Citation
Telecommunications reform in Africa and the United States

Material Information

Title:
Telecommunications reform in Africa and the United States
Creator:
Hamilton, Jacqueline Marie
Publication Date:
Language:
English
Physical Description:
vi, 104 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Countries ( jstor )
Financial investments ( jstor )
Infrastructure investments ( jstor )
Main lines ( jstor )
Mathematical dependent variables ( jstor )
Mathematical variables ( jstor )
Political institutions ( jstor )
Subscriptions ( jstor )
Telecommunications ( jstor )
Telephones ( jstor )
Dissertations, Academic -- Economics -- UF ( lcsh )
Economics thesis, Ph. D ( lcsh )
City of Sanford ( local )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2002.
Bibliography:
Includes bibliographical references (leaves 99-103).
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Jacqueline Marie Hamilton.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
028812577 ( ALEPH )
50408174 ( OCLC )

Downloads

This item has the following downloads:


Full Text











TELECOMMUNICATIONS REFORM IN AFRICA AND THE UNITED STATES


By

JACQUELINE MARIE HAMILTON



















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


2002















ACKNOWLEDGEMENTS

I am frequently asked why I decided to focus my research on issues of

telecommunications development. The answer is that it all began with the chair of my committee, Dr. Sanford Berg. As a result I would like to first express my sincere appreciation to him for helping me develop an interest in an area that is so important to a country's development. Dr. Berg has gone way beyond his duties as chair of my committee. He has contributed immensely to my development as an individual as well as a researcher. I am very grateful for the role that he has played so well as my great motivator and advisor.

I have been very fortunate to have the assistance of Dr. David Figlio, Dr. Chunrong Ai and Dr. Grant Thrall as members of my committee. They have been extremely helpful at every step of the way and have contributed significantly to my development as an economist.

I would also like to thank Drs. Jonathan Hamilton, Larry Kenny and David Sappington for their helpful comments and advice throughout my time as a graduate student at the University of Florida. I was extremely lucky to be given the opportunity to study in such a welcoming and helpful environment. I am also grateful to Dr. Luis Gutidrrez, who has always been very willing to assist me, especially in the beginning stage of my dissertation.

Finally, I have to thank Derek Horrall, who has always been a rock that supports me despite his own challenges as a Ph.D. student. He has motivated and inspired me at the time when I most needed validation. Derek, along with the rest of my family, has motivated and supported me through steadfast belief in my ability. For this I am eternally grateful.


ii
















TABLE OF CONTENTS

page
A CKN O W LED G EM N TS............................................................................................................... ii

A BSTRACT .....................................................................................................................................v

CHAPTERS

I INTRO DU CTION .........................................................................................................................I

1.1 Privatization of the Telephone Industry ........................................................................1
1.2 Telecom m unications Sector Com petition ................................................................. 1
1.3 Technological Innovation........................................................................................ 2
1.4 Applications .......................................................................................................... 2

2 INSTITUTIONS, POLITICAL REGIME AND ACCESS TO TELEOMMUNICATIONS
IN FRA STRU CTU RE IN A FRICA .....................................................................................4

2.1 Introduction ...................................................................................................................4
2.2 Recent Research ........................................................................................................ 8
2.3 Hypothesis.....................................................................................................................9
2.3.1 Institutions and N etw ork A ccess............................................................... 9
2.3.2 Political System s and N etw ork A ccess .................................................... 11
2.3.3 Cultural Considerations and N etw ork A ccess......................................... 11
2.4 The State of Telecom m unications in A frica........................................................... 13
2.5 The D ata ......................................................................................................................15
2.5.1 D efining the D ata .................................................................................... 15
2.5.2 D escriptive Results.................................................................................. 17
2.6 The M odel ...................................................................................................................21
2.6.1 O m itted V ariables.................................................................................... 22
2.6.2 A lternative Specifications ...................................................................... 24
2.6.3 Country Effect and Sam ple Selectivity ................................................. 25
2.7 Estim ated Results ................................................................................................... 26
2.8 Conclusions .................................................................................................................34

3 ARE MAIN LINES AND MOBILE TELEPHONES SUBSTITUTES OR COMPLEMENTS?
EV ID EN CE FRO M A FRICA ...........................................................................................35

3.1 Introduction .................................................................................................................35
3.2 Recent Em pirical Studies ........................................................................................ 39
3.3 Hypothesis.................................................................................................... ............42
3.3.1 Mobile Competition and Main-Line performance.................................. 42
3.3.2 O ther Considerations............................................................................. 43
3.4 Background .................................................................................................................43


iii









3.4.1 Com petition Policies ............................................................................... 43
3.4.2 M obile versus Fixed Lines ...................................................................... 45
3.5 The D ata ......................................................................................................................46
3.5.1 D efining the D ata .................................................................................... 46
3.5.2 D escriptive Results.................................................................................. 48
3.6 The Em pirical M odel............................................................................................. 49
3.6.1 M odel Estim ation ................................................................................... 49
3.6.2 Sam ple Selectivity.................................................................................. 51
3.6.3 Endogeniety............................................................................................. 52
3.6.4 The Dependent V ariable Redefined ........................................................ 54
3.6.5 Estim ated Results .................................................................................... 56
3.7 Conclusions .................................................................................................................63

4. SUBSIDIZING INIVERSAL TELEPHONE SERVICE AT THE STATE LEVEL .............65
4.1 Introduction .................................................................................................................65
4.2 A Brief H istory of the Lifeline Plan ........................................................................ 67
4.3 Related Research .................................................................................................... 68
4.4 The D eterm inants of Lifeline A doption Rates ........................................................ 70
4.4.1 The Network Externality Versus the Interest Group View .....................70
4.4.2 Other Determ inants of Lifeline A doption ............................................... 73
4.5 Em pirical Specification and D ata........................................................................... 73
4.5.1 Statistical M odel of Policy A doption ...................................................... 75
4.5.2 Stationary H azard Rate........................................................................... 76
4.5.3 N on-Stationary H azard Rate ................................................................... 76
4.6 D ata Estim ation and Results.................................................................................... 77
4.7 Conclusions .................................................................................................................90

5. CON CLU D IN G REM A RK S ................................................................................................. 92

APPENDICES

A D A TA AN A LY SIS FO R CH A PTER 2 ................................................................................. 94

B DA TA A N A LY SIS FO R CH A PTER 3.................................................................................. 97

REFEREN CES...................................................................................................... .............. ... .99

BIO G RA PH ICA L SK ETCH ............................................................................................... .......104

















iv















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



TELECOMMUNICATIONS REFORM IN AFRICA AND THE UNITED STATES By

Jacqueline Marie Hamilton

May 2002

Chairperson: Dr. Sanford Berg
Major Department: Economics


My research is reported in three essays that focus on issues involving telecommunications development since the 1980s, from the perspective of developing and developed regions. In looking at issues from both viewpoints we are able to develop a broader perspective regarding specific issues involved in the improvement of telephone access. The research gives insight into the workings of telecom development and could serve to inform both private and public investment decisions.

The first essay explores particular institutional issues that exist in Africa and develops a relation between telecommunications development and institutions in that region. Specifically, this essay analyzes telephone penetration in thirty-seven African countries, by looking at the impact of the institutional environment, political conditions and legal systems on access to fixedand wire line telephones. The analysis suggests that strong institutions can promote investment in telephone infrastructure, particularly in countries that historically base their legal systems on the French civil code. In addition, a country's advancement democratically does not appear to improve telephone access in Africa. Finally, high per-capita GDP is associated with improved


v









telecommunications performance but by itself is not as strong as one might expect in comparison with the results of other cross-national studies. This essay sets the stage for further analysis regarding technological development (of cellular service) in that region.

The second essay examines the relation (substitutes or complements) between mobile and fixed-line telephone development in a region with very low main line access and the potential for rapid growth in cellular service. The relationship is modeled by accounting for reverse causality between main line and mobile phones. The results challenge the belief that cellular plays a different role in developing countries than it does in a developed region like the United States. The analysis suggests that mobile phones act as a competitive force, thereby encouraging fixedline providers to improve access.

The third essay examines telecommunications reform in the United States, where access is close to universal but where is still tremendous concern about increasing penetration as well as making access affordable for everyone. Many states have adopted policies that attempt to achieve maximum access within each state. In this essay I attempt to characterize states according to the timing of their adoption of the Lifeline plan. Together, the studies demonstrate the usefulness of economic modeling for understanding the impacts of public policy on telecommunications infrastructure.


vi















CHAPTER 1
INTRODUCTION

Reform of the telecommunications industry in developed countries such as the United States has created a basis for reforming the industry in many developing countries. During the 1980s, the view of the place of telecommunications in the market changed. Developing countries began to believe that telephone sector performance could be improved only if the sector was completely reformed. Popular avenues for reform included privatization and liberalization of the industry, technological innovation and the introduction of competition.

1.1 Privatization of the Telephone Industry

The degree of and the approach to reform of the telephone industry differed across countries, but the trend in the 1980s moved toward an ownership structure that was at least partially privatized. It was still believed that the sector was best served through a monopoly or through significant if not complete government presence. At least initially governments continued to play a significant role in the provision of the service. In many instances, full privatization was accompanied by licenses that guaranteed a long exclusivity period to large multi-national corporations.

Privatization of the telecom industry typically improved sector performance in terms of increased efficiency and access. It was, however, increasingly debated whether even better performance could be generated through sector competition.

1.2 Telecommunications Sector Competition

In an effort to continue improving sector performance, many countries committed to allow some degree of competition within the sector. Examples such as the break up of AT&T and the introduction of competition in the early 1980s suggested that the telecommunications


I






2


industry may not be a natural monopoly and that competition could further improve sector performance. In the United States, competition in the provision of telephone services was done mostly in long distance rather than local provision of wire line service. In addition there were multiple cellular service providers. For most developing countries, however, competition was primarily introduced in the cellular market.

1.3 Technological Innovation

The development of cellular technology is one of the biggest innovations in the

telecommunications industry. It quickly asserted its place within the developed world and proceeded to invade the developing world at an astounding rate. Many developing countries saw cellular usage as a possible solution to the problem of low access rates. As the technology continues to improve and the cost of its provision falls, cellular emerges as a very important avenue for increasing telephone subscription rapidly. While sector reform has resulted in a noticeable increase in telephone access, the biggest growth is in cellular usage, and the prospect for increased telephone access in the developing world seems less daunting.

1.4 Applications

This dissertation examines issues of reform in both developed and developing countries. Typically, privatization initiatives as well as competition and technological advancement have been introduced in both developed and developing countries at different paces and at different times. Many of the achievements, experiences and expertise within developed countries have been offered as examples of how sector reform may be implemented. Experience across developing countries is also shared as part of the ongoing attempt to improve the provision of telecommunications provision.

With this in mind, I examine different issues of reform chapters 2-4 of this dissertation, in both developing and developed countries. In chapter 2, I provide a detailed analysis of the institutional, economic, political, cultural and demographic environment for telephone investment






3


for thirty-seven African countries. The results indicate that significant reform of the institutional environment is essential for any significant improvement in sector performance.

Chapter 3 highlights cellular technology as one of the most important developments in the telecom sector. It assesses the role of cellular phones within the context of extremely low telephone connection rates in twenty-three African countries. I tested the hypothesis that cellular phones and mainline telephones are substitutes in consumption. I found the hypothesis to be true only after achieving a critical cellular subscription rate.

Chapter 4 continues to assess the implication of telecom reform by focusing on policy

reforms within the context of a developed country. In this chapter I used state-level data from the United States to characterize states based on the timing of their adoption of a low-income program called the Lifeline plan.

Chapter 5 presents the general conclusions.















CHAPTER 2
INSTITUTIONS, POLITICAL REGIME AND ACCESS TO TELECOMMUNICATIONS INFRASTRUCTURE IN AFRICA

2.1 Introduction

Empirical evidence regarding the effect of institutions and political regime exists, but little is empirically established for developing economies in Africa.' Most analyses of telecommunications reform in Africa are found in case studies and country reports, such as those by Chance, Booz, Allen and Hamilton, Inc. (1998), Laidlaw and Parkinson (1995), Frempong and Atubra (2001) and Onwumechili (2001). Policy inferences for the African telecommunications sector have relied on research done in other developing regions, such as Latin America and the Caribbean,2 and conclusions are sometimes drawn from studies that lump Africa with countries from dissimilar regions.

A study specific to Africa is warranted because of the continent's diverse cultural, ethnic, economic and political institutions, its persistent internal conflicts and its lack of development relative to other regions. By considering Africa alone, we reduce the chance of drawing conclusions that might be true in one developing region but not another. It is also easier to account for heterogeneity across countries in a single region than for countries in different regions.



1 I would like to thank Chunrong Ai, Sanford Berg, David Figlio, Grant Thrall, Jonathan Hamilton, David Sappington, Larry Kenny and Cezley Sampson for helpful comments on the earlier version of this study. Remaining errors of fact or interpretation are the author's responsibility.
2 Some cross-national studies of developing areas, such as Singh (2000), Gutierrez and Berg (2000) and Wallsten (2001), either do not include data on Africa or they lump Africa with other emerging economies.


4






5


Research such as that by Sachs and Warner (1997) concludes that the slower growth

observed in Africa is not explained any differently from growth in other regions. More recently, however, Block (2001) found important differences in Africa compared to other countries. For instance, he found that lack of openness to trade affects growth more adversely in Africa than in other low- and middle-income countries. By extension, this finding supports the circumscribed study of Africa, so that policy prescriptions may be based on relations that are particularly applicable to Africa, as opposed to generic prescriptions that may not be appropriate.

Cross-national studies on African telecommunications development have been difficult because of data limitations, since serious reform began only in the mid- to late 1980s. With data from thirty-seven African countries during 1985-91, this study helps fill a gap in research on Africa by using panel data techniques to perform a cross-national evaluation of the effects of institutions, political systems and cultural issues (proxied by legal systems) on telephone penetration rates.

The results from the empirical analysis suggest that high per capita GDP is associated with improved telephone penetration but, by itself, is not as strong as one might expect on the basis of other cross-national studies. In addition, a strong institutional framework can enhance investment in telephone infrastructure. Such a framework involves a respect for property rights, which yields perceptions of contractual security and reduced likelihood of expropriation. Contrary to expectations, countries with similar institutional quality are likely to have higher access to telephones under the French legal system rather than the English common law code. Finally, a more democratic country is likely to have lower access to a telephone network than a less democratic country with similar characteristics.

The results are important, both from the perspective of the private investor as well as the public operator. Consideration of demography and affordability on penetration rates is essential for the analysis of the cost effectiveness and potential profitability of investing in the region. From a policy perspective, it is important for the decision maker to understand the relation






6


between the institutional environment, polity, culture and telecom investment in order to implement policies that create an environment conducive to telecommunications investment. The environment would include a regulatory system that benefits both providers and users of basic telephone service.

Classifying the institutional and political issues sets the background for an exploration of infrastructure development in Africa from a more technical perspective. Such issues include the role of growing mobile access in countries with relatively low penetration rates. I analyzed this issue in Hamilton (2002) by modeling the endogeniety between main line and mobile telephones.

Despite the influx of new telecommunications services like mobile technology, voice mail, call waiting, as well as increased access to fixed-line telephony, access to basic telecommunications' in Africa is still very limited (Kerf and Smith, 1996). Investment in basic telephony in Africa is far below the level in Latin America, the Caribbean and Asian Pacific regions. In 1985, Africa's penetration rate was, on average, 10.45 per 1,000 people, more than five times lower than that of Latin America and the Caribbean. By 1997, despite a growth rate of 147 percent, Africa's access rate remained more than five times lower than in Latin America and the Caribbean.

Teledensity in Africa has been low partly because of relatively low income levels.

Between 1980 and 1990, Africa suffered a decline in its already low income levels. Between 1985 and 1997, per capita income in Africa was, on average, US $915, compared to more than US $2,000 in Latin America and the Caribbean and US $5,000 in Asia. Although low, teledensity has been increasing in Africa and reached 25.82 telephones per 1,000 people in 1997. This average (unweighted by population) is skewed upward since many countries have much weaker performance. If Seychelles, with an access rate of 203.64, Mauritius with 227.65 and



3 Basic telecommunications, telephony, telephone access, penetration and network access refer to both mobile and main line telephones unless otherwise specified.






7


15.66 in 1997. So despite recent innovations in the sector and general economic growth, access to telecommunications in Africa is still limited. As in most developing regions, telephone lines are concentrated in the cities, with only limited access for rural areas. Nevertheless, despite the poor quality of infrastructure compared to other developing countries, the opportunities for telecommunications development in Africa are substantial.

While there has been a tremendous amount of investment and reform elsewhere in the world, Africa was largely ignored until recently. The sector is currently predominantly stateowned, but some governments have embarked on reform programs, most of which involve two elements: gradual commercialization by separating operational management from government ministries and the transfer of responsibility for regulation away from government ministries to independent agencies. Privatization options being considered include public offers for sale to financial institutions, sale to private investors and employees, private sale to strategic investors, or divestiture and management contracts with foreign operators.

The change occurring in the region is often obscured by political constraints that limit a government's desire or ability to make policy commitments to promote the development of the sector (Mustafa et al. 1997). Analysts like Kerf and Smith (1996) argue that special attention must be given to establishing stable and independent regulatory agencies that can provide credibility for investors, legitimacy for consumers, and more efficient sector performance. The creation of suitable regulatory systems is important because the success of the restructuring process depends heavily on the credibility and consistency of that reform. After surveying recent studies, I outline several hypotheses and then briefly review the state of telecommunications in Africa. The basic data are examined in Section 5, followed by the model development and estimation. The concluding section summarizes the results.






8


2.2 Recent Research

Research concerning the influence of political stability on investments has a long history. Three decades ago, Bennett and Green (1972) sought to identify the role of politics by testing the hypothesis of a negative relation between the allocation of U. S. marketing investment throughout the world and the level of national political instability in different countries. They also tested whether such a negative relation existed only in less developed countries. They found that political instability does not discourage investment in marketing activities. If this finding were applicable to infrastructure investments, then political factors could be viewed as playing only a minor role in utility sectors; however, irreversible fixed costs makes these capital-intensive sectors sensitive to political instability.

Studies of infrastructure since 1972 have found that political conditions do, in fact,

influence the level of investment that takes place abroad (Haan and Siermaan 1995, Bergara et al. 1998, Henisz 2000, Svensson 1998). Research has shown that (telecommunication) investments will be larger when there are strong political constraints on government officials.4

Knack and Keefer (1995,1997), Guiterrez and Berg (2000), Henisz (2000) and Singh (2000) are among studies finding a relation between the institutional environment, economic growth and infrastructure investment. They have shown that the greater the policy



4The framework for utility investment is more consistent and predictable when there are more independent checks on government's executive power. According to Svensson (1998), political factors affect infrastructure development, but the effect is indirect in that the political environment merely provides the channel through which private investment flows. He found that private investment is restricted when those in power lack incentives to undertake legal reforms that protect property rights and encourage investment. Weak property rights lead to the reallocation of resources away from taxable activities, which reduces future governments' tax revenues and ability to spend. In not reforming the legal system, the current government affects future governments. By protecting its current constituency, the government neglects the provision of a favorable environment for investment. Haan and Siermaan (1995) stressed that, although the relation between democracy and economic growth is not robust, high levels of economic growth cannot take place in an environment where democratic rights are repressed. Since investment contributes significantly to the rate of economic growth, the same relation might be expected between investment and democracy.






9


uncertainty, the smaller the level of investment. Among poorer countries, those with stronger institutional bodies that enforce law and protect property rights have better prospects for private investment and increased payoff for public infrastructure investment.

Considerations other than political and institutional ones have been identified as

determinates of infrastructure investment, including economic indicators that capture the standard of living and other features of the economy. Higher income levels make foreign investment more attractive, as does expected growth and predictability of that growth. Alesina and Perotti (1996) looked at the more complicated question of how the distribution of income may determine the level of investment. Using data from 1960 to 1985, they found that income inequality increases sociopolitical instability by fueling discontent. This in turn creates uncertainty in the political and economic environment, which reduces investment.

Applying the "catch up" theory of growth, Antonelli (1993) provided an explanation for the ability of some poor countries to catch up. By taking advantage of technology in advanced countries, latecomers can experience rapid growth in telecommunications development. He concluded that new investment plays a significant role in the diffusion of advanced telecommunications in countries with high rates of growth in GDP and telecommunications infrastructure.

2.3 Hypotheses

2.3.1 Institutions and Network Access

A positive correlation is expected between network access and institutions. Institutions are the formal and informal rules that guide human interactions, whether they are social, political or economic (North 1990). They reduce uncertainty by establishing a stable framework for human relationships. Institutions involve not only rules but also enforcement, which usually involve the state. Investment is likely to be encouraged in an environment where participants understand the rules of the game and where the risk of losses is minimized (North 1986). Rules






10


and regulations that guide the telecommunication sector protect both the private operator and the public manager. Private investors appreciate strong institutions that discourage governments from reneging on promises. Investment in telecommunications involves the commitment of large sunk costs, and private investors are exposed to the risk of expropriation of their property (Levy and Spiller 1994).

The same point applies to government investment. Ministries responsible for investment are less likely to be able to raise capital (from the national budget or through the issuance of project-specific bonds) if political instability is high. Why go to all the effort of planning and coordinating major new initiatives when the next government might reverse the process or give the construction contract to a political supporter? Shifting patronage and civil strife do not provide a firm foundation for long-term government investment. Furthermore, if prices are below cost, additional subscribers just mean a larger deficit. Theories of political economy suggest that the political and institutional environment in Africa has not been conducive to telecom investment (Goldsmith 2001).

Governments that tend to be unstable characterize Africa for the most part, and

government policies are often unstable as well. South Africa is an example of how stronger institutions can affect telecommunications investment. Since 1990 and the establishment of political freedom and stability in South Africa, many parts of the nation that previously had minimal access to telephones now have relatively good access. The state-owned operator is committed to privatization of the industry and to increased access for citizens (International Telecommunications Union [ITU] 1998).

The transformation in South Africa highlights the interdependence of institutions and organizations. As discussed by North (1990), economies perform differently over time not just because of institutions and organizations themselves, but also because of their interactions, which determine the direction of institutional change. A significant improvement in the development of Africa's telecom infrastructure is therefore conditional on strong credible political institutions






I1


that protect against illegitimate rent seeking and create safe and healthy environments for investment.

2.3.2 Political Systems and Network Access

To control for the effect of political structures and party system, I introduce a proxy for democracy taken from the Polity III Democracy Index5 described in Section 5. Theory suggests a direct relationship between democracy and investment. According to this view, the democratic process and the existence of civil liberties generate conditions most suitable to economic development. In a democracy, political actions are accountable to the press and the public; therefore, a government's ability to manipulate policies to suit political ends may be limited (Haan and Siermann 1995). This argument is particularly applicable to investments in telecommunications, which cannot be easily redeployed. In countries with a non-democratic system of government, the risk associated with investment is high unless the government can make a credible commitment not to expropriate capital assets or unduly limit returns.' Over the sample period, state-owned firms provided most telecommunication services in Africa.

2.3.3 Cultural Considerations and Network Access

Although people and regions in Africa are diverse, a common history of colonization is unifying. Colonization remains a legacy in the form of cultural, social and political influences of foreign powers that help to define these countries today. Often the laws and norms of a country draw on those of the former colonial power. According to cultural theories of institutions, a society's behavior, its actions and governments are shaped by its beliefs and shared values.



5 The democracy score of the Polity III index is based on five components: The competitiveness of political participation, weighted 0.3; the regulation of political participation, competitiveness and openness of executive recruitment, each weighted 0.1; and constraints on the chief executive, weighted 0.4 (Jaggers and Gurr 1995).
6 Goldsmith (2001) found that African countries with less political risk tend to have more open governments and low levels of political corruption. This results in better functioning of the democratic process and encourages leaders to pursue policies that are less shortsighted.






12


The efficient functioning of a society requires that individuals (economic agents) believe that their institutions are credible and efficient (La Porta et al. 1999). One aspect of the cultural norms of a country is its legal system. Countries that have an interventionist legal system restrict the ability of investors to buy, sell and engage in efficient contracting, which affects risks and the return on investments. The view of the new institutional economics supports an economic environment in which the government is relatively non-interventionist, providing a promising environment for investment (Knack and Keefer 1995).

Using these theories, I hypothesize that the impact of the institutional variables in the

model will vary, depending on the legal system in each country. With this in mind, the data were divided according to legal code. 7 Incorporating the legal system as a variable allows the countries in the sample to be divided according to the traditional tendencies of the government regarding intervention. LaPorta et al. (1999) argue that French civil law tends to be used by the state as a means of expanding its power and offers relatively less protection for individuals than British common law. Governments that utilize the French civil code are therefore expected to be more reluctant to reduce their control in the telecom sector, which may be regarded as a key area that has to be guarded closely.8 Because British common law traditionally offers more protection for individuals and limits the powers of governments, a country with French legal traditions is expected to have lower access to telephones inasmuch as control hampers investment and lowers the efficiency of sector performance.



Specifically, I divide the countries into three groups: (1) countries that utilize the English common law legal system, (2) those that utilize the French civil code and (3) all other countries.

8 So far, there is no evidence that countries with French civil legal systems are less likely to privatize their telecommunications network. Up to 1997 the number of countries with a tradition of French civil legal background that had privatized (or promised to privatize) exceeded those of English common law background by one country.






13


2.4 The State of Telecommunications in Africa

Tables 2.1 and 2.2 trace the development of telecommunications in Africa from 1985 to 1997. Many African countries are now concerned with telecom infrastructure for the same reasons that they limited foreign and private ownership in the past. Telecom investment is strategic and can contribute to a country's economic development. Most African countries, however, still require a substantial increase in investment in telecom infrastructure to even catch up with other developing regions.

Table 2.1: Network Access. Africa's Place Relative to Other Developing Regions (19851997)


Region 1985 1991 1997
Africa 10.4 15.0 25.8
Latin America and the 51.4 68.2 135.8
Caribbean
Asia 83.5 111.6 234.7

Table 2.2: Growth in Network Access: Africa Relative to LAC and Asia (1985-1997) Region 1985 1991 1997
Africa 4.6 (44%) 10.0 (63%) 14.6 (145%)
Latin America and the 16.8 (33%) 61.2 (82%) 84.4 (146%)
Caribbean
Asia 28.1 (34%) 111.0 (90%) 151.2 (181%)


Note: The first number in each cell represents growth in telephone access per 1,000 individuals; the percentage change is in parentheses.

Over the period 1985-97, Africa consistently displays lower network access than Asia and Latin America and the Caribbean. In 1997, access in Africa was more than five times less than in Latin America and the Caribbean and more than nine times less than in Asia. Despite growth at an increasing rate over the years, Africa has failed to catch up with Latin America and the Caribbean and Asia, where access has also grown. However, the rate of growth indicates that Africa has recognized the importance of developing its telecom sector with both mobile and wire line provision. As part of their commitment to increase telephone access in Africa, some countries have agreed to allow some degree of private participation in wire line provision. Table 2.3 reflects this change, showing that it is no longer a foregone conclusion in Africa that government must operate telecommunications in order to meet national objectives. An asterisk indicates those






14


countries that have made a commitment to privatize even though ownership remains with the

state thus far.9

Table 2.3 Ownership of Wire Line Service in Africa
Country Ownership Ownership
(Amount and year privatized) (Amount and year privatized)
Algeria 100% state-owned Morocco* 100% state-owned
Botswana* 100% state-owned Niger* 100% state-owned
Cameroon 100% state-owned Nigeria 100% state-owned
Congo* 100% state-owned Rwanda 100% state-owned
Cote d'Ivoire Privatized (51%, 1997) Sierra Leone 100% state-owned Egypt* 100% state-owned South Africa Privatized (30%, 1997)
Gabon Privatized (39% by 1998) Tanzania Privatized (_ by 1998)
Ghana Privatized (30%, 1996) Togo 100% state-owned
Kenya* 100% state-owned Tunisia 100% state-owned
Madagascar Privatized (34%, 1995) Uganda* 100% state-owned
Malawi 100% state-owned Zambia* 100% state-owned
Mali* 100% state-owned Zimbabwe* 100% state-owned
* indicates countries that have committed to privatize in the near future. Source: Leblanch-Woher and Lewington (2000).

Many socialized entities have not performed well. The extent of investment required in

developing countries is usually too large and expensive for the government to manage on its own,

which is one reason infrastructure (telecommunications) development in Africa has been slow

and sometimes nonexistent. Indeed, there is evidence of strong performance following

privatization in many developing nations, including those in Latin America and the Caribbean

(Guiterrez and Berg 2000).

Once privatization is accomplished, modernization and development can increase the

efficiency and availability of service. Although telecommunication infrastructure is largely stateowned in Africa, some countries are in the process of reforming the sector, following world

trends as well as demands by international lending agencies. Joint ventures are a typical first

step. The extent to which governments commit to privatization and sector development depends



' Even those countries reluctant to give up control of wire line provision have allowed some degree of private participation in cellular provision. For more on cellular privatization and competition, see Hamilton (2002), which analyzes the role of mobile competition in the development of fixed line telephony.






15


on economic and political considerations, as well as the risk environment of the country. The

increased commitment to private investment shown in Table 2.3 indicates that African countries

have come to accept that there are benefits from privatization. Some of these include increased

services and quality of service, as well as improved access at lower cost and the availability of

additional capital and management skills.10 Even when the privatized entities are essentially

monopolies, consumers can benefit from reduced prices with proper regulation of monopolistic

power (Noll 1999). The threat of potential competition may be enough to induce telecom

investment.

2.5 The Data

2.5.1 Defining the Data

The 1985-97 study period was determined by constraints on the data available for some

of the variables. Definitions of variables and sources of the data are given in Table 2.4.

Table 2.4: Definitions and Sources
Variable Definition Source
NETWORK Is the dependent variable. It is the (# of main World Bank's World
telephone lines/population)* 1,000 (# of cellular Development
subscriptions/population)* 1,000. Indicators (1999).
Economic and demographic variables: LOGGDPC The natural logarithm of a country's per capita World Bank's World
gross domestic product (GDP) lagged one Development
period. The base year is 1990. Indicators (1999).
LTRADE Imports plus exports as a fraction of GDP World Bank's World
lagged one period. Development
Indicators (1999).
URBAN Urban population/total population. That World Bank's World
is the percentage of the total population that Development
resides in urban areas. Indicators (1999).


'0 For instance, Ghana has experienced noticeable improvements in telecom penetration since liberalization and privatization in the mid- to late 1990s. The number of direct lines increased by 26% in 1997 alone, compared to growth of 2-3 %in the early 1990s (Frempong and Atubra 2001).






16


Table 2.4 Continued
Variable I Definition Source
Institutional variables:
CORRUP Corruption within the political system on a IRIS-3 file of International
scale of 0 to 6. Larger values indicate less Country Risk Guide (ICRG)
corruption. data, 1982-97, constructed
by Stephen Knack and the
IRIS Center, University of
Maryland.
LAW Index of law-and-order tradition. It measures IRIS-3 file of International
the degree to which citizens of a country are Country Risk Guide (ICRG)
willing to support established institutions to data, 1982-97, constructed make and implement laws and adjudicate by Stephen Knack and the disputes. The index ranges from 0 to 6. A IRIS Center, University of
high point indicates a strong law-and-order Maryland.
tradition.
BUREAU Index of bureaucratic quality on a scale IRIS-3 file of International
ranging from 0 to 6. It measures the extent to Country Risk Guide (ICRG)
which a country's bureaucracy is able to data, 1982-97, constructed govern without drastic changes in policy, or by Stephen Knack and the interruption in government services. High IRIS Center, University of
index values indicate a strong bureaucracy. Maryland.

CONTRACT The risk to foreign businesses, contractors IRIS-3 file of International
and consultants that government will modify Country Risk Guide (ICRG)
a contract in the form of repudiation, data, 1982-97, constructed postponement or scaling down. The index by Stephen Knack and the ranges from 0 to 10. A high point index IRIS Center, University of
signifies less likelihood that a contract will Maryland.
be modified.
EXPROP The risk of expropriation of private IRIS-3 file of International
investments in terms of outright confiscation Country Risk Guide (ICRG)
or forced nationalization. The index is rated data, 1982-97, constructed from 0 to 10. Higher index points signify by Stephen Knack and the less likelihood of investment expropriation. IRIS Center, University of Maryland.

ICRG Following Knack and Keefer (1995), I IRIS-3 file of International
created a 50-point index from five of the Country Risk Guide (ICRG)
ICRG variables. CORRUP, LAW and data, 1982-97, constructed
BUREAU were converted to a 10-point scale by Stephen Knack and the by multiplying each by 5/3. These were IRIS Center, University of
summed with CONTRACT and EXPROP. Maryland.

Variables indic ating party system (freedom measures) and legal l gacy
DEMOC Democracy, an indicator of regime type on a Polity III: Regime type and
scale of 0 to 10. This data is assumed to be political authority, 1800-1994,
constant between 1994 and 1997. Jaggers and Gurr (1996).






17


Table 2.4 Continued
Variable Definition Source
DEMOC- A single summary measure of political Polity III: Regime type and
AUTO regime, this variable is calculated by political authority, 1800-1994,
subtracting the index of autocracy score from Jaggers and Gurr (1996).
the democracy index score. It is measured
on a scale of -10 to 10, where, 0 to -10
indicates autocratic regimes and 1-10 indicates
democracies.
POLRIGHT From the Freedom in the World Survey, From the Freedom in the
1998-99. POLRIGHT is an index that World Survey, 1998-99.
measures the extent to which people
participate freely in the political process or the rights of all adults to vote and compete
for public office. The index is measured on a
scale of I to 7. Smaller numbers represent
greater freedom.
ENGINST ENG*ICRG, where ENG is a dummy Legal system is taken from
variable equal to 1 if the country has an CIA World Factbook, 1999.
English common law tradition and 0
otherwise.
OINST OTHER*ICRG, where OTHER is a dummy Legal system is taken from
variable equal to 1 if the country does not CIA World Factbook, 1999.
utilize English or French legal systems and 0
otherwise.
ENGPOL ENG*DEMOC-AUTO. Legal system is taken from
CIA World Factbook, 1999.
OPOL OTHER*DEMOC-AUTO. Legal system is taken from
CIA World Factbook, 1999.

2.5.2 Descriptive Results

Table 2.5 summarizes the variables used in the models. The differences in means vary

across the samples. Confirming the hypothesis that countries with a history of the French civil

code tend to be more interventionist, potentially restricting the ease of investments, indicators of

political freedom are stronger when the legal system is based on English rather than French law.

The negative sign on the difference in POLRIGHT indicates stronger political freedom in the

English nations, which overall tend to have a stronger institutional framework (based on ICRG),

but not significantly so. Of the ICRG measures, only LAW and BUREAU have significantly


different means across the sub-samples.






18


Table 2.5 Sample Descriptive Statistics, 1985-97
Variable Mean and Mean and Mean and Difference in
Standard Standard Standard Means Across
Deviation. Full Deviation. Deviation. Legal Systems
Sample (481) English French Civil
Common Law Code Countries
Countries (169) (182)
NETWORK 13.573 14.669 12.150 2.519
(20.128) (25.294) (15.203)
LTRADE 60.501 56.251 61.088 -4.837***
(27.373) (25.124) (23.828)
URBAN 34.276 30.129 38.733 -8.604*
(15.720) (10.730) (16.491)
CORRUP 2.827 2.914 2.928 -0.014
(1.105) (1.010) (0.851)
LAW 2.679 2.764 2.796 -0.032
(1.168) (1.070) (0.877)
BUREAU 2.629 2.715 2.758 -0.043
(1.118) (1.269) (0.939)
CONTRACT 5.098 5.342 5.360 -0.018
(1.784) (1.756) (1.541)
EXPROP 6.081 6.491 6.227 0.264
(1.901) (1.781) (1.676)
ICRG 24.737 25.822 25.725 0.097
(7.171) (6.922) (4.851)
POLRIGHT 5.325 5.231 5.335 -0.104
(1.577) (1.516) (1.368)
DEMOC 0.353 1.982 1.115 0.867
(10.812) (3.123) (2.180)
DEMOC- -2.933 -2.663 -3.972 1.309***
SAUTO (5.429) (5.805) (4.489)


Standard deviations are in parentheses; sinfcn t1,* = sinfctat5,* significant at 10%.

Mean network access (unweighted by population) is only 1.2 times greater in nations

utilizing English common law, which is not statistically different from the French countries. The results in Table 2.5 suggest that the difference in institutional factors across regions is not enough to create a difference in telephone access. LOGGDPC, LTRADE and URBAN are all significantly higher in nations utilizing the French civil code. Taking this into consideration along with the information on institutions, I view the raw data as suggesting that countries with stronger institutions tend to have higher access to telephones, even when their economies are less open and they have lower per capita income and a relatively small urban population. The






19


difference in institutional quality (ICRG) is, however, not big enough to create any significant differences in network access.

Table 2.6 provides summary statistics based on differences in legal tradition and

institutional quality. An ICRG index above the sample average of 24.74 defines high institutional quality, while countries scoring below this average are considered to have low institutional quality. The numbers in parentheses indicate percentage growth in network access between 1985 and 1997.

Table 2.6: Regional Differences, Institutional Quality and Network Access. 1985-1997 Legal system Institutional Quality
High Low Overall
English Common Law 21.30, (128) 8.01, (44) 14.66, (118)
English minus South Africa 10.99, (153) 8.01, (44) 8.09, (126)
French Civil Code 13.45, (172) 9.94, (83) 12.20, (146)
French minus Arab Nations 7.19, (134) 3.82, (63) 5.93, (124)
All Countries 16.99, (133) 5.49, (78) 13.57, (112)

As expected, countries with stronger institutions on average have higher access to

telephones. In this sample, countries with high institutional quality have 3.7 times greater access than countries with low ICRG scores. This result conforms to expectations, unlike the observations across English and French legal origins. According to Chong and Zanforlin (2000), countries with a system based on the French civil code appear to have lower institutional quality. This suggests that countries of French origin in our sample should have lower access to telephones. However, the data in Table 2.6 suggest that this is not true for the sample of countries used in this analysis when institutional quality is low.

Growth in telephone access is also higher in French nations, even when countries of English common law origin have higher penetration levels. The implication of Chong and Zanforlin (2000) can be extended only when institutional quality is high. In this case, countries with an English common law tradition have higher network access than countries using the French civil code. This suggests that legal origins may influence institutional quality, but the legal system itself has less of an influence on telephone access than institutional quality.






20


Countries with strong institutions have higher access than those with weaker institutions, regardless of their legal system.

The difference in access across legal origins is at most 9 per 1,000 individuals when

institutional quality is either high or low, but the difference across institutional qualities is a little over 11 per 1,000. When South Africa is excluded, the growth rate in the English common law countries increased so that the difference between the two groups of countries was negligible. This suggests that the countries with low access, relative to South Africa, will eventually achieve comparable access levels.

Table 2.7 compares access on the basis of differences in countries' legal tradition and polity. The entire sample period, 1985-97 is used when necessary, but the last five years, 199297, is used when this better reflects the political tendencies of a country, since many countries that were autocratic in earlier years are adjusting to a polity with more democratic characteristics. In addition, a significant amount of the development in telecom happened during the last five years of the sample period.

Following Jaggers and Gurr (1995), countries are defined as democratic if the autocratic index subtracted from the democratic index is greater than or equal to 1. Otherwise, the political system is defined as autocratic. Countries scoring 1-6 are primarily democratic but put some limits on political participation and civil liberties (e.g., Zambia and Ghana). Scores above 6 represent highly democratic nations, such as Botswana, Madagascar, and South Africa. Countries scoring 0 to -6 are autocratic in nature but allow some degree of political freedom (e.g., Egypt and Zimbabwe) while those scoring -7 and lower are fully autocratic (e.g., Cameroon). Many African countries fall in this penultimate category.

In general, democratic nations tend to have higher access than non-democratic ones, but the difference in growth is very small. This is attributable entirely to countries using the English common law code.






21


Table 2.7: Regional Differences, Political Systems and Network Access

Legal Legacy Political System
Democratic Non-Democratic All
English Common Law 24.23, (127) 8.67, (104) 14.66, (118)
English without South Africa 6.91, (198) 8.67, (104) 8.09, (126)
French Civil Code 10.62, (105) 12.76, (143) 12.20, (146)
French without Arab Nations 4.55, (130) 6.41, (81) 5.93, (124)
All Countries 13.93, (126) 10.83, (129) 13.57, (112)


Note: Numbers in parentheses are percentage growth in network access between 1985 and 1997.

The slightly higher growth in non-democratic countries results from a higher growth rate in nations using the French civil code. When the government is democratic, countries under an English legal system have significantly higher access than those under the French system. The opposite is true when the political system is not democratic. One implication of this is that countries under the French civil code may have less experience working under democratic systems. Indeed, leaders can make more efficient decisions regarding telecom development under a system that is more authoritarian, and the benefits from investing in telecommunications are more likely to accrue to the politician (Goldsmith 2001). The role of legal systems is further explored in the estimation of the model.


2.6 The Model

NETWORK (per 1,000 people) is the dependent variable used to estimate the effects of institutions, political regime and legal origin on investment in basic telecommunications in Africa. Although data on dollar investment in telecommunications is sometimes directly available, I decided to use telephone lines per capita as the dependent variable because per-capita dollar investment can be a misleading indicator of the degree of accessibility. For example, a $100 investment in a country with rugged terrain is likely to generate fewer telephones than in a country where it is easier to lay telephone cables. Since the data set consists of countries with different (and often unique) characteristics, this measure would be more useful if the research focus was on cost effectiveness rather than on access. NETWORK per capita is therefore a more suitable measure.






22


Because the data used in this study spans the period 1985-97, there are multiple

observations on each country and the explanatory variables are time-series related. That is, they vary over time. To take account of this time-series component while maintaining the crosssection, both components are pooled to develop a data set consisting of 37 countries over the period 1985-97. The advantage of pooling the data is that it generates a larger number of data points, which increases the degrees of freedom and may reduce colinearity among independent variables (Hsiao 1986). Cross-sectional estimation provides information relevant to a single time period, but this study is more concerned with cross-variation over time. By pooling the data and applying panel data techniques, I am able to make references about cross-variation of variables over time. Equations 1 is specified to accommodate the pooled data as follows:

(1) NETWORKCT =a+ 8K XCT +6CT ,

Where XCT is a vector of independent variables for country C in year T. These variables are LOGGDPC, LTRADE, URBAN, CORRUP, LAW, BUREAU, CONTRACT, EXPROP, and DEMOC. All these variables are defined in Table 3. F is a random error term.

2.6.1 Omitted Variables

Since the data represent a panel of different countries, it is highly likely that unobserved differences across countries will affect telephone investment. If these unobserved variables are not accounted for, the interpretation and use of the estimated equation may be unreliable (Studemund 1992). Omitted variables will cause a bias in the estimated coefficients of the other explanatory variables. For example, the coefficient of LTRADE in Equation 1 represents the change in NETWORK caused by a one-unit change in LTRADE, when the values of all other right-hand-side variables are held constant. If a variable is omitted, it is not included as an independent variable and is not held constant for the calculation and interpretation of the coefficient of LTRADEcT. Thus, some of the variation in the omitted variable will be incorrectly attributed to LTRADE. The efficiency and experience of multi-national operators assisting the






23


local companies is an example of omitted variables. A more experienced operator is likely to provide better training to managers, technicians and engineers as reform of the sector takes place (Henisz and Delios 2002).

Privatization and regulatory reform are other variables that could be included in the

model. Privatization of wire line provision, however, started only in the mid- to late 1990s for the countries that have committed to privatize. Agencies dedicated to the regulation of telecommunications in the context of privatizations were only in the formatory stages for most countries toward the end of the sample period. As a result, enough observations of privatization and regulatory reform did not exist throughout the sample period. A country-specific constant ac is introduced to account for the effects of those omitted variables that are specific to individual countries.

It is also possible for effects that differ across countries to exhibit variation through time. For example, telephone penetration in each country may be affected by the unobservable worker quality (ability), which may vary from country to country. Countries with a more efficient or able workforce or provider are better able to make telephones accessible than less efficient ones. Worker ability is also likely to vary through time. Worker quality may improve through education, for instance, or through learning by doing. The trend through time is thus likely to affect penetration. Year effects (OT) are introduced to account for such trends through time. The new equation is:

(2) NETWORKCT = ac + JK XCT +'T +CT,

The fixed-effect model is estimated by using deviations from the group mean. In so doing, Equation 2 will pick up any constant differences occurring on a country-specific level. Finally, the standard errors are adjusted to account for the systematic correlation across units in the same group.






24


2.6.2 Alternative Specifications

Equation 2 is estimated by using two specifications of institutional variables and three alternatives to measure political freedom in order to check the sensitivity of the results to alternative specifications. Equation 2 is first estimated with the components of the ICRG index (CORRUP, LAW, BUREAU, CONTRACT, EXPROP) as individual independent variables. They are then treated as a single variable by using the average of the five components and thus aggregating the index to avoid possible problems of correlation between the individual components. The aggregate ICRG index, however, assumes that some factors that affect the institutional environment are more important than others, and so must be weighted more heavily. The problem with this approach is that different conditions across countries may warrant component weights that vary from the ones imposed by the index. The implicit weights may also vary through time, so the aggregate ICRG index may contain considerable measurement error. Using the individual components of the index provides a check against potential biases that may result from the use of subjective component weighting in the aggregated index.

Equation 2 is also estimated with three different political variables. First, DEMOC is used to capture the degree of political freedom enjoyed by a country. This is then replaced with POLRIGHT so results can be compared to check whether the measure of freedom used makes a difference. While the two sets of variables are constructed to measure the freedom of a country, there may be noticeable differences in what is captured in reality. The POLRIGHT index does not rate governments, but rather the rights and freedoms enjoyed by individuals in each country. The DEMOC index rates governments by measuring factors such as constraints on the chief executive and the openness of executive recruitment. Differences in the construction of the two measures of freedom make it likely that the economic freedom measure contains some information that is not in POLRIGHT.

DEMOC-AUTO is the third measure of political freedom. The first measure (DEMOC) is rated on a scale of 0 -10, where 0 denotes limited democracy and 10 denotes a high level of






25


democracy. In this use of the democracy index, a low level of democracy implies high autocracy. Yet, if low democracy correlated perfectly with high autocracy, then Ghana, with a democracy score of 3, should have autocracy score of around 7, rather than 2. This kind of separate measurement of democracy (DEMOC) and autocracy (AUTO) is difficult to interpret when democracy is not a close inverse of autocracy, so I follow the Jaggers and Gurr (1995) approach and adopt a single summary measure of political regime (DEMOC-AUTO) to replace DEMOC.

Finally, to explore the argument that the legal tradition of a country may affect telephone development through its institutional framework, I introduce four new variables: ENGINST, OINST, ENGPOL and OPOL. These are defined in Table 3. Equation 2 is re-estimated using these variables in all the specifications discussed above.

2.6.3 Country Effect and Sample Selectivity

Because the sample consists of only 37 of the 55 countries in Africa, the results may not be representative of the 18 excluded countries. To check for sample selectivity, I compared the countries in the sample with those excluded on observable factors such as network access and per capita GDP. Table A3 provides a summary of the comparison. Noticeable difference exist between the two groups of countries in network access per capita GDP as well as trade. The first two columns of Table A3 suggest systematic differences between the countries.

It is also possible that some countries in the sample individually influence the results in a way that may not be true for other countries in the sample. For instance, South Africa stands apart from other countries in the sample because of its comparatively developed telecommunications sector (CIA World Factbook 1999). South Africa has the highest telephone penetration in the sample, and arguably sets the pace for sector development in the region. South Africa was among the first African nations to start the process of telecommunications reform in terms of modernizing and expanding services, introducing private participation and setting up regulatory bodies. South Africa has received much attention from the rest of the world as an example for other African countries. If other African nations view South Africa in a similar






26


manner, then South Africa may be influencing the rest of the countries in the sample and thereby the regression estimates."

Most of the excluded countries are low-income countries with relatively low network access, but a few countries such as Seychelles and Mauritius have much higher income levels than most countries in the sample. Access rates in these two countries exceed that of every country in the sample. If we exclude these outliers, the difference in network access (between excluded and included countries) is just over 4 per 1,000 individuals. The difference in per capita GDP remains noticeable and falls in the lower income range for both included and excluded countries.

I also compared regression results from the excluded countries with the entire group of all countries in Africa. Table A4 shows that the results are quantitatively the same except that LTRADE is not significant in the smaller sample. On the basis of these rough sensitivity analyses, I feel that the qualitative results of the sample can be extended to the eighteen excluded countries, at least in a general manner, and that most of Africa can be described accurately with the results of the estimations.

2.7 Estimated Results

Table 2.8 shows the main effects of institutional factors on network development. Bureaucratic quality is significantly correlated with network development with an unexpected sign. The bureaucratic quality index may not be differentiating between the rules that define how the game is played and the players (the bureaucrats). Analysts often assume that the objective of



"To account for this possibility, I re-estimated Equation 2 without South Africa and then sequentially removed every country (one at a time) to see if there was any influence on the results. The results where changes did occur are reported in Table A2 in the Appendix and show that Algeria, Burkina Faso, Democratic Republic Congo, Cote d'Ivoire, Guinea, Guinea Bissau, Morocco, Mozambique, Uganda, South Africa and Zimbabwe may be individually affecting the Network regressions in that LGDPC became insignificant. In the case of Morocco, EXPROP also became insignificant. All specifications are re-estimated without these countries (reduced sample).






27


bureaucrats is to make decisions that maximize their own individual utilities. Unless there are strong incentives for them to improve industry performance, they may not encourage it. Personnel in a strong bureaucracy recognize how certain decisions will affect them personally, while less efficient bureaucracies may not. In this case the index may be reflecting protection of the bureaucrats' position rather than the ability to protect contract and property rights. Stronger bureaucratic quality in this instance denotes greater personal protection.

Generally, the results show that well-defined and credible institutions are positively and significantly correlated with network development regardless of the definition of institutions used. If a country such as Niger at the average level of adherence to the rule of law (LAW) were to increase this quality by 2 units to arrive at the level held by Botswana, then network development would increase from 1.6 telephones per 1,000 people to almost 2.2.

This corresponds to an elasticity of 0.34 at the mean. A reduction in the risk associated with contract repudiation in Cote d'Ivoire to South African levels would increase network access in Cote d'Ivoire from 11.65 telephones per 1,000 inhabitants to 16.42.

The results with the ICRG index explain less of the variation in NETWORK access (the R2 is lower), and the coefficients tend to be different from the original regression (ICRG components). The results are, however, similar in that they indicate that strong institutions are important for NETWORK development.

Per capita GDP is not always significant but other economic and demographic variables (LTRADE and URBAN) are significantly correlated with network development. The trade variable, however, has an unexpectedly negative sign, indicating that the state may not respond to economic forces in the way private individuals would. Bleaney and Greenway (2001) examined the impact of exchange rate volatility and terms of trade on investment and growth in subSaharan Africa and found that growth is adversely affected if a country specializes in the export of primary products. If this finding is applicable to infrastructure investments, it could explain the negative sign on LTRADE.






28


Table 2.8: Panel Fixed-Effect Regression Controlling for Country-Level Differences Variable ICRG ICRG index: The POLRIGHT: An DEMOC-AUTO:
Components: sum of index of Single summary
Institutional CORRUP. LAW, political rights measure of political
factors BUREAU, used to measure regime, which is
measured by CONTRACT and place of the difference
the individual EXPROP is used democracy between the index
components of as the of democracy and
ICRG institutional the index of
variable autocracy
LOGGDPC 3.345 4.424*** 3.304 1.098
(1.527) (1.865) (1.558) (0.893)
LTRADE -0.108* -0.115* -0.100* -0.098*
(-6.328) (-6.131) (-6.351) (6.229)
URBAN 0.697* 1.053* 0.739* 0.813*
(7.068) (8.550) (7.508) (8.133)
CORRUP -0.012 0.027 0.070
(-0.024) (0.053) (0.137)
LAW 1.693* 1.732* 1.717*
(2.769) (2.922) (2.820)
BUREAU -3.666* -3.732* -3.595*
(-3.169) (-3.191) (-3.099)
CONTRACT 2.170* 1.984* 1.967*
(5.086) (4.867) (4.497)
EXPROP 0.334 0.437 0.546
(0.996) (1.306) (1.582)
DEMOC -0.043** -0.036***
(-2.059) (-1.630)
ICRG 0.613*
(7.094)
DEMOC- -0.248*
AUTO (4.565)
POLRIGHT 0.538***
(1.884)
N 421 421 421 421
R2 (adjusted) 0.95 0.93 0.95 0.95
F-Stat 24.15 27.60 27.07 26.82
Note: The dependent variable is NETWORK, access to telephone lines per 1,000 people. The sample is 37 African nations over the period 1985-97. Since a test of the significance of a time trend could not be rejected, the time trend was excluded from estimations. * = significant at 1%,
** = significant at 5%, *** = significant at 10%.

The highly significant result on URBAN suggests that cost considerations with respect to

location may be important for some countries. Greater access to telephones in the city may have

a lot to do with its being cheaper to connect telephones to a network in urban as opposed to rural

areas. This idea is less easy to understand in the context of cellular service provision, which






29


entails less cumbersome and risky investments than wire line investments. Although cellular service is increasing in Africa, the bulk of the growth is in urban areas.

The regressions with the POLRIGHT index of political rights and DEMOC-AUTO provide results similar to those for the ICRG components estimation, indicating that political institutions will have an impact on telecom development in the region. Contrary to our prediction, however, the coefficient on measures of democracy is usually negative (positive for POLRIGHT). The negative (positive for POLRIGHT) correlation makes sense if we assume the perspective outlined by Haan and Siermann (1995). The existence of checks and balances on executives can hamper development as politicians take actions to placate pressure groups at the expense of investment in infrastructure. In addition, most democracies in Africa are transitional and thus fragile.

Table 2.9 shows the results when ENGINST, OINST, ENGPOL, and OPOL are included in the models. The results are qualitatively the same, except that LOGGDPC and EXPROP are always significant. POLRIGHT is no longer significant. The marginal effect of greater institutional quality on network access is increased when the countries involved have a tradition of French civil law compared to English common law code and other legal systems. This differs from expectations, but supports the results from the summary statistics when institutional quality is low to begin with.

This result might also be picking up the effect of something else that was not captured in the model. It is difficult to define countries according to only three types of legal systems, since most countries often utilize some other major system, such as customary or tribal laws. While countries fall clearly in one or the other of the categories used in the analysis, it fails to capture these other aspects of the legal system. While the level of political freedom is important to network development, there is no evidence that it has different effects according to origin of the legal system (OPOL is significant only in one specification). In conditions like these, the demands of pressure groups may be met by succumbing to a "pork barrel" regime that generates






30


Table 2.9: Panel Fixed-Effect Regression Controlling for Country-Level Differences and Accounting for Legal Traditions
Variable ICRG components: ICRG index: The POLRIGHT: DEMOC-AUTO:
Institutional factors sum of CORRUP, An index of Single summary
measured by the LAW, BUREAU, political rights measure of political individual CONTRACT and used to measure regime, which is the
components of EXPROP is used as place of difference between the
ICRG. the institutional democracy. index of democracy and
variable. the index of autocracy
LOGGDPC 5.040** 5.139** 4.816** 4.177***
(2.106) (1.999) (2.055) (1.845)
LTRADE -0.103* -0.106* -0.095* -0.090*
(5.662) (-5.377) (-5.722) (-5.382)
URBAN 0.614* 0.970* 0.649* 0.739*
(5.935) (7.466) (6.327) (7.131)
CORRUP 0.521 0.518 0.636
(0.990) (0.988) (1.232)
LAW 2.068* 2.046* 2.085*
(3.633) (3.640) (3.707)
BUREAU -2.639*** -2.798** -2.761***
(-1.855) (-2.023) (-1.929)
CONTRACT 2.418* 2.268* 2.220*
(5.588) (5.526) (4.939)
EXPROP 0.953* 0.965* 1.091*
(2.639) (2.722) (3.003)
DEMOC -0.042** -0.039***
(-1.957) (-1.730)
ICRG 1.132*
(7.189)
DEMOC- -0.271*
AUTO (4.796)
POLRIGHT 0.385
(1.378)
ENGINST -0.487** -0.695* -0.447** -0.445**
(-2.442) (-3.566) (-2.342) (-2.228)
OINST -0.568** -0.764* -0.499** -0.516**
(-2.511) (-3.542) (-2.307) (-2.423)
ENGPOL -0.014 0.010 -0.008 0.005
(-0.959) (-0.587) (-0.606) (0.332)
OPOL 0.025 0.001 0.026 0.048*
(1.457) (0.013) (1.526) (2.677)
N 421 421 421 421
R2 (adjusted) 0.95 0.93 0.95 0.95
F-Stat 20.71 20.91 21.60 22.30
Note: * = significant at 1%. ** = significant at 5%, *** = significant at 10%. Since a test of the significance of a time trend could not be rejected the time trend was excluded from estimations. Again, the dependent variable is access to telephone network per 1,000 people (NETWORK) with data from 37 African countries in the 1985-97 period.






31


popularity at the expense of increased productivity. This kind of behavior can undermine (telecom) investment, especially when the majority of investments involve public undertakings.

Tables 2.10 and 2.11 show the results of the NETWORK regression when eleven

countries (listed in footnote 10) are excluded from the sample. The results for the most part are as expected. LOGGDPC and EXPROP are now significant in every specification. The sensitivity test of country exclusion had shown that the inclusion of these eleven countries influenced the relationship between LOGGDPC and NETWORK to make it unexpectedly insignificant.

With an estimated real GDP growth of 6.5% in 1999, Botswana has become one of the fastest growing nations in Africa, while most other African countries are experiencing little, if any, growth. For Botswana, GDP and telecom development have been progressing together, and so are expected to be correlated. Other countries have also been modernizing and expanding their telecom infrastructure despite an unimpressive growth in GDP. South Africa for instance, has the most modem infrastructure in the region, yet its growth in GDP has not been as impressive as Botswana's. In 1999, South Africa's estimated GDP growth was only 0.6 percent. The slow growth in GDP at the same time that network access is expanding could mean that the two are not closely related. This may be one reason that the influence of South Africa was making LOGGDPC insignificant.

Although the size of the coefficients varied from those in the full sample, the results are qualitatively the same after taking into considering the exclusion of some countries. An exception is CORRUP and LAW, which changed significance in some specifications of the reduced sample.






32


Table 2.10: Panel Fixed-Effect Regression Controlling for Country-Level Differences, Reduced Sample
Variable ICRG ICRG index: The sum POLRIGHT: DEMOC-AUTO:
components: of CORRUP, LAW, An index of Single summary
Institutional BUREAU, political measure of political
factors CONTRACT and rights used to regime, which is the
measured by EXPROP is used as the measure difference between
the individual institutional variable. place of the index of
components of democracy. democracy and the
ICRG. index of autocracy
LOGGDPC 17.338* 15.788* 16.652* 15.404*
(6.021) (5.214) (6.043) (5.497)
LTRADE -0.100* -0.089* -0.085* -0.079*
(-5.442) (4.888) (5.119) (-4.972)
URBAN 0.788* 1.061* 0.876* 0.908*
(7.433) (9.411) (8.425) (8.250)
CORRUP 0.678 0.843** 0.839***
(1.573) (1.963) (1.931)
LAW 0.284 0.250 0.434
(0.491) (0.458) (0.743)
BUREAU -2.021* -2.005* -2.205*
(-2.982) (-3.169) (-3.423)
CONTRACT 1.033** 0.723*** 0.692***
(2.409) (1.787) (1.606)
EXPROP 1.283* 1.479* 1.637*
(3.588) (4.105) (4.493)
DEMOC -0.051** -0.052**
(2.038) (-2.114)
ICRG 0.452*
(5.098)
DEMOC- -0.305*
AUTO (5.648)
POLRIGHT 1.096*
(6.032)
N 280 280 280 280
R2 (adjusted) 0.93 0.922 0.94 0.94
F-Stat 21.80 31.77 23.86 23.88
Note: * = significant at 1%. ** = significant at 5%, *** = significant at 10%. The dependent variable is access to telephone network per 1,000 people (NETWORK) during 1985-97. Since a test of the significance of a time trend could not be rejected, the time trend was excluded from estimations.






33


Table 2.11: Panel Fixed-Effect Regression Controlling for Country-Level Differences and Accounting for Legal Traditions, Reduced Sample Variable ICRG ICRG index: The POLRIGHT DEMOC-AUTO:
components: sum of CORRUP, : An index Single summary
Institutional LAW, BUREAU, of political measure of political
factors measured CONTRACT and rights used regime, which is the by the individual EXPROP is used to measure difference between
components of as the institutional place of the index of
ICRG. variable. democracy. democracy and the
index of autocracy
LOGGDPC 18.013* 16.743* 17.415* 16.160*
(6.341) (5.514) (6.326) (5.775)
LTRADE -0.097* -0.085* -0.086* -0.080*
(-5.142) (-5.659) (-4.973) (4.629)
URBAN 1.487* 1.575* 1.565*
(3.024) (3.323) (3.309)
LAW 0.597 0.464 0.087
(1.048) (0.841) (1.422)
BUREAU -0.868 -1.101*** -1.312**
(-1.412) (-1.726) (-2.155)
CONTRACT 1.815* 1.438* 1.388*
(3.541) (2.849) (2.638)
EXPROP 1.813* 1.880* 2.048*
(4.399) (4.677) (5.016)
DEMOC -0.047*** -0.051**
(-1.814) (1.982)
ICRG 0.851*
(4.357)
DEMOC- -0.298*
AUTO (-5.068)
ENGINST -0.694* -0.511** -0.566* -0.562*
(-3.253) (-2.313) (-2.778) (2.770)
ENGINST -0.694* -0.511** -0.566* -0.562*
(-3.253) (-2.313) (-2.778) (2.770)
OINST -0.454** -0.383 -0.302 -0.423*
(-2.154) (-1.576) (-1.489) (-2.053)
ENGPOL -0.028 -0.036** -0.009 0.003
(-1.527) (-2.344) (0.715) (-0.188)
OPOL 0.038*** 0.032 0.043** 0.059**
(1.715) (1.562) (1.963) (2.514)
N 280 280 280 280
R2 (adjusted) 0.94 0.92 0.94 0.94
F-Stat 18.06 19.58 20.30 19.61


Note: * significant at 1%. ** = significant at 5%, *** = significant at 10%. The dependent variable is access to telephone network per 1,000 people (NETWORK) during 1985-97. Since a test of the significance of a time trend could not be rejected, the time trend was excluded from estimations.






34


2.8 Conclusions

Previous studies have examined other developing regions, including Asia, Latin America and the Caribbean. It is important consider Africa as well and not depend solely on conclusions based on empirical work for other developing regions.

The empirical results support the theory that credible institutions, including stable

political structures, are important driving forces behind the surge of modernization in Africa's telecommunications sector. The strong institutional results should serve as a signal to governments of the importance of creating and maintaining well-functioning political and regulatory institutions. To benefit from new technological innovations and competition, the institutional framework needs to adjust. Significant political reform will be required to mitigate risks to fragile and new democracies in Africa.

The results also show that the origin of a country's legal system is correlated with

network development. Controlling for other factors, countries with similar institutional quality tend to have higher access to telephones if their system is derived from the French civil code rather than English common law and other legal traditions.

More populated urban areas are associated with higher access to basic telephony. There may or may not be economies of density to be gained from providing service in a highly populated area. Generally, the existing low penetration levels, along with the generally small elasticities, suggest that African nations require huge adjustments in their investment climate to achieve access levels comparable to other developing countries in a short period of time. Nevertheless, with the present drive toward competition and privatization, changes can already be observed.















CHAPTER 3
ARE MAIN LINES AND MOBILE PHONES SUBSTITUTES OR COMPLEMENTS?
EVIDENCE FROM AFRICA

3.1 Introduction

Mobile telephone subscriptions have been growing rapidly since the 1980s in both

developing and developed regions. Subscriptions to fixed telephones have also grown, but at a slower rate than cellular in many regions of the world (ITU 1999). For instance, by 1997, mobile subscriptions in Lebanon accounted for 76 percent of total telephone subscriptions (World Development Indicators 2000). If cellular continues to grow rapidly, it is likely that its subscription will surpass fixed-line access in the near future. Currently, developing countries are experiencing the highest levels of mobile growth (ITU 1999).

The increasing use of mobile telephony has implications for main-line access in

developed countries and in regions where access to traditional wire-line telephones is relatively low. The growth of mobile subscription may reflect its role as a substitute for main lines. However, it is not uncommon for calls to be connected between a fixed line and a mobile telephone, so the services may in fact be complementary to each other.

Despite the growing importance of mobile telephony, very little is empirically established regarding its position vis-A-vis fixed-line telephony as stimulus to connectivity. Since the fastest growth in mobile is occurring in developing countries, this study examines its role in fixed-line development in Africa, illustrating the impact of mobile provision in many developing regions. As one of the emerging markets, Africa shows high growth and increasing competitiveness in mobile communications. It is also a region with very low access to fixed-line telephones. During the 1980s, cellular provision was practically nonexistent in Africa. Today, virtually all countries


35






36


in the region have access cellular service, and many have at least two operators, one of which is usually privately owned. These characteristics are evident in many developing countries.

Typically, cellular usage and main line access are both growing rapidly. In many cases, regions with rapid growth in one service also experience rapid growth in the other (compared to regions with slower growth). Regardless of the pace of growth, or level of telephone access, mobile usually grows faster than main line access. For instance, South Africa enjoys growth in main line access of about 39 per thousand, per year, but enjoys growth of 4.5 per thousand, per year in mobile access. The trend is similar in cases where access to main-telephone lines tend to be relatively low. In Tanzania, main line access grew by less than 0.1 per thousand people, per year, while mobile subscriptions increased by 0.3 per thousand each year. Looking at South Africa and Tanzania together, suggest that individuals may be using mobile, not necessarily as a substitute for fixed lines, but for other reasons such as joint use. In other words, as individuals become more sophisticated in their use of telephones, they tend to behave more like users in developed countries.

While mobile growth in Tanzania may appear to be negligible, in actuality it is

substantial, given that mobile usage in that country started only in 1995. By 2001, mobile subscriptions in Tanzania outweighed that of main line access. This story is also true for Morocco, which enjoys relatively high main line access. Between 1985 and 1997 main-line access grew by an average of 2.9 per thousand individuals per year, while mobile access increased by only 0.27 per thousand individuals per year. By 2001, however, access to mobile telephones surpassed that of main lines, at least in urban areas. This occurred as a result of the introduction of privatization and competition in the mobile sector. Mobile access, therefore, appears to be growing rapidly across the continent of Africa, regardless of the stage of main line development. For this reason, it is difficult to determine what the relation is between the two services by just comparing access trends.






37


Today, virtually all countries in the region have access to cellular service, and many have at least two operators, one of which is privately owned. This is happening as a first step towards telecom sector privatization, since most countries, although allowing private participation in cellular services, still support state-owned main line provision. In these markets, the only potential threat to fixed line provision is competition from cellular. The question of the role of mobile is thus interesting in the case where the mobile provider is privatized and not owned by the incumbent. Real competition can only occur in a situation like this, because if the incumbent is also the sole provider of mobile services, then the potential for competition is diminished to the extent that the incumbent controls the mobile/main line trade off from the supply side. In this scenario, the relation in consumption would be difficult to determine. Even in cases where mobile provision is privatized and competition exist, conceptually it remains unclear whether mobile is a substitute or complement for main lines in consumption.

A popular argument for the view that mobile usage is substituting for fixed lines is its prevalence where access to main lines is low or unreliable.'2 For instance, cellular phones may be an attractive alternative where it is difficult to install fixed-line networks. Because mobile networks can be installed more rapidly than fixed networks, they can alleviate waiting time for potential subscribers (Minges 1999) and reduce unsatisfied demand.

The use of pre-paid cards by mobile users also supports the view that mobile is a

substitute for main lines. With the use of pre-paid cards, users who otherwise would not qualify for a phone can now access the service. This is especially important for users in developing regions where it is not uncommon for people to lack credit histories. For individuals with poor credit histories, the option to pre-pay is an important development, since people are not



1 According to the ITU (1999) report, mobile telephones are used in developed countries to complement existing fixed lines but are emerging as a substitute for fixed lines in developing countries.






38


automatically disqualified from using the service because of bad credit. The pre-paid alternative also provides the opportunity for individuals to manage their telephone expenses, since the number of calls that can be made is restricted. The result is that more people than ever now have access to mobile telephones. The increased supply of cellular service allows an attractive alternative to fixed-line telephony.

Just like users in developed countries, some people in developing areas are attracted to mobile not because there are no alternatives, but because of the convenience of mobile phones (Frempong and Atubra 2001); furthermore, mobile phones are often used in conjunction with fixed-line telephones. Fixed lines are usually used at home, while mobile phones are used to keep in touch with home or office when individuals are on the road. In fact, calls from cellular phones are commonly made to fixed lines as opposed to another cellular phone (Jha and Majumdar 1999). In such instances, mobile and fixed lines are used in a complementary fashion. A larger fixed-line penetration increases the value of mobile service.

The high cost of cellular service relative to the nominal price of main lines can make

cellular phones unlikely to substitute for main lines.3 Even though the cost of cellular has been falling, it remains high relative to the cost of fixed telephones. In regions where income is low, cellular phones may be out of reach for most consumers. In cases like this, mobile subscriptions may be confined to the wealthy, a relatively small group within a country. The preceding suggests that the role of mobile is ambiguous, at least at a conceptual level. Thus, the question of substitutability versus complementarity has to be solved empirically.



13 The price of main-line service has not generally been a strong factor affecting its demand because the price elasticities of demand for main lines tend to be low. Ahan and Lee (1999) found that the demand for mobile is positively correlated with per capita GDP, although the price effects on mobile tend to be weak. Their research suggests that even where income is low and costs high, subscribers may not respond very much to the price of mobile. If this is true, then the high cost of mobile to the subscribers should not be a strong enough factor to seriously restrict the number of individuals willing to become cellular subscribers. It is therefore feasible that individuals unable to access main lines would be willing to use mobile phones whenever they are available.






39


3.2 Recent Empirical Studies

A large body of research involving the development of telephone network has focused on the role of institutions. These researchers found that a strong institutional framework is essential for network development and expansion. Other studies examining the effects of new market conditions for utilities have focused mainly on industry privatization and competition. The general consensus is that competition and privatization tend to be associated with increased efficiency and growth. However, privatization is not always associated with network expansion.







Table 3.1: Overview of Empirical Analyses of Telecommunications Development (International Studies)


Period/viethod


Questions


Conclusions


1947-1987 U.S. and the U.K.



1960-1992 - World Non-linear least squares


Henisz and Zelner (2001) 1975-1994 - World Non-linear least squares


Guiterrez (1999) Ros (1999)


1980-97-Latin America Panel data analysis



1986-1995 - World Fixed-effects model


Kwoka (1993) Henisz (1998)


What are the effects of privatization, competition, and regulatory reform on telecommunications penetration and efficiency levels?

What are the effects of privatization and competition on telephone network and efficiency?


What is the incremental impact of of privatization and competition on the total factor productivity growth of AT&T and British Telecom?

How do cross-national differences in the levels of checks and balances on executive discretion (created by variations in political structures) affect the relative rates of basic telephone infrastructure deployment?

How does the role of political constraints on executive discretion determine cross-national variation in the diffusion of basic telecom infrastructure?


Stuuy


AT&T's total factor productivity increased more rapidly as it faced stronger competition. Privatization is positively associated with gains in productivity.

There is a direct relation between penetration rates (given political constraints) and future penetration growth rates. Countries with poor penetration will begin to "catch up". Political and institutional factors are significantly related to telephone deployment.

Countries with poor telephone penetration begin to "catch up" quickly if the country's institutional environment is able to limit policy change. Political constraints result in reduced rent seeking, which in turn leads to reduced investment risk and increased growth in basic infrastructure.

The regulatory framework is positively associated with telephone network deployment and efficiency levels. Privatization is not a significant determinant of penetration.

Countries with greater than 50% of their telecommunications asset in the private expansion sector have higher penetration levels and growth. Privatization does not appear to lead to higher growth in telephone penetration in countries with per capita GDP of less than $10,000. Competition has no effect on network expansion. Both privatization and competition are positively associated with network efficiency.


0









StQ eidMto uestions Concl~usions


Latin America and Africa Fixed effect


Jha and Majumdar( 1999)


Guiterrez & Berg (2000)


Singh (2000) 1


1980-1985 OECD


1986, 1990, 1995
Latin America Panel fixed effect


Asia


How do regulation, privatization and competition affect telephone network expansion?


What is the effect of mobile diffusion on the productive efficiency of the Telecom sector?


How do the economic, demographic, institutional and regulatory factors affect the level of telephone penetration in Latin America?

What are the effects of institutional environment (which establishes property rights) on network expansion and efficiency?


Wallsten (200 1)


" Singh (2000) is more of a descriptive than an empirical paper in the sense that he used case studies rather than formal statistical techniques that are typically used in empirical papers.


Regulation by itself has no statistically significant impact on fixed-line penetration. Penetration along with competition is positively and significantly correlated with network development. Competition by itself is also important.

Cellular technology diffusion is positively associated with productive efficiency. Productive efficiency is enhanced when the monopoly operator is privatized and when the competitive environment is liberalized.

These factors play important and positive roles in the modernization of the sector. A strong regulatory framework has a significant and positive impact on penetration growth.

Market competition is better for growth than privatization alone. Network expansion and efficiency occur when adequate property rights and enforcement mechanisms are in place.


Study Period/Method


Questions


Conclusions






42


All of these studies except Wallsten (200 1)"5 examine the role of competition in basic services, rather than competition between mobile and fixed-line provision. Ros (1999) included some African countries in his data set to provide insight into the role of competition on main-line access as well as efficiency in the telecom sectors of these countries. The present study examines competition solely from the perspective of developing regions, using data from Africa. In addition, it recognizes the importance of mobile as a competitive force in a context where more competition occurs via cellular provision than via privatization of incumbent wire-line providers.

The role of competition in the telecommunications sector is seldom viewed from the perspective of how mobile provision affects basic service. However, the astounding growth of mobile communications makes it one of the largest forces in emerging markets. A better understanding of whether mobile is a substitute for or a complement of basic services is important for policy makers as they attempt to reform the telecommunications market to improve access and increase efficiency.

3.3 Hypotheses

3.3.1 Mobile Competition and Main-Line Performance

The relation between mobile competition and access to main line telephones is ambiguous. Mobile and main line correlation may be perceived as positive or negative, depending on whether they are complements or substitutes. When the two services are



" Wallsten (2001) looked at the relation between mobile and fixed-line access, but the data used have two main drawbacks. The first is that it lumps together two dissimilar regions with different levels of growth and telephone development in terms of main-line access. It is likely that there are larger variations across these regions, which may affect how confidently the results are to be interpreted regarding any one region. In addition, only 15 of the 55 countries in Africa were included in the sample, and five of those had monopoly mobile providers. These monopolies are usually at least partially owned by the incumbent wire-line firm. A single country from Northern Africa was included, which raises a question as to whether the sample was representative of the African region as a whole. In contrast, this study focuses on a single region and uses a more representative sample of countries from all parts of Africa. Instead of using the number of cellular providers to measure mobile competition, this study uses the actual level of mobile subscriptions.






43


complements, the increased usage of cellular phones will be associated with increased access to main lines. On the other hand, if mobile subscriptions are substituting for fixed lines (for instance, by covering unsatisfied demand for fixed lines), a negative relation is expected. 16

3.3.2 Other Considerations

The relation between telephone access and other factors, such as per capita GDP,

institutions and political regime has been established in other studies.'" The general consensus is that strong institutions and stable political systems are positively correlated with increased access to basic telecommunications. Likewise, as individual income increases, the demand for telephones should rise, especially in situations where individuals are poor to begin with and suffer from lack of access because of the inability to pay. The relation between per capita GDP and telephone access may, however, be small in situations where lack of access is attributable to insufficient supply rather than low demand.

3.4 Background

3.4.1 Competition Policies

Historically competition in basic services has been nonexistent in Africa. Even though there has been some privatization in the 1990s, providers of basic services remain monopolies in most of Africa.'8 In contrast, the mobile sector in Africa opened rapidly to competition, and



16 It is possible for mobile and main lines to be strategic substitutes rather than substitutes in consumption. Fixed-line penetration in Africa may be growing because incumbent providers think that increasing access is the best response to the competition it faces from mobile providers. If the services are strategic substitutes then the relation between the two will be positive. Likewise the two services may be complements in consumption or strategic complements. In the first instance individuals get connected because of an increase in the value of being connected as more and more people have access to fixed lines. It is important to be able to distinguish the different types of substitutes and complements for the purposes of proper interpretation of the relationship between mobile and main lines.

'7 See Henisz (1998), Gutierrez (1999), Hamilton (2002) and Singh (2000). '8 By 1997, Cote d'Ivoire, Gabon, Ghana, Madagascar, South Africa and Tanzania had partially privatized their communications, but most were still largely state-owned (Hamilton 2002).






44


Africa was one of first regions to adopt mobile service (Leblanch-Wohrer and Lewington, 2000).

As Table 3.2 shows, a number of countries currently sustain multiple (at least two)

providers (over a wide range per capita GDP).

Table 3.2 The Level of Mobile Competition in Africa by 1997
Country Legally Country Legally Country Legally
Permissible Permissible Permissible
Competition Competition Competition
Algeria M Ethiopia M Niger C
Angola M Gabon M Nigeria C
Benin M Gambia, The M Rwanda M
Botswana Ghana C Sao Tome
Burkina Faso M Guinea C Senegal C
Burundi C Guinea Seychelles
Bissau
Cameroon M Kenya M Sierra Leone C
Cape Verde C Lesotho M Somalia
C. African C Liberia South Africa C
Rep
Tchad D Libya M Sudan
Comoros Madagascar C Swaziland M
D. R. of C Malawi M Tanzania C
Congo
Congo M Mali M Togo D
Cote d'Ivoire C Mauritania M Tunisia M
Djibouti D Mauritius M Uganda D
Egypt D Morocco D Zambia C
E. Guinea M Mozambique M Zimbabwe C
Eritrea C Namibia M
M-monopoly, D-Duopoly, C-More than two providers Source: Leblanc-Wohrer and Lewington (2000) and Minges (1999)

One has to be cautious, however, in using the increase in cellular networks across Africa

as an indication of the growth in private-sector participation and competition. In fact, some

cellular ventures (Angola and Algeria for example) are still state-owned or monopolies.

Furthermore, the existence of cellular operators may not be a perfect measure of

competition since cellular service is often available only to the wealthy, who represent only a

small portion of African economies. The price of the service may limit its ability to act as a






45


strong competitive force where incomes are low. Nevertheless, in villages call aggregators can play a role in providing access (if not ownership) for low-income demanders. '9

3.4.2 Mobile versus Fixed Lines

As cellular networks continue to emerge, both fixed-line and wireless services have been growing at a rapid pace in Africa. Table 3.3 shows the relative size of a country's mobile and main-line markets using the 23 countries in our sample. Table 3.3: Relation Between Mobile and Main Lines 1987 1992 1997
Total Main-Line Penetration (per 1,000) 11.29 15.702 22.341
Total Mobile Subscription (per 1,000) 0.004 0.056 2.565
Ratio of Mobile to Main Line 0.00033 0.00359 0.111
Percent of New Mobile to Main Line 1.100 37.800
Mobile as a Percent of Total Subscription 0.033 0.358 10.285

Between 1987 and 1997, access to fixed lines almost doubled, growing from 11.29 per 1,000 inhabitants to 22.341 per 1,000. In 1987, South Africa, Tunisia and Algeria accounted for more than 50 percent of all fixed-line telephones in the sample. Ownership patterns had shifted little by 1997, so that South Africa, Tunisia, Botswana and Egypt accounted for 56 percent of main-line access across the sample. The trend shows that countries that were doing relatively well in early years continue to lead the way in the late 1990s. Some of the same countries appear to be leaders in the mobile market. Over the same period, (1987-97), mobile subscription grew from close to 0 per 1,000 people to 22.34 per 1,000 people. South Africa, Tunisia, Morocco and Gabon are among the countries that have the highest subscription rates. Gabon moved from


19Main-line access and mobile subscriptions tend to be significantly higher in middle-income countries compared to low-income countries in Africa. (Table 3.6 gives a comparison of telephone subscriptions in middle- and low-income countries in Africa). However, over time, as the price of cellular provision falls, this becomes less true. Uganda and Cote d'Ivoire are examples of countries with per capita income below $1,000 and with at least two cellular operators. Cellular operators can be potential threats to incumbent firms, since they can increase penetration at relatively low cost per additional subscriber. The threat of competition may be enough to give the incumbent the incentive to improve service. The potential threat is enough to provide the impetus for telecom growth.






46


having zero mobile subscriptions in 1987 to having the second highest subscription level of 8.35 per one thousand individuals. South Africa, by far, has had the most success in increasing access to mobile phones. By 1997 access to mobile phones in South Africa was 36 per thousand, which accounted for sixty-three percent of all cellular subscription in Africa. This rapid rate of growth in cellular subscriptions may be attributed to the influx of new mobile providers, both foreign and local.

While mobile subscriptions still lag behind fixed-line access, the gap between the two has been closing over time. The ratio of new mobile subscriptions to new main-line access is increasing over time to just under forty percent in 1997.

3.5 The Data

3.5.1 Defining the Data

Annual data used in the analysis represent 23 African countries, 1985-97.2o The time period and number of countries used was determined by constraints on the data available for some of the variables. The role of mobile competition in determining fixed-network development is assessed by using main line per 1,000 inhabitants as the dependent variable. The variables used in this analysis are shown in Table 3.4.



20 The countries included in the sample are Algeria, Botswana, Cameroon, Republic of Congo, Cote d'Ivoire, Egypt, Gabon, Ghana, Kenya, Madagascar, Malawi, Mali, Morocco, Niger, Nigeria, Sierra Leone, South Africa, Tanzania, Togo, Tunisia, Uganda, Zambia and Zimbabwe.






47


Table 3.4: Data Definitions and Sources


Variable Definition Source
Main lines per This is the first dependent variable. It is World Bank's World
1,000 inhabitants the (# of main telephone Development Indicators
lines/population)* 1,000. (1999).
Demand for main This is the second dependent variable. World Bank's World lines It is the (# of main telephone lines + # of Development Indicators
unmet applications/population)* 1,000. (1999).
Per capita GDP (Real Gross domestic product in World Bank's World
constant 1990 U.S. dollar)/population. Development Indicators (1999).
Institutional Institutional features which capture IRIS-3 file of International
Factors country risk factors. It is the sum of the Country Risk Guide (ICRG)
following: corruption of the political data, 1982-1997 constructed system (CORRUP), rule of law (LAW), by Stephen Knack and the bureaucratic quality (BUREAU), IRIS center.
security of contract (CONTRACT) and risk of expropriation (EXPROP).
Economic Freedom Two variables are used from this index: Economic Freedom of the government operations (GOVOP), World annual report, 1997.
which measure the extent to which Compile by James Gwartney personal choice and markets, rather than and Robert Lawson. political planning and coercion direct resources. The second variable is
discriminatory taxes, (DISCTAX), which measures the extent to which government protects property rights in terms of low transfers and subsidies.
Democracy Indicator of political regime type on a Polity III: Regime Change and
scale of 0 to 10, 0 being the least Political Authority 1800-1994. democratic. (DEMOC).2 Compiled by Keith Jaggers
and Robert Gurr.
Mobile (# of cellular World Bank's World
Subscriptions subscriptions/population)* 1,000. Development Indicators
(MOBILE). (1999)
Urban percent (urban population)/total population World Bank's World
(URBAN) Development Indicators
(1999).
Trade (imports+exports)/GDP lagged one World Bank's World
period. (LTRADE.) Development Indicators
(1999).


2 The level of Democracy is assumed to be constant between 1994 and 1997.






48


3.5.2 Descriptive Results

Table 3.5 summarizes access to telephones in based on regional differences and income. The data indicate North Africa having lower access to both main lines and mobile than SubSaharan Africa. The higher level of mobile subscription in sub-Saharan Africa is primarily due to access in South Africa (36.95 in 1997), compared to the highest rate in North Africa of 2.7 in Tunisia in the same year. South Africa is also the reason for the high fixed-line access in SubSaharan Africa. Once South Africa is excluded, mobile access in sub-Saharan Africa falls, but remain above that of North Africa. Although the difference seems negligible, the pattern suggests that countries with low access to fixed-line telephones tend to have relatively high access to cellular phones. This result concurs with the view that mobile is satisfying unmet demand for main lines.

Table 3.5 also shows that main line access is about nine times higher in middle-income

countries compared to low income ones. When South Africa is excluded from the sample, mobile subscription falls by almost one third, but access in middle-income countries remain higher than that of low-income countries. This pattern suggests that the ability to pay for mobile may be an important consideration even while cellular costs are falling. One explanation for this may be that providers incorrectly perceive that lower incomes reduce the willingness of individuals to pay for the service. In other words, they may view this as a factor that limits market size. Such providers would therefore target higher income markets. Lower access in this case would be a supply side problem.

Finally, Table 3.6 provides summary statistics based on regional differences and the level of mobile competition within a country by 1997. As expected, countries that allow mobile competition invest more in mobile communications. Interestingly, these same countries, on average, tend to invest less in main lines. According to the data, countries that allow mobile competition invest in mobile approximately two times as much as countries with monopolies in mobile service. At the same time mobile monopoly countries on average invest in main lines by






49


just under one and one half times more than countries that allow mobile competition. One

implication of this is that mobile competition allows cellular providers to be better able to pick up

the slack where main lines are under-supplied.

The difference between mobile subscriptions in low-income countries is miniscule when

compared to that of middle income ones. This result suggests that the ability of customers to pay

appears to be an important factor in determining mobile access.

Table 3.5: Regional Differences, Income and Telephone Access 1985-1997
North Africa Sub-Saharan Africa Sub-Sahara without
South Africa
Middle Income 34.63 45.794 25.420
(0.234) (2.685) (0.852)
Low Income - 4.580 4.580
(0.068) (0.068)
The first number in each cell represents average main-line penetration. The numbers in parentheses are mobile subscription per thousand individuals.

Table 3.6: Main-Line Penetration, Income and Mobile Provision
Market Structure for Mobile
Monopoly Competition
Middle Income 32.50 87.069
(0.528) (6.354)
Low Income 3.349 6.00
(0.021) (0.062)
All Income 21.569 10.736
(0.338) (0.527)
The first number in each cell represents average main-line penetration. The numbers in parentheses are mobile subscription per thousand individuals.

3.6 The Empirical Model

3.6.1 Model Estimation

Panel data estimation techniques are used to analyze the impact of MOBILE competition

on MAINLINE access. The model is estimated by using pooled data from 1985 to 1997. The

data set begins in 1985 because the mid-1980s marked the take-off of rapid telecom reform in

Africa as well as the introduction of cellular phones

Before estimating the panel data, some cross-sectional analyses are conducted to compare

the attributes of the explanatory variables at different points in time. The impact of MOBILE on






50


MAINLINE development is assessed for a cross-section of countries in 1987, 1993 and at the end of the sample period in 1997.

The following equation was estimated for each of the three years:

(1) MAINLINE = a + /K X + 6 ,

where X is a vector of independent variables, containing LOGGDPC, LTRADE, URBAN, CORRUP, LAW, BUREAU, CONTRACT, EXPROP, GOVOP, DISCTAX, DEMOC and MOBILE.

(2) AMINLINE = a + PK AX +,

where, A indicates that the variables are expressed as their 1997 values minus those of 1987.

Equation 3 is estimated to account for cross-country variation with the panel data approach.

(3) MAINLINECT = aCT + PK XCT + 4CT + 6CT where, XCT is the same vector of independent variables as in equation 1 for country C in year T. aCT are country dummies that account for unobserved differences across countries that may affect the dependent variable. 5CT captures year effects that take care of possible variations of the omitted variables through time. Equation 3 is estimated by expressing the variables as deviations from their group means. This approach picks up any constant differences on a country-specific level through time. The model is estimated by using fixed effects. A parsimonious specification in which mobile subscription per 1,000 individuals is the only righthand-side variable is first estimated to establish the relation between mobile and main line assess,



22 1987 as opposed to 1985 was used for the first cross-sectional estimation because there were not enough observations of MOBILE in 1985 to generate meaningful results. 23 The fixed-effect model assumes that the counties in the sample each have characteristics that are unique and do not change over time. These differences are captured in differences in the constant term.






51


and between mobile and main line demand in isolation of other independent variables. The standard errors of all estimations are adjusted for cross-observation error dependence with the Huber/White variance estimator.

3.6.2 Sample Selectivity

Only twenty-three out of more than fifty countries in Africa are included in the sample used in this analysis. This raises the concern that the results of the study may not be generalized to include the excluded countries. To check this possibility, we compare the excluded countries in cross-section with those included in the analysis along observable lines. Per capita GDP, main line and mobile access, trade and urban percentage are used as the basis for comparison. Table A in the appendix compares the overall means of these variables with and without South Africa as well as with and without Seychelles and Mauritius.4 Table A shows that at the mean, the included countries tend to be similar to those excluded. For instance, the biggest difference between main line access is only about 2 per 1,000 individuals. The difference in mobile access is negligible at less than 0.1 per 1,000. Average per capita GDP tends to be slightly higher in the included countries, but both groups tend to have an average income in the lower income category.

To further check the possibility of selectivity bias reduced fixed effects regressions

(using the same observable variables) were conducted for both groups of countries. Table B in the appendix shows the results. Apart from the trade variable, which is insignificant for the countries excluded, the regression results were quantitatively similar. Finally, the reduced regression was conducted using all the countries in Africa. Again, the results were quantitatively similar (in terms of sign and significance), to the results using just the 23 countries in our sample. This



24 South Africa, Seychelles and Mauritius have significantly higher access to main lines than the rest of Africa (107, 203,and 195, respectively, in 1997). These countries also tend to have higher income levels than the rest of Africa. Except for Botswana, they are the only other upper-middleincome countries in Africa.






52


informal sensitivity analysis (by itself), suggest that selectivity may not be an important problem, at least qualitatively.

3.6.3 Endogeniety

Some analysts argue that some reverse causality exists between GDP per capita and

infrastructure (telecom) investment. (Colier and Gunning 1999). Kerf and Smith (1996) suggest that the poor quality of Africa's infrastructure constrains private investment in other activities and thus is an obstacle to economic growth. That is, countries with proper infrastructure are expected to attract investors, which in turn generate a higher per capita GDP. (Madden and Savage 1998). This potential is unlikely to be present because the region studied is comprised of nations that are plagued with instability and represent relatively high-risk projects. An increase in telecommunications investment will not be sufficient to attract non-infrastructure investment that would significantly affect GDP.

MOBILE captures the competition effect in markets where direct telephony is a

monopoly. Competition is, however, a result of regulatory considerations, and monopoly firms often have control over the regulatory regimes in Africa through the ministry. Often there is no separation of regulatory functions and operation because both may be operating under the same ministry. Mobile and direct lines are therefore jointly determined. If MOBILE is endogenously determined, it will be correlated with the error term. Under this condition, ordinary least squares (OLS) estimations will tend to attribute changes in the dependent variable (MAINLINE) caused by the error term to MOBILE. As a result, the coefficient on MOBILE will be biased upward or downward, depending on the sign of the correlation between MOBILE and the error term. In addition, if MOBILE and MAINLINE are jointly determined, MOBILE cannot be considered fixed in repeated samplings and there is a potential bias in all the estimated coefficients. Consider the coefficient of URBAN (pA) for instance. Note that P3 is supposed to be the estimated effect of URBAN on MAINLINE, holding MOBILE and all other right-hand-side variables constant. MOBILE is, however, not held constant when changes in MAINLINE takes place.






53


Therefore, 03 may actually measure some mix of the effects of both URBAN and MOBILE. Because of the potential problems that may result when OLS is used, it is vital to consider alternative estimation techniques to reduce the simultaneity bias.

We use the instrumental variable (IV) technique to reduce the potential biases in the estimation of Equation 3. To use this technique, we replace MOBILE with a new variable (instrumental variable), which should be highly correlated with MOBILE but uncorrelated with the error term.

Until 1997 only six of the twenty-three countries in the sample had allowed some private participation in fixed-line telephone service provision. All governments, except for Cote d'Ivoire, which sold 51% of its shares to the private sector in 1997, had maintained majority ownership. Private ownership of wire-line phones was therefore practically nonexistent in the sample. In contrast, private individuals are usually licensed to provide cellular service. The nature of ownership made it more likely that private sector credit directed toward investment in telecommunications would go into mobile communications rather than fixed-line development. Thus, when main lines are publicly owned, credit to the private sector will not determine public investment. We identified PRIVCRED (the ratio of private sector credit to GDP) as an instrumental variable for MOBILE and use two-stage least squares to estimate Equation 3.25

It may be argued that an increase in PRIVCRED implies a reduced capacity for the

government to fund telephone investment. If true, PRIVCRED would be correlated with both MAINLINE and MOBILE. Historically, however, telecom reform is financed largely by loans from international lending agencies as part of their mandate to aid in the improvement of



25 Although PRIVCRED is defined as credit to the private sector as a fraction of GDP, it may also randomly include some credit to the public sector as well. Measurement errors introduced by this are assumed to be captured as part of the random error term since the inclusion of public sector credit is not systematic.






54


infrastructure in lesser-developed countries. Local investors in mobile communications are likely to be more dependent on domestic credit.26

3.6.4 The Dependent Variable Redefined

As discussed in section 3.3.1, it is not entirely clear whether MOBILE is a substitute or a complement for MAINLINE. The supply of mobile phones have grown rapidly, mainly because many governments tend to be more willing to introduce privatization and competition in the market for cellular rather than fixed lines. While the supply of mobile has been growing, so has the supply of fixed lines and more people are using both fixed lines and wireless phones. It is, however, not clear how much of the new cellular subscriptions are by individuals who also have access to fixed-line telephones.

A positive sign could indicate that the two services are complements. In this case, it may be that the same individuals have access to both fixed lines and cellular phones. A positive sign may, however, result even if the services are substitutes in consumption. Cellular operators can be potential threats to incumbent firms, since they can increase penetration at a relatively low cost per additional subscriber, and in less time. The threat of competition may be enough to give the incumbent the initiative to improve service. It is also unusual to observe a decline in the level of teledensity in countries where access is already low. Both MOBILE and MAINLINE access are



26 Because only one instrument is identified, a formal test of its validity could not be conducted. Instead, we performed a sensitivity analysis by comparing the R2s of the restricted regression (without PRIVCRED as a regressor) with the unrestricted regression (including PRIVCRED), using MAINLINE as the dependent variable. The R2 remained unchanged once PRIVCRED was added to the regression. Likewise, a test of first-stage correlation was conducted with MOBILE as the dependent variable. When PRIVCRED was added as an explanatory variable, the R2 improved noticeably, by 8%. The results of these tests indicate that PRIVCRED may be used as an instrument for MOBILE. The results are of course based on the current specification of MAINLINE. It may be possible that PRIVCRED would not survive these tests for a different specification of mainline development.






55


therefore expected to grow. It is therefore not entirely clear how to interpret a positive coefficient

on MOBILE when we estimate the model using a supply side definition of MAINLINE.

To check our interpretation of the relation between MOBILE and MAINLINE, we

redefine the dependent variable to be actual total access demand for wire line telephones, rather

than just satisfied demand. The new dependent variable is DEMAND, which measures the total

number of applications for fixed-line service.28 The equation estimated is:

(4) DEMANDCT = aCT + 8K XCT + (5CT + ECT,29

Equation 4 measures the effect of satisfied demand for mobile on total demand for main

lines. Since the model is expressed in terms of demand relations, we can infer more clearly the

relation between mobile and fixed lines. If the coefficient on MOBILE is positive, then the result

can be interpreted to mean that fixed lines and cellular phones are complements. On the other




2MAINLINE access per 1,000 inhabitants is a measure of how many telephone lines the respective governments or incumbent wire line firm actually make available to the public. This is usually lower than its demand. In essence main line access measures the portion of demand for main lines that have been met. Since there is usually a shortage, this is in effect satisfied demand. Ros (1999) argued that the low number of main lines is a supply rather than demand-side constraint. Because of the absence of rate rebalancing, residential access prices are likely to be below their economic costs. An increase in supply is overwhelmed by demand, mainly because of long waiting lists as well as low price elasticities of demand. 28 This is just main line access per thousand individuals waiting list per thousand. Waiting list captures the number of applications for connection to a main telephone line that have been unmet. Specifically, DEMAND applications* 1000 , where applications is unmet demand for
population
main lines + the supply of main lines.
29 The dependent variable is now a demand side variable and so is not determined by policy. Although MOBILE is still a policy variable, it is not simultaneously determined with the new dependent variable (DEMAND). If MOBILE was the total demand for cellular phones, then the two variables would be simultaneously determined, by virtue of the fact that MOBILE and DEMAND for main lines are related services. MOBILE, however, measures just the part of its demand that is satisfied. In other words, it measures demand constrained by actual supply of cellular phones. It is therefore not a complete measure of demand. It is in fact the supply of cellular phones, which is captured by this variable. If this is true, then we can assume that although MOBILE is a policy variable, it is not jointly determined with total main line demand.






56


hand, a negative coefficient may be interpreted to suggest that the two services are substitutes.

An increase in MOBILE usage could have externality effects, resulting in increased

MAINLILNE demand. The growing MAINLINE usage could, however, have the same effects. To separate the externality effects of MOBILE usage from that of MAINLINE usage itself, we introduce MAINLINE access lagged one period (LMAINLINE) as an independent variable in the DEMAND equation.

Finally, is possible that as mobile subscriptions increase, its relation with demand for wire-lines may change. To account for this possibility, we introduce mobile squared (SQMOBILE) as an independent variable in the demand equation.

3.6.5 Estimated Results

Table 3.7 presents the results of the cross-sectional estimations. In 1987, MOBILE, was not significantly associated with MAINLINE access, but had a large positive coefficient. This insignificant relation may have occurred because although growing, the level of MOBILE subscriptions was too small in 1987 to have any real effect on MAINLINE development. By 1997, however, MOBILE had cemented its presence in the African telecom market enough to be significantly and positively associated with MAINLINE. This is also shown in the last column of Table 3.7, which captures how the growth in MOBILE access between 1987 and 1997 affected fixed-line growth for a cross-section of countries. This could suggest that while in earlier years cellular provision was inconsequential in the telecommunications market, it is becoming more of a competitive force as its usage grows. The real impact of MOBILE on MAINLINE, however, appears to be very small between 1987 and 97. In this estimation, the mobile coefficient corresponds to an elasticity that is close to zero.

Generally, the large changes in the size of the coefficients as well as variation in their

significance levels over the three years reported, indicate that the variables included in the model are not constant overtime. These cross-sectional results are also suffer from omitted variable bias, so the coefficient size and significance may not be precise when country specific elements






57


are omitted. The fixed-effect model that is later estimated takes care of the omitted variable

problem and accounts for any constant level differences in the explanatory variables over time.

Table 3.8 shows the result of the regression in which MOBILE is isolated from all other

independent variables. The results indicate a positive and significant correlation between mobile

usage, and the two dependent variables (MAINLINE and DEMAND).

Table 3.7: Cross-Sectional Regression Results for 1987', 1997 and the Change Between 1985 and 1997. (Dependent Variable: MAINLINE per 1,000 individuals) Variable 1987 Cross- 1993 Cross- 1997 Cross- Change between
section. section section. 1987 and 1997
Cross-section.

LOGGDPC -3.548 10.814* 10.772** 22.126
(-0.875) (1.875) (2.308) (1.363)
LTRADE -0.104 -0.005 -0.229** -0.160**
(-1.107) (-0.027 (-2.025) (2.208)
URBAN 0.559** 0.024 0.599** 0.426
(2.261) (0.592) (2.044) (0.861)
CORRUP 2.391 11.526** 3.348 4.930**
(0.766) (2.213) (0.891) (1.985)
LAW -6.811*** -3.353 3.307 -1.416
(-3.401) (-1.528) (0.989) (-0.496)
BUREAU 4.551* -7.969*** -6.312** -4.379**
(1.800) (-2.872) (-2.510) (-1.777)
CONTRACT 3.057 0.550 1.798 5.475***
(1.211) (0.212) (1.195) (2.732)
EXPROP -0.900 2.885 2.946 -0.263
(-0.308) (0.821) (1.242) (-0.125)
GOVOP -0.150 1.015** 0.932* 1.046
(-0.353) (2.022) (1.627) (1.130)
DISCTAX -0.531* 0.135 -0.371 0.788*
(-1.638) (0.446) (-0.996) (1.636)
DEMOC 4.049** -1.401 -0.102 -1.975**
(2.546) (-1.535) (-0.010) (-2.183)
MOBILE 185.949 31.986* 1.308*** 0.503**
(1.531) (1.678) (4.135) (2.474)
N 23 23 23 23
R2 _ 0.90 0.87 0.95 0.85
F Stat 10.29 10.79 107.14 49.61


Note: a: Prior to 1987 there was not sufficient data on MOBILE to run cross-sectional regressions. T-stats are shown in brackets * significant at 10%. ** = significant at 5%, significant at 1%.

Tables 3.9 and 3.10 show the results of the pooled time series, cross-sectional

estimations. Table 3.9 shows the fixed and instrumental variable estimation results using






58


MAINLINE as the dependent variable. All estimations show a positive and significant correlation between MOBILE and MAINLINE. This result suggests that cellular subscriptions may be playing a complementary role to fixed telephone lines. It is also reasonable to interpret the positive sign in terms of competition within the industry.30 The use of cellular phones act as a competitive force, which encourages increased investment in direct lines.

It is possible that main lines and mobile phones may be complements sometimes and substitutes in other cases. For instance it may be possible for mobile and main lines to be complements in consumption in countries where income is relatively high and less so in areas with lower income levels. To further explore this potential differential relationship, the relation between main lines and mobile, we interact MOBILE with per capita GDP (MOBINC), and with a regional dummy which equals 1 if countries are Sub-Saharan and 0 if they are Northern African (MOBSUB). The results show that the interaction of MOBILE with per capita GDP is insignificant. However, the marginal effect of increased mobile subscription on mainline access is more than 2 times lower in Sub-Saharan Africa, than in Northern Africa. For instance it is not uncommon for professionals to own both mobile and fixed lines. At the same time, it is becoming increasingly commonplace to have access to one or the other for various different reasons. Mobile can play the role of both a substitute and a complement of main line in the same country at different points in time.



30 Ahan and Lee (1999) estimated the demand for mobile networks using data from sixty-four countries in both developed and developing countries and found that mobile demand is positively correlated with the number of fixed lines per person. 31 When South Africa is excluded, the marginal effect of increasing MOBILE on MAINLINE is reduced when income is higher. (Second column of Table C in the appendix). Since MAINLINE is a supply side measure, this result may be indicating that providers of fixed service will find it strategic to focus more on cellular provision in high income areas, rather than on fixed line expansion, which is relatively time consuming. From the demand side perspective, (where the dependent variable is DEMAND in Table C), MOBINC is not significant and the result for MOBSUB is quantitatively the same as for the other estimations in Tables 3.9 and 3.10.






59


Table 3.8: Parsimonious Specification of Panel Regressions Controlling for Country-Level Differences. Data: 23 countries, 1985-97
Variable Dependent Variable: Dependent Variable:
MAINLINE access per DEMAND for main lines per
1,000 people 1,000 people
MOBILE 0.93 1*** 0.853***
(4.951) (5.150)
N 299 239
R2 with fixed effects and other 0.91 0.93
independent variables32
R2 with only fixed effects 0.89 0.89
F-Stat 24.51 26.52


Note: t-stats are shown in brackets *=significant at 10%. **=significant at 5%,**= significant at 1%.

Although the various estimations show MOBILE as highly significant, large changes are necessary to affect main-line development noticeably. Using the fixed-effects coefficients from Table 3.9, if Gabon increased its cellular subscriptions by 10 per 1,000 people, main-line access would increase in that country by 22.83 telephones per 1,000 inhabitants. This corresponds to an elasticity of only 0.204 in 1997." A low mobile elasticity indicates that direct line and cellular subscriptions are not close complements. Cellular might be too costly for individuals with per capita incomes below $1,000. Only the elite can afford cellular telephones. In addition, many countries' regulatory frameworks are not yet equipped to deal with the emerging competitive markets.



32 The R2s of 0.91 and 0.93 indicates that the model explains much of the variation in the dependent variables. Specifically up to four percent more of the variation in MAINLINE and DEMAND is explained by MOBILE.

3 An increase of 22.83 does not represent a large increase in MAINLINE. This is because mobile subscription in Africa was only 2.67 per 1,000 individuals while MAINLINE access was 22.34. This is 8.7 times higher than mobile access. When mobile is increased by 10, mainline access in Gabon increased by only 2.2 times more than the increase in mobile even though mainline access was almost nine times higher in 1997. 34 MAINLINE elasticity is calculated using 1997 values for each country. That is, P*[MOBILE / MAINLINE].






60


Table 3.9: Panel Regressions Controlling for Country-Level Differences. Data: 23 countries, 1985-97 Dependent Variable: MAINLINE access per 1,000 people

Variable Fixed Effects Fixed effects with Instrumental Variable:
Interactions PRIVCRED used as a
proxy for MOBILE.
LOGGDPC 7.593*** 6.646** 8.262***
(2.740) (2.398) (3.063)
LTRADE -0.115*** -0.111*** -0.113***
(-6.252) (6.300) (-5.996)
URBAN 0.624*** 0.641*** 0.647***
(7.043) (7.307) (7.027)
CORRUP -0.067 -0.167 -0.270
(-0.138) (-0.342) (-0.553)
LAW 1.295** 0.905** 1.265**
(2.500) (1.972) (2.073)
BUREAU -2.296*** -1.696*** -1.500**
(-4.310) (-3.508) (-1.805)
CONTRACT 2.119*** 1.819*** 2.000***
(4.811) (4.406) (4.411)

EXPROP 0.060 0.242 0.068
(0.156) (0.687) (0.177)
GOVOP 0.422*** 0.412*** 0.493***
(2.975) (2.975) (3.245)
DISCTAX 0.184*** 0.162** 0.176***
(2.658) (2.378) (2.330)
DEMOC -0.441*** -0.407*** -0.467***
(-4.302) (4.088) (-4.241)
MOBILE 0.438*** 5.229*** 0.799**
(4.579) (3.930) (2.211)
MOBINC -0.0002
(-1.486)
MOBSUB -3.923**
(-2.520)
LMAINLINE
N 299 299 289
R2 with fixed effects 0.97 0.97 0.97
and other variables
R2 with only fixed 0.89 089 0.89
effects
F-Stat 32.55 34.91 304.73
Note: T-stats are shown in brackets * = significant at 10%. ** = significant at 5%, *** = significant at 1%. A test of the significance of a time trend could not be rejected. As a result, the time trend was excluded from our estimations.

Results from the instrumental variable estimation, with PRIVCRED used as an

instrument for MOBILE, show that the impact of the competition variable (MOBILE) remains

positive and statistically significant. Table 3.10 shows the results using DEMAND for main lines






61


as the dependent variable. The results for MOBILE competition are quantitatively the same as those in the MAINLINE estimations. The positive and significant coefficient on MOBILE in this case lend more support for the view that cellular and main line telephones may in fact be complements.

The second column of Table 3.10 accounts for the increase in value that may result from MAINLINE usage. Once the externality effect of MAINLINE usage is accounted for (by introducing LMAINLINE), the result indicates that the two services may in fact be complements in consumption. The positive coefficient on MOBILE may be interpreted to mean that an increase in satisfied demand for MOBILE is associated with an increase in total demand for main lines and the two services are used together. The result may be capturing the effect where two or more telephone users connect with each other, with at least one party using a cellular phone.

The coefficient on SQMOBILE (although small), suggests that when mobile usage is low, (below 35 per thousand), it may be playing the role of a complement for main line. However, as usage becomes more widespread, its role switches to be that of a substitute. This scenario does a good job of describing the evolution of cellular usage in these developing economies, and concurs with earlier suggestions that it is possible for the service to be play both the role of a complement and that of a substitute. As a new innovation in the mid to late-1980s, cellular service was relatively expensive. It is quite possible that at this stage, only professionals and private individuals with higher incomes (usually in urban areas) would utilize the service.



" We also ran a fixed-effects regression (result not reported) using the six countries in the sample where there is a single cellular provider, who is also the incumbent. The result of the regression using MAINLINE as the dependent variable shows a positive (but insignificant) coefficient on MOBILE. If the incumbent is also the only cellular provider, then the competitive pressure of a substitute that would result in an increase in main line access (when the mainline and mobile are substitutes) does not exist. In cases where the incumbent is the clear provider of mobile a negative correlation between main line access and cellular subscriptions would indicate that the two services might be substitutes. If the correlation is positive, it lends support to the view that the positive correlation found in our estimations indicate that the services are complements in consumption.






62


Table 3.10: Panel Regressions Controlling for Country-Level Differences. Data: 23 countries, 1985-97
Dependent Variable: DEMAND for main lines per 1,000 people
Variable Fixed Fixed Fixed Fixed Fixed Effects Fixed Effects
Effects Effects effects Effects with with with lagged
without with lagged with lagged SQMOBILE SQMOBILE
lagged MAINLINE Interact- MAINLINE and
MAINLINE ions and Interactions
Interactions
LOGGDPC 3.957 3.661 2.717 2.827 4.020 3.230
(1.109) (1.215) (0.784) (0.960) (1.323) (1.085)
LTRADE -0.089*** -0.057*** -0.083*** -0.055*** -0.057*** -0.055***
(-3.646) (-2.884) (-3.535) (-2.797) (-2.881) (2.798)
URBAN 0.951*** 0.887*** 0.961*** 0.898*** 0.886*** 0.897***
(10.149) (10.274) (10.421) (10.429) (10.336) (10.492)
CORRUP 0.046 0.600 -0.130 0.443 0.695 0.550
(0.069) (1.165) (-0.196) (0.835) (1.344) (1.030)
LAW 0.632 0.359 0.245 0.111 0.238 -0.013
(1.127) (0.736) (0.459) (0.237) (0.492) (-0.027)
BUREAU -1.597** -1.311** -1.047 -0.952* -1.275** -0.924
(-2.148) (-2.174) (-1.437) (1.606) (-2.117) (-1.555)
CONTRAC 1.941*** 1.653*** 1.616*** 1.449*** 1.669*** 1.467***
T (3.868) (4.112) (3.328) (3.692) (4.150) (3.728)
EXPROP -0.211 -0.354 0.048 -0.167 -0.362 -0.178
(-0.544) (-1.162) (0.131) (-0.562) (-1.193) (-0.599)
GOVOP 0.629*** 0.631*** 0.600*** 0.611*** 0.623*** 0.603***
(3.170) (3.992) (3.084) (3.932) (3.954) (3.884)
DISCTAX 0.132 0.088 0.077 0.053 0.089 0.055
(1.218) (0.875) (0.711) (0.526) (0.880) (0.536)
DEMOC -0.402*** -0.283** -0.333** -0.244** -0.285** -0.248**
(-2.759) (-2.337) (-2.348) (2.043) (-2.344) (-2.060)
MOBILE 0.496*** 0.399*** 3.966*** 2.775*** 0.770*** 3.226***
(5.280) (5.591) (3.397) (2.837) (3.732) (3.360)
SQMOBILE -0.011** -0.010**
(-2.299) (-2.310)
MOBINC 0.00003 0.00002 -0.00004
(0.148) (0.110) (-0.228)
MOBSUB -3.541*** -2.418** -2.336**
(-2.816) (-2.434) (2.406)
LMAIN 0.171*** 0.160*** 0.168*** 0.158***
LINE (3.189) (3.059) (3.166) (3.043)
N 239 238 239 238 238
R' with 0.98 0.98 0.98 0.99 0.99 0.99
fixed effects
and other
variables
R2 with only 0.92 0.92 0.92 0.92 0.92 0.92
fixed effects
F-Stat 41.11 67.14 42.25 67.41 57.98 59.87
Note: T-stats are shown in brackets * significant at 10%. ** = significant at 5%, * * * = significant at 1%. A test of the significance of a time trend could not be rejected. As a result, the time trend was excluded from our estimations.






63


These users tend to use cellular service in conjunction with fixed line phones. As the relative price of cellular service falls over time, usage spreads to other users, including rural and lower income subscribers. When mobile becomes available in regions where main line access is non-existent or low, it becomes a substitute for main line. Currently, the net effect of mobile on main line demand is greater than zero at the level of mobile access for most countries in the data set. As such, the complementary effect outweighs the substitution effect. However, as mobile subscription increases, its role as a substitute will begin to dominate.

Generally, the results show that both access to MAINLINES and the DEMAND for main lines tend to be higher where income levels are relatively high. The institutional environment is important for telecommunications development (Bergara, Henisz and Spiller 1998). The institutional variables are also important in determining main-line demand. Rule of law (LAW), security of contracts (CONTRACT), and role of markets (GOVOP) all affect penetration in a positive manner. Political institutions appear to be important, but democracy has a negative coefficient. Clearly, modeling these features warrants more attention in future research. In particular, modifications could be made to consider the possibility that the measures of the institutional environment used, may not readily combine in a linear model for their impact upon the relation between main line and mobile subscription.

3.7 Conclusions

Some research has found evidence that competition generally tends to improve industry performance and productivity. Studies that looked at the telecommunications industry have also found competition in basic services is associated with increased telecom growth and development. This study examine the role of MOBILE competition rather than competition in basic services. This issue is of interest particularly since mobile competition tends to be more widespread than competition in basic services. Many countries are privatizing, but most have stopped short of allowing multiple provision. Mobile competition is expanding its role in many developing countries, even where access to fixed-line telephones is very low.






64


The empirical results support the theory that credible institutions, including stable legal structures are important driving forces behind the surge of modernization in Africa's telecommunications sector. The effect of mobile competition on main-line development and demand, though relatively weak is significant. At different stages of cellular development, mobile play the role of both a substitute for, and a complement of main line demand. From the supply side perspective, the pressure on incumbent firms to improve service provision increase with mobile operators present, possibly leading to improvements in main-line quality as well. The empirical results presented here suggest that competition is important in fostering telecom development: it induces investment by wire-line investors. As more private participation and competition are allowed, competition is expected to have a much larger effect on wire-line investments.

The important lesson is that, although there is some substitution between mobile and main line, mobile's role as a complement dominates. As a result, mobile is not just picking up the slack where demand for main lines is unmet. A market for cellular phones exists beyond reducing the waiting list for traditional wire line phones. This possibility is important for private investors and governments, since the potential market size for both types of service is immense. People in developing countries do not demand cellular phones only as a second-best solution to becoming connected. A significant impetus for mobile demand could simply be due to its convenience (or its implications for social status). Thus, fixed-line providers should not expect a reduction in demand, even as cellular usage continues to expand. Competition from mobile can do much to improve main-line access.















CHAPTER 4
SUBSIDIZING UNIVERSAL TELEPHONE SERVICE AT THE STATE LEVEL

4.1 Introduction

Historically, the Federal Communications Commission (FCC) has pursued the goal of universal service (defined as maximum telephone subscribership) with various systems of crosssubsidization from toll to local telephone service. This system of subsidizing local calls by taxing toll calls resulted in large deviations between long distance and local rates, and researchers have questioned whether these cross-subsidies were achieving the universal service goals. Some find that universal service had little to do with states' adoption of subsidy programs.

In an attempt to bring telephone tariffs closer to economic cost, the FCC introduced a

subscriber line charge in 1984 that was supposed to shift part of the responsibility for access cost to the customer (Johnson 1988). The subscriber pays a fixed charge per month per line. The resulting increase in the flat rate for telephone service created concerns that the goal of universal service might be undermined. To encourage maximum subscribership, the FCC then initiated low-income programs to mitigate the effects of the higher rates and encourage subscribership. Among these programs was the Lifeline plan, established by the FCC in 1984. The aim of this plan is to aid in the promotion of universal service by helping low- income individuals afford monthly telephone costs.36



36 This plan allows for a reduction in fixed telephone service charges equal to the subscriber line charge (SLC) for eligible households. Each qualifying household receives assistance for a single telephone. The FCC provided a 50% reduction in SLC, provided that the participating state fund a reduction in basic local telephone rate equal to 50% of the SLC. Some states provided initiatives beyond the federal requirement by allowing low-income households additional support. As a result, the number of households eligible for support increased in those states.


65






66


The purpose of this paper is to characterize states according to their timing of adoption of the Lifeline plan (using U.S. data between 1984 and 1997). Specifically, I explore whether telephone subscription rates or interest group pressure or both influence the likelihood of adoption. The results show that interest group pressure is important, but also that states do respond to subscription levels in determining their adoption decisions. The FCC did succeed in getting the states with the lowest penetration rates to adopt the policy.37

Knowledge of the adoption patterns of the states is important because it can provide information regarding the pace of adoption and spread of future regulatory plans. It provides useful information for accessing whether the achievement of certain objectives requires regulation by the FCC (Donald and Sappington 1995). These issues are important even when access to the telephone network is relatively high. From a business perspective, telecommunications is important because of its potential influence on the performance of the economy. From a social welfare perspective, access is important to the extent that individuals can be connected to the community and for emergency assistance. It is important to pursue this research for these reasons, even though the adoption of the Lifeline plan was made mandatory in 1997."



" There has been much discussion regarding the success of the Lifeline program as a tool to aid in the achievement of the universal service goals of the FCC. Garbaz and Thompson (1997) examine the effect of this policy on improving subscription rates. This research does not attempt to provide an answer to this particular question. That is, it does not address the question of how the Lifeline plan could be successfully implemented. Rather, this research should serve to inform regulators regarding the potential for the success of future regulatory plans based on state characteristics.

3 In 1997, the Universal Service Order modified the Lifeline program to make it available in all states. The state-matching requirement was also modified and the federal support amount was increased. Beginning January 1, 1998, Lifeline customers received $3.50 in federal support without a state-matching requirement. An additional $1.75 federal support is provided if states further reduce intra-state rates by at least $3.50. Prior to the 1997 Universal Service Order, the Lifeline program required a matching local rate reduction that had to be approved by the state utility commission. Since 1998, the states are no longer required to provide matching reductions.






67


4.2 A Brief History of the Lifeline Plan

Following the initiation of the optional Lifeline plan in 1984, the FCC revised the plan in 1985 to provide a reduction in fixed telephone charges equal to twice the amount of the SLC. Under the revised plan, the FCC provides a waiver of the full SLC up to the amount of reduction provided by the state. In order to participate, the state has to fund a reduction in the local service rate at least equal to the amount of the SLC. Eligible individuals have to satisfy a means test, based on income. The states are also required to establish procedures that verify that only eligible households are benefiting from the plan.

Federal assistance for the Lifeline plan was funded through the carrier common line charge up to 1989. Beginning April 1989, funding was done through direct billing of interexchange carriers (IXCs) by the National Exchange Carrier Association. The IXCs are responsible for paying Lifeline assistance of at least 0.05 percent of the nationwide presubscribed lines. The revenue from this billing is used to pay certified local exchange carriers for matching Lifeline assistance costs. The federal portion of the Lifeline funds is funded from revenues from IXCs according to the number of pre-subscribed lines they serve.

The Lifeline benefit is available to individuals verified by the state as eligible for a state public assistance program in accordance with a means test. Eligibility is based on participation in any one of the following five public assistance programs: Medicaid, public housing, energy assistance program, food stamps and supplemental security income.39 The FCC specifies these public assistance programs, but each state may set its own guidelines by choosing one or more of the five public programs or by using additional programs to determine qualification for Lifeline assistance.



39 From Weinhaus et al. (2000). This research was conducted by the Telecommunications Industry Analysis Project (TIAP).






68


Up to 1997, adoption of the Lifeline plan was optional. As a result, some states adopted the plan while others did not. By 1989, 26 states had adopted the plan at varying points in time. The earliest adoption was by California in 1983. This was more than a year before the FCC initiated its program. Other early adopters include Arkansas, New York, Vermont, Oregon, North Carolina, the District of Colombia and Arizona.

After surveying related research, I outline the determinants of Lifeline adoption rates. Sections 5 and 6 develop the empirical model specification and present the data and empirical results. The concluding section summarizes the results.

4.3 Related Research

Stigler's (1971) seminal paper, formalized by Peltzman (1976), marked the beginning of a growing body of work devoted to the economic theory of regulation. In these models, interest groups lobby for the regulator to institute policies to their benefit. The regulator in turn arbitrates among the various interest groups in a utility-maximizing fashion. Related work includes Joksow (1974), in which the regulator maximizes his utility by minimizing criticism from interest groups.

Gary Becker (1983) and Dewey (2000) made some of the more recent contributions to the economic theory of regulation. Becker and Dewey extended earlier research to show that, where the majority rules, the success of an interest group is not wholly dependent on its relative size. They showed that small groups are not necessarily at a disadvantage, as there is a direct link between the efficiency with which political groups exert pressure and the benefits accrued by the particular interest group, regardless of its size. More efficient groups are likely to be winners while less efficient ones are likely to be losers.

Other economists have concentrated on testing some of these theories empirically. The general consensus is that interest group pressure is an important determinant of the choice of regulatory policy. Kaserman, Mayo and Pacey (1993) showed that the likelihood of deregulating long distance telephone service increases with business usage and when residential interests are less extensive.






69


Research on regulation began focusing on incentive regulation in later years as some states switched from rate-of-return regulation to flexible pricing. Mathios and Rogers (1989) found that telephone prices are lower in states where price flexibility is allowed than in states where rate of return is practiced. Donald and Sappington (1995) found that a state's decision to replace rate-of-return regulation with incentive regulation is directly related to local service rates, and particularly with high or low allowed earnings under rate-of-return regulation and high growth in the state's urban population. Berg and Jeong (1991, 1994) studied the determinants and impacts of cost-component incentive regulation on heat rates in the electric utility industry. They found that heat rates' efficiency improved with incentive regulation and that high prices relative to average total cost may be responsible for triggering cost-component incentive regulation. The latter finding is in keeping with that of Donald and Sappington (1995) in the telecommunications industry.

Research assessing the effect of later subsidy schemes intended to promote increased access to telephone service in the United States has also been done. Examples of these studies include Garbacz and Thompson (1997), Eriksson, Kaserman and Mayo (1997), and Kaserman, Mayo and Flyn (1990). These studies show that the cross-subsidy mechanism used to achieve universal service has failed to do so. There appears to be little or no causal relationship between these subsidies and the growth in subscription rate.

One argument used to support the ineffectiveness of these subsidies has to do with the

inelastic nature of telephone demand. To achieve even small increases in telephone subscription, rates have to be significantly reduced. Taylor (1980) compiled a comprehensive review of telephone demand studies. These suggest that price and income elasticities tend to be small, but generally increase with distance. If this is true, then subsidizing local service is not going to be effective in adding households to the telephone network. In addition, the reduction in access attributable to the increase in long distance price will undermine any increase in subscription that might have resulted from lower local prices.






70


However, Cain and Macdonald (1991), using 1987 data, argued that there is evidence that the use of low-income subsidies may help to improve subscription rates. They found that price has a noticeably stronger effect on access demand than previously believed, particularly at low levels of income. They argue that the elasticities have increased since 1980. This finding contradicts work done by Kaserman et al. (1990, 1997) and Garbacz et al. (1997), who obtain opposite results using post-1980 data. The question regarding the effectiveness of these subsidy programs is therefore still unanswered. The FCC and state regulators continue to implement these programs, the latest of which includes the low-income programs. However, it is also important to determine the impact of federal initiatives in states. The present research uses data incorporating more recent years to investigate the effect of subscription and pressure groups on Lifeline adoption probabilities and characterizes the states according to the timing of their adoption.

4.4 The Determinants of Lifeline Adoption Rates

4.4.1 The Network Externality Versus the Interest Group View

There are two views used to explain when states will adopt the subsidy program. The first, which seems to be the underlying assumption of the FCC, is that the Lifeline program is designed to motivate states with low subscribership level to increase penetration rates by offering subsidies to eligible households. We refer to this as the network externality view. Consumers benefit as more people are connected to the network, but private individuals do not consider other consumers' value when making their decisions to subscribe. In other words, consumers do not internalize the externality effect. The result is a divergence between the economic and socially desirable outcome. In maximizing the social good, regulators support policies that promote increased subscription because of this divergence.

On the basis of this view, states with lower subscription are expected to adopt the lowincome policy more quickly. In other words, states with a more severe need for subsidization will adopt earlier. Adoption of the Lifeline plan would add individuals to the network who, ceteris






71


paribus, would not have subscribed, thereby increasing overall subscribership. Households' quality of life is assumed to be better off with access to the network and ability to communicate with other members of the society and make calls in times of emergency. The welfaremaximizing regulator will be more strongly motivated to institute policies that increase subscribership to the extent that that welfare will be enhanced. From this perspective, regulators are welfare-maximizing agents. The externality effect is captured by telephone subscription (SUB84) in the empirical specification section. If this hypothesis that regulators respond to the universal service goal by adopting Lifeline is true, then SUB84 should be negative and significant. The rationale of the FCC is that regulators should aim to achieve maximum subscribership. If universal service appears to be threatened, the regulator is likely to adopt the lifeline plan more quickly than if telephone subscription is relatively high. Otherwise, we cannot reject the interest group view that universal service had little to do with the adoption of Lifeline.

Subscription at the beginning of the data set (prior to the signing of letters of certification by the FCC) is used instead of annual subscription because of potential reverse causality between adoption probabilities and subscription levels. If individuals who otherwise would not be connected to the telephone network are connected, then telephone subscription will increase when the plan is implemented. Estimation of simultaneous models in this framework is extremely problematic. Thus, to avoid problems of simultaneity in the estimation of adoption patterns across states, the subscription rate in 1984 is used as an explanatory variable instead of annual rates.4

The second view used to explain the timing of Lifeline adoption is more in line with the theories of regulation proposed by Stigler (1971) and Peltzman (1976). We refer to this as the



40Garbaz and Thompson (2001) suggest that the Lifeline effect on penetration rate was no different from zero. If this is true, then the reverse causality between telephone subscription and the Lifeline policy may not exist.






72


interest group view. Unlike the network externality view, this view considers state commissions as self-utility-maximizing agents. That is, they will mandate the Lifeline subsidy plan if it serves their interests to do so.

Regulators are subject to customer votes that may affect their future. As such, regulators favor actions that ensure favorable support from various interest groups. Since it is not always possible to satisfy all interest groups, the regulator will end up taking an action that satisfies the strongest group. Since increasing the rates on long distance calls funds the Lifeline program, the interest groups include long distance users and local customers. Assume that telephone users consist of residential customers only. Individuals who are eligible for the program, but who would have subscribed in its absence anyway, experience redistribution in their expenditures and there is no impact on subscribership. Conversely, subscribers not eligible for the subsidy are penalized by higher long distance rates. The only consumer beneficiaries are the subscribers who would not be connected to the network without the subsidy. Both beneficiaries and losers cast votes that affect the future of the regulator. The timing of state commissions' adoption of the Lifeline subsidy will depend on the relative strength of these interest groups. If long distance users are represented more strongly in earlier years than in later years in a state, then it is unlikely that regulators in that state will be early adopters. The growth in poverty (POVGROW), the fraction of the population that is non-metropolitan (RURAL) and the relative size of the nonmetropolitan population (PRATIO) are introduced in the empirical specification of the model to account for the strength of the relevant interest groups. The interest group view supports a positive relation between the time to adoption and these variables. Poor households should welcome local rate reductions immediately since they are likely to qualify for assistance.

Regulators may behave differently because of their utility preferences irrespective of

pressure group support. Regulatory type may therefore affect adoption probabilities. Following Kaserman et al. (1993), I introduce the dummy variable REGELECT to account for regulatory type. Whether the regulator is appointed or elected may influence the commission's interest and






73


therefore the decision to adopt the low-income program. An appointed regulator is more likely to reflect the view of government officials, which may or may not be in the interest of a pressure group. The elected official, however, may be more likely to act in a pro-voter fashion since his/her tenure depends on voter preferences.

4.4.2 Other Determinants of Lifeline Adoption

AVGRATE is introduced as a measure of the price of basic local telephone service in each state. We expect that adoption of the Lifeline plan is more likely where local prices are higher.4'

Just as the cost of telephone service to consumers (AVGRATE) is likely to influence adoption rate, it is also possible that the implementation of the policy may face budgetary constraints. The size of the regulator's budget will therefore influence the pattern of Lifeline adoption across states insofar as it determines what activities are affordable to the commission. The relative attractiveness of a rate reduction is diminished if implementation and administration costs are high. Higher costs are likely to result in a delay of policy adoption. RABUDGET and RABUGPS and RABUGC are used as a measure of the costs that the regulator can afford to incur.

4.5 Empirical Specification and Data

In this section, I use duration analysis to empirically test whether Lifeline adoption

decisions within states depend on telephone subscription rates, interest group pressure, or both. The data consists of 49 states and the District of Colombia. (Hawaii was excluded because data was unavailable for key variables).

Prior to 1997, it was not mandatory for states to adopt this policy. As a result some states adopted while others did not. Table 4.1 categorizes states according to their adoption



4'1 According to Donald and Sappington (1995), commissioners may be willing to consider alternative regulatory plans that reduce or limit further increase in charges for basic telephone service in states where local rates are high.






74


decision over the sample period. The explanatory variables used in the regressions are described in Table 4.2.


Table 4.1: States Included in the Data Set
Early Adopters Lat
Arizona, Arkansas, California, Ala Colorado, Idaho, Maine, Co
Montana, Maryland, Minnesota, Ge Michigan, Missouri, Nevada, Ma
New Mexico, New York, Mi
North Carolina, North Dakota, Ohio, Ok Oregon, Rhode Island, Texas, Sol
Utah, Vermont, Virginia, Washington, Sot West Virginia, Washington D.C. Wi


e Adopters Non-adopters
ska, Alabama, Delaware, Indiana,
nnecticut, Florida, Iowa, Kentucky,
orgia, Illinois, Kansas, Louisiana, Nebraska ssachusetts, New Hampshire,
ssissippi, New Jersey
lahoma, Pennsylvania, ith Carolina,
ith Dakota, Tennessee, sconsin, Wyoming


Note: Information for the table was taken from http/ftp.fcc.gov/Bureaus/Common_ carrier/reports/FCC.statelink/monitor/mrp7-0.pdf March 15, 2001. A state is classified as operating under Lifeline in a particular year as long as the FCC signed letters of certification some time that year. Early adopters are defined as states that adopted the Lifeline plan between 1984 and 1989. Late adopters include states that adopted the Lifeline plan in 1990 or later. Non-adopters are states that did not adopt the plan during the sample period 1984-97.


Table 4.2: Variables Definitions and Sources
Variable Definition Source
SUB84 Percentage of household that http/ftp.fcc.gov/Bureaus/commoncarrier
subscribe to the telephone network in /reports/FCC.State__link/monitor/mr97each state. 0.pdf. March 15, 2001.

POVGROW Growth in the fraction of households Bureau of Census.
below the poverty line in each state.
PRATIO Non-metropolitan population divided U.S. Department of Commerce,
by metropolitan population. Statistical Abstract of the United States.
RURAL Non-metropolitan population divided U.S. Department of Commerce,
by population. Statistical Abstract of the United States
REGELECT Dummy variable, equal to 1 if state The Book of the States, 1984 -1997.
commissioners are elected and zero Council of State Governments.
otherwise.
AVGRATE The average of basic (single, National Association of Regulatory
unmeasured) monthly local service Utility Commissioners, Bell Operating rate charged to residential consumers Companies exchange service telephone
in the smallest and largest urban rate. December 31, 1984-1997.
exchange.
RABUDGET Real annual public utilities National Association of Regulatory
commission (PUC) budget. Utility Commissioners Yearbook.
RABUGPS Real budget per subscriber, measured National Association of Regulatory
by the real PUC budget weighted by Utility Commissioners Yearbook.
the fraction of the population that
resides in metropolitan areas.






75


Table 4.2 Continued
Variable Definition Source
RABUGC Real PUC budget divided by the state National Association of Regulatory
population. Utility Commissioners Yearbook.

4.5.1 Statistical Model of Policy Adoption

States are defined as non-adopters prior to initiating the Lifeline plan and as adopters once the program has been implemented. A state's probability of adopting Lifeline is related to its characteristics listed above as exogenous variables in Table 4.2. Variations in Lifeline adoption dates are attributable to cross-sectional differences in these characteristics. Adoption probability is specified as:

P(state i adopts the lifeline plan at time t) = f(Xi, ti)42

where X, is the set of exogenous variables that affect state i's adoption of Lifeline, and ti is state i's time to adoption. The time to adoption is distributed over (0, 00). Time 0 is the first year that a state became at risk of adoption. Some observations are right censored; that is, there are states that are still non-adopters at the end of the sample period.

Estimations are conducted using hazard rate models.43 The hazard rate (hi (t)) is the

probability that state i will adopt Lifeline at time t conditional on having not adopted the policy before t. Hazard rate models are appropriate for this research because they provide information on the likelihood of transition from one condition to another. The structure of hazard rate models are such that the unconditional probability that state i will adopt at time t is

I
ft) = hi(t) exp -h(s)ds (8)




42 This definition of adoption probability follows much of the theory of duration models. For detailed discussions on duration models, see Green (1997), Amemiya (1985). 43 Rose and Joksow (1990) used hazard rate models to analyze technology diffusion. Hazard rate models are used to estimate the instantaneous probability of leaving the present state.






76


and the probability of state i not adopting the policy is


I - Fi(t)= exp -h(s)dsj. (9)


The Lifeline hazard rate is estimated by using maximum likelihood. To take account of right censoring of the data, a likelihood function of the following form is estimated.

L 7 =funcensoredobservations fi(t)l esrdbeains( it)(0
L =en redbse (1 - Fit)) ( 10)

Assumptions have to be made regarding the nature of the hazard rate in order to estimate the model.

4.5.2 Stationary Hazard Rate

The model is fist estimated under the assumption that the likelihood of adopting the

lifeline plan does not change throughout the entire sample period. This assumption means that the hazard rate is exponentially distributed. (Amemiya 1985). The hazard rate depends on the set of exogenous variables (Xi), such that

hi = exp(a+Xfp) (11)

Where a and P are unknown parameters. Denote states i =(l ...n) as states that completed their non-adoption spells of duration ti. States i=(n+l ... N) are right-censored at the end of the sample period (T) because they never adopted Lifeline during the sample period. The likelihood function is

n N
L H exp(a + Xpl)exp(- ti exp(a + X4)). H exp(- ti exp(a + Xp3)) (12)
i=1 In+ I

4.5.3 Non-Stationary Hazard Rate

The model is then estimated under the assumption that the relative probabilities of

adoption across states change through time. The non-stationary hazard model controls for trends or patterns of lifeline adoption in each state. In early years, there are more uncertainties surrounding the plan, in terms of how it should be implemented as well as the costs and benefits






77


associated with implementations and monitoring. This uncertainty may reduce the adoption probabilities and result in slow adoption at first by some states.

Over time, critics of low-income subsidies have emerged. Some doubt whether the lowincome programs are achieving their desired objective (Kaserman, Mayo and Flynn 1990 and Garbaz and Thompson 1997). At the same time, there appears to be some degree of success. If states that adopted Lifeline are viewed as successful, then other states are likely to adopt more quickly. It is not clear in theory whether the hazard rate is increasing or decreasing. The hazard rate is assumed to follow a Weibull distribution. This assumption allows estimation without having to know ex ante the exact direction of change of the hazard rate. The Weibull distribution allows the hazard rate to be comprised of a time component along with other explanatory variables. The hazard rate is defined as

hi= at "-I exp(Xg), (13)


where at"-' specifies the evolution of the hazard rate over time. If a > 1: > 0, and the



hazard rate is increasing. If a =1: =0, and the hazard rate is constant. If a < : < 0, 9t 9t

and the hazard rate is decreasing over time. The likelihood function is L = at,"' exp(XpJ) x exp(- exp(Xi)t" f exp(- exp(Xp)T). (14)
t+l

4.6 Data and Estimation Results

The statistical models described above are based on time-varying characteristics of each state. They are estimated by using duration analysis data on Lifeline adoption decisions of forty-nine states and the District of Colombia, between 1984 and 1997.

Table 4.3 reports means and standard deviations of the variables used in the sample by Lifeline adoption dates.






78


Table 4.3: Sample Descriptive Statistics, 1984-97 By Lifeline Adoption Date
Variable Full Sample Early Adopters Late Adopters Non-Adopters
(N=371 N=99 N=160 N=I 12
REGELECT 0.315 0.140 0.463 0.250
(0.465) (0.349) (0.500) (0.435)
RABUGET 8350.317 9176.280 9938.401 4957.307
(9588.435) (9819.811) (11276.71) (3963.799)
RABUGPS 6156.419 7016.017 6955.307 3494.205
(8311.68) (9022.227) (9192.528) (4078.854)
RABUGC 2.274 2.308 2.662 1.619
(1.985) (1.795) (2.469) (0.814)
AVGRATE 12.465 12.875 12.348 12.334
(2.610) (2.90) (2.694) (2.260)
SUB84 91.538 91.523 90.466 93.079
(3.674) (3.098) (4.161) (2.767)
POVGROW -0.016 -0.021 -0.058 -0.009
(2.817) (3.274) (2.610) (2.603)
RURAL 0.366 0.354 0.367 0.381
(0.220) (0.237) (0.213) (0.211)
PRARTIO 0.845 0.918 0.868 0.711
(0.882) (1.108) (0.872) (0.440)
Note: Standard deviations are shown in parentheses.

While there are variations in the means across the three sub-samples, the differences

between early and late adopters seem small. For instance, the differences in SUB84 between

early and late adopters is only one percentage point and the differences in AVGRATE and

RURAL are minute. Unexpectedly, the subscription rate is higher in states that adopted the

Lifeline plan early. However, if we compare adopters (early and late) to non-adopters the pattern

is as expected (SUB84 is higher in non-adopting states) and the difference in subscription rates

across the states is more noticeable. Likewise, the difference in REGELECT across the three

sub-samples is noticeable, but the means are all closer to zero than to 1, which indicates that more

commissioners are appointed rather than elected.

While differences in the budget of adopters are small, the noticeably smaller allocation to

non-adopters suggests that the availability of resources is likely to determine the timing of the

policy adoption. Likewise, the higher growth in poverty among non-adopters suggests that the

ability to pay for telephone service is a function in determining whether to adopt the Lifeline






79


plan, but that as the need base expands, undertaking the Lifeline plan may be too great a challenge and would be less likely to succeed. The data suggest that states with relatively large rural population are likely to adopt the plan later. This may be a reflection of the potential that a powerful (though numerically small) interest group may have on the future of the regulator, in terms of their efforts to generate votes.

Before presenting the regression estimates, I chart the diffusion path using the Weibull and exponential distributions that are assumed for the statistical analysis. Figures 4. land 4.2 fit the survival functions evaluated for a state close to the means of the covariates over the sample period. Both functions illustrate at least a slight increase in the hazard of adoption over time.

The graphs show that at least one state adopts the Lifeline plan each year. This suggests that a state's likelihood of adopting the Lifeline plan in any given year (conditional on its duration up to that point) is increasing over time. In other words, a state is more likely to adopt the Lifeline plan in the current year than it was in previous years. By 1990 close to half the states had adopted the plan. The survival functions of the two parametric models (Weibull and exponential) suggest slightly different hazards of adoption. By imposing a particular structure on the data, these models may distort the estimated likelihood of survival. As a result, the KaplanMeier (non-parametric) estimate is also presented. This representation of the survival functions may be more accurate because it imposes fewer restrictions on the data.

Figure 4.3 plots the Kaplan-Meier estimates of the survivor function. The function shows that the likelihood of adopting the Lifeline plan is relatively small during the first three years (1984-86). From 1986 to 1988 (corresponding to analysis time 3-5) the hazard of adoption appeared fairly constant and relatively large, but flattens slightly subsequently towards the end of the period. The general pattern suggests that continued adoption of the Lifeline plan by the remaining twenty-eight states was likely to continue past 1988 even without the mandatory adoption rules of 1997.






80


With this basic understanding of the Lifeline adoption pattern, we now turn to the

regression estimates of diffusion patterns. Tables 4.4 and 4.5 present the estimates of the Weibull and the exponential diffusion models for the entire data set, while Tables 4.6 and 4.7 report the estimations using data, which excludes the state of California.






81











1.0































0.24




1.0 Duration 13.0


Figure 4.1: Survivor Function-Webull Regression Model


I






82










0.90






























0.27




1.0 13.0
Duration


Figure 4.2: Survivor Function-Exponential Regression Model




83










1.00







0.75







0.50






0.25





0.00



0 5 10 15

Durafion


Figure 4.3: Survivor Function-Kaplan-Meier Estimate






84


Tables 4.4 and 4.5 report the estimates of the Weibull and exponential adoption models.

The results are similar across the two model specifications. States that start out with low

telephone subscription rates are likely to adopt the Lifeline plan earlier than states with higher

household subscription rates. A one-unit increase in telephone subscription rate cuts the hazard

(likelihood) of adoption by roughly 90 percent. This result is consistent with the view that the

Lifeline plan is a response to universal service goals. That is, states are likely to adopt the

Lifeline plan in order to boost low subscription rates. However, none of the variables intended to

capture the effect of interest groups is significant, except for the exponential estimation of

RURAL. This suggests that alternative proxies of interest groups may be required. Basic local

rates appear to have a positive impact on adoption probabilities in all but two specifications.

Table 4.4: Weibull Hazard Estimates of Lifeline Adoption Probabilities Variables RABUDGET RABUGPS RABUGC RURAL PRATIO
REGELECT -0.227*** -0.067*** -0.206*** -0.073* -0.080*
(-3.061) (-3.312) (-3.832) (-3.076) (-3.141
RABUDGET 1.000***
(5.171)
RABUGPS 1.00***
(3.602)
RABUGC 1.123*** 1.179*** 1.163***
(4.357) (3.635) (3.334)
AVGRATE 1.148** 1.168** 1.114 1.180* 1.132
(2.013) (2.061) (1.586) (1.900) (1.559)
SUB84 -0.905** -0.907** -0.891*** -0.885*** -0.896**
(-2.217) (-2.219) (-2.773) (-2.851 (-2.422)
POVGROW 1.155 1.185 1.158
(1.521) (1.200) (1.543)
RURAL -0.100
(-1.510)
PRATIO 0.688
(-0.700)
N 267 194 243 195 195
Log -34.969 -25.650 -32.145 -27.003 -28.110
Likelihood
cc 1.713 1.788 1.716 1.774 1.728
Note: t-statistics are shown in parentheses. ***=significant at the 1% level; ** significant at the 5% level; and *= significant at the 10% level.






85


Table 4.5: Exponential Hazard Estimates of Lifeline Adopti n Probabilities Variables RABUDGET RABUGPS RABUGC RURAL PRATIO
REGELECT -0.313*** -0.101*** -0.281*** -0.099*** -0.101***
(-3.155) (-3.626) (3.998) (-3.079 (-3.159)
RABUDGET 1.00***
(6.497)
RABUGPS 1.000***
(4.231)
RABUGC 1.123*** 1.187*** 1.174***
(4.655) (3.950 (3.526)
AVGRATE 1.092* 1.100* 1.063 1.109* 1.077
(1.932) (1.803) (1.308) (1.725) (1.260)
SUB84 -0.914** -0.911 -0.902*** -0.888*** -0.900**
(-2.267) (-2.245) (-2.680) (2.767 (-2.352)
POVGROW 1.171 1.190 1.169
(1.564) (1.281) (1.554)
RURAL -0.118-*
(-1.642
PRATIO 0.685
(-0.819)
N 267 194 243 195 195
Log -37.939 -28.253 -35.231 -29.429 -30.343
Likelihood II


Note: t-statistics are shown in parentheses. ***=significant at the 1% level; **= significant at the 5% level; and *= significant at the 10% level.

The hazard ratio estimates show that a one-unit increase in basic local rates can increase the hazard of adoption by up to 1.17 times. This result is expected. If local rates are high, they can act as a deterrent to subscribers. The Lifeline plan may be an attractive avenue for providing local rate discounts to select consumers.

The negative, although insignificant, effect of RURAL suggests that the preference will be not to adopt the lifeline plan in heavily rural areas. This result makes sense if we assume that potential users believe that the likelihood of benefiting from the program is low. Garbaz and Thompson (2001) suggest that 95-100 percent of all households receiving Lifeline subsidies would have been on the telephone network without any subsidy. Alternatively, the higher cost of service for less dense areas reduces the suppliers' interest in expanding penetration. The revenue per phone is less than the cost of providing the service.






86


The consistently significant and positive estimates of RABUGET, RABUGPS, RABUGC are according to our expectations. Regulators with access to a larger budget have greater access to information and resources, and are better able to determine to determine the pros and cons of adopting a new policy. In addition, states with larger per capita budgets are in a position to adopt the policy even where the cost per subscriber is relatively high. A one-unit increase in the percapita budget of the regulator can increase the likelihood of failure by as much as 1.2 times. The Weibull estimate of cX" is greater than 1, which suggests that the hazard rate increases through time. When the hazard ratio estimate is used on RABUGC, an cc value of 1.716 indicates that are

1.5 times more likely to adopt the Lifeline plan after seven years states than after three years.

States in which the regulators are elected rather than appointed are likely to adopt the

Lifeline plan later. The estimated hazard ratios suggest that the hazard of adoption is up to a third less if the regulator is elected rather than appointed. Donald and Sappington (1995) suggest that, appointed regulators are more likely to be insulated from the direct action of voters, and so may be more willing to experiment with new policies than if they were elected. Elected regulators are less likely to be reelected if their policies are unpopular and so may prefer to observe the success of a policy in other states before adoption in their own state.

California adopted the Lifeline plan in 1983, one year prior to the institution of the

optional Lifeline plan by the FCC. To check whether the special circumstances of California had any impact on the results, I re-estimated both the Weibull and exponential duration models. The results are presented in Tables 4.6 and 4.7, and are not qualitatively different from those that included California.



ca is an estimated parameter that tells whether the data used in the estimation has a hazard rate that is monotonically increasing or decreasing through time. It estimates the shape of the hazard rate through time.






87


Table 4.6: Weibull Hazard Estimates of Lifeline Adoption Probabilities Excluding California
Variables RABUDGET RABUGPS RABUGC RURAL PRATIO
REGELECT -0.229*** -0.075*** -0.197*** -0.070*** -0.078***
(-2.949) (-3.358) (-3.753) (-3.061) (3.108)
RABUDGET 1.000***
(4.882)
RABUGPS 1.000***
(3.060)
RABUGC 1.129*** 1.182*** 1.167***
(4.539) (3.569) (3.344)
AVGRATE 1.169** 1.185** 1.145** 1.214** 1.167**
(2.205) (2.162) (2.054) (2.151) (1.964)
SUB84 -0.905** -0.909** -0.890*** -0.886*** -0.897**
(-2.158) (-2.198) (-2.695) (2.810) (-2.344)
POVGROW 1.154 1.184 1.155
(1.526) (1.179) (1.502)
RURAL -0.115
(-1.363)
PRATIO -0.712
(-0.620)
N 266 193 242 194 194
Log Likelihood -33.539 -24.918 -29.765 -24.972 -25.922
a 1.819 1.888 1.839 1.920 1.872
Note: t-statistics are shown in parentheses. ***=significant at the 1% level; **= significant at the 5% level; and *= significant at the 10% level.

Table 4.7: Exponential Hazard Estimates of Lifeline Adoption Probabilities Excluding California
Variables RABUDGET RABUGPS RABUGC RURAL PRATIO
REGELECT -0.328*** -0.106*** -0.288*** -0.102*** -0.104***
(-3.065) (-3.503) (-3.921) (-3.059) (-3.117)
RABUDGET 1.000***
(6.476)
RABUGPS 1.000***
(3.362)
RABUGC 1.126*** 1.189*** 1.177***
(4.888) (3.919) (3.542)
AVGRATE 1.020** 1.100* 1.080* 1.125** 1.097*
(2.093) (1.855) (1.734) (1.965) (1.611)
SUB84 -0.916** -0.912** -0.904** -0.892*** -0.903**
(-2.231) (-2.238) (2.621) (2.696) (-2.262)
POVGROW 1.171 1.189 1.167
(1.556) (1.269) (1.496)
RURAL -0.142
(-1519)
PRATIO -0.710
(-0.754)






88


Table 4.7 Continued
Variables RABUDGET RABUGPS RABUGC RURAL PRATIO
N 266 193 242 194 194
Log -37.164 -27.881 -33.657 -28.111 -28.854
Likelihood
Note: t-statistics are shown in parentheses. ***=significant at the 1% level; **= significant at the 5% level; and *= significant at the 10% level.

To translate the effect of the independent variables on the likelihood of adoption on their meaning for the timing of adoption, I report elasticities up to the time of adoption with respect to each of the same independent variables. The results are reported in Tables 4.8 and 4.9.

All elasticities are reported at the sample means. The point estimates of the elasticites are similar across the Weibull and exponential models. Even though the t-statistics on the exponential estimations for AVGRATE elasticities are small, the point estimates are quite similar to their corresponding Weibull estimation.

Telephone subscription has by far the greatest impact on timing of the adoption of the Lifeline plan. On the basis of the Weibull estimation, a 1 percent increase in telephone subscription rate will lead to a 5-8 percent increase in the duration time (that is, an increase in the length of time that a state continues without adopting the policy). This means that a state that adopted the plan in 1990 would adopt four years later (1994) if the subscription rate in that state increased by 1 percent. The time until adoption is also sensitive to the local telephone rate. A 10 percent increase in the average local telephone rate can reduce the duration time by up to 13 percent. Telephone demand is believed to be inelastic, meaning that telephone prices really did not influence the extent of a subscriber's use of the telephone network. If this is true, then it would seem that policymakers should not respond to subscription rates in their decision to adopt the Lifeline plan. However, the Lifeline policy is designed to target states with low subscription rates attributable to unaffordable telephone rates. It may be that, once individuals are connected to the network, their response to changes in local rates is small, but the decision to get connected might be very responsive to the monthly expenditures that come with being connected. In other






89


words, individuals who are more price-sensitive may be the ones who are not connected to the

network.

Table 4.8: Weibull Elasticity Esti ates of the Survival Time of Lifeline Variables RABUDGET RABUGPS RABUGC RURAL PRATIO
REGELECT 0.301*** 0.483*** 0.314*** 0.468*** 0.464***
(3.57) (3.83) (4.44) (3.41) (3.42)
RABUDGET -0.144***
(-4.68)
RABUGPS -0.177***
(-3.83)
RABUGC -0.148*** -0.201*** -0.189***
(-4.75) (-3.61) (-3.25)
AVGRATE 1.010** -1.087*** -0.786* -1.169** -0.896**
(-2.56) (-2.78) (-1.93) (-2.40) (-1.86)
SUB84 5.340** 4.961** 6.183** 6.318** 5.820**
(2.13) (2.00) (2.50) (2.36) (2.09)
POVGROW 0.015 0.030 0.019
(-1.34) (-1.08) (-1.35)
RURAL 0.461
(1.49)
PRATIO 0.166
(0.69)
Note: t-statistics are shown in parentheses. ***=significant at the 1% level; **= significant at the 5% level; and *= significant at the 10% level. All elasticities are evaluated at the means of the independent variables.

Compared to SUB84 and AVGRATE, elasticities of the variables measuring the interest

group effect on the timing of adoption tend to relatively small. If we view regulators as self

interested parties, then REGELECT is the only measure of interest group pressure that has a

significant impact on the timing of adoption of the Lifeline plan. States with elected regulators

have a duration time that is 1/3 of a percent greater than it would be if the regulators were

appointed. This result suggests that even though elected official may be more likely to impose

pro-voter reforms, their interests closely coincided over the Lifeline policy.






90


Table 4.9: Exponential Elasticity Estimates of the Survival Time of Lifeline Variables RABUDGET RABUGPS RABUGC RURAL PRATIO
REGELECT 0.404*** 0.734*** 0.433*** 0.734*** 0.729***
(3.16) (3.63) (4.00) (3.08) (3.16)
RABUDGET -0.241***
(-6.50)
RABUGPS -0.296***
(4.23)
RABUGC -0.254*** -0.371*** -0.347***
(-4.65) (-3.95) (-3.53)

AVGRATE -1.100* -1.159* -0.760 -1.306* -0.932
(-1.93) (-1.80) (-1.31) (-1.72) (-1.26)
SUB84 8.204** 8.485** 9.398*** 10.872*** 9.671**
(2.27) (2.25) (2.68) (2.77) (2.35)
POVGROW 0.028 0.054 0.035
(-1.56) (-1.28) (-1.55)
RURAL 0.759*
(1.69)
PRATIO 0.289
1 (0.82)


Note: t-statistics are shown in parentheses. ***=significant at the 1% level; **= significant at the 5% level; and *= significant at the 10% level. All elasticities are evaluated at the means of the independent variables.

4.7 Conclusions

The results suggest that states with relatively low telephone penetration tended to lead the U.S telephone industry in the adoption of the Lifeline plan. All regression estimates show that states with low telephone penetration rates are more likely to be among the early adopters. This result suggests that the efforts of the FCC to encourage states with low penetration levels to adopt the Lifeline plan appear to have worked. It appears that states adopt the plan as a means of achieving universal service goals; whether the adopting states benefited significantly in terms of increased penetration is an entirely different question.

The results suggest that, if the Lifeline plan has not resulted in providing access to individuals who otherwise would not have access to the telephone network, there might have been a problem at the implementation stage of the plan. The question of how the program could be effectively implemented therefore remains. It is important to find the answer to this question for the purposes of future policy application.






91


The analysis provides some evidence that telephone rates may be an important factor

influencing the time of adoption. Adoption of the Lifeline plan is likely to take place earlier if the local telephone rate is relatively high.

Finally, the influence of stakeholders' interest in influencing adoption decisions can be supported to the extent that regulators are self-utility maximizing agents. The results strongly suggest that states where regulators are elected, rather than appointed, are less likely to be among early adopters. However, variables that are used as proxies for the effect of interest groups generally were not significantly different from zero, although they had the expected signs. This suggests that alternative proxies for interest group effect may be required.















CHAPTER 5
CONCLUDING REMARKS

This dissertation has discussed issues involved in the provision of telephone service and in the accomplishment of increased connectivity, in the United States and Africa. By looking at issues in these two regions, it became clear that the basic aim of improving sector performance is an ongoing one, regardless of the level of sector development.

From chapter 2, the main lesson is that in order for there to be significant improvement in telephone access level in Africa, there has to be a tremendous improvement in the risk environment in the region. This is true whether public or private investors will undertake telecom investments. Like private investment, public investment is likely only if there are positive returns to that investment. Private investors have to be concerned with the risk of losing their investments or having the returns confiscated. Public investors are interested in the risk environment insofar as it reflects the likelihood of their being around to recoup the benefits (politically or otherwise) of the investments.

Analysis of the data reveals that the access level is extremely low in Africa, but the large waiting lists indicate that the market potential in that region is immense. With the rapid expansion of cellular usage, the options for improving telephone access have increased. Through its potential for competition, mobile provision encourages wire line providers to expand their service. In addition, as a substitute for main line phones, mobile telephones provide an alternative to policymakers or regulators to extend telephone access through cellular provision. This policy implication is important because it suggests that low access levels in Africa may be resolved more quickly than if the sole process for doing so was via main line provision. The potential for using cellular service in this role suggests that its importance will continue to grow


92






93


in the region and may be seen as the solution to low connectivity that is widespread throughout Africa.

With new developments in the achievement in telecommunications reform also comes new challenges, as demonstrated in my analysis of Lifeline policy reform in the United States. Despite the high telephone access level in the United States, there is still room for improvement. Chapter 4 contributes to the understanding of the reform environment by characterizing states within the United States according to the timing of their adoption of the Lifeline plan. Understanding the adoption patterns provides information regarding the potential success or failure of future regulatory plans.



















Table Al: Countries I
Country Algeria Angola Botswana Burkina Faso Cameroon Congo, Dem. Rep. Congo, Rep. Cote d'Ivoire Egypt
Ethiopia Gabon The Gambia Ghana Guinea Guinea Bissau Kenya Liberia Libya Madagascar Malawi Mali Morocco Mozambique Namibia Niger Nigeria Senegal Sierra Leone Somalia South Africa Sudan Tanzania Togo Tunisia Uganda Zambia Zimbabwe


APPENDIX A
DATA ANALYSIS FOR CHAPTER 2 ncluded in the Sample


Legal System Official Language
French, Islamic Arabic
Portuguese Civil Portuguese
Roman-Dutch English
French Civil French
French Civil French
Belgian Civil French
French Civil French
French Civil French
English, Islamic Arabic
Transitional (regional and national courts) English French Civil French
English Common English
English Common English
French Civil French
N/A Portuguese
English Common English
Anglo American and Customary English
Italian Civil and Islamic Arabic/English
French Civil French
English Common English
French Civil French
Islamic, French, Spanish Arabic
Portuguese Civil Portuguese
Roman-Dutch English
French Civil French
English Common English
Portuguese Civil French
English Common English
N/A Arabic/English
English, Roman-Dutch English
English Common Arabic/English
English Common English
French Civil French
French Civil French
English Common English
English Common English
English Common, Roman-Dutch English


Source: CIA World Factbook 1999.


94




Full Text

PAGE 1

TELECOMMUNICATIONS REFORM IN AFRICA AND THE UNITED STATES By JACQUELINE MARIE HAMILTON 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 2002

PAGE 2

ACKNOWLEDGEMENTS I am frequently asked why I decided to focus my research on issues of telecommunications development. The answer is that it all began with the chair of my committee, Dr. Sanford Berg. As a result I would like to first express my sincere appreciation to him for helping me develop an interest in an area that is so important to a countryÂ’s development. Dr. Berg has gone way beyond his duties as chair of my committee. He has contributed immensely to my development as an individual as well as a researcher. I am very grateful for the role that he has played so well as my great motivator and advisor. I have been very fortunate to have the assistance of Dr. David Figlio, Dr. Chunrong Ai and Dr. Grant Thrall as members of my committee. They have been extremely helpful at every step of the way and have contributed significantly to my development as an economist. I would also like to thank Drs. Jonathan Hamilton, Larry Kenny and David Sappington for their helpful comments and advice throughout my time as a graduate student at the University of Florida. I was extremely lucky to be given the opportunity to study in such a welcoming and helpful environment. I am also grateful to Dr. Luis Gutierrez, who has always been very willing to assist me, especially in the beginning stage of my dissertation. Finally, I have to thank Derek Horrall, who has always been a rock that supports me despite his own challenges as a Ph.D. student. He has motivated and inspired me at the time when I most needed validation. Derek, along with the rest of my family, has motivated and supported me through steadfast belief in my ability. For this I am eternally grateful. 11

PAGE 3

TABLE OF CONTENTS page ACKNOWLEDGEMNTS ii ABSTRACT v CHAPTERS 1 INTRODUCTION 1 1 . 1 Privatization of the Telephone Industry 1 1.2 Telecommunications Sector Competition 1 1.3 Technological Innovation 2 1.4 Applications 2 2 INSTITUTIONS, POLITICAL REGIME AND ACCESS TO TELEOMMUNIC ATION S INFRASTRUCTURE IN AFRICA 4 2.1 Introduction 4 2.2 Recent Research 8 2.3 Hypothesis 9 2.3.1 Institutions and Network Access 9 2.3.2 Political Systems and Network Access 1 1 2.3.3 Cultural Considerations and Network Access 1 1 2.4 The State of Telecommunications in Africa 13 2.5 The Data 15 2.5.1 Defining the Data 1 5 2.5.2 Descriptive Results 17 2.6 The Model 21 2.6.1 Omitted Variables 22 2.6.2 Alternative Specifications 24 2.6.3 Country Effect and Sample Selectivity 25 2.7 Estimated Results 26 2.8 Conclusions 34 3 ARE MAIN LINES AND MOBILE TELEPHONES SUBSTITUTES OR COMPLEMENTS? EVIDENCE FROM AFRICA 35 3.1 Introduction 35 3.2 Recent Empirical Studies 39 3.3 Hypothesis 42 3.3.1 Mobile Competition and Main-Line performance 42 3.3.2 Other Considerations 43 3.4 Background 43 iii

PAGE 4

3.4.1 Competition Policies 43 3.4.2 Mobile versus Fixed Lines 45 3.5 The Data 46 3.5.1 Defining the Data 46 3.5.2 Descriptive Results 48 3.6 The Empirical Model 49 3.6.1 Model Estimation 49 3.6.2 Sample Selectivity 51 3.6.3 Endogeniety 52 3.6.4 The Dependent Variable Redefined 54 3.6.5 Estimated Results 56 3.7 Conclusions 63 4. SUBSIDIZING INIVERSAL TELEPHONE SERVICE AT THE STATE LEVEL 65 4.1 Introduction 65 4.2 A Brief History of the Lifeline Plan 67 4.3 Related Research 68 4.4 The Determinants of Lifeline Adoption Rates 70 4.4.1 The Network Externality Versus the Interest Group View 70 4.4.2 Other Determinants of Lifeline Adoption 73 4.5 Empirical Specification and Data 73 4.5.1 Statistical Model of Policy Adoption 75 4.5.2 Stationary Hazard Rate 76 4.5.3 Non-Stationary Hazard Rate 76 4.6 Data Estimation and Results 77 4.7 Conclusions 90 5. CONCLUDING REMARKS 92 APPENDICES A DATA ANALYSIS FOR CHAPTER 2 94 B DATA ANALYSIS FOR CHAPTER 3 97 REFERENCES 99 BIOGRAPHICAL SKETCH 104 IV

PAGE 5

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 TELECOMMUNICATIONS REFORM IN AFRICA AND THE UNITED STATES By Jacqueline Marie Hamilton May 2002 Chairperson: Dr. Sanford Berg Major Department: Economics My research is reported in three essays that focus on issues involving telecommunications development since the 1980s, from the perspective of developing and developed regions. In looking at issues from both viewpoints we are able to develop a broader perspective regarding specific issues involved in the improvement of telephone access. The research gives insight into the workings of telecom development and could serve to inform both private and public investment decisions. The first essay explores particular institutional issues that exist in Africa and develops a relation between telecommunications development and institutions in that region. Specifically, this essay analyzes telephone penetration in thirty-seven African countries, by looking at the impact of the institutional environment, political conditions and legal systems on access to fixedand wire line telephones. The analysis suggests that strong institutions can promote investment in telephone infrastructure, particularly in countries that historically base their legal systems on the French civil code. In addition, a countryÂ’s advancement democratically does not appear to improve telephone access in Africa. Finally, high per-capita GDP is associated with improved v

PAGE 6

telecommunications performance but by itself is not as strong as one might expect in comparison with the results of other cross-national studies. This essay sets the stage for further analysis regarding technological development (of cellular service) in that region. The second essay examines the relation (substitutes or complements) between mobile and fixed-line telephone development in a region with very low main line access and the potential for rapid growth in cellular service. The relationship is modeled by accounting for reverse causality between main line and mobile phones. The results challenge the belief that cellular plays a different role in developing countries than it does in a developed region like the United States. The analysis suggests that mobile phones act as a competitive force, thereby encouraging fixedline providers to improve access. The third essay examines telecommunications reform in the United States, where access is close to universal but where is still tremendous concern about increasing penetration as well as making access affordable for everyone. Many states have adopted policies that attempt to achieve maximum access within each state. In this essay I attempt to characterize states according to the timing of their adoption of the Lifeline plan. Together, the studies demonstrate the usefulness of economic modeling for understanding the impacts of public policy on telecommunications infrastructure. vi

PAGE 7

CHAPTER 1 INTRODUCTION Reform of the telecommunications industry in developed countries such as the United States has created a basis for reforming the industry in many developing countries. During the 1980s, the view of the place of telecommunications in the market changed. Developing countries began to believe that telephone sector performance could be improved only if the sector was completely reformed. Popular avenues for reform included privatization and liberalization of the industry, technological innovation and the introduction of competition. 1.1 Privatization of the Telephone Industry The degree of and the approach to reform of the telephone industry differed across countries, but the trend in the 1980s moved toward an ownership structure that was at least partially privatized. It was still believed that the sector was best served through a monopoly or through significant if not complete government presence. At least initially governments continued to play a significant role in the provision of the service. In many instances, full privatization was accompanied by licenses that guaranteed a long exclusivity period to large multi-national corporations. Privatization of the telecom industry typically improved sector performance in terms of increased efficiency and access. It was, however, increasingly debated whether even better performance could be generated through sector competition. 1.2 Telecommunications Sector Competition In an effort to continue improving sector performance, many countries committed to allow some degree of competition within the sector. Examples such as the break up of AT&T and the introduction of competition in the early 1980s suggested that the telecommunications 1

PAGE 8

2 industry may not be a natural monopoly and that competition could further improve sector performance. In the United States, competition in the provision of telephone services was done mostly in long distance rather than local provision of wire line service. In addition there were multiple cellular service providers. For most developing countries, however, competition was primarily introduced in the cellular market. 1.3 Technological Innovation The development of cellular technology is one of the biggest innovations in the telecommunications industry. It quickly asserted its place within the developed world and proceeded to invade the developing world at an astounding rate. Many developing countries saw cellular usage as a possible solution to the problem of low access rates. As the technology continues to improve and the cost of its provision falls, cellular emerges as a very important avenue for increasing telephone subscription rapidly. While sector reform has resulted in a noticeable increase in telephone access, the biggest growth is in cellular usage, and the prospect for increased telephone access in the developing world seems less daunting. 1.4 Applications This dissertation examines issues of reform in both developed and developing countries. Typically, privatization initiatives as well as competition and technological advancement have been introduced in both developed and developing countries at different paces and at different times. Many of the achievements, experiences and expertise within developed countries have been offered as examples of how sector reform may be implemented. Experience across developing countries is also shared as part of the ongoing attempt to improve the provision of telecommunications provision. With this in mind, I examine different issues of reform chapters 2-4 of this dissertation, in both developing and developed countries. In chapter 2, 1 provide a detailed analysis of the institutional, economic, political, cultural and demographic environment for telephone investment

PAGE 9

3 for thirty-seven African countries. The results indicate that significant reform of the institutional environment is essential for any significant improvement in sector performance. Chapter 3 highlights cellular technology as one of the most important developments in the telecom sector. It assesses the role of cellular phones within the context of extremely low telephone connection rates in twenty-three African countries. I tested the hypothesis that cellular phones and mainline telephones are substitutes in consumption. I found the hypothesis to be true only after achieving a critical cellular subscription rate. Chapter 4 continues to assess the implication of telecom reform by focusing on policy reforms within the context of a developed country. In this chapter I used state-level data from the United States to characterize states based on the timing of their adoption of a low-income program called the Lifeline plan. Chapter 5 presents the general conclusions.

PAGE 10

CHAPTER 2 INSTITUTIONS, POLITICAL REGIME AND ACCESS TO TELECOMMUNICATIONS INFRASTRUCTURE IN AFRICA 2.1 Introduction Empirical evidence regarding the effect of institutions and political regime exists, but little is empirically established for developing economies in Africa. 1 Most analyses of telecommunications reform in Africa are found in case studies and country reports, such as those by Chance, Booz, Allen and Hamilton, Inc. (1998), Laidlaw and Parkinson (1995), Frempong and Atubra (2001) and Onwumechili (2001). Policy inferences for the African telecommunications sector have relied on research done in other developing regions, such as Latin America and the Caribbean, 2 and conclusions are sometimes drawn from studies that lump Africa with countries from dissimilar regions. A study specific to Africa is warranted because of the continentÂ’s diverse cultural, ethnic, economic and political institutions, its persistent internal conflicts and its lack of development relative to other regions. By considering Africa alone, we reduce the chance of drawing conclusions that might be true in one developing region but not another. It is also easier to account for heterogeneity across countries in a single region than for countries in different regions. 1 I would like to thank Chunrong Ai, Sanford Berg, David Figlio, Grant Thrall, Jonathan Hamilton, David Sappington, Larry Kenny and Cezley Sampson for helpful comments on the earlier version of this study. Remaining errors of fact or interpretation are the authorÂ’s responsibility. 2 Some cross-national studies of developing areas, such as Singh (2000), Gutierrez and Berg (2000) and Wallsten (2001), either do not include data on Africa or they lump Africa with other emerging economies. 4

PAGE 11

5 Research such as that by Sachs and Warner (1997) concludes that the slower growth observed in Africa is not explained any differently from growth in other regions. More recently, however, Block (2001) found important differences in Africa compared to other countries. For instance, he found that lack of openness to trade affects growth more adversely in Africa than in other lowand middle-income countries. By extension, this finding supports the circumscribed study of Africa, so that policy prescriptions may be based on relations that are particularly applicable to Africa, as opposed to generic prescriptions that may not be appropriate. Cross-national studies on African telecommunications development have been difficult because of data limitations, since serious reform began only in the midto late 1980s. With data from thirty-seven African countries during 1985-91, this study helps fill a gap in research on Africa by using panel data techniques to perform a cross-national evaluation of the effects of institutions, political systems and cultural issues (proxied by legal systems) on telephone penetration rates. The results from the empirical analysis suggest that high per capita GDP is associated with improved telephone penetration but, by itself, is not as strong as one might expect on the basis of other cross-national studies. In addition, a strong institutional framework can enhance investment in telephone infrastructure. Such a framework involves a respect for property rights, which yields perceptions of contractual security and reduced likelihood of expropriation. Contrary to expectations, countries with similar institutional quality are likely to have higher access to telephones under the French legal system rather than the English common law code. Finally, a more democratic country is likely to have lower access to a telephone network than a less democratic country with similar characteristics. The results are important, both from the perspective of the private investor as well as the public operator. Consideration of demography and affordability on penetration rates is essential for the analysis of the cost effectiveness and potential profitability of investing in the region. From a policy perspective, it is important for the decision maker to understand the relation

PAGE 12

6 between the institutional environment, polity, culture and telecom investment in order to implement policies that create an environment conducive to telecommunications investment. The environment would include a regulatory system that benefits both providers and users of basic telephone service. Classifying the institutional and political issues sets the background for an exploration of infrastructure development in Africa from a more technical perspective. Such issues include the role of growing mobile access in countries with relatively low penetration rates. I analyzed this issue in Hamilton (2002) by modeling the endogeniety between main line and mobile telephones. Despite the influx of new telecommunications services like mobile technology, voice mail, call waiting, as well as increased access to fixed-line telephony, access to basic telecommunications 3 in Africa is still very limited (Kerf and Smith, 1996). Investment in basic telephony in Africa is far below the level in Latin America, the Caribbean and Asian Pacific regions. In 1985, AfricaÂ’s penetration rate was, on average, 10.45 per 1,000 people, more than five times lower than that of Latin America and the Caribbean. By 1997, despite a growth rate of 147 percent, AfricaÂ’s access rate remained more than five times lower than in Latin America and the Caribbean. Teledensity in Africa has been low partly because of relatively low income levels. Between 1980 and 1990, Africa suffered a decline in its already low income levels. Between 1985 and 1997, per capita income in Africa was, on average, US $915, compared to more than US $2,000 in Latin America and the Caribbean and US $5,000 in Asia. Although low, teledensity has been increasing in Africa and reached 25.82 telephones per 1,000 people in 1997. This average (unweighted by population) is skewed upward since many countries have much weaker performance. If Seychelles, with an access rate of 203.64, Mauritius with 227.65 and 3 Basic telecommunications, telephony, telephone access, penetration and network access refer to both mobile and main line telephones unless otherwise specified.

PAGE 13

7 15.66 in 1997. So despite recent innovations in the sector and general economic growth, access to telecommunications in Africa is still limited. As in most developing regions, telephone lines are concentrated in the cities, with only limited access for rural areas. Nevertheless, despite the poor quality of infrastructure compared to other developing countries, the opportunities for telecommunications development in Africa are substantial. While there has been a tremendous amount of investment and reform elsewhere in the world, Africa was largely ignored until recently. The sector is currently predominantly stateowned, but some governments have embarked on reform programs, most of which involve two elements: gradual commercialization by separating operational management from government ministries and the transfer of responsibility for regulation away from government ministries to independent agencies. Privatization options being considered include public offers for sale to financial institutions, sale to private investors and employees, private sale to strategic investors, or divestiture and management contracts with foreign operators. The change occurring in the region is often obscured by political constraints that limit a governmentÂ’s desire or ability to make policy commitments to promote the development of the sector (Mustafa et al. 1997). Analysts like Kerf and Smith (1996) argue that special attention must be given to establishing stable and independent regulatory agencies that can provide credibility for investors, legitimacy for consumers, and more efficient sector performance. The creation of suitable regulatory systems is important because the success of the restructuring process depends heavily on the credibility and consistency of that reform. After surveying recent studies, I outline several hypotheses and then briefly review the state of telecommunications in Africa. The basic data are examined in Section 5, followed by the model development and estimation. The concluding section summarizes the results.

PAGE 14

8 2.2 Recent Research Research concerning the influence of political stability on investments has a long history. Three decades ago, Bennett and Green (1972) sought to identify the role of politics by testing the hypothesis of a negative relation between the allocation of U. S. marketing investment throughout the world and the level of national political instability in different countries. They also tested whether such a negative relation existed only in less developed countries. They found that political instability does not discourage investment in marketing activities. If this finding were applicable to infrastructure investments, then political factors could be viewed as playing only a minor role in utility sectors; however, irreversible fixed costs makes these capital-intensive sectors sensitive to political instability. Studies of infrastructure since 1972 have found that political conditions do, in fact, influence the level of investment that takes place abroad (Haan and Siermaan 1995, Bergara et al. 1998, Henisz 2000, Svensson 1998). Research has shown that (telecommunication) investments will be larger when there are strong political constraints on government officials. 4 Knack and Keefer (1995,1997), Guiterrez and Berg (2000), Henisz (2000) and Singh (2000) are among studies finding a relation between the institutional environment, economic growth and infrastructure investment. They have shown that the greater the policy 4 The framework for utility investment is more consistent and predictable when there are more independent checks on governmentÂ’s executive power. According to Svensson (1998), political factors affect infrastructure development, but the effect is indirect in that the political environment merely provides the channel through which private investment flows. He found that private investment is restricted when those in power lack incentives to undertake legal reforms that protect property rights and encourage investment. Weak property rights lead to the reallocation of resources away from taxable activities, which reduces future governmentsÂ’ tax revenues and ability to spend. In not reforming the legal system, the current government affects future governments. By protecting its current constituency, the government neglects the provision of a favorable environment for investment. Haan and Siermaan (1995) stressed that, although the relation between democracy and economic growth is not robust, high levels of economic growth cannot take place in an environment where democratic rights are repressed. Since investment contributes significantly to the rate of economic growth, the same relation might be expected between investment and democracy.

PAGE 15

9 uncertainty, the smaller the level of investment. Among poorer countries, those with stronger institutional bodies that enforce law and protect property rights have better prospects for private investment and increased payoff for public infrastructure investment. Considerations other than political and institutional ones have been identified as determinates of infrastructure investment, including economic indicators that capture the standard of living and other features of the economy. Higher income levels make foreign investment more attractive, as does expected growth and predictability of that growth. Alesina and Perotti (1996) looked at the more complicated question of how the distribution of income may determine the level of investment. Using data from 1960 to 1985, they found that income inequality increases sociopolitical instability by fueling discontent. This in turn creates uncertainty in the political and economic environment, which reduces investment. Applying the “catch up” theory of growth, Antonelli (1993) provided an explanation for the ability of some poor countries to catch up. By taking advantage of technology in advanced countries, latecomers can experience rapid growth in telecommunications development. He concluded that new investment plays a significant role in the diffusion of advanced telecommunications in countries with high rates of growth in GDP and telecommunications infrastructure. 2.3 Hypotheses 2.3.1 Institutions and Network Access A positive correlation is expected between network access and institutions. Institutions are the formal and informal rules that guide human interactions, whether they are social, political or economic (North 1990). They reduce uncertainty by establishing a stable framework for human relationships. Institutions involve not only rules but also enforcement, which usually involve the state. Investment is likely to be encouraged in an environment where participants understand the rules of the game and where the risk of losses is minimized (North 1986). Rules

PAGE 16

10 and regulations that guide the telecommunication sector protect both the private operator and the public manager. Private investors appreciate strong institutions that discourage governments from reneging on promises. Investment in telecommunications involves the commitment of large sunk costs, and private investors are exposed to the risk of expropriation of their property (Levy and Spiller 1994). The same point applies to government investment. Ministries responsible for investment are less likely to be able to raise capital (from the national budget or through the issuance of project-specific bonds) if political instability is high. Why go to all the effort of planning and coordinating major new initiatives when the next government might reverse the process or give the construction contract to a political supporter? Shifting patronage and civil strife do not provide a firm foundation for long-term government investment. Furthermore, if prices are below cost, additional subscribers just mean a larger deficit. Theories of political economy suggest that the political and institutional environment in Africa has not been conducive to telecom investment (Goldsmith 2001). Governments that tend to be unstable characterize Africa for the most part, and government policies are often unstable as well. South Africa is an example of how stronger institutions can affect telecommunications investment. Since 1990 and the establishment of political freedom and stability in South Africa, many parts of the nation that previously had minimal access to telephones now have relatively good access. The state -owned operator is committed to privatization of the industry and to increased access for citizens (International Telecommunications Union [ITU] 1998). The transformation in South Africa highlights the interdependence of institutions and organizations. As discussed by North (1990), economies perform differently over time not just because of institutions and organizations themselves, but also because of their interactions, which determine the direction of institutional change. A significant improvement in the development of AfricaÂ’s telecom infrastructure is therefore conditional on strong credible political institutions

PAGE 17

11 that protect against illegitimate rent seeking and create safe and healthy environments for investment. 2.3.2 Political Systems and Network Access To control for the effect of political structures and party system, I introduce a proxy for democracy taken from the Polity III Democracy Index 5 described in Section 5. Theory suggests a direct relationship between democracy and investment. According to this view, the democratic process and the existence of civil liberties generate conditions most suitable to economic development. In a democracy, political actions are accountable to the press and the public; therefore, a governmentÂ’s ability to manipulate policies to suit political ends may be limited (Haan and Siermann 1995). This argument is particularly applicable to investments in telecommunications, which cannot be easily redeployed. In countries with a non-democratic system of government, the risk associated with investment is high unless the government can make a credible commitment not to expropriate capital assets or unduly limit returns. 6 Over the sample period, state-owned firms provided most telecommunication services in Africa. 2.3.3 Cultural Considerations and Network Access Although people and regions in Africa are diverse, a common history of colonization is unifying. Colonization remains a legacy in the form of cultural, social and political influences of foreign powers that help to define these countries today. Often the laws and norms of a country draw on those of the former colonial power. According to cultural theories of institutions, a societyÂ’s behavior, its actions and governments are shaped by its beliefs and shared values. 5 The democracy score of the Polity III index is based on five components: The competitiveness of political participation, weighted 0.3; the regulation of political participation, competitiveness and openness of executive recruitment, each weighted 0.1; and constraints on the chief executive, weighted 0.4 (Jaggers and Gurr 1995). 6 Goldsmith (2001) found that African countries with less political risk tend to have more open governments and low levels of political corruption. This results in better functioning of the democratic process and encourages leaders to pursue policies that are less shortsighted.

PAGE 18

12 The efficient functioning of a society requires that individuals (economic agents) believe that their institutions are credible and efficient (La Porta et al. 1999). One aspect of the cultural norms of a country is its legal system. Countries that have an interventionist legal system restrict the ability of investors to buy, sell and engage in efficient contracting, which affects risks and the return on investments. The view of the new institutional economics supports an economic environment in which the government is relatively non-interventionist, providing a promising environment for investment (Knack and Keefer 1995). Using these theories, I hypothesize that the impact of the institutional variables in the model will vary, depending on the legal system in each country. With this in mind, the data were divided according to legal code. 7 Incorporating the legal system as a variable allows the countries in the sample to be divided according to the traditional tendencies of the government regarding intervention. LaPorta et al. (1999) argue that French civil law tends to be used by the state as a means of expanding its power and offers relatively less protection for individuals than British common law. Governments that utilize the French civil code are therefore expected to be more reluctant to reduce their control in the telecom sector, which may be regarded as a key area that has to be guarded closely. 8 Because British common law traditionally offers more protection for individuals and limits the powers of governments, a country with French legal traditions is expected to have lower access to telephones inasmuch as control hampers investment and lowers the efficiency of sector performance. 7 Specifically, I divide the countries into three groups: (1) countries that utilize the English common law legal system, (2) those that utilize the French civil code and (3) all other countries. 8 So far, there is no evidence that countries with French civil legal systems are less likely to privatize their telecommunications network. Up to 1 997 the number of countries with a tradition of French civil legal background that had privatized (or promised to privatize) exceeded those of English common law background by one country.

PAGE 19

13 2.4 The State of Telecommunications in Africa Tables 2. 1 and 2.2 trace the development of telecommunications in Africa from 1985 to 1997. Many African countries are now concerned with telecom infrastructure for the same reasons that they limited foreign and private ownership in the past. Telecom investment is strategic and can contribute to a countryÂ’s economic development. Most African countries, however, still require a substantial increase in investment in telecom infrastructure to even catch up with other developing regions. Table 2.1: Network Access. AfricaÂ’s Place Relative to Other Developing Regions (19851997) Region 1985 1991 1997 Africa 10.4 15.0 25.8 Latin America and the Caribbean 51.4 68.2 135.8 Asia 83.5 111.6 234.7 Table 2.2: Growth in Network Access: Africa Relative to LAC and Asia (1985-1997) Region 1985 1991 1997 Africa 4.6 (44%) 10.0 (63%) 14.6(145%) Latin America and the Caribbean 16.8 (33%) 61.2 (82%) 84.4 (146%) Asia 28.1 (34%) 111.0 (90%) 151.2(181%) Note: The first number in each cell represents growth in telephone access per 1,000 individuals; the percentage change is in parentheses. Over the period 1985-97, Africa consistently displays lower network access than Asia and Latin America and the Caribbean. In 1997, access in Africa was more than five times less than in Latin America and the Caribbean and more than nine times less than in Asia. Despite growth at an increasing rate over the years, Africa has failed to catch up with Latin America and the Caribbean and Asia, where access has also grown. However, the rate of growth indicates that Africa has recognized the importance of developing its telecom sector with both mobile and wire line provision. As part of their commitment to increase telephone access in Africa, some countries have agreed to allow some degree of private participation in wire line provision. Table 2.3 reflects this change, showing that it is no longer a foregone conclusion in Africa that government must operate telecommunications in order to meet national objectives. An asterisk indicates those

PAGE 20

14 countries that have made a commitment to privatize even though ownership remains with the state thus far. 9 Table 2.3 Ownership of Wire Line Serv ice i n Af rica Country Ownership (Amount and year privatized) Ownership (Amount and year privatized) Algeria 1 00% state-owned Morocco* 1 00% state-owned Botswana* 1 00% state-owned Niger* 100% state -owned Cameroon 100% state-owned Nigeria 100% state-owned Congo* 1 00% state-owned Rwanda 1 00% state-owned Cote dÂ’Ivoire Privatized (51%, 1997) Sierra Leone 1 00% state-owned Egypt* 1 00% state-owned South Africa Privatized (30%, 1997) Gabon Privatized (39% by 1998) Tanzania Privatized! by 1998) Ghana Privatized (30%, 1996) Togo 100% state -owned Kenya* 100% state-owned Tunisia 100% state-owned Madagascar Privatized (34%, 1995) Uganda* 100% state-owned Malawi 1 00% state-owned Zambia* 1 00% state-owned Mali* 1 00% state-owned Zimbabwe* 1 00% state-owned * indicates countries that have committed to privatize in the near future. Source: Leblanch-Woher and Lewington (2000). Many socialized entities have not performed well. The extent of investment required in developing countries is usually too large and expensive for the government to manage on its own, which is one reason infrastructure (telecommunications) development in Africa has been slow and sometimes nonexistent. Indeed, there is evidence of strong performance following privatization in many developing nations, including those in Latin America and the Caribbean (Guiterrez and Berg 2000). Once privatization is accomplished, modernization and development can increase the efficiency and availability of service. Although telecommunication infrastructure is largely stateowned in Africa, some countries are in the process of reforming the sector, following world trends as well as demands by international lending agencies. Joint ventures are a typical first step. The extent to which governments commit to privatization and sector development depends 9 Even those countries reluctant to give up control of wire line provision have allowed some degree of private participation in cellular provision. For more on cellular privatization and competition, see Hamilton (2002), which analyzes the role of mobile competition in the development of fixed line telephony.

PAGE 21

15 on economic and political considerations, as well as the risk environment of the country. The increased commitment to private investment shown in Table 2.3 indicates that African countries have come to accept that there are benefits from privatization. Some of these include increased services and quality of service, as well as improved access at lower cost and the availability of additional capital and management skills. 10 Even when the privatized entities are essentially monopolies, consumers can benefit from reduced prices with proper regulation of monopolistic power (Noll 1999). The threat of potential competition may be enough to induce telecom investment. 2.5 The Data 2.5.1 Defining the Data The 1985-97 study period was determined by constraints on the data available for some of the variables. Definitions of variables and sources of the data are given in Table 2.4. Table 2.4: Dei Initions and Sources Variable Definition Source NETWORK Is the dependent variable. It is the (# of main telephone lines/population)* 1,000 (# of cellular subscriptions/population)* 1 ,000. World BankÂ’s World Development Indicators (1999). Economic and demographic variables: LOGGDPC The natural logarithm of a countryÂ’s per capita gross domestic product (GDP) lagged one period. The base year is 1990. World BankÂ’s World Development Indicators (1999). LTRADE Imports plus exports as a fraction of GDP lagged one period. World BankÂ’s World Development Indicators (1999). URBAN Urban population/total population. That is the percentage of the total population that resides in urban areas. World BankÂ’s World Development Indicators (1999). 10 For instance, Ghana has experienced noticeable improvements in telecom penetration since liberalization and privatization in the midto late 1990s. The number of direct lines increased by 26% in 1997 alone, compared to growth of 2-3 %in the early 1990s (Frempong and Atubra 2001).

PAGE 22

16 Table 2.4 Continued Variable Definition Source Institutional variables: CORRUP Corruption within the political system on a scale of 0 to 6. Larger values indicate less corruption. IRIS-3 file of International Country Risk Guide (ICRG) data, 1 982-97, constructed by Stephen Knack and the IRIS Center, University of Maryland. LAW Index of law-and-order tradition. It measures the degree to which citizens of a country are willing to support established institutions to make and implement laws and adjudicate disputes. The index ranges from 0 to 6. A high point indicates a strong law-and-order tradition. IRIS-3 file of International Country Risk Guide (ICRG) data, 1982-97, constructed by Stephen Knack and the IRIS Center, University of Maryland. BUREAU Index of bureaucratic quality on a scale ranging from 0 to 6. It measures the extent to which a countryÂ’s bureaucracy is able to govern without drastic changes in policy, or interruption in government services. High index values indicate a strong bureaucracy. IRIS-3 file of International Country Risk Guide (ICRG) data, 1982-97, constructed by Stephen Knack and the IRIS Center, University of Maryland. CONTRACT The risk to foreign businesses, contractors and consultants that government will modify a contract in the form of repudiation, postponement or scaling down. The index ranges from 0 to 10. A high point index signifies less likelihood that a contract will be modified. IRIS-3 file of International Country Risk Guide (ICRG) data, 1982-97, constructed by Stephen Knack and the IRIS Center, University of Maryland. EXPROP The risk of expropriation of private investments in terms of outright confiscation or forced nationalization. The index is rated from 0 to 10. Higher index points signify less likelihood of investment expropriation. IRIS-3 file of International Country Risk Guide (ICRG) data, 1982-97, constructed by Stephen Knack and the IRIS Center, University of Maryland. ICRG Following Knack and Keefer (1995), I created a 50-point index from five of the ICRG variables. CORRUP, LAW and BUREAU were converted to a 1 0-point scale by multiplying each by 5/3. These were summed with CONTRACT and EXPROP. IRIS-3 file of International Country Risk Guide (ICRG) data, 1982-97, constructed by Stephen Knack and the IRIS Center, University of Maryland. Variables indicating party system (freedom measures) and legal legacy DEMOC Democracy, an indicator of regime type on a scale of 0 to 10. This data is assumed to be constant between 1994 and 1997. Polity III: Regime type and political authority, 1800-1994, Jaggers and Gurr (1996).

PAGE 23

17 Table 2.4 Continued Variable Definition Source DEMOCAUTO A single summary measure of political regime, this variable is calculated by subtracting the index of autocracy score from the democracy index score. It is measured on a scale of-10 to 10, where, 0 to -10 indicates autocratic regimes and 1-10 indicates democracies. Polity III: Regime type and political authority, 1800-1994, Jaggers and Gurr (1996). POLRIGHT From the Freedom in the World Survey, 1998-99. POLRIGHT is an index that measures the extent to which people participate freely in the political process or the rights of all adults to vote and compete for public office. The index is measured on a scale of 1 to 7. Smaller numbers represent greater freedom. From the Freedom in the World Survey, 1998-99. ENGINST ENG*ICRG, where ENG is a dummy variable equal to 1 if the country has an English common law tradition and 0 otherwise. Legal system is taken from CIA World Factbook, 1999. OINST 0THER*ICRG, where OTHER is a dummy variable equal to 1 if the country does not utilize English or French legal systems and 0 otherwise. Legal system is taken from CIA World Factbook, 1999. ENGPOL ENG*DEMOC-AUTO. Legal system is taken from CIA World Factbook, 1999. OPOL OTHER*DEMOC-AUTO. Legal system is taken from CIA World Factbook, 1999. 2.5.2 Descriptive Results Table 2.5 summarizes the variables used in the models. The differences in means vary across the samples. Confirming the hypothesis that countries with a history of the French civil code tend to be more interventionist, potentially restricting the ease of investments, indicators of political freedom are stronger when the legal system is based on English rather than French law. The negative sign on the difference in POLRIGHT indicates stronger political freedom in the English nations, which overall tend to have a stronger institutional framework (based on ICRG), but not significantly so. Of the ICRG measures, only LAW and BUREAU have significantly different means across the sub-samples.

PAGE 24

18 Table 2.5 Sample Descriptive Statistics, 1985-97 Variable Mean and Standard Deviation. Full Sample (481) Mean and Standard Deviation. English Common Law Countries ( 1 69) Mean and Standard Deviation. French Civil Code Countries (182) Difference in Means Across Legal Systems NETWORK 13.573 (20.128) 14.669 (25.294) 12.150 (15.203) 2.519 LTRADE 60.501 (27.373) 56.251 (25.124) 61.088 (23.828) -4.837*** URBAN 34.276 (15.720) 30.129 (10.730) 38.733 (16.491) -8.604* CORRUP 2.827 (1.105) 2.914 (1.010) 2.928 (0.851) -0.014 LAW 2.679 (1.168) 2.764 (1.070) 2.796 (0.877) -0.032 BUREAU 2.629 (1.118) 2.715 (1.269) 2.758 (0.939) -0.043 CONTRACT 5.098 (1.784) 5.342 (1.756) 5.360 (1.541) -0.018 EXPROP 6.081 (1.901) 6.491 (1.781) 6.227 (1.676) 0.264 ICRG 24.737 (7.171) 25.822 (6.922) 25.725 (4.851) 0.097 POLRIGHT 5.325 (1.577) 5.231 (1.516) 5.335 (1.368) -0.104 DEMOC 0.353 (10.812) 1.982 (3.123) 1.115 (2.180) 0.867 DEMOCAUTO -2.933 (5.429) -2.663 (5.805) -3.972 (4.489) 1 309*** Standard deviations are in parentheses; * = significant at 1%,** = significant at 5%,*** = significant at 10%. Mean network access (unweighted by population) is only 1 .2 times greater in nations utilizing English common law, which is not statistically different from the French countries. The results in Table 2.5 suggest that the difference in institutional factors across regions is not enough to create a difference in telephone access. LOGGDPC, LTRADE and URBAN are all significantly higher in nations utilizing the French civil code. Taking this into consideration along with the information on institutions, I view the raw data as suggesting that countries with stronger institutions tend to have higher access to telephones, even when their economies are less open and they have lower per capita income and a relatively small urban population. The

PAGE 25

19 difference in institutional quality (ICRG) is, however, not big enough to create any significant differences in network access. Table 2.6 provides summary statistics based on differences in legal tradition and institutional quality. An ICRG index above the sample average of 24.74 defines high institutional quality, while countries scoring below this average are considered to have low institutional quality. The numbers in parentheses indicate percentage growth in network access between 1985 and 1997. Table 2.6: Regional Differences, Institutional Quality and Network Access. 1985-1997 Legal system Institutional Quality High Low Overall English Common Law 21.30, (128) 8.01, (44) 14.66, (118) English minus South Africa 10.99, (153) 8.01, (44) 8.09, (126) French Civil Code 13.45, (172) 9.94, (83) 12.20, (146) French minus Arab Nations 7.19, (134) 3.82, (63) 5.93, (124) All Countries 16.99, (133) 5.49, (78) 13.57, (112) As expected, countries with stronger institutions on average have higher access to telephones. In this sample, countries with high institutional quality have 3.7 times greater access than countries with low ICRG scores. This result conforms to expectations, unlike the observations across English and French legal origins. According to Chong and Zanforlin (2000), countries with a system based on the French civil code appear to have lower institutional quality. This suggests that countries of French origin in our sample should have lower access to telephones. However, the data in Table 2.6 suggest that this is not true for the sample of countries used in this analysis when institutional quality is low. Growth in telephone access is also higher in French nations, even when countries of English common law origin have higher penetration levels. The implication of Chong and Zanforlin (2000) can be extended only when institutional quality is high. In this case, countries with an English common law tradition have higher network access than countries using the French civil code. This suggests that legal origins may influence institutional quality, but the legal system itself has less of an influence on telephone access than institutional quality.

PAGE 26

20 Countries with strong institutions have higher access than those with weaker institutions, regardless of their legal system. The difference in access across legal origins is at most 9 per 1,000 individuals when institutional quality is either high or low, but the difference across institutional qualities is a little over 1 1 per 1,000. When South Africa is excluded, the growth rate in the English common law countries increased so that the difference between the two groups of countries was negligible. This suggests that the countries with low access, relative to South Africa, will eventually achieve comparable access levels. Table 2.7 compares access on the basis of differences in countriesÂ’ legal tradition and polity. The entire sample period, 1985-97 is used when necessary, but the last five years, 199297, is used when this better reflects the political tendencies of a country, since many countries that were autocratic in earlier years are adjusting to a polity with more democratic characteristics. In addition, a significant amount of the development in telecom happened during the last five years of the sample period. Following Jaggers and Gurr (1995), countries are defined as democratic if the autocratic index subtracted from the democratic index is greater than or equal to 1. Otherwise, the political system is defined as autocratic. Countries scoring 1-6 are primarily democratic but put some limits on political participation and civil liberties (e.g., Zambia and Ghana). Scores above 6 represent highly democratic nations, such as Botswana, Madagascar, and South Africa. Countries scoring 0 to -6 are autocratic in nature but allow some degree of political freedom (e.g., Egypt and Zimbabwe) while those scoring -7 and lower are fully autocratic (e.g., Cameroon). Many African countries fall in this penultimate category. In general, democratic nations tend to have higher access than non-democratic ones, but the difference in growth is very small. This is attributable entirely to countries using the English common law code.

PAGE 27

21 Table 2.7: Regional Differences, Political Systems and Network Access Legal Legacy Political System Democratic Non-Democratic All English Common Law 24.23, (127) 8.67, (104) 14.66, (118) English without South Africa 6.91, (198) 8.67, (104) 8.09, (126) French Civil Code 10.62, (105) 12.76, (143) 12.20, (146) French without Arab Nations 4.55, (130) 6.41, (81) 5.93, (124) All Countries 13.93, (126) 10.83, (129) 13.57, (112) Note: Numbers in parentheses are percentage growth in network access between 1985 and 1997. The slightly higher growth in non-democratic countries results from a higher growth rate in nations using the French civil code. When the government is democratic, countries under an English legal system have significantly higher access than those under the French system. The opposite is true when the political system is not democratic. One implication of this is that countries under the French civil code may have less experience working under democratic systems. Indeed, leaders can make more efficient decisions regarding telecom development under a system that is more authoritarian, and the benefits from investing in telecommunications are more likely to accrue to the politician (Goldsmith 2001). The role of legal systems is further explored in the estimation of the model. 2.6 The Model NETWORK (per 1,000 people) is the dependent variable used to estimate the effects of institutions, political regime and legal origin on investment in basic telecommunications in Africa. Although data on dollar investment in telecommunications is sometimes directly available, I decided to use telephone lines per capita as the dependent variable because per-capita dollar investment can be a misleading indicator of the degree of accessibility. For example, a $100 investment in a country with rugged terrain is likely to generate fewer telephones than in a country where it is easier to lay telephone cables. Since the data set consists of countries with different (and often unique) characteristics, this measure would be more useful if the research focus was on cost effectiveness rather than on access. NETWORK per capita is therefore a more suitable measure.

PAGE 28

22 Because the data used in this study spans the period 1985-97, there are multiple observations on each country and the explanatory variables are time-series related. That is, they vary over time. To take account of this time-series component while maintaining the crosssection, both components are pooled to develop a data set consisting of 37 countries over the period 1985-97. The advantage of pooling the data is that it generates a larger number of data points, which increases the degrees of freedom and may reduce colinearity among independent variables (Hsiao 1986). Cross-sectional estimation provides information relevant to a single time period, but this study is more concerned with cross-variation over time. By pooling the data and applying panel data techniques, I am able to make references about cross-variation of variables over time. Equations 1 is specified to accommodate the pooled data as follows: (1) NETWORK CT = a + J3 K X CT +e CT , Where X CT is a vector of independent variables for country C in year T. These variables are LOGGDPC, LTRADE, URBAN, CORRUP, LAW, BUREAU, CONTRACT, EXPROP, and DEMOC. All these variables are defined in Table 3. e is a random error term. 2.6.1 Omitted Variables Since the data represent a panel of different countries, it is highly likely that unobserved differences across countries will affect telephone investment. If these unobserved variables are not accounted for, the interpretation and use of the estimated equation may be unreliable (Studemund 1992). Omitted variables will cause a bias in the estimated coefficients of the other explanatory variables. For example, the coefficient of LTRADE in Equation 1 represents the change in NETWORK caused by a one-unit change in LTRADE, when the values of all other right-hand-side variables are held constant. If a variable is omitted, it is not included as an independent variable and is not held constant for the calculation and interpretation of the coefficient of LTRADE C tThus, some of the variation in the omitted variable will be incorrectly attributed to LTRADE. The efficiency and experience of multi-national operators assisting the

PAGE 29

23 local companies is an example of omitted variables. A more experienced operator is likely to provide better training to managers, technicians and engineers as reform of the sector takes place (Henisz and Delios 2002). Privatization and regulatory reform are other variables that could be included in the model. Privatization of wire line provision, however, started only in the midto latel990s for the countries that have committed to privatize. Agencies dedicated to the regulation of telecommunications in the context of privatizations were only in the formatory stages for most countries toward the end of the sample period. As a result, enough observations of privatization and regulatory reform did not exist throughout the sample period. A country-specific constant ac is introduced to account for the effects of those omitted variables that are specific to individual countries. It is also possible for effects that differ across countries to exhibit variation through time. For example, telephone penetration in each country may be affected by the unobservable worker quality (ability), which may vary from country to country. Countries with a more efficient or able workforce or provider are better able to make telephones accessible than less efficient ones. Worker ability is also likely to vary through time. Worker quality may improve through education, for instance, or through learning by doing. The trend through time is thus likely to affect penetration. Year effects (S T ) are introduced to account for such trends through time. The new equation is: (2) NETWORK CT -a c + (3 K X CT + 8 r + e ct , The fixed-effect model is estimated by using deviations from the group mean. In so doing, Equation 2 will pick up any constant differences occurring on a country-specific level. Finally, the standard errors are adjusted to account for the systematic correlation across units in the same group.

PAGE 30

24 2.6.2 Alternative Specifications Equation 2 is estimated by using two specifications of institutional variables and three alternatives to measure political freedom in order to check the sensitivity of the results to alternative specifications. Equation 2 is first estimated with the components of the ICRG index (CORRUP, LAW, BUREAU, CONTRACT, EXPROP) as individual independent variables. They are then treated as a single variable by using the average of the five components and thus aggregating the index to avoid possible problems of correlation between the individual components. The aggregate ICRG index, however, assumes that some factors that affect the institutional environment are more important than others, and so must be weighted more heavily. The problem with this approach is that different conditions across countries may warrant component weights that vary from the ones imposed by the index. The implicit weights may also vary through time, so the aggregate ICRG index may contain considerable measurement error. Using the individual components of the index provides a check against potential biases that may result from the use of subjective component weighting in the aggregated index. Equation 2 is also estimated with three different political variables. First, DEMOC is used to capture the degree of political freedom enjoyed by a country. This is then replaced with POLRIGHT so results can be compared to check whether the measure of freedom used makes a difference. While the two sets of variables are constructed to measure the freedom of a country, there may be noticeable differences in what is captured in reality. The POLRIGHT index does not rate governments, but rather the rights and freedoms enjoyed by individuals in each country. The DEMOC index rates governments by measuring factors such as constraints on the chief executive and the openness of executive recruitment. Differences in the construction of the two measures of freedom make it likely that the economic freedom measure contains some information that is not in POLRIGHT. DEMOC-AUTO is the third measure of political freedom. The first measure (DEMOC) is rated on a scale of 0 -10, where 0 denotes limited democracy and 10 denotes a high level of

PAGE 31

25 democracy. In this use of the democracy index, a low level of democracy implies high autocracy. Yet, if low democracy correlated perfectly with high autocracy, then Ghana, with a democracy score of 3, should have autocracy score of around 7, rather than 2. This kind of separate measurement of democracy (DEMOC) and autocracy (AUTO) is difficult to interpret when democracy is not a close inverse of autocracy, so I follow the Jaggers and Gurr (1995) approach and adopt a single summary measure of political regime (DEMOC-AUTO) to replace DEMOC. Finally, to explore the argument that the legal tradition of a country may affect telephone development through its institutional framework, I introduce four new variables: ENGINST, OINST, ENGPOL and OPOL. These are defined in Table 3. Equation 2 is re-estimated using these variables in all the specifications discussed above. 2.6.3 Country Effect and Sample Selectivity Because the sample consists of only 37 of the 55 countries in Africa, the results may not be representative of the 18 excluded countries. To check for sample selectivity, I compared the countries in the sample with those excluded on observable factors such as network access and per capita GDP. Table A3 provides a summary of the comparison. Noticeable difference exist between the two groups of countries in network access per capita GDP as well as trade. The first two columns of Table A3 suggest systematic differences between the countries. It is also possible that some countries in the sample individually influence the results in a way that may not be true for other countries in the sample. For instance, South Africa stands apart from other countries in the sample because of its comparatively developed telecommunications sector (CIA World Factbook 1999). South Africa has the highest telephone penetration in the sample, and arguably sets the pace for sector development in the region. South Africa was among the first African nations to start the process of telecommunications reform in terms of modernizing and expanding services, introducing private participation and setting up regulatory bodies. South Africa has received much attention from the rest of the world as an example for other African countries. If other African nations view South Africa in a similar

PAGE 32

26 manner, then South Africa may be influencing the rest of the countries in the sample and thereby the regression estimates. 11 Most of the excluded countries are low-income countries with relatively low network access, but a few countries such as Seychelles and Mauritius have much higher income levels than most countries in the sample. Access rates in these two countries exceed that of every country in the sample. If we exclude these outliers, the difference in network access (between excluded and included countries) is just over 4 per 1,000 individuals. The difference in per capita GDP remains noticeable and falls in the lower income range for both included and excluded countries. I also compared regression results from the excluded countries with the entire group of all countries in Africa. Table A4 shows that the results are quantitatively the same except that LTRADE is not significant in the smaller sample. On the basis of these rough sensitivity analyses, I feel that the qualitative results of the sample can be extended to the eighteen excluded countries, at least in a general manner, and that most of Africa can be described accurately with the results of the estimations. 2.7 Estimated Results Table 2.8 shows the main effects of institutional factors on network development. Bureaucratic quality is significantly correlated with network development with an unexpected sign. The bureaucratic quality index may not be differentiating between the rules that define how the game is played and the players (the bureaucrats). Analysts often assume that the objective of 11 To account for this possibility, I re-estimated Equation 2 without South Africa and then sequentially removed every country (one at a time) to see if there was any influence on the results. The results where changes did occur are reported in Table A2 in the Appendix and show that Algeria, Burkina Faso, Democratic Republic Congo, Cote dÂ’Ivoire, Guinea, Guinea Bissau, Morocco, Mozambique, Uganda, South Africa and Zimbabwe may be individually affecting the Network regressions in that LGDPC became insignificant. In the case of Morocco, EXPROP also became insignificant. All specifications are re-estimated without these countries (reduced sample).

PAGE 33

27 bureaucrats is to make decisions that maximize their own individual utilities. Unless there are strong incentives for them to improve industry performance, they may not encourage it. Personnel in a strong bureaucracy recognize how certain decisions will affect them personally, while less efficient bureaucracies may not. In this case the index may be reflecting protection of the bureaucratsÂ’ position rather than the ability to protect contract and property rights. Stronger bureaucratic quality in this instance denotes greater personal protection. Generally, the results show that well-defined and credible institutions are positively and significantly correlated with network development regardless of the definition of institutions used. If a country such as Niger at the average level of adherence to the rule of law (LAW) were to increase this quality by 2 units to arrive at the level held by Botswana, then network development would increase from 1.6 telephones per 1,000 people to almost 2.2. This corresponds to an elasticity of 0.34 at the mean. A reduction in the risk associated with contract repudiation in Cote dÂ’Ivoire to South African levels would increase network access in Cote dÂ’Ivoire from 1 1.65 telephones per 1,000 inhabitants to 16.42. The results with the ICRG index explain less of the variation in NETWORK access (the R 2 is lower), and the coefficients tend to be different from the original regression (ICRG components). The results are, however, similar in that they indicate that strong institutions are important for NETWORK development. Per capita GDP is not always significant but other economic and demographic variables (LTRADE and URBAN) are significantly correlated with network development. The trade variable, however, has an unexpectedly negative sign, indicating that the state may not respond to economic forces in the way private individuals would. Bleaney and Greenway (2001) examined the impact of exchange rate volatility and terms of trade on investment and growth in subSaharan Africa and found that growth is adversely affected if a country specializes in the export of primary products. If this finding is applicable to infrastructure investments, it could explain the negative sign on LTRADE.

PAGE 34

28 Table 2.8: Panel Fixed-Effect Regression Controlling for Country-Level Differences Variable ICRG Components: Institutional factors measured by the individual components of ICRG ICRG index: The sum of CORRUP. LAW, BUREAU, CONTRACT and EXPROP is used as the institutional variable POLRIGHT: An index of political rights used to measure place of democracy DEMOCAUTO: Single summary measure of political regime, which is the difference between the index of democracy and the index of autocracy LOGGDPC 3.345 (1.527) 4 424 *** (1.865) 3.304 (1.658) 1.098 (0.893) LTRADE -0.108* (-6.328) -0.115* (-6-131) 0 . 100 * (-6.351) -0.098* (6.229) URBAN 0.697* (7.068) 1.053* (8.550) 0.739* (7.508) 0.813* (8.133) CORRUP 0.012 (-0.024) 0.027 (0.053) 0.070 (0.137) LAW 1.693* (2.769) 1.732* (2.922) 1.717* (2.820) BUREAU -3.666* (-3.169) -3.732* (-3.191) -3.595* (-3.099) CONTRACT 2.170* (5.086) 1.984* (4.867) 1.967* (4.497) EXPROP 0.334 (0.996) 0.437 (1.306) 0.546 (1.582) DEMOC -0.043** (-2.059) -0.036*** (-1.630) ICRG 0.613* (7.094) DEMOCAUTO -0.248* (4.565) POLRIGHT 0.538*** (1.884) N 421 421 421 421 R 2 (adjusted) 0.95 0.93 0.95 0.95 F-Stat 24.15 27.60 27.07 26.82 Note: The dependent variable is NETWORK, access to telephone lines per 1,000 people. The sample is 37 African nations over the period 1985-97. Since a test of the significance of a time trend could not be rejected, the time trend was excluded from estimations. * = significant at 1 %, ** = significant at 5%, *** = significant at 10%. The highly significant result on URBAN suggests that cost considerations with respect to location may be important for some countries. Greater access to telephones in the city may have a lot to do with its being cheaper to connect telephones to a network in urban as opposed to rural areas. This idea is less easy to understand in the context of cellular service provision, which

PAGE 35

29 entails less cumbersome and risky investments than wire line investments. Although cellular service is increasing in Africa, the bulk of the growth is in urban areas. The regressions with the POLRIGHT index of political rights and DEMOC-AUTO provide results similar to those for the ICRG components estimation, indicating that political institutions will have an impact on telecom development in the region. Contrary to our prediction, however, the coefficient on measures of democracy is usually negative (positive for POLRIGHT). The negative (positive for POLRIGHT) correlation makes sense if we assume the perspective outlined by Haan and Siermann (1995). The existence of checks and balances on executives can hamper development as politicians take actions to placate pressure groups at the expense of investment in infrastructure. In addition, most democracies in Africa are transitional and thus fragile. Table 2.9 shows the results when ENGINST, OINST, ENGPOL, and OPOL are included in the models. The results are qualitatively the same, except that LOGGDPC and EXPROP are always significant. POLRIGHT is no longer significant. The marginal effect of greater institutional quality on network access is increased when the countries involved have a tradition of French civil law compared to English common law code and other legal systems. This differs from expectations, but supports the results from the summary statistics when institutional quality is low to begin with. This result might also be picking up the effect of something else that was not captured in the model. It is difficult to define countries according to only three types of legal systems, since most countries often utilize some other major system, such as customary or tribal laws. While countries fall clearly in one or the other of the categories used in the analysis, it fails to capture these other aspects of the legal system. While the level of political freedom is important to network development, there is no evidence that it has different effects according to origin of the legal system (OPOL is significant only in one specification). In conditions like these, the demands of pressure groups may be met by succumbing to a “pork barrel” regime that generates

PAGE 36

30 Table 2.9: Panel Fixed-Effect Regression Controlling for Country-Level Differences and Accounting for Legal Traditions Variable ICRG components: Institutional factors measured by the individual components of ICRG. ICRG index: The sum of CORRUP, LAW, BUREAU, CONTRACT and EXPROP is used as the institutional variable. POLRIGHT: An index of political rights used to measure place of democracy. DEMOC-AUTO: Single summary measure of political regime, which is the difference between the index of democracy and the index of autocracy LOGGDPC 5.040** (2.106) 5.139** (1.999) 4.816** (2.055) 4.177*** (1.845) LTRADE -0.103* (5.662) -0.106* (-5.377) -0.095* (-5.722) -0.090* (-5.382) URBAN 0.614* (5.935) 0.970* (7.466) 0.649* (6.327) 0.739* (7.131) CORRUP 0.521 (0.990) 0.518 (0.988) 0.636 (1.232) LAW 2.068* (3.633) 2.046* (3.640) 2.085* (3.707) BUREAU -2.639*** (-1.855) -2.798** (-2.023) -2.761*** (-1.929) CONTRACT 2.418* (5.588) 2.268* (5.526) 2.220* (4.939) EXPROP 0.953* (2.639) 0.965* (2.722) 1.091* (3.003) DEMOC -0.042** (-1.957) -0.039*** (-1.730) ICRG 1.132* (7.189) DEMOCAUTO -0.271* (4.796) POLRIGHT 0.385 (1.378) ENGINST -0.487** (-2.442) -0.695* (-3.566) -0.447** (-2.342) -0.445** (-2.228) OINST -0.568** (-2.511) -0.764* (-3.542) -0.499** (-2.307) -0.516** (-2.423) ENGPOL -0.014 (-0.959) 0.010 (-0.587) -0.008 (-0.606) 0.005 (0.332) OPOL 0.025 (1.457) 0.001 (0.013) 0.026 (1.526) 0.048* (2.677) N 421 421 421 421 R" (adjusted) 0.95 0.93 0.95 0.95 F-Stat 20.71 20.91 21.60 22.30 Note: * = significant at 1%. ** = significant at 5%, *** = significant at 10%. Since a test of the significance of a time trend could not be rejected the time trend was excluded from estimations. Again, the dependent variable is access to telephone network per 1,000 people (NETWORK) with data from 37 African countries in the 1985-97 period.

PAGE 37

31 popularity at the expense of increased productivity. This kind of behavior can undermine (telecom) investment, especially when the majority of investments involve public undertakings. Tables 2.10 and 2. 1 1 show the results of the NETWORK regression when eleven countries (listed in footnote 10) are excluded from the sample. The results for the most part are as expected. LOGGDPC and EXPROP are now significant in every specification. The sensitivity test of country exclusion had shown that the inclusion of these eleven countries influenced the relationship between LOGGDPC and NETWORK to make it unexpectedly insignificant. With an estimated real GDP growth of 6.5% in 1999, Botswana has become one of the fastest growing nations in Africa, while most other African countries are experiencing little, if any, growth. For Botswana, GDP and telecom development have been progressing together, and so are expected to be correlated. Other countries have also been modernizing and expanding their telecom infrastructure despite an unimpressive growth in GDP. South Africa for instance, has the most modem infrastructure in the region, yet its growth in GDP has not been as impressive as BotswanaÂ’s. In 1999, South AfricaÂ’s estimated GDP growth was only 0.6 percent. The slow growth in GDP at the same time that network access is expanding could mean that the two are not closely related. This may be one reason that the influence of South Africa was making LOGGDPC insignificant. Although the size of the coefficients varied from those in the full sample, the results are qualitatively the same after taking into considering the exclusion of some countries. An exception is CORRUP and LAW, which changed significance in some specifications of the reduced sample.

PAGE 38

32 Table 2.10: Panel Fixed-Effect Regression Controlling for Country-Level Differences, Reduced Sample Variable ICRG components: Institutional factors measured by the individual components of ICRG. ICRG index: The sum of CORRUP, LAW, BUREAU, CONTRACT and EXPROP is used as the institutional variable. POLRIGHT: An index of political rights used to measure place of democracy. DEMOCAUTO: Single summary measure of political regime, which is the difference between the index of democracy and the index of autocracy LOGGDPC 17.338* (6.021) 15.788* (5.214) 16.652* (6.043) 15.404* (5.497) LTRADE -0.100* (-5.442) -0.089* (4.888) -0.085* (5-119) -0.079* (-4.972) URBAN 0.788* (7.433) 1.061* (9.411) 0.876* (8.425) 0.908* (8.250) CORRUP 0.678 (1.573) 0.843** (1.963) 0.839*** (1.931) LAW 0.284 (0.491) 0.250 (0.458) 0.434 (0.743) BUREAU -2.021* (-2.982) -2.005* (-3.169) -2.205* (-3.423) CONTRACT 1.033** (2.409) 0.723*** (1.787) 0.692*** (1.606) EXPROP 1.283* (3.588) 1.479* (4.105) 1.637* (4.493) DEMOC -0.051** (2.038) -0.052** (-2.114) ICRG 0.452* (5.098) DEMOCAUTO -0.305* (5.648) POLRIGHT 1.096* (6.032) N 280 280 280 280 R" (adjusted) 0.93 0.922 0.94 0.94 F-Stat 21.80 31.77 23.86 23.88 Note: * = significant at 1%. ** = significant at 5%, *** = significant at 10%. The dependent variable is access to telephone network per 1,000 people (NETWORK) during 1985-97. Since a test of the significance of a time trend could not be rejected, the time trend was excluded from estimations.

PAGE 39

33 Table 2.11: Panel Fixed-Effect Regression Controlling for Country-Level Differences and Accounting for Legal Traditions, Reduced Sample Variable ICRG components: Institutional factors measured by the individual components of ICRG. ICRG index: The sum of CORRUP, LAW, BUREAU, CONTRACT and EXPROP is used as the institutional variable. POLRIGHT : An index of political rights used to measure place of democracy. DEMOC-AUTO: Single summary measure of political regime, which is the difference between the index of democracy and the index of autocracy LOGGDPC 18.013* (6.341) 16.743* (5.514) 17.415* (6.326) 16.160* (5.775) LTRADE -0.097* (-5.142) -0.085* (-5.659) -0.086* (-4.973) -0.080* (-4,629) URBAN 1.487* (3.024) 1.575* (3.323) 1.565* (3.309) LAW 0.597 (1.048) 0.464 (0.841) 0.087 (1.422) BUREAU -0.868 (-1.412) -1.101*** (-1.726) -1.312** (-2.155) CONTRACT 1.815* (3.541) 1.438* (2.849) 1.388* (2.638) EXPROP 1.813* (4.399) 1.880* (4.677) 2.048* (5.016) DEMOC -0.047*** (-1.814) -0.051** (1.982) ICRG 0.851* (4.357) DEMOCAUTO -0.298* (-5.068) ENGINST -0.694* (-3.253) -0.511** (-2.313) -0.566* (-2.778) -0.562* (2.770) ENGINST -0.694* (-3.253) -0.511** (-2.313) -0.566* (-2.778) -0.562* (2.770) OINST -0.454** (-2.154) -0.383 (-1.576) -0.302 (-1.489) -0.423* (-2.053) ENGPOL -0.028 (-1.527) -0.036** (-2.344) -0.009 (0.715) 0.003 (-0.188) OPOL 0.038*** 0-715) 0.032 (1.562) 0.043** (1.963) 0.059** (2-514) N 280 280 280 280 R 2 (adjusted) 0.94 0.92 0.94 0.94 F-Stat 18.06 19.58 20.30 19.61 Note: * = significant at 1%. ** = significant at 5%, *** = significant at 10%. The dependent variable is access to telephone network per 1,000 people (NETWORK) during 1985-97. Since a test of the significance of a time trend could not be rejected, the time trend was excluded from estimations.

PAGE 40

34 2.8 Conclusions Previous studies have examined other developing regions, including Asia, Latin America and the Caribbean. It is important consider Africa as well and not depend solely on conclusions based on empirical work for other developing regions. The empirical results support the theory that credible institutions, including stable political structures, are important driving forces behind the surge of modernization in AfricaÂ’s telecommunications sector. The strong institutional results should serve as a signal to governments of the importance of creating and maintaining well-functioning political and regulatory institutions. To benefit from new technological innovations and competition, the institutional framework needs to adjust. Significant political reform will be required to mitigate risks to fragile and new democracies in Africa. The results also show that the origin of a countryÂ’s legal system is correlated with network development. Controlling for other factors, countries with similar institutional quality tend to have higher access to telephones if their system is derived from the French civil code rather than English common law and other legal traditions. More populated urban areas are associated with higher access to basic telephony. There may or may not be economies of density to be gained from providing service in a highly populated area. Generally, the existing low penetration levels, along with the generally small elasticities, suggest that African nations require huge adjustments in their investment climate to achieve access levels comparable to other developing countries in a short period of time. Nevertheless, with the present drive toward competition and privatization, changes can already be observed.

PAGE 41

CHAPTER 3 ARE MAIN LINES AND MOBILE PHONES SUBSTITUTES OR COMPLEMENTS? EVIDENCE FROM AFRICA 3.1 Introduction Mobile telephone subscriptions have been growing rapidly since the 1980s in both developing and developed regions. Subscriptions to fixed telephones have also grown, but at a slower rate than cellular in many regions of the world (ITU 1999). For instance, by 1997, mobile subscriptions in Lebanon accounted for 76 percent of total telephone subscriptions (World Development Indicators 2000). If cellular continues to grow rapidly, it is likely that its subscription will surpass fixed-line access in the near future. Currently, developing countries are experiencing the highest levels of mobile growth (ITU 1999). The increasing use of mobile telephony has implications for main-line access in developed countries and in regions where access to traditional wire-line telephones is relatively low. The growth of mobile subscription may reflect its role as a substitute for main lines. However, it is not uncommon for calls to be connected between a fixed line and a mobile telephone, so the services may in fact be complementary to each other. Despite the growing importance of mobile telephony, very little is empirically established regarding its position vis-a-vis fixed-line telephony as stimulus to connectivity. Since the fastest growth in mobile is occurring in developing countries, this study examines its role in fixed-line development in Africa, illustrating the impact of mobile provision in many developing regions. As one of the emerging markets, Africa shows high growth and increasing competitiveness in mobile communications. It is also a region with very low access to fixed-line telephones. During the 1980s, cellular provision was practically nonexistent in Africa. Today, virtually all countries 35

PAGE 42

36 in the region have access cellular service, and many have at least two operators, one of which is usually privately owned. These characteristics are evident in many developing countries. Typically, cellular usage and main line access are both growing rapidly. In many cases, regions with rapid growth in one service also experience rapid growth in the other (compared to regions with slower growth). Regardless of the pace of growth, or level of telephone access, mobile usually grows faster than main line access. For instance, South Africa enjoys growth in main line access of about 39 per thousand, per year, but enjoys growth of 4.5 per thousand, per year in mobile access. The trend is similar in cases where access to main-telephone lines tend to be relatively low. In Tanzania, main line access grew by less than 0. 1 per thousand people, per year, while mobile subscriptions increased by 0.3 per thousand each year. Looking at South Africa and Tanzania together, suggest that individuals may be using mobile, not necessarily as a substitute for fixed lines, but for other reasons such as joint use. In other words, as individuals become more sophisticated in their use of telephones, they tend to behave more like users in developed countries. While mobile growth in Tanzania may appear to be negligible, in actuality it is substantial, given that mobile usage in that country started only in 1995. By 2001, mobile subscriptions in Tanzania outweighed that of main line access. This story is also true for Morocco, which enjoys relatively high main line access. Between 1985 and 1997 main-line access grew by an average of 2.9 per thousand individuals per year, while mobile access increased by only 0.27 per thousand individuals per year. By 2001, however, access to mobile telephones surpassed that of main lines, at least in urban areas. This occurred as a result of the introduction of privatization and competition in the mobile sector. Mobile access, therefore, appears to be growing rapidly across the continent of Africa, regardless of the stage of main line development. For this reason, it is difficult to determine what the relation is between the two services by just comparing access trends.

PAGE 43

37 Today, virtually all countries in the region have access to cellular service, and many have at least two operators, one of which is privately owned. This is happening as a first step towards telecom sector privatization, since most countries, although allowing private participation in cellular services, still support state-owned main line provision. In these markets, the only potential threat to fixed line provision is competition from cellular. The question of the role of mobile is thus interesting in the case where the mobile provider is privatized and not owned by the incumbent. Real competition can only occur in a situation like this, because if the incumbent is also the sole provider of mobile services, then the potential for competition is diminished to the extent that the incumbent controls the mobile/main line trade off from the supply side. In this scenario, the relation in consumption would be difficult to determine. Even in cases where mobile provision is privatized and competition exist, conceptually it remains unclear whether mobile is a substitute or complement for main lines in consumption. A popular argument for the view that mobile usage is substituting for fixed lines is its prevalence where access to main lines is low or unreliable. 12 For instance, cellular phones may be an attractive alternative where it is difficult to install fixed-line networks. Because mobile networks can be installed more rapidly than fixed networks, they can alleviate waiting time for potential subscribers (Minges 1999) and reduce unsatisfied demand. The use of pre-paid cards by mobile users also supports the view that mobile is a substitute for main lines. With the use of pre-paid cards, users who otherwise would not qualify for a phone can now access the service. This is especially important for users in developing regions where it is not uncommon for people to lack credit histories. For individuals with poor credit histories, the option to pre-pay is an important development, since people are not 12 According to the ITU (1999) report, mobile telephones are used in developed countries to complement existing fixed lines but are emerging as a substitute for fixed lines in developing countries.

PAGE 44

38 automatically disqualified from using the service because of bad credit. The pre-paid alternative also provides the opportunity for individuals to manage their telephone expenses, since the number of calls that can be made is restricted. The result is that more people than ever now have access to mobile telephones. The increased supply of cellular service allows an attractive alternative to fixed-line telephony. Just like users in developed countries, some people in developing areas are attracted to mobile not because there are no alternatives, but because of the convenience of mobile phones (Frempong and Atubra 2001); furthermore, mobile phones are often used in conjunction with fixed-line telephones. Fixed lines are usually used at home, while mobile phones are used to keep in touch with home or office when individuals are on the road. In fact, calls from cellular phones are commonly made to fixed lines as opposed to another cellular phone (Jha and Majumdar 1999). In such instances, mobile and fixed lines are used in a complementary fashion. A larger fixed-line penetration increases the value of mobile service. The high cost of cellular service relative to the nominal price of main lines can make cellular phones unlikely to substitute for main lines. 13 Even though the cost of cellular has been falling, it remains high relative to the cost of fixed telephones. In regions where income is low, cellular phones may be out of reach for most consumers. In cases like this, mobile subscriptions may be confined to the wealthy, a relatively small group within a country. The preceding suggests that the role of mobile is ambiguous, at least at a conceptual level. Thus, the question of substitutability versus complementarity has to be solved empirically. 13 The price of main-line service has not generally been a strong factor affecting its demand because the price elasticities of demand for main lines tend to be low. Ahan and Lee (1999) found that the demand for mobile is positively correlated with per capita GDP, although the price effects on mobile tend to be weak. Their research suggests that even where income is low and costs high, subscribers may not respond very much to the price of mobile. If this is true, then the high cost of mobile to the subscribers should not be a strong enough factor to seriously restrict the number of individuals willing to become cellular subscribers. It is therefore feasible that individuals unable to access main lines would be willing to use mobile phones whenever they are available.

PAGE 45

39 3.2 Recent Empirical Studies A large body of research involving the development of telephone network has focused on the role of institutions. These researchers found that a strong institutional framework is essential for network development and expansion. Other studies examining the effects of new market conditions for utilities have focused mainly on industry privatization and competition. The general consensus is that competition and privatization tend to be associated with increased efficiency and growth. However, privatization is not always associated with network expansion.

PAGE 46

Table 3.1: Overview of Empirical Analyses of Telecommunications Development (International Studies) 40 3 O' U P u r£ oo P O' 12 — c i 03 ron 2p o uy OP C 0-h d 03 T3 0-. O'* a 03 JL3 a 03 o >% 03 •s "cj *— > 03 03 03 .g C K _o 4—1 1 c o _o -*-< 03 i_, 0 1 03 O 03 u c 03 H 03 c o S 3 CP H £ -*-» o *— » 03 o 03 _c e (2 d -a -o o 03 u c d o t-. O* cd 03 03 3 cr C/3 O § 3 cr C/3 ^ 03 I « CM £ y ° § 2z > 03 , JH £ s 2 c m oo as as as os *~~4 N^' >> 03 N 09 -o O 3 c/5 £ B 03 P £ § 2| o o rs 03 *03 N d 03 *0 'qj t: -a 0 o £ E 1 £3 y d o sc as os N o p ‘3 O as a as o oC

PAGE 47

41 c o CJ fi o U 2 ‘£| , \g c J 3 o 3 £ w> ~ D 4> c o *c« a G a. x Q* 4> o £ ° g o l-o! Kgg G O *55 o «P 4= T3 0) O >> JD G O U £ § £ O a) o W «§ .2: t £ -5 o u 3 O O T3 G ^ o H •a c 03 £ G £ '•§ E 3 Q o 4> 43 *-» o TD G3 03 _ l-i o o .4) C ta « 03 O, M _G C/3 U. o T3 ox) p Vh > 43 G .ss ^ 3 9 3 G C/2 O w 3 c ^ B -2 4= § C U 3~ u 3 -S,-a * •S C/3 4> 43 O a, 03 a. £ o -a o 4 = 15 a .© *B a/ cu | S jg *3 ’G ^ < £ U c < "O • , u « ^ .B i-J to b in oo o\ T Q o cj oo pj 2 o 1^3 ON ON CJ ,u „ « £ CJ © § -o On g ^ ON C X CJ3 V©' d "flS 00 'G G On ™ cd J G CX E c 03 G 4 ) > 4 > -a •o 3 3 o (N G a> ca £ ON O' Os 03 G E 3 •o d 03 03 40 o o o *-. O E C/3 © o o w 43 OX) G c/3 techniques that are typically used in empirical papers.

PAGE 48

42 All of these studies except Wallsten (2001) 15 examine the role of competition in basic services, rather than competition between mobile and fixed-line provision. Ros (1999) included some African countries in his data set to provide insight into the role of competition on main-line access as well as efficiency in the telecom sectors of these countries. The present study examines competition solely from the perspective of developing regions, using data from Africa. In addition, it recognizes the importance of mobile as a competitive force in a context where more competition occurs via cellular provision than via privatization of incumbent wire-line providers. The role of competition in the telecommunications sector is seldom viewed from the perspective of how mobile provision affects basic service. However, the astounding growth of mobile communications makes it one of the largest forces in emerging markets. A better understanding of whether mobile is a substitute for or a complement of basic services is important for policy makers as they attempt to reform the telecommunications market to improve access and increase efficiency. 3.3 Hypotheses 3.3.1 Mobile Competition and Main-Line Performance The relation between mobile competition and access to main line telephones is ambiguous. Mobile and main line correlation may be perceived as positive or negative, depending on whether they are complements or substitutes. When the two services are 15 Wallsten (2001) looked at the relation between mobile and fixed-line access, but the data used have two main drawbacks. The first is that it lumps together two dissimilar regions with different levels of growth and telephone development in terms of main-line access. It is likely that there are larger variations across these regions, which may affect how confidently the results are to be interpreted regarding any one region. In addition, only 15 of the 55 countries in Africa were included in the sample, and five of those had monopoly mobile providers. These monopolies are usually at least partially owned by the incumbent wire-line firm. A single country from Northern Africa was included, which raises a question as to whether the sample was representative of the African region as a whole. In contrast, this study focuses on a single region and uses a more representative sample of countries from all parts of Africa. Instead of using the number of cellular providers to measure mobile competition, this study uses the actual level of mobile subscriptions.

PAGE 49

43 complements, the increased usage of cellular phones will be associated with increased access to main lines. On the other hand, if mobile subscriptions are substituting for fixed lines (for instance, by covering unsatisfied demand for fixed lines), a negative relation is expected. 16 3.3.2 Other Considerations The relation between telephone access and other factors, such as per capita GDP, institutions and political regime has been established in other studies. 17 The general consensus is that strong institutions and stable political systems are positively correlated with increased access to basic telecommunications. Likewise, as individual income increases, the demand for telephones should rise, especially in situations where individuals are poor to begin with and suffer from lack of access because of the inability to pay. The relation between per capita GDP and telephone access may, however, be small in situations where lack of access is attributable to insufficient supply rather than low demand. 3.4 Background 3.4.1 Competition Policies Historically competition in basic services has been nonexistent in Africa. Even though there has been some privatization in the 1990s, providers of basic services remain monopolies in most of Africa. 18 In contrast, the mobile sector in Africa opened rapidly to competition, and 16 It is possible for mobile and main lines to be strategic substitutes rather than substitutes in consumption. Fixed-line penetration in Africa may be growing because incumbent providers think that increasing access is the best response to the competition it faces from mobile providers. If the services are strategic substitutes then the relation between the two will be positive. Likewise the two services may be complements in consumption or strategic complements. In the first instance individuals get connected because of an increase in the value of being connected as more and more people have access to fixed lines. It is important to be able to distinguish the different types of substitutes and complements for the purposes of proper interpretation of the relationship between mobile and main lines. 17 See Henisz (1998), Gutierrez (1999), Hamilton (2002) and Singh (2000). 18 By 1997, Cote dÂ’Ivoire, Gabon, Ghana, Madagascar, South Africa and Tanzania had partially privatized their communications, but most were still largely state-owned (Hamilton 2002).

PAGE 50

44 Africa was one of first regions to adopt mobile service (Leblanch-Wohrer and Lewington, 2000). As Table 3.2 shows, a number of countries currently sustain multiple (at least two) providers (over a wide range per capita GDP). Table 3.2 The Level of Mobile Competition in Africa by 1997 Country Legally Permissible Competition Country Legally Permissible Competition Country Legally Permissible Competition Algeria M Ethiopia M Niger C Angola M Gabon M Nigeria C Benin M Gambia, The M Rwanda M Botswana Ghana C Sao Tome Burkina Faso M Guinea C Senegal C Burundi C Guinea Bissau Seychelles Cameroon M Kenya M Sierra Leone C Cape Verde C Lesotho M Somalia C. African Rep c Liberia South Africa c Tchad D Libya M Sudan Comoros Madagascar C Swaziland M D. R. of Congo C Malawi M Tanzania c Congo M Mali M Togo D Cote dÂ’Ivoire C Mauritania M Tunisia M Djibouti D Mauritius M Uganda D Egypt D Morocco D Zambia C E. Guinea M Mozambique M Zimbabwe C Eritrea C Namibia M M-monopoly, D-Duopoly, C-More than two providers Source: Leblanc-Wohrer and Lewington (2000) and Minges (1999) One has to be cautious, however, in using the increase in cellular networks across Africa as an indication of the growth in private-sector participation and competition. In fact, some cellular ventures (Angola and Algeria for example) are still state-owned or monopolies. Furthermore, the existence of cellular operators may not be a perfect measure of competition since cellular service is often available only to the wealthy, who represent only a small portion of African economies. The price of the service may limit its ability to act as a

PAGE 51

45 strong competitive force where incomes are low. Nevertheless, in villages call aggregators can play a role in providing access (if not ownership) for low-income demanders. 19 3.4.2 Mobile versus Fixed Lines As cellular networks continue to emerge, both fixed-line and wireless services have been growing at a rapid pace in Africa. Table 3.3 shows the relative size of a countryÂ’s mobile and main-line markets using the 23 countries in our sample. Table 3.3: Relation Between Mobile and Main Lines 1987 1992 1997 Total Main-Line Penetration (per 1,000) 11.29 15.702 22.341 Total Mobile Subscription (per 1,000) 0.004 0.056 2.565 Ratio of Mobile to Main Line 0.00033 0.00359 0.111 Percent of New Mobile to Main Line 1.100 37.800 Mobile as a Percent of Total Subscription 0.033 0.358 10.285 Between 1987 and 1997, access to fixed lines almost doubled, growing from 1 1.29 per 1,000 inhabitants to 22.341 per 1,000. In 1987, South Africa, Tunisia and Algeria accounted for more than 50 percent of all fixed-line telephones in the sample. Ownership patterns had shifted little by 1997, so that South Africa, Tunisia, Botswana and Egypt accounted for 56 percent of main-line access across the sample. The trend shows that countries that were doing relatively well in early years continue to lead the way in the late 1990s. Some of the same countries appear to be leaders in the mobile market. Over the same period, (1987-97), mobile subscription grew from close to 0 per 1,000 people to 22.34 per 1,000 people. South Africa, Tunisia, Morocco and Gabon are among the countries that have the highest subscription rates. Gabon moved from l9 Main-line access and mobile subscriptions tend to be significantly higher in middle-income countries compared to low-income countries in Africa. (Table 3.6 gives a comparison of telephone subscriptions in middleand low-income countries in Africa). However, over time, as the price of cellular provision falls, this becomes less true. Uganda and Cote dÂ’Ivoire are examples of countries with per capita income below $1,000 and with at least two cellular operators. Cellular operators can be potential threats to incumbent firms, since they can increase penetration at relatively low cost per additional subscriber. The threat of competition may be enough to give the incumbent the incentive to improve service. The potential threat is enough to provide the impetus for telecom growth.

PAGE 52

46 having zero mobile subscriptions in 1987 to having the second highest subscription level of 8.35 per one thousand individuals. South Africa, by far, has had the most success in increasing access to mobile phones. By 1997 access to mobile phones in South Africa was 36 per thousand, which accounted for sixty-three percent of all cellular subscription in Africa. This rapid rate of growth in cellular subscriptions may be attributed to the influx of new mobile providers, both foreign and local. While mobile subscriptions still lag behind fixed-line access, the gap between the two has been closing over time. The ratio of new mobile subscriptions to new main-line access is increasing over time to just under forty percent in 1997. 3.5 The Data 3.5.1 Defining the Data Annual data used in the analysis represent 23 African countries, 1985-97. 20 The time period and number of countries used was determined by constraints on the data available for some of the variables. The role of mobile competition in determining fixed-network development is assessed by using main line per 1,000 inhabitants as the dependent variable. The variables used in this analysis are shown in Table 3.4. 20 The countries included in the sample are Algeria, Botswana, Cameroon, Republic of Congo, Cote dÂ’Ivoire, Egypt, Gabon, Ghana, Kenya, Madagascar, Malawi, Mali, Morocco, Niger, Nigeria, Sierra Leone, South Africa, Tanzania, Togo, Tunisia, Uganda, Zambia and Zimbabwe.

PAGE 53

47 Table 3.4: Data Definitions and Sources Variable Definition Source Main lines per 1 ,000 inhabitants This is the first dependent variable. It is the (# of main telephone lines/population)* 1 ,000. World BankÂ’s World Development Indicators (1999). Demand for main lines This is the second dependent variable. It is the (# of main telephone lines + # of unmet applications/population)* 1 ,000. World BankÂ’s World Development Indicators (1999). Per capita GDP (Real Gross domestic product in constant 1990 U.S. dollar )/population. World BankÂ’s World Development Indicators (1999). Institutional Factors Institutional features which capture country risk factors. It is the sum of the following: corruption of the political system (CORRUP), rule of law (LAW), bureaucratic quality (BUREAU), security of contract (CONTRACT) and risk of expropriation (EXPROP). IRIS-3 file of International Country Risk Guide (ICRG) data, 1982-1997 constructed by Stephen Knack and the IRIS center. Economic Freedom Two variables are used from this index: government operations (GOVOP), which measure the extent to which personal choice and markets, rather than political planning and coercion direct resources. The second variable is discriminatory taxes, (DISCTAX), which measures the extent to which government protects property rights in terms of low transfers and subsidies. Economic Freedom of the World annual report, 1997. Compile by James Gwartney and Robert Lawson. Democracy Indicator of political regime type on a scale of 0 to 10,0 being the least democratic. (DEMOC). 21 Polity III: Regime Change and Political Authority 1800-1994. Compiled by Keith Jaggers and Robert Gurr. Mobile Subscriptions (# of cellular subscriptions/population)* 1 ,000. (MOBILE). World BankÂ’s World Development Indicators (1999) Urban percent (urban population)/total population (URBAN) World BankÂ’s World Development Indicators (1999). Trade (imports+exports)/GDP lagged one period. (LTRADE.) World BankÂ’s World Development Indicators (1999). 21 The level of Democracy is assumed to be constant between 1994 and 1997.

PAGE 54

48 3.5.2 Descriptive Results Table 3.5 summarizes access to telephones in based on regional differences and income. The data indicate North Africa having lower access to both main lines and mobile than SubSaharan Africa. The higher level of mobile subscription in sub-Saharan Africa is primarily due to access in South Africa (36.95 in 1997), compared to the highest rate in North Africa of 2.7 in Tunisia in the same year. South Africa is also the reason for the high fixed-line access in SubSaharan Africa. Once South Africa is excluded, mobile access in sub-Saharan Africa falls, but remain above that of North Africa. Although the difference seems negligible, the pattern suggests that countries with low access to fixed-line telephones tend to have relatively high access to cellular phones. This result concurs with the view that mobile is satisfying unmet demand for main lines. Table 3.5 also shows that main line access is about nine times higher in middle-income countries compared to low income ones. When South Africa is excluded from the sample, mobile subscription falls by almost one third, but access in middle-income countries remain higher than that of low-income countries. This pattern suggests that the ability to pay for mobile may be an important consideration even while cellular costs are falling. One explanation for this may be that providers incorrectly perceive that lower incomes reduce the willingness of individuals to pay for the service. In other words, they may view this as a factor that limits market size. Such providers would therefore target higher income markets. Lower access in this case would be a supply side problem. Finally, Table 3.6 provides summary statistics based on regional differences and the level of mobile competition within a country by 1997. As expected, countries that allow mobile competition invest more in mobile communications. Interestingly, these same countries, on average, tend to invest less in main lines. According to the data, countries that allow mobile competition invest in mobile approximately two times as much as countries with monopolies in mobile service. At the same time mobile monopoly countries on average invest in main lines by

PAGE 55

49 just under one and one half times more than countries that allow mobile competition. One implication of this is that mobile competition allows cellular providers to be better able to pick up the slack where main lines are under-supplied. The difference between mobile subscriptions in low-income countries is miniscule when compared to that of middle income ones. This result suggests that the ability of customers to pay appears to be an important factor in determining mobile access. Table 3.5: Regional Differences, Income and Telephone Access 1985-1997 North Africa Sub-Saharan Africa Sub-Sahara without South Africa Middle Income 34.63 45.794 25.420 (0.234) (2.685) (0.852) Low Income 4.580 4.580 (0.068) (0.068) The first number in each cell represents average main-line penetration. The numbers in parentheses are mobile subscription per thousand individuals. Table 3.6: Main-Line Penetration, Income and Mobile Provision Market Structure for Mobile Monopoly Competition Middle Income 32.50 87.069 (0.528) (6.354) Low Income 3.349 6.00 (0.021) (0.062) All Income 21.569 10.736 (0.338) (0.527) The first number in each cell represents average main-line penetration. The numbers in parentheses are mobile subscription per thousand individuals. 3.6 The Empirical Model 3.6.1 Model Estimation Panel data estimation techniques are used to analyze the impact of MOBILE competition on MAINLINE access. The model is estimated by using pooled data from 1985 to 1997. The data set begins in 1985 because the mid-1980s marked the take-off of rapid telecom reform in Africa as well as the introduction of cellular phones Before estimating the panel data, some cross-sectional analyses are conducted to compare the attributes of the explanatory variables at different points in time. The impact of MOBILE on

PAGE 56

50 MAINLINE development is assessed for a cross-section of countries in 1987, 22 1993 and at the end of the sample period in 1997. The following equation was estimated for each of the three years: (1) MAINLINE a + J3 K X + e , where X is a vector of independent variables, containing LOGGDPC, LTRADE, URBAN, CORRUP, LAW, BUREAU, CONTRACT, EXPROP, GOVOP, DISCTAX, DEMOC and MOBILE. (2) A MAINLINE = a + fl K AX + e , where, A indicates that the variables are expressed as their 1997 values minus those of 1987. Equation 3 is estimated to account for cross-country variation with the panel data approach. (3) MAINLINE CT = a CT + (3 K X CT + 5 cr + s CT , where, Xct is the same vector of independent variables as in equation 1 for country C in year T. ct CT are country dummies that account for unobserved differences across countries that may affect the dependent variable. 8 cr captures year effects that take care of possible variations of the omitted variables through time. Equation 3 is estimated by expressing the variables as deviations from their group means. This approach picks up any constant differences on a country-specific level through time. The model is estimated by using fixed effects. 23 A parsimonious specification in which mobile subscription per 1,000 individuals is the only righthand-side variable is first estimated to establish the relation between mobile and main line assess, 22 1987 as opposed to 1985 was used for the first cross-sectional estimation because there were not enough observations of MOBILE in 1985 to generate meaningful results. “ 3 The fixed-effect model assumes that the counties in the sample each have characteristics that are unique and do not change over time. These differences are captured in differences in the constant term.

PAGE 57

51 and between mobile and main line demand in isolation of other independent variables. The standard errors of all estimations are adjusted for cross-observation error dependence with the HuberAVhite variance estimator. 3.6.2 Sample Selectivity Only twenty-three out of more than fifty countries in Africa are included in the sample used in this analysis. This raises the concern that the results of the study may not be generalized to include the excluded countries. To check this possibility, we compare the excluded countries in cross-section with those included in the analysis along observable lines. Per capita GDP, main line and mobile access, trade and urban percentage are used as the basis for comparison. Table A in the appendix compares the overall means of these variables with and without South Africa as well as with and without Seychelles and Mauritius. 24 Table A shows that at the mean, the included countries tend to be similar to those excluded. For instance, the biggest difference between main line access is only about 2 per 1 ,000 individuals. The difference in mobile access is negligible at less than 0.1 per 1,000. Average per capita GDP tends to be slightly higher in the included countries, but both groups tend to have an average income in the lower income category. To further check the possibility of selectivity bias reduced fixed effects regressions (using the same observable variables) were conducted for both groups of countries. Table B in the appendix shows the results. Apart from the trade variable, which is insignificant for the countries excluded, the regression results were quantitatively similar. Finally, the reduced regression was conducted using all the countries in Africa. Again, the results were quantitatively similar (in terms of sign and significance), to the results using just the 23 countries in our sample. This ' 4 South Africa, Seychelles and Mauritius have significantly higher access to main lines than the rest of Africa (107, 203, and 195, respectively, in 1997). These countries also tend to have higher income levels than the rest of Africa. Except for Botswana, they are the only other upper-middleincome countries in Africa.

PAGE 58

52 informal sensitivity analysis (by itself), suggest that selectivity may not be an important problem, at least qualitatively. 3.6.3 Endogeniety Some analysts argue that some reverse causality exists between GDP per capita and infrastructure (telecom) investment. (Colier and Gunning 1999). Kerf and Smith ( 1 996) suggest that the poor quality of AfricaÂ’s infrastructure constrains private investment in other activities and thus is an obstacle to economic growth. That is, countries with proper infrastructure are expected to attract investors, which in turn generate a higher per capita GDP. (Madden and Savage 1998). This potential is unlikely to be present because the region studied is comprised of nations that are plagued with instability and represent relatively high-risk projects. An increase in telecommunications investment will not be sufficient to attract non-infrastructure investment that would significantly affect GDP. MOBILE captures the competition effect in markets where direct telephony is a monopoly. Competition is, however, a result of regulatory considerations, and monopoly firms often have control over the regulatory regimes in Africa through the ministry. Often there is no separation of regulatory functions and operation because both may be operating under the same ministry. Mobile and direct lines are therefore jointly determined. If MOBILE is endogenously determined, it will be correlated with the error term. Under this condition, ordinary least squares (OLS) estimations will tend to attribute changes in the dependent variable (MAINLINE) caused by the error term to MOBILE. As a result, the coefficient on MOBILE will be biased upward or downward, depending on the sign of the correlation between MOBILE and the error term. In addition, if MOBILE and MAINLINE are jointly determined, MOBILE cannot be considered fixed in repeated samplings and there is a potential bias in all the estimated coefficients. Consider the coefficient of URBAN (p 3 ) for instance. Note that (3 3 is supposed to be the estimated effect of URBAN on MAINLINE, holding MOBILE and all other right-hand-side variables constant. MOBILE is, however, not held constant when changes in MAINLINE takes place.

PAGE 59

53 Therefore, p 3 may actually measure some mix of the effects of both URBAN and MOBILE. Because of the potential problems that may result when OLS is used, it is vital to consider alternative estimation techniques to reduce the simultaneity bias. We use the instrumental variable (IV) technique to reduce the potential biases in the estimation of Equation 3. To use this technique, we replace MOBILE with a new variable (instrumental variable), which should be highly correlated with MOBILE but uncorrelated with the error term. Until 1997 only six of the twenty-three countries in the sample had allowed some private participation in fixed-line telephone service provision. All governments, except for Cote d Ivoire, which sold 51% of its shares to the private sector in 1997, had maintained majority ownership. Private ownership of wire-line phones was therefore practically nonexistent in the sample. In contrast, private individuals are usually licensed to provide cellular service. The nature of ownership made it more likely that private sector credit directed toward investment in telecommunications would go into mobile communications rather than fixed-line development. Thus, when main lines are publicly owned, credit to the private sector will not determine public investment. We identified PRIVCRED (the ratio of private sector credit to GDP) as an instrumental variable for MOBILE and use two-stage least squares to estimate Equation 3. 25 It may be argued that an increase in PRIVCRED implies a reduced capacity for the government to fund telephone investment. If true, PRIVCRED would be correlated with both MAINLINE and MOBILE. Historically, however, telecom reform is financed largely by loans from international lending agencies as part of their mandate to aid in the improvement of 25 Although PRIVCRED is defined as credit to the private sector as a fraction of GDP, it may also randomly include some credit to the public sector as well. Measurement errors introduced by this are assumed to be captured as part of the random error term since the inclusion of public sector credit is not systematic.

PAGE 60

54 infrastructure in lesser-developed countries. Local investors in mobile communications are likely to be more dependent on domestic credit. 26 3.6.4 The Dependent Variable Redefined As discussed in section 3.3.1, it is not entirely clear whether MOBILE is a substitute or a complement for MAINLINE. The supply of mobile phones have grown rapidly, mainly because many governments tend to be more willing to introduce privatization and competition in the market for cellular rather than fixed lines. While the supply of mobile has been growing, so has the supply of fixed lines and more people are using both fixed lines and wireless phones. It is, however, not clear how much of the new cellular subscriptions are by individuals who also have access to fixed-line telephones. A positive sign could indicate that the two services are complements. In this case, it may be that the same individuals have access to both fixed lines and cellular phones. A positive sign may, however, result even if the services are substitutes in consumption. Cellular operators can be potential threats to incumbent firms, since they can increase penetration at a relatively low cost per additional subscriber, and in less time. The threat of competition may be enough to give the incumbent the initiative to improve service. It is also unusual to observe a decline in the level of teledensity in countries where access is already low. Both MOBILE and MAINLINE access are 26 Because only one instrument is identified, a formal test of its validity could not be conducted. Instead, we performed a sensitivity analysis by comparing the R's of the restricted regression (without PRIVCRED as a regressor) with the unrestricted regression (including PRIVCRED), using MAINLINE as the dependent variable. The R 2 remained unchanged once PRIVCRED was added to the regression. Likewise, a test of first-stage correlation was conducted with MOBILE as the dependent variable. When PRIVCRED was added as an explanatory variable, the R 2 improved noticeably, by 8%. The results of these tests indicate that PRIVCRED may be used as an instrument for MOBILE. The results are of course based on the current specification of MAINLINE. It may be possible that PRIVCRED would not survive these tests for a different specification of mainline development.

PAGE 61

55 therefore expected to grow. It is therefore not entirely clear how to interpret a positive coefficient on MOBILE when we estimate the model using a supply side definition of MAINLINE. 27 To check our interpretation of the relation between MOBILE and MAINLINE, we redefine the dependent variable to be actual total access demand for wire line telephones, rather than just satisfied demand. The new dependent variable is DEMAND, which measures the total number of applications for fixed-line service. 28 The equation estimated is: (4) DEMAND CT = a CT + fi K X CT + S CT + s CT , 29 Equation 4 measures the effect of satisfied demand for mobile on total demand for main lines. Since the model is expressed in terms of demand relations, we can infer more clearly the relation between mobile and fixed lines. If the coefficient on MOBILE is positive, then the result can be interpreted to mean that fixed lines and cellular phones are complements. On the other ~ 7 MAINLINE access per 1,000 inhabitants is a measure of how many telephone lines the respective governments or incumbent wire line firm actually make available to the public. This is usually lower than its demand. In essence main line access measures the portion of demand for main lines that have been met. Since there is usually a shortage, this is in effect satisfied demand. Ros ( 1 999) argued that the low number of main lines is a supply rather than demand-side constraint. Because of the absence of rate rebalancing, residential access prices are likely to be below their economic costs. An increase in supply is overwhelmed by demand, mainly because of long waiting lists as well as low price elasticities of demand. “ 8 This is just main line access per thousand individuals waiting list per thousand. Waiting list captures the number of applications for connection to a main telephone line that have been unmet. main lines f applications ^ population the supply of main lines. Specifically, DEMAND = *1000 , where applications is unmet demand for " 9 The dependent variable is now a demand side variable and so is not determined by policy. Although MOBILE is still a policy variable, it is not simultaneously determined with the new dependent variable (DEMAND). If MOBILE was the total demand for cellular phones, then the two variables would be simultaneously determined, by virtue of the fact that MOBILE and DEMAND for main lines are related services. MOBILE, however, measures just the part of its demand that is satisfied. In other words, it measures demand constrained by actual supply of cellular phones. It is therefore not a complete measure of demand. It is in fact the supply of cellular phones, which is captured by this variable. If this is true, then we can assume that although MOBILE is a policy variable, it is not jointly determined with total main line demand.

PAGE 62

56 hand, a negative coefficient may be interpreted to suggest that the two services are substitutes. An increase in MOBILE usage could have externality effects, resulting in increased MAINLILNE demand. The growing MAINLINE usage could, however, have the same effects. To separate the externality effects of MOBILE usage from that of MAINLINE usage itself, we introduce MAINLINE access lagged one period (LMAINLINE) as an independent variable in the DEMAND equation. Finally, is possible that as mobile subscriptions increase, its relation with demand for wire-lines may change. To account for this possibility, we introduce mobile squared (SQMOBILE) as an independent variable in the demand equation. 3.6.5 Estimated Results Table 3.7 presents the results of the cross-sectional estimations. In 1987, MOBILE, was not significantly associated with MAINLINE access, but had a large positive coefficient. This insignificant relation may have occurred because although growing, the level of MOBILE subscriptions was too small in 1987 to have any real effect on MAINLINE development. By 1997, however, MOBILE had cemented its presence in the African telecom market enough to be significantly and positively associated with MAINLINE. This is also shown in the last column of Table 3.7, which captures how the growth in MOBILE access between 1987 and 1997 affected fixed-line growth for a cross-section of countries. This could suggest that while in earlier years cellular provision was inconsequential in the telecommunications market, it is becoming more of a competitive force as its usage grows. The real impact of MOBILE on MAINLINE, however, appears to be very small between 1987 and 97. In this estimation, the mobile coefficient corresponds to an elasticity that is close to zero. Generally, the large changes in the size of the coefficients as well as variation in their significance levels over the three years reported, indicate that the variables included in the model are not constant overtime. These cross-sectional results are also suffer from omitted variable bias, so the coefficient size and significance may not be precise when country specific elements

PAGE 63

57 are omitted. The fixed-effect model that is later estimated takes care of the omitted variable problem and accounts for any constant level differences in the explanatory variables over time. Table 3.8 shows the result of the regression in which MOBILE is isolated from all other independent variables. The results indicate a positive and significant correlation between mobile usage, and the two dependent variables (MAINLINE and DEMAND). Table 3.7: Cross-Sectional Regression Results for 1987\ 1997 and the Change Between 1985 and 1997. (Dependent Variable: MAINLINE per 1,000 individuals) Variable 1987 Crosssection. 1993 Crosssection 1997 Crosssection. Change between 1987 and 1997 Cross-section. LOGGDPC -3.548 10.814* 10.772** 22.126 (-0.875) (1.875) (2.308) (1.363) LTRADE -0.104 -0.005 -0.229** -0.160** (-1.107) (-0.027 (-2.025) (2.208) URBAN 0.559** 0.024 0.599** 0.426 (2.261) (0.592) (2.044) (0.861) CORRUP 2.391 11.526** 3.348 4.930** (0.766) (2.213) (0.891) (1.985) LAW -6.811*** -3.353 3.307 -1.416 (-3.401) (-1.528) (0.989) (-0.496) BUREAU 4.551* -7.969*** -6.312** -4.379** (1.800) (-2.872) (-2.510) (-1.777) CONTRACT 3.057 0.550 1.798 5.475*** (1.211) (0.212) 0.195) (2.732) EXPROP -0.900 2.885 2.946 -0.263 (-0.308) (0.821) (1.242) (-0.125) GOVOP -0.150 1.015** 0.932* 1.046 (-0.353) (2.022) (1.627) (1.130) DISCTAX -0.531* 0.135 -0.371 0.788* (-1.638) (0.446) (-0.996) (1.636) DEMOC 4.049** -1.401 -0.102 -1.975** (2.546) (-1.535) (-0.010) (-2.183) MOBILE 185.949 31.986* 1.308*** 0.503** (1.531) (1.678) (4.135) (2.474) N 23 23 23 23 R 2 0.90 0.87 0.95 0.85 F Stat 10.29 10.79 107.14 49.61 Note: a: Prior to 1987 there was not sufficient data on MOBILE to run cross-sectional regressions. T-stats are shown in brackets * = significant at 10%. ** = significant at 5%, *** = significant at 1%. Tables 3.9 and 3. 10 show the results of the pooled time series, cross-sectional estimations. Table 3.9 shows the fixed and instrumental variable estimation results using

PAGE 64

58 MAINLINE as the dependent variable. All estimations show a positive and significant correlation between MOBILE and MAINLINE. This result suggests that cellular subscriptions may be playing a complementary role to fixed telephone lines. It is also reasonable to interpret the positive sign in terms of competition within the industry. 30 The use of cellular phones act as a competitive force, which encourages increased investment in direct lines. It is possible that main lines and mobile phones may be complements sometimes and substitutes in other cases. For instance it may be possible for mobile and main lines to be complements in consumption in countries where income is relatively high and less so in areas with lower income levels. To further explore this potential differential relationship, the relation between main lines and mobile, we interact MOBILE with per capita GDP (MOBINC), and with a regional dummy which equals 1 if countries are Sub-Saharan and 0 if they are Northern African (MOBSUB). The results show that the interaction of MOBILE with per capita GDP is insignificant. However, the marginal effect of increased mobile subscription on mainline access is more than 2 times lower in Sub-Saharan Africa, than in Northern Africa. For instance it is not uncommon for professionals to own both mobile and fixed lines. At the same time, it is becoming increasingly commonplace to have access to one or the other for various different reasons. 31 Mobile can play the role of both a substitute and a complement of main line in the same country at different points in time. 0 Ahan and Lee (1999) estimated the demand for mobile networks using data from sixty-four countries in both developed and developing countries and found that mobile demand is positively correlated with the number of fixed lines per person. 31 When South Africa is excluded, the marginal effect of increasing MOBILE on MAINLINE is reduced when income is higher. (Second column of Table C in the appendix). Since MAINLINE is a supply side measure, this result may be indicating that providers of fixed service will find it strategic to focus more on cellular provision in high income areas, rather than on fixed line expansion, which is relatively time consuming. From the demand side perspective, (where the dependent variable is DEMAND in Table C), MOBINC is not significant and the result for MOBSUB is quantitatively the same as for the other estimations in Tables 3.9 and 3.10.

PAGE 65

59 Table 3.8: Parsimonious Specification of Panel Regressions Controlling for Country-Level Differences. Data: 23 countries, 1985-97 Variable Dependent Variable: MAINLINE access per 1,000 people Dependent Variable: DEMAND for main lines per 1,000 people MOBILE 0.931*** 0.853*** (4.951) (5.150) N 299 239 R 2 with fixed effects and other independent variables 32 0.91 0.93 R 2 with only fixed effects 0.89 0.89 F-Stat 24.51 26.52 Note: t-stats are shown in brackets * = significant at 10%. ** = significant at 5%, *** = significant at 1%. Although the various estimations show MOBILE as highly significant, large changes are necessary to affect main-line development noticeably. Using the fixed-effects coefficients from Table 3.9, if Gabon increased its cellular subscriptions by 10 per 1,000 people, main-line access would increase in that country by 22.83 telephones per 1,000 inhabitants. 33 This corresponds to an elasticity of only 0.204 in 1997. 34 A low mobile elasticity indicates that direct line and cellular subscriptions are not close complements. Cellular might be too costly for individuals with per capita incomes below $1,000. Only the elite can afford cellular telephones. In addition, many countries' regulatory frameworks are not yet equipped to deal with the emerging competitive markets. 3 " The R 2 s of 0.91 and 0.93 indicates that the model explains much of the variation in the dependent variables. Specifically up to four percent more of the variation in MAINLINE and DEMAND is explained by MOBILE. 33 An increase of 22.83 does not represent a large increase in MAINLINE. This is because mobile subscription in Africa was only 2.67 per 1,000 individuals while MAINLINE access was 22.34. This is 8.7 times higher than mobile access. When mobile is increased by 10, mainline access in Gabon increased by only 2.2 times more than the increase in mobile even though mainline access was almost nine times higher in 1997. 34 MAINLINE elasticity is calculated using 1997 values for each country. That is, P*[MOBILE / MAINLINE],

PAGE 66

60 Table 3.9: Panel Regressions Controlling for Country-Level Differences. Data: 23 countries, 1985-97 Dependent Variable: MAINLINE access per 1,000 people Variable Fixed Effects Fixed effects with Interactions Instrumental Variable: PRIVCRED used as a proxy for MOBILE. LOGGDPC 7.593*** (2.740) 6.646** (2.398) 8.262*** (3.063) LTRADE -0.115*** (-6.252) -0.111*** (6.300) -0.113*** (-5.996) URBAN 0.624*** (7.043) 0.641*** (7.307) 0.647*** (7.027) CORRUP -0.067 (-0.138) -0.167 (-0.342) -0.270 (-0.553) LAW 1.295** (2.500) 0.905** (1.972) 1.265** (2.073) BUREAU -2.296*** (-4.310) -1.696*** (-3.508) -1.500** (-1.805) CONTRACT 2 i (4.811) 1.819*** (4.406) 2.000*** (4.411) EXPROP 0.060 (0.156) 0.242 (0.687) 0.068 (0.177) GOVOP 0.422*** (2.975) 0.412*** (2.975) 0.493*** (3.245) DISCTAX 0.184*** (2.658) 0.162** (2.378) 0.176*** (2.330) DEMOC -0.441*** (-4.302) -0.407*** (4.088) -0.467*** (-4.241) MOBILE 0.438*** (4.579) 5.229*** (3.930) 0.799** (2.211) MOBINC -0.0002 (-1.486) MOBSUB -3.923** (-2.520) LMAINLINE N 299 299 289 R z with fixed effects and other variables 0.97 0.97 0.97 R" 1 with only fixed effects 0.89 089 0.89 F-Stat 32.55 34.91 304.73 Note: T-stats are shown in brackets * = significant at 10%. ** = significant at 5%, *** = significant at 1%. A test of the significance of a time trend could not be rejected. As a result, the time trend was excluded from our estimations. Results from the instrumental variable estimation, with PRIVCRED used as an instrument for MOBILE, show that the impact of the competition variable (MOBILE) remains positive and statistically significant. Table 3.10 shows the results using DEMAND for main lines

PAGE 67

61 as the dependent variable. The results for MOBILE competition are quantitatively the same as those in the MAINLINE estimations. The positive and significant coefficient on MOBILE in this case lend more support for the view that cellular and main line telephones may in fact be complements. 35 The second column of Table 3.10 accounts for the increase in value that may result from MAINLINE usage. Once the externality effect of MAINLINE usage is accounted for (by introducing LMAINLINE), the result indicates that the two services may in fact be complements in consumption. The positive coefficient on MOBILE may be interpreted to mean that an increase in satisfied demand for MOBILE is associated with an increase in total demand for main lines and the two services are used together. The result may be capturing the effect where two or more telephone users connect with each other, with at least one party using a cellular phone. The coefficient on SQMOBILE (although small), suggests that when mobile usage is low, (below 35 per thousand), it may be playing the role of a complement for main line. However, as usage becomes more widespread, its role switches to be that of a substitute. This scenario does a good job of describing the evolution of cellular usage in these developing economies, and concurs with earlier suggestions that it is possible for the service to be play both the role of a complement and that of a substitute. As a new innovation in the mid to late1980s, cellular service was relatively expensive. It is quite possible that at this stage, only professionals and private individuals with higher incomes (usually in urban areas) would utilize the service. 35 We also ran a fixed-effects regression (result not reported) using the six countries in the sample where there is a single cellular provider, who is also the incumbent. The result of the regression using MAINLINE as the dependent variable shows a positive (but insignificant) coefficient on MOBILE. If the incumbent is also the only cellular provider, then the competitive pressure of a substitute that would result in an increase in main line access (when the mainline and mobile are substitutes) does not exist. In cases where the incumbent is the clear provider of mobile a negative correlation between main line access and cellular subscriptions would indicate that the two services might be substitutes. If the correlation is positive, it lends support to the view that the positive correlation found in our estimations indicate that the services are complements in consumption.

PAGE 68

62 Table 3.10: Panel Regressions Controlling for Country-Level Differences. Data: 23 countries, 1985-97 Dependent Variable: DEMAND for main lines per 1,000 people Variable Fixed Effects without lagged MAINLINE Fixed Effects with lagged MAINLINE Fixed effects with Interactions Fixed Effects with lagged MAINLINE and Interactions Fixed Effects with SQMOBILE Fixed Effects with lagged SQMOBILE and Interactions LOGGDPC 3.957 (1.109) 3.661 (1.215) 2.717 (0.784) 2.827 (0.960) 4.020 (1.323) 3.230 (1.085) LTRADE -0.089*** (-3.646) -0.057*** (-2.884) -0.083*** (-3.535) -0.055*** (-2.797) -0.057*** (-2.881) -0.055*** (2.798) URBAN 0.951*** (10.149) 0.887*** (10.274) 0.961*** (10.421) 0.898*** (10.429) 0 . 886 *** (10.336) 0.897*** (10.492) CORRUP 0.046 (0.069) 0.600 (1.165) -0.130 (-0.196) 0.443 (0.835) 0.695 (1.344) 0.550 (1.030) LAW 0.632 (1.127) 0.359 (0.736) 0.245 (0.459) 0.111 (0.237) 0.238 (0.492) -0.013 (-0.027) BUREAU -1.597** (-2.148) -1.311** (-2.174) -1.047 (-1.437) -0.952* (1.606) -1.275** (-2.117) -0.924 (-1.555) CONTRAC T 1.941*** (3.868) 1.653*** (4.112) 1.616*** (3.328) 1.449*** (3.692) 1.669*** (4.150) 1.467*** (3.728) EXPROP 0.211 (-0.544) -0.354 (-1.162) 0.048 (0.131) -0.167 (-0.562) -0.362 (-1.193) -0.178 (-0.599) GOVOP 0.629*** (3.170) 0.631*** (3.992) 0.600*** (3.084) 0.611*** (3.932) 0.623*** (3.954) 0.603*** (3.884) DISCTAX 0.132 (1.218) 0.088 (0.875) 0.077 (0.711) 0.053 (0.526) 0.089 (0.880) 0.055 (0.536) DEMOC -0.402*** (-2.759) -0.283** (-2.337) -0.333** (-2.348) -0.244** (2.043) -0.285** (-2.344) -0.248** (-2.060) MOBILE 0.496*** (5.280) q 399 *** (5.591) 3.966*** (3.397) 2.775*** (2.837) 0.770*** (3.732) 3.226*** (3.360) SQMOBILE 0 . 011 ** (-2.299) 0 . 010 ** (-2.310) MOBINC 0.00003 (0.148) 0.00002 ( 0 . 110 ) -0.00004 (-0.228) MOBSUB -3.541 *** (-2.816) -2.418** (-2.434) -2.336** (2.406) LMAIN LINE 0.171*** (3.189) 0.160*** (3.059) 0.168*** (3.166) 0.158*** (3.043) N 239 238 239 238 238 R 2 with fixed effects and other variables 0.98 0.98 0.98 0.99 0.99 0.99 R 2 with only fixed effects 0.92 0.92 0.92 0.92 0.92 0.92 F-Stat 41.11 67.14 42.25 67.41 57.98 59.87 Note: T-stats are shown in brackets * = significant at 10%. ** = significant at 5%, *** = significant at 1%. A test of the significance of a time trend could not be rejected. As a result, the time trend was excluded from our estimations.

PAGE 69

63 These users tend to use cellular service in conjunction with fixed line phones. As the relative price of cellular service falls over time, usage spreads to other users, including rural and lower income subscribers. When mobile becomes available in regions where main line access is non-existent or low, it becomes a substitute for main line. Currently, the net effect of mobile on main line demand is greater than zero at the level of mobile access for most countries in the data set. As such, the complementary effect outweighs the substitution effect. However, as mobile subscription increases, its role as a substitute will begin to dominate. Generally, the results show that both access to MAINLINES and the DEMAND for main lines tend to be higher where income levels are relatively high. The institutional environment is important for telecommunications development (Bergara, Henisz and Spiller 1998). The institutional variables are also important in determining main-line demand. Rule of law (LAW), security of contracts (CONTRACT), and role of markets (GOVOP) all affect penetration in a positive manner. Political institutions appear to be important, but democracy has a negative coefficient. Clearly, modeling these features warrants more attention in future research. In particular, modifications could be made to consider the possibility that the measures of the institutional environment used, may not readily combine in a linear model for their impact upon the relation between main line and mobile subscription. 3.7 Conclusions Some research has found evidence that competition generally tends to improve industry performance and productivity. Studies that looked at the telecommunications industry have also found competition in basic services is associated with increased telecom growth and development. This study examine the role of MOBILE competition rather than competition in basic services. This issue is of interest particularly since mobile competition tends to be more widespread than competition in basic services. Many countries are privatizing, but most have stopped short of allowing multiple provision. Mobile competition is expanding its role in many developing countries, even where access to fixed-line telephones is very low.

PAGE 70

64 The empirical results support the theory that credible institutions, including stable legal structures are important driving forces behind the surge of modernization in AfricaÂ’s telecommunications sector. The effect of mobile competition on main-line development and demand, though relatively weak is significant. At different stages of cellular development, mobile play the role of both a substitute for, and a complement of main line demand. From the supply side perspective, the pressure on incumbent firms to improve service provision increase with mobile operators present, possibly leading to improvements in main-line quality as well. The empirical results presented here suggest that competition is important in fostering telecom development: it induces investment by wire-line investors. As more private participation and competition are allowed, competition is expected to have a much larger effect on wire-line investments. The important lesson is that, although there is some substitution between mobile and main line, mobileÂ’s role as a complement dominates. As a result, mobile is not just picking up the slack where demand for main lines is unmet. A market for cellular phones exists beyond reducing the waiting list for traditional wire line phones. This possibility is important for private investors and governments, since the potential market size for both types of service is immense. People in developing countries do not demand cellular phones only as a second-best solution to becoming connected. A significant impetus for mobile demand could simply be due to its convenience (or its implications for social status). Thus, fixed-line providers should not expect a reduction in demand, even as cellular usage continues to expand. Competition from mobile can do much to improve main-line access.

PAGE 71

CHAPTER 4 SUBSIDIZING UNIVERSAL TELEPHONE SERVICE AT THE STATE LEVEL 4.1 Introduction Historically, the Federal Communications Commission (FCC) has pursued the goal of universal service (defined as maximum telephone subscribership) with various systems of crosssubsidization from toll to local telephone service. This system of subsidizing local calls by taxing toll calls resulted in large deviations between long distance and local rates, and researchers have questioned whether these cross-subsidies were achieving the universal service goals. Some find that universal service had little to do with states' adoption of subsidy programs. In an attempt to bring telephone tariffs closer to economic cost, the FCC introduced a subscriber line charge in 1984 that was supposed to shift part of the responsibility for access cost to the customer (Johnson 1988). The subscriber pays a fixed charge per month per line. The resulting increase in the flat rate for telephone service created concerns that the goal of universal service might be undermined. To encourage maximum subscribership, the FCC then initiated low-income programs to mitigate the effects of the higher rates and encourage subscribership. Among these programs was the Lifeline plan, established by the FCC in 1984. The aim of this plan is to aid in the promotion of universal service by helping lowincome individuals afford monthly telephone costs. 36 This plan allows for a reduction in fixed telephone service charges equal to the subscriber line charge (SLC) for eligible households. Each qualifying household receives assistance for a single telephone. The FCC provided a 50% reduction in SLC, provided that the participating state fund a reduction in basic local telephone rate equal to 50% of the SLC. Some states provided initiatives beyond the federal requirement by allowing low-income households additional support. As a result, the number of households eligible for support increased in those states. 65

PAGE 72

66 The purpose of this paper is to characterize states according to their timing of adoption of the Lifeline plan (using U.S. data between 1984 and 1997). Specifically, I explore whether telephone subscription rates or interest group pressure or both influence the likelihood of adoption. The results show that interest group pressure is important, but also that states do respond to subscription levels in determining their adoption decisions. The FCC did succeed in getting the states with the lowest penetration rates to adopt the policy. 37 Knowledge of the adoption patterns of the states is important because it can provide information regarding the pace of adoption and spread of future regulatory plans. It provides useful information for accessing whether the achievement of certain objectives requires regulation by the FCC (Donald and Sappington 1995). These issues are important even when access to the telephone network is relatively high. From a business perspective, telecommunications is important because of its potential influence on the performance of the economy. From a social welfare perspective, access is important to the extent that individuals can be connected to the community and for emergency assistance. It is important to pursue this research for these reasons, even though the adoption of the Lifeline plan was made mandatory in 1997. 38 37 There has been much discussion regarding the success of the Lifeline program as a tool to aid in the achievement of the universal service goals of the FCC. Garbaz and Thompson (1997) examine the effect of this policy on improving subscription rates. This research does not attempt to provide an answer to this particular question. That is, it does not address the question of how the Lifeline plan could be successfully implemented. Rather, this research should serve to inform regulators regarding the potential for the success of future regulatory plans based on state characteristics. 38 In 1997, the Universal Service Order modified the Lifeline program to make it available in all states. The state-matching requirement was also modified and the federal support amount was increased. Beginning January 1, 1998, Lifeline customers received $3.50 in federal support without a state-matching requirement. An additional $1.75 federal support is provided if states further reduce intra-state rates by at least $3.50. Prior to the 1997 Universal Service Order, the Lifeline program required a matching local rate reduction that had to be approved by the state utility commission. Since 1998, the states are no longer required to provide matching reductions.

PAGE 73

67 4.2 A Brief History of the Lifeline Plan Following the initiation of the optional Lifeline plan in 1984, the FCC revised the plan in 1985 to provide a reduction in fixed telephone charges equal to twice the amount of the SLC. Under the revised plan, the FCC provides a waiver of the full SLC up to the amount of reduction provided by the state. In order to participate, the state has to fund a reduction in the local service rate at least equal to the amount of the SLC. Eligible individuals have to satisfy a means test, based on income. The states are also required to establish procedures that verify that only eligible households are benefiting from the plan. Federal assistance for the Lifeline plan was funded through the carrier common line charge up to 1989. Beginning April 1989, funding was done through direct billing of interexchange carriers (IXCs) by the National Exchange Carrier Association. The IXCs are responsible for paying Lifeline assistance of at least 0.05 percent of the nationwide presubscribed lines. The revenue from this billing is used to pay certified local exchange carriers for matching Lifeline assistance costs. The federal portion of the Lifeline funds is funded from revenues from IXCs according to the number of pre-subscribed lines they serve. The Lifeline benefit is available to individuals verified by the state as eligible for a state public assistance program in accordance with a means test. Eligibility is based on participation in any one of the following five public assistance programs: Medicaid, public housing, energy assistance program, food stamps and supplemental security income. 39 The FCC specifies these public assistance programs, but each state may set its own guidelines by choosing one or more of the five public programs or by using additional programs to determine qualification for Lifeline assistance. 39 From Weinhaus et al. (2000). This research was conducted by the Telecommunications Industry Analysis Project (TIAP).

PAGE 74

68 Up to 1997, adoption of the Lifeline plan was optional. As a result, some states adopted the plan while others did not. By 1989, 26 states had adopted the plan at varying points in time. The earliest adoption was by California in 1983. This was more than a year before the FCC initiated its program. Other early adopters include Arkansas, New York, Vermont, Oregon, North Carolina, the District of Colombia and Arizona. After surveying related research, I outline the determinants of Lifeline adoption rates. Sections 5 and 6 develop the empirical model specification and present the data and empirical results. The concluding section summarizes the results. 4.3 Related Research StiglerÂ’s (1971) seminal paper, formalized by Peltzman (1976), marked the beginning of a growing body of work devoted to the economic theory of regulation. In these models, interest groups lobby for the regulator to institute policies to their benefit. The regulator in turn arbitrates among the various interest groups in a utility-maximizing fashion. Related work includes Joksow (1974), in which the regulator maximizes his utility by minimizing criticism from interest groups. Gary Becker (1983) and Dewey (2000) made some of the more recent contributions to the economic theory of regulation. Becker and Dewey extended earlier research to show that, where the majority rules, the success of an interest group is not wholly dependent on its relative size. They showed that small groups are not necessarily at a disadvantage, as there is a direct link between the efficiency with which political groups exert pressure and the benefits accrued by the particular interest group, regardless of its size. More efficient groups are likely to be winners while less efficient ones are likely to be losers. Other economists have concentrated on testing some of these theories empirically. The general consensus is that interest group pressure is an important determinant of the choice of regulatory policy. Kaserman, Mayo and Pacey (1993) showed that the likelihood of deregulating long distance telephone service increases with business usage and when residential interests are less extensive.

PAGE 75

69 Research on regulation began focusing on incentive regulation in later years as some states switched from rate-of-retum regulation to flexible pricing. Mathios and Rogers (1989) found that telephone prices are lower in states where price flexibility is allowed than in states where rate of return is practiced. Donald and Sappington ( 1 995) found that a stateÂ’s decision to replace rate-of-retum regulation with incentive regulation is directly related to local service rates, and particularly with high or low allowed earnings under rate-of-retum regulation and high growth in the state s urban population. Berg and Jeong (1991, 1994) studied the determinants and impacts of cost-component incentive regulation on heat rates in the electric utility industry. They found that heat rates' efficiency improved with incentive regulation and that high prices relative to average total cost may be responsible for triggering cost-component incentive regulation. The latter finding is in keeping with that of Donald and Sappington (1995) in the telecommunications industry. Research assessing the effect of later subsidy schemes intended to promote increased access to telephone service in the United States has also been done. Examples of these studies include Garbacz and Thompson (1997), Eriksson, Kaserman and Mayo (1997), and Kaserman, Mayo and Flyn ( 1 990). These studies show that the cross-subsidy mechanism used to achieve universal service has failed to do so. There appears to be little or no causal relationship between these subsidies and the growth in subscription rate. One argument used to support the ineffectiveness of these subsidies has to do with the inelastic nature of telephone demand. To achieve even small increases in telephone subscription, rates have to be significantly reduced. Taylor (1980) compiled a comprehensive review of telephone demand studies. These suggest that price and income elasticities tend to be small, but generally increase with distance. If this is true, then subsidizing local service is not going to be effective in adding households to the telephone network. In addition, the reduction in access attributable to the increase in long distance price will undermine any increase in subscription that might have resulted from lower local prices.

PAGE 76

70 However, Cain and Macdonald (1991), using 1987 data, argued that there is evidence that the use of low-income subsidies may help to improve subscription rates. They found that price has a noticeably stronger effect on access demand than previously believed, particularly at low levels of income. They argue that the elasticities have increased since 1980. This finding contradicts work done by Kaserman et al. (1990, 1997) and Garbacz et al. (1997), who obtain opposite results using post1980 data. The question regarding the effectiveness of these subsidy programs is therefore still unanswered. The FCC and state regulators continue to implement these programs, the latest of which includes the low-income programs. However, it is also important to determine the impact of federal initiatives in states. The present research uses data incorporating more recent years to investigate the effect of subscription and pressure groups on Lifeline adoption probabilities and characterizes the states according to the timing of their adoption. 4.4 The Determinants of Lifeline Adoption Rates 4.4.1 The Network Externality Versus the Interest Group View There are two views used to explain when states will adopt the subsidy program. The first, which seems to be the underlying assumption of the FCC, is that the Lifeline program is designed to motivate states with low subscribership level to increase penetration rates by offering subsidies to eligible households. We refer to this as the network externality view. Consumers benefit as more people are connected to the network, but private individuals do not consider other consumers' value when making their decisions to subscribe. In other words, consumers do not internalize the externality effect. The result is a divergence between the economic and socially desirable outcome. In maximizing the social good, regulators support policies that promote increased subscription because of this divergence. On the basis of this view, states with lower subscription are expected to adopt the lowincome policy more quickly. In other words, states with a more severe need for subsidization will adopt earlier. Adoption of the Lifeline plan would add individuals to the network who, ceteris

PAGE 77

71 paribus, would not have subscribed, thereby increasing overall subscribership. HouseholdsÂ’ quality of life is assumed to be better off with access to the network and ability to communicate with other members of the society and make calls in times of emergency. The welfaremaximizing regulator will be more strongly motivated to institute policies that increase subscribership to the extent that that welfare will be enhanced. From this perspective, regulators are welfare-maximizing agents. The externality effect is captured by telephone subscription (SUB84) in the empirical specification section. If this hypothesis that regulators respond to the universal service goal by adopting Lifeline is true, then SUB84 should be negative and significant. The rationale of the FCC is that regulators should aim to achieve maximum subscribership. If universal service appears to be threatened, the regulator is likely to adopt the lifeline plan more quickly than if telephone subscription is relatively high. Otherwise, we cannot reject the interest group view that universal service had little to do with the adoption of Lifeline. Subscription at the beginning of the data set (prior to the signing of letters of certification by the FCC) is used instead of annual subscription because of potential reverse causality between adoption probabilities and subscription levels. If individuals who otherwise would not be connected to the telephone network are connected, then telephone subscription will increase when the plan is implemented. Estimation of simultaneous models in this framework is extremely problematic. Thus, to avoid problems of simultaneity in the estimation of adoption patterns across states, the subscription rate in 1984 is used as an explanatory variable instead of annual rates. 40 The second view used to explain the timing of Lifeline adoption is more in line with the theories of regulation proposed by Stigler (1971) and Peltzman (1976). We refer to this as the 40 Garbaz and Thompson (200 1 ) suggest that the Lifeline effect on penetration rate was no different from zero. If this is true, then the reverse causality between telephone subscription and the Lifeline policy may not exist.

PAGE 78

72 interest group view. Unlike the network externality view, this view considers state commissions as self-utility-maximizing agents. That is, they will mandate the Lifeline subsidy plan if it serves their interests to do so. Regulators are subject to customer votes that may affect their future. As such, regulators favor actions that ensure favorable support from various interest groups. Since it is not always possible to satisfy all interest groups, the regulator will end up taking an action that satisfies the strongest group. Since increasing the rates on long distance calls funds the Lifeline program, the interest groups include long distance users and local customers. Assume that telephone users consist of residential customers only. Individuals who are eligible for the program, but who would have subscribed in its absence anyway, experience redistribution in their expenditures and there is no impact on subscribership. Conversely, subscribers not eligible for the subsidy are penalized by higher long distance rates. The only consumer beneficiaries are the subscribers who would not be connected to the network without the subsidy. Both beneficiaries and losers cast votes that affect the future of the regulator. The timing of state commissionsÂ’ adoption of the Lifeline subsidy will depend on the relative strength of these interest groups. If long distance users are represented more strongly in earlier years than in later years in a state, then it is unlikely that regulators in that state will be early adopters. The growth in poverty (POVGROW), the fraction of the population that is non-metropolitan (RURAL) and the relative size of the nonmetropolitan population (PRATIO) are introduced in the empirical specification of the model to account for the strength of the relevant interest groups. The interest group view supports a positive relation between the time to adoption and these variables. Poor households should welcome local rate reductions immediately since they are likely to qualify for assistance. Regulators may behave differently because of their utility preferences irrespective of pressure group support. Regulatory type may therefore affect adoption probabilities. Following Kaserman et al. (1993), I introduce the dummy variable REGELECT to account for regulatory type. Whether the regulator is appointed or elected may influence the commissionÂ’s interest and

PAGE 79

73 therefore the decision to adopt the low-income program. An appointed regulator is more likely to reflect the view of government officials, which may or may not be in the interest of a pressure group. The elected official, however, may be more likely to act in a pro-voter fashion since his/her tenure depends on voter preferences. 4.4.2 Other Determinants of Lifeline Adoption AVGRATE is introduced as a measure of the price of basic local telephone service in each state. We expect that adoption of the Lifeline plan is more likely where local prices are higher. 41 Just as the cost of telephone service to consumers (AVGRATE) is likely to influence adoption rate, it is also possible that the implementation of the policy may face budgetary constraints. The size of the regulatorÂ’s budget will therefore influence the pattern of Lifeline adoption across states insofar as it determines what activities are affordable to the commission. The relative attractiveness of a rate reduction is diminished if implementation and administration costs are high. Higher costs are likely to result in a delay of policy adoption. RABUDGET and RABUGPS and RABUGC are used as a measure of the costs that the regulator can afford to incur. 4.5 Empirical Specification and Data In this section, I use duration analysis to empirically test whether Lifeline adoption decisions within states depend on telephone subscription rates, interest group pressure, or both. The data consists of 49 states and the District of Colombia. (Hawaii was excluded because data was unavailable for key variables). Prior to 1 997, it was not mandatory for states to adopt this policy. As a result some states adopted while others did not. Table 4.1 categorizes states according to their adoption 41 According to Donald and Sappington (1995), commissioners may be willing to consider alternative regulatory plans that reduce or limit further increase in charges for basic telephone service in states where local rates are high.

PAGE 80

74 decision over the sample period. The explanatory variables used in the regressions are described in Table 4.2. Table 4.1: States Included in the Data Set Early Adopters Arizona, Arkansas, California, Colorado, Idaho, Maine, Montana, Maryland, Minnesota, Michigan, Missouri, Nevada, New Mexico, New York, North Carolina, North Dakota, Ohio, Oregon, Rhode Island, Texas, Utah, Vermont, Virginia, Washington, West Virginia, Washington D.C. Late Adopters Alaska, Alabama, Connecticut, Florida, Georgia, Illinois, Kansas, Massachusetts, Mississippi, Oklahoma, Pennsylvania, South Carolina, South Dakota, Tennessee, Wisconsin, Wyoming Non-adopters Delaware, Indiana, Iowa, Kentucky, Louisiana, Nebraska New Hampshire, New Jersey Note: Information for the table was taken from http/ftp.fcc.gov/Bureaus/Common_carrier/reports/FCC.state_link/monitor/mrp7-0.pdf. March 15, 2001. A state is classified as operating under Lifeline in a particular year as long as the FCC signed letters of certification some time that year. Early adopters are defined as states that adopted the Lifeline plan between 1984 and 1989. Late adopters include states that adopted the Lifeline plan in 1990 or later. Non-adopters are states that did not adopt the plan during the sample period 1984-97. Table 4.2: Variables Definitions and Sources Variable Definition Source SUB 84 Percentage of household that subscribe to the telephone network in each state. http/ ftp. fee . gov/Bureaus/ common_carrier /reports/FCC.State_link/monitor/mr970.pdf. March 15, 2001. POVGROW Growth in the fraction of households below the poverty line in each state. Bureau of Census. PRATIO Non-metropolitan population divided by metropolitan population. U.S. Department of Commerce, Statistical Abstract of the United States. RURAL Non-metropolitan population divided by population. U.S. Department of Commerce, Statistical Abstract of the United States REGELECT Dummy variable, equal to 1 if state commissioners are elected and zero otherwise. The Book of the States, 1984 -1997. Council of State Governments. AVGRATE The average of basic (single, unmeasured) monthly local service rate charged to residential consumers in the smallest and largest urban exchange. National Association of Regulatory Utility Commissioners, Bell Operating Companies exchange service telephone rate. December 31, 1984-1997. RABUDGET Real annual public utilities commission (PUC) budget. National Association of Regulatory Utility Commissioners Yearbook. RABUGPS Real budget per subscriber, measured by the real PUC budget weighted by the fraction of the population that resides in metropolitan areas. National Association of Regulatory Utility Commissioners Yearbook.

PAGE 81

75 Table 4.2 Continued Variable Definition Source RABUGC Real PUC budget divided by the state population. National Association of Regulatory Utility Commissioners Yearbook. 4.5.1 Statistical Model of Policy Adoption States are defined as non-adopters prior to initiating the Lifeline plan and as adopters once the program has been implemented. A stateÂ’s probability of adopting Lifeline is related to its characteristics listed above as exogenous variables in Table 4.2. Variations in Lifeline adoption dates are attributable to cross-sectional differences in these characteristics. Adoption probability is specified as: P(state i adopts the lifeline plan at time t) = f(Xj, tj) 42 where Xj is the set of exogenous variables that affect state iÂ’s adoption of Lifeline, and t, is state iÂ’s time to adoption. The time to adoption is distributed over (0, oo). Time 0 is the first year that a state became at risk of adoption. Some observations are right censored; that is, there are states that are still non-adopters at the end of the sample period. Estimations are conducted using hazard rate models. 43 The hazard rate (hi (t)) is the probability that state i will adopt Lifeline at time t conditional on having not adopted the policy before t. Hazard rate models are appropriate for this research because they provide information on the likelihood of transition from one condition to another. The structure of hazard rate models are such that the unconditional probability that state i will adopt at time t is m= ( 8 ) 42 This definition of adoption probability follows much of the theory of duration models. For detailed discussions on duration models, see Green (1997), Amemiya (1985). 43 Rose and Joksow (1990) used hazard rate models to analyze technology diffusion. Hazard rate models are used to estimate the instantaneous probability of leaving the present state.

PAGE 82

76 and the probability of state i not adopting the policy is 1 Fi(t) = exp i | h(s)ds V o ( 9 ) The Lifeline hazard rate is estimated by using maximum likelihood. To take account of right censoring of the data, a likelihood function of the following form is estimated. i=fl uncensoredobservations fm censoredobservations (1 Fit)) ( 10 ) Assumptions have to be made regarding the nature of the hazard rate in order to estimate the model. 4.5.2 Stationary Hazard Rate The model is fist estimated under the assumption that the likelihood of adopting the lifeline plan does not change throughout the entire sample period. This assumption means that the hazard rate is exponentially distributed. (Amemiya 1985). The hazard rate depends on the set of exogenous variables (Xj), such that hi = exp(or + Xfl) (H) Where a and p are unknown parameters. Denote states i =(1 . . .n) as states that completed their non-adoption spells of duration tj. States i=(n+l . . .N) are right-censored at the end of the sample period (T) because they never adopted Lifeline during the sample period. The likelihood function is L = El ex P( a + ^V?)exp(U exp (a + X/3)). exp(U exp(or + Xifi)) ( 1 2) <=1 i=n + 1 4.5.3 Non-Stationary Hazard Rate The model is then estimated under the assumption that the relative probabilities of adoption across states change through time. The non-stationary hazard model controls for trends or patterns of lifeline adoption in each state. In early years, there are more uncertainties surrounding the plan, in terms of how it should be implemented as well as the costs and benefits

PAGE 83

77 associated with implementations and monitoring. This uncertainty may reduce the adoption probabilities and result in slow adoption at first by some states. Over time, critics of low-income subsidies have emerged. Some doubt whether the lowincome programs are achieving their desired objective (Kaserman, Mayo and Flynn 1990 and Garbaz and Thompson 1997). At the same time, there appears to be some degree of success. If states that adopted Lifeline are viewed as successful, then other states are likely to adopt more quickly. It is not clear in theory whether the hazard rate is increasing or decreasing. The hazard rate is assumed to follow a Weibull distribution. This assumption allows estimation without having to know ex ante the exact direction of change of the hazard rate. The Weibull distribution allows the hazard rate to be comprised of a time component along with other explanatory variables. The hazard rate is defined as hi = cct a ~ x exp(A^), ( 13 ) i 5h where cct a 1 specifies the evolution of the hazard rate over time. If a > 1 : — >0 and the St hazard rate is increasing. If a = 1 : — = 0 , and the hazard rate is constant. If a < 1 : — < 0 St St and the hazard rate is decreasing over time. The likelihood function is L = n exp(Ai/?) x exp(exp (X/3)t“ )n exp(exp (Xij3)T ) . (14) <=i /=«+ 1 4.6 Data and Estimation Results The statistical models described above are based on time-varying characteristics of each state. They are estimated by using duration analysis data on Lifeline adoption decisions of forty-nine states and the District of Colombia, between 1984 and 1997. Table 4.3 reports means and standard deviations of the variables used in the sample by Lifeline adoption dates.

PAGE 84

78 Table 4.3: Sample Descriptive Statistics, 1984-97 By Lifeline Adoption Date Variable Full Sample Early Adopters Late Adopters Non-Adopters (N=371 N=99 N=160 N=1 12 REGELECT 0.315 0.140 0.463 0.250 (0.465) (0.349) (0.500) (0.435) RABUGET 8350.317 9176.280 9938.401 4957.307 (9588.435) (9819.811) (11276.71) (3963.799) RABUGPS 6156.419 7016.017 6955.307 3494.205 (8311.68) (9022.227) (9192.528) (4078.854) RABUGC 2.274 2.308 2.662 1.619 (1.985) (1.795) (2.469) (0.814) AVGRATE 12.465 12.875 12.348 12.334 (2.610) (2.90) (2.694) (2.260) SUB84 91.538 91.523 90.466 93.079 (3.674) (3.098) (4.161) (2.767) POVGROW -0.016 -0.021 -0.058 -0.009 (2.817) (3.274) (2.610) (2.603) RURAL 0.366 0.354 0.367 0.381 (0.220) (0.237) (0.213) (0.211) PRARTIO 0.845 0.918 0.868 0.711 (0.882) (1.108) (0.872) (0.440) Note: Standard deviations are shown in parentheses. While there are variations in the means across the three sub-samples, the differences between early and late adopters seem small. For instance, the differences in SUB84 between early and late adopters is only one percentage point and the differences in AVGRATE and RURAL are minute. Unexpectedly, the subscription rate is higher in states that adopted the Lifeline plan early. However, if we compare adopters (early and late) to non-adopters the pattern is as expected (SUB84 is higher in non-adopting states) and the difference in subscription rates across the states is more noticeable. Likewise, the difference in REGELECT across the three sub-samples is noticeable, but the means are all closer to zero than to 1, which indicates that more commissioners are appointed rather than elected. While differences in the budget of adopters are small, the noticeably smaller allocation to non-adopters suggests that the availability of resources is likely to determine the timing of the policy adoption. Likewise, the higher growth in poverty among non-adopters suggests that the ability to pay for telephone service is a function in determining whether to adopt the Lifeline

PAGE 85

79 plan, but that as the need base expands, undertaking the Lifeline plan may be too great a challenge and would be less likely to succeed. The data suggest that states with relatively large rural population are likely to adopt the plan later. This may be a reflection of the potential that a powerful (though numerically small) interest group may have on the future of the regulator, in terms of their efforts to generate votes. Before presenting the regression estimates, I chart the diffusion path using the Weibull and exponential distributions that are assumed for the statistical analysis. Figures 4. land 4.2 fit the survival functions evaluated for a state close to the means of the covariates over the sample period. Both functions illustrate at least a slight increase in the hazard of adoption over time. The graphs show that at least one state adopts the Lifeline plan each year. This suggests that a stateÂ’s likelihood of adopting the Lifeline plan in any given year (conditional on its duration up to that point) is increasing over time. In other words, a state is more likely to adopt the Lifeline plan in the current year than it was in previous years. By 1990 close to half the states had adopted the plan. The survival functions of the two parametric models (Weibull and exponential) suggest slightly different hazards of adoption. By imposing a particular structure on the data, these models may distort the estimated likelihood of survival. As a result, the KaplanMeier (non-parametric) estimate is also presented. This representation of the survival functions may be more accurate because it imposes fewer restrictions on the data. Figure 4.3 plots the Kaplan-Meier estimates of the survivor function. The function shows that the likelihood of adopting the Lifeline plan is relatively small during the first three years (1984-86). From 1986 to 1988 (corresponding to analysis time 3-5) the hazard of adoption appeared fairly constant and relatively large, but flattens slightly subsequently towards the end of the period. The general pattern suggests that continued adoption of the Lifeline plan by the remaining twenty-eight states was likely to continue past 1988 even without the mandatory adoption rules of 1997.

PAGE 86

80 With this basic understanding of the Lifeline adoption pattern, we now turn to the regression estimates of diffusion patterns. Tables 4.4 and 4.5 present the estimates of the Weibull and the exponential diffusion models for the entire data set, while Tables 4.6 and 4.7 report the estimations using data, which excludes the state of California.

PAGE 87

81 Figure 4.1: Survivor Function-Webull Regression Model

PAGE 88

Fraction Surviving 82 1.0 Duration 13.0 Figure 4.2: Survivor Function-Exponential Regression Model

PAGE 89

Fraction Surviving 83 Figure 4.3: Survivor Function-Kaplan-Meier Estimate

PAGE 90

84 Tables 4.4 and 4.5 report the estimates of the Weibull and exponential adoption models. The results are similar across the two model specifications. States that start out with low telephone subscription rates are likely to adopt the Lifeline plan earlier than states with higher household subscription rates. A one-unit increase in telephone subscription rate cuts the hazard (likelihood) of adoption by roughly 90 percent. This result is consistent with the view that the Lifeline plan is a response to universal service goals. That is, states are likely to adopt the Lifeline plan in order to boost low subscription rates. However, none of the variables intended to capture the effect of interest groups is significant, except for the exponential estimation of RURAL. This suggests that alternative proxies of interest groups may be required. Basic local rates appear to have a positive impact on adoption probabilities in all but two specifications. Table 4.4: Weil jull Hazard Estimates of Lifeline Adoption Probabilities Variables RABUDGET RABUGPS RABUGC RURAL PRATIO REGELECT -0.227*** (-3.061) -0.067*** (-3.312) -0.206*** (-3.832) -0.073* (-3.076) -0.080* (-3.141 RABUDGET 1.000*** (5.171) RABUGPS 1.00*** (3.602) RABUGC 1.123*** (4.357) j J79*** (3.635) 1.163*** (3.334) AVGRATE 1.148** (2.013) 1.168** (2.061) 1.114 (1.586) 1.180* (1.900) 1.132 (1.559) SUB84 -0.905** (-2.217) -0.907** (-2.219) -0.891*** (-2.773) -0.885*** (-2.851 -0.896** (-2.422) POVGROW 1.155 (1.521) 1.185 (1.200) 1.158 (1.543) RURAL -0.100 (-1.510) PRATIO 0.688 (-0.700) N 267 194 243 195 195 Log Likelihood -34.969 -25.650 -32.145 -27.003 -28.110 a 1.713 1.788 1.716 1.774 1.728 Note: t-statistics are shown in parentheses. ***=significant at the 1% level; **= significant at the 5% level; and *= significant at the 10% level.

PAGE 91

85 Variables RABUDGET RABUGPS RABUGC RURAL PRATIO REGELECT -0.313*** (-3.155) -0.101*** (-3.626) -0.281*** (3.998) -0.099*** (-3.079 -0.101*** (-3.159) RABUDGET 1.00*** (6.497) RABUGPS 1.000*** (4.231) RABUGC 1 123*** (4.655) 1.187*** (3.950 1.174*** (3.526) AVGRATE 1.092* (1.932) 1.100* (1.803) 1.063 (1.308) 1.109* (1.725) 1.077 (1.260) SUB 84 -0.914** (-2.267) -0.911 (-2.245) -0.902*** (-2.680) -0.888*** (2.767 -0.900** (-2.352) POVGROW 1.171 (1.564) 1.190 (1.281) 1.169 (1.554) RURAL -0.11 8-* (-1.642 PRATIO 0.685 (-0.819) N 267 194 243 195 195 Log Likelihood -37.939 -28.253 -35.231 -29.429 -30.343 the 5% level; and *= significant at the 10% level. The hazard ratio estimates show that a one-unit increase in basic local rates can increase the hazard of adoption by up to 1.17 times. This result is expected. If local rates are high, they can act as a deterrent to subscribers. The Lifeline plan may be an attractive avenue for providing local rate discounts to select consumers. The negative, although insignificant, effect of RURAL suggests that the preference will be not to adopt the lifeline plan in heavily rural areas. This result makes sense if we assume that potential users believe that the likelihood of benefiting from the program is low. Garbaz and Thompson (2001) suggest that 95-100 percent of all households receiving Lifeline subsidies would have been on the telephone network without any subsidy. Alternatively, the higher cost of service for less dense areas reduces the suppliersÂ’ interest in expanding penetration. The revenue per phone is less than the cost of providing the service.

PAGE 92

86 The consistently significant and positive estimates of RABUGET, RABUGPS, RABUGC are according to our expectations. Regulators with access to a larger budget have greater access to information and resources, and are better able to determine to determine the pros and cons of adopting a new policy. In addition, states with larger per capita budgets are in a position to adopt the policy even where the cost per subscriber is relatively high. A one-unit increase in the percapita budget of the regulator can increase the likelihood of failure by as much as 1 .2 times. The Weibull estimate of a 44 is greater than 1, which suggests that the hazard rate increases through time. When the hazard ratio estimate is used on RABUGC, an a value of 1 .716 indicates that are 1 .5 times more likely to adopt the Lifeline plan after seven years states than after three years. States in which the regulators are elected rather than appointed are likely to adopt the Lifeline plan later. The estimated hazard ratios suggest that the hazard of adoption is up to a third less if the regulator is elected rather than appointed. Donald and Sappington (1995) suggest that, appointed regulators are more likely to be insulated from the direct action of voters, and so may be more willing to experiment with new policies than if they were elected. Elected regulators are less likely to be reelected if their policies are unpopular and so may prefer to observe the success of a policy in other states before adoption in their own state. California adopted the Lifeline plan in 1983, one year prior to the institution of the optional Lifeline plan by the FCC. To check whether the special circumstances of California had any impact on the results, I re-estimated both the Weibull and exponential duration models. The results are presented in Tables 4.6 and 4.7, and are not qualitatively different from those that included California. 44 a is an estimated parameter that tells whether the data used in the estimation has a hazard rate that is monotonically increasing or decreasing through time. It estimates the shape of the hazard rate through time.

PAGE 93

87 Table 4.6: Weibull Hazard Estimates of Lifeline Adoption Probabilities Excluding California Variables RABUDGET RABUGPS RABUGC RURAL PRATIO REGELECT -0.229*** (-2.949) -0.075*** (-3.358) -0.197*** (-3.753) -0.070*** (-3.061) -0.078*** (3.108) RABUDGET 1.000*** (4.882) RABUGPS 1.000*** (3.060) RABUGC 1.129*** (4.539) 1.182*** (3.569) 1.167*** (3.344) AVGRATE 1.169** (2.205) 1.185** (2.162) 1.145** (2.054) 1.214** (2.151) 1.167** (1.964) SUB84 -0.905** (-2.158) -0.909** (-2.198) -0.890*** (-2.695) -0.886*** (2.810) -0.897** (-2.344) POVGROW 1.154 (1.526) 1.184 (1.179) 1.155 (1.502) RURAL -0.115 (-1.363) PRATIO -0.712 (-0.620) N 266 193 242 194 194 Log Likelihood -33.539 -24.918 -29.765 -24.972 -25.922 a 1.819 1.888 1.839 1.920 1.872 Note: t-statistics are shown in parentheses. ***=significant at the 1% level; **= significant at the 5% level; and *= significant at the 10% level. Table 4.7: Exponential Hazard Estimates of Lifeline Adoption Probabilities Excluding California Variables RABUDGET RABUGPS RABUGC RURAL PRATIO REGELECT -0.328*** (-3.065) -0.106*** (-3.503) -0.288*** (-3.921) -0.102*** (-3.059) -0.104*** (-3.117) RABUDGET 1.000*** (6.476) RABUGPS 1.000*** (3.362) RABUGC 1.126*** (4.888) 1 189*** (3.919) 1.177*** (3.542) AVGRATE 1.020** (2.093) 1.100* (1.855) 1.080* (1.734) 1.125** (1.965) 1.097* (1.611) SUB 84 -0.916** (-2.231) -0.912** (-2.238) -0.904** (2.621) -0.892*** (2.696) -0.903** (-2.262) POVGROW 1.171 (1.556) 1.189 (1.269) 1.167 (1.496) RURAL -0.142 (-1519) PRATIO -0.710 (-0.754)

PAGE 94

88 Table 4.7 Continued Variables RABUDGET RABUGPS F IABUGC RURAL PRATIO N 266 193 242 194 194 Log Likelihood -37.164 -27.881 -33.657 -28.111 -28.854 Note: t-statistics are shown in parentheses. ***=significant at the % level; **= significant at the 5% level; and *= significant at the 10% level. To translate the effect of the independent variables on the likelihood of adoption on their meaning for the timing of adoption, I report elasticities up to the time of adoption with respect to each of the same independent variables. The results are reported in Tables 4.8 and 4.9. All elasticities are reported at the sample means. The point estimates of the elasticites are similar across the Weibull and exponential models. Even though the t-statistics on the exponential estimations for AVGRATE elasticities are small, the point estimates are quite similar to their corresponding Weibull estimation. Telephone subscription has by far the greatest impact on timing of the adoption of the Lifeline plan. On the basis of the Weibull estimation, a 1 percent increase in telephone subscription rate will lead to a 5-8 percent increase in the duration time (that is, an increase in the length of time that a state continues without adopting the policy). This means that a state that adopted the plan in 1990 would adopt four years later (1994) if the subscription rate in that state increased by 1 percent. The time until adoption is also sensitive to the local telephone rate. A 10 percent increase in the average local telephone rate can reduce the duration time by up to 13 percent. Telephone demand is believed to be inelastic, meaning that telephone prices really did not influence the extent of a subscriberÂ’s use of the telephone network. If this is true, then it would seem that policymakers should not respond to subscription rates in their decision to adopt the Lifeline plan. However, the Lifeline policy is designed to target states with low subscription rates attributable to unaffordable telephone rates. It may be that, once individuals are connected to the network, their response to changes in local rates is small, but the decision to get connected might be very responsive to the monthly expenditures that come with being connected. In other

PAGE 95

89 words, individuals who are more price-sensitive may be the ones who are not connected to the network. Table 4.8: Wei )ull Elasticity Estimates of the Survival Time of Lifeline Variables RABUDGET RABUGPS RABUGC RURAL PRATIO REGELECT 0.301*** (3.57) 0.483*** (3.83) 0.3 14*** (4.44) 0.468*** (3.41) 0.464*** (3.42) RABUDGET -0 144*** (-4.68) RABUGPS -0 177*** (-3.83) RABUGC -0.148*** (-4.75) -0.201*** (-3.61) -0.189*** (-3.25) AVGRATE 1.010** (-2.56) -1.087*** (-2.78) -0.786* (-1.93) -1.169** (-2.40) -0.896** (-1.86) SUB 84 5.340** (2.13) 4.961** (2.00) 6.183** (2.50) 6.318** (2.36) 5.820** (2.09) POVGROW 0.015 (-1.34) 0.030 (-1.08) 0.019 (-1.35) RURAL 0.461 (1-49) PRATIO 0.166 (0.69) Note: t-statistics are shown in parentheses. ***=significant at the 1% level; **= significant at the 5% level; and *= significant at the 10% level. All elasticities are evaluated at the means of the independent variables. Compared to SUB84 and AVGRATE, elasticities of the variables measuring the interest group effect on the timing of adoption tend to relatively small. If we view regulators as selfinterested parties, then REGELECT is the only measure of interest group pressure that has a significant impact on the timing of adoption of the Lifeline plan. States with elected regulators have a duration time that is 1/3 of a percent greater than it would be if the regulators were appointed. This result suggests that even though elected official may be more likely to impose pro-voter reforms, their interests closely coincided over the Lifeline policy.

PAGE 96

90 Table 4.9: Exponential Elasticity Estimates of the Survival Time of Lifeline Variables RABUDGET RABUGPS RABUGC RURAL PRATIO REGELECT 0.404*** (3.16) 0.734*** (3.63) 0.433*** (4.00) 0.734*** (3.08) q 729*** (3.16) RABUDGET -0.241*** (-6.50) RABUGPS -0.296*** (-4.23) RABUGC -0.254*** (-4.65) -0.371*** (-3.95) -0.347*** (-3.53) AVGRATE -1.100* (-1.93) -1.159* (-1.80) -0.760 (-1.31) -1.306* (-1.72) -0.932 (-1-26) SUB84 8.204** (2.27) 8.485** (2.25) 9.398*** (2.68) 10.872*** (2.77) 9.671** (2.35) POVGROW 0.028 (-156) 0.054 (-1-28) 0.035 (-1-55) RURAL 0.759* (1.69) PRATIO 0.289 (0.82) Note: t-statistics are shown in parentheses. ***=significant at the 1% level; **= significant at the 5% level; and *= significant at the 10% level. All elasticities are evaluated at the means of the independent variables. 4.7 Conclusions The results suggest that states with relatively low telephone penetration tended to lead the U.S telephone industry in the adoption of the Lifeline plan. All regression estimates show that states with low telephone penetration rates are more likely to be among the early adopters. This result suggests that the efforts of the FCC to encourage states with low penetration levels to adopt the Lifeline plan appear to have worked. It appears that states adopt the plan as a means of achieving universal service goals; whether the adopting states benefited significantly in terms of increased penetration is an entirely different question. The results suggest that, if the Lifeline plan has not resulted in providing access to individuals who otherwise would not have access to the telephone network, there might have been a problem at the implementation stage of the plan. The question of how the program could be effectively implemented therefore remains. It is important to find the answer to this question for the purposes of future policy application.

PAGE 97

91 The analysis provides some evidence that telephone rates may be an important factor influencing the time of adoption. Adoption of the Lifeline plan is likely to take place earlier if the local telephone rate is relatively high. Finally, the influence of stakeholdersÂ’ interest in influencing adoption decisions can be supported to the extent that regulators are self-utility maximizing agents. The results strongly suggest that states where regulators are elected, rather than appointed, are less likely to be among early adopters. However, variables that are used as proxies for the effect of interest groups generally were not significantly different from zero, although they had the expected signs. This suggests that alternative proxies for interest group effect may be required.

PAGE 98

CHAPTER 5 CONCLUDING REMARKS This dissertation has discussed issues involved in the provision of telephone service and in the accomplishment of increased connectivity, in the United States and Africa. By looking at issues in these two regions, it became clear that the basic aim of improving sector performance is an ongoing one, regardless of the level of sector development. From chapter 2, the main lesson is that in order for there to be significant improvement in telephone access level in Africa, there has to be a tremendous improvement in the risk environment in the region. This is true whether public or private investors will undertake telecom investments. Like private investment, public investment is likely only if there are positive returns to that investment. Private investors have to be concerned with the risk of losing their investments or having the returns confiscated. Public investors are interested in the risk environment insofar as it reflects the likelihood of their being around to recoup the benefits (politically or otherwise) of the investments. Analysis of the data reveals that the access level is extremely low in Africa, but the large waiting lists indicate that the market potential in that region is immense. With the rapid expansion of cellular usage, the options for improving telephone access have increased. Through its potential for competition, mobile provision encourages wire line providers to expand their service. In addition, as a substitute for main line phones, mobile telephones provide an alternative to policymakers or regulators to extend telephone access through cellular provision. This policy implication is important because it suggests that low access levels in Africa may be resolved more quickly than if the sole process for doing so was via main line provision. The potential for using cellular service in this role suggests that its importance will continue to grow 92

PAGE 99

93 in the region and may be seen as the solution to low connectivity that is widespread throughout Africa. With new developments in the achievement in telecommunications reform also comes new challenges, as demonstrated in my analysis of Lifeline policy reform in the United States. Despite the high telephone access level in the United States, there is still room for improvement. Chapter 4 contributes to the understanding of the reform environment by characterizing states within the United States according to the timing of their adoption of the Lifeline plan. Understanding the adoption patterns provides information regarding the potential success or failure of future regulatory plans.

PAGE 100

APPENDIX A DATA ANALYSIS FOR CHAPTER 2 Table Al: Countries Included in the Sa m ple Country Legal System Official Language Algeria French, Islamic Arabic Angola Portuguese Civil Portuguese Botswana Roman-Dutch English Burkina Faso French Civil French Cameroon French Civil French Congo, Dem. Rep. Belgian Civil French Congo, Rep. French Civil French Cote dÂ’Ivoire French Civil French Egypt English, Islamic Arabic Ethiopia Transitional (regional and national courts) English Gabon French Civil French The Gambia English Common English Ghana English Common English Guinea French Civil French Guinea Bissau N/A Portuguese Kenya English Common English Liberia Anglo American and Customary English Libya Italian Civil and Islamic Arabic/English Madagascar French Civil French Malawi English Common English Mali French Civil French Morocco Islamic, French, Spanish Arabic Mozambique Portuguese Civil Portuguese Namibia Roman-Dutch English Niger French Civil French Nigeria English Common English Senegal Portuguese Civil French Sierra Leone English Common English Somalia N/A Arabic/English South Africa English, Roman-Dutch English Sudan English Common Arabic/English Tanzania English Common English Togo French Civil French Tunisia French Civil French Uganda English Common English Zambia English Common English Zimbabwe English Common, Roman-Dutch English Source: CIA World Factbook 1999. 94

PAGE 101

Table A2: Panel Fixed-Effects Regressions with One Country Excluded (Dependent Variable: Network Access Per 1,000 Inhabitants) Country Explanatory Variables Excluded 95 i fcul <*> ~a . . < OZ U O 2 w Q CL O PZ CL X w H u < K f— z o u p < w 3 p CQ £ < eu 5 2 o u z < ua Q < oZ u CL Q U o o si ON K 3 NO _ ON Ov 3 o — 1 r— VO 00 3 2 ^ Os PO o o csi 02 s 1 NO ^3 * 00 3 no 3 2 d 2 (N m _ NO • m TT 8 NO gS $ o NO O 2 p" d ^ NO * — 7 ~T no NO . 8^ * 00 s £ NO r-s P* ON VO r}™ oo (N £ K C o g £ 3 'O d 2 r2 o On 3 2 3 s /-tv ^ 3 o rNO NO §o If d 2 P 2 (N * i/“) O 3 o 2 °° o .2 O 03 ao P N

PAGE 102

96 Table A3: Comparison of Countries Based on Means Countries Included in Sample Countries Excluded from Sample Countries Excluded minus Seychelles and Mauritius. NETWORK 13.573 22.700 8.823 PER CAITA GDP 860.257 1039.379 499.221 LTRADE 60.501 89.085 83.711 URBAN 34.276 33.68 31.696 Table A4; Comparison of Network Regression Based on Observable Variables. All Countries All Countries Countries Countries minus Seychelles Excluded from Excluded minus and Mauritius Sample Seychelles and Mauritius. PER CAITA 32.232* 7.815* 72.057* 6.934* GDP (4.089) (4.152) (3.301) (2.664) LTRADE -0.057*** -0.044* -1.108 0.006 (-1.902) (-3.643) (-1.091) (0.254) URBAN 1.530* 1.258* 1.588* 1.070* (8.839) (10.269) (4.086) (3.819) N 614 588 188 162

PAGE 103

APPENDIX B DATA ANALYSIS FOR CHAPTER 3 Table Bl: Comparison of Means across Counties Included in and Excluded from the Sample Variable Countries Included Countries Excluded Countries Included Without South Africa Countries Excluded without Seychelles and Mauritius MAINLINE 15.206 16.350 11.94 9.338 MOBILE 0.439 0.349 0.170 0.100 Per capita GDP 1,000.507 811.978 882.776 492.73 LTRADE 59.570 77.908 60.059 74.065 URBAN 35.547 32.286 34.938 31.194 Table B2: Comparison of Fixed-Effect Regressions across Counties Included in and Excluded from the Sample. Dependent variable: MAINLINE per 1,000 individuals. Variable Countries Included Countries Excluded Countries Included Without South Africa Countries Excluded without Seychelles and Mauritius All of Africa Africa without South Africa, Mauritius and Seychelles MOBILE 0.818*** (6.028) 4.476 (11.104)* 1.0671** (2.189) 1.958*** (6.445) 2.435*** (3.356) 1.781*** (4.032) LOGGDP 12.367*** (4.276) 17.867 (3.954)* 13.807*** (4.848) 3.916** (2.556) 22.530*** (5.285) 7.607*** (4.778) LTRADE -0.052*** (-3.067) -0.027 (-1.072) -0.049*** (-2.900) -0.080 (-0.645) -0.038** (-1.987) -0.026** (-2.549) URBAN 1.059*** (12.318) 0.989* (4.621) 1.009*** (12.052) 0.826*** (3.800) 1.017*** (8.707) 0 945 *** (10.272) Note. The t-stats are shown in brackets. *** = significant at 1%. ** = significant at 5%. 97

PAGE 104

98 Table B3: Pooled Time-Series and Cross-Sectional Regression Controlling for CountryLevel Differences. Data: 1985-97 Excluding South Africa. Variable Dependent variable: MAINLINE per 1,000 people Dependent variable: DEMAND for main lines per 1,000 people Fixed-Effect Estimation Fixed effect estimation with interactions Fixed effect estimation with interactions Fixed effect estimation with lagged main line and interactions LOGGDPC 9 572*** (3.579) 8.797*** (3.285) 2.718 (0.784) 2.827 (0.960) LTRADE -0.115*** (-6.262) 0 . 112 *** (-6.272) -0.083*** (-3.535) -0.551*** (-2.797) URBAN 0.577* (6.156) 0.595*** (6.382) 0.961*** (10.421) 0.898*** (10.429) CORRUP 0.421 (0.962) 0.344 (0.766) -0.130 (-0.196) 0.443 (0.835) LAW 1.187** (2.365) 0.783* (1.753) 0.245 (0.459) 0.111 (0.237) BUREAU -2.839*** (4.942) -2.250*** (-4.234) -1.047 (-1.437) -0.952* (-1.606) CONTRACT 1.982*** (4.773) 1.700*** (4.339) 1.616*** (3.328) L449*** (3.692) EXPROP 0.312 (0.860) 0.465 (1.385) 0.048 (0.131) -0.167 (-0.562) GOVOP 0.347** (2.299) 0.299** (2.322) 0.600*** (3.084) 0.611*** (3.932) DISCTAX 0.118* (1.771) 0.113* (1.759) 0.080 (0.711) 0.053 (0.526) DEMOC -0.398*** (-3.816) -0.381*** (-3.708) -0.334** (-2.348) -0.244** (2.403) MOBILE 1.098** (2.652) 6 . 121 *** (5.309) 3.966*** (3.397) 2 775 *** (2.837) MOBINC -0.0004* (1.823) 0.00003 (0.148) 0.00002 ( 0 . 110 ) MOB SUB -3.512** (-2.320) -3.541*** (-2.816) -2.418** (2.434) LMAINLINE 0.160*** (3.059) N 286 286 239 238 ~R ! ~ 0.94 0.95 0.98 0.99 Note: T-stats are shown in brackets. ***= significant at 1%. ** = significant at 5%, * = significant at 10%. A test of the significance of a time trend could not be rejected. As a result, the time trend was excluded from estimations.

PAGE 105

REFERENCES Ahan, Hyungtaik, and Myeong-Ho Lee. 1999. An Econometric Analysis of the Demand for Access to Mobile Telephone Networks. Information Economics and Policy 1 1 : 297-305. Alesina, Alberto, and R. Perotti. 1996. Income Distribution, Political Instability, and Investment. European Economic Review 40: 1203-28. Anremiya, Takeshi. 1985. Advanced Econometrics. Cambridge, Massachusetts: Harvard University Press. Antonelli, C. 1993. Investment, Productivity, and Growth. Adoption in Advanced Telecommunications. Journal of Economic Behavior and Organization 20: 227-45. Becker, Gary, 1983. A Theory of Competition Among Pressure Groups for Political Influence. Quarterly Journal of Economics 98: 371-400. Bennett, P. D., and R. Green. 1972. Political Instability as a Determinant of Direct Foreign Investment in Marketing. Journal of Marketing Research 19: 182-86. Berg, Sanford, and Jinook Jeong. 1991. An Evaluation of Incentive Regulation for Electric Utilities. Journal of Regulatory Economic 3: 45-55. . 1994. An Evaluation of Incentive Regulation: Corrections. Journal of Regulatory Economics 6: 321-328. Bergara, M., W. Henisz, and P. T. Spiller. 1998. Political Institutions and Electric Utility Investments: A Cross-Nation Analysis. California Management Review 40 (2): 18-35. Bleaney, Michael, and David Greenway. 2001. The Impact of Terms of Trade and Real Exchange Rate Volatility on Investment and Growth in Sub-Saharan Africa. Journal of Development Economics 65: 491-500. Block, Stephen. 2001. Does Africa Grow Differently? Journal of Development Economics 65: 443-67. Cain, Paul, and James Macdonald. 1991. Telephone Pricing Structures: The Effects on Universal Service. Journal of Regulatory Economics 3: 293-308. Chance, C., Booz, Allen, and Hamilton, Inc. 1998. Case Study on the Impact of the Changing International Telecommunications Environment: Uganda. London: Booz, Allen, and Hamilton, Inc. Chong, Alberto, and Lisa Zanforlin. 2000. Law Tradition and Institutional Quality: Some Evidence. Journal of International Development 12: 1057-1068. 99

PAGE 106

100 CIA World Factbook. 1999. Washington DC: Central Intelligence Agency. Dewey, James. 2000. More is Less? Regulation in a Rent Seeking World. Journal of Rgulatory Economics 18(2): 95-1 12. Donald, Stephen, and David Sappington. 1995. Explaining The Choice Among Regulatory Plans in the U.S. Telecommunications Industry. Journal of Economic and Management Strategy 4 (2): 237-265. Eriksson, Ross, David Kaserman and John Mayo, 1997. Targeted and Untargeted Subsidy Schemes: Evidence of Post-Divestiture Efforts to Promote Universal Telephone Service. Unpublished Draft, University of Tennessee. Freedom in the World Survey 1998-1999. Washington D.C: Freedom House. Frempong, G. K., and W. H. Atubra. 2001. Liberalization of Telecoms: The Ghanaian Experience. Telecommunications Policy 25: 197-210. Garbacz, Christopher, and Herbert Thompson Jr. 1997. Accessing the Impact of FCC Lifeline and Link-Up Programs on Telephone Penetration. Journal of Regulatory Economics 11: 67-78. . 2001 . Universal Service Versus Universal Competition: A Review Article of Crandall and Waverman. Journal of Regulatory Economics 19(1): 93-96. Goldsmith, Arthur. 2001. Risk, Rule and Reason: Leadership in Africa. Public Administration and Development 21: 77-87. Green, William. 1997. Econometric Analysis . Upper Saddle River, New Jersey: Prentice Hall. Gutierrez, L. 1999. Three Essays on Latin American Telecommunications Reform. Ph.D. Dissertation. University of Florida, Department of Economics. Guiterrez, L., and S. Berg. 2000. Telecommunications Liberalization and Regulatory Governance: Lessons from Latin America. Telecommunications Policy 24: 865-84 Gwartney, James, and Robert Lawson. 1997. Economic Freedom of the World: 1997 Report. Vancouver: Fraser Institute. Haan, J., and C. Siermann. 1995. New Evidence on the Relationship Between Democracy and Economic Growth. Public Choice 86: 175-98. Hamilton, Jacqueline. 2002. Are Main Lines and Mobile Phones Substitutes or Complements? Evidence from Africa. Telecommunications Policy (In Press). Henisz, Withold. 1998. State Sector Opportunism and Opportunistic Use of the State. Unpublished Ph.D. Dissertation. University of California, Berkeley.

PAGE 107

101 • 2000. The Institutional Environment for International Investment. Safeguarding Against Sector Opportunism and Opportunistic Use of the State. Ph.D. Dissertation. University of California, Berkley. Henisz, Withold, and Bennet A. Zelner. 2001. The Institutional Environment for Telecommunications Investment. Journal of Economics and Management Strategy 10(1): 12347. Henisz, Withold, and Andrew Delios. 2002. Learning about the Institutional Environment. Strategic Management (In Press). Hsiao Cheng. 1986. Analysis of Panel Data. New York: Cambridge University Press. http/ftp.fcc.gov/Bureaus/Common_carrier/reports/FCC.state_link/monitor/mrp97-0.pdf. March 15,2001. International Telecommunications Union. 1998. African Telecommunication Indicators 1998. Geneva: ITU. International Telecommunications Union. 1999. World Telecommunication Development Report 1999. Mobile Cellular. Executive Summary. Geneva: ITU. Jaggers, Keith, and Robert Gurr. 1995. Tracking Democracy’s Third Wave with Polity III Data. Journal of Peace Research 32(4): 469-82. • 1996. Polity III: Regime Change and Political Authority, 1800-1994. 2 nd ICPSR version (computer file). Ann Arbor, Michigan: Inter-University Consortium for Political and Social Research. Jha Raghbendra, and Sumit K. Majumdar. 1999. A Matter of Connections:OECD Telecommunications Sector Productivity and the Role of Cellular Technology Diffusion. Information Economics and Policy 1 1 : 243-269. Johnson, Leland. 1988. Telephone Assistance Programs for Low-Income Households. A Preliminary Assessment. Santa Monica, California: The Rand Publication Series. The Rand Corporation. Joskow, Paul. 1974. Inflation and Environmental Concern: Structural Change in the Process of Public Utility Price Regulation. Journal of Law and Economics 291-327. Kaserman, David, John Mayo and Joseph Flynn. 1990. Cross-Subsidization in Telecommunications: Beyond the Universal Service Fairy Tale. Journal of Regulatory Economics 2: 231-249. Kaserman, David, John Mayo, and Patricia Pacey. 993. The Political Economy of Deregulation: The Case of Intrastate Long Distance. Journal of Regulatory Economics 5: 49-63. Kerf, M., and W. Smith. 1996. Privatizing Africa 's Infrastructure: Promise and Challenge. Washington, D.C: World Bank Group.

PAGE 108

102 Knack, Stephen. IRIS-3 File of International Country Risk Guide data, 1982-1997. IRIS Center, University of Maryland. Knack, Stephen, and P. Keefer. 1997. Why DonÂ’t Poor Countries Catch Up? A Cross-National Test of an Institutional Explanation. Economic Enquiry 35: 590-602. Knack, Stephen, and P. Keefer. 1995. Institutions and Economic Performance: Cross-Country Tests Using Alternative Measures. Economics and Politics 7: 207-27. Kwoka , John E. 1993. The effects of Divestiture, Privatization and Competition on Productivity in U.S. and U.K. Telecommunications. Review of Industrial Organization 8: 49-61. La Porta, Rafael, F. Lopez-de-Silanes, A. Shleifer, and R. Vishny. 1999. The Quality of Government. Journal of Law, Economics and Organization 15 (1): 222-82. Laidlaw, Bruce, and Christine Parkinson. 1995. Africa Telecommunications Policy Study: Uganda Report. Washington, D.C: World Bank Group. LeblancWohrer, Marion, and Ilka Lewington. 2000. Utility Regulation in Africa and the Middle East. Regulation of Liberalized Utilities and Network Industries. Utility Regulation 2000 Series 4. Levy, Brian, and Pablo T. Spiller. 1994. The Institutional Foundations of Regulatory Commitment: A Comparative Analysis of Telecommunications Regulation. Journal of Law, Economics and Organization 10: 201-46. Mathios, Alan, and Robert Rogers. 1989. The Impact of Alternative Forms of State Regulation of AT&T on Direct-Dial, Long Distance Telephone Rates. Rand Journal of Economics 20(3): 437-453. Minges, Michael. 1999. Mobile Cellular Communications in the Southern African Region. Telecommunications Policy 23: 585-93. Mustapha, M., B. Laidlaw and M. Brand. 1997. Telecommunications Policies for Sub-Saharan Africa. Washington, D.C: World Bank Group. National Association of Regulatory Utility Commissioners. Bell Operating Companies Exchange Service Telephone Rates. Washington , DC: NARUC, 1984-1990. National Association of Regulatory Utility Commissioners. Yearbook. Washington, DC: NARUC, 1995-1996, 1994-1995, 1993-1994, 1992-1993, 1991-1992. Noll, Roger. 1999. Telecommunications Reform in Developing Countries. Working Paper 9910. Washington, D.C.: AEI-Brookings Joint Center for Regulatory Studies. North, Douglas. 1990. Institutions, Institutional Change and Economic Performance. Cambridge: Cambridge University Press 1986. The New Institutional Economics. Journal of Institutional and Theoretical Economics 142 (1): 230-37.

PAGE 109

103 Onwumechili, Chuka. 2001. Dream or Reality? Providing Universal Access to Basic Telecommunications in Nigeria. Telecommunications Policy 25: 219-31. Peltzman, Sam. 1976. Toward a More General Theory of Regulation. Journal of Law and Economics 19: 211-240. Ros, Austin J. 1999. Does Ownership or Competition Matter? The Effects of Telecommunications Reform on Network Expansion and Efficiency. Journal of Regulatory Economics 15: 65-92. Rose, Nancy, L., and Paul Joskow. 1990. The Diffusion of New Technologies: Evidence from the Electric Utility Industry. RAND Journal of Economics 21(3): 354372. Sachs, Jeffery, and Andrew M. Warner. 1997. Sources of Slow Growth in African Economies. Journal of African Economies 6 (3): 335-76. Singh, J.P. 2000. The Institutional Environment and Effects of Telecommunication Privatization and Market Liberalization in Asia. Telecommunications Policy 24: 885-906. Stigler, George. 1971. The Theory of Economic Regulation. Bell Journal of Economics and Management Science 2: 3-21. Studemund, A. H. 1992. Using Econometrics. A Practical Guide. New York: Harper Codings. Svensson, Jakob. 1998. Investment, Property Rights and Political Instability: Theory and Evidence. European Economic Review 42: 131 7-4 1 . Taylor, Lester D. 1980. Telecommunications Demand: A survey and Critique. Cambridge, Massachusetts: Ballinger Publishing Company. The Book of the States. Lexington, Kentucky: Council of State Governments, 1984-1997 United States Department of Commerce. Statistical Abstract of the United States. Washington, D.C.: Bureau of the Census, 1984-1988, 1990, 1992, 1994, 1996 Wallsten, Scott. 2001. Competition, Privatization and Regulation in Telecommunications Markets in Developing Countries: An Econometric Analysis of Reforms in Africa and Latin America. Journal of Industrial Economics 49(1): 1-19. Weinhaus, Carol, Tom Wilson, Gordon Calaway, Robert Kwiatkowski, Mark Lemler, Dan Harris, Eugene Goldrick, Pat McLamey, Sally Simmons, Pamela Russell. 2000. Closing the Gap: Universal Service for Low-Income Household. Boston, Massachusetts: Telecommunications Industries Analysis Project. World Development Indicators on CD-ROM. 1999. Washington, D.C: World Bank Group. World Development Indicators on CD-ROM. 2000. Washington, D.C: World Bank Group.

PAGE 110

BIOGRAPHICAL SKETCH Jacqueline Marie Hamilton was bom in St. Ann, Jamaica, where she spent her childhood. Later she attended the University of the West Indies and received a Bachelor of Science degree in management studies and a Master of Science in economics. After graduating in 1 996, she worked with the Central Bank of Jamaica. A year later she joined the Ph.D. economics program at the University of Florida. There she had the opportunity to gain some teaching experience. She also worked with the Public Utility Research Center at the University. 104

PAGE 111

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Sanford Berg, Chairman Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Associate Professor of Economics Chunrong Ai I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. David Figlio Associate Professor of Economics I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. This dissertation was submitted to the Graduate Faculty of the Department of Economics in the College of Business Administration and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. May 2002 Dean, Graduate School