CENTRO INTERNATIONAL DE MEJORAMIENTO DE MAIZ Y TRIGO
INTERNATIONAL MAIZE AND WHEAT IMPROVEMENT CENTER
IMPEDIMENTS TO TECHNICAL PROGRESS
ON SMALL VERSUS LARGE FARMS
During the past four years, CIMMYT has been associated with a
number of farm-level studies of the adoption of new wheat and maize
varieties and fertilizer use. Most of these studies have now been pub-
lished, and it is our purpose here to summarize some of the findings
with respect to impediments to farmer adoption of new varieties.
Farm-to-farm differences in the date of adoption or the extent
of adoption of new technology might be explained by a number of factors.
The studies reviewed here considered farm-to-farm differences in the
cost of acquiring and processing information, differences in the physical
productivity of the new technology, differences in incentives due to
tenure arrangements, differences in the cost of inputs, differences in
prices received for the product, differential aversion to risk among
farmers, and scale effects associated with farm or enterprise size.
Policies to stimulate more rapid or more extensive adoption will vary
in their success depending upon which of these factors are the most
important in limiting the rate and extent of adoption. Of particular
relevance to the topic of this session is the extent to which small farmers
lag behind large farmers in adoption, and which of the above factors
might account for that lag.
Paper prepared for the American Agricultural Economics Association
meetings, August 1976. Ricard Perrin is an Associate Professor of
Economics at North Carolina State University, formerly an Associate
Economist with CIMMYT, Mexico; Don Winkelmann is an Economist,
CIMMYT (International Maize and Wheat Improvement Center), Mexico.
Views expressed in this paper are not necessarily those of CIMMYT
or North Carolina State University.
The studies reviewed here were directed toward factors
explaining differences in adoption of new varieties and fertilizer
among farmers surveyed in a number of countries in 1972 and 1973.
Since seeds and fertilizer are nearly completely divisible, one might
ask why farm size effects might be expected at all. First there are
likely to be economies of size in transactions costs. It takes about
the same effort for the small and large farmers to acquire and
evaluate information about a new technique, to learn how to use it
and to make the purchase transaction. For an increase in net returns
of $ 20 per hectare, the effort might be quite worthwhile for a
10-hectare farmer, but not for a 1-hectare farmer. These transaction
costs will fall as community experience with and information about
the new technique becomes more common. Because of this, the
phenomenon of economies of size in transaction costs can lead to a transaction
lag in small farmer adoption even though ultimate adoption levels
may be the same as on large farms. Hence, we will more likely
observe size effects during the early phases of the adoption cycle.
Furthermore, experimentation with new techniques involves
the risks of the unknown, usually involving additional investment,
and small farmers may be less able to undertake such risks. This
risk effect can lead to lower equilibrium levels of adoption, or to
a lag in adoption by smaller farmers until the risks of the unknown
are reduced with experience in the area.
A third potential source of size effects is that smaller farmers
may face higher input costs. For example, quantity discounts might
be available or government subsidies of information, credit or inputs
may favor larger farmers.
The Study Areas
CIMMYT has been associated with maize adoption studies in
Kenya (Gerhart), Colombia (Colmenares)., El Salvador (Cuti6), and
Mexico (Perrin), and with wheat adoption studies in Tunisia (Gafsi),
and Turkey (Demir). In addition, V.S. Vyas undertook, at CIMMYT's
request, a review of the studies of wheat adoption which had been
conducted in India. While the primary focus of these studies was the
adoption of improved varieties, fertilizer use was also examined to
some extent in each. The percent of surveyed farmers who had
adopted improved varieties at the time of the studies ranged from
20 percent among Colombian maize growers to 67 percent among
Kenyan maize growers (though adoption rates for subregions were
as high as 100 percent).
The adoption trends shown in Figure 1 help to provide the
dynamic contexts of the adoption processes at the time the studies
were conducted. Adoption of hybrid maize in Western Kenya shows
the most stable growth pattern, with use increasing steadily at
the rate of an additional 8 percent of the farmers adopting each
year after the hybrids were introduced in 1964. This rate slackened
somewhat in the year immediately preceding the survey. Increases
in the adoption of hybrid maize in El Salvador and Colombia have
Coastal Turkey, Wheat, 1972 study
................ W. Kenya, Maize, 1973 study
--- El Salvador,Maize, 1972 study
---- Colombia, Maize, 1972 study ...
60- Tunisia,Wheat, 1973 study .
14 // /
"- .4* /
+- *.**' -/*
6 7 6 70 7727
00 .. ":. /^ .. .
0..6 20 -8697 7 27
Fig.1 Trends in adoption of new varieties in the study areas.
- 4 -
been both more modest and more irregular. While the earliest
hybrids were introduced into El Salvador in the late 1950's, not
until after the introduction of H3 and H5 (after 1966) did the in-
crease in use of hybrids become steady. In Colombia, on the other
hand, the level of hybrid use has been relatively static in the past
High-yielding varieties of wheat were first introduced com-
mercially in India in 1966-67. There followed a rapid adoption of
these varieties by about 30 percent of the farmers within two years.
The adoption level began to rise again after 1970, as irrigation
facilities were extended and improved, until nearly 60 percent of
Indian wheat was in new varieties by 1973. The data reported by
Vyas were largely gathered in 1969-1971, after the initial surge of
adoption had slowed. In the coastal spring wheat areas of Turkey,
very rapid adoption of Mexican varieties occurred the first two
years following their introduction in 1967. A second very rapid rise
in the use of new varieties occurred in 1972 and 1973 after the
introduction of the Russian winter variety Bezostaya into the winter
wheat producing areas of Thrace and Marmara. In Tunisia, on the
other hand, adoption of the Mexican bread wheat varieties increased
only slowly from their introduction in 1968 until 1970, after which it
fell hi 1973, reportedly due to problems with seed quality. The
niajority of new varieties which have been adopted there are high-
yielding varieties of durum wheats (durums constitute three-fourths
of the national wheat acreage).
In each of the survey studies, the population to be studied
was prestratified by agro-climatic zone as determined in collabora-
tion with agricultural scientists. The samples were later post-
stratified into size groupings. Analytical procedures included both
multivariate analyses and interpretation of two-way tables. The
summary here is restricted to a general interpretation of the results
of somewhat diverse analytical approaches to a similar issue.
The Effects of Farm Size and Other Factors on Variety Adoption
In Table 1 we present the levels of adoption of new varieties
by farm size groupings within agro-climatic zones of the study areas.
In about half of these situations, there appears to be little relation-
ship between farm size and adoption. This occurs in areas of Kenya,
India, and Turkey where adequate time has elapsed since introduction
and nearly all farmers have adopted regardless of size, and in the
medium and high valleys of Colombia where few farmers have
adopted but little additional diffusion has been taking place (Figure 1).
In these areas equilibrium levels of adoption appear to have been
realized, and we would consequently expect little correlation between
farm size and adoption decisions.
The multivariate analyses in the cited studies offer some
insight into which factors are important in explaining farm-to-farm
differences in adoption behavior in areas where equilibrium adoption
TABLE 1.- Farm size and percent of farmers adopting new varieties
Number Farm size Percent of farmers adopting
Crop and area of limits by farm size
farmers (ha) Small Medium Large
1) high, wet
2) high, dry
3) low, dry
2) Aegean valley
3) S. Marmara valley
4) Thrace hillside
1) low rainfall 17
2) high rainfall 20
Sources: Gerhart, Colmenares,
Demir and Gafsi.
Cuti6, Perrin, Vyas,
- 6 -
levels have not been realized. The factors considered varied from
study to study, but included zone and topographic variables to 'capture
differences from farm to farm in yield superiority of the new vari-
eties, plus variables representing differences in farmer incentives
arising from input and product prices, differences in farmer percep-
tions of or reactions to risk, and farm size effects. Due to the dif-
ferences from study to study in variables used and analytical tech-
niques, we can present here only a brief summary (Table 2) of the
results of these multivariate analyses. For each factor included in
the analysis, either a yes or a no is entered in Table 2; yes if the
factor was significant statistically (t-ratios of 1.5 or more) with a
coefficient sufficiently large to be of practical significance, and a no
A cursory glance at Table 2 shows that productivity factors--
agro-climatic zone and topography--are the most consistent in ex-
plaining why some farmers adopt new varieties and others do not.
While these factors were generally far more important in explaining
behavior than any others, we are convinced that much more of
farmers' adoption behavior could have been explained by productivity
considerations, had more accurate measurement of agro-climatic
factors related to productivity been possible. This became clear
in retrospect, as for example when we re-examined some villages
within a few miles of one another in the Aegean area of Turkey
where farmer adoption patterns seemed to make little sense. We
found that the elevation on one valley village was just enough higher
than the others that frost problems precluded the use of new varieties,
and no farmers were using them, though some had tried. Within a
lower valley village, the land parcels were divided between two types
of valley land-- a lower alluvial terrace and an upper alluvial terrace.
The upper terrace soils were -shallower and lighter than those of the
lower terrace, and soil water holding capacity was so low that farmers
preferred the local varieties, which yield as well as the new varieties
under limited rainfall conditions where little fertilizer is used.
Thus, within a small geographic area, we had observed three
villages, ostensibly similar, with markedly different patterns of
adoption--no adopters in one village, nearly all adopters in a second,
and a mixed pattern of adoption in a third. Yet with a better insight
into agro-climatic factors affecting the productivity of new varieties
versus the old, this pattern of behavior was understandable quite
apart from considerations of information, prices and risks. It was
not that agro-climatic factors were not considered prior to the study,
but rather that relatively subtle agro-climatic gradients can lead to
dramatic changes in farmer behavior. Gerhart's Kenya study and an
as yet unfinished maize study in Pakistan reinforce this observation,
as do studies with rice (Barker). It is also a point stressed by
Evenson (p. 201) and his co-workers. Gerhart (p. 26) concluded
that "if anything, this study shows how very site-specific agri-
cultural technology is, even within one region of one (admittedly
very diverse) country'.
Factors explaining within-region variability in
decisions to adopt improved varieties.
Country: Kenya Colombia El Salvador Veracruz Tunisia Turkey
Crop: Maize Maize Maize Maize Wheat Wheat
% Adopters: 67% 20% 36% 34% 31% 64%
Agro-climatic zone yes yes no yes
Topography yes yes yes yes yes
Schooling yes yes yes no no
Extension visits no yes
Demonstrations yes yes no
Age no no
Tenure no no no
Credit-use yes no no yes
Credit-availability yes no
Co-op membership yes
Variety discounts yes noa/
Market sales of crop no no-
Perceived yield risk no no yes
Use of drought:crops -yes
Off-farm income no no no no
Farm size-b no yes no yes no yes
Sources: Gerhart, Colmenares
, Cutib, Perrin, Gafsi, and
a/ Market participation was important in one
of four Turkish
b/ Farm size was important in one of three Colombian, two
of four Turkish regions.
These experiences force us to recognize that within any farming
area, there exists a wide range of expected yield increments from a
given new variety or new technology. The differences can be the result
of gradients in soil depth, texture or other characteristics, differences
in quantity of rainfall or irrigation water, differences in night-time
low or daytime high temperatures in certain seasons, differences in
disease incidence related to these factors, and so on. In a truly
homogeneous agro-climatic zone, the frequency distribution of expected
yield increments across farmers' fields would be bunched rather
tightly about the mean. For practical purposes, however, either for
developing and recommending new technologies or for analyzing farmer
behavior, we must deal with groups of farmers for whom the frequency
distribution of expected yield increments will not be very compact due
to agro-climatic variability. Because of this, it will often be difficult to
identify to what extent productivity factors are affecting farmers'
decisions versus differences in economic factors such as information,
tenure, input and product markets, scale economics, etc.
Turning now to the effect of these economic factors in the six
studies, with some exceptions farmer behavior was not found to be
significantly or consistently related to them. One exception was the
availability or use of credit. In four of the six studies in which it
was considered, this variable was significantly related to the adoption
of high-yielding varieties. However, even though farmers using credit
were more likely to be using new varieties (44 percent more likely
in Colombia, for example), this does not necessarily imply that the
existence of credit programs is critical to farmer adoption decisions.
Where a new variety or new technology is marginally profitable, the
subsidies implied by government credit programs could be expected
to affect many farmers' decisions. But if the new technology is profit-
able even at unsubsidized capital and input prices, farmers could be
expected to adopt the technology and to purchase the inputs through
the credit program if it in fact provided capital and inputs at sub-
sidized prices. In the latter case, we would see a high correlation
between the use of credit and adoption, even though the credit program
itself had little effect on the adoption decision.
Market conditions accounted for very little of observed adoption
behavior. Differences in price discounts for new wheat varieties from
village to village in Tunisia were significant in explaining farmer
adoption decisions, but this was the only case in which product market
circumstances or market participation were related significantly to new
variety adoption. The availability of seed appeared to be a limiting
factor in adoption of winter wheats in Turkey and durum wheats in
Tunisia, where the introduction of new varieties was fairly recent and
seed multiplication had not kept pace with farmer demand.
Differences in the cost of acquiring and processing information
as measured by extension activity and schooling, had only a slight
relationship to adoption. Extension visits or demonstration attendance
appeared to increase the probability of adoption by 10 to 15 percent in
certain subregions of Colombia and Turkey, less or none in other
- 9 -
areas. Each additional year of schooling increased the probability of
adoption by an estimated 8 percent in Colombia, less in other areas.
The question of whether risk aversion is affecting adoption
decisions was not really very satisfactorily addressed in these studies,
and it is difficult to imagine how one can do so. In three studies,
farmer perception of the risk of low crop yields was used as an index
of farmers' risk aversion, while in Kenya the existence of drought
resistant "famine" crops was used. While these indexes were in general
correlated with adoption, the interpretation of this result is not
straightforward. It is true that where crop failures are more common,
risks are greater. But it is also true that where crop failures are
more common, average yields will be lower. Where farmers perceive
high risks of crop failure, reluctance to adopt could be evidence of
risk aversion or simply evidence that the average returns to adoption
are too low. To determine which of these factors are contributing to
the observed correlation, we would require more specific information
than was available about farmer perceptions of expected yields versus
yield risks for both new and old varieties.
This brings us again to the question of the relationship between
farm size and adoption decisions. We previously noted that in about
half of the agro-climatic situations of Table 1, farm size appeared to
have an important effect on adoption. Yet in the multivariate analyses,
farm size was not correlated with adoption decisions in some of these
areas. In the low elevation areas of Colombia, where Table 1 indicates
an important size effect, the multivariate analysis showed that once
credit and extension contacts are considered, the residual effect of
size was minor, with each hectare increase in size being associated
with only one-tenth of 1 percent increase in the probability of adoption.
After a careful consideration of extension and credit activities, Colme-
nares concluded that the orientation of these programs toward larger
farmers contributed substantially to the greater adoption rate among
larger farmers. In Tunisia, it appears that smaller farms' lower
adoption rates are due to differences in topography, credit use and
local price discounts, rather than to factors related to farm size
per se. The multivariate analyses of hybrid maize adoption in El
SalvaWor likewise fail to show any effect of farm size per se, though
it is not clear yet what factors may be contributing to te apparent
size effect revealed in Table 1. In India, Vyas reports that most of
the observed differences in variety adoption by farm size were due to
differences in the availability of irrigation,water, and to a seed
distribution policy which initially favored larger farmers.
In the remaining areas in which size effects are notable in
Table 1, variety introductions were quite recent in Veracruz, Mexico,
(maize) and in the South Marmara and Thrace areas of Turkey, and
the multivariate analyses indicated definite size effects associated
with adoption even when other factors were considered. Thus, of all
the areas considered, only in the Aegean area of Turkey did there
appear to be significant size effects in adoption which were not
plausibly explained by differences in productivity, differences in ef-
fective market prices for inputs or products, or by the early stage
of adoption in which small farms can be expected to lag behind larger
farmers because of economies of size in transaction costs in evaluating
and acquiring the new varieties.
By contrast, the scale effect in fertilizer adoption in these
studies was more pronounced. Space precludes a summary of the
analyses of fertilizer adoption decisions, but there appeared to be
significant size effects in fertilizer adoption in nearly half of the
agro-climatic situations considered. Apparently fertilizer use is
more heavily influenced by farm size than is variety use, as might
be expected due to the greater investment at risk with fertilizer.
The adoption studies with which CIMMYT has been associated
have shown that to a limited extent, differences in farmer adoption
behavior can be explained by differences in information, in the
availability of inputs, in market opportunities for the crop, and dif-
ferences in farm size and farmer risk aversion or risk perception.
The pattern of adoption among large and small farms is generally
consistent with the proposition that small farms may lag behind
larger farms in the early stages of adoption, but soon catch up.
Our impression from these studies is that the most pervasive ex-
planation of why some farmers don't adopt new varieties and fertil-
izer while others do, is that the expected increase in yield for some
farmers is small or nil, while for others it is significant, due to
differences (sometimes subtle) in soils, climate, water availability,
or other biological factors.
Agricultural technology is more site-specific than we were
led to believe by some of the early successes with wheat and rice
varieties. The early high-yielding wheat varieties were adapted to
extensive production areas and were widely adopted. It appears to
us that these early successes will not be easily duplicated on such
a scale. In the areas not already dominated by new varieties, the
factors limiting yields are so disparate and complex as to make it
unlikely that any single new variety can repeat the success of the
early releases from international breeding programs.
Government policies to reduce the cost of information, the
cost of inputs or the impact of risk could be expected to influence
the decisions of those farmers whose expected yield increases are
now marginal. But the experience in the areas of these .studies
suggests that these policies are not likely to have a large impact in
increasing the number of farmers who adopt new technologies.
Significant advances in adoption will not occur until significant
advances are made in technologies which will increase yields in the
agro-climatic environment of those farmers not presently adopting.
This will require greater attention to the environment for which
technologies are being developed than was necessary for the early
successes in cereal breeding.
- 11 -
Barker, Randolph, et al. Changes in Rice Production Technology in
Selected Areas -oT Asia. International Rice Research Institute,
Los Bafios, Philippines, 1975.
Colmenares, J. Humberto. Adoption of Hybrid Seeds and Fertilizers
Among Colombian Corn Growers. Abridged by CIMMYT. Cen-
tro Internacional de Mejoramiento de Mafz y Trigo, Mexico
Cuti6, Jesus. Diffusion of Hybrid Corn Technology, The Case of
El Salvador. Abridged by CIMMYT. Centro Internacional de
Mejoramiento de Mafz y Trigo, Mexico City, 1976.
Demir, Nazmi. The Adoption of New Bread Wheat Technology in
Selected Regions of Turkey. Abridged by CIMMYT. Centro
International de Mejoramiento de Mafz y Trigo, Mexico City,
Evenson, Robert. Technology generation in agriculture, Chapter 8
in Lloyd G. Reynolds, ed., Agriculture and Development,
Yale University Press, 1975.
Gafsi, Salem. Green Revolution, The Tunisian Experience. Abridged
by CIMMYT. Centro Internacional de Mejoramiento de Mafz
y Trigo, Mexico City, 1976.
Gerhart, John. The Diffusion of Hybrid Maize in West Kenya. Abridged
by CIMMYT. Centro Internacional de Mejoramiento de Mafz y
Trigo, Mexico City, 1975.
Perrin, Richard. New Maize Technology and Its Adoption in Veracruz,
Mexico. Centro Internacional de Mejoramiento de Mafz y Trigo,
M/exico City, forthcoming.