b328190d6b42a2b7f8051400894aca79
01e1f9e75284ca6a32aaf62f56b5f25f2a13623d
F20101202_AAAUFE corton_m_Page_029.tif
703818897fe7033626004e9cd5350e00
b04fcdb6edba8863436ef8ced47c6eaebfc3def0
F20101202_AAAUEQ corton_m_Page_006.tif
6f7d91e9f27511b24f503b09ba510688
5e45ba6aca5979b4020adfffa1c743145ac2a0c9
1647 F20101202_AAATZK corton_m_Page_110.txt
145e649b0b530a6648a6c9259162a381
4a86316e1bbdd28db6d967b4e2a284cac2b6c500
F20101202_AAATYW corton_m_Page_025.jp2
f271141361a6c0fb25c716d4e6acfe18
5354a2dfa83cf314aa439c5fee055a60aa3175f2
F20101202_AAAUFF corton_m_Page_030.tif
476ba273ffbf1dde941f2664de9056a6
84aac68f72219320f88886a32488dee96ac5f94f
F20101202_AAAUER corton_m_Page_007.tif
a3d7e4f58542dedd45da6aec134c26b1
ca2ef6de783907ccb00d5ca5a210569f9ecb988e
6556 F20101202_AAATZL corton_m_Page_029thm.jpg
996a9a30d37270b11e0341d32dbd138f
460d6beaba4b385f1f0d9202f3e02678c9c9ca79
F20101202_AAATYX corton_m_Page_108.tif
f1f8954a7ef9218ddc912bdcd3a54f41
bb368e5941a61372a1aad139713a66f47b22ed47
F20101202_AAAUFG corton_m_Page_032.tif
0e8189d83de8ba6472a9fc2621839a2d
006a09d0287120b571c1b02008681de1d6828458
F20101202_AAAUES corton_m_Page_008.tif
771b6c2a30e34d7ead3f0f27decb01e9
8113b79d4a9a2b0ab802c9af3cfa8c45ac6b6e79
2231 F20101202_AAATZM corton_m_Page_011.txt
c18509be06606380b2ff1f1ec29fc013
071b0166961e4736454e67b2f718b9449690962b
F20101202_AAAUFH corton_m_Page_041.tif
db74b9b467378a72842cd77b503e428f
7143b76c2ebb9a2c702717e4650aeca81df6497f
F20101202_AAAUET corton_m_Page_014.tif
2fb59b501f3a58c48b91a35c9acfd2f5
ea32b660e78234deff1dcf43ce9a21e203860afa
2458 F20101202_AAATYY corton_m_Page_082.txt
ce66cb4d3c90f4f42c79bc90aacc59cf
d0e56baefb9c791c54d79bb3a1db86b1c92a2648
F20101202_AAAUFI corton_m_Page_045.tif
ad58c0cafd49df104d71c1832ac7cb25
70729ddc79f0cc4e1bf3125d41dbfee79e6f3a9c
F20101202_AAAUEU corton_m_Page_015.tif
e4302f0fa50eaa92c2d9cf5b6bb24363
d034b9f9c7974d4243d62d80f48f20041f912c82
20706 F20101202_AAATZN corton_m_Page_008.QC.jpg
644efa4d0e720468411f110dfccbd00d
4693a550da310b9aa02ddcfab569225144afda53
90514 F20101202_AAATYZ corton_m_Page_078.jpg
eb8a5c1f27ae1f2491ace6e63fd85ca0
3bd46522983718fc19303171790fbd99b5b56cd5
F20101202_AAAUFJ corton_m_Page_049.tif
d6e6ca22f736b2c4bf416710d98e0dc3
193c762623fdeed4e25d9e8c40f06dc1b9c58595
F20101202_AAAUEV corton_m_Page_016.tif
e447681646f2fc69566e0d94a08c9074
3fe5de185b1ebbdf69307b4eb109d51e54777b21
46167 F20101202_AAATZO corton_m_Page_014.pro
b051dfcee45f33217eb4bfa3eae4f821
727293b7cfc801e095a5874e09a922434adc6d76
F20101202_AAAUFK corton_m_Page_050.tif
350dc08e4a12f62d4bcf46f8311e2782
8ebde313e4fbcfb101695391930a4799783a16cc
F20101202_AAAUEW corton_m_Page_018.tif
7f01d4202e94fbc70ad07d2bcc3766e4
1121865902bcd0690230ac9b799949e0b98ba7d7
53695 F20101202_AAATZP corton_m_Page_072.pro
c5e431d88c16ee06b49fce5f02616ed0
16d11e30e25d1a201b1ea3800e6241cf6e753d24
F20101202_AAAUFL corton_m_Page_051.tif
e6a20d700de9f72e95d83aa5ad7db19e
e7d4ad4b10a3fa04878671c0090af66c79d75bc9
28920 F20101202_AAATZQ corton_m_Page_064.QC.jpg
a85edf37c879edac8a94d051b98d462c
1ae1490b2ae4c70c903c6a96d6a0bffd16f4dddb
F20101202_AAAUFM corton_m_Page_052.tif
616c002d07d671598fcf073949a29f05
a7af2504c461dfa61caaecdaa565b9e4489488c6
F20101202_AAAUEX corton_m_Page_019.tif
09e296263ac2c012fca5abca4b25720d
55ae2ca89f3cabfacbc5827ed3817ab8f3eefbd3
28240 F20101202_AAATZR corton_m_Page_038.QC.jpg
baf60a34e3fb0231bb7b991f8fd5f533
4bd4c1a01ff935bb7e875f676be2801e93bbf0b5
F20101202_AAAUGA corton_m_Page_080.tif
f77ede5155c74e02504f3d7ff40d61f0
c3ba2d8b7e6294c768e6b9bfdc3ef477de51d48f
F20101202_AAAUFN corton_m_Page_053.tif
1c1cb80abf55b82907dd7d7b364a4bf8
ec067eebbd1a2fcb6b327a60d0d3e4145dc91a1c
F20101202_AAAUEY corton_m_Page_021.tif
f0d5286502add633ec454849d79fe80d
c1b0727af8533fec629ca2178340bba1c4373846
F20101202_AAATZS corton_m_Page_118.tif
6b8e6028f3a05cca8d8d2e397b627bbd
4dc9e82a377efcaec7c63023c0e049b53ac48354
F20101202_AAAUGB corton_m_Page_084.tif
56579e34bd96ac860cc54db589e0af18
9270045238199a99386713c100c989ca01114422
F20101202_AAAUFO corton_m_Page_055.tif
dd4885f1aa2f2568a0904558b20d368e
69c91981976359ea8fdda999d7ec1c6435bf2449
F20101202_AAAUEZ corton_m_Page_022.tif
2bd704e3defdc97e6acad8d28b9975ce
fa4a37a8725682e4a7ad2036df574089531b516c
83338 F20101202_AAATZT corton_m_Page_098.jpg
7b321c6aeec4fdc84fc1df13015ab434
89f8f7cdb13b725ba6ddaeb8ca8e533c50e549ba
F20101202_AAAUGC corton_m_Page_085.tif
d86b4e0995677c5ea8ba4bd520ef70a4
3d50077ef0f1348a681e9ea03a9bcff5298d2be6
F20101202_AAAUFP corton_m_Page_056.tif
719737ee25f049bf79a28864fb77319f
88ff31476c0afdd22a96a7ed157b0e501b872008
F20101202_AAATZU corton_m_Page_068.txt
bf1760337b729f16f6ebe9aac7a81062
8526913a1ed3f9d0ecb7f3f5f2d857191836ddec
F20101202_AAAUGD corton_m_Page_087.tif
c0b1978f96045a04e7b8946e37f3d2a0
025f7b191bf128c97a2af100c75fa34d4f0f7f03
F20101202_AAAUFQ corton_m_Page_057.tif
626c7fbe4e2df7d4b1b00af1fd03be78
10f24e206a0813c7f34a81b102d056133d3b111d
55316 F20101202_AAATZV corton_m_Page_037.pro
825746d23c3e9c55293f5842ec892aa4
883450d087b0aa9e3ded2257bcddfd5edc91da9f
F20101202_AAAUGE corton_m_Page_089.tif
d0f6a3e3d8bac7c5629b67e18ad2dcc9
21a20af6f903fcd77540395820be966b7a0e016d
F20101202_AAAUFR corton_m_Page_058.tif
621fa26b164a55f770e3cb1fee2c8524
260932266be044013690c892ef32dc77b8dbf04e
1046647 F20101202_AAATZW corton_m_Page_069.jp2
e8c2ec863a9c596926ae6f61115cecd4
fffcafe86d6a493407f11854e4511cc69f727dae
F20101202_AAAUGF corton_m_Page_091.tif
7e8118968f185d8176009475d60e084d
c5ed720fb5c9dbc24283cda13ba7a123cbb4771e
F20101202_AAAUFS corton_m_Page_060.tif
8da786ec41f3264de169239b4ed302cd
c7654c356eca5f7e01ec57557589793c310c3cec
F20101202_AAATZX corton_m_Page_102.tif
21bf653041104c7f139a432a30f18f25
b3cfa45e5d8e9acae281e73eeba3373f799f1c31
F20101202_AAAUGG corton_m_Page_094.tif
3600260f33058440ec1aeb84a6ce5f81
07020e518202a658ea63c4407db8d977285148a8
F20101202_AAAUFT corton_m_Page_061.tif
361a105213bde525a81e243674e1632e
b9f8dcd0a3d6bd5fd6a11a51c70b796a63c701b0
22196 F20101202_AAATZY corton_m_Page_085.pro
3219e03182d78a4170c2cd145d2d5769
5f9c16aab604ec617737ba0fd77b518899beafb0
F20101202_AAAUGH corton_m_Page_095.tif
e6bba0befc562224ea05ff53cd757445
21e58b710afd2168b56bb77b6e7ab76b2125d3e8
F20101202_AAAUFU corton_m_Page_063.tif
fc114c35daf16791d831d4aba34c0402
6d8e7b65ee7dede59b0d9ebba7e68c43a229a894
51462 F20101202_AAATZZ corton_m_Page_089.pro
bb2d944c9cf21576eba415ee2ac3f742
9b70122cb204aa00511ba74eb6e276be29221588
F20101202_AAAUGI corton_m_Page_096.tif
d521504700b7c65981102aebde0ef3c6
36e387b8fb549e544f4668f9a494850098ee809f
F20101202_AAAUFV corton_m_Page_065.tif
d9253e6828733c44917f350b5449f9d7
5de32ca4ecc156e450c652e9e892ea06cf96e938
F20101202_AAAUGJ corton_m_Page_098.tif
28101a378fc8ac6774b91f5a07962ef1
d6fdb4e96c649e771e2ec67e429dfcdd200856bc
F20101202_AAAUGK corton_m_Page_099.tif
68639e9602e9871e25b0d58eba7c22a5
e03923f24e1b9c662681b870f6f44547f012c827
F20101202_AAAUFW corton_m_Page_070.tif
c060168b2a24849a61ac3c77a9e2a51e
a85b4ca63fdb090d6b9846bf96a86c4f1f623ab6
F20101202_AAAUGL corton_m_Page_100.tif
4340311bb31e03fc227bcef9a7564fa8
39f955ae75a4e0442d74ebb04c70df50e690fca8
F20101202_AAAUFX corton_m_Page_074.tif
751469a402ed3f3dafa30f8d669b6b07
90a3348ed7fd5df2948639a64c0dd4cb3e670a9e
49481 F20101202_AAAUHA corton_m_Page_008.pro
6d0ba3d6e5d6e6e71caefed9ce79fe6d
9b4db8deab2f080b07b7aff1329dfa82d94e75c7
F20101202_AAAUGM corton_m_Page_101.tif
251f5ce16e45e78afceab7439c74c31a
3a1616904b35c7667c9c631566d7dd6719ff002e
40262 F20101202_AAAUHB corton_m_Page_018.pro
2f2cf4aa65118eb9a7a3b257d34fd0d0
cfbef3452e74022cd8d773d5f7b2c2dd4f147bc2
F20101202_AAAUGN corton_m_Page_103.tif
699a8d8801ad6eaf98380700d89b675b
d54337b4612b8c19939c162ae78cbeb158f2a4f6
F20101202_AAAUFY corton_m_Page_075.tif
e294af4298871dbc62031c6d8fcd908f
946f4c7af93e7c60d1c761b6eb58e2b4dafb1ac2
55075 F20101202_AAAUHC corton_m_Page_019.pro
ea95178a31227cecd73be2f76cf12667
16b21c1e375a3593929b0085666b9acf59ae03b6
F20101202_AAAUGO corton_m_Page_105.tif
96e0b8d809c53e883a6587fc2eade511
d054183a5773122211b438287c7767813bc84689
F20101202_AAAUFZ corton_m_Page_078.tif
982b1f729f1ed7e45e50650a9e8a98b4
e3237170773af45372ffa18574ad997540c3b23e
51006 F20101202_AAAUHD corton_m_Page_022.pro
035949284a49f09d9c3a313dc5eb073d
7c1c2c02362a32db3adb687e836df64e030eb207
F20101202_AAAUGP corton_m_Page_106.tif
7421f8939b4a149824004e5acb567ba6
84c1a773b236854b590931abe648c3fb5d822383
8807 F20101202_AAAUHE corton_m_Page_026.pro
c0c8580dac0d5aefac13c77d899a1be4
5c22315c3de3d4570b36edcd20b010ded39625d6
F20101202_AAAUGQ corton_m_Page_107.tif
0dcf19fcdc672e99ae960384b38727e2
9dbf43a0f11cb359ba4e38d9ef0959f7eee86e9a
51971 F20101202_AAAUHF corton_m_Page_027.pro
de7595f3cdaa49b272f2a0777627037d
35df757dce0a8f21185b486c5ac95634ac397303
F20101202_AAAUGR corton_m_Page_109.tif
931a61ee0e5e861b036d82083f02466a
9e197577e90f67f1673fb2b198894b045ab171b3
53629 F20101202_AAAUHG corton_m_Page_029.pro
2b377518ef866b02490a95aed0b7d87d
b720e7faae1173b27e609987671515345e516546
F20101202_AAAUGS corton_m_Page_111.tif
f310179a08a9a554950e11a8055c8647
bd13b1c6a7f182ce9ba390eea5489dce129a1f72