f/ ,, O 4-
ON-FARM RESEARCH TO DEVELOP TECHNOLOGIES APPROPRIATE TO FARMERS
Derek Byerlee *
Juan Carlos Martinez
Paper presented at the Conference of the International Association
of Agricultural Economists, Banff, Canada, Sept. 1979
* The authors are economists with the International Wheat and Maize
Improvement Center (CIMIYT), Londres 40, M6xico 6, D.F. Biggs works
in the South Asian Region, Collinson in Eastern Africa, Martinez in
Central America and Moscardi in the Andean Region. Byerlee, Harrington
and Winkelmann are based in Mexico. Views expressed are not necessarily
those of CIBMYT.
ON-FARM RESEARCH TO DEVELOP TECHNOLOGIES APPROPRIATE TO FARMERS
It is now widely accepted that technological change is a necessary al-
though by no means sufficient condition for agricultural development. It is
clear that despite the widespread diffusion of new wheat and rice varieties,
many new technologies are not being widely used by farmers because they do
not fit the particular circumstances of farmers for whom they are intended.
This is despite the fact that considerable public expenditures are often made
to provide the infrastructure such as credit and markets to enable the farmer
to adopt these technologies.
This paper attempts to synthesize our experiences with national research
programs and with CIMrIYT's wheat and maize programs in developing research
methodologies to ensure that agricultural technologies generated by scarce
research resources are consistent with the circumstances of target farmers.
It emphasizes ex ante collaboration of technical and social scientists in
on-farm research both in diagnosing farmers' problems and demands for tech-
nology and in the development and testing in farmers' fields of those technol-
ogies which appear to meet these problems.
Traditional Approaches to Agricultural Research: Agricultural research is gener-
ally characterized by the gap between the researcher and the farmerI/. On the
one hand much research has been guided by disciplinary interests. Although the
importance of the problem to the farmer is sometimes advocated in determining
research priorities, no explicit means of identifying priorities is employed.
On the other hand, even research aimed at farmer problems has traditionally
used a top-down approach that is, it is conducted on research stations under
conditions quite different to those of farmers and then passed to extension for
promotion to farmers. Although the problems of extrapolating these results to
farmers have been recognized (e.g. Swanson), the movement of research to farmers'
fields has been slow. An exception is experimentation on fertilizers which has
long been conducted on farmers' fields in many countries; but practices under
which these experiments are conducted (e.g. weed control, land preparation) are
often quite different from those of local farmers.
The economist in this process has usually been a late .actor. Large scale
involvement began with production function analysis of agronomic (usually ferti-
lizer) experiments. In some cases this led to the collaboration of agronomists
and agricultural economists in the ex ante design of fertilizer experiments
(e.g. Hoffnar and Johnson). More recently in developing countries, the economists
and other social scientists have been even later participants through ex post
studies of technology adoption. Increasingly these studies reflect the fact that
recommended technologies are not appropriate to particular farmer circumstances/.
However, these adoption studies have largely been conducted by economists outside
of agricultural research institutions and as a result there has been little
immediate feedback to decision-making on research priorities for developing im-
Toward Integrated On-Farm Research Programs: There is now increasing emphasis
on collaboration of technical and social scientists in on-farm research to bridge
the gap between researchers and farmers. The general approach embraced in
various degrees by various institutions (e.g. Norman, CIIIYTT, Dillon et al,
-The assertions in this section arise from our experience in many countries.
There are of course many exceptions to these generalizations.
-/See for example the Perrin and Winkelmann review of several such adoption
Hildebrand (ICTA, Guatemala), Navarro (CATIE) has three important components:
(a) The approach emphasizes solving farmer problems specific in time
and location. It begins with an ex ante identification of current
farmer problems and possible technological solutions to these
problems that are feasible under the natural and socio-economic
circumstances faced by farmers and that are consistent with national
(b) Technologies appropriate to farmers are then developed and evaluated
by experimentation in farmers' fields under farmers' conditions.
(c) Farmers' experiences with the new technology are monitored and this
information fedback to research decision-making.
There is now growing support for the central role of on-farm research in
national agricultural research programs. However the actual methodologies for
implementing this type of research vary quite widely. In searching for such a
iechodoliogy we have been conscious of tne need for s~-eVral basc criteria;
a) The research should be well focused to enable quick payoffs to relatively
limited research resources. This means research is highly location specific
and focuses on technology for an important or potentially important crop or
crop mixture (in our case wheat or maize) rather than considering all crops
and crop technologies in the system as variables. However, the identification
and evaluation of potential technologies for a single crop must be made in the
context of the farming system. Often small farmers operating in imperfect
factors markets in an uncertain environment will operate highly complex systems
to meet an overriding food security objective. While yield increasing technol-
ogies are important, technologies for the target crop which have total system
benefits are also necessary (e.g. an earlier variety to allow two crops per
b) The farmer's decision-making with respect to technology is coidiLioned by
natural circumstances such as soils and climate and by economic circumstances
such as resource endowments and access to markets. To understand this complex
of factors a multi-disciplinary research team usually consisting of an agrono-
mist and an economist is needed to plan research with farmers.
c) Technological adoption by farmers is a learning by doing process that pro-
ceeds in small steps. The on-farm research should therefore set as an objective
the generation of a few best-bet technological components. Furthermore, it is
not possible to provide precise recommendations to each farmer but recommend-
ations can be made which are generally relevant to representative groups of
d) On-farm research should be part of a broader program to improve crop pro-
duction and farmers incomes and therefore must be closely linked to experiment
station research, policy making and extension.
e) The methodology of on-farm research should be practical and replicable in
the context of scarce research resources of developing countries. It should
be relatively cheap in implementation and enable a fast turn-around in results.
The remainder of this paper summarizes the methodology we have found to
meet these criteria. Figure 1 shows the specific steps in the methodology*
Choice of Target Region and Crop: Initially the choice of the crop and region
for on-farm research program must be justified against the objectives of national
policy and the resources and logistics available for the research. If national
policy dictates that low income farmers should be priority beneficiaries of
research expenditures, then the on-farm research will focus on a region where
low income farmers are concentrated and where technologies are available with
* This methodology is described in detail in our Manual "Planning Technologies
Appropriate to Farmers;Concepts and Procedures".
potential to increase production of a crop which is important in the farming
system of these low income farmers. In this process researchers will want to
also consider the future perspective as well as the present for example,
will there be an adequate market for the increased production. This initial
matching of the likely outcome of the research with national development
priorities helps in allocation scarce research resources.
Collecting Information on Farmer Circumstances: The research then focuses on
an understanding of farmer circumstances in the target region. This phase has
both diagnostic and descriptive objectives. The primary objective is of course
to diagnose the problems and constraints to crop production in the area in
order to prescreenn" from a wide range of possible technological components a
few "best-bet" components for experimentation in farmers' fields. Information
from this diagnostic stage can also be used to guide experiment station research
The diagnostic effort also uncovers particular constraints at the farm level
which are the result of policy decisions or problems in policy implementation
(e.g. problems of input availability, product marketing and credit). Secondly
a description of farmer circumstances enables farmers in the target region to
be tentatively classified into relatively homogeneous groups or recommendation
domains for which the same recommended technology will be generally applicable.
Also this description of current farmers' practices and fields is important to
establish representative farmers' practices and sites for on-farm experiments.
In the context of this paper, fanner circumstances are all those factors
which bear on farmers decisions with respect to technology for the target crop.
These include natural factors such as climate, soils and pests and socio-economic
factors including the environment of markets, infrastructure and land tenure
and Lhe fanmers' own goals and resource constraints. Often these factors
influence the farmer's choice of a crop technology through interactions within
the farming system, i.e. resource competition, crop rotations and multiple
cropping, risk management, and food preferences. To the extent that these
interactions are important, information must be obtained on other activities
of the farming system.
Three sources of information on fanner circumstances are used: (a) back-
ground information on the farmers' environment, usually from secondary sources,
(b) interviews with farmers, and (c) observations in farmers' fields. Typically
available background agro-climatic and socio-economic data are first collected
and analysed. For example, in dryland areas rainfall data from several sites
are checked for important differences across the region and for periods of
major rainfall uncertainty. A team--usually an agronomist and an economist--
will then spend one to four weeks touring the region in an exploratory survey
of farmers and other persons linked to the farming community such as merchants,
inputs suppliers and extension agents. At this stage informal interviews of
farmers and visits to farmers' fields are conducted. A questionnaire is not
used although the data is collected in a systematic manner against a mental
"checklist" of issues and problems. Efforts are made to talk to traditional
leaders who can often explain traditional practices, to innovative farmers
who may or may not be working closely with the extension service and to farmers
encountered by chance on the tour. The researchers begin by trying to obtain a
broad perspective on the fanning system. As the exploratory survey proceeds,
the interviews become more focused on specific problems and hypotheses to
explain farmers' practices for the target crop within the context of the larger
farming system. Also at this stage tentative definitions of recommendation
domains are formulated and potential technological components are identified.
The primary role of this exploratory survey is to place the researcher
in the farmers' fields in direct communication with the farmer and to help
design a sharply focused formal one-contact survey of farmers in the region
in order to quantify and verify what has been learned in the exploratory
survey and investigate some critical problems in more depth. Because a random
sample of farmers is interviewed, the use of certain practices can be quanti-
fied and hypotheses on the reasons for these practices formally tested.
Furthermore relatively little emphasis is placed on quantifying farmers'
resource use and allocation in order to infer technological needs. Rather the
questions aim to exploit the farmers own intimate knowledge and experience of
his environment in order to identify these needs. As a result, many questions
(again based on exploratory survey information) elicit subjective types of
information such as preferences about specific varietal characteristics.
The implementation of the formal survey the training of enumerators,
sampling and field work generally follows standard procedures for this type
of work (e.g. Collinson). Farmers are stratified as far as possible by the
tentatively defined recommendation domains and about 30-60 farmers are
interviewed in each recommendation domain. The questionnaire is designed to
be completed in 45 to 90 minutes.
Data Analysis and Prescreening Technologies: Data are analysed quickly
after the survey (with a maximum of 3 months) usually using hand tabulation
sometimes supplemented by computer analysis. Descriptive tabulations of
farming systems and cultural practices with respect to factors such as rain-
fall and farm size, are used to refine boundaries of recommendation domains.
These recommendation domains will only be broad classification of farmers and
much heterogeneity will still remain within each domain. The descriptive
tabulations also provide a profile of reprsntative farmers' practices and
fields in each domain for the design of on-farm experiments.
The diagnosis of research priorities is made in the following steps:
a) Important reasons for farmers using current technologies are listed, b)
Priority problems and constraints in the target crop are identified on the
basis of these reasons, results of field observations and farmers' opinions
and perceptions, c) Possible solutions to these problems and constraints are
noted on the basis of practices of innovative farmer, on-station research and
results and agronomic-economic expertise, d) All changes in the farming system
implied by each solution including associated costs, labor needs and risks are
listed. Based on researchers' understanding of the current farming system,
those changes which are subjectively felt to be inconsistent with farmers'
circumstances are eliminated (e.g. cash costs too high, unacceptable risks,
conflicts with present multiple cropping systems), e) Partial budgets follow-
ing Perrin et al are constructed assuming a priori best guestimates of yield
responses for these technological components, f) In each recommendation domain
a few best-bet components arising from this "prescreening" exercise are chosen
for on-farm experiments.
This prescreening process emphasizes identifying technologies which use
resources and inputs available to farmers and have short run pay-offs.
However, experiments may be included with a longer run horizon. For example,
such experiments may provide information on the desirability of making a new
On-Farm Experimentation: Space limitations prevents a detailed discussion of
on-farm experimentation here. It serves three purposes. First, it enables
technology to be developed and tested under farmers' conditions. Most
experiment stations are managed in such a way that over time, soil structure,
fertility, weeds, pests and diseases are quite different from those in farmers'
fields. Second, the technology can be developed and tested over a variety of
environments and cultural practices representing a region. Finally, the
farmer and extension can be actively involved in the process of developing and
demonstrating technological components.
Several types of on-farm experiments are implemented by the same multi-
disciplinary team responsible for the surveys in order to test the three to
four priority technological components arising out of the diagnostic studies
(Violic et al, Winkelmann and Moscardi). Exploratory experiments are 2n
factorials with levels of each factor set at the farmers' level and at
substantially higher levels to look at main effects and first-order inter-
actions of each factor. On the basis of these experiments, experiments with
one or two factors (deending on interactions noted in the exploratory exper-
iments) are designed to find recommended levels of these factors in terms or
income and risks_/. Finally factors are combined to verify tentatively
recommended technologies in comparison to the farmers' technology. These
verification experiments often serve as the basic design for extension-run
demonstrations. To ensure relevance all these experiments are conducted under
conditions determined in the formal survey to be representative of local
Dynamics of On-Farm Research: On-farm research is a continuous learning process.
After each cycle of on-farm research, information from surveys and experiments
are integrated and analysed and strategies for the following cycle formulated.
Special purpose surveys may be organized, particularly to monitor how farmers
use the recommended technologies when they themselves must pay all costs and
accept the risks. This provides important feedback to the research program.
If cooperating farmers are accepting the technology then it can be promoted
through extension and demonstrations, and new technological components,
previously of lower priority, can be incorporated into the research program.
On the other hand where farmers reject, or significantly change the recommended
technology, an understanding of why this is so could influence the design of
future experiments. Likewise, experimental results may help modify recommend-
ation domains or identify new technological components not previously considered
Linkages of On-Farm Research, Experiment Station Research and Policy: Effective-
ness of on-farm research can be greatly strengthened by maintaining close link-
ages with research on experiment stations. Experiment station research focuses
on developing and screening new technological components that is, research
that usually requires greater control (e.g. development of new varieties) or
would be risky when done on farmers' fields (e.g. screening new herbicides).
Promising new technological components arising from station research are then
submitted to experimental evaluation in farmers' fields. The results of on-farm
research also help establish priorities for station research. For example, a
knowledge of farmer circumstances can guide plant breeders in deciding between
yield, earliness, specific disease resistance, grain type, and storability in
-/Procedures for analysing experimental data and making recommendations are
-described in Perrin et al (1976).
On-farm research is also conducted in a specific policy context that
might guide the selection of target farmers and technologies consistent with
national goals. In addition, in most countries there is a shortage of micro-
level information for policy analysis. The detailed information on farmers'
circumstances and technological responses under farmers' conditions generated
by on-farm research can be important for identifying changes in policy and
policy implementation that would complement the introduction of improved
technologies (e.g. increasing the availability of specific inputs).
Finally, the impact of on-farm research is increased if researchers
responsible for on-farm research are institutionalized within the agricultural
research establishments with appropriate incentives and logistics to work in
a multi-disciplinary team on priority farmer problems. This will require
agricultural research programs to include economists as an integral part of
the research staff. Our work initially focused on demonstrating to research
aiiinisraors the value of these on-farmi rsear-c proiedres a Ti potti i-.ial
contribution of economists (e.g. Collinson, 1978). The emphasis has now
shifted to assisting national research programs to develop their own capacity
for on-farm research. It has been shown that well trained four year agricultural
graduates are capable of implementing these on-farm research procedures in a
CIMMYT "Planning Technologies Appropriate to Farmer Circumstances:
Concepts and Procedures", Economics Program, CIMMYT, Mexico
Collinson, M.P. Farm Management in Peasant Agriculture, Praeger,
New York, 1972.
Collinson, M.P. "Demonstration of an Interdisciplinary Approach to
Planning Adaptive Agricultural Research Programs; Serenje District,
Zambia", Report No. 3, CIvI IYT Eastern African Economics Programs,
Dillon, John L. et al, Farming Systems Research at the International
Agricultural Research Centers, Technical Advisory Group, CGIAR
World Bank, Washington, 1978.
Hildebrand, P.E. "Generando Tecnologia para Agricultores Tradicionales:
Una Metodologia Multidisciplinaria", ICTA, Guatemala, 1976.
Hoffnar, G.R. and G.L. Johnson, Summary and Evaluation of the Cooperative
Agronomic-Economic Experimentation at Michigan State University,
Michigan State Experiment Station, Research Bulletin 11; E. Lansing, 1966.
Navarro, L.A. "Dealing with Risk and Uncertainty in-Crop Production; A
Lesson from Small Farmers", CATIE, Turrialba, Costa Rica, 1977.
Norman, D.W., "Farming Systems Research to Improve the Livelihood of Small
Farmers", American Journal of Agricultural Economics 60(5); 813-818, 1978.
Perrin, R. and D. Winkelmann, "Impediments to Technical Progress on Small
vs. Large Farms", American Journal of Agricultural Economics, 58(5)
Perrin, R. et al, From Agronomic Data to Farmer Recommendations, CIMvYT,
Swanson, E.R., "Problems of Applying Experiment Results to Commercial
Practice", Journal of Farm Economics, 39(2); 382-389, 1957.
Violic, A. et al, Maize Training in the International Maize and Wheat
Improvement Center, CLMIYT, Mexico, 1976.
Winkelmann, Donald and Edgardo Moscardi, "Aiming Agricultural Research at
the Needs of Farmers", CIMMYT, Information Bulletin, 1979.
FIGURE 1. OVERVIEW OF AN INTEGRATED RESEARCH PROGRAM
Choice of Target
Farmers and Research
Priorities based on
Policy Related Problems
Background --> Exploratory <-- Formal
Data Survey Survey
Reasons for Practices
Component Effects and Interactions
Levels of Components
Verification of Technologies
Extend Technologies Implications Analysis
Formulate Recommendations Integrate Survey and
Demonstrate Recommendations -->Experimental data to--
Monitor Farmers' Experiences plan future activities
Priorities for Station