WORKSHOP ON METHODOLOGICAL ISSUES FACING SOCIAL SCIENTIST IN ON-FARM/FARMING
Report of rapporteur on session: "General overview of FSR and potential
role of social scientists"
(Paper delivered by David Norman, KSU)
Major issues raised in main session and small division group were:
Many participants were not happy with upstream/downstream research con-
cept however, many used the terms correctly.
During the discussion group meetings the only acceptable alternative
appeared to be long-run research vs. short-run research.
Following the rapporteurs' reports other alternatives were presented
Support research and directly applied research
Generation research and adoption research
Developmental research and adaptive research
Source research and adaptive research
No cornensus was evident.
2. ON-FARM RESEARCH AND FARMING SYSTEMS RESEARCH
This distinction was clarified for many participants by the paper by
Norman. However, doubt persisted in some minds.
On-farm research is not necessarily farming systems research (much
on-farm research is commodity research or reductionist research).
Farming systems research is not necessarily all on-farm.
3. ROLE OF ON-FARM RESEARCH
At CIMMYT, we feel that on-farm research:
a) permits the derivation .of recommendations appropriate to farmer
b) permits feedback to guide upstream (long-term) research.
c) producer guidelines for policy changes.
Obviously, CIMMYT is focused in more on the national program situation
than some other programs.
4. DISTINCTIVE FEATURES OF FARMING 'SYSTTES R'FJ.EC" ;
a) Recognizes technical, economic and socio-cultural constraints
b) Crosses both discipline and commodity lines.
It is not unique is one component of the system. FSR is not compe-
titive with reductionist and commodity research it is complementary
5. COST EFFECTIVENESS OF ON-FARM RESEARCH
Resources are always short particularly in the national program situa-
tion so this aspect featured in all thediscussions.
It was claimed that on-farm research more efficient than on-station
research in terms of technology adopted per unit of money spent. It
is probably time to produce numbers on this, but again it is not an
It is time to demonstrate, that cost effectiveness of on-farm research
can be improved further before funds for such activities dry up.
A healthy trend is that on-farm research is being tailored to farmers'
conditions more and more i.e. it is no longer simply on-station research
transferred to the farmers' fields.
6. ROLE OF THE SOCIAL SCIENTIST
Again, the need is for complementarity between biological scientists,
economists and other social scientists it is not a perfectly binary
Various roles were given by different participants (these roles were
generally for the economist rather than the non-economic social scien-
tists) as follows:
a) to evaluate technology developed by the biological scientist dan-
gerous for the social scientist to set himself up as an arbiter but
the point was made that the economist is often specially qualified
to be the coordinator).
b) to provide survey techniques
c) the economist is sensitive to the time frame and to the concept of
d) the social scientist has an important role in policy issues
In the regular sessions the role of the non-economic social scientists
was largely ignored until the last day. However, in the small group
discussions a strong case was presented for the anthropologist to play ;.
a role in farming systems research because the anthropologists approach
is holistic at the technical, economic and socio-cultural levels. It
was felt that the anthropologist should play an important role at the
policy level. The anthropologist could also play a vital role in informal
methods of analysis where information is largely qualitative rather
In some quarters, it was felt that the major role of the social scientist
is to make the biologists aware of things they have failed to take into
account but not to make complications unnecessarily. When we are
working at the national program level simplicity is essential.
Surary of discussions of whole group and small
group on "On-Farm Experimentation".
1. Approach. Cropping System or Component Crops
Three alternative approaches were suggested:
a) Study the whole cropping system from the start.
b) Start by looking at the cropping system and deciding whether
one can change one component without major changes in the
system as a whole. If this is possible, it is justifiable
to concentrate on one component, but if a large interaction
between components exists, one must work on the system as a
c) Start with a pre-determined commodity and decide what changes
need to be made in the production of that commodity, then
decide if these changes can be acccmodated in the system or
sub-system (e.g. weeds for fodder).
2. Number of Experimental Variables in the Experimental Program
Numbers quoted during the discussion varied between three and thirty.
The actual number will depend to a large degree on the background
technical knowledge and the results of a survey.
Once the number has been decided these shall be included in screening
trials. Trials suggested were 2n (n = 3 to 8) and 3n (n = 2-4). Where
more experimental variables are to be treated in the whole program than
can be accommodated in a single trial, variables should preferably 'overlap'
(i.e. be included in more than one trial with different variables).
3. Levels of Non-Experinental Variables (NEV's) or Background Matrise
Three possibilities: Farmers level
Same intermediate level. This may be
a future package or profitable level.
The choice depends on the client (farmer or policy maker and/or the
time dependency of the research.
NEV's may be set at non-limiting levels for longer-run research or
demonstrations for policy makers (upstream), e.g. evaluation of yield
depression and importance of a disease or demonstration of the effect of
an unavailable input.
NEV's should be set at farmers level in trials aimed at near-term
or short-run recommendations, but may be raised to a probable 'future
package' level as the research continues.
4. Level of Farmer Practices Actual Farmer or Average Farmer
The average farmer probably does not exist and therefore trials should
be placed on the farms of representative farmers and replicated over sites,
although this replicaton does increase the expense of the program.
The more heterogenous the zone or recommendation domain, the more
replication over sites is needed, and the more 'noise' or variation in
control yields, but this variation may be a useful component in making
overall recommendations for the domain, and may help to re-define the
5. Package or Single Component
The decision depends on the profitability, acceptability and interaction
between components. Where a package is advocated, the priority of each
component, based on the above three points, should be known.
It my be a good strategy to use a near variety as a vehicle for a
package to avoid the problem of telling to the farmer that his normal
practices are deficient. However, sane members stressed that there is
nothing wrong with telling farmers that one is looking for changes to
increase productivity and profit, and therefore there is little risk of
farmers rejecting recommendations off-hand.
6. Farmer Participation and Contribution
The norms of decency should always be observed with the farmer and
the work to be conducted on his land should always be explained to him.
However, his actual participation in the trial program depends on the
stage of research. In developmental or 'upstream' research his direct
input is not very important, but becomes increasingly important from
adaptive trials to verification and he becomes a principal participant
in the demonstration phase.
7. Extension Agent Participation
The extension agent is preferably involved from the initial survey
all the way down through the process, with more responsibility for the
work as it gets closer to the diffusion stage. In this way the extension
agent is more able to appreciate the product he is being asked to sell.
It is important to try and break down the barrier between research and
8. Methodological Points
There are a number of other methodological or 'nuts and bolts' issues
which need to be addressed by the experimenter and determined within
the bounds of the restriction placed on him. (As outlined at the end of
Dr. A. Violic's paper). These are the most appropriate designs, plot
sizes, number of replications per site, number of sites per experiment,
number of experiments per site, selection of sites, selection of
collaborating farmers, research and extension personnel involvement,
proportion of developmental and adaptive research in an on-farm research
system, factors for which the research station is representative,
highest coefficients of variation (C'V's) acceptable, probability level
accepted for rejection of the null hypothesis, etc.
'THE FARMER AS THE RESEARCHERS' CLIENT
(Conclusions from topic presentation, general discussion and specific
Discussion Leader: Michael Collinson
Specific Discussion Group:
Rapporteur: Luis Navarro
CIMMYT, April 3, 1980
- 1 -
Farming Systems characteristics and complexities are tied -to the phy-
sical environment characteristics as well as to the market and Institutional
support to agriculture system developmental stage and influences.
Farming systems research should explicitly consider the characteristics
and possible influences of markets and Institutional support system on the
target farming systems at all stages.
Such consideration should'be specially contemplated in making explicit
the assumptions underlying the socio-economic evaluations of would be "im-
Closely related is the need to include, in evaluation, the long term
macro type of implications, of would be improvedd technology" (i.e. on re-
sources productive capacity conservation) along with the short or medium
term benefits to target farmers (i.e. on income).
The identification of target group of farmers is a need for the effi-
ciency of the whole process of technology development adoption by farmers
under the FSR approach.
Such need stems out of the impossibility (cost wise) of treating every
farmer independently and from the proven impracticality of attempting to
arrive to blanket type of recommendations.
Target group of farmers is a group considered homogeneous on the basis
of predetermined characteristics.
Traditionally, target groups of farmers have been determined on the
basis of their location on homogeneous type of physical environments (i.e.
on land with the same production potential).
Observable F.S. reflects the weighting farmers' give to all circumstances
of the production environment surrounding them,, both natural and economic.
Homogeneous F.S. determination is a good possibility to identify target
groups of farmers.
Given the unstable characteristics of the socio-economic component of
the "total" farm environment, the F.S. criteria to define target groups should
be a complement to homogeneous zone definition all acting as a refinement of
the body of knowledge. Also as a device to distinguish differences in poten-
tial, within and between groups of farmers pesently operating the same F.S.
Ideally excluding institutional mandates influences FSR planning
process should be initiated with a whole F.S. improvement orientation.
Posterior steps in the FST may prove mo~e efficient to concentrate
efforts on a specific commodity or system component. Evaluation criteria
should still consider implication of particular type of research results
on the whole FS.
Previous thought suggests the FSR approach as a link between the tra-
ditional research and extension efforts. Problems and priority identifi-
cation help to draw on existing body of knowledge at the research level
or provide the "upstream" research groups with problems to target on.
Extrictly extension groups would benefit from the "downstream" research
effort of FSR which could provide them with appropriate technologies for
transferring to determined groups of farmers.
Technological "pieces" versus "packages" is not an issue provided every
"piece" independently presuited or not is consistent with net negative
and it does not imply an effect on the FS.
RAPPORTEUR'S REPORT ON FORMAL SURVEYS
Rapporteur: D. Ramakrishnaiah
1. The group discussed at length, the advantages and disadvantages
of formal survey methods as compared to informal surveys. It was pointed
out that each method of data collection has its own merits and demerits,
and one method may work better in a particular situation than others. All
surveys-whether informal-or formal are a part of continuum in gathering
A majority in the group felt that these survey methods are not mutually
exclusive that is informal and formal surveys are not substitutes, but they
complement each other. However, there are a few reservations on this point,
in light of much undigested formal survey data.
2. The question of representativeness was raised in collecting time
series data using a few villages as a sample, as in the case of ICRISAT.
It was pointed that great care should be exercised in sleeting representa-
tive villages on the basis of same basic characteristics which the researcher
feels important as criteria for extrapolation of results.
3. The question was also raised about the representativeness of a non-
random sample in informal surveys. The general feeling was that if an in-
formal sondeo type survey was followed by a formal survey using a random
sample, one can preserve the representativeness of the sample to a greater
4. As regarding the frequency and uniformity of data collection the
general feeling was that infrequent data collection allows us to cover a large
sample whereas frequent data collection allows us to gather indepth information,
covering a smaller sample. However, it was strongly felt that the frequency
of data collection cannot be uniform for all types of data and there should
be a degree of flexibility in this regard, depending on the type of data needed.
5. Same expresses the new that the intensive multiple-visit formal
methods of data collection as used in sane international research programmes
are extremely expensive and cannot be taken up by the national research
programmes with their many resource limitations. But others felt that cost
is not a major factor, but that time may prohibit national programmes from
undertaking intensive formal surveys.
6. On the question of the minimum data requirements for designing on-
farm research in national research programmes, the majority feeling was
that it depends on the background information already available. However,
the least that should be done is a SONDEO type survey.
7. The reliability of data collected through formal surveys was dis-
cussed at length. The general agreement was that the reliability increases
with increasing rapport between the farmer and the interviewer. However,
unreliable yield data is a persistent problem with any survey method
(formal or informal) and no general consensus could be reached as to how
to improve the reliability. Crop cutting was an alternative but same
pointed out that even crop cutting data is not accurate in some instances
there was a wide variation of the crop cutting results as compared to the
actual harvest in a plot. However, it was felt that there is scope to imr
prove the crop cutting techniques to obtain accurate information on yields.
8. On the procedure of asking right questions, the question was raised
as to what is the process that encourages one to ask right questions. The
general feeling was that an exploratory SONDEO type survey is a pre-requisite
for framing relevant and right questions in the following formal survey.
9. Small or medium sample-size, one-shot, well-focused formal surveys
were felt by some to be valuable in the design of on-farm experiments when
representativeness is an issue, or where credibility of results to non-parti-
cipants is important. These surveys were also felt to be important in ob-
taining data for the analysis of experimental results.
Report of Discussion:
ANALYSIS OF EXPERIMENTAL DATA FOR
DESIGNING TECHNOLOGIES FOR FARMERS
Analysis of variance, farm budgeting and LP are the techniques most fre-
quently employed by the workshop participants in the analysis of experimental
data. Some times all three are used in conjunction with ANOVA sorting out
statistically significant factors, budget analysis identifying economically
feasible activities and LP used to identify optimal solutions and explore
implications various changes in technology. In some cases budget analysis is
the principle analytical tool and LP is not employed.
Budget analysis is a most simpler technique than LP. However, in addi-
tion to not being able to identify optimal solutions if used in isolation,
it provides little insight into
1) implications of variability over time
2) implications of factors which are in less
than perfectly elastic supply.
3) implications of input rationing behavior
In addition to identifying treatments which generate highest net returns,
it is useful also to look a net returns to inputs which may be difficult to value
but are in very scarce supply such as water or cash. If there is a substantial
difference among the best treatments as measured against these different criteria,
then input rationing may determine which technology is most acceptable to farmers.
The use of current factor prices in budget analysis assumes that factors
are in elastic supply. This may not be the case, however, in the face of rapid
adoption of labor using technology. There is no methodology for estimating short
run labor or other factor supply curves. In lieu of this, it is important to
supplement budget analysis with time profiles of factor use and supply availabi-'.
lities. Labor, power, and water balances should be simulated in order to
anticipate factormarket implications. Wben potential bottlenecks are observed
then varying input prices in a sensitivity analysis of budgets can provide
Season to season variability was identified as an important problem for
analysis. It was noted that the yields of new and old technology need not be
highly correlated on a season to season basis; this makes relative comparison
from a short time series of data hazardous. One way to overcame some of the
problem is to use historical time series on site specific variables such as
weather to help us simulate instability. At ICRISAT they are modeling surface
run-off behavior in order to provide Information to those developing waters
conservation technology. In some cases LP models are being used to analyse
risk associated with particular technologies. ICRISAT is also using E-V
analysis combined with a risk attitude study to examine profit-risk tradeoffs.
Sane concern was expressed-about the capacity of national programs to
use these sophisticated techniques. Same simple comparisons of frequency
distribution of yields with minimum economic requirements may provide important
insights into implications of instability. The use of stochastic dominance
techniques was mentioned but not developed.
It was noted that experiments may be giving us very biased information
a) There is a high correlation between management
and yield; and
b) Considerably less management is applied to target
group farms relative to experiments.
The Issue of using crop cut from an adjacent farmer managed field as a
measure of experimental boas was raised. Simply deflating experimental results
by a proportional factor may introduce substantial error, however, due to the
presence of interactions. It was generally observed that if large differences
between farmer managed fields and agronomists simulations of farmer practice
was observed then there was strong likelihood of a specification error or
missing experimental variable.
Other topics whcih the discussion group thought important but about which
there was little discussion were:
a) valuing crop residues
b) valuing household resources
c) measuring impact of tenancy on farm management
INFORMAL SURVEY METHODS
Repporteur: Juan Carlos Martinez
As we may recall, the discussion was centered around the strength and
weaknesses of the informal survey methods, and the conditions under which they
may be particularly appropriate.
The discussions emphasized the sondeo as a particular kind of informal
survey. However, this is not the pattern we would like to follow in this occasion.
We will rather try to speak of informal surveys as such.
This brief presentation is divided in two parts. In the first part we
would like to mention a couple of points which came up in the small discussion
group, that in our understanding have not been properly covered in the general
discussion and yet could shed some additional light on the topic under consid-
eration. In the second part we would like to re-emphasize three features of
the methods that we believe deserve to be mentioned in this closing statement.
In relation with the first part the points to be mentioned are the
1) The first one relates to the fact that main advantages and dis-
advantages of the informal survey methods cannot be assessed independently of
the particular sequence in which the method is placed. For example, these
will not be the same for an informal survey immediately followed by on-farm
experiments than for an informal survey followed by a formal survey and then
followed by experiments. Without pretending to be exhaustive the most common
alternative sequences will be:
Informal survey--on-farm experiments.
Informal survey--bformal survey->on-farm experiments.
Informal survey--on-farm experiments and on-farm records.
Informal survey--.exploratory on-farm experiments and formal survey.
2) The second point is that whenever the comparison between informal
and formal surveys or sondeo and formal surveys was made, a good sondeo was
compared with a bad formal survey. Many of the examples mentioned corresponded
to holistic formal surveys, very descriptive in nature.
As opposed to this, we are looking to the minimum set of information
needed to orient our on-farm research. Being our surveys operationally instru-
mental to the orientation of the on-farm research program, we are not particu-
larly concern about general socioeconomic or farm management surveys, but rather
about well focused production surveys. These are the ones which should enter
in the comparisons if these are made.
On the second part of this brief intervention we would like to re-emphasize
certain points already covered in the large group discussion. First, even though
we are not looking for consensus, in this particular case we could agree that
low cost associated with minimum time requirements will be a characteristic which
in higher or lesser degree will be associated with the method depending on the
particular sequence it is supposed to fit. Second, if we look at the main dis-
advantages, they appear to be associated with the non-existence of sampling
procedures within the method, which strongly qualify the ability ot make infer-
ences and the concept of representativeness that we may need in some dimensions
of our recommendation domain (representative farmers in an area; prevailing
practices for a given production pattern, etc.). We may notice however that this
disadvantage will not be so for the particular cases (sequencies) in which a
formal survey follows the informal one. For those cases for which the sampling
problem is relevant, it could be worth to consider the costs and benefits
associated with incorporating sampling procedures into the method.
Finally, we would like to reiterate a point which is not only associated
with the informal survey methods, but rather it was present along all the topics
considered. For informal survey methods to render its benefits it is imperative
that it be carried as a team effort, involving the collaborative work of biological
and social scientist.