Farming Systems Research
in West African Agricultural Development
Professor of Agricultural Economics
Farming Systems Workshop
January 12-15, 1981
USAID Contracts AFR-C-1257, 1258 and 1472
Farming Systems Research
in West African Agricultural Development
There are two problems which might be called basic: the self-
sufficiency problem requiring an increase in the productivity of the
agricultural sector, either to attain and maintain food self-suffi-
ciency or, using its comparative advantage, to attain economic self-
sufficiency. In the future of the food supply problem in this region,
John Mellor (1980) is rather gloomy:
"The most difficult food problems of the next decade (the 1980's)
are likely to occur in Subsaharan Africa. Demand, generated by high
population growth rates (among the highest in the world) and rapid
urbanisation, will be difficult to meet in the face of declining growth
rates in area cultivated."
A second basic problem is that an increase in the productivity and
so, hopefully, the income of the rural sector (which constitutes the
majority of the population) is needed to increase demand for goods and
services from the urban areas. Only in this way can economic develop-
ment on a national scale be achieved (W. Arthur Lewis, 1980).
Reviews of the large investments in livestock and agricultural
development projects (Uma Lele, 1975, Development Alternatives Inc.,
1977?,CILSS/Club du Sahel/OECD, 1980) indicate that a high proportion
of the efforts fail. This is mainly attributed to the promotion of
technical packages or of farm practices which are either not profitable
or not feasible for the farmer to adopt under existing circumstances.
A second major constraint is lack of discussion with the farmers by those
planning the projects and a lack of knowledge of the existing farming
The concept is being accepted that even if the presently proposed
rural developments were financed, this would not result in a satisfactory
level of rural development. Improvements are required in conception
and execution before satisfactory results will be obtained.
Nonetheless, there are examples of "operations" which have succeeded.
Cotton production in Central and West Africa* has increased at the rate
of 10.4% a year over the last 15 years and between from 1954-1970
coffee production increased at over 8% a year. Maize production has
approximately kept pace with the increase in population (2 1/2% a year).
Millet and sorghum production has increased at less than the rate of
increase in population (1 1/2%/year), a decline of 1% a year in per
capital production. This probably indicates a decline in demand for
sorghum and millet, replaced in part by wheat and rice. This needs study
because otherwise attempts to increase the marketing and production of
sorghum and millet will come to naught (CILSS et al, 1980; W. Arthur
Lewis, 1964); such attempts are called for in the strategy for develop-
ment of rainfed agriculture.
"Farming" in this region we consider to consist of the exploitation
by the farm families of all the ecological niches available to them,
whether they be rainfed or irrigated agriculture, livestock production
or fishing. The farming system is considered as the set of economic
activities of the farm family including their activities both as a
production and as a consumption unit. There can, of course, be a dis-
cussion on what comprises the family farming unit, an extended or a
nuclear family; the definition of the unit as those who eat from the
common pot is attractive but in some cases it changes between the rainy
season and the dry season. A serious study of family organisation,
which is necessary to understand the farming system, will reveal the
appropriate unit for study.
The activities within the farming system within a region depend
upon the types of "land" available for example rainfed, basfonds,
flood plain recession, river bank recession, irrigated and so on and
the soil type, usually the texture, sandy, silty and clay. The treat-
ment of a field also depends upon whether it is very close to the house
(e.g. kitchen garden), close enough to be manured regularly, more re-
mote or a bush field. The relative importance of each of the "types"
(excluding irrigation) will vary from year to year depending upon the
local rainfall and its distribution or upon the flood, which in turn
depends upon the rainfall somewhere else in the river basin.
Families differ in their access to the different types of land,
for example due to caste, the lineage, and the seniority in the lineage
and perhaps ethnic group. Families also differ in the quantity and
quality of their available resources other than land labor, capital
(short term inputs and durable inputs) and in their managerial ability
to use those resources productively.
The concept of a suitable technical package or practice must there-
fore be related to the type of farming appropriate to each ecological
niche exploited and to the type of farm, represented by the quantity
and quality of the resources available to them. This implies a defini-
tion of the recommendation domains for the different ecological niches
and also a typology of the farms. This, of course, becomes quite com-
plex. When, for example does a recommendation domain end and another
The analysis of this situation can be made using traditional and
modern farm management techniques, assuming adequate understanding and
data. However, these are only rarely available in this region and then
for specific ethnic groups and specific areas. Where the data are avail-
able modelling seems to be particularly suitable to exploring the
boundaries between recommendation domains, and types of farm.
Modelling can also be useful in the so-called "upstream" Farming
Systems Research (FSR) (Norman et al, 1980), whose objective is stated
as "finding out how to overcome major constraints common to a range
of farming systems extending across one or more geographic zones,"
by developing proto type solutions to increase productivity. The extent
and the range of constraints and the effect of different possible sol-
utions on productivity can be studied by modelling.
SThe West African research done by Purdue mostly falls into the
category of the descriptive and diagnostic of downstream FSR as defined
by Norman (1976) and Gilbert et al (1980) (see fig 1). Four stages are
defined as follows:
1. the descriptive or diagnostic state in which the "system"
is investigated in the context of the total environment to
identify the constraints and the flexibility in the system
(labor and other unused resources). The goals and motiva-
tions of the farmers and their family members might also be
studied and further constraints or motivation in improving
the system identified.
2. the design stage in which a range of strategies is identified
that may be relevant in dealing with the constraints identified
3. the testing stage in which a few promising alternatives from the
design stage are evaluated under farm conditions (a) under joint
researcher and farmer control and (b) only under farmer control.
4. the extension stage in which the strategies in the testing
stage are demonstrated to extension and the successful strategies
The conception of the Purdue research arose from the frustration of
trying to identify, design and evaluate projects with the current level
of understanding of farming systems.
An attempt was made to carry out detailed input-output surveys,
using the Norman (1974) model in the major ecological zones within the
region. The only other attempt of this nature was that proposed in
1964 by Sautter and Pelissier, which naturally was more oriented towards
geography. Our attempt was also much more constrained in funds, in
the qualifications and experience of the personnel and the length of
time available. It was a completely new type of research for us. It
was oriented towards socioeconomics and the study of the farm family
in all of its economic activities. The input-output format, success-
fully executed, would permit modelling of one year's data. We recog-
nised the limitations of a single year or even two years of data but
our funding and the contract system forced us to accept the limitation.
As the institution involved is a university, training would
clearly be one of the products of the contract.
The sites were selected in discussion with the national govern-
ments, USAID Missions and AID/Washington. In each case our research
workers were assigned to a national or local agency. The areas
selected were as follows:
Matam and Podor on the Senegal River
Thies/Bambey and Kaolak in the Peanut Basin
Zuigenchor, lower Casamance
Mopti region, Mali
Sikasso including basfond production
Kaya, Upper Volta
Volta Valley, AVV Villages, Upper Volta
Bentange, in the AVV region, Upper Volta, Niger
NDjamena, irrigated gardening.
Some other studies were made including:
a preliminary study of goal programming in the Senegal peanut basin
a study of the effect of a decorticater and a mill in the work of
women, near Bambey, Senegal
an agricultural sector model for Senegal, comparing policy
a flexible linear programming farm model for the Sahelian countries
a handbook for cost/benefit analysis of irrigation projects in the
Partial support was given to a CRED study of food purchases in
Dakar and a rural area near Bambey, Senegal.
In addition the research also provided:
a preliminary model for development project monitoring providing
the basis for a book being written under a separate contract.
a system of microcomputer data processing for FSR.
training of 11 ex PCV's in the MS program in agricultural economics,
4 other MS and 2 Ph.D.'s in agricultural economics at Purdue, 6
Ph.D.'s or 3rd cycle doctorats in other institutions.
training of seven professional staff in the project (already having
training of 6 francophone african students to the MS level.
It is meaningless to try to summarize the results of the research
program in the time available at this meeting so the comments will be
restricted to the methodology. Sampling is always a problem in socio-
economic surveys; it was not proposed that the survey samples would be
representative of the village or of the region. However, typically a
village census was made listing major variables; subsequently a random
sample was selected from the different strata identified in the census.
It was recognized that only after the surveys had been made would it be
possible to draw a representative sample.
Draft census and survey forms were supplied to the research workers,
at least to serve as a scope of work. There was a lack of uniformity in
the forms actually used and in the degree of coding used by the survey
workers and the researchers in the field. Ideally, in hind sight, we
would have profited from more uniformity in the forms used and failure
to encode the data either by the survey worker or an additional clerical
worker at the time of observation or shortly thereafter caused many problems
and some loss of data as well as a tremendous loss of time. We would
now recommend recording the data on microcomputers using floppy disks,
stringy floppy (high speed tapes) or similar devices. The program de-
signed for recording the data should include an editing feature or the
data should be subjected to editing or "cleaning" in the capital city
as soon as possible after recording.
With the capacity of microcomputers available today, at relatively
modest prices ($3000-5000), tabulations can be provided in the field only
a few weeks after the data has been collected. In this way the field
researcher can identify problems and can clarify results that are not
The comprehensiveness of the data obviously depends upon the ob-
jective of the survey. A baseline or descriptive FSR survey requires in-
put-output data for the economic activities of the family, and sales and
purchases, at least of food stuffs. The greatest problems tend to be
associated with field identification and measurement, yield measurement
and family budgets. In our case, it was not always possible to survey
and measure the yield of every field and sampling was used. Some re-
searchers obtained good budget data and others did not. This was not
associated with ability to speak the local language but with the quality
of the survey workers their instruction and supervision.
We used two different types of organisation of the supervision in
gathering data, with a technician living in the village or nearby and
supervising 3 to 6 survey workers (six is probably too many) and with a
technician living further away, sometimes using an intermediate level of
local supervisor. It seems that for baseline studies we have a greater
probability of really understanding the farming system using the former
system than the latter. For more limited follow up surveys the latter
has a lower cost and with well qualified and trained interviewers and
supervisors can work satisfactorily.
A major variable which was not considered at the time of sampling
is managerial ability of the head of the family. Predictive measures
of this are not known; all measures normally being based upon performance.
It is probable that an estimate could be made by the estimates of certain
defined indicators, perhaps yield per hectare or yield produced per worker,
made by other villagers or, in the cotton zone, by the extension workers.
Any attempt to develop production functions (equations predicting
the relationship between inputs and output) is much more likely to
succeed if the farmers can be categorised in terms of managerial ability.
Such quantitative relationships form the basis of farm management advice
and of mathematical modelling, which permits easy and rapid testing of
the effect of changes in the farming system.
Averaging of the results usually disguises the most interesting
information. For example, on average at most of our research sites
intensiveness (e.g. use of inputs such as fertilizer) in the production
of cereals and peanuts does not seem to be as profitable as extensifica-
tion (e.g. increasing the area cultivated per worker). Cotton production
in some respects is an exception. However, it is clear that some farmers
can profitably intensify their production of cereals. This is a very im-
portant relationship to understand, since intensification is proposed
as the appropriate technology in almost all agricultural development
Once the typology and the nature of the production functions are
adequately defined or a mathematical model has been developed, the
survey can be quite specific and gather only the data required for solving
the problem. Major problems will continue to be measurement of the field
areas and yields.
Our research did not attempt to study the year-to-year variability
between the yields for the different types of land farmed by the different
family members with different crop associations. The difference in
the labor and other inputs into the different types of farming (e.g. rain-
fed, basfonds, flood recession, irrigated, etc.) from year to year.
Understanding of this variability is an important step in understanding
the decision making process of the farmer family.
Another subject which we attempted to address was the relative im-
portance of different goals in the family and their effect on the farming
system. The start made in our research was carried on in other studies
in Latin America and in the Philippines. For lack of better definition
of goals, the error is not believed to be very great from modelling
the family farm in our surveys in West Africa assuming a profit max-
imizing goal, with a constraint on the head of the household to plan to
produce enough food to meet family needs and a constraint on his risk
taking. This was done in the Sahel Farm Model.
A major fault in our work and that of others in FSR is the long de-
lay in publishing the results of our research. To be an effective part
of a FSR effort it will be necessary to provide preliminary analyses soon
after harvest and fairly detailed analysis in time to play the activities
for the next rainy season. We believe that our methodology would permit
this if the data are encoded and recorded while the survey is proceeding
and the data are "cleaned" shortly after being recorded. This appears
to be a matter of organisation rather than technology. It certainly
imposes too great a delay to gather the data in the field and bring it
to the University for encoding, recording and cleaning. In fact only
the most sophisticated analyses need to be done outside the country in
which the data are collected.
A problem which has not been sufficiently studies to date is the
management of the survey data on the microcomputer. In a baseline survey,
if the data are summarized, for example, by time period for each farm and
the time-period summaries are accumulated over the season, the data manage-
ment problems are minimized. If every line of data from the survey forms
continues to be used in the analysis throughout the season, rather large
storage capacity is required (on a mainframe computer) 2400 foot magnetic
tapes are used. Even so starting each seasonal analysis using all the
lines of data, requires a very large data management system. We used
Bill Elbring's Tapelog Utility System and still had problems from the
vast number of lines of data stored. The basics of suitable data manage-
ment systems are available for microcomputers but they have not yet been
used in analysis of FSR.
Identifying and Testing Innovations
In the conventional model (fig 1), following the identification of
constraints in the baseline survey the research workers in the station are
expected to propose solutions which can be tested in villages and on
farms. It is certainly not desirable to propose the field experiments
before the constraints have been analysed. It is also quite likely that
the research workers cannot propose the solutions to the constraints
quickly or even in a few years. For example, in the US Midwest there has
been no success in solving the constraint of maize yields in stations or
on the best managed farms for about 15-20 years, inspite of considerable
efforts and investments in both the public and the private sector.
This does not mean that the average productivity, yield per hectare
or whatever, is not increasing; as the average and below average farmers
catch up with the best farmers improvements are still apparent. The
same is certainly true in several countries in West Africa; these are
groups of elite farmers, who, when they can reliably obtain modern inputs,
have developed systems for their profitable employment. Improvements
can be made using FSR to identify the best practices appropriate to the
different strata in the typology and to show the practices being used on
farms to the appropriate groups of farmers and to extension workers.
This has been done in developed countries and has been proposed for
developing countries (e.g. Norman et al, 1980, pp 14). It leads to the
exploitation of indigenous knowledge.
The Purdue research was a socioeconomic study to provide a better
understanding of farming systems and an economic evaluation of the improved
practices being promoted by extension workers. It would have profited
from more assistance from biological scientists in measuring levels of
biological variables, such as pest and disease infestation, mineral defi-
ciencies and so on.
Uses of FSR
FSR is in danger of becoming considered as the panacea for failures
in agricultural development. Some large projects are being proposed,
perhaps exceeding capabilities to man them. As has been mentioned above,
the assumption that stage 2 of the conventional farming systems model,
"research" solving the problems causing constraints within the term of
the project, is generally not valid. Under African conditions there seem
to be few, if any, examples of a 5-year research program answering a major
constraint. In irrigated rice production the situation is different,
both in S.E. Asia and Africa. There seems to be a much wider adaptation of
varieties presumably because of the greater degree of control of conditions.
In discussions at Purdue with biological scientists (INTSORMIL
Workshop, May, 1980) the agronomists frequently insisted that the constraints
to increased productivity were not agronomic, presuming that varieties and
technologies existed for increasing yields; the fact that these technologies
were not profitable in the hands of most farmers did not convince the
agronomists that more agronomic research was required. Other agronomists
frankly stated that it was very difficult to improve on the local varieties
unless the system of production, level of fertility etc. could be improved.
In fact, it seems that resistance to pests and diseases could be'improved
by crossing exotic resistant material with traditional varieties, which
often already have a certain level of resistance. Specificity may be a
problem: for example, resistance to Indian strains of striga does not im-
part resistance to African strains. For these reasons, I am not very
optimistic that the traditional agricultural research in food grains, with
a major emphasis on breeding, work calendar, fertilizer and tillage methods
will be any more productive in the future than in the past (Anthony et al,
1980). FSR can certainly serve as a link between research workers and the
farmers and as a means of collecting and cataloging the indigenous know-
ledge and making biological research workers aware of its significance
(e.g. crop associations).
In the West African region large sums are being invested in Pest
Management programs, which mainly consist of training and setting up of a
system of scouts to find pest infestations and spray teams to combat the
pests. A limited amount of research is carried out to study the losses
caused by the pests. It is clear that pests and diseases cause considerable
losses-in crop production and there seem to be two important areas of inter-
vention for socioeconomic and/or FSR: inventorying the indigenous knowledge
of farmers on pests, their habits and their relative importance* and
participating in integrated pest management field studies to identify
the economic threshold for intervention with spraying or dusting and the
economics of the alternatives in the integrated pest management approach.
Mathematical modelling has been used for this purpose in developed
counties. I believe that this may be the best chance for increasing
productivity in food grains production in this region within the next
decade. It is important to note that the level of managerial ability of
farmers will affect their ability to handle some of the alternative pest
The socioeconomic research workers using their data in mathematical
models can identify research priorities in terms of the research efforts
that would produce the greatest benefits. A similar approach can help
agricultural policy makers to study the alternatives open to them and
the likely results of their actions.
The FSR approach, or its socioeconomic component, is being used in
several agricultural development projects as a tool for monitoring.
Projects include the AVV in Upper Volta, and IBRD projects in Northern
Nigeria, and in Tanzania. With the introduction of the micorcomputer the
average delay of three years in the analysis of data (for example in the
IBRD project monitoring units) can be eliminated and the results can be
produced (as in the AVV unit) in time for the next agricultural season.
This provides a means for the staff of a development project to bring the
strengths of farm management research and FSR to bear on improving the
content and effectiveness extension program. The development operation
has the great advantages of having its agents distributed throughout the
area of intervention. These agents could be taught to work with monitoring
at the sector and perhaps subsector level. Research institutions seem to
have a greater problem in getting qualified people to live and work in
rural areas, perhaps because the level of qualification is excessively
high (i.e. ingenieur or matrice).
Socioeconomic research can also provide the development operation
scientists with a means of dialog with the farmers, a typology of farmers,
a catalog of indigenous knowledge including the practices of elite farmers
in all classes of the typology, and a means of measuring the impact of
the operation in different areas. Weaknesses in the extension chain can
be identified and hopefully strengthened, because it is not sufficient to
have an appropriate technical package, it must also be delivered appro-
priately and the pricing system must be such that the farmers find the ex-
tension program leads them to the greatest profit. For example, if an
improved system of rice production is proposed but the price of paddy is
such that selling paddy does not bring the greatest return to the limiting
resource (usually labor), the farmers will not sell much rice but will sell
maize or peanuts or whatever brings them the greatest expected return.
*Reports on a Nigerian project for combatting post harvest losses by killing
insects in granaries but the farmers considered that their greatest losses
arose from termite damage in the wooden legs of the granaries causing
the whole granary to collapse exposing the grain to the weather, rodents
and birds. This caused a reorientation of the project.
As has been proposed by the FSU of the SAFGRAD/OAU project, it is
certainly possible to train some people in the villages to keep farm
accounts, for their own farm and perhaps for a neighbor. Farm accounts
have been the basis for farm management extension and for farm policy in
most developed countries. So it is not too early to start them in this
region. In Indiana we use a discussion of the farm account with the
farmer, and a comparison between his farm and others in his strata in
the typology, as a means of teaching the farmer the principles of agricul-
tural economics, so that he is better able to make decisions on his own.
There is little doubt on the stagnation of African agriculture over
the last 15 years and the decline in the per capital production. This is
predicted to cause "the most difficult food problems of the next decade"
the 1980's by Mellor (1980). Cotton production in the Sahelian states and
coffee in the coastal states are the two exceptional performers in agricul-
ture. The good performance of cotton and the poor performance of peanuts
and other crops has not yet been explained.
The conception and methodology of the Purdue Farming Systems Research
in West Africa under AID contracts AFR 1257 and 1258 has been briefly
reviewed and the methodolical experience in developing, encoding, recording,
correcting and analysis of the data discussed. The microcomputer has now
become an import tool in this type of research.
The potential of the microcomputer to shorten the analysis time from
2-3 years later to within the same year as the survey opens up new poten-
tial for FSR and socioeconomic research in agricultural development. The
methodology can serve both and also can make agricultural and livestock
development project monitoring a practical reality.
FSR as normally defined requires a rapid solution of constraints to
increased productivity to come from agronomic research; experience in
developing and developed countries causes one to doubt that this is really
feasible. However, in the short and medium term indigenous knowledge and
integrated pest management offer possible alternatives or complements.
There is an urgent need for a practical system of classification of
farms and farming in order to develop a set of recommendations suitable for
each class. In developed and developing countries recommendations for
improved practices for the average farmer are not useful.
Mathematical modelling of type farms can help in classification, in
developing recommendations, in establishing research priorities and in
formulating farm policy.
Anthony, K.R.M., Johnston, B.F., Jones, W.O. and Uchendu, V.C. Agricul-
tural Change in Tropical Africa, Cornell University Press, 1980.
CILSS/Club du Sahel/OECD Strategy for Drought Control and Development
in the Sahel, OECD DAC, Sahel D (80) 102, September 1980.
Development Alternatives Inc., Report of a Study on the Success of Rural
Development Projects for USAID/TAB, about 1977.
Gilbert, E.H., Norman, D.W. and Winch, F.E. Farming Systems Research:
A Critical Appraisal, M.S.U. Rural Development Paper, No. 6, 1980.
Lele, U. Design for Rural Development. Lessons from Africa, IBRD/
Johns Hopkins University Press, 1975.
Lewis, J.A. Reflections on Nigeria's economic growth, Paris, 1967.
Mellor, J.W. Commentary: Africa Depressing Trends and a Difficult
Task, IFPRI Report, Vol 2 (2) 1, May 1980.
Morris, W.H.M., editor, INTSORMIL Workshop Proceedings, Purdue University,
May 22-23, 1980, in process.
Norman, D.W. Farming Systems Research in the Context of Mali. In Pro-
ceedings of the Farming Systems Workshop, Bamako, Nov. 16-20, 1976.
Norman, D.W. An economic survey of 3 villages in Zaria Province. Ahmadu
Bello University, Institute of Agricultural Research, Samaru,
Misc. Paper No. 37, 1974.