Planning technologies appropriate to farmers

Material Information

Planning technologies appropriate to farmers concepts and procedures
International Maize and Wheat Improvement Center
Place of Publication:
México D.F
Economics Program, Centro Internacional de Mejoramiento de Maíz y Trigo
Publication Date:
Physical Description:
75 p. : ill. ; 28 cm.


Subjects / Keywords:
Agriculture -- Research ( lcsh )
Agricultural innovations ( lcsh )
bibliography ( marcgt )
non-fiction ( marcgt )


Includes bibliographies.
General Note:
"Preliminary version being circulated for comment."

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact Digital Services ( with any additional information they can provide.
Resource Identifier:
08095551 ( OCLC )


This item has the following downloads:

Full Text

T Londres 40, Apdo. Postal 6-641, Mexico 6, D.F. M6xico

16. ._-




Prepared by:

The CIMMYT Economics Program*

For Comment

* The views expressed in this inanual do not necessarily represent those of
CIMMYT. This is a preliminary version beino circulated for comment. A
.revised version will be prepared for CIMMYT publication in 1980.


- What this Manual is About
- The Need for New Procedures
- A Preview of the Manual
- A Note to the User


Overview of Research Procedures to Develop Technologies for Farmers
On-Farm Research
Experiment Station Research
Policy Context of Agricultural Research
The Place of this Manual in the Overall Research Procedures

Farmer Circumstances as a Basis for Planning Research
Definition of Farmer Circumstances
Decisions Required for Planning a Research Program
Identifying Farmers'Problems and Prescreening Technological Components
for On-Farm Experiments
Grouping Farmers into Recommendation Domains
Establishing Representative Practices and Sites for On-Farm Experiments
Identifying Problems for Experiment Station Research and Policy

A Checklist of Information on Farmer Circumstances
Description of Farmers' Technology for the Target Crop
Identification of Limiting Factors
Natural Circumstances
External Socio-Economic Circumstances
Resource Constraints and Goals of Farmers
Farmers Goals
Farming System Interactions


Overview of the Procedures
Sources of Information on Farmer Circumstances
A Sequence of Steps
Implementing the Research Procedures
Adjusting the Procedures to Fit the Researchers' Circumstances

Assembling Background Information on the Region
Sources of S2condary Data
Analyzing Secondary Information

The Exploratory Survey
The Exploratory Survey Process
Assembling information in the Exploratory Survey
Integration and Evaluation of the Exploratory Survey Data












Analyzing the Survey Data
Refining Recommendation Domains
Assembling Information on Farmers' Practices
Diagnostic Analysis of Farmer Circumstances
Methods for Tabulating Information
Weighting Procedures

Prescreening Potential Technological Components
Identifying Limiting Factors
Identifying Alternative Solutions to Limiting Factors
Listing All Changes to Farmers from Using Technological Comppnents
Economic Costs and Benefits to Potential Technological Components
Matching Potential Technological Components to Farmer Circumstances

Examples of Using a Knowledge of Farmer Circumstances to Plan Research
Planning On-Farm Experiments on Maize in East Africa
Planning On-Farm Trials in the Andean Region
Guiding Research on Tropical Maize Varieties in Dry Areas of Eastern

The Formal Survey-The Questionnaire
General Rules for Developing a Questionnaire
General Guidelines for Asking Questions
Guidelines to Obtaining Specific Types of Information
Finalizing the Questionnaire

Sampling for the Formal Survey
Random Sampling Procedures
Sample Size
Examples of Sampling

Implementing the Formal Survey
The Interviewer
Gaining Farmers' Cooperation
The Interview
Survey Implementation









We have prepared this Manual for profession-
als involved in research on improving agricultural
technology for farmers. We believe that it will be
useful to both biological scientists and social
scientists and that parts of the Manual, especially
Chapters 1, 2, 4, 11, and 12 will also be of interest
to those who administer agricultural research
Agricultural research should have as one of its
basic purposes the formulation of technologies
which can be widely used by farmers. Our purpose
in this manual is to present procedures which will
facilitate that effort particularly in the planning
stage. To a lesser but still important degree, we
want to relate the goals of agricultural research to
those of the larger society.
Two themes are central to the Manual. The
first is that effective research on agricultural
technology starts and finishes with the farmer. The
second is that integration of the perceptions of
biological scientists and social scientists is an
essential element in such research.

The Need for New Procedures
Although many farmers in developing coun-
tries are using improved varieties few farmers are
following in their entirety the recommendations
made by researchers and extension workers. Why
this occurs is the subject of a large body of litera-
ture. Some argue that farmers are at fault, some
that extension is ineffective, others that credit is
unsuitable, and some that inputs are not available
in a timely way. A less frequently heard explana-
tion is that the recommended technologies them-
selves are simply not appropriate for farmers.
Certainly one or the other of these explana-
tions is valid at some time and place. But a number
of recent experiences have shown even the
poorest farmers-presumably the most tradition-
bound and usually those with least access to
information, inputs, and markets-adopting certain
technologies while rejecting others. Based on
research on the diffusion of new cereals technol-

ogies1 in many countries, our own experiences
and the reports of many others, we concluded
that farmers often do not adopt recommendations
because they are not suitable for them. The
adoption of new technology hinges on many
interrelated factors. In general, farmers seek
technologies that increase their incomes while
keeping risks within reasonable bounds under
the circumstances within which the farmer has
to apply that technology, e.g. the resources
available to the farmer, the climatic, soils and
topographic characteristics of his land, pest and
disease complex of the crop and the input and
product markets in which he operates. We
concluded that recommendations are often not
consistent with these circumstances of farmers.
In conjunction with biological scientists
in CIMMYT and national research programs, we
began to search for concepts and procedures
which would lead to technologies well adapted to
farmers' needs. These procedures would have to
integrate information on the many natural and
economic circumstances that dominate farmer
responses to alternative technologies. Moreover,
to be useful to national research programs, these
procedures should not require more research
resources than are usually available.
This Manual focuses on only that part of
these procedures concerned with planning experi-
mental research to develop technologies for
farmers. We believe that the concepts and guide-
lines presented will help to develop technologies
which will be widely adopted.

A Preview of the Manual
This Manual treats concepts and procedures
for planning technologies for a single crop within
the farmer' total cropping system. While the
Manual features examples from maize and wheat
(sometimes in crop mixtures), the procedures can
be readily applied to other crops and cropping
systems. Although we emphasize biological

1/ See the series of CIMMYT adoption studies. A summary is given by R.K. Perrin and Donald Winkelmann in
"Impediments to Technical Progress on Small Versus Large Farms", American Journal of Agricultural Economics

technologies the procedures can be applied to
the development of mechanical technologies.
We divide the Manual into three parts.
Part I provides an overview of the concepts of
a collaborative research process to deliver tech-
nologies appropriate to farmers and of the types
of information regarding farmer circumstances
that are needed for planning this research. Part
II describes a set of procedures with examples
for obtaining information from farmers at
relatively low costs. Part III then provides
procedures and examples for incorporating this
information into the design of a research program.

A Note to the User
Before we proceed, we emphasize that we
present here a set of guidelines for planning
research to develop technologies appropriate
to farmers. These have evolved from our
experiences with farmers and researchers in many
countries. However we fully expect these guide-
lines to be improved through the experience of
other researchers and we hope users of this
Manual will contribute procedures and examples
from their own research so that we can improve
future editions of this Manual.



In chapter 1 we present an overview of the
organization of a research program that aims to
develop technologies for farmers. We then note in
chapter 2 the types of decisions that researchers
must make in order to plan such a program and
how a knowledge of the researchers' clients, the

farmers, is critical to each type of decision. Chapter
3 then discusses in more detail the type of infor-
mation on the farmer clientele that will be impor-
tant in this decision-making. This then leads to Part
II on procedures for obtaining this information.



The procedures described in this Manual are
part of a collaborative research process based on
the cooperation of applied scientists of different
disciplines and farmers, to develop technologies
which are appropriate to farmer circumstances
and which help to meet the goals. of national
Now let us expand on the concepts contained
in this statement. First, a technology is a combi-
nation of all the management practices for pro-
ducing or storing a given crop or crop mixture.
Each practice is defined by the timing, amount
and type of various technological components
such as seed-bed preparation, fertilizer use or
weeding. A subsistence farmer who uses no
purchased inputs is nevertheless using a technology
-sometimes quite complex. We are particularly
concerned with developing technologies appro-
priate to the circumstances of target groups of
farmers. Farmer circumstances are all those
factors which affect farmers decisions with respect
to a crop technology-their natural environment
(such as rainfall), their economic environment
(such as product markets) and their own goals,
preferences and resources constraints. If technol-
ogies are appropriate to farmer circumstances
they will, by definition, be rapidly adopted by
We also seek a technology that helps meet
the national policy goals of government. Most
governments desire increases in cereals production.

-therefore any technology which increases pro-
duction and is rapidly adopted by farmers will
help meet this goal. Most governments also have
goals of reducing income inequalities. This
may require technologies adapted to small farmers
or to poorer regions or that provide cheap food to
low income urban consumers.
Applied scientists-that is, those scientists
from different disciplines working to solve im-
mediate and high priority problems-are, with
farmers, the main actors in this research process.
In most cases these scientists should include one
biological scientist, usually an agronomist, to
integrate the physical and biological aspects of
crop production, and a social scientist, usually an
agricultural economist, to integrate various
aspects of the farmers' resource endowments,
goals and market environment. These disciplines
may be supplemented where there are specialized
problems-for example an entomologist might
participate in solving a particular insect problem.
We belive that it is essential that the agronomist
and agricultural economist collaborate in all
phases of the research and that major decisions
such as the design of on-farm experiments are
made jointly.
With these concepts as background, Figure 1.1
gives an overview of an integrated research program
to develop technologies for farmers. At the base
of these procedures is on-farm research. This
research is however, linked to two other important


Figure 1.1. Overview of an Integrated On-Farm Research Program


Identification of
Policy Issues


National Goals, supply of
inputs, credit, markets,

Choice of Target Farmers
and Research Priorities

Identification of Problems
for Station Research



1. Plan
Obtain a knowledge and
understanding of farmer
circumstances and problems
to plan experiments.

2. Experiment
Conduct experiments in
farmers' fields to formulate
improved technologies 4<-
under farmers' conditions.

3. Recommend
Analyze experimental
results in light of farmer
circumstances to formulate
farmer recommendations.

4. Promote
Demonstrate improved
technologies to farmers.*--

5. Monitor
Determine farmers'
experience with

New Components Incorpo-
rated in On-Farm Research

factors in developing technologies-experiment
station research which emphasizes the develop-
ment of new technological components such as
new varieties, and policy which sets much of the
economic environment such as national goals,
input prices and supply, product markets and
infrastructure in which researchers and farmers
make decisions.

1.1 On-Farm Research
Agronomic conditions on experiment sta-
tions, with intensive management practices and
locations that are not representative of the area,
are often quite different from those in farmers'
fields. Through on-farm research-research con-
ducted in farmers' fields with the participation of
farmers,-technologies can be formulated and
evaluated under farmers' conditions. Technologies

can also be evaluated at many locations by experi-
mentation in farmers' fields. Moreover on-farm
research is an important means of increasing
communication between researchers and farmers.
Various activities or stages of on-farm
research are indicated in Figure 1.1. In the
planning stage, research teams, usually consisting
of an agronomist and an economist, try to describe
and understand farmer circumstances. This infor-
mation is used to identify priority technological
components which have potential to increase
production (or reduce costs) and which are
consistent with the circumstances of target groups
of farmers. Of course it is often easy to identify
many technological components but the essential
task at. this stage is to identify priorities since
research resources are limited and farmers, due to
scarce capital and risk avoidance usually have a
limited capacity to absorb large changes in tech-

Developing and screening
new technological compo-
nents (e.g. varieties, new
herbicides, pesticides)




nologies at one time. These priority components
are then the focus of the experimentation stage of
the research program which aims to formulate
improved technologies that is,- to construct from
known technological components, technologies
that improve upon farmers' existing technologies.
These experiments are conducted in farmers' fields
so that technology is formulated under conditions
similar to those in which farmers will use them.
Technologies are then recommended to farmers
after careful testing against farmers' technologies
in many locations and after economic analysis of
the results, using procedures described in a previ-
ous manual "From Agronomic Data to Farmer
Recommendations"' .
The final phases of the on-farm research are
to promote the recommendations and monitor
farmers' experiences. Promotion of the technology
is generally the role of extension although in this
process the design of demonstration plots logically
flows from the research experiments. Monitoring
farmers' reaction to the technologies when they
themselves pay the cost of inputs and bear the
risks is an important feedback mechanism to the
research process. If farmers are accepting the
technology, researchers can turn to other problems.
If they are rejecting or substantially modifying
the technology, then an understanding of why
farmers modify or reject the technology might
lead to a change in recommendations and even
to changes in the experiments.
This on-farm research process is essentially
dynamic as information is accumulated about
farmer circumstances, about performance of
various technologies in experiments and about
farmers' experiences with the technologies.
Over time some problems might be solved (or
discarded because of a lack of solution) and new
problems added. The system provides for contin-
uing improvement in technologies as researchers
apply information gained from past research cycles
to plan future research.

1.2 Experiment Station Research
With a strong on-farm research program, re-
search on experiment stations is primarily aimed
at developing new technological components
which require more controlled conditions such as
the development of new varieties. Also,experiment
station research can be used to screen technologi-
cal components that might have undersirable
effects on farmers' fields, such as herbicides that

might leave residuals. Promising technological
components arising out of experiment station
research are further refined and evaluated in on-
farm experiments for their appropriateness to
The flow of information between on-farm
research and research stations is two-way.
Information generated by on-farm research is
important for guiding experiment station research.
For example, information on farmer circumstances
and from on-farm experiments may provide guid-
ance on the type of variety that performs well
under farmer conditions and that conforms
to farmer preferences for maturity, yield, tastes
and storage quality.

1.3 Policy Context of Agricultural Research
Referring back to Figure 1.1 we see that
another important influence on agricultural re-
search are the policies which shape the economic
environment in which researchers and farmers
make decisions. Policies here refer to actions and
rules of governments implemented in order to
meet regional or national development goals.
Many policies influence the production
decisions df farmers. Some policies affect these
decisions directly, such as the policy to make
available to farmers only compound fertilizers
and not single nutrient fertilizers. Most policies
influence farmer behavior indirectly through their
effects on prices for inputs (e.g. through subsidies)
or for products (e.g. through marketing boards).
These influences of policy on farmers' decision
making in turn have implications for agricultural
research. In countries where herbicides are expen-
sive or difficult to obtain, researchers might orient
research on weed control problems differently
from that in a country where herbicides are cheap
and available.
Policies may also influence research decisions
directly. For example, many governments express
the desire to make the distribution of real income
more equal. This might influence the orientation
of research programs toward poorer regions and
farmers if most of the poor are in agriculture or
toward regions with high production potential if
most of the poor are in urban areas. In fact,
most countries have many geographical regions
and insufficient research resources to initiate
research programs in all regions. Measuring charac-
teristics of regions against national priorities such
as increased production and income distribution is

1/ See Perrin et al, From Agronomic Data to Farmer Recommendations; An Economics Training Manual, CIMMYT,1976.

one factor affecting the choice of target farmers
for a research program.
Agricultural research, and particularly on-
farm research programs, can also provide valuable
information to the policy maker that might
encourage a change in policies to facilitate the
introduction of improved technologies to farmers.
For example, on-farm experiments may demon-
strate the superiority of a given input which is not
available to farmers because of import restrictions.
The information about farmer circumstances
that is gathered in the course of on-farm research
may have other important implications for policy.
In some cases there may be important discrepan-
cies between stated policy goals and policy imple-
mentation and farmers are in a unique position to
identify where such discrepancies occur. For
example, in one maizeproducing area a stated
policy was to provide inputs and credit at low
interest rates to all farmers. However, discussions
with farmers revealed that because of considerable
paperwork, late arrival of credit and high costs for
inputs provided by the credit agency, it was not
in the interest of the farmers to seek credit and
inputs from this agency. In another case it was
found that farmers had to pay a considerable
premium on the black market in order to obtain

fertilizer despite the existence of an official
fertilizer subsidy.

1.4 The Place of this Manual in the Overall
Research Procedures
This chapter has described a general set of
research procedures in which farmers play a key
role. This Manual focuses on the planning stage
of on-farm research during which knowledge
and understanding of farmer circumstances is
obtained, farmers' problems are identified and
potential technological components to solve
these problems are narrowed to a few priority
components for on-farm experiments. In this
process, information that is useful to guide
experiment station research and policy analysis
is also obtained. This planning stage is part of
an on-farm research program, Which in turn is
part of a broader program of agricultural research
and policy analysis needed to improve production
and incomes of farmers. We believe that this
critical stage of explicitly considering the farmer
as the primary client in agricultural research
decisions provides an essential input into the
organization and effectiveness of agricultural
research programs.



1. Hildebrand, Peter "Generating Technology for Traditional Farmers: A Multidisciplinary Approach";
Institute de Ciencia y Tecnologfa Agricolas, Guatemala, C.A., December, 1976.

(Describes methodology and implementation of an on-farm research approach in the Guatemalan
Agricultural Research Institute)

2. Dillon, John L. et al "Farming Systems Research at the International Agricultural Research
Centers" TAC/CGIAR, World Bank, Washington, 1978.

(Reviews procedures for farming systems research at various national and international research

3. CIMMYT "Summary Report Wheat Training On-Farm Activities, 1978" and CIMMYT "Maize
Training: Report of Off-Station Experiments 1977 and Summary of Selected Experiments, 1973-

(Contain results of experiments in farmers fields conducted by CIMMYTs wheat and maize training

4. CIMMYT "The Puebla Project: Seven Years of Experience 1967-1973", El Batan, Mexico, 1974.

(A small farmer maize production project using technologies developed from on-farm experiments)



In the introduction to this Manual we said
that successful research begins with the farmer-
that is, planning research must explicitly take
into account the circumstances of farmers for
whom the technology is intended, in this chapter
we define further what we mean by farmer
circumstances and then show how information on
farmer circumstances can be used in planning

2.1 Definition of Farmer Circumstances
Farmer circumstances in this Manual are
defined as those set of factors that affect farmers'
decisions with respect to the use of crop technol-
ogies (in our case wheat or maize). Expressed this
way, farmer circumstances explain both a farmer's
current technology as well as his decisions about
changes in that technology. Various farmer cir-
cumstances are shown in Figure 2.1. They include
natural and socio-economic circumstances. Socio-
economic circumstances can be further divided
into those that are internal to the farmer and over
which he has some control (e.g. his goals and
resources) and those which condition his external
economic environment (e.g. markets).
Almost all farmers have a goal of increasing
income, broadly defined to include production
for home consumption. Generally too, small
farmers have a security goal of meeting subsistence
requirements of their preferred foods. They
generally also want to avoid taking risks that
might endanger their subsistence or cash sources
of income.
Farmers have relatively fixed endowments
of resources of land, family labor and capital
which thay can allocate to meet these goals.
(Capital resources here include both durable
equipment and cash availability). Farmers may
allocate these resources to different uses. Within
limits they may also adjust the amount of a
resource-for example, they may use some of
the cash resources to hire more land or labor

Many circumstances also define the economic
environment in which farmers make decisions.
These include the prices and price variability
for inputs and outputs, access to inputs and
product markets, land tenure systems, credit
facilities, physical infrastructure and so on.
While this economic environment is largely
outside of the control of a farmer, it is influenced
by many policy decisions such as distribution of
inputs, pricing policy and infrastructural devel-
opment. A large number of natural circumstances
also condition the farmer's decision making, such
as soil slope, depth, climate, weeds and pests.
The farmer generally makes decisions accep-
ting external, natural and economic factors such
as rainfall and prices as fixed, although he may be
able to modify their effects. For example, a
farmer may know that he has soils of different
fertility and decide to plant crops which meet
subsistence food preferences on his best soils to
meet his goal of food security. Many external
factors, particularly rainfall and prices, are
variable and unknown to the farmer when he
makes decisions. They provide an element of risk
to farmer decision making. In Figure 2.1 those
factors which are major sources of uncertainty are
marked with a dotted line. Risk may have impor-
tant effects on farmers' decision making. For ex-
ample, although a farmer may not be able to
predict rainfall he is aware of the degree of vari-
ability and therefore takes actions such as planting
a crop at several dates to avoid the risk of low
rainfall at a particular period in a crop cycle.
Most of these factors have direct effects on
farmers' decisions about a technology for a
given crop. Late season frosts might cause farmers
to seek an earlier variety to avoid risks. An expen-
sive herbicide encourages farmers to use a labor
intensive weeding method such as a hoe in place
of a herbicide. Many factors affect the choice of
a technology for the target crop because of
interactions in the farming system (again refer
to Figure 2.1). The farming system is here defined

Figure 2.1 Various Circumstances Affecting Farmers' Choice of a Crop Technology



Farmers' Goals-Income,
food preferences, risk.
Resource Constraints
-Land, labor, capital.


Markets Institutions
Product Land Tenure
Inputs Credit





Soils Type


- - PCircumstances which are often major sources of uncertainty for decision-making.

as the totality of production and consumption
decisions of the farm-household, including the
choice of crop, livestock and off-farm enterprises,
and food consumed by the household. For
example, a farmer may choose to plant maize
late because he is planting beans earlier, in order
to avoid disease problems in beans later in the
season. Or he may plant an early variety of
maize in order to have food early in the season
before other crops mature. Examples of interac-
tions in the farming system affecting choice of
a crop technology are many, and we will illustrate
these interactions throughout the Manual. The
point here is that crop technologies often result
from decisions made for the farming system as
a whole so that planning technologies for a
specific crop requires knowledge of important
interactions in the farming system which potential-
ly influence that crop. We shall refer to these as
system interactions.

In the same way that farmer circumstances
determine a current crop technology, they
also important in a farmers' decision to change his
technology. Conflict of a change in a technology
with anyone of the circumstances of farmers may
lead to rejection of that technology by the farmer.
High yielding Mexican bread wheats were rejected
in one country because they were not suited to
the soil conditions there. (Work is moving forward
to develop new varieties which can tolerate those
soils). Farmers in another area reject maize
varieties which mature too late for the planting of
the next crop. Farmers reject fertilizer recommen-
dations which maximize yields because these are
not consistent with either their income increasing
or risk avoiding objectives.
These examples have indicated to us clearly
that farmers reject technologies not because they
are conservative or ignorant, but because they
rationally weigh the changes in incomes and risks


Farming System Technology for
Target Crop
Cropping Pattern
Food Supply System Time, Method,
Labor Hiring, interactions / Amount for
etc. Various Practices.

-- hv

associated with these given technologies under
their natural and economic circumstances-and
correctly decide that for them the technology
does not pay. Our task then is to show how to
incorporate a knowledge of farmer circumstances
into the design of technologies so that they are
consistent with farmer circumstances-the subject
of the next section.

2.2 Decisions Required for Planning a Research
Researchers must make a series of decisions
in planning an on-farm experimental program.
First, they must decide what problems are going
to be investigated and which technological compo-
nents will be included in experiments to address
these problems. At the same time, the researchers
must determine if all farmers in the region are
sufficiently alike to allow a common set of experi-
ments and a common recommendation. If there
are significant differences among farmers, re-
searchers must somehow divide farmers into more
homogeneous groups and design experiments for
each group. For each technological component
included in experimentation, the levels, timing
and type of input or practice must be chosen.
Then for each set of experiments, researchers must
determine the levels of non-experimental variables
or those variables which are fixed for all treat-
ments in the experiments. Finally, the researchers
must choose sites on which to locate the experi-
ments. The circumstances of the farmers for
whom the technology is intended will be a key
factor in all of these decisions.

2.3 Identifying Farmers' Problems and
Prescreening Technological Components for
On-Farm Experiments
Farmers face many constraints which
directly limit production and incomes, such as
weeds, pests, diseases, inferior genotypes and
drought. Few research programs can investigate all
of these problems, so priorities must be established
to choose for research those few problems which
are most important in limiting farmers' production
and incomes and for which technological compo-
nents exist that promise immediate solutions to
these problems. For each important problem
there may be several technological components
available that contribute to its solution. For
example, a weed problem might be reduced by
changing rotations, time and method of land
preparation and planting or seeding rate, or
through improved manual weeding techniques
or use of a herbicide. In planning experiments it

is necessary to prescreen from these various
components those few "best bet" components
which have a high probability of success. Since
the final choice of components for on-farm
experiments must be compatible with farmer
circumstances, a knowledge of these circumstances
is essential not only to identify problems but
also to prescreen technological components.
Information on farmer circumstances also helps
define levels over which to experiment for the
technological component. If fertilizer is expensive,
rainfall. is variable, and farm size is small, the
relevant range of levels for on-farm fertilizer
trials will be lower than where each condition is
more favorable for fertilizer use.

2.4 Grouping Farmers into Recommendation
It is true that no two farmers have identical
circumstances and therefore identical needs for
technology. It is also true that a research program
cannot be established to provide recommendations
for each farmer. It is therefore necessary to
classify farmers with similar circumstances into
recommendation domains groups of farmers for
whom we can make more or less the same recom-
mendations. At least a tentative delineation of
these recommendation domains is necessary in
planning on-farm research, since the research
priorities and consequent experiments might be
different in each domain.
Clearly, the number of recommendation
domains depends on the amount of variation in
farmers' circumstances-the more variation the
more domains needed-and on the amount of
research resources-the more resources the more
domains we can afford. The final decision on the
number of domains will be a trade-off between
these two factors. However, it is well to remember
that the researcher need not seek precise recom-
mendations but general guidelines which the
farmer can adjust to his own circumstances. In
general, we need relatively few recommendation
Recommendation domains can be defined
on the basis of the various farmer circumstances.
They may be determined by variations in the
natural circumstances of the farmer such as
rainfall, soils or diseases. A given region may
contain many agro-climatic environments. These
are areas where a crop exhibits roughly the same
biological expression so that we would obtain, for
example, similar varietal or fertilizer responses,
everything else being equal. These agro-climatic
environments are, however, often modified by

socio-economic circumstances that produce
different recommendation domains. For example,
close to a large town, maize may be grown largely
for sale as fresh ears while further away it is a
subsistence grain. Such differences may impose
modifications on varietal selection and planting
date. More commonly, even if all locations are in
the same agroclimatic environment the resources
endowments of farmers may lead to different
technological needs. For example, small farmers
with scare capital relative to labor and who place
more emphasis on food security may follow quite
different cropping patterns and practices from
large farmers in the same agro-climatic environ-
At times a recommendation domain may
result from a complex interaction of agro-climatic
and socio-economic factors. For example, within
an agro-climatic environment for maize there
may be different disease incidences for beans
which cause farmers in one part of the agro-
climatic environment to plant beans early, there-
fore delaying maize plantings. In this case recom-
mendation domains may result from natural
circumstances (i.e. diseases) affecting bean
production and an economic circumstance (i.e.
labor scarcity) translating this effect onto maize
Both agro-climatic environments and recom-
mendation domains are not necessarily continuous
geographical areas. Fdr example, two neighboring
farmers may be in different recommendation
domains because of large differences in available
resources. Even within a farm there may be
different recommendation domains due to varia-
tion in soil type or topography.
It is clear then that a knowledge of farmer
circumstances and how they affect crop technol-
ogies will be a necessary element in defining these
recommendation domains.

2.5 Establishing Representative Practices and
Sites for On-Farm Experiments
One important reason for conducting experi-
ments in farmers' fields is to be able to formulate
technologies under farmers' conditions. Informa-
tion on farmers' practices helps design experiments
in which non-experimental variables reflect farmers'
conditions. For example, in a research program
emphasizing variety, fertilizer and weed control,
non-experimental variables such as time and

method of land preparation, seeding rate and
pest control should be maintained at farmers'
levels to reflect the results of variety, fertilizer
and weed control under farmers' conditions.* In
this case, of course, the chosen experimental
variables of the program variety, fertilizer and
weed control will reflect the most limiting factors.
Likewise it is important that sites for on-farm
experiments are representative of most farmers in
a recommendation domain with respect to soils,
crop rotations topography, location and farm
size. While it is often easier to choose sites that
avoid travel or are identified by cooperating exten-
sion personal these will not usually be representa-
tive of farmers in the area.
If farmers interplant maize and beans while
researchers do not, then weed control recom-
mendations arising from research may not be
appropriate for farmers, and in the absence of
effective weed control the profitability of fertilizer
recommendations can be markedly altered. In
one country most research on maize fertilization
was done on one type of soil. Investigation into
farmer circumstances showed that little maizo
was planted on this type of soil; virtually all of
it was planted on a soil where the recommendations
were entirely inappropriate.
With base practices at the levels of represen-
tative farmers, the researcher can be sure that his
recommendations are appropriate for farmers.
However, if new and profitable levels of the
experimental components are not identified, then
the researcher must make experimental variables
of some of those components he thought were of
lesser importance and which he had held at farmer
levels. In one country, efforts to formulate
appropriate new practices for wheat took the
planting date as practiced by farmers. The effort
was not notably successful until planting dates
were moved well forward (quite feasible for Those
farmers); then new technologies emerged which
increased profits and reduced risk to farmers.

2.6 Identifying Problems for Experiment Station
Research and Policy
So far we have emphasized using a knowledge
of farmer circumstances to guide on-farm experi-
mentation. But as we showed in Chapter 1, on-
farm research is closely linked to experiment
station research and to policy decisions. The
knowledge of farmer circumstances obtained in

* There is not universal agreement among researchers on appropriate levels of non-experimental variables in on-farm exper;-
ments. Some argue that these ; should be set high enough that the expression of experimental variables is not limited.

on-farm research therefore plays another role in
guiding these two activities.
One of the major activities conducted on
experiment station is the development of new
varieties. Knowledge of farmer circumstances is
important for identifying the priorities to be
attached to various breeding objectives. Do
farmers need earlier varieties to increase cropping
intensity or reduce late season weather risks? Do
they need varieties with specific insect or lodging
resistance? Or do they need to improve storability
because of difficulties in the marketing system?
The answers to these questions depend on the
circumstances of farmers for whom the variety is
Sometimes this information on farmer
circumstances will have to be quite detailed.
In one country farmers regularly strip the lower
leaves from their growing maize to feed animals.
Researchers had demonstrated that leaf stripping
reduced yields notably, hence had recommended
against the practice. Furthermore, the researchers
were working on new varieties of maize having
a more streamlined plant, shorter and with
fewer leaves but with less buffering to leaf strip-
ping. However, experiments conducted using
farmers' time and method of leaf stripping
showed that in fact existing varieties, with some
redundancy and buffering, permitted leaf strip-

ping with little effect on yields. With information
on the value of leaves and the real yield loss when
stripping is combined with the existing varieties,
researchers now have a measure of the amount by
which yields of grain must be increased if farmers
are to adopt new varieties which do not tolerate
Information on farmer circumstances also
helps identify policy problems which may impede
successful introduction of new technologies.
in one country, decision makers believed that
insecticides were easily available to all farmers.
However, information obtained from farmers
demonstrated that this was not the case at all;
some insecticides were available in one place,
some in another, and the distribution of insecti-
cides did not at all coincide with the distribution
of insects. This information demonstrated to
administrators the need to reexamine the input
distribution system. Often, information from
research will show policy makers the potential
benefits from changing policies. For example, if
fertilizer is in short supply, they may want to con-
duct some experiments to provide information to
policy makers on fertilizer response. These be-
come experiments for recommendations to
policy makers, not for recommendations to
farmers since farmers don't initially have access to
the input.



In this chapter we set out a "checklist" of
information on farmer circumstances which is
useful in planning experiments. We call it a check-
list because it provides a systematic way of
arraying the information we need and is therefore
a reference for Part II of the Manual on procedures
for obtaining this information. Of course, only a
part of this checklist of information will be
relevant in any given situation.
The checklist of information is organized in
three parts. First, we need information todescribe
farmer practices relating to the target crop. This
is a base for assesing potentially improved prac-
tices and particularly for designing on-farm exper-
iments that reflect farmer practices. Second, we
must diagnose those factors in the target crop
which limit farmers' production and incomes in
order to identify research priorities. Finally, we
have to understand those natural and economic
circumstances which influence farmers' decisions
with respect to existing practices in order to
prescreen technological components to fit these
Figure 3.1 shows the relationship between
these farmer practices, limiting factors, and natu-
ral and economic circumstances. On the right
hand side are farmers' management practices
which are determined by the various natural and
economic circumstances of farmers. On the left
hand side are the problems or factors which limit
production and income-that is, the proximate
causes such as weeds, fertility, or moisture-which
are a function of the natural circumstances of the
farmer. The effect of these limiting factors is
modified by the practices the farmer adopts. Here
- it is important to know the way in which various
practices affect these limiting factors. For example,
to what extent does a weed problem reflect
farmers' practices of rotations, land preparation
or timing and method of weeding.
These relationship can be illustrated in more
detail by Table 3.1 which shows the many
potential effects of farmer circumstances on
farmers' choice of a crop technology or farming
system. On the left are listed various management

practices for the target crop (in this case, maize)
and important farming system practices. On top
of the table are some of the various circumstances
that we have discussed in Chapter 2. The food
security goal is represented by the first two
columns. The income goal and the fact that
income can be increased by changing the
productivity of the various fixed resources are
shown in the next three columns. On the right
are the various external natural and economic
circumstances that create hazards or risks for
farmers. Some farmer circumstances such as to-
.pography, soil type, and tenure, input distribution,
etc. have been omitted from this table to keep it
from becoming too large.
A variety of farmer circumstances may
influence any one practice. Take the example of
the number of plantings of the crop in one season,
listed as practice No.9. It is asterisked against six
potential circumstances: (A) preferred food staple,
(B) food needs at specific times of the year, (D)
labor scarcity, (E) rainfall uncertainty, (G) floods,
(I) pests. Several plantings of a crop made over a
period in the rainy season may be a practice
influenced by several of these asterisked circum-
stances at the same time. In our example the
staggered planting of maize will prolong the
availability of green maize cobs for roasting-a
favorite food in many communities (influence
A). At the same time, a very early planting of an
area of maize, before the main crop, can give an
early harvest of new food at a time when stored
supplies from last year's harvest may be running
low (Influence B). Although a given time of
planting may give the highest yield per acre of
maize, the labor available to the farmer may
limit his ability to prepare seedbeds and establish
maize at that time (Influence D). By staggering
plantings of his maize crop over a two month
period he may be able to establish three times the
area he could establish by planting at the technical-
ly optimum time, and the increased area may
more than compensate for the decline in yields
per acre of the later planted crop. Furthermore,
in areas where rainfall uncertainty is a dominant

Figure 3.1 Relationship of Farmer Circumstances, Farmers' Management Practices and Factors Limiting

Natural Circumstances
(e.g. rainfall, weeds, diseases, insects,

modifying <-

Production Limiting Factors
(e.g. moisture, weeds, fertility,
diseases, insects)

Economic Circumstances
(e.g. food needs, resource constraints,

Farmer decision making

Farmers' Management Practices
(e.g. time of planting, weeding,

hazard, several plantings reduce the probability
of losing the whole crop (Influence E). If a period
of drought strikes when the first planting is at
flowering (a period of very high water demand),
losses from that planting will be heavy. Subse-
quent plantings, at earlier stages of maturity which
have lower water requirements will be less affected,
so risk of losses from rain failure will have been
reduced. Finally, floods, pest and disease attacks
may require replanting (Influences G, I). For any
given group of farmers only 'some of these poten-
tial relationships will be important. For example,
Table 3.1 shows that for one study area (denoted
by a circled asterisk) the need for food at certain
times and labor scarcity were major factors affect-
ing the number of plantings made.

3.1 Description of Farmers' Technology for the
Target Crop.
The first category of information is a
description of the management practices that
farmers follow for the target crop. Table 3.2 is
a checklist of the various management practices
for the target crop. It follows the sequence of

operations for producing the crop-land prepara-
tion, planting, thinning, weeding, fertilization-
through to post-harvest operations and seed
selection for the next crop. Not all operations are
relevant in a given situation, e.g. irrigation in a
rainfed production area. Note also that many
practices require very specific information, both
in the timing and sequence of operations. The
example of leaf stripping in the previous chapter
showed the need to be quite precise about the
farmers' method of leaf stripping. In another case,
the timing of irrigation in relation to the type of
nitrogen fertilizer and application methods was
important in understanding fertilizer efficiency
Practices for other crops and information
on the farming system will also be important
where they bear on practices on the target crop.
These are discussed in Section 3.8.

3.2 Identification of Limiting Factors
In order to choose technological components
which will be adopted by farmers, we must be



A-M ., .
gI *

Choice of soil type 1 *

Time of pl r ting 4 *
Method of planting 5
Spacing 6
Plant population 7
.Variety used 8
Number of plantings made 9
Intercropping 10
Relay cropping 11 *
Frequency and timing of weeding 12
Use of fertilizer 13
Method and time of harvest 14
Method of processing & storage 15
Use of herbicide 16
Use of insecticides 17


Growing preferred foods 18
Growing non-preferred foods 19
Crop rotation 20
Renting of land 21
Hire of labor and machines 22
Labour recriprocity 23
Winter land preparation 24 *
Firing or bush fallow 25

Farmer circumstances potentially influencing choice of a management practice

I ; riet r'J ms.i aijj overnin choice of practices in one study area
M M M.. ... i lI I Imm

able to diagnose the constraints on farmers' pro-
duction and income. We shall call these limiting
factors or problems. A major set of these factors
will be those that limit yield-that is, those
proximate factors such as a) genotype, b) fertility
and other soil-related problems such as salinity,
c) weeds, d) diseases and insects, e) stand
establishment or density, f) moisture and g)
rainfall, etc. Here we are only referring to
proximate causes of yield losses. Of course each
of these problems might be a manifestation of
more general problems; for example, disease is
often due to a susceptible variety and weeds to
scarcity of labor at the time of weeding. These
factors are explored later in relating farmers'
practice to their circumstances.
Increased production and income might also
be possible through more intensive cropping. For

example, an earlier variety of maize might enable
another crop (not necessarily maize) to be planted
in the same season. Finally, there may be other
ways in which farmers' income might be increased.
These include reduced storage losses (since most
small farmers store considerable portions of grain
for home consumption) and improved grain quality
(either for home consumption use or for sale).
Income might also increase by reducing the cost
of an operation such as where wages are high and
herbicides have potential to reduce the cost of
weed control.
In order to rank research priorities, it will
be necessary to assess the extent of the loss
associated with each limiting factor or constraint.
We will also need to assess the incidence of these
losses. Some factors may be relatively constant
from year to year (e.g., weeds or the potential for


Land Preparation
Sequence of operations
Timing of each operation in relation
to rains
Equipment used in each operation
Variation in method with seasonal

Variety(ies) used
Density and spacing
Density and spacing of interplanted crops
Time of planting in relation to rains,
frosts, etc.
Spread of planting dates
Sequence of interplanting crops
Method of planting (hills, broadcast, etc.)
Method of covering seed
Practices of replanting part or whole

Target density
Use of thinnings

Number of weedings
Timing of each in relation to planting
Equipment used in weeding
Use of herbicides (type, rate, timing and
method of application)
Use of weeds

Type of fertilizer(s) including organic
Rate(s) of application
Number and timing of applications
Equipment used for application
Method of application (e.g. broadcast,
furrows, etc.)

Pest Control
Method of control (type, rate, equipment)
Timing of control

Method of irrigation
Frequency and timing of irrigation

Timing of harvest in relation to maturity
Method of harvesting
Use of leaves and tops for animals
Timing and method of picking leaves and tops
Use'of stalks

Method of threshing/shelling
Timing of threshing/shelling
Method and quantity stored
Disposal of produce (stored, sold, etc.)
Use of crop in local foods

Seed Selection
Time of selection
Criteria for selection
Special seed production or storage methods
Seed treatment

increased cropping intensity) while others may
lead to large losses in occasional years (e.g.
diseases or drought). These latter losses have an
additional cost since they increase farmers' risks.
Finally, in order to propose technological solutions
to a particular limitation we also must have specific
details of the problem,-e.g. type of weeds, type
of pests or specific nutrient deficiencies.

3.3 Natural Circumstances
Natural circumstances influence farmers'
decisions by imposing particular biological con-
straints on the crop (e.g. the pattern of rainfall
affects decisions on time of planting). Natural
circumstances-particularly weather-also create
an environment of uncertainty which risk averting
farmers must take into account.
Climate: Often the major climatic factor
affecting farmers' decisions is rainfall. The amount
and within-year distribution over time of rainfall
indicates the potential for the crop in question,
the length of the growing season and the potential
planting dates. Year to year variability in rainfall
indicates the level of risk faced by farmers and
months when this risk might present special man-
agement problems. In some cases too much rain
may be the critical problem during some months
or some years. Other climatic variables are also
potentially important. Early or late frosts may be
the limiting factor on the growing season and a
major risk to farmers. Often a combination of
climatic factors may be critical. For example, late
planting when rains are secure may increase the
risk of frost later in the season.
Soils and Topography: Differences in soils
and topographies affect farmers' management
practices. Varied topography or soil within a
farm are usually exploited by farmers. Valley
bottoms will often support a longer growing
season but may become waterlogged at the
height of the rainy season. Alternatively hillsides
may be less suitable in drier seasons or may create
particular problems of erosion management. Op-
portunities for mechanization of land preparation
or weeding are affected by soil texture and topog-
Pests and Diseases: The incidence of insect
and diseases will often be associated with climatic
variables (e.g. humidity in the case of wheat
rusts). As with climate, the variation in pest and
disease incidence across years may be impor-
tant in understanding risks faced by farmers.
Management practices of farmers often relate to
pest or disease problems. Farmers may follow
particular rotations to reduce the incidence of

these problems or may time their planting of the
crop so that climatic conditions are not so
favorable to the disease/pest. Problems with
storage pests might also lead to particular practices
such as selling immediately after harvest or
planting early to obtain an early source of food
supply. Land preparation practices are often
critical to incidence of weed problems.

3.4 External Socio-Economic Circumstances
Many socio-economic circumstances affect
the external environment in which farmers
make decisions. Here we consider those external
circumstances over which individual farmers
have little control.
Land Tenure: Land tenure affects incentives
for land improvement and use of new inputs.
Landlord/tenant systems may lead to adverse
soil management practices or to disincentives to
use a new input if the rental is paid as a share of
the harvest. Inheritance systems may lead to
fragmented landholding with adverse consequences
for mechanization. A knowledge of land tenure
systems in the area can therefore serve as a useful
background to understanding farmers' practices.
Land Settlement Patterns: Patterns of
settlement, whether individual homesteads or
villages, also have implications for understanding
management practices. Sometimes particular crops
that require intensive management or that play an
important role in people's diet are located closer
to the village.
Extension: The extension service is a possible
source of technical information for farmers. The
influence of extension on farmer practices will
depend on the amount of contact between
extension workers and local farmers and on the
relevance of the technical information to farmer
SCredit: Credit availability and cost will have
a bearing on farmers' purchases of cash inputs,
use of hired labor and selling/storage strategy
for the crop. The likely severity of farmers' cash
constraints and its implications for using cash
inputs are determined by the availability of
credit for agriculture from formal (banks) and
informal (money lenders) sources and the interest
rates of each. The effectiveness of bank credit
depends on the procedures for obtaining a bank
loan and for what purposes loans are given.
Physical Infrastructure: Physical infra-
structure affects access to input and product
markets. Transportation to market towns,
particularly in the wet season, affects product
marketing. Availability of irrigation and the

conditions under which tarmers receive water
(e.g. who owns the facilities and controls
distribution) will be important in understanding
irrigation practices.
Product Markets: It is important to
understand the market faced by the farmer,
both in the sale of his crop and the purchase of
food staples. This will affect farmers' storage and
selling strategies as well as risks associated with
cash incomes versus subsistence production.
Factors to consider here are the major marketing
channels for the crop in question, the seasonal
and annual variations in price levels over recent
years, the spread between the producer price
and consumer price for the product, government
price guarantees and availability of milling facili-
ties for processing subsistence consumption.
Labor and Machinery Market: Information
is needed on the local labor market, such as the
available farm labor force from local sources (e.g.
landless workers), competing labor opportunities
(e.g. industrial jobs) and important streams of
seasonal migration into or out of the area. In many
areas labor resources can be supplemented by
hiring machinery, depending on the availability
and cost of machinery services. The availability of
hired labor and outside job opportunities will help
identify farmer labor constraints and alternative
employment opportunities, which in turn will
have a bearing on practices such as planting and
weeding, where timing and amount of labor are
often critical.
Input Markets: Information on the various
distribution channels for farm inputs, on prices,
price trends and availability of key inputs is
important in understanding farmers' use of
inputs and in designing technologies using
purchased inputs.

3.5 Resource Constraints and Goals of Farmers
Land: Land resources available to farmers
influence such practices as type of crop rotation
(e.g. length of fallow), soil management practices
(e.g. use of organic manures) and use of machinery.
Measures of land scarcity are the intensity of land
cropped in a given year and trends in price or
rental value of land in the area. In areas where
land is becoming very scarce, research may be
needed on fertility, water management, crop
rotations and multiple cropping.
it is important to know the system of
rotation followed including amount and length of
fallow (in either shifting or permanent cultiva-
tion) and sequences of rotations of specific crops
in the farming system. It is often useful to relate

variations in these patterns to differing populations
pressures, pest-disease incidence, topography or
soil type.
Cash: For most small farmers cash is a
constraint on using new inputs (at least at some
period of the year); farmers' actions often reflect
cash constraints. Cash constraints may be reflected
in practices such as selling home-produced food
soon after harvest at low prices and then
repurchasing food at higher prices at a later
date. Other effects of cash constraints may be
working off-farm at periods when there is a labor
shortage on his own farm or taking loans in the
informal credit market at certain times of the
year on relatively unfavorable terms. Identifying
behavior such as the above can help to establish
both the magnitude and timing of cash constraints.
The nature and timing of the cash constraint
is best seen through a calendar of cash flows that
indicates seasonal inflows due to farm sales.and
other sources of cash income (e.g. off-farm
employment) and seasonal outflows, such as
input purchases and other necessary expenditures-
such as food purchases and school fees.
Family Labor: Family labor is one of the
major inputsfor small farmers. Seasonal shortages
of labor may have major impacts on farmers'
practices. This can be gauged by determining
first, the busiest periods of the year and the
type of work done during these periods, and
second, those periods and type of work for which
farmers hire labor. This alerts us to look for
practices such as staggered plantings or problems
such as weeds, related to labor shortages.
Capital: Labor shortages can be overcome
to some extent by equipment and it is therefore
necessary to know what power sources/equipment
farmers have available.

3.6 Farmers Goals
A primary goal of farmers is to increase
incomes. This is achieved through increased
productivity of the resources-land and capital-
discussed above. Farmers' income goals are how-
ever strongly conditioned by food preferences
and risk aversion.
Food Consumption and Preferences: If the
crop of interest is an important part of home
consumption, it may be necessary to know
something about seasonal food supplies and
food processing and consumption patterns and
preferences. These may influence farmers cropping
patterns, choice of variety, planting dates and
storage and marketing strategies. Farmers often
grow security crops to act as substitutes for the

preferred foods. There are often differences in
varieties with respect to suitability for local
processing methods and local tastes. If food has
to be purchased, then the cash requirements
might lead to other practices and problems, such
as untimely or inadequate weeding due to insuffi-
cient cash to hire labor.
Risks and Risk Management: In most areas,
an understanding of farmer practices requires an
understanding of the overall risk situation of the
farmer and what management strategies can be
used in the face of those risks. Uncertainty
arises from both the natural and economic
circumstances of farmers.
For the major crops in the system, we need
to know the frequency and causes of crop failure
and the severity of each in terms of food and cash
needs. It is also necessary to know the specific
nature of the problem. If the failure is due to
rainfall, we need to know if the problem was
caused by a late start to the rains, early finish
or a mid-season dry period? If it is due to a pest
or disease, we need to know the timing of the
problem and the conditions under which it is
most prevalent. For each problem farmers may
follow insurance practices to reduce the risk.
For example, farmers may stagger planting to
reduce the effect of rainfall unreliability. They
may follow particular rotations to conserve
moisture or reduce pest problems. Finally farmers
may take specific measures when the problem
occurs, e.g. spraying for an insect problem or
replanting with a short season crop.when early
rains fail.
Uncertainty in product markets also affects
management practices. Variability in prices may
lead to insurance strategies such as crop diversi-
fication or storage.

3.7 Farming System Interactions
Many of the influences of farmer circum-
stances on management practices discussed in this
chapter are direct influences (e.g. rainfall affecting
time of planting) but many are influences oper-
ating through interactions in the farming system.
These interactions are often overlooked.
-Some farming system interactions are direct
interactions between enterprises. The most
common example is interactions between livestock
and crop enterprises. Crop enterprises often
provide forage .for livestock. In one area farmers
planted 110,000 maize plants per hectare and
thinned almost one third of them for livestock
feed. In another area, farmers planted a maize
variety which was developed for higher altitudes

in place of the recommended variety. In this
case the variety farmers chose to plant was more
suitable for leaf stripping to-provide forage for
animals. It is also a common practice for farmers
in dryland wheat producing areas to leave weedy
fallow for livestock feed; this has adverse effects
on moisture availability for the following wheat
crop. If animals are also used for land preparation
then availability of forage often affects the avail-
ability and strength of traction animals, with
important consequences for method and timing
of land preparation.
Many farming system interactions also occur
through competition for scarce resources. In
areas where more than one crop is planted per
year, crops often compete over time for the same
land. In one irrigated area wheat is planted
several weeks after the optimum time to obtain
highest yields because farmers have to wait for
their cotton to mature. In many areas, enterprises
compete at certain times of the year for scarce
labor and/or cash resources (since cash resources
can be used to hire labor). In such cases farmers
may reject such labor intensive practices as
thinning or they may delay such operations as
weeding beyond a time that would maximize
Finally farming system interactions occur
through efforts of farmers to manage available
resources to meet goals of supplies of preferred
foods and risk avoidance. Managing food supplies
may result in practices such as early planting or
use of an early variety. Risk avoidance is expressed
through crop diversification and planting of crops
which are less risky although perhaps less preferred
as foods and/or less profitable.
The importance of some of these interactions
in explaining farmers' practices for a given crop
means we may require detailed information on
the farming system. For example, if crop-livestock
interactions are important, we may need to gather
information on the livestock sector to learn about
seasonal forage sources, markets for forages, etc.
Or if labor is a constraint at a given period, we may
need to focus on operations in other crops at that
time, in order to understand why these operations
are performed when they are. We may also need
to investigate the supply of hired labor and sea-
sonal migration.
An understanding of the effect of these
interactions on current practices will of course also
be important in prescreening new technological
components. It might mean that we have to
carefully evaluate the effects of new components

on forage availability or labor requirements at explore each of these system interactions, but we
certain periods of the year. It is not possible to should be alert to their potential importance and
provide a checklist of the information needed to focus our efforts accordingly.



1. Cleave, John H. "Decision-Making on the African Farm". In Contributed Papers Read at the 16th
International Conference of Agricultural Economists. University of Oxford, 1979.

(A good review of the many factors or circumstances influencing farmers' decision-making)

2. Collinson, M.C. "Demonstration of an Interdisciplinary Approach to Planning Adaptive Agricul-
tural Research Programs; Serenje District, Zambia"; CIMMYT East Africa Report No. 3, Nairobi.

(Traces the many influences on farmers' management practices on maize in one study area of



In Part I, we have established the vital
importance of knowing and understanding
farmer circumstances for planning agricultural
research to develop technologies appropriate to
farmers. We now ask the question: how can we
obtain this information cheaply and efficiently?
Part II describes a set of procedures that we have

found -useful in obtaining this information.
However, this is not intended as a recipe since a
recipe presumes a given situation. Rather we want
to provide a set of guidelines and principles to
help researchers make decisions for their specific



4.1 Sources of Information on Farmer
Information on farmer circumstances can be
obtained in several ways. First, there are secondary
sources such as published census data or unpub-
lished rainfall data. Second, there is information
obtained by interviewing farmers. These interviews
may be conducted relatively informally by conver-
sations with farmers or more formally by inter-
viewing the farmer using a written questionnaire.
Also the information may be obtained in one
interview or in a series of interviews with a farmer.
Third, information can be obtained by direct
observations by researchers in farmers' fields.
Secondary sources of information should be
used when they are available and reliable. However,
rarely will there be adequate data about farmer
circumstances-their farming systems, resource use,
problems and constraints-for planning research.
A large part of an information-gathering effort
will have to be devoted to obtaining the informa-
tion directly from the farmer and his fields.
Informal interviews with farmers and other
persons knowledgeable of farmer circumstances

are particularly valuable when conducted by the
researchers themselves. They place the researchers
in direct contact with the farmer. Because the
researchers are free to structure the interview
depending on the responses of each farmer, they
can quickly learn and understand the farming
system and cropping practices. Also the informal
setting of the interview often makes it easier to
obtain sensitive or complicated information.
The formal interview, conducted with a
questionnaire, has the advantage of obtaining
a standard set of information from each farmer.
Because a fixed questionnaire is used, well-trained
intermediaries or interviewers may be used to
conduct interviews thereby making it possible
to obtain information from a large number of
farmers. Furthermore, farmers can be chosen
randomly to provide a representative sample of
farmers. In this way the researchers can obtain
a more complete picture of farmer circumstances
in the area. They are also able to communicate
these results to others quantitatively. For example,
they may be able to report to policy makers
responsible for fertilizer distribution that a



1. Assembling Background

2. Exploratory

3. Formal Survey



Interviews (informal) by researchers
with farmers, input suppliers,
merchants, bankers, etc.
Field observations on crop.

Interviews (formal) by trained
interviewers with farmers. May
include field observations.



Initially broad but
narrowing as survey

Usually specific and well
focused on important
information defined in the
exploratory survey.

specific percentage of farmers received fertilizer
Finally, direct field observation by experi-
enced agronomists is valuable in identifying proxi-
mate factors limiting production in farmers' fields.
However, direct observation will rarely be suffi-
cient for this purpose because observed problems
will be specific to the stage in the crop growth
cycle and the seasonal conditions in which the
observations are made.

4.2 A Sequence of Steps
Figure 4.1 illustrates a sequence of steps
employing these various methods of obtaining
information on farmer circumstances. First the
researchers assemble and analyze background
data from secondary sources and note those
factors which will guide questioning of farmers
(e.g. months of low rainfall suggest high risks to
farmers or areas without roads suggest serious
product marketing problems). The assembly of
background information is followed by an
exploratory survey in which the researcher, the
agronomist and the economist, work together as
a team. They traverse the region to informally
observe farmers' fields and interview farmers and
other persons, such as merchant or extension
agents, familiar with farming in the area. No
questionnaire is used, but the interviews are more
than a casual conversation since the researchers
have a systematic checklist of information to be
obtained and hypotheses to be tested in the
interviews. During this exploratory survey, the

researchers also tentatively identify recommenda-
tion domains.
The exploratory survey is used to give
researchers a first hand knowledge of farmers
and their problems and to provide a base for
organizing a formal survey using trained inter-
viewers to administer a questionnaire (and some-
times field observations) to a randomly chosen
group of farmers. We emphasize that this formal
survey should be conducted only after a thorough
exploratory survey. The exploratory survey is
necessary to decide what is the important infor-
mation that should be collected in the formal
survey, how the questions should be asked and
how to choose a representative sample of farmers.
(See Chapter 6).
There are many types of formal surveys
ranging from statistical surveys involving quick
visits to a large number of farmers to in-depth
studies on a small number of farmers through
periodic visits. For the purposes of planning
agricultural research, a single-interview, sometimes
supplemented by field observations, of a relatively
small number of farmers is usually sufficient.
Furthermore, the questions will focus on details
of farmers' management practices and will be
quite different from many standard surveys for
cost of production and agricultural statistics.
The information from all these various
sources must then be assembled in a form suitable
for planning research. Critical factors limiting
production are listed, possible technological
solutions are identified, farmers' production prac-


Year 1 Months Year 2 Months
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12

First Crop Cycle a Second Crop Cycle

Gathering Secondary Data on

Exploratory Survey

Formal Survey
Analysing Survey Data

On-Farm Experiments

Reporting Results and
Planning Next Cycle's


-------- ---- -----------------------------

6 wks

8 wks b/

-- 8 wks

4 -a

20-25 wks
6 wks

8 wks

- P=Planting; H=Harvesting
- Dotted lines indicate that these activities may be less than full time for the period.

- Preparation activities include sampling of farmers', design of questionnaires, training, etc. for the formal survey and site selection,
assembly of materials etc. for experimentation.


tices to be used as a base for the experiments are
noted and policy-related problems are identified.

4.3 Implementing the Research Procedures
The process of obtaining information on
farmer circumstances should be part of an ongoing
process to plan and execute on-farm experiments.
The same research team-the agronomist and the
economist-will then be responsible for carrying
on each of these activities over time in one or
more target regions. The team should, however,
be available full-time for this program of surveys
and experiments. At times, it will be necessary
for the research team to include, at least on a
part-time basis, researchers from other specialities
depending on the type of problems identified.
For example, if insects are a particular problem,
an entomologist might participate in the field
work and research planning. Or the contribution
of an agricultural engineer would be valuable
where there are special problems of water or
machinery management.
The researchers will also need assistants both
for the survey work and experimentation work.
There are benefits from using the same assistants
in both phases of the work. These assistants
need not have high levels of education; rather,
their practical knowledge of agriculture and
their ability to work in the field and communicate
with farmers are more important attributes.
Chapter 9 describes in more detail some desirable
characteristics of assistants. Finally, the research
progranmwill require ready access to transportation
since the success of the program depends on fre-
quent field visits by the researchers through-out
a region.
In an ideal situation, the exploratory survey
is conducted during the crop cycle in order to
perform the important task of field observations
on the target crop. For both maize and wheat, the
time around flowering is a good time to observe
problems in farmers' fields. A typical sequence of
operations for an on-farm research program is
shown in Figure 4.2. The period after the explor-
atory survey is used to design the questionnaire
and train enumerators and prepare for the formal
survey. However, the informal dialogue between

researchers and farmers continues in a less inten-
sive way throughout the research program. The
formal survey is executed in the period after
harvest; a time when farmers will have fresh in
their minds the previous crop cycle and also a
time in most agricultural areas when farmers are
not very busy. This timing will, however, preclude
field observations in the formal survey. The data
from the formal survey is then analyzed to design
an experimentation program for the following

4.4 Adjusting the Procedures to Fit the
Researchers' Circumstances
Just as different technologies should be
designed to fit different circumstances of farmers,
so must the procedures used to obtain information
from farmers be specific to the circumstances of
each team of researchers. The circumstances of
researchers might depend on the resources avail-
able for the research and the types of problems to
be investigated. The process described above is
suggested as a model to be followed when a new
on-farm research program is being planned.
In practice, many variations in the sequence
of steps have been tried in several situations.
For example, the crop cycle in which the
surveys are conducted can be used to obtain some
preliminary experimental information. In this
case, experiments based on whatever information
is available are conducted at the same time as
information on farmer circumstances is collected.
Information from both the experiments and the
surveys are pooled to plan research for the next
cycle. In another situation, with few resources
and little time, it was decided to focus on
exploratory surveys in a few regions in order to
tentatively plan on-farm experiments and then
later conduct some short, well focused formal
surveys in those regions at the same time as the
first round of experiments. In yet another situation,
the researchers already had considerable secondary
data but lacked some critical information such
as timing of operations on the crop. In this case, a
small special purpose survey was designed to
obtain this missing information.



At this stage secondary data on the target
region is assembled and analyzed from diverse
sources. This provides useful background on the
region for beginning an exploratory survey. -

5.1 Sources of Secondary Information
Secondary information can be obtained
from government sources such as maps, regular and
ad hoc reports, and from other sources such as
reports of a research organization. Since many of
these sources will only be available in the national
or regional capital, the process of assembling sec-
ondary data may involve travel to these centers.
Here are some examples of these types of
Agro-Climatic Data: Monthly rainfall and
temperature data are usually available from indi-
vidual weather stations or from the national
weather service.
Topographic Data: Topographic maps of a
scale of about 1:50,000 are available from carton
graphic units in nearly every country. They are
extremely valuable in defining the area and in
sampling and conducting field operations.
Soils Data: Soil maps are often available
from soil survey units and help define variation
in soil type that might result in differences in
cropping patterns, incidence of drought or
waterlogging and fertility problems. Available
chemical and physical analyses of soil might
also help in decisions on fertilizer experiments-
particularly in the case of phosphorous and
Population Data: The latest population
census will provide data for local government
units-Jor villages. When urban areas are excluded,
these data provide a measure of population density
and the variation in land pressure in the area. This
alerts us to possible problems of declining soil
fertility or erosion.

Production Data: Agricultural census data
provides at a minimum, acreage and yield data for
major crops in each local government unit. Data
on variation in cropping patterns and yields across
the region will guide later questioning in the field.

Price and Market Data: Information on
quantity, prices and distribution of inputs,
production and credit can often be obtained from
reports of public and private agencies operating in
the region such as banks, seed production agencies
and marketing boards. Time-series data on pro-
duction prices might alert us to changes in crop-
ping patterns. Seasonal price data might indicate
particular storage or risk problems for farmers.
Research Data: Reports of previous research
conducted in the region are particularly valuable
since they often contain more detail and better
quality data than official censuses. Data from
farm-level surveys and previous on-farm experi-
ments will often be directly relevant to the task
of planning future research.

5.2"Analyzing Secondary Information
All secondary information should be analyzed
for variations across the region. At this stage it
should be possible to form some tentative
hypotheses about recommendation domains. if
climatic data shows sharp variations in rainfall
across the region, we might check the production
statistics to determine if this variation in rainfall
results in different cropping patterns or yields.
Data on factors which cause uncertainty in farmer
decision-making such as rainfall and prices should
also be analyzed for year to year variations;

'Example 5.1:
At one site, a histogram of monthly average
rainfall shown in Figure 5.1 indicated that January

1/ Local government units are collections of villages with the same local governments. They are variously called counties.
shires, district councils, municipios, chiefdoms, etc.

Figure 5.1 Average Monthly Rainfall Profile




rainfall was low. In this case, the researchers
subjectively estimated that given local soil and
temperature conditions, maize would need at
least 75 mm of rain in January to support a
January planting. Examination of historical
rainfall data showed that this only occurred in 10
out of 26 years, (i.e. four years in ten) indicating
that January planting is quite risky. Various
strategies are possible to reduce this risk such as
later planting of an early variety or planting of
sorghum in January followed by early maize. The
point is that crude analysis of rainfall data alerts
us to be on the lookout for some risk-avoiding
strategies of farmers.

Example 5.2:
In a highland maize-producing area where
some supplemental irrigation was available, rainfall
data and data on frost incidence were particularly
valuable in establishing the degree of risk faced by
farmers. Table 5.1 shows that maize planted in
September has the lowest frost risk but also has a
considerable drought risk at the time of planting.
In this case the data indicated that farmers with
supplemental irrigation to enable September
planting and avoid both rainfall and frost risks,
would likely use quite different practices than
farmers without irrigation who would have to
weigh drought and frost risks.



Probability of less Probability of 20C
Month than 30 mm rainfall frost exposure if
in month-j planted in month-2

o/o O/o
August 90 27
September 30 13
October 10 25
November 0 39

1/ Rainfall below 30 mm was estimated to be insufficient to support a maize planting.
2/ Probability of frost from the first month after planting to one month before harvest.



The exploratory survey is essential to ob-
taining information about farmer circumstances.
In some ways it is more important than the formal
survey since it places researchers in direct contact
with farmers and enables them to observe first-
hand the farmers' crop and cultural practices. The
exploratory survey must be initiated before the
formal survey since one of its principal objectives
is to help design this formal survey. However, the
type of researcher-farmer dialogue we recommend
here should be a continuous process through the
phase of experimentation.
The essential task of the exploratory survey
is to quickly gather information through informal
interviews with many people-and particularly
farmers-in order to arrive at a tentative descrip-
tion of farmer practices and problems limiting
production and an understanding of how circum-
stances influence farmers' practices. This informa-
tion is useful in tentatively defining recommenda-
tion domains and identifying potentially improved
One of the major objectives of the explor-
atory survey is to help design a well-focused
formal survey to verify and quantify what is
learned in the exploratory survey. The exploratory
survey helps in the design of the formal survey by:
a) identifying important topics bearing on
research planning that need to be included in
the formal survey;
b) ensuring that written questions in the
formal survey are asked in a way that can
be understood by farmers;
c) designing and testing a sampling scheme;
d) publicizing the forthcoming research program
including formal surveys and experiments.
Finally, the exploratory survey is used to
collect important information that may be too
sensitive or complicated to include in a formal
survey. With researchers themselves asking the
questions in informal conversations, more complex
dialogues can be pursued. Examples of such
information are: the way farmers reason about
particular problems, interest rates and borrowings
in the informal credit market, and crop sales.

6.1 The Exploratory Survey Process
The exploratory survey is a recursive learning
process of assembling information on farmer
circumstances against the checklist in Chapter 3,
evaluating the information obtained to determine
where further efforts should be focused, and then
returning to the field to obtain this information.
As the survey proceeds, the checklist is
narrowed by eliminating information that is not
relevant to understanding farmers' practices in the
area. Finally the list is narrowed to variables and
relationships which will be verified and quantified
in the formal survey.
The exploratory survey is conducted by the
researchers themselves-the agronomist and the
economist-working as a team. They traverse
the target region observing farmers' fields, and
interviewing farmers and other persons, such as
merchants, with specialized knowledge of agricul-
ture in the area. It is best done when the target
crop is in the ground so that problems can be
observed in the field. The amount of time spent
on the exploratory survey will vary from two to
three weeks depending on the size and complexity
of the region and the previous local knowledge of
the researchers. It is our experience that researchers
rarely spend sufficient time on this rewarding and
essential task.

6.2 Assembling Information in the Exploratory
The assembly of information in the explor-
atory survey follows the checklist given in Chapter
3 although not necessarily in that order. Informa-
tion on production practices of farmers is collected
at two levels. First, we want to know what are the
general practices of farmers in the area. Second,
for most practices we find it useful to explore
variations in the practices with respect to (a)
variation across farmers in the region, (b) variation
across years or seasons and (c) long-term trends.
in each case, trying to understand why such
variation occurs will help in understanding why
farmers use certain practices. Variations among
farmers help in defining recommendation domains.

Variations across years are important to assess
farmers' risk management strategies. Long-term
trends (i.e. those traditional practices that are
being discarded and those new practices that are
becoming common) help establish how farmers are
reacting to a changing external environment (such
as population pressure or market opportunities).
This description of farmer practices is, of
course, part of the effort to understand why
farmers follow certain practices. Researchers al-
ready will have some secondary data on some of
these influences which will guide the assembly of
information in the exploratory survey. For exam-
ple if it is known from the analysis of rainfall data
that rainfall is low and unreliable, then attention
will be focused on farmer practices to conserve
moisture and reduce risk.
Natural circumstances affect practices
through, defining the potential for the crop in
the region and through risk. In many situations
it is the risk element of natural circumstances
and how farmers react to this risk, which will
require emphasis in the exploratory survey.
On the economic circumstances, emphasis is
placed on identifying those circumstances which
create particular constraints and/or risks for
farmers. For example, what inputs are not avail-
able when needed; under what conditions is credit
available; or when do high prices and poor avail-
ability of the food staple occur. Of course, natural
and economic circumstances influencing farmers'
practices cannot be analyzed separately since
there will often be important interactions. The
example cited earlier in which farmers weed
maize late because they are planting beans is a
situation in which natural circumstances of
disease dictate the planting time of beans, while
economic circumstances of labor shortages lead to
late weeding of maize.
The guiding principle in assembling informa-
tion on farmer circumstances to explain farmers'
practices should be that if a significant number of
farmers in a region are using a particular practice,
farmers have a good reason for using this practice.
That is, farmers in choosing to use a given practice
are rationally reacting to elements of their natural
and economic circumstances and it is the challenge
to the researchers to uncover what are those
circumstances. The easy solution is, of course, to
assume that farmers use an apparently "bad"
practice because they are irrational, traditional or
ignorant; then we can ignore these practices in the
design of improved technology. Unfortunately, all
too often we will find that the farmer will not use
the "improved" practices because they conflict

with the very circumstances that we originally
failed to understand.
The bulk of the exploratory work will
consist of interviews with farmers. An effort
should be made to interview a broad cross-section
of farmers. Farmers who hold positions of
traditional leadership ofter have very useful
perceptions about the reasons behind traditional
practices and how these practices have changed
over time. Farmers identified by the extension
service will often have tried recommended
technologies and therefore will have information
and opinions about problems and potentials for
these technologies. So-called "innovative" farmers,
or farmers who have successfully developed their
own improved technologies will be valuable
sources of information on potential technologies
for farmers in the area. Finally, efforts should be
made to identify farmers who are roughly
representative of different recommendation
domains in the region.
Most of the interviews should be carried out
in the farmers' field in order to relate questions to
observations in the field. Interviews should be
conducted in a relaxed manner. Use of pencil
and paper should be avoided or restricted to
noting specific data such as quantities or names
of products and varieties. However, the researchers
should immediately note down all relevant
information after leaving the farmer.
Because of the informal nature of the survey,
little difficulty is usually experienced in gaining
farmers' cooperation as long as researchers are
respectful toward farmers. In the case of a longer
more structured interview, the purpose of
obtaining the information should be first
explained; otherwise the interview can be con-
ducted as a conversation with a passing farmer.
However, it will not be worthwhile in an explor-
atory survey to try to convince an uncooperative
farmer to be interviewed-rather another farmer
should be sought. (Chapter 9 provides more dis-
cussion on gaining cooperation of farmers).
Efforts should be made to identify the
primary decision maker in the household with
respect to a certain crop or practice. For example,
if women are responsible for weeding maize, then
it will be desirable to talk to women to discuss
weeding practices. In some cultures this may be
difficult if all members of the research team are
Interviews with farmers will range from a
casual conversation to in-depth interviews over a
broad range of topics. Clearly, it is possible to
cover only a part of the information listed in our

checklist in one interview with a farmer. What
information is included will depend on what
practices a farmer is following, problems he is
experiencing, and the degree of cooperation
encountered. For example, a farmer who is
experiencing difficulty in completing the quantity
and quality of weeding desired (observed in the
field visit) might be asked detailed questions
about the hired labor market, competing labor
demands in other crops, timing of operations,
etc. A farmer who is particularly cooperative
might provide information on several topics,
including sensitive information such as cash
flows and loans. In general it is useful to ask some
general questions about the target crop and the
farming system and then use the responses to
decide what specific areas will be emphasized.
It is not necessary to focus the questions
on the practices of a specific farmer. In fact,
much can be gained-particularly in interviews
with traditional leaders-by discussing practices
and variations commonly followed by farmers in
the area. For these types of questions, interviews
with groups of farmers can be particularly valuable
in gaining rough estimates of the frequency of use
of various practices.
Field observations are important in the
exploratory survey. Here the efforts of the re-
search team agronomist will be important. How-
ever, field observations need to be complemented
by questions to farmers. The interpretation of
problems in the field often depends on a knowl-
edge of farmer practices such as previous crop,
time of planting, variety and time and method of
weed control. Also farmers may see a problem
from a different point of view. In many areas
they recognize yield losses from weeds but also
value weeds as fodder. Furthermore, the problems
observed in a field at one point in time in one
crop cycle may occur only rarely. Questions can
be directed to farmers to determine other prob-
lems or the frequency of a given problem over
The farmers' knowledge and opinion of
current recommended practices are very useful
in identifying those critical circumstances that
have been most important in accepting or
rejecting components of these technologies. This
is particularly the case if farmers have tried a
practice and then rejected it. This is clear evidence
that the practice did not accord with farmer cir-
cumstances. Reasons why this was so, provide
valuable guides in designing more appropriate

In addition to farmers there will be many
other persons in the region who can provide
valuable information on specific aspects of farmer
circumstances and/or can help in implementing
the exploratory and formal surveys. Local govern-
ment officials should be contacted early in the
survey to ensure that they are familiar with the
scope and purposes of the research program. A
failure to inform any local institutions which has
influence with local farmers jeopardizes the success
of the program. Local officials can also be useful
in planning the formal survey. They may be able
to help develop lists of farmers or villages for
sampling, recruit local enumerators and find
accommodations for interviewers.
in any area there are a number of people
who are linked to farmers in the area and who
have specialized knowledge of some aspects of
local farmers' circumstances. These include (a)
agricultural extension agents, (b) government
marketing agents and private buyers, (c) input
suppliers, (d) machinery contractors, (e) bankers
and credit agentsand (f) land reform and irrigation
agencies. Marketing agents can provide information
on marketing channels, seasonal and annual price
variation and marketing margins. Extension agents
can provide experiences with recommended tech-
nologies. Input suppliers know the availability
and sales volumes of various products. Further-
more, they can be useful in delineating recom-
mendation domains or in establishing character-
istics which are very common or unusual in the
After each day's work, it is useful for the
researchers to evaluate what they have learned,
formulate new hypotheses and determine what
are the key gaps and conflicts in their understand-
ing which should be explored in further interviews.
The matching of the information obtained against
the checklist is valuable for this purpose. Discus-
sions of tentative findings and hypotheses with
local officials and extension agents often helps.

6.3 Integration and Evaluation of the Exploratory
Survey Data
As the exploratory survey continues, the
information obtained is integrated and evaluated
to guide further exploratory survey work and to
design the formal survey.
Description of Present Practices: The re-
searchers, as a result of their informal interviews,
should establish a list of management practices
Used for the target crop as well as for other crops
and activities in the farming system which impinge

on the target crop. They should note the appar-
ently widespread practices as well as those which
vary considerably in the area. They should attempt
to establish the characteristics of farmer circum-
stances which seem to be associated with the use
of a given practice. Toward the end of the explor-
atory survey, it should also be possible to give
approximate frequencies of use for a given practice
among the target population (e.g., 0-10 per cent,
10-25 per cent, 25-50 per cent, 50-75 per cent,
75-100 per cent of farmers).
Hypotheses to Explain Present Practices: An
important part of the exploratory survey is to
formulate hypotheses on reasons for farmers'
practices. In many cases several circumstances
may bear on a particular farmer practices. For
example, in one area farmers were found to
stagger their planting of maize, usually making
three plantings in a season. Three hypotheses for
this practice in tentative order of importance
were formulated: (a) a larger area can be planted as
labor is a limiting factor at the planting period,
(b) there is a dry period three months after the
start of the rains and late plantings may survive
this period better than plantings that flower at
that time, and (c) early plantings give an early
supply of new food and are particularly important
when the previous harvest is poor.
In another area with two maize crops per
year maize was observed to be harvested well past
the point where the grain was considered dry
enough for harvesting under normal conditions.
To account for this practice, it was hypothesized
that a labor shortage existed at the normal har-
vesting time because of the need to plant next
cycle's maize.
In both examples the tentative hypotheses
helps formulate questions for the formal survey
to test each hypothesis. In the first example,
questions were included in the formal survey on
labor availability at planting time, crop losses
for each planting due to drought, seasonal food
supplies and farmers response to a poor harvest.
Practices and priorities of other activities in
the farming system may also influence manage-
ment practices for the target crop. In one short-
season maize area with increasing population
pressure, it was found that farmers were often
planting late, well after the rains had started,
resulting in reduced yields. It was hypothesized
(in discussion with livestock officers of the area)
that because of reduced grazing area and feed
shortages at the end of the dry season, oxen
were weak at the preferred period of land
preparation and farmers were delaying ploughing

until the rains had softened the ground, to make
ploughing easier for the oxen. The formal survey
was designed to test the hypothesis by including
questions on trends in ownership of oxen and
forage availability.
Establishing Tentative Recommendation
Domains: The variation in farmer circumstances
in the region is a basis for identifying recom-
mendation domains. The hypotheses developed
above help to establish those natural and economic
circumstances that lead to variations in practices
among farmers in an area. A useful starting point
is to note major variations in current farmers'
practices (including cropping systems) in the
region. These variations are then related to the
circumstances hypothesized to influence the
particular practice (see Example 6.1). Recall, how-
ever, that we are only interested in variations that
might make a substantial difference in recom-
mendations to farmers. Where such variations exist
but are very gradual, the boundaries of recommen-
dation domains will be arbitrary. Rainfall varia-
tions are often quite gradual over a region so that
there will be no sharp distinction between wetter
and drier areas.

Example 6.1:
A South American highland maize-producing
valley was divided into four recommendation
domains shown in Table 6.1. Altitude, market loca-
tion, and availability of irrigation were major fac-
tors defining recommendation domains. As alti-
tude increased among irrigated farms, the vegeta-
tive cycle became longer and planting dates later
(to avoid early frost). Associated with these
changes, however, was a decrease in a leaf disease
problem of maize one of the factors limiting
yields. At the lower altitudes there was also a
small group of farmers near to a large town who
sold green corn resulting in somewhat different
practices. Finally there was a small group of
farmers who did not have irrigation and had to
plant later when rains were more reliable. The
management practices of this group of farmers
were less intensive because of. the high risks
in this case information obtained in the
exploratory survey on practices such as time of
planting, intensity of input use and selling of green
maize and problems such as leaf diseases could
be related to data obtained from secondary
sources on natural circumstances (altitude and
water availability) and economic circumstances
(location near a market) to provide well defined
recommendation domains.

Table 6.1: Example of Defining Recommendation Domains.
Recommendation Irrigated Main Planting Vegetative Disease Maize
Domain Altitude Rainfed Dates Cycle Incidence Disposal

I 2400-2600 Irrigated Aug-Oct 150 Very high Sold as green
II 2600-3000 Irrigated Sept-Nov 180 High Subsistence
III 3000-3500 Irrigated Oct-Nov 210 Mod. Subsistence
IV 2600-3500 Rainfed Nov-Dec 195 Mod. Subsistence
(300-600 mm) grain

Prescreening of Technological Components:
The prescreening of technological components
that can potentially solve identified problems
is discussed in detail in Part III where data from
both the exploratory and formal surveys are
utilized. However it is useful to begin the prescre-
ening process in the exploratory survey in order
to sharpen questions for the formal survey.
A list should be drawn up of factors limiting
production and incomes and of potential manage-
ment components to overcome these constraints.
Each new, or potentially new management
component is evaluated in the light of the
difference between it and the present practice.
Hypotheses are developed on the feasibility of
changing each management component from the
present to the new practice. This process yields
two sets of hypotheses: on reasons for present
practices and on the acceptability of the changed
practices (see Example 6.2). These can be tested
and verified in the formal survey.
Designing the Formal Survey: The explor-
atory survey work has as an important objective,
the design of a formal survey of farmers in the
region. Most importantly, as has been shown in
the examples above, it narrows down the data to
be collected in the formal survey to that which are
essential for understanding present practices and

prescreening technologies. In this way the ques-
tions in the formal survey are sharpened. The
exploratory survey work should also be used as a
vehicle for (a) designing a sampling frame, (b) pub-
licizing the formal survey and (c) determining
local terminology and measures, all of which con-
tribute to the success of the formal survey.

Example 6.2:
In one maize-growing area weeds were
identified as a major factor limiting production.
The exploratory survey established that the first
weeding was being conducted after weeds had
already done much of the damage. It was hypothe-
sized that a labor shortage and wet weather
conditions often prevented timely weedings.
Researchers felt that a pre-emergence herbicide
applied at the time of planting would offer
potential benefits to farmers. As a result, the
formal survey focused on information (see
Chapter 11) which would be needed to evaluate
the suitability of such a herbicide. Questions
were related to factors such as time of weeding,
labor required for weeding, use of hired labor,
intercropping practices and cash availability at



1. Hildebrand, Peter "Summary of the Sondeo Methodology used by ICTA"; Institute de Ciencia y
Tecnologfa Agrfcolas, Guatemala, C.A., June, 1979.

(Describes the methodology of an exploratory survey approach to understand farmer circumstances
in a situation in which no formal survey is used)



The purpose of the formal survey is to
quantify information and test hypotheses formu-
lated in the exploratory survey. Variations in
farmer practices in the region can be quantified
and hypotheses or reasons for the use of these
practices can be more formally tested. The
essential characteristic of the formal survey is
that a uniform set of data are obtained from a
relatively large number of farmers representative
of the region. This is achieved through use of a
written questionnaire discussed in this chapter,
and a random sample of farmers, discussed in the
next chapter.
We emphasized in the previous chapter that
the questionnaire is developed from the explor-
atory survey. There is no "standard" question-
naire for this type of survey but rather the
questionnaire is specific to a given region and
set of research objectives. Nonetheless, given the
general objectives for understanding farmer cir-
cumstances outlined in Chapter 2, most question-
naires will have some sections in common. For
example, to obtain information on representative
practices of farmers to be used as a basis for
on-farm experiments, the questionnaire will
normally include a section on the timing and
methods of farmers' practices-from land prepara-
tion to post-harvest operations-for the target crop
However, the specific information solicited will
vary from area to area. Surveys in irrigated areas,
for example, will include questions on water
In this chapter we provide some general
principles for developing the formal survey
questionnaire. Many examples of questions are
included to illustrate these principles. However,
once again we emphasize that each example
was developed for a specific situation and these
examples are not necessarily intended for general

7.1 General Rules for Developing Questionnaire
Organizing the Questionnaire:
Questions should be included in a sequence
that begins with specific questions on crop
practices which the farmer will find easy to

answer and proceeds to more sensitive and
difficult questions. Here is a suggested sequence:
(a) Screening questions to determine if the
farmer fits the requirements of the sample. For
example if the sample is restricted to producers
of the target crop, a question is included to find
out if that farmer grows the crop.
(b) Facts about management practices used on
the target crop, i.e. land preparation to post-
harvest operations including use of inputs.
(c) Facts about disposition of the target crop-
e.g. yields, marketing, storage and use of crop
(d) Opinions about specific management prac-
tices and the severity of hazards, problems and
constraints for the target crop.
(e) Important facts and opinions about the total
farming system which bear on the target crop-e.g.
labor bottlenecks, crop sequences and rotations,
livestock manure for crops, food preferences,
seasonal consumption patterns and cash flows.
These groups of questions should be orga-
nized into sections of the questionnaire in such a
way that the questionnaire has a logical flow.
There should be no need to frequently change
topics or to flip pages back and forth
during the interview. At the beginning of each new
section of the questionnaire, there should be a
sentence to be read by the interviewer to introduce
the topic. For example, the section on crop dis-
posal might be introduced as follows: "Now we
would like to talk about how you use the maize
that you produce".
Language of the Questionnaire: It is common
to find that the language spoken by farmers
differs from the official language of the country
or region. If this is the case the questions should
be asked in the local language by an interviewer
whose native tongue is that language (see Chapter
9). If the spoken language is widely written in the
area, the questionnaire should be written in that
language. Otherwise, the questionnaire should be
written in the common written language and
translated by the interviewer during the interview.
In both cases the translation should be thoroughly
checked, preferably by a senior researcher profi-

cient in both languages. In particular, questions
that solicit opinions have to be very carefully
translated to ensure that the meaning of the
question is correctly conveyed.
Length of the Questionnaire: The length of
the questionnaire depends on the objectives of
the survey and the complexity of the farming
system in the area of study. As a general rule,
the questionnaire should be completed in less
than 90 minutes to avoid fatigue on the part of
the farmer. In our experience, a thorough explor-
atory survey enables the design of a questionnaire
that can be completed in about one hour. It is
desirable to avoid trying to obtain information
for several objectives in the one survey. Rather
the quality of the information is improved if we
focus only on important information useful for
planning research.
The length of the interview can also be
reduced by subsampling the population for parts
of the questionnaire where information from the
total sample is not required. For example, in one
survey researchers felt they wanted more in-depth
information on two topics: marketing activities
and family' labor used in off-farm work. The
questionnaire was designed so that one half the
sample, chosen at random, was asked in detail
about off-farm work and the other half was
questioned on marketing activities.

7.2 General Guidelines for Asking Questions
There are several guidelines to keep in mind
in asking individual questions. The questions
should be written, as they are to be asked. (This
does not mean that the interviewer should read
the questions). For example, a section on fertil-
izer use might include (after establishing that the
farmer uses urea) a question such as:

"Can you tell us where you bought urea and the
price you paid for it last cycle?
Store Town_ Price /kg"

Factual questions should be specific to a par-
ticular crop season. For example ask, "Did you
apply fertilizer to wheat this year?" rather than,
"Do you use fertilizer on wheat?". This latter
question will tend to give a bias toward fertilizer
use, since the farmer will often answer positively
even if he only rarely uses fertilizer. After finding
out if a practice was used in a specific cycle, it is
sometimes useful to ask if that practice is normally
used depending on whether year to yearvariability
was noted in the exploratory survey.
Questions must be asked in a way that is

easy for the farmer to answer. For example,
allow the farmer to express his crop production
and area in local units rather than standard kg/ha
units. The conversion to standard units should be
made after the interview.
It is nearly always preferable to permit open
responses where the farmers answers in his own
words. However, preceded questionnaires are an
efficient way to record responses. An example
of a preceded open response is:

"What did you do with the crop residues after
Burnt it
Ploughed it in
Left it on top of the soil
Fed it to the animals _
Other (sp)

In this example the farmer is given an open
response question but several likely answers
(based on the exploratory survey) are provided
and the interviewer simply marks the appropriate
response thus saving time in writing. The preceded
question also ensures that the interviewer elicits
a specific answer. Note, however, that preceded
questions should include "Other (specify)" to
record unusual answers (e.g. the farmer sold the
Tables are a convenient way to ask sets of
similar questions to obtain factual information (see
Examples 7.1 and 7.2).

7.3 Guidelines to Obtaining Specific Types of
Some types of information are difficult to
record and will often be unreliable unless special
precautions are taken. These problems arise for
two reasons. First, the farmer may not know the
answer to the question because he can't remember
(e.g. amount of labor used for an operation) or
because he is not accustomed to quantifying the
variable in question (e.g. land area in some
regions). Second, the farmer may know but does
not give the correct information because the
question is not asked properly (e.g. "How long
did it take you to plant this field of wheat?" may
not elicit mandays of labor provided by helpers)
or because of the sensitive nature of the informa-
tion (e.g. data on loans, sales, etc).
In many cases these problems can be avoided
by careful questioning. In some cases it may be
.necessary to either omit this information or to
employ more costly methods if the information
is really needed. Some guides for obtaining specific

Example 7.1:
The table below was used to obtain data on
land preparation and sowing practices for wheat.
In this table the interviewer first asks "Did
you plough this wheat field this season?" and
circles the response11. If a ploughing was done,
he then asks "When did you do it?" and tries to
obtain data accurate to the week of the month.
He then proceeds to ask how it was ploughed and
fills in the information on power sources and

implements. Finally he obtains information on
ownership of the equipment and then continues
to the first harrowing.

Example 7.2:
Information on the farming system was
gathered through the table shown in Table 7.1.
The exploratory survey had established that choice
of variety and seedbed type were important in
managing labor scarcity, food supply and rainfall

(For the interviewer: Obtain the following information for the identified field for this season)

Operation Operation When Method Implement Is Animal or
Performed Used Tractor Rented/

Ploughing Y/N -1234 M/A/T R/O
First Harrowing Y/N 1234 M/A/T R/O
Second Harrowing Y/N 1234 M/A/T R/O
Third Harrowing Y/N 1234 M/AT R/O
Seeding Y/N 1234 M/A/T R/O
Covering Y/N 1234 M/A/T R/O
(sp ) 1234 M/A/T R/O

Y=Yes Put month & M=Manual Specify R=Rented
KEY N=No :ircle week of A=Animal Implement O=Owned
month T=Tractor

types of information are given in the following
Land inputs: In many traditional agricultural
areas, particularly in areas where the bush-fallow
system is used, farmers do not have a measure of
the land area (either in local or standard units). In
this case it will be desirable to measure the area
of the target crop at least for a subsample of
farmers since a good deal of other information
on input use and production must be expressed
on a per area basis to be meaningful. Methods
using a compass and tape are quite cheap although
time consuming. Further readings at the end of
this chapter provide references describing various
field measurement methods.
Labor inputs: Accurate labor input data are
almost always difficult to collect in one interview
because it is almost impossible for farmers to
recall precisely labor inputs for specific crop

operations. For the purposes of planning technol-
ogies, labor information may be needed for
specific crop operations affected by potential
technologies or for determining seasonal labor
constraints on family labor use.
All new technologies require some changes
in farmers' labor inputs. Fertilizers require labor
for application. Herbicides substitute for labor
for hand weeding. Increased yields require
additional labor for harvesting. Labor inputs for
specific operations are therefore required for
prescreening technologies for experimentation
and for making farmer recommendations on the
basis of the results of these experiments. Note
that this is not a cost of production study and
therefore only labor inputs effected by technol-
ogies being considered should be collected in
order to keep the questionnaire manageable.
An approximate estimate of specific labor

1/ This question relates to a specific field. The problem of interviewing farmers with many fields of the crop is discussed later.

__ L


(For each of the following crops grown by the farmer fill in details in the table below for the last crop cycle.
Take one crop at a time, maize first. Ask the farmer how many separate plantings he made, fill in details
for each planting one by one.)

Crop Planting Month Month Variety Type of Area in Month of Month of
(Check Started Comp!etcd Secdbed Acres Use from Crop
each Planting Planting M/R/F/I/ (approx) Field b Harvest

Maize* 3

Beans 3
Finger 1
Millet 2_
Sweet 1
Potato 2

Cassava 2 2
Old Cassava-/

a! M= Mounds, R= Ridge, F=Flat
I/ Note month when begin using new crop (e.g. green maize).
c/ For old cassava-give month and year planted.

- MM M m MMi M M M

Table 7.1

inputs can be made by carefully questioning the
farmer about the labor, per unit of area or better
for a given field, usually required for that opera-
tion. (If the field area is not known, it must be
measured too). In general this procedure will
lead to an underestimate of labor inputs. The
error can be reduced by ensuring that the farmer:
1) is familiar with the unit area or field referred
to; 2) includes all labor (himself, family, hired)
and particularly labor that is often overlooked
such as transporting inputs to the field, 3)
understands that the question refers to units of
adult labor; 4) understands clearly the operation
in question, e.g. is it labor for the first weeding
or all weedings?; 5) separates labor for the opera-
tion in question from other operations performed
simultaneously, e.g. fertilizer that is applied at
the same time as planting or weeding (see Example ,'
A second important aspect of labor use in
areas of relative labor scarcity is the existence of
periods of the year when family labor is fully
occupied and of other periods of slack labor.
Knowledge of when these periods occur is impor-
tant in promoting new technologies since it is
desirable to promote technologies which reduce
labor requirements at busy periods and utilize
more labor during slack periods. However it is
impossible to construct a detailed labor profile
Example 7.3:
In an area where farmers commonly grow
maize on the same field each season, the farmers
-were-asked questions specific to a field about
labor for hand weeding in a specific field. Inter-
viewers first asked how many men normally
participated in weeding that field (men in this
area did the weeding). They are then asked if
this number of persons worked a normal day,
how long would it take them to complete the
first weeding, second weeding, etc.1/ Responses
were recorded thus:

No. men normally No. days if all men
weed this field work full time

First Weeding

Second Weeding

Third Weeding
'ISPS^^S^.Z ESss1 ** = === ~ca.

for the whole farm household for the whole year
in one interview and a simplified approach is
suggested (see Exarimple 7.4). First the busy
period of the year is identified by asking farmers
the months when there is the most work to be
done/. Then the tasks must be identified that
have to be performed during those months, both
on the farm and off the farm. This will provide a
good picture of what operations the farmer would
have to reduce in the busy period in order to
adopt a technology which requires added labor at
this period. Information on what tasks the farmer
feels he has difficulty completing on time also
helps in identifying labor constraints.
Purchased inputs: It has been our experience
that data on inputs such as seed, chemicals and
equipment require special care. Some rules to
follow are:
a) be familiar with local units to be sure
that the quantities expressed in local units can be
converted to standard units. Also ensure that
actual and not recommended rates are reported;
b) check that the input was applied to the
total field and not a part of the field;
c) for chemical inputs ask the number of
applications and find out if the quantity applied
is for one or for all applications.
d) find out the type of input. Often in the
case of fertilizers and chemicals, it will be neces-
sary to look at the container.
Example 7.5 reproduces a part of a question-
naire used to obtain data on insecticide use in
Field Versus Crop Data: In areas where farmers
commonly have more than one field of a given
crop, a decision has to be made whether data
on crop management practices are to be collected
by crop or by field, and if by field for some or for
all fields, Again the exploratory survey must be
used to make this decision. if it appears that fields
of the crop are similar with respect to location,
topography, soils and rotations, and farmers are
applying the same practices to each field, then
information by crop will be satisfactory. On the
other hand, if fields differ physically or in man-
agement practices then data should not be collect-
ed by crop. Data can be collected by field but it is
time consuming to collect information for more
than two or three fields. In some cases, fields of
similar characteristics can be merged into groups.

In some cases men, women and children regularly participate in a task, and if for example, children are felt to be less
efficient in the task than adults, then it may be necessary to record labor inputs of each type of labor separately.
It is common to ask when farmers hire most labor in order to determine the busiest month. However, because of the
supply of labor or availability of cash, farmers may hire more labor at times other than the busiest month.

For example, during the exploratory survey in
one area farmers were found to be making two
separate plantings of maize-one for early green
maize consumption and the other primarily for
sale of grain. In the formal survey, information
was collected for each planting-regardless of the
number of fields used for each crop. Finally
there will be cases where farmers have many
fields and the differences between each field
are so marked that groups cannot be easily
defined. In these cases, information will have to
be collected for only one or two fields. These
fields can be chosen randomly from all the
farmers' fields but greater weight should be

given to larger fields, thus providing a better
representation of farmers' practices-l'.
Rotations / Multiple Cropping Patterns :
Farmers' desired rotation and cropping patterns
over -time often differ from what they actually
practice. For example, they may desire a wheat/
oats/fallow rotation but in practice they often
change because of weather conditions such as
drought or because of specific cash needs. There-
fore, in addition to asking general rotation
practices, it is often necessary to trace the history
of a specific plot to determine how land is being
used over time (see Example 7.6).

Example 7.4:
For a tropical maize growing area with two
crops of maize per year, the following questions
were asked:
In what months or period of the year do you
and your family have to work hardest?

Jan Feb Mar April. ay June
July_ Aug Sep Oct Nov- Dec-

Example 7.5:
In one tropical maize area it was found in
exploratory survey work that farmers applied
insecticides several times each cycle but often
using different insecticides and dosages on a
variable part of the crop. Information for each
application was collected in tabular form as


What work is done in these months/periods?


Crop and Task

Usually completed
on time?

1. 1. YN
1. 2. Y/N
3. 3. YI/N
1. 1. Y/Nt
2. 2. _2. Y/N-
3. 3. Y/Nf
1. 1. Y/N
3. 2. 2. YLIL.
3. 3. Y/N

* In this question the interviewer was asked to
combine two or more consecutive months into

L priods.


Name of






j/ Weighting can be done in the following manner. If a farmer has three plots of area-0.9 ha, 0.5 ha, 1.5 ha-then each plot
is assigned numbers as follows: Plot 1-0 to 9; Plot 2-10 to 14;and Plot 3-15 to 29; with the interval proportional to the
size of the plot. Another person (e.g. the farmer) is then asked to choose a number between 0 and 30 and the plot corre-
sponding to that number is chosen.


Number of

__ I __

Agronomic Data: Agronomic data on density,
intercropping practices, nutrient deficiencies, soil
type, germination problems, type of weeds and
incidence of diseases and insects are useful for
diagnosing proximate factors limiting production.
Generally, this information can best be obtained
by direct observation in the farmers' fields. If the
interview is conducted in the farmers' fields
during the growing season, then interviewers can
be taught to estimate variables such as soil color
and texture, percent damaged plants (with reasons
for damage), extent and type of weeds and (in the
case of maize) planting density (see Example 7.7).
The interviewer should of course ensure that the
observations are taken in a representative part of
the field.
Many agronomic problems occur in different
stages of crop development. In our experience,
field visits at the stage of flowering in maize or
wheat have been the most timely. However, even
here, there is usually a need to ask the farmer
supplementary questions about other stages of
the crop (e.g. leaf diseases which appear near
crop maturity or insect/disease problems of the
ear). The farmer will of course have a local name
for the insects, diseases and weeds which should
have been noted in the exploratory survey.
Finally, it is useful to supplement the field
observations by asking farmers whether particular
problems noted in the field are also common in
other seasons.

Example 7.6:
In a tropical maize growing area a complex
system of intercropping and relay cropping is
practiced. The following table elicited the three-
year history of one field (usually farmers only
planted one field to maize in a given season).
"For this maize field please list the crops
that you have planted in this field in the last
three years. Begin with the previous crop.


Date Planted Date Harvested
(month/year) (month/year)


Example 7.7:
The following agronomic data was collected
by interviewers in farmers' fields. Particular
emphasis was placed on training the interviewers
in choosing representative sites in the field and
identifying weeds, pests and insects:
"If the interview is conducted in or near
to the farmer's field, choose a representative
part of the field and take the following

1. Density: In the farmer's field, choose four re-
presentative points at random and take the follow-
ing data at each point-!. (Tape measures are pro-
vided for the purpose):


between 5

between 5

Number of
plants in 5


2. Weeds
a) Most important weed types
b) Percentage of ground covered with weeds

c) Height of the weeds
3. Insects
a) Insects currently present
b) Insect that were present and caused
c) Insect damage: Serious
Some Not important
4. Diseases
a) Diseases currently present
b) Disease damage: Serious
Some Not important
5. Soils: Color? _Texture?

6. Topography: Fat Sloping
Steep Hill top
River flood plane

7. Nutrient deficiencies observed:

7. Nutrient deficiencies observed: -----

8. Other observations:.

1/ Density in plants per hectare can then be calculated as (50,000 x plants in 5 hills)/(distance between 5 hills x distance
between 5 rows).

Production and Yield: Production and yield
figures given by farmers are often quite unreliable
particularly where (1) there are no standard units
of measure, (2) the crop is continuously consumed,
sold or used to pay labor over the harvest period,
or (3) taxation or local .customs makes this
sensitive information.
In surveys for the purpose of planning
research, accurate production and yield data
are usually not essential. In areas where farmers
are familiar with the concept of yield, direct
questions will provide an approximate yield.
In other cases rough estimates can be obtained
by carefully asking questions about various end
uses of production such as amount stored,
consumed, sold, given as labor payments, gifts,
etc. Yield estimates obtained through these
methods are useful to provide a guide to average
farmer yields and their variability. They will
not be suitable for trying to explain yields in
terms of the practices used by farmers. Where
accurate yield data is required, direct measurement
techniques are available and are described in
further readings listed at the end of the chapter.
Other uses of a crop are often forgotten-
for example, uses of weeds, crop residues and
leaves for animals. Sometimes it will not be
possible to quantify this output but it is important
to take note of such practices for the evaluation
of technologies which might affect these by-
Finally, experience shows that farmers
generally underestimate sales data since it concerns
cash inflows and is therefore considered sensitive
information. Because of this, data on subsistence
grain consumption, estimated as the difference
between production and sales, are correspondingly
inflated and should be treated cautiously.
Cash Flows: Information on cash flows is
difficult to obtain because of its complexity and
*sensitivity. However for the purpose of designing
technologies appropriate to farmers, it is often
useful to know the time and level at which cash
constraints occur in the cropping season. A
knowledge of the 'farming system obtained in
the exploratory or formal survey-what crops
are sold and when-will provide a rough guide.
More direct methods can be used as shown in

Example 7.8:
Cash flow was collected using the table in
Table 7.2. In this case, the data was collected at
the end of the interview when the farmer's

particularly where (1) there are no standard units
of measure, (2) the crop is continuously consumed,
sold or used to pay labor over the harvest period,
or (3) taxation or local customs makes this
sensitive information.

Subjective Data-Obtaining Farmers' Opinions:
Information on what farmers do is objective and
usually quantifiable. However, subjective data on
farmers' opinions and perceptions about problems
and technologies require different handling. The
identification of major problems perceived by
farmers is done in the exploratory survey. The
role of the questionnaire is to obtain estimates of
how widespread are those problems and opinions
and whether there are differences between groups
of farmers. Therefore it is not usually useful in a
formal survey to ask broad questions about the
problems in producing a given crop. The questions
should be much more specific. For example, what
are the best varieties with respect to yield, storage,
processing, drought tolerance, disease and insect
resistance, or what are the advantages of applying
fertilizer after planting. Farmers may also be asked
to express preferences about varieties. Do they
prefer earlier (later) varieties to higher yields and
why? Subjective questions should be asked in
a way that respects farmer's practices. Questions
such as "Why do you do intercropping?" are
asked from a biased viewpoint of researchers who
work only with monocultures. A preferred ques-
tion is "What would be the disadvantages to you
if you planted maize alone?".
Examples 7.9 and 7.10 show questions that
were successful in obtaining subjective data. Both
these examples depend on a thorough exploratory
survey in order to list the problems and opinions
that should be included in these questions.
7.4 Finalizing the Questionnaire
On the basis of a thorough exploratory
survey, a good questionnaire can be developed.
Still, it is always necessary to pre-test the question-
naire before producing a final version. This pre-
testing enables the researchers to determine
which questions are not easily understood by the
farmer and therefore should be redrafted; to
check the time required for completing the
questionnaire; and to test the sampling procedure.
The adequacy of the data in the questionnaire
should be pre-tested first in the office by filling
in two or three sets of plausible answers and
further by going to the field to interview five
or so farmers. Some or all of these farmers may
be selected to pre-test the sampling procedure as

well. The researchers must conduct or be present
at these interviews so that they can better note
the problems and determine the necessary
After the pre-testing, it is useful to tabulate

the responses using the methods of data tabulation
suggested in Chapter 10. This will be a test to
ensure that the data provided by the question-
naire satisfied our needs for information on
farmer circumstances.


CASH INCOME, SOURCE AND LEVELS: Now I have a few questions about where you earned cash
since the end of last year's (1977) rains. (Go through the sources first then deal with each source checked.

SOURCES Maize Bean G/Nut Bana- Fruit Other Chick- Eggs Fam- Govt. Shop Casual Pen- Other
Sold Sold Sold na vog. prod. ens Sold ily work Trade Labor sion
Sold Sold Sold Sold work.

Received Cash
Period Received

All through the year


Less than 10 -M-
10-30 -M-
30-50 -M-
50-100 -M-
100-200 -M-
200-300 -M-
300-400 -M-
More than 400 -M-
Amount unknown

Example 7.9:
The following series of questions was asked
about maize varieties: "What are the advantages
of Madero ("improved" variety) compared to
the local variety?"

Doesn't fall over with the winds
Early maturing
Better yield
Tastes better
Other (sp)

What are some problems with Madero compared
to the local variety?

Low yield Poor taste

Insect damage to grain
Doesn't resist drought
Insect damage to leaves
More weed problems
Other (sp)

(These questions were then supported by some comparisons between Madero and the local variety)

Madero Local No dif-
Variety ference
Which variety yields better when there is a drought?
Which variety yields better when there are strong winds?
Which variety matures earliest?
Which variety stores well for the longest period?
Which variety makes the best Irio (local maize food)

Example 7.10:
The importance and frequency of various
climatic hazards to tropical maize was gauged
from the following questions:

1. Which was the worst season for farming over
recent years?

1977 1976

1975 1974

Long rains
Short rains

2. What was the main cause of these difficulties?

3. Why was this the worst season and what was
the effect on you and your family?

4. In which years recently have you experienced
the following problems?


77 176

Aware of

Not a

i. Heavy late rains rot maize ears
2. Heavy early rains spoils maize seeds in the field
giving poor germination
3. Early finish to rain so that maize matures poorly
4. Poor germination of finger millet
5. Early finish to rains-late planted beans fail

I _



1. Kearle, B. (ed) Field Data Collection in the Social Sciences, Agricultural Development Council,

(Chapter 5 discusses issues and experiences in questionnaire development)

2. Bernsten, R. Design and Management of Agricultural. Research: A Guide for Agricultural Re-
searchers; Draft Paper, International Rice Research Institute, 1979.

(Offers useful guides to questionnaire development)

3. Collinson, M.C. Farm Management in Peasant Agriculture, Praeger, New York, 1972.

(Methods for collection of specific types of data especially land, labor and output)

4. Norman, D.W. "Methodology and Problems of Farm Management Investigations; Experiences
from Northern Nigeria" African Rural Economy Paper No. 8, Michigan State University, 1973.

(Reviews experiences in questionnaire development for conducting farm level studies)



Before beginning a survey a basic decision
must be reached about the population of farmers
of interest. Generally we are interested in im-
proving technologies for those farmers currently
growing the crop. Sometimes our interest is
broader. If the target crop is a new crop or not
widely grown, the population of interest are all
those farmers who could potentially grow the crop
especially if a technology were available to make
the crop attractive to these farmers. In this
chapter, we assume the more usual situation where
the survey population consists of farmers already
growing the crop. However, the procedures can be
easily modified to include populations of farmers
based on other criteria.
The unit of interest in sampling is those
members-of the farm family who make decisions
about technologies. This applies even in cases
discussed later in this chapter where plots or
fields are sampled rather than farmers. In this case
the field is only used as a convenient means for
identifying farmers who then become the focus of
the survey.
Because it is not possible to interview all
farmers in the target group, we interview a part
or a sample of the farmers and use the information
obtained from this sample of farmers to make
statements or inferences about all farmers in the
population. That is, we describe cropping patterns
and management practices, use of inputs, and
production problems for the target crop for the
whole population using information from the
sample. Our objective in sampling is therefore to
select at reasonable cost, a group of farmers which
is roughly representative of farmers in the popula-
tion. A representative sample must be selected at
random-that is each unit in the population or
subgroup of the population has an equal or known
chance.of selection. A representative sample must
be of a certain minimum size in order to confi-
dently make statements about the population as a
whole. However, as size increases, so do costs
therefore, sample sizes must be kept within-
reasonable bounds.

In this chapter we describe practical sampling
methods which we have found to give represen-
tative samples. These methods provide several
alternatives, at least one of which should be
appropriate in a given situation. The relative
advantages of each method are also given to help
in making decisions on which one to use.

8.1 Stratification
Stratification of the population is the process
of dividing the population into relatively homoge-
neous subgroups called strata, and then taking
separate samples from each group or strata.
For example, let us assume we are surveying an
area with two distinct groups of farmers with
different management practice-small subsistence-
oriented farmers who comprise 90 per cent of
the farmers in the area, and large commercial
farmers who are only 10 per cent of all farmers.
If we take a random sample of 100 farmers, we
expect about 90 small and 10 large farmers. (In
fact there is a 45 per cent probability that we
would get less than 10 large farmers). In this
case we probably have more small farmers than
needed to represent this group while only 10(and
perhaps less) large farmers is hardly sufficient to
be representative of this type of farmer. A more
efficient procedure is to stratify the population
into small and large farmers and choose say 50
small farmers and 30 large farmers,1thus reducing
total sample size and providing more information
on the large farmers at little or no sacrifice of
information about small farmers.
It is convenient to stratify as far as possible
using the recommendation domains previously
delineated-that is, groups of farmers of similar
agro-climatic and socio-economic circumstances
for which the same technological alternatives
can be recommended. In the case of recom-
mendation domains based on agro-climatic
characteristics such as rainfall, elevation and
sometimes soils, it is usually possible to divide
the region into distinct geographical areas for

1/ The actual number chosen will depend on the relative variation in each farm size group. See Section 8.3.

the purpose of sampling. However, with other
agro-climatic characteristics such as sloping
versus flat land, this may not be possible since
these different agro-climatic environments may
even occur within a single farm.
One of the most common socio-economic
characteristics used in defining recommendation
domains is farm size. However, stratification by
farm size requires a knowledge of farm size prior
to sampling and this is not always available. If
recommendation domains are defined on the
basis of location-usually proximity to large
town(s)-there is no difficulty in dividing the
area into strata for sampling purposes.
In summary, try to stratify the population
by recommendation domains prior to sampling.
In many cases only a partial stratification is
possible. For example, with recommendation
domains based on rainfall and farm size, it might
only be possible to stratify by rainfall since data
on farm size are not available prior to sampling.-l

8.2 Random Sampling Procedures
Random sampling is a selection procedure
which ensures that every unit of the population
or a strata of the population has an equal chance
of being selected. Random sampling is best done
with a table of random numbers such as the one
included in the Appendix.
Simple Random Sampling: In this method
every farmer in the population or in each strata
(if the population is stratified) is listed and a
table of random numbers is used to select the
farmers to be interviewed. This is a very simple
method. It's main disadvantage is that a reliable
list of all farmers in a region is usually not available.
Lists kept for tax purposes, for example, are
often incomplete.
Two-Stage Sampling: In this method a
random sample of villages is taken from a list
of villages in the region or in each strata and then
a sample of farmers is selected randomly from a
list of farmers in each selected village (see Example
8.1). Using this procedure, it is only necessary to
construct lists of farmers for the selected villages.
In addition, interviewing costs are also reduced
because of the geographic clustering of the
farmers. Counties or townships, cooperatives, land
reform units, or districts, as well as villages may
be used in the first phase of sampling.

Field Sampling: At times it may be more
convenient to sample fields rather than farmers.
SThe cultivator of the field is identified and then
interviewed. Fields may be randomly sampled
by several methods. Topographic maps and aerial
Photos of scale 1:50,000 or less are available
for many areas (although often difficult to
locate) and can be used for sampling. Any strata
based on factors such as rainfall or location are
first identified on the maps. Points on the map
are then selected by randomly drawing pairs
of coordinates (three digits will usually suffice).
In the field, each point is located and the
cultivator of that field then becomes the selected
farmer (see Example 8.2). If detailed maps are
not available then points may be randomly
located on more general maps (e.g. a road map).
Then in the field that point is very approximately
located and a field chosen randomly from all
nearby fields-for example by walking a fixed
distance in a certain direction.
The main advantage of field sampling is
that it avoids making lists of farmers or villages.
However, because the field is the unit of sampling,
larger farmers have a higher probability of being
selected and weighting of data is required when
summaries are made of average farmer practices
in the area (see Chapter 10). Moreover these
procedures may be costly to implement where
travel is difficult and inpractical or where it is
necessary to establish contact with local leaders
such as village officials before interviewing a
farmer. A modification of the above procedure
using two-stage sampling solves some of these
problems. In this method a segment (e.g. one
square kilometer) is selected at random on a
map using similar procedures to those above. In
the field, all farmers who have fields within the
segment are identified and if this number is too
large, a random sample is chosen from this
group of farmers. If the segment is relatively
large, each farmer has equal probability of being
chosen. Moreover, travel time is again reduced
by the clustering of farmers.

8.3 Sample Size
A representative sample not only must be
random but must also be large enough to reflect
all farmers in the region. Well known statistical
rules for determining sample sizes on the basis of

1/ Where farmers cannot be conveniently stratified, quota sampling may be useful. A quota is set for each stratum and
farmers are randomly selected, visited and asked a screening question to determine which strata the farmer belongs to.
If the quota is already met, the interviewer proceeds to the next farmer in the sample until the quota is met for each

the variability within the sample cannot be
formally applied. Nonetheless consideration of
variability within the target region is important
in determining sample size. As a general rule we
have found that 40-50 farmers for each recom-
mendation domain will usually reflect quite well
the circumstances of farmers in that recommen-
dation domain. Where it is not possible to stratify
by recommendation domain for sampling purposes,
try to choose a sample size that will result in at
least this number of farmers in each recommenda-
tion domain.
The sample sizes suggested here may be
adjusted according to the amount of variability
in the population. In an area where there is much
variability within recommendation domains, for
example, due to mountains and where any further
disaggregation would create too many domains,
the sample size should be increased. On the other
hand, in an irrigated land reform area with similar
size farms and agro-climatic characteristics, a
smaller sample size may be in order. Note that the
sample size depends on the variability within the
population and not on the size of the population.
The percentage of farmers sampled may vary
substantially between regions or recommendation
Finally, sample size must conform to the
time and cost constraints of the survey. In general
we recommend that total sample size be reduced
to fit these constraints by focusing on fewer
recommendation domains rather than reducing
the number of farmers sampled per recommenda-

tion domain to unacceptably low levels. (e.g. less
than 25 farmers per recommendation domain).
Recommendation domains considered less impor-
tant, and therefore omitted, can always be in-
cluded in future research plans.

8.& Non-response
It is common to find farmers away from
home at the time scheduled for the interview. If
these farmers are omitted from the survey, the
results will be biased toward the type of farmers
who are at home most of the time. Those who are
often gone from home could be those who have
part-time work off the farm; those who are
community leaders; those who leave frequently
for machinery repairs or purchase of inputs;
or those who idle in village coffee houses or
taverns. It is clearly worth some special effort
to ensure that these types of farmers have the
same probability of being in the sample as do the
In practically no case should a not-at-home
be dropped from the sample on the basis of one
attempted interview. In few cases would more
than two return calls be cost effective. The
choice will depend upon the cost of return calls
and the number of other not-at-homes encountered
in the sample. Ocassionally non-responses will be
due to the farmer's refusal to cooperate. In our
experience this is not common in a well managed
survey. Procedures to ensure farmers' cooperation
are discussed in the next Chapter.

Example 8.1:
Two-Stage-Sampling: in one tropical maize
growing area this method enabled considerable
cost savings. Farmers in the target population had
been grouped into villages based on land reform
units. At the head of each village was a local
official who had lists of all farmers in that village.
The following steps were taken in sampling.
First, a complete list of all villages in the region
and the number of farmers in each was drawn-up.
Second, villages were stratified as far as possible
by three previously defined recommendation
domains-heavy vertisol soils on flat land, the
same soil type but on hillsides, and lighter
alluvial soils on flat land. In this case villages

could only be distinguished by soil type as most
villages with heavy soils included both sloping
and flat land. Third, a relatively small number
of. villages (eleven for the first two recommen-
dation domains combined, and four for the
third) were selected randomly so that the proba-
bility of selection of a village was proportional to
the number of farmers in each village.-1
Finally, each chosen village was visited and a
list of maize farmers was drawn up with local
officials from available lists of all farmers. Ten
farmers were chosen at random from this list.
This gave a total sample of 150 farmers-110 in
the first two recommendation domains and 40

1/ 'The procedure used was to cumulate the number of farmers in each village with village 1 with 55 farmers receiving the
numbers 1-55, village 2 with 33 farmers 58 to 88 (55 +33 = 88), village 3 with 75 farmers the numbers 89 to 143 and
so on. Random numbers were then drawn to choose villages. If the number 69 is drawn then village 2 is chosen. In this
procedure the probability of a village being chosen increases proportionally with the number of farmers in that village,
but each farmer in the area has equal probability of being chosen.


in the third (although only 130 were finally
interviewed because of not-at-homes, etc). The
sample size in the third recommendation domain
was smaller than the others because of the relative
uniformity of the domain. The two-stage proce-
dure saved costs because (a) selected farmers were

clustered into a few villages reducing travelling
time; (b) the purpose of the survey could be
explained to local officials at the time of sample
selection, and (c) lists of villages and farmers were
already available.

Example 8.2:
Field Sampling: A joint survey of maize
and potatoes was conducted in a highland valley
where these crops were dominant. Recom-
mendation domains were largely based on altitudes
and a target sample size was established as follows:


Valley Floor
East Slope
West Slope
High Elevation


3100-3500 m.
3500-3950 m.
3500-3950 m.
3950 m. +

Crop Size

Maize 125
Potatoes/wheat 60
Potatoesibarley 50
Bitter potatoes 20

In this case, sample size was based on
relative heterogeneity of each domain leading
to a larger sample in the valley floor. The sample

size in the high elevation zone was reduced
because of logistic difficulties and its relative
unimportance in crop production.
Fields were chosen using a regional map
(1:100,000 scale) and with random numbers
specific points were selected until the target
sample size was reached for each recommendation
domain. A field of the target crops could usually
be found at that point and if this was not the
case the farmer with the closest field was
in this case, field sampling was appropriate
because reliable lists of farmers were not available
and it was not deemed necessary to approach
farmers by first seeking support from local



1. Raj, Des The Design of Sample Surveys, McGraw Hill, New York, 1972.

(Chapters 1-9 provide detailed discussion of application of sampling theory to survey design)

2. Kearle, B. (ed) Field Data Collection in the Social Sciences, ADC, New York, 1976.

(Chapter 3 reviews various sampling approaches used in micro-level surveys in Africa)



With a questionnaire developed and a sample
drawn, the formal survey is ready for implemen-
tation. Successful implementation requires a cadre
of capable interviewers, the farmers' cooperation,
correct completion of the interview and close
supervision by the researchers of these activities.
In this chapter we shall look at each of these in

9.1 The Interviewer
The interviewer is the middleman between
the researchers and the farmers. The quality of
the interviewer is one of the most important
factors in conducting a successful survey.
Number of Interviewer: Even with our recom-
mendation that the questionnaire be designed to
be completed in less than one hour, it is our
experience that interviewers will only average
about one to two interviews per day. The remain-
ing time is spent locating farmers, waiting for
public transportation, and conducting "return
visits". In a typical survey of say 120-150 farmers
(e.g. 40-50 farmers in each of three recommenda-
tion domains) we would need about three inter-
viewers to complete the survey in a month. The
survey could be completed more quickly by using
more interviewers in a shorter period of time, say
12 enumerators for a period of a week. However,
the quality of the data will usually be less because
each interviewer has less opportunity to develop
his skills through on-the-job training and will also
receive less intense supervision.
Selection of Interviewers: The researcher
should personally recruit the interviewers. Four
characteristics are important in selecting inter-
viewers: a) motivation to work hard and honestly,
b) ability to fill the questionnaire correctly
(usually determined by some minimum level of
schooling and intelligence), c) ability to commu-
nicate with the farmers in the local language, and
d) knowledge of local agriculture and respect for
farmers and rural people. Some of these charac-
teristics such as language ability, ability to
complete the questionnaire, and knowledge of
local agriculture can be evaluated in recruitment
interviews and in training but others such as

motivation and honesty must depend on personal
assessments by trusted acquaintances.
In the ideal situation, the research program
employs its own research assistants recruited on
the basis of the above characteristics. These
research assistants are then available not only for
survey work but also for other activities of the
research program, particularly work on on-farm
experiments and demonstrations.
Where such research staff is not available,
the best choice is usually to hire, on a temporary
basis for the survey, sons of local farmers who
have completed at least a primary school education
and who are literate. During school vacations,
high school students (again farmers' sons) or
local school teachers can also be employed.
Although university students have been widely
used in surveys, they are unsuitable if they lack
respect for farmers and rural customs.
Training the Interviewers: The training
period for this type of survey will depend on the
type of interviewer but will vary from two days
for research staff already familiar with survey
work to five days or more for temporary staff.
The training period should include the following:
a) The purposes of the survey should be fully
explained including an explanation of how the
data'will be used in planning on-farm experiments
and experiment station research.
b) The questionnaire should be explained
question by question. (Many explanatory notes
should also appear on the questionnaire.) The
sampling procedure should be explained and local
terminology and units of measure discussed.
c) The plan for field operations should be
explained, including instructions on the inter-
viewer's responsibilities for screening respondents
and action to be taken for nonrespondents.
d) Interviewing techniques should be
-described and practiced among each other.
e) If the interviewers are unfamiliar with
the area, they should be taken on a tour and given
background information on the agriculture, social
structures and government development activities
in the area and introduced to relevant local

f) Field interviews should be conducted by
the interviewers both in the presence of one of
the researchers and alone. The respondents should
not be part of the sample, nor should the data be
used. These interviews may be made a part of
the tour of the area, if one is conducted.
The effectiveness of the training and later
field work can be greatly increased by developing
an interviewer's manual. This manual should be
comprehensive and cover all the points covered
in training, i.e. purpose of the survey, explanation
of. each question, logistics and interviewing
techniques. Also the manual should be a ready
reference of common diseases and pests, chemicals
and varieties available in the area. The interviewer
should be required to have this manual with him
at all times.
The training period is a good time to "weed
out" undesirable interviewers. In fact, it is best to
hire interviewers on the condition that they
successfully complete the training course. In this
case allow for dropouts at the time of recruitment.

9.2 Gaining Farmers' Cooperation
Farmers' cooperation is essential to the
success of the survey. This cooperation is gained
at two levels: a) through support of local leaders
and government officials and b) by correctly
introducing the survey to the farmer. The support
of local leaders and officials is best obtained during
the exploratory survey through personal visits
by the researchers to explain the purpose of the
work. These local officials can then be asked to
help explain the work to the farmers in the sample.
Also, where two-stage sampling is used and selected
farmers are clustered into villages, it is often
helpful to have group meetings with the farmers
in each chosen village for the purpose of
explaining survey objectives and enlisting their
To obtain the cooperation of individual
farmers, the interviewer should introduce himself,
explain for whom he is working, and explain
fully the need to have information about farmers'
production practices and problems to help
direct research work. While the potential benefits
of the information to farmers as a whole may be
mentioned, each farmer should understand that
he will receive no special consideration as a
result of participation in the survey, although
he may be invited to visit the closest experimental
site. The farmer should be told that he was
selected on a lottery basis and that all the
information he provides will be kept strictly

It is best to interview the farmer when
and where it is convenient for him. If he is to be
interviewed at home, early morning and late
afternoon usually will be less disruptive to his
work schedule. However farmers will often
perceive more interest if the interviewer is willing
to go to the field for the interview. This also
provides an opportunity for direct observation
of the crop in the field. If the farmer is very
busy, offer to help him for a while before begin-
ning the interview. Above all, treat the farmer
with respect.
When these steps are taken, it has been our
experience that farmers are very willing to co-
operate with this type of survey. Gifts or
remunerations are not necessary except according
to local customs. Problems of cooperation usually
arise when the researchers do not inform local
leaders, do not explain the purposes of the survey
to the farmer or do not treat the farmer with

9.3 The Interview
In general, interviews should be conducted
with the primary decision-maker for the target
crop. In some cases this may be the women of
the household, in which case female interviewers
may be more suitable. In any case, women may
play an important role in crop production and be
responsible for decisions concerning certain
cultural practices (e.g. weeding) and food proces-
sing and consumption. In this case it may be
better to ask questions related to these activities
directly to the women, provided it is done with
the consent of the household head.
The interviewing should be as relaxed and
informal as possible. The farmer is most comfort-
able sitting down in his house or under the shade
of a tree in his field without the presence of other
people. The interviewer can help by conducting
the interview as a conversation. He should know
the questionnaire so well that he memorizes
individual questions and does not laboriously read
them. The farmer is encouraged to talk about
certain topics with gentle direction from the
interviewer. Additional information or unusual
information may be recorded in space provided
on the questionnaire.
The interviewer should ensure that the
farmer understands the question but should not
inject his own opinion. He must be alert for
responses which are irrelevant, vague, improbable,
or inconsistent with previous responses. When
such responses are noted, the interviewer should
probe further by asking related questions which

will help to clarify such responses. The interviewer
must guard that these probes do not suggest
answers, as the respondent may acquiesce to the
suggestion as being the easiest way to solve the
Before terminating the interview, the
questionnaire should be reviewed to ensure that
all information is complete. Interviewers should
record responses for all questions. If a question is
not applicable--e.g. a question about method
of fertilizer application for farmers who don't
use fertilizer-then NA should be inserted. If the
farmer doesn't wish' or cannot respond to a
question, then NR for non-response may be
recorded on the questionnaire.

9.4 Supervision
Experience shows that constant effective
supervision is critical to the success of a survey.
The researchers) must be in the area during the
period of the survey acting as field supervisor
throughout. He or she should, to the extent
possible, collect and field edit the questionnaires
on a daily basis. Field editing consists of checking
through the questionnaire for legibility, complete-
ness, and consistency. Regular and frequent field
editing allows the supervisor to discuss the problem
with the interviewer while the interview is fresh in
his mind, and also provides motivation for the in-
terviewer since he knows his work will be quickly
and thoroughly examined.
In addition, the supervisor should spot-
check the work in the field to determine that
interviewers are conducting interviews when and
where scheduled. Often it is useful to reinterview
a few farmers on an informal basis to check that
the interviewer is doing his work correctly. Finally
the researcher learns much from this intensive
supervision which helps him later to interpret
the data.
9.5 Example of Survey Implementation
In one successful survey a sample of 130
maize farmers clustered in eight villages was
interviewed in an area of strong traditional
village leadership. At the beginning of the research
work, letters were sent from the head of the
agricultural research institute to the local govern-
ment official in the area presenting the researchers,
explaining the purposes of the survey and asking
him to advise the village leader in each selected
village about the work. (A two-stage sampling
procedure was used). During the exploratory
survey, the government official was personally
visited by the researchers and asked to accompany

them on a tour of the area and present the
researchers to the village officials. The purposes
of the survey were explained personally to each
village leader and if it was a Friday when the
farmers (who were largely Moslem) were in the
village, a meeting was also held with farmers.
Meanwhile four interviewers with two substi-
tutes were recruited by informal contacts. These
interviewers were 18-25 year old sons of local
small farmers, had completed primary schooling
and had been recommended to the researchers for
their personal characteristics (e.g. enthusiasm,
intelligence, agreeable personality). All four inter-
viewers successfully completed a five day training
course and the two substitutes were not needed.
This training course consisted of a full explanation
of the purpose of the survey, the role and charac-
ter of the sponsoring agency, the way in which
collected data would be used, and methods of
gaining a respondent's cooperation. They were
then given a question-by-question explanation of
the questionnaire, with possible ambiguities
pointed out and clarified. Subsequently, they
conducted mock interviews with each other; the
researcher and other enumerators criticized each
performance. Finally, they were assigned inter-
views with local farmers, the results of which were
carefully scrutinized. These trial interviews pro-
vided further opportunities for pre-testing the
questionnaire, and several changes suggested by
farmers or enumerators were incorporated at the
final hour.
During the survey, the interviewers were
introduced to the village leader by the researchers
and a list of randomly selected farmers was given
to the leader who was then asked to advise each
farmer of the coming survey. At this time the
survey was administered to the village leader to
demonstrate its nature and ensure cooperation
of other farmers. (The information was not used
in later analysis). Because the survey was timed at
a slack period just before harvest, many farmers
were in the village. Those who worked in fields at
a reasonable distance from the village were often
interviewed in the field. The village leader was
asked to provide accommodation for the inter-
viewers during the survey in that village.
The supervisor divided his time between
advance work in the next village, collecting and
editing previous questionnaires on a daily basis,
and surprise visits to interviewers in the field. He
also informally revisited some farmers to check
on the quality of the information obtained by
one interviewer in whom he did not have full

The survey was completed by four inter-
viewers in four weeks. Only one of the farmers
refused to cooperate after the purposes of the
survey were explained. Researchers were pleased
with the high quality of data obtained which

they attributed to the training, personality and
local knowledge of the interviewers and the
constant supervision by one of the senior re-



1. Kearle, B. Field Data Collection in the Social Sciences, ADC, New York, 1976.

(Chapters 7-10 contain many useful experiences in survey implementation)

2. Bernsten, R. Design and Management of Survey Research, Draft Paper, International Rice Re-
search Institute, 1979.

(Chapters 8-10 provide guides for survey implementation)

3. Spencer Dunstan S.C. "Micro-level Farm Management and Production Economics Research Among
Traditional African Farmers: Lessons from Sierra Leone", Africa Rural Employment Study No.3,
Michigan State University, 1972.

(An overview of implementation procedures used in farm level surveys in Sierra Leone)

PART ill


Once the formal survey is completed, the
information on farmer circumstances must be
analyzed and used in planning research to develop
technologies appropriate to farmers. It is impor-
tant that this analysis be conducted quickly so
that the result can be immediately incorporated
into decisions on crop research.
The analysis of the survey data is best
undertaken according to the objectives listed in
Chapter 2 for collecting information about farmer
circumstances. First, there are the descriptive
objectives of (a) refining boundaries of recommen-
dation domains and (b) describing characteristics
of farmers, their management practices and their
fields to help guide the choice of representative
sites and practices for on-farm experimentation.
Second, there are the diagnostic objectives to (a)
identify the relevant problems of farmers and


prescreen technological components for inclusion
in on-farm experiments, (b) to identify problems
and constraints of farmers that should guide
research on experiment stations such as the
development of new varieties, and (c) to identify
the implications for policies relating to credit and
input distribution and marketing, which support
the introduction of new technologies.
Chapter 10 focuses on methods for analyzing
the survey data to meet the descriptive objectives
and to test hypotheses on farmers' practices
and problems. Chapter 11 presents a set of proce-
dures for prescreening technological components
for on-farm research and establishing priorities for
-varietal development. Finally, some examples of
the application of these procedures are given in
Chapter 12.



In this chapter, we show how information
can be extracted from the survey to meet each of
the survey objectives. In addition, we describe
methods for efficiently assembling this informa-

10.1 Refining Recommendation Domains
Recall that recommendation domains have
been tentatively defined in the exploratory
survey on the basis of both agro climatic and socio-
economic factors. A first step in the data analysis
is to check that variation in farmers' practices do
correspond with these domains. On the basis of
this checking it might be necessary to combine
recommendation domains, create new ones or
simply adjust boundaries. This is done by observing
whether variation in current farming systems and
crop management practices in the region are
related to those agro-climatic and economic
circumstances hypothesized from the exploratory
survey to be important in determining priorities.

Practices to consider are: importance of various
crops and varieties, intercropping practices,
rotations, planting method and dates, tillage
techniques, yields and crop disposal. This can be
done by arranging data for farmers in the sample
according to each circumstance hypothesized to
be important and then looking for a tendency for
some of these practices to be related with that
factor. For example, if rainfall is hypothesized to
be important, data for farmers is arranged accord-
ing to approximate rainfall gradients to observe
any changes in farming systems and practices. If
farm size is hypothesized to affect management
practices, farmers should be arranged in order of
increasing farm size to look at variation in vari-
ables such as crops grown, crop rotations, tillage
practices, planting method, intercropping or
production problems with increasing farm size
(see Example 10.1).
This analysis will usually allow more precise
drawing of boundaries of recommendation domains.

Example 10.1:
In one barley/wheat producing area it was
clear in the exploratory survey that larger farmers
were ploughing earlier and using drills to plant
but it was not possible to determine at what farm
size these practices became common. By arranging
data from the formal survey in order of increasing
farm size, it was decided to include farmers with
over 20 ha. as a separate group of farmers. Sixty

per cent of farmers above 20 ha. were ploughing
immediately after harvest compared to only
eighteen per cent with less than 20 ha. Eighty per
cent of farmers with over 20 ha. planted with a
drill but all farmers with less than 20 ha. broadcast
their seed. It was clear in this case that farmers
with more than 20 hectares were generally using
different practices.

10.2 Assembling Information on Farmers' Practices:
A description of farmers and their practices
in each recommendation domain area is used to
help select sites for on-farm research that are
representative in terms of soils, tillage techniques,
rotations, topography, etc., and then to provide
representative practices in terms of time and meth-
od of planting, weeding, etc. for the conducting
experiments on these sites.
The information in Table 10.1 shows the
type of information that is needed here to describe
practices of farmers in each recommendation
domain. This tabulation is relatively simple once
recommendation domains have been established.
Descriptive statistics are assembled for each type
of variable. These may be either frequencies or
means. Variables which are not quantified must
be presented as frequencies-e.g. type of variety,
power source for land preparation (manual,
animal, tractor) etc. Histograms are a convenient
way to present frequencies. Variables which are
quantified, e.g. seeding rate, area, labor inputs,
yield may be presented as means as well as
frequencies to show the variability within the
sample. For many inputs it is useful to complete
two statistics (a) the frequency (percentage)
of farmers using the input, and (b) the average
rate of use of the input for those farmers who
use it.
Table 10.1 shows that each type of tabulation
is performed to meet certain objectives of the
survey. For example, to establish representative
practices for the on-farm research, tabulations
are needed on intercropping, density, all field
operations, and use of specific inputs. This infor-
mation should be quite specific. For example, tim-
ing of field operations or input application is im-
portant in representing farmers' practices in on-farm
experiments. (Tabulation of survey data on farmers'
practices is shown in Examples 10.2 and 10.3).

Example 10.2:
Table 10.2 shows an example of assembling
descriptive statistics by recommendation domains
from a highland maize producing area. There are
clearly some practices such as a common density
and intercropping systems that are followed by
most farmers. There are also practices which vary
importantly by recommendation domain. For
example in recommendation domain two, the
farmers are generally larger and sell more maize
than other farmers. In recommendation domain
4, feeding of weeds to animals is less important.
This tabulation then suggests that base
practices for on-farm experiments be: intercrop-
ping with beans, 2-3 seeds/hill at a spacing of
60 cm x 60 cm, without fertilizer and insecticide.
The high proportion of farmers feeding weeds to
animals suggests that this is a factor to consider in
evaluating weed control technologies.

Example 10.3:
Table 10.3 shows results of a survey of barley
producers where farmers have been divided into
four, recommendation domains based on three
factors: rainfall, intercropping with a perennial
(preventing machinery use), and farm size. This
table shows basic differences between land
preparation and seeding practices, and inputs
by recommendation domain. Use of improved
varieties, fertilizer and herbicides is common
in the higher rainfall areas particularly for large
farmers. Use of tractors for land preparation is
minimal in the second recommendation domain.
From the same survey a histogram showing
frequency of use of varieties has been constructed
in Figure 10.1. Again in recommendation domain
three, farmers use later varieties because of their
more favorable climatic circumstances.

__ _. __ __.. I_ _~_ __ ____


Type of information Details of Tabulationsa/ Use of Information

General farmer data
Farm size
Land tenancy
Crops grown

Specific field data


Density and Spacing

Basic operations

Land preparation




Crop Production/Disposal
Yield of crop/crop mixture
Crop disposal
Use of crop residues
Use of weeds

Price data
Price received for crop

Price to hire labor
Price to hire machines
(if relevant)
Prices of various inputs

Average (and freq.)
Percent x type
Average area x crop



Average (and freq.)

Per cent x type

Average (and freq.)

Average dates (and freq.)
Per cent x* type power
Per cent x type implement
Per cent x times performed
Average labor (only if relevant)

Per cent using input
Per cent x type
Per cent x times applied
Average date(s) applied
Per cent x method applied
Average labor/application

Average (and freq.)
Per cent x type
Per cent x type
Per cent x type

Average (and freq.)

Relative importance of
different farm groups.
Refine RD
Help choose representative
farmer for trials.

Refine RD
Choose representative site
for trials.
Establish intercropping and
and density for trials.

1. Establish representative
practices for trials

1. Base data on per cent
using new inputs before
2. Establish representative
practices for trials.

1. Refine RD
2. Base data on current pro-
duction and disposal

Pre-screen technologies for

Economic analysis of
experimental data

a/ Averages are means of the variables. Per cent x type are percentages of farmers with specific character-
istics such as land owner, share cropper, renter in the case of land tenancy or maize, maize-beans,
maize-sorghum, maize-sorghum-beans in the of intercropping.

Sample Tabulation of Some Descriptive Statistics on Farmers' Practices by

Recommendation Domain-Highland Maize

Recommendation Domain

Variable 1 2 3 4 5 6 All

General Farm Data
Number of farmers 53 33 26 20 58 30 230
Average farm size (ha.) 1.06 2.19 0.71 2.70 1.60 0.92 1.46
Per cent cultivated area in maize
(ha.) 85 68 42 56 81 98 71
Off-farm job-per cent yes 60 36 38 30 38 37 42
Per cent of off-farm jobs
that are fulltime 56 55 70 67 64 91 64

Field Data for Maize
Slope-Flat (per cent) 60 63 73 80 79 80 72
-Steep (per cent) 40 37 27 20 21 20 28
Intercropping maize-beans-
per cent practice 92 70 100 100 91 100 92
Density-Distance between
rows (cm.) .71 .70 .65 .60 .64 .70 .66
-Distance between
hills (cm.) .62 .72 .67 .59 .56 .59 .62
-Seeds/hill 2-3 2-3 2-3 2-3 2-3 2-3 2-3

Crop Production-Disposal
Per cent selling maize 11 42 0 3 22 7 16
Per cent feeding weeds to
animals 79 76 54 87 71 73 74

Use of Inputs
Per cent use any fertilizer 75 42 42 93 76 76 70
Of users-per cent manure 98 93 82 82 75 100 88
-per cent urea 0 0 9 7 7 0 4
-per cent 10-30-0 3 7 9 11 18 0 8
Per cent use insecticide 6 3 0 3 5 7 5

Table 10.2

Sample Tabulation of Production Practices from a Survey of Barley Producers

Recommendation Domain "
Low Rainfall Higher Rainfall
Variable 1 2 3 4
Solo Inter- Small Large
Cropped cropped 20 ha. 20 ha. +

General Farm Data
Average area barley (ha.) 4.6 4.7 1.9 57.1
Per cent flat land 95 44 12 50
For rented fields-per cent cash 21 0 0 50
-per cent shares 79 100 100 50

Basic Operations
Per cent ploughing 100 56 100 100
Per cent plough after harvest 49 2 18 60
Per cent harrow one or more times 100 41 50 100
Sowing-per cent broadcast 88 100 100 20
--per cent drill 12 0 0 80
Covering broadcast seed-per cent
with harrow 100 25 79 100
-per cent
with plough 0 75 21 0

Variety-per cent using improved
variety 46 2 75 100
Fertilizer-per cent apply nitrogen 4 2 40 80
Herbicide-per cent use 2-4D ) (2 932 100
Production (Disposal)
Yield (tons/ha.) .91 .60 1.00 2.30
Per cent sold 29 52 70 100

Figure 10.1 Histogram Showing Dittribution of Varieties Sown by Recommendation

S Loca! variety
S Late matu ing improved variety


Other improved varieties


!y, i^
,-. i .-
.,* '" i i r '

**,' i ., ,,*- ..
'" :

_^j __ ^j _,___ri _________ -
'.- .,
.. ... ... r _2 ,

Sols cropped


Low Rainfsll

RD-3 RD-4
Small Frmsn (< 20 ha.) Lsrge Farms (< 20 ha.)
Higher Rainfall

'Table 10.3




10.3 Diagnostic Analysis of Farmer Circumstances
The next stage of the data is to look for
particular aspects of farmer circumstances that
influence current practices and that might guide
the choice of technological components and their
level for on-farm experimentation. This diagnostic
work also helps identify problems for on-station
research and provides evidence for changing
policies relating to credit, input distribution and
marketing that would facilitate introduction
of these technologies.
In all cases several types of information are
assembled. Most important are the opinions
expressed by farmers themselves about specific
issues. Through a thorough exploratory survey
and careful design of the questionnaire, the survey
should elicit information for each recommendation
domain about farmers' opinions and preferences
with respect to particular varieties, use of inputs
and method and timing of cultural practices. Also
helpful are responses on common circumstances
leading to crop losses, e.g. drought (at particular
times in the crop cycle), specific insects and
diseases, and difficulties with obtaining inputs
or marketing produce (see Example 10.4). If the
survey includes field observations, tabulation of
data on the incidence of problems in farmers'
field is also valuable.
Data on the farming system-for example
periods of labor bottlenecks (see Example 10.5),
food preferences, rotations-give the researchers a
better understanding of the farming system into

Example 10.4:
In one survey farmers were asked about the
incidence of certain climatic risks in various
crops. In Table 10.4 it is clear that rainfall, both
too much and too little, is dominant among

Table 10.4 .

which new technologies must be introduced.
Finally cross tabulations are important in
testing hypotheses about farmers' behavior. These
hypotheses will have been formulated over the
course of the survey work and particularly in the
exploratory survey. For example, if a considerable
variation in time of planting is observed with
likely consequences for yield, we might cross-
tabulate time of planting with factors hypothesized
to affect time of planting such as the previous
crop in the field, method of land preparation,
ownership of equipment, or period of food
scarcity (leading to earlier planting). Or chemical
fertilizer use might be hypothesized to be related
to type of rotation followed, type of land, use of
organic fertilizers or availability of credit. Choice
of variety might depend on the end use of the
crop, local soil type or intercropping practices.
The hypothesis on circumstances leading to the
use of a given practice will arise from the explor-
atory survey. (Cross tabulation shown in Example
Obviously the range of cross-tabulation is
infinite. The important point is that the specific
tabulation chosen should reflect hypotheses of
the researchers about the use of a given practice.
Statistical testing such as the Chi-squared test
may be employed in this analysis to provide a
degree of confidence in the observed associations.
Also at this stage it is often helpful to combine
data across recommendation domains to ensure
a large enough sample.

farmers' risks. The incidence of these risks suggests
care in choosing technologies for experimentation.
For example, varieties should have a good resis-
tance to ear rot.


Type of Climatic Risk

Heavy late rains-maize cobs rot
Rains finish early -late planted beans fail
Rains finish early -poor grain filling of maize
Heavy early rains-poor germination of finger millet
Heavy early rains-poor germination of maize because
of waterlogging

Per cent of Farmers
Reporting Problem
in past three seasons



_ ___ I ~___ __



Example 10.5:
Farmers' response to a question about the
busiest months of the year is shown in Figure
10,2. Clearly November-December is a peak work
period end farmers will favor technologies which
reduce their work in these months and tend to
reject those that add work at this time.

per cent of farmers
mentioning a given
month a "busy
ons' in maize pro-

Figure 10.2 Distribution of Busy Month .


1 2 3 4 5

6 7 8 9 i0 11 12


Example 10.6:
Two examples of cross-tabulation to test
hypotheses are provided by one study in a tropical
area where maize was a relatively new crop. In
the exploratory survey a range of planting dates
and methods was noted. It was indicated that
farmers used three planting methods-planting
on mounds, planting on ridges and planting on
the flat. It was hypothesized that farmers planted
on mounds before the onset of rains in November
on low-lying areas. Later the labor constraint of
the November and December period and the
reduced probability of water lodging after the
initial heavy rains were factors leading to planting
on the flat. The cross-tabulation of type of seed
bed preparation and time of planting using data
from the survey shown in Table 10.5 verifies the
hypothesis. In the same study, it was hypothesized

Table 10.5

that maize was rapidly replacing the traditional
crop, finger millet, and in particular, that young
people preferred maize for their staple dish. In
fact, cross-tabulations of grain preference in
Table 10.6 show a strong preference to millet
among older household heads but about equal
preference for each grain among younger age

Cross-tabulation of Type of Seedbed and Time of Planting


i- -16

Seedbed Types
Planting time Mounded Ridged Flat

Before November 81 6 13
Early November 26 48 26
Late November 4 61 35
Early December 8 12 80
Late December 24 76
January 11 33 56

-h --- i ---I--------------------^-C1--l----l"l'

~LI__I I

"~~""""1"--~^I`----------- -------

Cross-Tabulation of Expressed Food Preference and Age of Household Head

10.4 Methods for Tabulating Information
Several methods are available to tabulate
survey information and the method used will
depend on the size of the sample and the time
and resources available for the task. All methods
require that all questionnaires be pre-edited to
convert quantities to standard units, clarify incon-
sistencies and discard any questionnaires where
inconsistencies are too large or non-response is
too high. It is also desirable (again regardless of
method used) to list out data for each important
variable on a sheet of paper to help find errors
and note the range of response.
Tabulating Directly From the Questionnaire:
In this method the questionnaires are arranged by
recommendation domains and opened to a given
page. Data are simply taken by flipping through
the questionnaire and counting the number of
farmers using a given practice or averaging with a
pocket calculator the amounts of a given variable
(e.g. seed rate) over all the farmers in each domain.
When all relevant variables on that page are
tabulated, all questionnaires are opened to a
new page and the-process continued.
The advantage of this approach is that it
can be done immediately with only a pocket
calculator. The main disadvantage is that for
larger sample sizes, say over a hundred, it is time
consuming unless clerical helpers are available.
If cross-tabulation is to be used extensively, it
is also cumbersome even for smaller samples
because variables to be cross-tabulated are often
on different pages of the questionnaire. Moreover
the addition of helpers raises the problem of errors
in the manual calculations and the need for
strict supervision.
Sorting Strips: These are made from com-
puter cards, cardboard file folders, etc. and data

for one farmer are written across the top of each
card. They can be made very cheaply and in
little time. The main advantage of this approach
is that it allows the farmers to be re-ordered
into different groups by farm size, tenancy,
geographical area, etc. For this reason it is most
useful in refining the recommendation domains
and in cross-tabulation. The main disadvantage
is that the data must be written out from the
questionnaire prior to tabulation. Generally once
recommendation domains are established most
type of tabulation are easier to make directly
from the questionnaires.
Use of Computer: In this case all data is
coded onto computer key-punch forms according
to prearranged codes (e.g. planting by hand = 1,
by machine = 2, etc.) and then punched onto
computer cards. It is also possible to precede
the questionnaire in such a way that data is
punched directly on to computer cards. Once on
computer cards, standard statistical packages or
specially written programs will quickly tabulate
the data. It is beyond the scope of this manual
to describe computer processing procedures
but a supplement to this manual is in preparation
which described how to code the information
and also gives a simple computer program easily
adapted to most computers for doing simple
tabular analysis.
The main advantage of the computer is that
once the coding is done the tabulation can be
done very quickly even for very large samples,
given the availability of a suitable program and
computer. Moreover, it enables statistical tests
of differences between groups of farmers to be
easily performed. The main disadvantages are:

Age of Farmer
Under 30 31-40 41-50 51-60 over 60

Per cent Prefer
Millet 50 63 27 59 78

Maize 50 27 63 41 22

Table 10.6

(a) the tendency to overlook errors or relation-
ships in the data that would be revealed in manual
analysis, and (b) the cost in time and money to
translate the data onto cards and to become
familiar with a program suitable for doing the
analysis. In general this latter disadvantage is
outweighed if the sample is over 100 farmers and
key punching, programming time and computer
time are readily available.
SEven where the data set is large enough and
computers and programs are available there are
benefits to completing a partial manual tabulation
before the computer analysis. First, it will enable
a quick summary of important information such
as -representative practices and incidence of
problems, -that can be quickly used in making
decisions about on-farm experiments. It is nearly
always quicker to do such a quick analysis
manually rather than by computer. Later, com-
puter analysis can still be useful to check the
previous analysis, provide more disaggregated
analysis, analyze less important information and
formally test hypotheses. Second, the manual
tabulation is useful in getting a "feel" for the data.
If the researchers participate in the tabulation they
note new relationships and define new hypotheses.
Finally a computer listing of data is convenient
for manual tabulation, particularly for simple
descriptive statistics such as number of farmers
using a given input.

10.5 Weighting Procedures
In many cases not al! farmers should be
given equal weight in the tabulation. The type
of weighting depends on the type of data and
method of sampling.
Tabulation by Farmer: Generally we are
interested in the average practices of farmers
in a given recommendation domain. if the
sampling procedure employs simple random
sampling of farmers then simple averages over
the sampled farmers is the appropriate method.
However, where sampling is by field then as we
noted, common sampling methods lead to a
higher probability of choosing larger farmers.
For this reason, when averages per farmer are
required the smaller farmers in an area sample
must be given greater weight and larger farmers a
smaller weight.
It is easiest to group farmers into farm sizk
groups of approximately equal numbers for the
purpose of weighting. In this case, if we have
groups then the proportion of farmers in each

group, pi is given by
Pi = (ni/xi ) / (? nj/lj)

where ni is the number of farmers in group i,
and xi is the simple average farm size for group i,
Then for calculating y, say the percentage of
farmers using fertilizer in the population, we
Simply weight the percentage of farmers using
fertilizer in each group

7i by pi (i.e. y= pi yi,).
Example 10.7 shows the weighting procedure in
one area sample.
Tabulation by Area: There are also many
variables that should be expressed on an area
basis, e.g. yields,fertilizer rates and labor inputs/ha.
so that larger farmers are weighted more heavily
to provide representative statistics of the region.
For example, in a random sample of farmers,
average yields should be computed not by
averaging yields of each farmer but by averaging
total production (yield x area) of each farmer and
then dividing by the average crop area over all
farmers. The resulting yield will be more represen-
tative of the region since it gives greater weight to
larger farmers which grow relatively more of the
crop. Likewise, fertilizer/ha and labor/ha should
be computed by averaging total fertilizer and
labor use and then dividing by average hectarage.
When field sampling is used, however, larger
farmers are already represented in the sample in
proportion to their area. In this case simple
averages of farmers' yields, fertilizer, etc. are
Weighting where Pre-Stratification was used:
If pre-stratification was used and the researcher
wishes to compute averages over groups for a
region, then the groups should be weighted by
the proportion of farmers in the population in
that group. For example, if a region with 75 per
cent dryland farmers and 25 per cent irrigated
farmers was pre-stratified into dryland and irri-
gated farmers and equal sample sizes are taken of
each type of farmer, then to compute a regional
average per farmer, dryland farmers should re-
ceive a .75 weight and irrigated farmers a weight
of .25.
In general however we are more interested
in representing average practices for each homo-
genous group or recommendation domain than in
calculating regional statistics.

Example 10.7:
A sample was taken using random coordi-
pates on a map of scale 1:50,000 and the resulting
farm sizes ranging from 0.5 ha to 40 ha were
grouped into three groups.

Average Farm Number of Number Using
Size in Farmers Fertilizer in
Group each Group


14 :


The denominator for the weighting is given

16 19 14-
by + j + 18.0 = 17.13. Therefore the
proportion of farmers in each group is (16/1.3)/
17.13= .72, (19/4.7)/17.13 = .25, and (14/18.0)/
17.13 = .06 respectively. That is the smallest
group makes up .72 of the farmers although they
own only 16/(16 +19 + 14) = .33 of the land.
In this case the percentage of farmers using
fertilizer in the population is (72x-2) + (25x-
16 19
+ ( 6 x ') = .22 although fertilizer is applied to
(2 12)/(16 +19 14)= .41 of the area.
(2 + 6 + 12)/(16 + 19 + 14) = .41 of the area.



1. Bernsten R. Design and Management of Survey Research, Draft Paper, International Rice Research
Institute, 1979.

(Discusses various methods for data analysis and also describes in detail the use of SPSS)

2. CIMMYT "The Use of Computers to Analyze Farm Survey Data" Economics Program (forthcoming)

(Describes coding procedures and data analyses using a small FORTRAN Program which can be
readily adapted to small or large computers. Also describes the use of SAS-Statistical Analysis
System-which is available in many computer systems)


0-1.9 ha
S 2.0-6.9 ha
7 ha and over

_ __ __



.The circumstances of most farmers are such
that they adopt technologies in small pieces,
usually one or two components at a time. They
do this because of (a) scarcity of capital, (b)
inability to withstand large risks and (c) a learning-
by-doing approach. A research program should
therefore initially aim to develop two or three
best bet technological components which have
relatively high pay-offs when added to the farmers'
technology. This strategy also benefits research
programs with limited resources which cannot
afford to investigate all possible components.
There are a number of steps in identifying
best bet technological components. These are:
(a) identifying for the target crop key factors
limiting farmers' production and income, (b)
identifying available technological components
by which those constraints may be overcome,
(c) listing all changes to the farmer that will
result by introducing these technological
components, (d) computing rough costs and
benefits to the farmer of these changes, and
(e) matching these changes against the relevant
circumstances of the farmer. This prescreening
process, involving both the agronomist and
agricultural economist, is a systematic way of
reducing the infinite variety of alternative
technological components down to a few best bet
components for immediate research.
Research priorities, of course, may be specific
to a given recommendation domain. However,
many research priorities will be general over all
recommendation domains in the region. In fact it
is even possible that the same research program
may be implemented over the whole region.
For example, if a region has two recommendation
domains based on Two distinct soil types, the same
research program might be implemented although
quite different fertilizer recommendations might
emerge for each soil type, hence justifying separate
recommendation domains. Generally we suggest
thinking about research priorities for-the target

region as a whole making appropriate adjustments
as necessary for the specific problems and
circumstances of farmers in each recommendation

11.1 Identifying Limiting Factors
As we explained in Chapter 3, we use
limiting factors as short-hand to refer to those
factors relating to the crop of interest which
currently limit farmers' incomes. These might
be factors which limit yield, reduce quality or
increase costs for the target crop. Also because
most small farmers store a large proportion of
their grain, storage losses may be an important
limitation. Or they may be factors such. as a late-
maturing variety which prevent planting of a
second crop immediately after harvest of the
target crop. Furthermore, at this stage we are
mainly interested in proximate limiting factors
such as weeds or insects, although we do need to
know how these factors are modified by farmers'
The process of identifying limiting factors
was an integral part of the exploratory and
formal survey. Agronomic observations in farmers'
fields with respect to weeds, pests and diseases,
are an essential starting point in recognizing these
factors. These, however, need to be supplemented
by informal and formal interviews with farmers
about their own perceptions and opinions of
limiting factors.
In noting these factors it is important to
be specific. For example, if it is an insect problem,
what type of insect is it? What parts of the plant
does it damage and when and in what years is
it most frequent? It is also useful to try to establish
what increase in production might be possible
by overcoming this constraint, as this helps rank
various factors.
In addition to identifying limiting factors it
is important to try to understand how farmers
react to these factors before proposing solutions.

11.2 Identifying Alternative Solutions to
Limiting Factors
To every limiting factor there is often a
solution and in many cases more than one. For
example, soil fertility problems can be alleviated
through rotations, land preparation practices, or
fertilizers. A weed control problem might be
approached through removal of weeds (e.g. more
efficient cultivation techniques or herbicides)
or through cultural practices which prevent
weed growth (e.g. land preparation or plant
density) or through changes in technologies in
other crops which free labor for weeding.
The researchers identify solutions to
problems based on the expertise of agronomist,
farmers and other specialists. In many cases some
innovative farmers in the area will have made
cheap and effective innovations which control
problems or exploit new potentials. The
exploratory survey and formal survey should
seek out such innovations. In other cases the
researchers might wish to call upon specialized
expertise to propose solutions to specific problems
(e.g. disease problems).
In seeking solutions to problems, it usually
should be assumed that most farmers are currently
using their resources quite efficiently and therefore
solutions must incorporate new components.
Here, hypotheses formulated about why farmers
use certain practices (see Chapter 3) will be
particularly useful. If there is a problem of time-
liness of weeding, is this because of natural factors
(e.g. too little or too much rain) or economic
factors (e.g. shortage of family labor or cash to
hire labor). Do certain problems (e.g. diseases)
relate to the type of rotation or time of planting
of the crop? If so why do farmers follow these
Thus in areas where farmers have considerable"
experience growing maize or wheat, experiments
on density, plant spacing, planting date, nitrogen
fertilizer levels, will not usually generate much
improvement over current farmer practices unless
combined with other changes such as a new var-
iety. In situations where the crop or the use of a
particular input is relatively new or where the
farmers' environment is changing very rapidly
over time, experiments to investigate levels and
timings of practices currently being used by
farmers may be useful.
If a solution is proposed for inclusion in
on-farm experiments for formulating farmer
recommendations, then inputs must be locally
available to the farmer. In many cases this might
eliminate the most efficient solution to the
problem. However, experiments might still be

planned using inputs not locally available in
order to provide information to policy makers on
the benefits of making the input available.

11.3 Listing All Changes to Farmers from Using
Technological Components
For each alternative technological component
identified as a solution to production limiting
problems, the researchers should now establish
a list of all changes to the farmer from using the
component. Here a knowledge of farmer circum-
stances will be important. It is important that the
changes be noted specifically, e.g. not only how
much additional labor but also when that labor
will be needed. (A listing of changes involved in
introducing a new weed control technology in one
area is shown in Example 11.1).

Example 11.1:
in one tropical maize area weeds were
identified as a problem arising out of farmers'
inability to conduct the first weeding early in
the crop season. Wet weather and labor shortages
were identified as reasons for this weeding practice.
A pre-emergence herbicide was becoming available
which promised to overcome this weeding
problem. The changes to farmers associated with
the use of this herbicide were as follows:

For hand weeding 12 mandays/ha in June/
July and 8 mandays/ha in August/September.
For herbicide 4 mandays/ha including car-
rying water. Labor reduced by about 15
mandays/ha with herbicide use. Labor in
peak June/July period reduced by 8 mandays/

About 75 per cent of labor for weeding is
hired, i.e. about M 700/ha in cash expenditures
in July/August. Herbicide use costs about
M 800/ha in June in cash expenditure. That is
herbicide use requires M 100 more in cash
and cash will be needed earlier in the season.

Multiple Cropping:
A small percentage of farmers intercropping

will have to change to sole cropping. Also
some residual effects are possible for those
farmers who plant beans immediately after

Probably some increase in yield to herbicide
use because of more timely weeding.

Animal Feed:
Minimal use of weeds from hand weeding.

11.4 Economic Costs and Benefits to Potential
Technological Components
The changes listed for each technological
component are now valued as far as possible
in terms of costs and benefits to farmers for all
components involving significant changes in costs.
As a.rough guide, if the total cost of the change
(i.e. all costs of inputs, labor, etc) is below the
equivalent of .2 tons of grain per hectare, this
economic analysis will not be useful because we
will not be able to measure such small changes,
nor will the farmer be able to note these yield
differences. In many cases this eliminates eco-
nomic analysis of changes in variety, planting
density and timing and method of application of
inputs, all of which are often (but not always)
low cost changes. Change involving chemical
inputs-herbicides, fertilizers and insecticides-
and equipment-method of land preparation,
planting, etc.-will usually require economic
The procedures for this economic analysis

Example 11.2:
In one wheat area a nitrogen deficiency was
noted. The field price of wheat (farmers' selling
price minus harvest and transport costs) was
M2.10/kg and the field price of urea (buying
price plus transport costs) was M3.50/kg. (urea
contains 46 per cent nitrogen). Labor to apply by
broadcasting was estimated at only one manday
/ha valued at M80/manday. Therefore the cost of
applying 60 kg N/ha was estimated at
(.60 x 3.50) +80 = M537. Assuming a cost of
capital of 35 per cent in this area and converting
to equivalent grain yield of wheat, the increase

are contained in the CIMMYT's manual From
Agronomic Data to Farmer Recommendations.
Cost data are not difficult to obtain and many
costs will be available from the survey work. It
is, however, quite difficult in many cases to
estimate benefits. Here the agronomist can choose
a "reasonable" level of the input and provide a
guestimate of what the yield increase might be.
Where there is much uncertainty different
estimates might be tried, e.g. a pessimistic and
optimistic estimate. Alternatively the researchers
can estimate the yield increase required to cover
the cost of the input and then decide what are
the chances of that yield increase being feasible.
Thistype of calculation is shown in Example 11.2.
Another guide to a possible level of inputs and
benefits will come from innovative farmers in
the area who might already be using the input.
For experiments which have the objective of
making immediate recommendations to farmers,
the current prices of the input should be used.
However, experiments which have a longer run
objective might be evaluated using other as-
sumptions on prices. For example, if fertilizer
is in short supply and has a black market price
well over the official price, the economic analysis
using the black market price might indicate that
fertilizer is not likely to pay, and therefore, should
not be included in on-farm experiments to provide
immediate recommendations. However, if the black
market price is temporary or if the researchers
want to show policy makers the benefits of in-
creased fertilizer availability, then they might still
include some fertilizer experiments in the re-
search experiments.

in yield of wheat to pay the cost of urea is
537 x 1.35 or about 350 kg/ha. Researchers
estimated that under farmers' conditions the
likely increase in yield in one recommendation
domain with higher rainfall would be at least
half a ton, thus justifying a nitrogen experiment.
The likely increase in yield in the lower rainfall
recommendation domain was less and probably
would only marginally pay the costs of the
fertilizer. A nitrogen experiment was still included
in this drier area but at only at one or two sites
and with somewhat lower levels of nitrogen.

_ -R


11.5 Matching Potential Technological
Components to Farmer Circumstances
Last but not least is the process of pre-
screening technological components against farmer
circumstances. Here all changes that farmers must
make in order to use each technological compo-
nent must be matched against farmer circum-
stances. This is particularly important in the case
of changes such as variety or time of planting, for
which no economic analysis of costs and benefits
was done. In all cases too we are aiert to possible
unacceptable levels of risk imposed by the com-
ponent on the farmer.

The matching of potential technological
components and farmer circumstances can best
be illustrated through examples. In Table 11.1 a
list of possible farmer circumstances is matched
against possible varietal characteristics for selecting
maize varieties for on-farm testing or for estab-
lishing breeding priorities. Those circumstances
favorable to a given variety's characteristics are
listed on the left and those unfavorable on the
right. To avoid redundancy we have listed a
circumstance only in one column. For example,
late season drought is favorable to earlier varieties
and it is understood that late season rains are not
favorable to an earlier variety. This is not meant
to be an exhaustive list of varietal characteristics
or circumstances.
Of course, yield is an overriding factor in
choosing varieties but the desirability of increased
yield may be modified by many other varietal
characteristics, e.g. earliness, grain type, height,
pest and disease resistance and storage quality.
For example an earlier variety might enable a
farmer to move to more intensive cropping,
e.g. two to three crops per year. An earlier
variety might also affect his risk situation e.g.
reducing risk in the case of late season drought
or frost. Storage quality is a characteristic which
is affected by the economic circumstances of the
farmer. If seasonal price swings are small and
farmers sell most maize, storage quality will be
less important to farmers, and vice-versa.
Table 11.2 matches a series of agronomic
research components in maize against farmer
circumstances. Many of these circumstances al-
ready have been considered and arise from the
list of changes involved in using the new techno-
logical component. Some changes will have been
included on the economic analysis of costs and
benefits but many will be difficult to value in this
economic analysis. For example in the experiment
to compare hand cultivation and chemical weed

control, several factors might conflict with farmer
circumstances even if the cost of herbicides is
lower than hand cultivation. Use of some herbi-
cides might not be compatible with the farmers'
rotation and intercropping practices. Alternatively
herbicides may entail a cash expense at a time
when cash is scarce. Weeds might have benefits as
animal feed. There might also be some benefits of
herbicide use if it enables more timely weeding-
for example, when the ground is very wet. These
types of costs and benefits are usually quite
difficult to value in monetary terms and a more
subjective accounting of the importance of
these changes to the farmer must be made.
Finally the proposed technological compo-
nents should be examined for their impact on cash
needs and labor requirements. In both cases the
level and timing of the requirements may be
important. The cash needs of a new technology
should, as far as possible, be minimized unless
there is an efficient credit program already oper-
ating. In general, packages which increase cash
expenses for a crop 50 per cent above cash
expenses of the current technology will create
problems for the farmer and will require
additional returns to.offset this need. Moreover,
cash expenses occurring at a time of cash in-flows
will be easier to meet than at a time when cash is
short and is needed to purchase food. Labor
inputs that occur at a particularly busy time may
also create difficulties. Higher density planting
of maize may not require much additional labor
but when combined with other parts of a
technology, such as fertilizer application at
planting, the total labor requirements for planting
and fertilization may increase by 50 per cent.
This increase could be critical to a farmer short of
labor (and cash) at planting time if there is little
flexibility in timing of planting because of weather
'So far we have considered varietal character-
istics or technological components as separate
entities. In practice, there will be strong inter-
actions between them so that we will want
to match groups of technological components
against farmer circumstances. For example,
nitrogen fertilizer may appear as a promising
component but only if a shorter variety less
susceptible to lodging is available. These two
components would then be considered together as
a potential technology.
The procedures described in this chapter
are a systematic way of screening technologies to
solve local problems. However the final choice
of technological components must be made by

the researchers in weighing the relative strengths
and weaknesses of each. Farmer circumstances
are by no means rigid. A technology that conflicts
with farmer circumstances such as labor con-
straints or drought risk may still be acceptable

to the farmer if the economic returns to the
technology are high and the conflicts are not
very severe. As the farmer may be willing to make
these trade-offs so should the researcher.


Varietal Circumstances which favor Circumstances which do not
Characteristics this varietal characteristic favor this varietal characteristic

Higher Yield All circumstances with modifications
as below.

Earlier 1. Potential for more intensive 1. Risk of mid-season drought
cropping. (e.g. flowering time)
2. Risk of early or late season 2. Move harvest into wet
drought or frost. period
3. Early season food shortage

Shorter 1. Lodging is a problem 1. Maize doubled prior to
harvest to facilitate drying.
2. More intensive technology
(e.g. N fertilizer) being

Stem Strength 1. Stem lodging problem.
2. Intercropping with climbing

Specific Insect/Disease 1. Specific insect/disease problem. 1. Cheap pesticide already
Resistance widely used.

Storage Quality 1. Subsistence production and 1. Maize largely a cash crop
traditional storage methods.
2. High seasonal price swings. 2. Insecticide used in storage.

Grain color & taste same as 1. Subsistence production.
local variety 2. Price differences based on
grain characteristics.

Easy to Shell 1. Subsistence production and
shelling by hand.


Common Types of Circumstances which favor Circumstances which do
Agronomic Experiments this type of experiment not favor this type of

Fertilizer (e.g. N,P levels)

Weed Control (e.g. Cultivation
v Herbicides)

Density/Spacing (farmers'
versus higher density and
closer spacing).

1. Intensive cropping systems
(2-3 crops/year) especially
continuous maize. Reduced
opportunities for fallow
because of population pressure.

2. New varieties available with
greater fertilizer response.

1. Obvious weed problems in
early growth stage (e.g. first
40 days) usually due to labor
bottleneck affecting the amount
and timing of weedings.

2. Cost of herbicide less than cost
of cultivation.

3. Hand weeding on time difficult
because of too much or too
little rain.

1. Availability of higher yielding,
shorter, smaller leaf varieties.

2. Farmers beginning to use
more intensive practices
(e.g. fertilizer).

3. Farmers beginning to use
machine planting.

1. Available and cheap sup-
plies of organic manures or
manures available within
the farm (no cash expense)

2. Highly variable rainfall with
considerable risk of low
yields or complete crop

1. Most weedings done by
family labor without cash
expenses (herbicides require

2. Maize intercropped with
broad leaf crop complicat-
ing application.

3. A crop immediately fol-
lowing maize is sensitive to
some herbicide residuals.

4. Weeds are fed to animals
or used for other purposes.

5. Water is not easily available
for herbicide application.

1. Considerable risk of

2. Intercropping is important.

3. Weeding by hand or animal
requiring sufficient row

TABLE 11.2 (Con't)

Common Types of
Agronomic Experiments

Circumstances which favor
this type of experiment

Circumstances which do
not favor this type of

Insecticide (application of
chemical insecticide)

Tillage Method (e.g. herbicide
zero v conventional tillage)

Time of Planting

Method and Time of Application
of Input (e.g. more precise
placement and split application
of fertilizer).

Obvious problem of insect damage
to farmer's maize in some seasons
(e.g. substantial reduction in density
of plants).

1. Problems preparing land on
time because of labor or
machinery shortage or weather.

2. Cost of herbicide use less than
cost of tractor hire (if tractors
are used).

1. Climate pattern suggests
flexibility in planting time.

2. Possibility to avoid hazards
such as disease, drought or
frost by changing date.

3. An earlier or later variety is
being introduced.

1. Most farmers are already using
the input.

Experiments for weed

1. Maize immediately follows
another crop.

2. Weeding/planting/harvesting
is shifted into a period of
serious labor shortages.

1. Change of method/time of
application would require
labor at the labor bottleneck

2. Input is expensive and needs to 2.
be more effectively utilized.

Method requires machinery.



1. Anderson, J.R. and B. Hardaker "Economic Analysis in the Design of New Technologies for
Small Farmers". In J. Dillon, G. Scobie and A. Valdez, (eds) Economics and the Design of Small
Farmer Technology, Iowa State, U.P., 1979.

(A good review of the limitations of various analytical techniques in prescreening technologies)





This chapter draws together various experi-
ences of planning research based on the procedures
presented in Chapter 11. Some of these experiences
emphasize planning an on-farm experimental
program. Others focus on priorities in varietal
development to be emphasized in on-station

12.1 Planning On-Farm :Experiments on Maize
in East Africa
Our first example is based on a tropical
maize growing area characterized by the recent
widespread adoption of hybrid sefd and fertilizer
into a farming system in which labor at planting is
one of the major bottlenecks.
In fact, most of the proximate factors
limiting maize production were due to the labor
problem. Many fields suffered from late season
moisture risks due to late planting. The 170-day
hybrids available to farmers required planting
at the beginning of rains to minimize moisture
risks later. Nonetheless, 50 per cent of plantings
were made with only 140 days of moisture avail-
ble. Although farmers started planting before the
rains on low lying areas, because of labor con-
straints they had to stagger plantings. In addition,
many fields were damaged by water-logging early
in the season because farmers switched from tra-
ditional ridge planting to planting on the flat
which required less labor and enabled farmers to
speed-up planting. Weeds were also a problem.
Fifty-five per cent of fields were weeded only
once and this was when the maize was at an
average height of 60 cms. In this case, weeding of
maize conflicted with later plantings of earlier
maturing subsistence crops finger millet and
beans. Also, the second fertilizer application was
made after the first weeding when maize was
already 75 cms high-again due to labor shortages.
Finally, unrelated to the labor shortage, many
maize fields suffered from stalk borer damage late
in the season although few farmers were using

One approach to alleviating the problems
of late planting, water-logging, weeds and late
fertilizer application would be to find ways of
reducing the labor constraint, such as use of
tractor or oxen cultivation or herbicide use.
However, there was evidence that farmers faced
a severe cash constraint and that solutions
requiring considerable cash would compete with
fertilizer use in maize. Fertilizer purchases
represented 54 per cent of cash production costs
for maize, and 25 per cent of farmers' cash in-
comes. Moreover, custom oxen and tractor services
were being used by a few farmers and it was felt
that other measures to increase cash incomes of
farmers would enable more of them to use these
The immediate solutions, therefore, centered
on selection of an earlier variety which could be
planted according to farmers' current planting
schedule and mature by the end of the rains.
Earlier varieties were available for testing on
farmers' fields. These varieties were somewhat
shorter, probably requiring higher densities,
hence variety X density experiments were included
to determine optimal densities. Experiments
were also designed to determine best use of
available fertilizer with earlier varieties, e.g. time
and method of application. Finally, an insecticide
experiment was designed to determine if there
was an economic response to insecticide treatment
of stalk-borers, These experiments were designed
for implementation on representative farmers'

12.2 Planning On-Farm Trials in the Andean
A survey of farmers in a highland maize
producing area of the Andes showed that one of
the major potentials in the area was an earlier
variety to enable farmers to plant a second crop-
peas or lentils. Sixty per cent of farmers preferred
an earlier variety even if yields were reduced.

By asking farmers about the trade-offs between
earliness and yield losses it was estimated that a
variety about 5 weeks earlier would best suit
their needs and that they would be willing to use
such a variety even if it yielded up to 20-25
per cent less than current varieties. Varieties
meeting these earliness/yield requirements were
selected from available varieties being developed
on-station and were included in on-farm varietal
experiments. Of course, in this' situation the
successful adoption of an early maize variety
might lead to a reduction of maize production
but, more importantly, farmers' incomes would
increase as a result of the second crop.
In addition to variety, researchers diagnosed
the proximate factors limiting production as
weeds, fertility and insect damage. Since weeds
were an important source of animal feed in the
area, it was not considered feasible to use herbicide
weed control methods until an alternative forage
source was found. One such source is the stripping
of maize leaves and tassel and the thinning of
maize plants. However, almost all maize is inter-
planted with local beans which, because of their
aggressive climbing habit, prevented leaf stripping.
It was therefore decided to look for beans with a
different growth habit, that would allow some
stripping. This bean type also gave more flexibility
in choosing early maize varieties which were not
adapted to intercropping with the local climbing
bean. At the same time the breeding program
began to look for maize varieties which provided
tillers which could be removed early in the crop
cycle to feed animals.
Most farmers were applying some animal
manure but this was insufficient to sustain high
maize yields. Few farmers were using chemical
fertilizer. The on-farm experiments therefore
included experiments to determine economic
doses of nitrogen and phosphorous.
Insect damage from ear worm was not a
major problem but still was felt to contribute to
a yield loss of from 10-15 per cent or about
200 kg/ha. Potential insecticide treatments were
then prescreened to identify treatments with a
cost of less than 200 kg/ha in grain equivalents.
Costs included in prescreening the insecticide
treatments included the cost of the insecticide,
the hand sprayer, the labor for application and a
25 per cent capital charge on these costs. The
procedure was similar to that given in Chapter 11.
The above experiments-variety, fertilizer
and insecticide-were designed so that the non-
experimental variables reflected farmers' practices.

information on farmer practices was obtained
from the survey and generally showed that
representative farmer practices were: maize
intercropped with beans, fertilization with animal
manure, no insect control and irrigation only
in some recommendation domains.
Finally the survey helped in choosing sites
for locating the experiment. Information on slope,
soil texture and irrigation helped establish
characteristics representative of farmers in the
region. Moreover, farmers were asked in the
survey to host an on-farm experiment. This
provided a long list of farmers from which to
choose sites.

12.3 Guiding Research on Tropical Maize Varieties
in Dry Areas of Eastern Africa
In a tropical maize area of Eastern Africa,
breeding efforts on maize had already focused
on finding earlier maize varieties to better fit the
relatively short period of 75 days of reliable
rains. Current farmers' varieties required 115-120
days to mature and therefore often suffered
severe losses when rains started late, when there
was a mid-season gap in rains or when the rains
finished early. A survey of farmer circumstances
in the area diagnosed other elements of the
farming system which reinforced the need for
emphasis on early varieties. First, farmers largely
depended on maize as an early source of food in
the critical period before other crops were
harvested. An earlier, more reliable harvest
would suit farmers of the area even better in satis-
fying food needs in this period. Second, early
maize varieties planted on low-lying areas would
increase the potential area and reliability of a
second crop, such as beans planted on residual
moisture immediately after maize. Third, the
planting of the main crop of an early variety of
maize could be done later when rains were more
reliable and relieve current labor bottlenecks for
planting and weeding that farmers experienced
with present varieties. This might enable greater
areas to be planted or better management practices
to be carried out, using the cash saved from
hiring labor in the peak labor period. Finally,
with an early variety the increased reliability of
a maize crop would reduce the need for planting
security crops such as sorghum and cassava and
again provide additional resources for increasing
the area and management of the preferred food
crop, maize.
The survey also uncovered other charac-
teristics of a desirable variety to farmers in the
area. These included resistance to lodging, since

ears of lodged plants were often damaged by
rats in the field, storage quality, since maize was
a staple food eaten throughout the year, and
palatability of the varieties when they were

processed into the preferred local maize fooas.
These characteristics could then be used to
prescreen early varieties of maize prior to testing
on farmers' fields.



* The following are empirical applications of the procedures described in this manual.

1. Collinson, M.C. "Demonstrations of an Interdisciplinary Approach to Planning Adaptive Agricultural
I Research Programs Serenje District, Zambia" Report No. 3 CIMMYT Eastern African Economics
Programs, Nairobi, 1978.

2. Harrington, Larry "Farmer Practices and Problems in Northern Veracruz" CIMMYT, El Batan,
Mexico, June,1979.

3. Benjamin Alan "The Agro-Economic Context of Maize Production in Three Valleys of the Peruvian
Sierra" (forthcoming) CIMMYT, El Batan, 1979.



Agro-Climatic Environments: Areas (not necessarily contiguous) where a crop exhibits
roughly the same biological expression, so that we would obtain, for example, similar
variety or fertilizer responses within a given environment, everything else being equal.

Base Practices: Management practices which are generally representative of practices of
farmers in a given Recommendation Domain. These practices serve as a reference for
comparing potentially improved technologies against farmers' present technology in
on-farm experiments.

Best-bet Components: Those components which result from the prescreening process that
promise significant increases in incomes at reasonable levels of risk within the resources
available to farmers. I

Exploratory Survey: A process by which the researchers traverse the target regions and
informally interview farmers and other persons knowledgeable of agriculture, in order
to arrive at a tentative understanding of farmers' existing technology for the target crop
and constraints limiting farmer's production and income.

Farmer Circumstances: All those factors which affect farmers' decisions with respect to use
of a crop technology. They include natural factors such as rainfall and soils and economic
factors such as markets, the farmers' goals, and resource constraints.

Farming System: The total of production and consumption decisions of the farm-household I
including the choice of crop, livestock and off-farm enterprises and food consumed.

Formal Survey: A survey of randomly chosen farmers who are interviewed by trained
interviewers using a written questionnaire in order to provide quantitative data on
farmer circumstances.

Limiting Factors: Those proximate factors such as weeds and pests which limit farmers'
production and incomes.

Management Practice: The actual use of a technological component defined in terms of the i
type, amount, and timing of the component.
New Technological Components: Practices or inputs which are yet to be developed or whose
performance under farmers' conditions cannot be predicted with confidence. Examples
are varieties yet to be created or new herbicides with which researchers have little or
no experience.

On-Farm Research: Research in farmers' fields with farmers involved to formulate improved
technologies. There are typically two types of interrelated activities: a) surveys of farmer
circumstances, and b) experiments.

Prescreening Technological Components: The process of choosing from many potential
components, a few components for on-farm experimentation which address critical
farmer problems and which are feasible given farmers' circumstances.

Random Sample: A sample drawn so that every unit in the population or sub-population has
an equal probability of being selected.


Recommendation Domain: A group of roughly homogeneous farmers with similar circum-
stances for whom we can make more or less the same recommendation. Recommendation
domains may be defined in terms of both natural factors (e.g. rainfall) and economic
factors (e.g. farm size).

Stratification: The process of dividing a population into relatively homogeneous subgroups
in order to increase sampling efficiency. Stratification follows as closely as possible the
definition of recommendation domains.

System Interactions: Interactions between different crops, livestock and non-farm enterprise
of the farming system which influence the choice of technology for the target crop-
Sfor example, the planting of a high density of maize so that thinnings can be used to
feed livestock.

Target Crop: A crop which is currently, or has potential to be, a major crop in the system
and for which there are available technologies with potential to increase farm production
and income. In this manual, the examples always refer to maize or wheat as the target

Target Region: A region chosen for an on-farm research program. The choice of the region
may depend on crop production potential, government goals with respect to income
distribution and the available infrastructure for doing research in the region.

Technological Components: A specific part of a technology such as variety, fertilizer, or

Technology: The combination of all the management practices used for producing or storing
a given crop or crop mixture.

Two-Stage Sampling: A sampling procedure in which sub-populations such as villages are
first selected and then units such as farmers chosen within each selected sub-population.