THE ECONOMIC RETURNS TO INSTITi'TIONAL
INNOVATIONS IN NATIONAL AGRICULTURAL RESEARCH:
ON-FARM RESEARCH IN IDIAP PANAMA
Juan Carlos Martfnez*
CIMMYT Economics Program Working Paper 04/83
* Regional Economist, CIMMYT Regional Economics Program for Central America and the Caribbean. The
activities of the Regional Program are supported by a grant from the Swiss government.
** Agricultural Economist, CIMMYT.
To the Caisan team, for its
personal and professional
commitment to the Panamanian
This work is not a result of our isolated intellectual efforts; it is
rather the outccme of our intense interaction with both, our colleagues
in CIMMYT and IDIAP, and our field experiences in on-farm research
The work has benefited from conrments made on an earlier draft. Our
thanks for that to Donald Winkelmann, Derek Byerlee, Edgardo Moscardi,
David Rohrbach, Larry Harrington, Mike Collinson, Grant Scobie, Lucio
Reca and Robert Tripp. We assume full responsibility for any remaining
limitation of the study.
Last, but not least, we want to extend our recognition to Jose Roman
Arauz and his team for their direct involvement in the field work; to
Pedro Santamaria for his assistance crunching numbers in the computer,
and to Maria Luisa Rodriguez for her patience and craftmanship in typing
numerous drafts in the iterative process leading to this final document.
PRE FA CE
In cooperation with researchers in national agricultural research
programs, CIMMYT has sought to develop procedures which help to focus
agricultural research squarely on the needs of farmers. The process
involves collaboration among biological and social scientists (for the
most part economists) in identifying groups of farmers for whom
technologies are to be developed, defining farmer circumstances and
problems, screening this information for appropriate research
opportunities, and then implementing the resulting research program on
experiment stations and in the fields of representative farmers. The
process emphasizes fixing priorities in research and identifying
solutions which are appropriate to farmer's circumstances.
The Instituto de Investigaciones Agropecuarias de Panama (IDIAP) was
created in 1975 with the basic goal of reaching Panamanian farmers with
technologies appropriate to their specific agroeconomic circumstances.
Given this goal, IDIAP agreed with CIMMYT on a cooperative effort in an
area-specific, on-farm research program. It was expected that work would
generate useful technology and, as well, would serve as a source of
methodological and organizational experience in this type of research.
The initial program was designed for a maize producing area under the
leadership of IDIAP and with technical support from CIMMYT. CIMMYT's
contribution drew on the experiences of other countries where national
program and CIMMYT staff were jointly engaged in farm-level research.
An earlier report, Institutional Innovations in National
Agricultural Research: On-farm Research within IDIAP, Panama, Martinez,
Juan Carlos and Jos6 RomAn Arauz, describes collaborative on-farm
research as a needed complement to experiment station research. It
concentrates on research in Caisan, a maize producing area, providing
details about the surveys and experimentation which led to the
introduction of an improved maize technology.
The study by Martinez and Sain presented here examines the benefits
and costs of on-farm research. In this paper, Martinez and Sain recognize
that the farming innovations under consideration could have come through
station research or through farmer experiments as well as through the
on-farm research process described. Each implies a different path through
time for development and diffusion of the innovation and for the rate at
which increases in productivity are achieved. These differences are a
focal point of the analysis. It was not possible to measure the method's
contribution to the relevancy of station research, another source of
benefits from on-farm work.
Some have been apprehensive about the cost efficiency of such
research. Martinez and Sain show a high rate of returns to this
investment in on-farm research, largely because the resulting technology
was taken up so rapidly by area farmers. The rapid diffusion, in turn,
was the result of a research process well tuned to their needs and
The Martinez and Sain report on returns to investments in on-farm
research is among the first of its kind. We hope that similar reports
will soon be available from other countries so as to add precision to
our understanding of the potential from such research.
Donald Winkelmann, Director
I. INTRODUCTION...................................................... 1
II. CAISAN: PRODUCTS AND INPUTS..................................... 3
III. METHODOLOGY................................................. 6
1. Objectives of the Evaluation and the Institutional Point of
View............................................ .... 6
2. Identification and Characterization of the Products to be
Evaluated ................................................... 8
3. Estimation of the Annual Flow of Research Benefits........... 10
3.1 Direct Impacts............................................. 10
3.2 Distributional Impacts...................................... 15
3.3 Indirect Impacts............................................ 15
3.4 The Regional Social Price of Maize................... ....... 16
4. Estimation of the Annual Flow of Research Costs............. 18
5. The Benefit/Cost Ratio and the Rate of Return................. 22
6. Estimation of Adoption Patterns............................. 23
7. The Economic Returns to the Methodological Innovation........ 26
IV. RESULTS......................................................... 32
1. Adoption Patterns........................................ 32
2. Estimation of the Annual Flow of Net Benefits ................ 36
2.1 Estimation of the Technology Induced Yield Increase......... 36
2.2 Net Benefits per Unit of Land............................... 39
2.3 The Annual Flow of Net Benefits............................ 39
3. The Annual Flow of Research Costs.......................... 40
4. Benefit Cost Ratio and the Rate of Return................... 40
V. CONCLUSIONS................................................ .... 42
APPENDIX............................................... ............. 48
BIBLIOGRAPHY .................................................... 51
THE ECONaOMIC RETURNS TO INSTITUTIONAL INNOVATIONS IN NATIONAL
AGRICULTURAL RESEARCH: ON FARM RESEARCH IN IDIAP PANAMA
Juan Carlos Martinez
It has been common practice in modeling technological change to
consider institutional development as an exogenous variable. Furthermore,
these models have usually hypothesized that causality runs from
technological to institutional change .(Ruttan, 1978). The fact that,
despite widespread promotion, many new technologies are not used by
farmers, has lead to the development of alternative models considering
institutional change as endogenous in the model with causality running in
both directions (De Janvry,-1978, Ruttan, 1978).
This study deals with the economic evaluation of certain innovative
methodologies for agricultural research, aimed at developing appropriate
technologies for target farmers in the near-term. These methodologies
were initially implemented on a trial basis by a recently, created
national research institute in Panama, and later institutionalized within
the research organization on the grounds that they significantly
contributed to the increased efficiency of public investment in
agricultural research. In the past, agricultural research in Panama was
conducted by different organizations including the Agricultural
Development Ministry (MIDA), the University of Panama, and various public
and private enterprises. In general, it was carried in agricultural
experimental -stations often under conditions quite different from those
faced by farmers.
At the same time, there was a general consensus among policy makers
that the existing research structure was generating an insufficient
I_/ Follaoing Ruttan "an institutional change occurs when there is a
change in (1) the behavior of a particular organization, (2) the
relationship- between such an organization and its environment, or
(3) in the rules that govern behavior and relationships in an
organization's environment" (1978, p. 329). In this definition the
term organization means a decision unit which exercises control of
amount of appropriate technology to impact effectively the technological
structure of the agricultural sector. This deficiency provoked interest.
in revising traditional research strategies. As a. consequence, the
Panamanian Institute of Agricultural Research (IDIAP) was created in 1975
with the main objective of increasing farmers' productivity and income
levels with special emphasis on farms of medium and small size.
A guideline of the institution was that of focusing research on
specific regions and crops for the development of technologies
appropriate to representative farmers in areas defined as high national
priorities. Research could thus be concentrated on the most important
farmer problems and the scarce resources of IDIAP used to best advantage.
Its activities were planned in a sequential pattern to permit
methodological adjustments as experience was gained and to provide a
framework for the training of a corp of national on-farm researchers.
In 1978, the first such program began in the area of Caisan with the
cooperation of CIVMYT and with a former CIMeYT. trainee assigned as
coordinator of the program. At the same time, the issues which would
shape IDIAP's institutional organization were being discussed and Caisan,
its first area-specific on-farm research program, was expected to be a
source of experience for the development of research procedures for
The Caisan program was planned and carried out strictly within the
limits of the human and financial resources normally available to IDIAP.
Thus, the cooperation of CIMMYT (development of procedures and in-service
training) was designed in such a way as to not exceed normal resource
allocation for area-specific programs.
The area of Caisan is located in' the northwest side of Panama,
involving 10,000 ha of territory and about 300 families. The most
important production system is a maize/bean rotation, which led to the
inclusion of both maize and beans as target crops for the research
program. A complete report of maize results, including surveys and
experiments carried between 1978 and 1982 can be found in Martinez and
Arauz (1983), and will not be covered in detail in this paper.
The main objective of the Caisan Project was to increase in the
near-term productivity and income of representative area farmers. Also,
Caisan represented a first step in a process which built up from on-farm
research actions towards an articulated on-farm research program. In this
framework, the methodological implications of the Caisan experience and
their spillover effect in terms of the institutionalization of on-farm
research within IDIAP was an important "output" expected from the
Project. Accordingly, the progress of the work was closely followed and
intensively discussed by IDIAP researchers and directing staff in
national meetings, field days and regional workshops. As a result of
these follow-up activities, and given the increasing emphasis in area
specific on-farm research in IDIAP, the institution decided to conduct an
evaluation of the cost efficiency of the on-farm research procedures used
in Caisan. This evaluation was expected to quantify the social rate of
return of IDIAP investment required to implement these on-farm research
procedures directed at one of the target crops of Caisan. With this basic
goal in mind, an evaluation study was designed and conducted for maize
within Caisan program by 1982. This paper reports on the qualitative
and quantitative results of this evaluation.
II. CAISAN: PRODUCTS AND INPUTS
The flow of basic inputs and outputs of the Caisan Project is
described in Figure 1. Two basic types of "outputs" resulted from the
implementation of Caisan. The first one is associated with the contribution
of Caisan to the institutionalization of on-farm research within IDIAP.
This contribution has been obtained through learning-by-doing, in-service
training and workshops, all based on Caisan experience. The resulting
_2/ On-farm research methodologies used in Caisan are not crop specific
and were equally applied to both target crops (maize and beans).
Although results on beans were less documented and sistematized at
the time of the evaluation (mid 1982), the impact of the program in
beans production has been at least equivalent, if not superior to
the one obtained in maize, being research costs at similar level for
FIGURE 1. THE CAISAN ON FARM RESEARCH PROGRAM. FLCW CHART OF BASIC INPUTS AND
Resource Allocation for
Area Specific Research
On Farm Research
Methodological Experience in OFR
Training; Seminars; Workshops
- Spillover Effect
- Institutionalization of
Direct and Indirect Impacts
- Increase in Yields
- Input Saving Effects
increase in IDIAP's capabilities to implement on-farm research procedures
in other areas of the country can be considered as an addition to the
national stock of knowledge and, hence, as an important positive
externality produced by the Project.
The other product is that related to the main objective of
increasing, through technology generation, productivity and income of
representative area farmers. In this sense, the products are the
technological alternatives generated by the program and recommended to
farmers. Benefits of these products are valuated through impacts
associated with actual farmer adoption of these alternatives. In other
words, the farmer is placed as the final judge of this 'process of
technology generation transfer, and accordingly, adoption is taken as
necessary condition for associating positive benefits with the
technological alternatives involved. At the time of the evaluation four
technological alternatives had been generated by the Program. Two of
them, chemical weed control and spatial arrangement and density, were of
yield increasing nature requiring some additional resources (costs) per
hectare for its adoption. The other two, zero-minimum tillage and no use
of fertilizers were basically input (costs) saving per hectare without
On the input side of the Project, two types of inputs can be
identified. First, CIMMYT contribution, which has been entirely composed
of procedures for on-farm research and training in the use of these
procedures. Both elements, research procedures and training have been
recently of increasing importance in the CINMYT cooperative work with
national research programs. The second type of inputs corresponds to
resources reassigned by IDIAP from the experimental station research to
on-farm research in Caisan. / These include human resources, fixed
assets (like vehicles and a mini tractor), the rental of a house in the
SThis cooperative work takes place through CIIYT Regional Programs.
See Juan Carlos Martinez (1982).
SOn-farm research is understood here as a liasson and needed
complement between the nore traditional station research on one
hand and the extension activities on the other.
area, and the different inpuEs (fertilizer, herbicides, insecticides)
required for carrying the on-farm experiments.
Figure 2 shows the sequence of decisions followed in the evaluation
of the Caisan Project. In the rest of this section the main
methodological problems associated with each stage are briefly described
and the particular decision taken regarding the case at hand is
1. Objectives of the Evaluation and the Institutional Point of View
Given the nature of the Project, its evaluation could be carried out-
assuming either a national or international perspective. Whether to
follow one or the other will basically depend on the objectives of the
The objective of the evaluation was the measurement of the cost
efficiency of the methodology applied in Caisan. Specifically the
national research institution, IDIAP, was interested in knowing the rate
of return of the investment required to iirplement area specific on-farm
research programs like Caisan as a needed complement to station research.
In this sense the objective of the evaluation can be stated as the
estimation of the rate of return to the Panamanian society of the
resources invested in implementing the OFR methodology used in Caisan.
Assuming an international perspective will imply that benefit and
costs will be traced beyond Panamanian borders. In this particular case,
on the benefit side spillover effects and increased methodological
experience accruing to other countries via CIMMYT international programs
should be considered as a positive externality. On the cost side, CIMMYT
resources allocated to this type of activities should also be considered,
spreading their cost among the various international programs of similar
nature implemented with CIMMYT cooperation. In other words, assuming an
international perspective would inply the evaluation, starting from
FIGURE 2.. STAGES OF THE EVACUATION PROCESS.
Definition of the
Definitions of the
Identification of Inputs
and Outputs to be
Estimation of the
Annual Flow of
Estimation of the
Annual Flow of
Panama, of the international system operating via CIMMYT in the area of
on-farm research. This does not appear consistent with the objectives of
the evaluation, and would be of little utility for IDIAP decision makers.
Assuming a national perspective has various implications. Spillover
effects would be confined within Panama, while, on the cost side, CIlMYT
could be visualized as part of the international environment faced by the
country, from which IDIAP, could occasionally, take elements without any
cost. This was the case of the contribution of CIMMYT to the program, in
terms of OFR rtthodologies and training, which accrued to IDIAP without
cost for the country. In the words of Kislev and Hoffman evaluating wheat
improvement in Israel: "From the point of view of the Israeli decision
maker who has to decide on the allocation of funds to research in the
country, free knowledge is part of the environment in which the local
research system is operating. The cost of producing this knowledge abroad
should not have any effect on his decision." (1978, p.173). In this case
the concept is extended to include not only knowledge captured and used
directly by Panama, but as well to include the free training and
methodological and professional services which complemented Panamanian
Since this appears to be more consistent with the specific
objectives of the evaluation, this was performed with a strict national
point of view. Accordingly only resources allocated to the Program by
IDIAP were considered.
2. Identification and Characterization of the Products To Be Evaluated
Agricultural research may be thought as a production process in
which inputs such as previous knowledge, research scientists,
laboratories, and research materials to produce certain products.
Definition and characterization of the outputs presents conceptual
problems which are, in general, dependant on the correct statement of the
objectives and institutional point of view of the evaluation (Scobie,
1979, Schuh and Tollini, 1979). The decision to perform the evaluation
front an strict national point of view, clarifies sane of the issues
outlined by Scobie and Schuch and Tollini but not all of them. As a
result of the research process two types of outputs were identified (see
previous section): 1) An addition to the national stock of knowledge
through increased capabilities and gain in experiences within IDIAP, and
2) an increase in productivity and income of area farmers through
As recognized in the literature, assigning a value to the first type
of product would imply an almost unafordable methodological burden in the
evaluation. This is so given that the product defined as a gain in the
stock of knowledge is intangible, and also due to the fact that there is
neither a market nor other feasible mechanism measuring how much society
values this kind of product. By considering only the product related to
technology generation it is possible to have an indication of the social
returns due to the methodological innovation applied in Caisan. This is
feasible if the evaluation is performed taking as a starting point the
prevailing technological situation (fanrer practices) which in turn can
be associated with the more traditional station research. -That is,
incremental Benefit-Cost arialysis applied to a case study (Caisan) would
serve as a proxi of the social gains of applying OFR methodologies. As
every other case of study generalization should be made with care.
In summary, due to methodological difficulties only technological
innovations generated by the program would be entered as products in the
evaluation. By comparing the benefits generated by this product with that
which would have resulted without the program, a reasonable measure of
the rate of return accrued from the OFR strategy followed in Caisan can
be achieved. In turn, this rate would be a proxi, assuming replicability
of the project by the national institution, of the returns to the second
type of product: the increased capability within IDIAP to conduct similar
area specific OFR programs in other regions of Panama.
5/ Traditionally the area was covered by the extension service from
3. Estimation of the Annual Flow of Research Benefits
3.1 Direct Impacts
Two main ex-post procedures have been used to evaluate the benefits
of agricultural research: the economic surplus (or index number) approach
and the production function approach. The choice between both methods
rests basically on the quantity and quality of the available data and on
the nature of the case at hand. -
Following Norton and Davis (1981), the production function approach
consists essentially in introducing into the production function a
variable as proxi for research and extension. Among the studies using
this approach at an aggregate level are those of Griliches (1964),
Evenson (1967, 1968), Cline (1975), Davis (1976), Kahlon (1977), and Lu,
Quance and Li (1978). While Peterson (1960, 1967) and Berdhal (1975) used
the approach at the individual product level.
The economic surplus approach, much less data demanding that the
previous one, has been widely used to estimate returns to agricultural
research in both developed and developing countries. The approach is
based on the concept of consumer and producer surplus. An excellent
discussion of both concepts can be found in Currie, Murphy and Schmitz
(1971), Hertford and Schmitz (1977), and Mishan (1968). A discussion
about the surpluses as measures of welfare changes can be found in
Boadway (1974). The work of Schultz (1953) and Griliches (1958) are
considered as pioneers in using the approach.
Griliches' analysis can be regarded as a special case of a more
general scheme presented by Peterson (1967). Figure 3 shows this general
There is also a differentiation in terms of marginal and average
rates. In the economic surplus approach what is estimated is an
average rate of return to the investment in agricultural research,
the production function approach will provide a marginal rate.
FIGURE 3. CHANGES IN CONSUMER AND PRODUCER SURPLUS
0 Q2 Q1
The displacement of the supply curve fran S1 to S2 due to the
vanishment of the innovation would reduce, social benefits (consumer and
producer surplus) by an amount equal to the area OAB which is regarded as
the net social benefits due to the innovation. The area OAB is in fact
the algebraic sum of the area P1PZAB, measuring the change in consumer
surplus and the area [(OBP1)-( (DA2) measuring the change in producer
Among other studies which have used the approach are those of
Schmitz and Seckler (1970); Hertford and Schmitz (1977); and Kislev
and Hoffman (1978). In a recent article, Wise (1981) shows that the
surplus method can be derived as a specific case of a more general
SThe authors showed that linear supply and demand functions provide
enough accuracy in the calculations without the complications due to
the non-linear specifications. In their words: "However differences
in the estimates of benefits provided by the more complicated
formulations and those presented here are small for usual values of
the key parameters. The main reason is that in all formulations the
critical determinant of the value of the benefits derived from
research is simply K.P.Q. or the percentage change in the value of
production attributable to research."
approach measuring benefits. The author also showed the equivalence of
both methods: when distributional aspects are introduced, social net
benefits can be obtained either as a sum of benefits or as a sum of
In order to measure the direct impacts generated by the adoption of
the new alternatives, the Wise benefit approach was chosen as the more
appropriate. Although the Wise benefit approach and the surplus method
have been shown to be equivalent the former was preferred because it
adjusts more naturally to the division of the technological alternatives
generated by the Program into two groups according to the type of impacts
resulting from its adoption.
Before proceeding to the application of the benefit approach, the
structure of regional supply and demand curves must be specified. Because
maize acreage affected by the program represents a small percentage of
total national maize acreage, it was considered that the project would
have a negligible impact on the national maize market, hence, its
evaluation may proceed at fixed prices. This assumes that the region
faces a perfectly elastic demand curve. Furthermore, the existence of a
minimum support price fixed at the national level provides further
motivation for considering the region as a "price taker", i.e. an
increase in regional production due to the Program will not affect the
maize price level.
In specifying the supply curve elasticity it should be noted that,
because of the nature of the on-farm research methodology, the short-run
supply curve should be used in evaluation of the impacts of improved
technological components. Although there are no available elasticity
estimates, same empirical evidence, (such as the fact that land and farm
labor resources remain fully employed despite short term price
variations) supports the hypothesis that the short run supply curve
is highly inelastic. Furthermore, the fact that maize is a component of
an important regional maize/bean crop rotation, the later being an
important cash crop in the region, would further sustain this hypothesis.
A highly inelastic short run maize supply curve would mean that in the
short run there is not a significant number of farmers leaving or
entering the maize production sector. In other words, maize acreage in
the region can be considered as fixed.
Surmarizing, the postulated regional Supply-Demand structure is one
of a perfectly elastic demand curve facing a perfectly inelastic short
run supply curve. Within the previous framework, four technological
alternatives generated by the Program will be considerated in terms of
their direct impact associated with impact farmer adoption. Table 1
describes them in contrast with farmers' practices prevailing in the
Recommendation Domain at the initial stage of the Program.
TECHNOLOGICAL COMPONENTS: FARMER PRACTICES AND
GENERATED BY THE PROGRAM
Technological Farner Practice Technological Alternatives Main Direct Impact
Caoponent Through Adoption (Per
Chemical Weed Application of 1 It/ha Application of 1-2 Its/ha
Control of 2,4-D, 30 Days After of Paraquat, 20-30 Days
Planting After Planting
or FIRST GROUP
Application of 1-2 kgs/ha
of Atrazine, 0-10 Days YIELD INCREASING
Spacing Arrangement Irregular Spacing 50,000 seeds per ha
and Density 40,000 Seeds per ha planted in Rows
Zero Minimum Plowing and Harrowing: Manual Chopping of Weeds
Tillage 3 Passes Followed by 1-2 It/ha of SECOND GROUP
Chemical Fertilizers Application of No Fertilizer Use
200 lbs of 10-30-10
The four alternatives have been classified in two groups according
to the nature of their impact in the production process. The first group
includes appropriate chemical weed control and a planting arrangement in
rows with increased density, implying a yield increasing effect and a net
addition of resources in order to achieve the higher yields. The second
group, includes zero-minimum tillage and no use of fertilizers which are
associated with an input saving effect without affecting yields. This
categorization allaos separate treatments for each group in measuring the
benefits associated with the alternatives involved. In the case of the
alternatives within the first group total benefits for each alternative
B = B -2 being
B1 = ARHII HAS Pm
B2 = AC H-HAS B1 B2, AR, AC> 0
BI, is the social value of additional maize production generated with
the diffusion of the new technological alternative involved.
AR, is the yield increase induced by the adoption of the alternative.
H, is the net proportion of maize acreage cultivated with the new
HAS, is the total maize acreage within the recommendation domain, which
is assumed to be fixed and estimated in around 1000 hectares.
B2, is the social value of the additional resources necessary to achieve
the increase in production, (AR-.H .AS),
AC, is the net addition in variable cost per unit of land -necessary to
achieve the increase in yields. Recombining previous equation B
could be written as:
B = H.HAS [ R.Pm AC]
That is, total benefit generated with each technological alternative
will be equal to the total maize acreage cultivated with the alternative
times the net benefit per hectare associated with its use.
For the particular case of alternatives in the second group, they do
not have any significative yield effect hence AR-- 0 and consequently
B1 = 0. Because they have an input saving effect, decreasing cost per
unit of land, AC
of these alternatives will be B = (-B2) = B
- For more details see Arauz, Martinez (1983), Chapters 5 and 6.
3.2 Distributional Impact
The distributional aspects of the Program can be assessed by
considering the four social groups identified by Wise (1982). Given
the nature of the Program these patterns of distributional impact appear
to be very simple. Consumer welfare does not change (no production
augmenting effect for second group of technologies, no price effect for
either group). While benefits of technological development will fully
accrue to innovators (adopters) in terms of producer surplus; those who
do not adopt the technology remain the same (relevant relative prices are
unaffected by the Program). Finally since there is no technological
treadmill effect 10 the number of farmers not adopting the technology
and leaving the sector will be negligible.
3.3 Indirect Impacts
The most important secondary impact of the Project is the effect of
zero (minimum) tillage on future soil erosion levels and consequently on
the natural fertility of the soil. This impact would be larger for
farmers whose maize plots are located on slopes of considerable
Little information is available about potential yield decrease due
to the reduction of soil fertility associated with erosion so
quantification of this impact is difficult. In this work no attempt is
made to measure it. Consequently, benefits are in this respect,
/ Consumers, producers who do not adopt the technology and leave the
sector, producers who do not adopt the technology and stay in the
sector, and producers who adopt the technology.
10/ Cochrane W. (1958) "Farm Prices: Myth and Reality" University of
11/ Following tradition in this type of evaluation when an option is
encountered the choice is made such that the associated benefits are
3.4 The Regional Social Price of Maize
Public policy in Panama with respect to maize production has been
directed in the last decade toward the goal of self sufficiency. The main
policy instrument used in the attempt to achieve this goal has been the
implementation of support prices for this product fixed above the
international price. As a consequence domestic production increased from
an annual average of 53,800 metric tons during the period 1971-75 to an
annual average of 68,600 metric tons during that of 1976-80. Concurrently
with this change, imports of maize decreased from an annual average of
19,500 metric tons to 11,400 metric tons during the same periods.
Although there is no direct empirical evidence about the annual
balance of the province of Chiriqui, indirect evidence support the
hypothesis that Chiriqui is a net exporter of the product The
situation for Panama and for the region is then depicted in Figures 4 and
In a closed economy and in absence of regulations Pe and Qe in
figure 4 would be the equilibril.m price and quantity for maize at the
national level. But Panama confronts a perfectly elastic international
supply curve given by P.O. where P. is the international price of maize.
'Then, in absence of regulations, the relevant supply curve becomes OAO.
with OQ2 domestically produced, OQ1 domestically consumed and (Q1-Q2)
imported. Once a support price P is established, domestic consumption
falls to OQ3, domestic production increases to OQ4 and imports decrease
to Q4Q3. The area ADE represents the social loss due to the excess cost
of domestic production of the amount Q4Q2, and the area CBF is the loss
12/ Maize production in Chiriqui is reportedly used as follows: 25.4% is
consumed in the farm, 4% is used as seed, 37.1% is used as feed and
the remaining 38% is sold out the farm. Given an annual average
total production of 11,661 m.t., it results that 4,431 m.t. are
annually sold by farms. Chiriqui urban population was, in 1980, of
91,017, considering an estimated consumption of 18.82 kg/person,
total urban consumption would be of 1,713 m.t. This would leave a
net positive balance of 2,178 m.t. Although there are some small
mills in the province, the bulk of Panama's mills are located out of
the province, hence the excess production over internal consumption
is likely sold out the region for consumption and/or proce rijig.
in consumer's surplus due to the reduction in consumption of the amount
National Level Regional Level
In figure 5, Or and Dr are the regional supply and demand curves,
respectively. Ps is the support price in Panama City and P = (Ps-TC) is
S S3 S
the support price net of transportation costs from the region to Panama
City. In this case, OQ1 is the amount produced by the region, QQ2 is the
amount consumed within the region and Q2Q1 is the amount the region
Given this situation and assuming that the region surplus production
is entirely exported to and consumed in Panama City the social price
of extra units of maize produced by the program will be given by the
import price of maize (CIF, Panama City) net of transportation costs from
1/ Panama City is the most distant point among those of potential
destination for the maize of the region. Hence this assumption
implies (via transport costs reduction) the lowest alternative with
respect to pricing the maize.
Caisan to Panama, i.e. p1 = P.-TC. This is so since in absence of
distortion in the foreign exchange market this price would reflect the
social opportunity cost of extra units of maize without accounting for
distortions in the internal market.
Introduction of a support price above the border price increases the
social opportunity cost of extra units of maize, hence the society
willingness to pay for additional units of maize from Caisan will also
increase. In this case the marginal value of extra units of maize from
Caisan will be given by the support price net of transportation costs,
i.e. p2 = p TC.
Fran a perspective of general equilibrium and welfare considerations,
maize pricing should be done under the assumption of non existence of
regulations and hence Pm should be used. However, social preferences in
terms of food security (through import substitution) are well established
and politically legitimized as a sustained policy goal which cannot be
ignored neither by IDIAP nor by any institution like IDIAP.
In other words, what could be considered with a global perspective
as a variable (i.e. agricultural policy) should be and has been
considered by IDIAP as a parameter, setting the framework within which
the institution could display its research policy and institutional
strategies. Then when considering haow efficiently IDIAP has been doing
this in the particular case of Caisan the maize price which should be
used is the one provided by P as a reflexion of the agricultural policy
framework faced by IDIAP. Accordingly, the resulting rate of return will
best reflect, in our view, the efficiency of OFR methodologies
implemented by IDIAP in Caisan. In any case, the decision was taken to
conduct the evaluation under both maize pricing options.
4. Estimation of the Annual Flow of Research Costs
Definition and characterization of inputs to the research process
presents similar problems to those described in the case of the products
(Scobie, 1979, Schuh and Tollini, 1979). Among the most comnon problems
are those arising when a particular resource is used to produce multiple
or joint products, the case of serendipity in the process, and the
correct specification of the knowledge stock participation in the
research process. Given the decision to perform the evaluation from a
strict national point of view, resources contributed by CIMtYT are
considered as free and do not enter the cost side of the evaluation. With
respect to resources contributed by IDIAP, Table 2 shaos their
composition in terms of basic items. The human resources figures
indicate the proportion of man-years used by the program for each of both
labor categories, for example during 1979 the program used. 8 man-years of
skilled labor and 1.45 man-years of semi-skilled labor. In the case of
fixed assets the figures reflect the proportion of the total annual
services provided by the asset which is assigned to the program i.e. in
1979, 50 percent of the annual services provided by the rented house is
assigned to the Project. Similarly 80 percent of the annual services
(hours per annum) provided by a pick-up truck is used by the Project.
Finally, as research materials are project specific, the table listed the
amounts of each one of the used in the program. -
The economic evaluation of the program requires first that the
project be evaluated against the without alternative and that resources
used be priced at their respective opportunity cost interpreted as the
value of the output the resource would produce in the activity from which
the resource is withdrawn.
Application of the with-without criterion requires the
identification, as component of the cost side of the evaluation, of only
/ While identifying research costs an interesting problem came out in
the discussion. Part of the cost of on-farm experimentation are
shared by the farmers (i.e. land, farmer time). Since OFR fits in
the production process, that is, it takes place while the farmer
production process takes place, these costs will be usually more
than outweighed by the benefits resulting from leaving the products
of the experimental plot with the cooperator farmer. While this
could still leave some doubts in terms of how it should be tested in
an evaluation of OFR, it has, from the perspective of research
strategy and management, a positive side associated with the fact
that a percentage, even though small, of research costs is paid by
those who will be the final recipients of research results.
additional resources necessary to conduct -the program. In the case of
fertilizers and pesticides they are program specific resources, hence
they are entirely charged to the program (see table 2). With respect to
fixed assets, the program used existing equipment and in addition a house
was rented. In the case of the equipment it was considered that program
accomplishment would require an additional time effort because increasing
off station work. As no reliable estimate of this increase existed the
total time proportion assigned to the project was taken as a proxi for
each piece of equipment used by the project (see table 2).
TABLE 2. COMPOSITION OF RESOURCES CONTRIBUTED BY IDIAP TO THE PROGRAM
1978 1979 1980 1981
A. Hnman Resources
1. Skilled Labour .40 .8 1.20 1.20
Labour .20 1.45 1.45 1.45
B. Fixed Assets
(Proportion of Total
1. Rented House .50 .50 .50 .50
Pick-up .65 .80 .80 .80
Mini-tractor .50 .50 .50
C. Research Materials
1. Fertilizers a/
Formula 12qq 12qq 12qq
Urea 6qq 6qq 6qq
Superphosphate 10qq 10qq lOqq
2. Herbicides b
Paraquat 10 G 10 Gl
Atrazine 8 kg 8 kg 8 kg
2-4-D 1 G1 1 G1 1 Gl
Dinitroaniline 1 Lt
Carbofurane 45 kg 45 4kg 45 kg
a/ One quintal (qq) is equivalent to 45 kg.
The listed amounts refer to commercial products.
All personnel involved in the project was already working for IDIAP.
Application of incremental analysis requires again the identification of
additional tiLe effort attributable to the program as the component of
the program's labor cost. The same criterion adopted for the equipment
was used in this case, the total time proportion assigned to the project
by each rmerber was taken as a proxi for the additional time effort
required by the accomplishment of the Project over the traditional
All resources allocated by IDIAP to the program including labor,
fixed assets and fertilizers and pesticides are evaluated at 1981 market
prices. This implies that market prices for these resources reflect their
opportunity costs. A brief justification for this procedure follows for
each resource category.
i. Labor--There is a general agreement in the welfare economics
literature that skilled labor wages approximately reflect the opportunity
cost of this type of labor (Irvin 1978). Since the program only used
skilled or semiskilled labor, 1981 wages were used to estimate the labor
cost incurred by the Caisan Project.
ii. Fixed Assets--Lack of accurate data availability precludes the
estimation of the total costs of using the equipment employed by the
project. The only data available was that of gasoline expenses. As the
gasoline price in Panama reflects its importation costs it was considered
that this price properly represented the opportunity cost of using it.
Ignoring other user costs would result in underestimation of the true
costs of the program. This is partially corrected by increasing the total
annual estimated cost by 10%. In the case of the rented house the annual
rent was considered as reflecting its opportunity cost.
iii. Pesticides and Fertilizers--The use of market prices was
justified in this case on the grounds that there are no subsidies with
respect to these resources. As such, domestic prices have historically
folla~ed world price fluctuations.
5. The Benefit/Cost Ratio and the Rate of Return
In order to calculate the benefit cost ratio and the rate of return
of the project, the rmthodology employed by Griliches (1958) -also used
by Kislev and Hoffman (1977) among others- was adopted. The benefit/cost
ratio is estimated as:
(1) B/C= (PANB+AFNB-AHFC)
PANB, are the past annual net benefits compounded to the base year
AFNB, are the annual future net benefits discounted to the base year
AFC, are the annual future maintenance costs discounted to the base year,
and PARC are the past annual research costs compounded to the base year.
Griliches presents the rate of return, r as (1958, p.425):
(2) r = k x B/C
where k is the discount rate used to estimate the B/C ratio. The rate of
return r, can be interpreted as the discount rate at which the stock of
costs yields an annual flow exactly equal to the annual flow of net
benefits, alternatively r may be interpreted as the discount factor at
which the annual flow of net benefits should be discounted in order to
yield a stock exactly equal to that of costs. In other words, a rate of
return of say 1.90 (190 percent) would indicate that each dollar spent in
the project would generate a future annual flow of 1.90 dollars of net
The criterion to judge a given project according to the B/C ratio is
that for the project to be acceptable this ratio should be larger or
equal to one. Then this last equation says that:
r = k if B/C = 1
That is, i k represents the social rate of tie preference betwe
That is, if k represents the social rate of tir!e preference between
present and future consumption, then equation 2 states that the average
rate of return for each dollar invested in the project (r), will be
larger equal or smaller than the social rate of preferences if the
discounted benefits are larger equal or smaller than the discounted (at
the same rate) costs.
6. Estimation of Adoption Patterns
An important element for measuring the benefits of agricultural
research is the estimation of the percentage of farmers (acreage) who
have adopted the new technology. The nature of the OFR methodology used
in Caisan presumed that appropriate technologies would be available in
the near term. This, in turn, would fulfil a necessary condition for
farmers' acceptance and accordingly speed up adoption.
A formulation commonly used to represent the diffusion of new
innovations is the logistic growth function or learning curve. The graph
of the function and its generic functional form are shown in figure 6.
FIGURE 6. THE LOGISTIC CURVE
y -- ---- --
y = K [ + exp (A + Bx)J
-1/ This section draws heavily on Martinez (1973).
Adaptation of this function to a context of- technology adoption
process is straightforward. Let h (t) be the maize acreage proportion
cultivated with the i-th technological alternative in year t. Then the
logistic function representing the difussion time pattern is:
hi (t)=K [ 1+exp-(A.+Bit) 1
In this equation Ki represents a constant called by Griliches the
"ceiling" of the adjustment function. That is Ki is the maximum expected
percentage of adoption of the technology. Ai is a parameter positioning
the curve in the time scale, while B. shows the rate of growth or rate of
acceptance of the innovation.
Martinez conceptualizes the diffusion process along the logistic
curve as succesive short-run equilibrium points between the supply and
demand for the new technology. In this context the value of the .ceiling
would be interpreted as the long run equilibrium. Quoting Griliches:
"While shift on the supply side determines the origin of the development,
the rate of development is largely a demand or acceptance variable"
(Martinez, 1973, p.81).
The rate of acceptance B, therefore, can be interpreted as
summarizing the demand conditions for the technology. As such, the
estimated value of B could be used as an indicator of the farmers' degree
of acceptability of the technology.
If enough data is available the logistic parameters A and B can be
estimated, after the model has been previously linearized, by the least
square method. If enough observations are not available, Martinez (op.
cit. p. 92) presents a method to obtain a crude estimation of the
parameters. The method consists in the simultaneous resolution of the
following two equation system:
hi(tl)= Ki [1+ exp-(A + Bitl)]
h (t )= Ki [1+ exp-(A + B.it)!
i 2 i 2.
To solve this system in terms of A and B it is necessary to have
information about the maximum expected adoption proportion K, and of the
adoption proportions in two points in time h (tl) and h (t2).
This has been the procedure followed for estimating the patterns of
adoption for each technological alternative generated by the Program. The
point hi (tl) was based in the initial formal survey implemented at the
planning stage of Caisan for assessment of farmer circumstances. The
other point h (t2) was h (1982); estimated with an especially designed
adoption survey carried out in July 1982. The survey methodology used in
this case was similar to that used in the assessment of farmer's
circumstances (yet with different objectives). That is, an informal
survey (done in May 1982) was carried to design a well focused formal
questionnaire. This questionnaire was filled out in June 1982 for a
random sample of 45 farmers within the recommendation domain of the
Finally, the value of the adoption ceiling was estimated based on
the knowledge and experience of regional technicians with the
acceptability of the alternatives offered, their potential impact in net
income and the relative degree of difficulty in their management. For
SOnce the values of K, h(t ) and h(t2) are known, the system is
reduced to a two equation-two unknown system which admit a unique
solution. System (3) can be rewriten as:
A + B tl = C1
A + B t2 = C2
C1 = In [h(tl)/K-h(tl)]
C2 = In [h(t2)/K-h(t2)]
The values of A and B which solve the system are calculated as:
A = Clt2 C2tl and B = C2 Cl
t2 tt2 t
example, the change from irregular planting to planting in rows is seen
by farmers as more "complicated" than the change in the type of weed
control. Furthermore, zero and minimum tillage requires the replacement
of usually contracted mechanical tillage by hand chopping plus
application of herbicides, activities which are usually performed with
own farm labor force.
In order to identify farmers who adopted the new technology,
definition of a set of relevant adoption criteria discriminantt
variables) for each new technological alternative is necessary. Once
these criteria are defined farmers are classified as full adopters if the
technology they are actually using agrees with all these criteria.
Similarly, partial adopters are those farmers whose current practices
only match some but not all of these criteria.
Partial adoption may occur if because of lack of information and/or
financial or market restrictions the farmer adopts only part of the new
technology. This case should not be confused with that of farmers who
decide to "test" the new technology. In this case he usually adopts the
new technology in full but applies it to only a fraction of his parcel.
In this way the farmer is able to cormare for himself the "advantages" of
the new technology against those of the conventional one.
Table 3 describes the adoption criteria defined for each of the four
7. The Econcmic Returns to the Methodological Innovation
As it was stated earlier, one of the objectives of the evaluation is
the estimation of the returns accrued to the application of the on-farm
methodology into the research process. In order to do so it is necessary
that the program be evaluated against the alternative, i.e., against
traditional station research, hence an adoption pattern for the
traditional approach needs to be assured.
TABLE 3. DISCRIMINANT VARIABLES AND ACCEPTANCE CRITERIA FOR EACH
1. Chemical Weed Control
1. If the farmer uses
chemical weed control
2. Type of Product
3. Application Time
2. If the control is
Gesaprim or Gramoxone
3. i) Gesaprim: 0-15
days after planting
ii) Gramoxone: 0-35
days after planting
4. i) Gesaprim l-3kg/ha
ii) Gramoxone: 1-3
TA2: Spacing 1. Planting Arrangement 1. If planting is
arrangement and accomplished in rows
2. Density 2. 45,000-60,000
TA3: Zero 1. Tillage system 1. If the farmer does
Tillage not use a mechanical .
2. Application of 2. If the farmer
herbicides applies herbicides
prior to planting
TA4: 1. Application of 1. If the farmer does
Fertilization fertilizers not apply fertili-
Following tradition, a pessimistic assumption which can be made
about this pattern, is that station research would be able to identify
exactly the sane research opportunities than those identified by OFR.
This assumes that. the only advantage of applying the new research
methodology is accounted by an extension effect. E_ In other words, the
only difference between both research strategies is given by a time lag
in the process of discovery, and dissemination of the same
technological components. -
Once this assumption is adopted the benefits to be estimated are a
function of the differential rates of adoption under both alternatives.
Following Lu (1981) several alternatives may be considered: these are
illustrated in Figure 7.
In order to estimate the incremental benefits due to the application
of the on-farm methodology in Caisan the shadow area under both curves
needs to be estimated. The problem is that the adoption curve under the
station research strategy (TSR in the figures) is an hypothetical one and
cannot be estimated directly. Hence 'it was decided to estimate upper and
lower bounds of benefit-cost ratios and rates of returns to the Project
by assuming different cut-off points of the benefit flow from the
Project. Figure 8 illustrates the approximation method used for the case
of a change in the adoption lag.
17/ The process of technological innovation can be decompossed into two
highly interrelated and complementary effects: a research effect
dealing with the creation of new technology and an extension effect
dealing with the dissemination of this technology. (Lu, 1981).
18/ In case of complex' farming systems, ignoring bioeconomic
interactions may lead to situations in which promising research
opportunities remain uncovered. OFR methodology implies the ex-ante
assessment of most promising research opportunities minimizing
therefore the probability of missing an important one. Hence the
assumption of ignoring the research effect in OFR will certainly
underestimate its contribution within the research process.
FIGURE 7. ALTERNATIVE EFFECTS OF ON-FARM RESEARCH (OFR) OVER TRADITIONAL
STATION RESEARCH (TSR)
i. The only change is in the year of
introduction of the new technological
iii. The only change is in the
adoption ceilingK of the new
ii. The only change is in the rate
of adoption of the new
iv. The only change is in the adoption
lag of the new technological
Percent of adoption
OFR= Adoption pattern of new technological components under OFR.
TSR= Adoption pattern of new technological component under traditional
t(1(1)l(1) t t1 time t (1)=tl(1) t
FIGURE 8. ESTIMATION OF BENEFITS ATRIBUTABLE TO OFR OVER TSR
Percent of adoption
OFR -I SR
h(t) -- _- __-
0 t(1) t* T
Given the estimated OFR curve, and the hypothetical TSR curve then a
cut off point t should be chosen such that the true unknown area ABC
would be approximated by the area AD. In order to do not overestimate
benefits the area D should be smaller than B plus C. By changing the cut
off point t it is possible to simulate different effects and positions
of the TSR curve, leading to upper and lower bounds in the benefits side
and hence in the rate of return of the program. The four cut off points
simulated in the evaluation are illustrated in figure 9.
The most likely changes between both methodologies (considering only
extension effects) would be a combined change in the year of introduction
and in the adoption lag of the new technological components. Hence, the
cut off point of 1982 would provide a very unlikely lower bound of the
benefits and returns to the OFR methodology applied in Caisan, while the
case of t =1985 would provide a fairly pessimistic estimate of benefits
and returns to the program.
A more realistic estimate of the benefits and returns to the OFR
methodologies would be given by the cut off point of 1990, in which case
some research effects would be also incorporated. Finally, the case of
t = illustrates the upper bound in benefits and returns to the
FIGURE 9. ALTERNATIVE HYPOTHESES FOR ESTIMATING THE CONTRIBUTION OF
OFR METHODOLOGY IN CAISAN
Percent of adoption
Percent of adoption
3. t*= 1990
program. As a matter of fact, this case is equivalent to remove the
assumption that traditional station research would have been able to
identify and effectively work with the same research opportunities
identified by the -OFR program in Caisan.
It should be stressed here that the simulation of benefits and
returns performed in this work does not imply that there are not benefits
to the society after a certain date. Social benefits due to
technological innovation are, once adopted, permanent.
1. Adoption Patterns
The initial step in the implementation of the evaluation methodology
described in previous section was given by the adoption surveys conducted
in 1982. The application of the adoption criteria resulted 'in the
percentage of maize acreage cultivated with each alternative. The same
procedure was carried for farmer practices from the 1978 survey. Table 4
presents the resulting percentages, as well as the estimates of maximum
adoption points (K ). In turn, these values lead to the 'estimation of
logistic adoption functions whose parameters Ai and Bi are also included
in Table 4. The shape of each logistic functions as well as their
location in the time scale are shown in Figure 10.
TABLE 4. ESTIMATED ADOPTION PARAMTERS
Alternatives t1 h(tl) h (1982) K. A B.
control- 1979 .082 .609 .90 3.3 1.0
2.Seeding arrangement 1979 .207 .627 .80 1.8 1.8
3.1.Zero tillage 1980 0 .188 .50 24.8 2.4
3.2.Minimum tillaqe 1980 0 .042 .25 23.9 2.2
4.No use of fertilizers 1979 .388 .795 .90 1.0 .1.0
FIGURE 10OADOPTION PATTERNS FOR THE RECOMMENDED TECHNOLOGICAL ALTERNATIVES
APPROPRIATE WEED CONTROL
80 82 84 86 88 90
1980 82 84 86
h,(t):.9 1xp-(- .046+3.68.t
1979 80 82 84 86 88 90
The values obtained for h. (1982) indicate a high percentage of
adoption for the alternatives -after only few years of project operation.
This confirms the hypothesis that technologies where appropriate to
farmer circumstances prevailing in the recommendation domain. Also, the
speed at which adoption took place, reflected by the value of the
parameter Bi, is telling us not only that the alternatives generated
where agroeconomically viable for representative farmers, but also that
they likely represented a solution for an important production problem
faced by farrrers and correctly identified as a research opportunity at
the planning stage of the Program. 19 This, as it stands, is an
important result of the evaluation. It is still to be seen in the rest of
this section whether or not they have been reached efficiently.
In order to proceed towards the estimation of benefits we need to
consider the values of h. (t.). The starting point for the diffusion due
to the Project is placed in different years according to the time of
release of the alternative. -In this sense, it was considered that there
was not diffusion, hence no .adoption due to the Project, of any
technological alternative until 1980. We are then interested in
clarifying the meaning of the positive values for hi(t,) and what role
have they play in the quantification of benefits.
The relatively high values for not use of fertilizer (38.8%) and
planting arrangement and density (20.7%) could be understood by the fact
that in the first case, some farmers were not having access to credit and
consequently not using fertilizer in spite of prevailing recommendations,
while the second one was also related to credit programs but in a
different sense, it correspond to farmers which had accepted the fully
mechanized services of PRCOM4EA, included in the credit package, and
consequently had conventional mechanized land preparation and planting,
in which case obviously arrangement, and also density were decided by
PRCOMECA which had the same calibration of the' planters for all the
services provided. Note that as the Project evolved, conventional
219/ The research hypothesis formulated at the planning stage can be seen
in Martinez and Arauz (1983), Section 4.
mecanized tillage and planting decreased as a result of the release of
the zero-minimum tillage alternative.
The remaining positive value for hi(ti) is a small percentage (8%)
in chemical weed control, corresponding to a non-systematic trial and
error process followed by farmers searching for better chemical weed
control alternatives. This value reflects the magnitude of the weeds
problem identified as a promising research opportunity by the
Project. Finally, the values of h(t1) for zero and minimum tillage
were zero, as there were no farmers using at t1 these alternatives. In
any case, in order to obtain the net annual adoption level due to the
Project, whenever a positive value for hi (t) was found it was discounted
from the adoption level reached in each year.
In other words, this net proportion was estimated along the time
scale of the logistic adoption function as H.(tj) = hi(tj) h (tl).
Table 5 shows the annual values of H.(t.) for each alternative. Zero
values for 1979 reflect the reasonable assumption already established
that no adoption due to the Program took place during this year.
TABLE 5. NET PROPORTION OF MAIZE CULTIVATED WITH EACH TECHNOLOGICAL
ALT E RN
H2 (ti) H3. (tj)
A T I V E
20/ For ore details
For nore details
and Arauz (1983), Section 4.
The ceiling (maximum adoption) is approximately achieved for every
alternative around 1985. The ceiling is continued to perpetuity for all
but fertilization. In this case it is assured that due to natural process
fertilizer application would start to have significant responses starting
in 1990, hence adoption of this alternative was assumed to produce
benefits only until 1990. -
2. Estimation of the Annual Flow of Net Benefits
Net benefits per hectare are estimated for each alternative
subtracting from the change in gross benefits the change in those costs
that vary. ( AC) Gross benefits are, in turn, calculated as the product
of the increase in yields due to the alternative (AR.), times the social
price of maize, (Pm).
This section is organized as follows: yield increases are estimated
in part 2.1; part 2.2 deals with the. estimation of net benefits per
hectare, and the annual flore of net benefits for each technological
alternative and for the entire program are calculated in part 2.3.
2.1 Estimation of the Technology Induced Yield Increase"
Although Hertford (1977) stressed some years ago that the most
important component in measuring the benefits from agricultural research
is the production shift parameter K (the parameter measures the relative
change in production A Q/Q due to the adoption of the new technology),
Dalrynple (1981) pointed out the existence of an imbalance between the
relatively sophisticated methodology developed to measure the area under
the supply-demand curves and the poor data base upon which the method is
SThe Project currently includes fertilizer level trials in continuous
plots to analyze what will happen in the medium run to the natural
fertility of the land as result of adoption by farmers of improved
(more intensive) production practices in weed control and spacial
arrangement and density.
Traditionally two sources of data have been used to measure the
relative change in yields: data gathered at experimental level and data
from aggregate average farm yields.
While data from aggregate average farm yields has the advantage of
being readily accessible; in most developing countries it is a very
unreliable estimate of yields at farm level. On the other hand, two
reasons are comnrnly quoted for the existence of an upward bias in the
yield gathered at experimental station level when used as farm level
yields estimators. The first one given by the fact that the level of
non-experimental variables is kept at what is considered an "optimum"
one, and the second one given by the existence of marked differentials in
In this work, a third source of data is used to estimate the
relative increase in yields at farm level: data gathered in on-farm
trials. In this case yield data is free of the first source of bias
because the level of non-experimental variables are kept at farmer level.
Hence, yield increases reflect more accurately increases due to the
change in experimental variables and cannot be partially attributed to
the interaction with high intensity use of other inputs.. Furthermore
experimental yields were adjusted not only to account for agroclimatic
and pest related risk factors but also for managerial factors.
In this sense, data used in this work reflects in the most accurate
possible way yield gains at farmer level due to the adoption of the
technological innovations. It is believed that if this source of data is
available (results of OFR experimentation), it should be preferred over
alternatives when estimating the gain in yields due to technological
Estimation of the yield increase induced by the adoption of
technological alternatives belonging to the first group was carried out
using the experimental results from all the three cycles of the project.
The first two years of experimentation were considered as normal, while
during 1981 yields were adversely affected by an epidemic attack of
Helminthosporium sp. Since epidemic attack of HIelminthosporium sp. in the
region is estimated to have a very low frequency of approximately one in
twenty years, a weighted average of both periods using these frequencies
as weights was estimated.
Another adverse factor which should be taking into account is that
of lodging. Results from the 1978 survey indicate a high frequency of
occurrence with partial crop losses. As there is no information about the
magnitude of partial damage, this effect was incorporated into the yields
calculation by assuming a total loss impact and a smaller occurrence
frequency of once every ten years. Adjusted yields are shown in Table 6,
where Do and H represent the farmer's practice for seed density and
herbicides application respectively where D1 and H1 represent the
reconnended levels in each case.
TABLE 6. YIELDS ADJUSTED BY RISKS FACTORS
H H Average
D 2.901 3.893 3.40
D1 3.682 4.841 4.26
Average 3.29 4.37
Source: Caisan On-Farm Trials 1979/81.
These numbers indicate a yield increase due to chemical weed control
of 1.08 ton/ha, while seed arrangement and density raises yields by .86
ton/ha. 22/ These impacts should be weighted by a "yield adjustment
coefficient" which account for differences between farmers and
experimental managerial conditions. In the case of Caisan this
coefficient was estimated by Arauz and Martinez (1983) to be 10 percent,
hence the final increase in yields to be used in the evaluation are:
Although a positive interaction effect was identified in some of the
trials, this impact was ignored in the evaluation.
AR1 = .972 ton/ha
AR2 = .774 ton/ha
for alternative 1 and 2, respectively.
2.2 Net Benefits per Unit of Land
Calculation of the net benefits per unit of land generated by each
technological alternative is accomplished by the method of partial
budgeting (Perrin et al 1976). Tables 10, 11, 12, 13, and 14 in the
appendix 1, provide details of performing this operation, a sunmary of
the results under the two pricing alternative is given in Table 7.
TABLE 7. NET BENEFITS PER UNIT OF LAND (DOLIARS/HA, 1981)
TA 1: Chemical TA 2: Spacing TA 3.1: TA 3.2: TA 4:
Social weed control arrangement Zero Minimum Fertili-
Pricing and density tillage tillage zation
1. Without 141.14 104.66 18.75 12.75 39.74
2. With 191.29 145.22 18.75 12.75 39.74
Taken 1981 prices the estimated social prices were:
P = without regulation = $149.50/metric ton., and
P2= with regulation = $201.09/metric ton.
2.3 The Annual Flow of Net Benefits
Once net benefits per unit of land and the total net acreage
cultivated with each alternative are estimated, calculation of the annual
flow of net benefits proceeds by multiplying both numbers for each
alternative. The annual flow for the entire program is procured by adding
up the individual flows over the alternatives.
Tables 14-a, 14-b, in the appendix, present these flows for both
social pricing alternatives. The benefits flows shown in these tables
correspond to those generated under the upper bound assumption of
perpetual flows. Alternative ones are estimated by choosing appropriate
3. The Annual Flow of Research Costs
Annual research costs elapsed from mid 1978 (last 5 months) to 1982.
It was assumed that after 1982 there would be a perpetual flow of
maintenance costs in order to keep yield constant against an adverse
nature (Griliches, 1958). A pessimistic annual estimation of these costs
amounted to 40 percent of the 1982 cost level. Table 8 shows the
estimated annual flow of research costs of the program.
TABLE 8. ANNUAL FLOW OF RESEARCH COSTS (IN DOLLARS)
1978 1979 1980 1981 1982 After
Annual 3,000 14,300 17,800 17,800 17,800 7,200
4. Benefit-Cost Ratio and the. Rate of Return
To estimate the Benefit-Cost ratio (B/C) and the rate of return (r),
1982 was chosen as base year. A discount factor of 15 percent, the same
used by USAID in the evaluation of Panamanian agricultural projects of
similar length, was employed to carried out the flows to the base year.
The benefit--cost ratio and the rate of return were estimated for the
four assurmmed cut off points of the flow of net benefits under both
pricing alternatives. The results of the calculations are summarized in
TABLE 9. BENEFIT-OST RATIOS A RATES OF R N 1/
TABLE 9. BENEFIT-ODST PATIOS AND PATES OF RETURN -
Pricing options t =1982 t =1985 t =1990 t =00
Social prices 3.11 7.86 12.93 16.36
without regulation (47%) (118%) (194%) (245%)
Social prices 4.05 10.30 16.99 21.69
with regulation (61%) (155%) (255%) (325%)
SValues within parenthesis are rates of return, without
parenthesis benefit-cost ratios.
In all cases, the rates of return substantially exceed the social
cost of investment capital. The level of the rates of return will depend
upon the assumption adopted about the true position of the tradition
station research curve (see section 3.7), ranging for the first pricing
option from the very unlikely 47 percent in the case of a cut off point
of 1982 to 245 percent in the case of a cut off point at infinite.
Using the authors' criteria, the case of t =1982 can be disregarded
as very unrealistic. Not only it ignores by assumption any "research
effect" in the implementation of the on-farm research methodology, but
also, it attributes an almost negligible "extension effect" which appears
to be inconsistent with the fact that in the past, actions of traditional
research and extension have not resulted in significant changes in
technological production patterns in the area, while as shown in the
adoption patterns, important technological changes have taken place soon
after the OFR starts rendering its first results.
A more realistic lower bound would be that of t =1985, while the
upper bound would be given by t =o. The first case would still represent
a purely "extension effect" as the only advantage of on-farm research
over traditional station research strategy. On the other hand, the upper
bound would represent a removal of the restrictive assumption that
traditional station would have generated in due time the same
technological alternatives as the OFR program. In other words, this cutt
off point stress the research effect of OFR by actually assuming that the
alternatives generated in Caisan would not have been available should the
OFR had not been in operation. The cut off point of t =1990, represents
a middle point. It compromises between stressing the "extension effect"
in scme components like the case of fertilizers and planting arrangement
and density where relatively high h(tl) values were found, and stressing
the "research effect" in others like herbicides and zero-minimum tillage
in which case the survey shows zero or near zero h(tl) values.
Summarizing, the rate of return due to the methodological innovation
in Caisan ranges between 118 and 245 percent, with a most likely value of
194 percent in the case of social prices without regulations and between
155 and 325 percent, with 255 percent as most likely value when
regulations are taken into account. The most likely values of 194 and
255 percent are revealing that each dollar invested up to the base year
in implementing the on-farm research methodology in Caisan, render a flow
of social net benefits of 1.94 and 2.55 dollars per year depending on the
In the last five years considerable progress has been accomplished by
national agricultural programs in the testing and developing of
operational methodologies for on-farm research, as well as in terms of
the amount of resources allocated to this type of activities. As this
process evolves, initial methodological and technical problems are
resolved and new ones take their place, among them that of the
institutionalization of on-farm research within national research
The starting point for this institutionalization process has been
the experience arising from ongoing area-specific, on-farm research
programs. These have usually been managed in the initial stage of this
process by ad-hoc technical groups from within the research structure.
From CIMMYT's perspective, the process builds from basic methodological
ideas, to on-farm research experiences, to the institutionalization of
these activities within the national program. In other words, it goes in
a bottom up approach, from on-farm research actions to an articulated
on-farm research program.
IDIAP and Caisan illustrate this process. The institutional
strategy of IDIAP provided the framework for the development of Caisan.
The progress of the Project and the methodological experiences arising
from it were closely followed by the national directing staff and
intensively discussed by researchers and directing staff in national
meetings, field days and regional workshops.
This in turn has lead to a reinforcement of the initial orientation
of IDIAP towards area-specific, on-farm research. Also, in the
methodological dimension, it provided concrete experiences, not only in
terms of what to do in on-farm research (surveys, experiments, etc.) but,
more important, how to do it, i.e., the informal survey leading to a
well-focused formal questionnaire, the prescreening of best-bet
technological components based on the assessment of farmer circumstances,
the management of experimental and nonexperimental variables within the
The present economic evaluation represents a contribution to this
process of institutionalization of on-farm research with some
methodological implications in terms of evaluation research; 2/ which in
this case has been adjusted to the particular requirements and
institutional circumstances of IDIAP.
An impact evaluation in the more traditional sense (for example
impact on yields, impact on production) could represent a necessary, yet
not sufficient condition for additional support of this activities on the
part of policy makers. Among other things, meeting sufficient condition
No work of similar nature evaluating on-farm research was found in
the review of the literature on economic evaluation of agricultural
for increasing support also requires dealing with the more controversial
issues of the cost efficiency of this methodological innovation in
agricultural research. This would be normally considered before the
process could be advanced toward the more complex stages of
institutionalizing on-farm research within the present national
structures. The evaluation has been carried incorporating both elements,
impact and cost efficiency, which are sinthetisized in the social rate of
return for the IDIAP investment required to implement the Caisan
In terms of economic evaluation methodology, section III showed how
the interaction between the "state of the arts" and the particular
requirements coming from the nature of on-farm research activities,
resulted in certain key decision for the evaluation methodology to be
implemented. Farmer adoption,' and consequently actual impact on
production, was required for associating any positive benefits to the
research process. The same information generated in the on-farm research
process was used in the evaluation. This included the initial planning
survey of 1978, and the trial results from three production cycles
adjusted by risk factors prevailing in the area. In addition to this, an
adoption survey specially designed to fit the objectives of the
evaluation, was implemented in 1982. With all this, the amount and
quality of the information used in the evaluation appear to be superior
to that used in evaluations of similar nature. In this sense, it is
believed that the information coming from appropriately conducted on-farm
experiments will close Dalrymple's gap between sophisticated evaluation
theory and poor data base (see section IV, 2.1).
Meaningful conclusions came out as a result of the evaluation
conducted. The level of adoption of the technological alternatives gave
indication of the degree in which they fitted circumstances of
24/ For example, the pioneer work of Griliches (1958), in a more
ambitious evaluation (hybrid corn in U.S.), assumes that yield
increase associated with the use of hybrids is of 15 percent over
pre-innovation yields. The equivalent information in the case of OFR
in Caisan is based in three cycles of on-farm trials, which
incorporates farmer practices as a base for the analysis.
representative farmers from Caisan. On the other hand, the speed at
which adoption took place indicated not only that the technologies
developed were agro-economically viable for representative farmers, but
also that they represented a solution to an important problem faced by
those farmers. This priority had been correctly identified in assessing
research opportunities in the planning stage of the program.
In other words, the results in terms of adoption speaks out of the
usefulness of the OFR methodology for reaching representative farmers
with appropriate technologies in the near term. In assessing "how
useful," the evaluation was required to enter into more quantitative
aspects dealing with the specific contribution of the methodological
innovation represented by OFR against the "without alternative"; that is,
what would have happened with purely traditional station research. This
way of approaching the evaluation is consistent with the conceptual role
attached to OFR as a needed complement of traditional station research.
Quantitative aspects of evaluating methodological innovations were
covered by considering three alternative assumptions about the relative
importance of the research effect of OFR over that of TSR.
The first assumption puts a zero weight to this effect attributing
.an extension effect as the only advantage of OFR over TSR. This provides
a lower bound for the economic returns to national investment on
methodological innovations. The assumption is represented by considering
that the flow of net benefits attributed to the programs last only to
In the other extreme of the scale the upper bound of the returns is
obtained by strongly weighting the research effect attributable to OFR:
technological alternatives engendered by the Project would not have been
generated otherwise. In this case, the flow of net benefits are extended
Finally, an intermediate case is considered by differentially
weighting the research effect by technological components. This provides
a measure of economic returns which appear as the more likely to occur.
In addition, two pricing scenarios for maize were considered. The
first one, consistent with a perspective of general equilibrium and
welfare considerations, assumed the existence of non regulations; the
other, more consistent with the specific objectives of this particular
evaluation, takes price policy as a parameter faced by IDIAP. In this
scenario, price policy is part of the framework (set of parameters)
within which the institution should display its research policy and
Results indicated that disregarding what pricing scenario is adopted
the lowest social benefits generated with the OFR methodology have been
comfortably above research costs. This lowest bound for the rate of
return ranges between 118 and 155 percent, depending on the pricing
scenario adopted, while in the case of the upper bound the rate
fluctuates between 245 and 325 percent.
Finally, when the most likely case of a net flow lasting to 1990 is
considered, the rate of return ranges between 194 and 255 percent.
These results reafirm the perception, based on this and other
experiences, 26 that the OFR methodologies used in Caisan are cost-
efficient in reaching target farmers with appropriate technologies in the
Consistently with proceeding results, IDIAP area-specific on-farm.
research activities, have gone through considerable expansion since 1978
when the Caisan program begun with only two national researchers. At
SAs a reassurance the benefit-cost ratio and rate of return for a
benefit flow lasting only to 1982 (the base year of the evaluation)
was calculated. The rate of return ranges in this case between 47%
and 61% depending on the pricing scenario. This indicated that even
after only less than four years of the beginning of the program,
social benefits basically accruing to farmers, outweighed research
26/ For example see Moscardi, E. (1983).
present these activities include five priority areas in agriculture,
involving the work of 24 national researchers, and three priority areas
in livestock with 21 researchers. As this expansion takes place, a set
of issues related to the institutionalization of on-farm research becomes
a matter of primary concern of IDIAP. As the central management moves
to cope with these issues, the institution comes closer to realizing its
full potential for the benefit of Panamanian farmers and the society as a
2 See Martinez and Arauz (1983), Section 5.
TABLE 10. TA1, CHEMICAL WEED CONTROL, NET BENEFITS PER UNIT OF LAND
A. Gross Benefits por ha. (A R x Pm)
B. Net Change in Variable Costs (A C)
b.l Value of Additional Resources
i. 2.0 kg Gesaprin x $8.13/kg
ii. 1.0 It Gramoxone x $5.5/it
Weighted average 1/
b.2 Value of Replaced Inputs
i. 1 It. 2-4-D x $2.58/it.
C. Cost of Capital (15% 8 Months)
D. Net Benefits per ha
Social price Social price
without regulation with regulation
$ 145.31/ha $ 195.46/ha
1/ The average was weighted by the acreage proportion controlled with
each product (weight for Gesaprim = .1166; w~:ight for Grarroxone
TABLE 11. TA2, SPACING ARRANGEMENT AND DENSITY
Social price Social price
without regulation with regulation
A. Gross Benefits per ha (AR x Pm) $115.71/ha $156.27/ha
B. Net Change in Variable Costs (AC) 11.05/ha 11.05
b.l. Value of Additional Resources
i. 3 kg/ha of maize seed x $0.35/kg 1.05/ha 1.05/ha
ii. 2 days/ha of labor x $5.00/day 10.00/ha 10.00/ha
b.2. Value of Replaced Inputs.
C. Cost of Capital (15% 8 months) 1.11/ha 1.11/ha
D. Net Benefits per ha. (A B) 103.55/ha 144.11/ha
TABLE 12. TAs 3.1, 3.2, ZERO TILLAGE AND MINIMUM TILLAGE /
Zero Tillage Minimum Tillage
A. Gross Benefits por ha. (A R x Pm)
B. Net Change in Variable Costs (A C) $ 18.75/ha $12.75/ha
b.l Value of Additional Resources
i. Two men days/ha x $5.00/day (hand 10.00/ha
ii. 1.5 It Gramoxone x $5.5/lt 8.25/ha 8.25/ha
iii. Two men days/ha x $5.00/day
(application of Herbicide) 10.00/ha 10.00/ha
iv. User cost of back pack sprayer 1.00/ha 1.00/ha
b.2 Value of Replaced Inputs
i. 3 machine hours/ha x $16.00/hour 48.00/ha
ii. 2 machine hours/ha x$16.00/hour 32.00
C. Cost of Capital (15% 8 Months) 1.88/ha 1.28/ha
D. Net Benefits per ha (A-B) 20.63/ha 14.03/ha
1/ Since technological alternatives 3.1, 3.2 and 4.0 are purely input
saving technologies without affecting yields the distinction between
alternative pricing of maize becomes irrelevant and is omitted from
TABLE 13. TA4, FERTILIZATION
A. Gross Benefits per ha (LR x Pm)
B. Net Change in Variable Costs (A C)
b.l. Value of Additional Resources
b.2. Value of Replaced Inputs.
i. 2qq/ha of mixed fertilizer x $17.37/qq
ii. 1 man day/ha x $5.00/day (application)
C. Cost of Capital (15% 8 months)
D. Net Benefits per ha. (A-B)
TABLE 14 FIL~ OF NET BENEFITS GENERATED BY THE PROGRAM.
a) Social Prices without Regulation
Technological 1 2 3 4 Total
yea ternatives Chemical Weed Spatial Arrange- Zero-Mininum Fertilization Caisan
Control ment and Density tillage Project
1980 15,901 4,038 0 7,431 27,370
81 43,201 30,237 565 '13,681 87,864
82 74,159 43,491 3,250 17,790 138,690
83 96,393 52,086 11,220 20,107 179,806
84 102,303 56,849 12,457 21,287 192,896
85 112,576 59,231 13,482 21,285 207,144
86 114,124 60,370 13,808 22,161 210,413
87 114,687 60,991 13,822 22,292 211,792
88 115,109 .61,198 13,822 22,336 212,465
89 115,109 61,405 13,822 22,336 212,672
90 115,109 61,405 13,822 22,336 210,672
1990 115,109 61,405 13,822 0 190,336
b) Social Prices with Regulation
Technological 1 2 3 4 5
year ternatives Chemical weed Spatial Arrange- Zero-Minimum Fertilization Total
Control ment and density tillage Caisan
1980 21,568 5,620 0 7,431 34,619
81 58,597 42,080 565 13,681 114,923
82 100,588 60,526 3,250 17,790 182,154
83 130,746 72,487 11,220 20,107 234,560
84 138,762 79,116 12,457 21,287 251,622
85 152,696 82,431 13,482 21,855 270,464
86 154,796 84,016 13,808 22,161 274,161
87 155,559 84,881 13,822 22,292 276,554
88 156,132 85,169 13,822 22,336 277,459
89 156,132 85,457 13,822 22,336 277,747
90 156,132 85,457 13,822 22,336 277,747
1990 156,132 85,457 13,822 0 255,411
1) Akino, M. and Y. Hayami. 1975. "Efficiency and Equity in Public
Research: Rice Breeding in Japan's Economic Development." American
Journal of Agricultural Economics, vol. 57 p. 1-10.
2) Arauz, J.R., y J.C. Martinez. 1983. "Desarrollando Tecnologia
Apropiada Para el Agricultor. Informed de Progreso del Programa de
Caisan en Panama." Serie de Estudios Especiales No. 1. IDIAP,
3) Boadway, R.W. 1974. "The Welfare Foundations of Cost-Benefit
Analysis." The Economic Journal, p. 926-39.
4) Bredhal, M.E. "The Productivity and Allocation of Research at U.S.
Agricultural Experiment Stations." Ph.D. Thesis, University of
5) Byerlee, Derek, Michael Collinson, et.al. "Planning Technologies
Appropriate to Farmers Concepts and Procedures." CIMMYT, Mexico,
6) Cochrane, W. 1958. "Farm Prices: Myth and-:Reality." University of
7) Currie, J.M., J.A. Murphy and A. Schmitz. 1971. "The Concept of
Economic Surplus and Its Use in Economic Analysis." Economic Journal
18, p. 741-98.
8) Dalrynple, Dana G. 1980. "Yield Information Needed for the Analysis
of Crop Improvement Research." In Evaluation of Agricultural
Research. Proceedings of a Workshop Sponsored by NC-148.
Miscellaneous Publication 8-1981, Minnesota Agricultural Experiment
Station, University of Minnesota.
9) De Janvry, A. 1978. "Social Structure and Biased Technical Change in
Argentine Agriculture." In Hans P. Birswanger, V.W. Ruttan and
others "Induced Innovation. Technology, Institutions, and
Development". J. Hopkins Press. Baltimore, p. 297-323.
10) Evenson, R.E. 1967. "The Contribution of Agricultural Research to
Production." Journal of Farm Economics vol. 49, p.1415-25.
11) Griliches, Z. 1958. "Research Costs and Social Returns: Hybrid Corn
and Related Innovations." Journal of Political Economics, vol. 66
12) Griliches, Z. 1964. "Research Expenditures, Education and the
Aggregate Production Function." American Economic Review, vol.54,
13) Hertford, R. and A. Schmitz. "Measuring Economic Returns to
Agricultural Research." In T.M. Arndt, D.G. Dalrymple, and V.W.
Ruttan eds. "Resource Allocation and Productivity in National and-
International Agricultural Research." Minneapolis: University of
Minnesota Press (1977).
14) Irvin, George. "Modern Cost-Benefit Methods." Harper and Rcw. Pub.
Inc. New York (1978).
15) Kahlon, A.S., H.K. Bal, P.N. Saxena, D. Jha. "Returns to Investment
in Research in India." In T.M. Arndt; D.G. Dalrymple and V.W. Ruttan
eds. "Resource Allocation and Productivity in National and
International Agricultural Research. Minneapolis. Minnesota Press
16) Kislev, Y. and M. Hoffman. "Research and Productivity in Wheat in
Israel." Journal of Development Studies 14 (1978) p. 166-81.
17) Lindner, R.K. and F.G. Jarret. "Supply Shifts and the Size of
Research Benefits." American Journal of Agricultural Economics, 60
(1978) p. 48-56.
18) Lu, Y.; L. Quance and C.L. Liu. "Projecting Agricultural
Productivity and its Economic Inpacts." American Journal of
Agricultural Economics. 60 (1978) p. 976-80.
19) Lu, Yao-Chi. "Ex Ante Evaluation of the Separ.-t Effects of
Research and Extension." In Evaluation of Agricultural Research.
Proceedings of a Workshop Sponsored by N.C-148. Miscellaneous
Publication 8, 1981, Minnesota Agricultural Experiment Station,
University of Minnesota.
20) Martinez,. J.C. "On The Econmnics of Technological Change: Induced
Innovation in Argentine Agriculture." Ph.D. thesis, Icva State
University, Ames, Icoa, (1973)..
21) Martinez, Juan Carlos. "Desarrollando Tecnologia Apropiad a las
Circunstancias del Productor: El Enfoque Restringido de Sistemas de
Producci6n." CI2,MYT, Reporte 1981.
22) Martinez, J.C., Arauz, J.R. "Institutional Innovations in National
Agricultural Research: On-fann Research within IDIAP, Panama."
CIMXYT Economics Program Working Paper 02/83.
23) Mishan, E.J. "What is the Producer Surplus?" American Economic
Review 58, (1968) p. 1269-82.
24) Moscardi, Edgardo, et. al. "Creating an On-Farm Research Program in
Ecuador. The Case of INIAP's Production Research Program." CIMMYT
Economics Program Working Paper, 01/83. 1983.
25) Norman, D.W. "Farming Systems Research to Inprove the Livelihood of
Small Farmers." American Journal of Agricultural Economics, 60(5):
26) Norton, George W. and Davis. "Review of Methods Used to Estimate
Returns to Agricultural Research." In Evaluation of Agricultural
Research. Proceedings of a Workshop Sponsored by NC-148.
Miscellaneous Publication 8-1981, Minnesota Agricultural Experiment
Station, University of Minnesota, 1981.
27) Perrin, R.K., D.L. Winkelmann, E.R. Moscardi and J.R. Anderson.
"From Agronomic Data to Farmer Reconmendations. An Economics
Training Manual." CIMMYT Information Bulletin 27 (1979).
28) Peterson, W.L. "Returns to Poultry Research in the United States."
Journal of Farm Economics 49 (1967) p. 656-69.
29) Ruttan, Vernon W. "Induced Institutional Change." In Hans P.
Birswanger, V.W. Ruttan and others "Induced Innovation. Technology,
Institutions, and Development." J. Hopkins Press. Baltimore 1978. P.
30) Schuh, Edward G. and Helio Tollini. "Costs and Benefits of
Agricultural Research: State of the Art and Implications for the
CGIAR." Consultative Group International Agricultural Research.
World Bank Staff Working Paper No. 360, Washington, D.C. 1979.
31) Schmitz, A. and D. Seckler. "Mechanical Agriculture and Social
Welfare: The Case of the Tomato Harvester." American Journal of
Agricultural Economics 52 (1970) p. 569-78.
32) Schultz, T.W. "The Economic Organization of Agriculture." New York
33) Scobie, Grant M. "Investment in International Agricultural Research:
Some Economic Dimensions." World Bank Staff Working Paper No. 361,
Washington, D.C. 1979.
34) Steiner, H.M. "Public and Private Investment. Socio Economic
Analysis." New York, John Wiley and Sons (1980).
35) Wise, W.S. "The Theory of Agricultural Research Benefits." Journal
of Agricultural Economics 32 (1981) p. 147-157.