|
On-I-arrm Research and Fa mcr Adoption of New Maize Technology
in Malang District, Indoresia
Larry Harrington
Ru!y Krisdiana
JIerianto
UR Malang Research institute for Food Crops (MARIF)
CIMMIYT Asian Regional Ecoronmics Program
I I P.O. Box 9-188, Bangkok 10900, Thailand
June, 1992
On-Farm Research and Farmer Adoption of New Maize Technology
in Malang District, Indonesia1
Larry Harrington 2
Ruly Krisdiana 3
Herianto 4
1. Opinions expressed are not necessarily those of CIMMYT or MARIF.
2. CIMMYT Economics Program, Bangkok, Thailand
3. Socioeconomics section, Malang Research Institute for Food Crops (MARIF)
4. Socioeconomics section, Malang Research Institute for Food Crops (MARIF).
Contents
Acknowledgements
1. Introduction and background 1
2. The MARIF maize on-farm research program 4
2.1 Farmers' circumstances: A summary 4
2.2 The research program and the recommendations 6
2.3 Research-extension links 9
3. The adoption survey: Some results 11
3.1 Characteristics of sampled farmers 12
3.2 Effects of planting date 18
3.3 Adoption of reduced seed rates 18
3.4 Adoption of shootfly control 23
3.5 Adoption of improved germplasm 28
3.6 Interactions among components sequential adoption 32
4. Logit analysis 34
4.1 Logit analysis: A review 34
4.2 Current method of shootfly control 35
4.3 Timing of adoption of chemical shootfly control 36
4.4 Current use of improved germplasm 37
5. Conclusions 41
5.1 Linking OFR activities with farmer adoption 41
5.2 Benefits earned from OFR 44
5.3 Future activities 47
References 50
List of Tables
1. Major Palawija production environments in Malang District 5
2. Recommendations developed by the OFR program 9
3. Some farmer characteristics, by farmer class 13
4. Factors associated with early vs late planting of the second maize crop 22
5. Seeds per hill, seed rate, and plant stand management in the second
maize crop, by farmer class 24
6. Shootfly problem and control, by farmer class 31
7. Adoption of three technical components in the second maize crop,
by farmer class 33
8. Model: Current use of chemical shootfly control 36
9. Model: Timing of adoption of chemical shootfly control (full model) 38
10. Model: Timing of adoption of chemical shootfly control
(restricted model) 39
11. Model: Current use of improved germplasm 40
12. Evidence on causal links between OFR program and farmer adoption 44
13. Economic benefits from OFR
(Shootfly control and reduced planted density) 46
14. Sensitivity analysis on economic benefits from OFR
(Shootfly control and reduced planted density)
Effect of expanded area of adoption, and additional years 48
List of Figures
1. East, Java, showing Malang study area and MARIF headquarters 3
2. Hypotheses on problems and causes 8
3. Farm size by farmer class 14
4. Varietal use by farmer class 15
5. Frequency of extension contact by farmer class 16
6. Mean farm size by frequency of extension contact 17
7. Expected shootfly damage as affected by planting date 19
8. Planting date of maize by cropping pattern 20
9. Planting date by variety 21
10. Change in seed rate per ha by farmer class 25
11. Cumulative adoption of carbofuran on maize, by farmer class 26
12. Shootfly control method, by variety 27
13. Use of carbofuran by farmer class 29
14. Adoption of improved maize germplasm (1984 vs. 1990) 30
Acknowledgements
The authors would like to acknowledge the work of the MARIF researchers
and technicians whose efforts contributed to the success of the OFR program de-
scribed in this paper. These include: Dr. Marsum Dahlan, Dr. Sudaryono, Ir Chamdi
Ismail, Ir Muchdar Soedarjo, Ir Herman Subagio, Ir Fachrur Rozi, Mrs. Ulfah Dah-
lan, and Mr. Cipto Prahoro. Dr. Charles van Santen and Ir M. J. P. van Staveren,
formerly of the ATA-272 project, were also active participants in the maize OFR
program. We would especially like to acknowledge the support and encouragement
provided by Dr. Soetarjo and Dr. Sumarno, former and current Directors of MARIF,
respectively. Helpful comments on an earlier drafat of this paper were provided by
Robert Tripp, Derek Byerlee, Charles van Santen, Hari Gurung and Christoph Mann.
Unfortunately, only the authors can be blamed for any remaining errors. Opinions
expressed are not necessarily those of MARIF, CIMMYT or the ATA-272 project.
Finally, we would like to express our gratitude to those extension agents and farmers
who helped us understand the complexity of maize farming in Malang District.
1 Introduction and background
Maize is a major staple food in Indonesia, especially in the Province of East
Java, where it is second in importance only to rice. In addition, it is an important
component of livestock feed, the demand for which is growing rapidly in Indonesia.
Maize production in Indonesia reached 6 million tons in 1986-88, having increased at
about 5.5% per year since 1973-77 (CIMMYT, 1990). Around 45% of Indonesian
maize production comes from East Java (Timmer, 1987).
Much of the increase in maize production at the national level can be traced to
improvements in maize yields, which have grown at about 4.5% per year between
1973-77 and 1984-88. In contrast, growth in maize harvested area over the same time
period was only around 1% per year (CIMMYT, 1990). Maize yield improvements in
East Java have more or less kept pace with those found in other Provinces (Timmer,
1987, Table 2.2).
Continued growth in maize production and productivity in East Java depends
largely on the development and dissemination of new agricultural technology. Possi-
ble sources of productivity growth include expansion in the area under improved
germplasm, including hybrids, and increases in the levels of fertilizer and other inputs
applied to maize. In addition, improvements in technical efficiency may be possible,
i.e., increasing yields or reducing unit costs, at constant input levels, through more
careful timing and placement of inputs, improved balance of plant nutrients, or adop-
tion of improved cultural practices.
It has been widely accepted that the development of new crop management
technology (especially technology to improve technical efficiency) requires research
with a farming systems perspective (OFR)1, conducted on farmers' fields with farmer
participation (Byerlee, Harrington and Winkelmann, 1982; Sumberg and Okali, 1988;
Manwan and Oka, 1990), and that this is particularly true in areas such as East Java,
where field crops are grown in rainfed upland environments in the context of complex
cropping patterns and farming systems. Recently, however, research managers and
donors have begun to have misgivings about the efficiency of OFR. While it has
undoubtedly helped researchers and extension workers to better understand farmers'
circumstances, OFR is increasingly viewed as having had little impact at the farm
level, i.e., as having failed to generate the kinds of technologies that farmers are will-
ing to adopt. Some observers are beginning to wonder whether OFR will be able to
repay the resources that have been invested in it (Tripp, et. al., 1990; Anderson,
1990).
The objective of this paper is to briefly describe one OFR program in East
Java, Indonesia, and present preliminary evidence that it has, in fact, had an impact
on farmers -- that new technology was developed that farmers have adopted. A
further objective is to examine the characteristics of adopters vs. non-adopters, and to
examine the adoption behavior of farmers and villages that participated in the OFR
program vs. those that did not. The program in question was conducted by the
Malang Research Institute for Food Crops (MARIF), between 1984 and 1989, and
focused on maize on rainfed landtypes in representative areas of Malang District
(Fig. 1).
1. OFR normally stands for "on-farm research". As used here, it stands for any research with a farm-
ing system perspective, regardless of whether it is labeled "on-farm research", "farming systems re-
search", "on-farm client-oriented research", etc.
Figure 1 East Java, showing Malang study area and MARIF headquarters
Source: Krisdiana, et. al., 1991, p. 145
2 The MARIF maize on-farm research program
The maize on-farm research program described in this section was begun by
MARIF with technical assistance from the Netherlands1 and from CIMMYT. Re-
search was carried out that addressed problems of pest control, plant stand manage-
ment, fertilizer management and varietal selection in maize cultivation. Solutions to
these problems were sought in light of various farming systems interactions. More
complete descriptions of this OFR program may be found in Dahlan, et. al., 1987 and
Krisdiana et. al., 1991.
2.1 Farmers' circumstances: A summary
Research has identified four different environments in Malang District suit-
able for the production of palawija (or secondary, non-rice food crops, a category
which includes maize). The program described in this paper focused on one of these
environments, characterized by "young volcanic soils" (YVS) (Table 1). The YVS
study area in Malang District includes about 30,000 ha of physical area (or almost
60,000 ha of harvested area, as most fields are double-cropped). The study area is felt
to be reasonably representative of an additional 150,000 ha of palawija area in East
Java.
Average annual rainfall in the YVS study area is around 2130 mm, most of
which falls between December and April. Average annual temperature is 24 degrees
C. Major soil categories include latosols (inceptisols) and regosols (entisols). Soils
are known to be low in organic matter and deficient in phosphate.
1. Project ATA-272, "Technical Cooperation Indonesia The Netherlands: Strengthening MARIF'.
In recent years, this has been implemented by the Royal Tropical Institute.
Table 1.
Major palawija production environments in Malang District
..-------------------------------------------------
Land Dominant Altitude Physical
Type Soils Crops mast Area (%)
..----------------------------------------..---------
Rainfed Limestone Maize, cassava < 600 43
grain Legumes,
sugarcane
Rainfed Young volcanic Maize, 400 700 37
dry seeded rice
Irrigated Alluvial and Transplanted 400 700 15
Young volcanic rice, maize
Rainfed Young volcanic Maize and 400 1500 5
and volcanic horticultural
ash crops
Source: Dahlan, et. at., 1987
Given the relatively favorable rainfall conditions, two crops are normally
grown per year. Within the study area, maize and upland rice are the dominant
crops, although cassava, sugar cane and grain legumes are also found. Common
cropping patterns include maize followed by maize, or upland rice followed by
maize. Cassava is restricted to field borders, or to an occasional row within the field.
Grain legumes are occasionally substituted for maize in the second crop.
Farms in East Java are small by any standard. In the study area, a typical
farm is composed largely of rainfed land suitable for field crops (tegal). Around a
third of study area farmers also produce small amounts of rice in bunded wetlands
or sawah. Most farmers have home garden plots or pekarangan. These are small
parcels near the home planted to numerous species, including perennials, and fea-
turing several layers of canopy.
Rice from sawah fields and citrus products, clove, bananas and other crops
from the pekarangan are sources of income that can be as important as that derived
from growing palawija on rainfed tegal fields. Most farms also raise livestock, espe-
cially cattle and poultry. In addition, many farms have family members active in off-
farm and non-agricultural labor markets, e.g., cutting and processing sugar cane, light
manufacture, small trading and retailing, and seasonal migration to urban centers.
Production of palawija, then, is a major source of income but is by no means
the only source. For many families, however, palawija and especially maize are the
major food staples. In the study area, many families consume more maize than rice.
Maize consumption is concentrated among lower-income families in rural areas, and
families without sawah suitable for rice production (Krisdiana et. al., 1991).
2.2 The research program and the recommendations1
Maize on-farm research was continuously conducted in the YVS study area
between 1984 and 1989. Activities included exploratory and formal surveys, a wide
array of on-farm trials, and laboratory tests of one kind or another. These activities
1. Details of the on-farm trial program, including a discussion of major maize productivity problems,
their causes, the importance of farming system interactions, and evidence on alternative means of
solving them, may be found in Krisdiana et al, (1991), or in a number of MARIF OFR working papers.
have taken MARIF staff into dozens of villages and have led to cooperation with
hundreds of farmers.
Research focused on four issues: early season insect damage (especially
shootfly or Atherigona spp.), plant stand management, soil fertility management,
and varietal selection. Diagnostic work led researchers to conclude that these
problems were interconnected.
Farmers' plant stand management features overplanting. This is used to help
compensate for expected shootfly damage. Overplanting, however, leads to inter-
plant competition and lodging. Lodging is further exacerbated by farmers' fertilizer
management practices. High levels of nitrogen, but little or no phosphate, are
commonly applied. The interactions among these problems, along with some of their
causes, are diagramed in Fig. 2.
On-farm trials of various kinds were used to define these problems and assess
alternative solutions. Agronomic, statistical and economic assessment of the results
of these trials led to the formulation of the recommendations shown in Table 2.
These recommendations should not be viewed as a package of technology, but rather
as alternatives to conventional farmers' practices. Researchers expected (based on
assessments provided by cooperating farmers) that shootfly control and plant stand
management changes would precede changes in maize variety or fertilizer manage-
ment.
Fig. 2: Hypotheses on Problems and Causes
Inefficiently high
doses of nitrogen
increase production
costs and reduce
profits.
Nutrient
deficiencies
(phosphate and
sulfur) reduce
maize yields.
Lodging reduces
maize yields.
Farmers use
unimproved maize
varieties, with
poor plant vigor
and low yield
potential.
Spindly stalks
and poor plant
vigor.
KEY: Problems affecting productivity are represented by
rectangles. Causes of these problems are represented as ovals
Government
subsidies on
nitrogen
fertilizer.
Table 2
Recommendations developed by the OFR program
Farmers' Suggested
Theme Practice Alternative
Shootfly Overplant (with seed Apply carbofuran 3%
control rates of over 40 kg/ha) granules (5 kg/ha)
to compensate for at planting time.
expected damage.
Plant Overplant as above. Reduce seed rate
stand Then remove damaged to 25 kg/ha. No
management plants over the course thinning.
of the crop cycle.
Some thinning.
Variety Traditional open- Improved OPV of the
pollinated variety, same grain type and
Fertilizer About 390 kg/ha of
Management urea (180 kg/ha nitro-
gen). No phosphate.
maturity (Arjuna), or
hybrid.
Around 90 kg/ha
nitrogen, with
about 46 kg/ha
phosphate.
Source: Krisdiana, et. at.. (1991).
2.3 Research-extension links
Farmer adoption of technology is often affected by extension efforts to inform
farmers. Good research-extension links are necessary to keep extension workers
current on research results -- and to aid researchers in staying current on farmers'
circumstances. It is not clear, however, whether farmer adoption of new maize
technology in Malang District (with the possible exception of varietal change) was
much accelerated by extension activity. Official research-extension mechanisms in
Indonesia that prevailed when the maize OFR program was active are described by
van Santen (1987):
"... formal lines of communication for passing relevant research results to
extension services are not direct. New technologies... are to be channeled via
a central "communication forum" consisting... [of] AARD, the BIMAS pro-
gram of the Directorate General of Food Crops (DGFC) and the Agency for
Agricultural Education, Training and Extension (AAETE). This "communica-
tion forum" refers the appropriate new technologies to the extension service,
which is organized on a provincial basis..." (p. 1).
Suggested recommendations, then, should in theory pass from researchers in
Malang District, to their superiors in Jakarta, through the central "communication
forum", through the extension offices in Jakarta, back to provincial extension authori-
ties, then finally to extension workers in Malang District.
During the early cycles of the MARIF on-farm research program, research-
extension links were imperfectly developed. The complexity of the research-
extension linkage mechanism described above inhibited the development of a
common understanding among researchers and extension workers of farming prac-
tices and possible alternatives. Alternative practices were developed by MARIF, but
lack of extension involvement raised the prospect of farmer adoption driven only by
farmer-to-farmer contact.
However, it would be mistaken to say that there was no contact at all between
MARIF researchers and extension workers. Since early cycles of research, research-
ers and extension workers cooperated in planning and implementing farmer field
days, focused on verification trial locations. Moreover, village level extension workers
were normally involved in the implementation of trials located in their village. What
was lacking, however, was active interaction regarding selection of experimental
themes, and the multiplication of effort by means of demonstration trials throughout
Malang District.
Informal contacts between MARIF researchers and district- and provincial-
level extension workers eventually led to proposals for modifying the linkage mecha-
nism. Improved contacts were made possible by the interest of a senior provincial-
level extension official. However, the expanded research-extension connection fo-
cused less on Malang District itself, and more on the possibility of extrapolating OFR
results to neighboring Districts with similar agroclimatic circumstances.
3 The adoption survey: Some results
In April, 1990, during the post-rainy season, a survey was conducted to ascer-
tain the level of adoption in Malang District of maize production recommendations
generated by the OFR program1. In addition, the survey aimed to measure the rate
and sequence of adoption among different farmer classes.
For the purpose of the survey, villages were divided into those that had at one
time collaborated with the OFR program ("collaborating villages")2 and those that
had not ("control villages"). In addition, farmers in collaborating villages were divided
into farmers that had directly cooperated with the OFR program ("cooperators") and
those that had not ("non-cooperators"). Thus, three farmer classes were formed:
1. Farmer adoption of soil fertility management recommendations was not studied in the adoption
survey. The survey was implemented before much of the fertilizer had been applied by farmers.
2. Of the 116 villages in Malang District, 16 had at one time or another cooperated with the OFR
program.
cooperators; non-cooperators in collaborating villages; and farmers in control villages.
Two-stage stratified random sampling was used. Villages were pre-stratified by OFR
collaboration, then three farmers per village were chosen at random. In total, sixty
farmers were selected. Sampled farmers were asked about farm resources and maize
production practices. Yield estimates were not obtained. Results have-been re-
weighted by the proportion of farmers in each stratum found in the complete District.
3.1 Characteristics of sampled farmers
Sampled farmers were found to operate farms of about 1.5 ha in size, includ-
ing sawah, tega, and pekarangan. Few farmers were found to grow maize on sawah.
Most farmers reported using an upland rice maize cropping pattern in their tegal
fields. Use of improved maize germplasm (Arjuna or a hybrid) was found to be
quite widespread.1 Planting dates were somewhat late, with almost half of the
maize planted in April (Table 3).
Cooperating farmers were somewhat different from other study area farmers
in a couple of ways. Farm size of cooperating farmers was somewhat larger, because
of greater sawah area (Fig. 3). Hybrid use was entirely restricted to farmers in
collaborating villages, though most farmers in all classes used Arjuna (Fig. 4). As
expected, cooperating farmers tended to have more contact with extension workers
(Fig. 5). And farmers with the most frequent contact with extension tended to have
considerably larger farms (Fig. 6).
1. It is possible that some farmers used more than one variety, e.g., on different fields. In this survey,
however, data were taken from only one field per farm and varietal use was ascertained for that field
only.
Table 3
Some farmer characteristics, by farmer class
Farmer class: a/
All
Farmers
Variable A B C (1990) b/
Sample size 12 30 18 60
Weight 0.01 0.13 0.86
Farm area:
Sawah (ha) ** c/ 0.78 0.27 0.03 0.07
Tegal (ha) 1.14 1.59 1.39 1.41
Pekarangan (ha) 0.09 0.04 0.07 0.07
Total (ha) 2.01 1.90 1.49 1.55
Grow maize on:
Sawah 8% 10% 0% 1%
Tegal 100% 100% 100% 100%
Cropping pattern:
Maize-maize d/ 33% 26% 29% 29%
Upland rice maize e/ 67% 74% 71% 71%
Maize variety + f/
Local 8% 19% 12% 13%
Arjuna 66% 61% 88% 84%
Hybrid 25% 19% 0% 3%
Planting Date ** g/
February 17% 8% 0% 1%
Early March 42% 7% 6% 6%
Late March 17% 29% 47% 44%
April 25% 55% 47% 48%
Family labor 1.7 1.5 1.4 1.4
a/ Group A = Farmers that directly cooperated with the OFR program;
Group B = Farmers that did not directly cooperate with the OFR pro-
gram, but who lived in the same village where the program was operat-
ing; Group C = Farmers living in villages with no contact with the OFR
program.
b/ Weighted by the proportion of each class in the population (row
labeled "weights")
c/ t-statistic for sawah area by farmer class (OFR cooperators vs.
non-cooperators) = 2.02, with 58 df, prob. = 0.02.
d/ Includes maize maize maize
e/ Includes upland rice maize maize
f/ Chi-square = 5.8 with 4 df, prob = 0.20
g/ Chi-square = 15.4 with 6 df, prob = 0.02
** = significant at the 0.05 Level; + = significant at the 0.20 level
2.5-
Fig. 3: Farm Size by Farmer Class
CCooperators Non-
-Non-Cooperators
2---
0.5--
0 -K------
B
Farmer Class
Tegal Sawah Pekarangan
SControl I
1.5--
Fig. 4: Varietal Use by Farmer Class
90
80--
Cooperators Non-Cooperators
70
60--
a) 40-
30" -. -
20
10
A B C
Farmer Class
M Local M Arjuna M Hybrid
Fig. 5: Frequency of Extension Contact
by Farmer Class
A B C
Farmer Class
Frequency is defined as the "normal" or "expected" number of extension visits per season.
16
rr -- T-r--m- T1 -1-77ri1-n I1 l TT f Ii T --P V 1 11'TJF
IF F- f [--F
Fig. 6: Mean Farm Size
by Frequency of Extension Contact
5.
4.5
4
3.5-
3-
S2.5-
2-
1.5
0.5
None One Two Three As Needed
Expected number of visits per season
17
3.2 Effects of planting date
In OFR diagnostic activities, it was suggested that early planting of post rainy
season maize would be less common in the upland rice maize pattern compared to
the maize maize pattern. Early planting of post rainy season maize, in turn, was
thought to lead to reduced levels of expected shootfly damage in that crop.1 Adop-
tion survey data support both of these hypotheses (Table 4, and Figs. 7 and 8). It is
possible that the adoption of chemical shootfly control has weakened the link be-
tween planting date and shootfly damage and has allowed farmers to use later maize
planting dates for the post-rainy season. This might help explain the observed expan-
sion of upland rice area during the rainy season.2 This expansion, however, may have
more to do with changes in markets and prices.
Hybrid maize was found to be rarely planted late (Fig. 9). This is mostly likely
due to the longer length of maturity of most hybrid materials.
3.3 Adoption of reduced seed rates
One of the problems addressed in the OFR program was interplant competi-
tion caused by high seed rates (overplanting). During the first diagnostic survey,
farmers were found to use an average of 3-5 seeds per hill, with a mean seedrate of 42
kg/ha (MARIF, 1985). In the 1990 adoption survey, farmers reported planting an
1. Things are, in fact, a bit more complex. Shootfly damage in post-rainy season maize is expected to
be more severe when maize is planted late -- but also when maize follows maize in the cropping pat-
tern. These factors partly offset each other.
2. Area in upland rice has apparently expanded since the 1984 diagnostic survey. At that time, about
one third of tegal area was found planted to upland rice in the rainy season, compared to well over two
thirds of the area in the 1990 adoption survey.
Fig. 7: Expected Shootfly Damage
As Affected by Planting Date
80
70
60-
E 50
u_ 40.------ -
C
1 40----------
20-
10
0
Feb Early March Late March April
Planting Date
Y-axis measures farmer opinion on whether shootfly is likely to be a problem for different planting dates.
19
Fig. 8: Planting Date of Maize
by Cropping Pattern
(Post-rainy Season)
5
5------
0- +H
0
Maize maize Upland rice maize
Cropping Pattern
Feb
Early March
Late March
April
5(
4i
S3i
1
E 3<
u 1
(D
Fig. 8: Planting Date of Maize
by Cropping Pattern
(Post-rainy Season)
#'
I-- I i I I
Planting Date by Variety
Feb Early March Late March April
Planting Date
SHybrid M Arjuna M6 Local
Fig. 9:
Table 4
Factors associated with early vs. Late planting
of the second maize crop
Early Late
Variable Feb March March April
Cropping Pattern ** a/
Maize maize 29% 11% 17% 41%
Upland rice maize 0% 14% 37% 48%
Variety ** b/
Hybrid 33% 22% 22% 3%
Arjuna 5% 14% 24% 40%
Local 0% 0% 78% 22%
Expect Shootfly Damage
on Maize + c/
Yes 40% 50% 68% 78%
No 60% 50% 32% 21%
a/ Planting date for second maize crop reported by farm-
ers. Chi-square = 14, 6 df, prob. = .003.
b/ Planting dates for different varieties as measured in
the adoption survey. Chi-square = 20.3, 6 df, prob = .002.
c/ Farmers' opinion on whether shootfly is normally a
problem, for different planting dates. Chi-square = 4.5, 3
df, prob. = .20.
** = significant at the .05 level. + = significant at
the .20 level.
average of just under 3 seeds per hill. Average seed rates were measured at only 29
kg/ha of seed, a 30% reduction over the last several years. More than half of the
farmers reported reducing seed rates, with most of the rest reporting no change. Only
a few farmers in the control villages reported increasing seed rates (Table 5 and Fig.
10).
Interestingly, there appears to be no relation between planting date and cur-
rent seed rates (data not shown). This may be due to a reduced need during late
seeding to control shootfly by overplanting, given the (now) widespread use of chemi-
cal control.
3.4 Adoption of shootfly control
The use of chemical control for shootfly has expanded rapidly in the study area
during the last few years. Most farmers report using carbofuran or other insecticides
for this purpose. The application dose for carbofuran users is around 3 kg/ha,
somewhat under the recommended dose of 5 kg/ha.
A comparison of cumulative adoption by farmer class shows that, as expected,
farmer cooperators tended to adopt carbofuran somewhat ahead of other farmers1
(Fig. 11). However, there appears to be a strong link between the use of local varie-
ties and the continued use of overplanting and thinning for insect control (Fig. 12).
Interestingly, continued use of overplanting and thinning is also associated with less
extension contact (data not shown).
1. Farmers were asked to report the first year in which they had used carbofuran for shootfly control
on maize.
Table 5
Seeds per hill, seed rate, and plant stand management
in the second maize crop, by farmer class
Farmer class: a/
Att
Farmers
A B C (1990) b/
12 30 18 60
0.01 0.13 0.86
Measured seeds per hill,
four weeks after planting,
current season (mean): c/
Seeds per hill at planting,
this season (farmer opinion)
(mean): + d/
Reduction in seeds per hill
compared with three years
ago (farmer opinion) ** e/
Median seed rate,
this season:
Seed rate this season
compared with three years
ago: + f/
Use the same seed rate
Have reduced the seed rate
Have increased the seed rate
2.25 2.33 2.45
2.71 2.81 2.94
1.83 1.13 0.89
27.1 28.6 30.0
45% 23%
55% 65%
0% 11%
a/ Group A = Farmers that directly cooperated with the OFR
program; Group B = Farmers that did not directly cooperate
with the OFR program, but who lived in the same village
where the program was operating; Group C = Farmers living in
villages with no contact with the OFR program.
b/ Weighted by the proportion of each class in the popula-
tion (row labeled "weights")
c/ t-statistic comparing means for classes A and B vs. C
equals 1.34, 58 df, prob. = .09.
d/ t-statistic comparing means for classes A and B vs. C
equals 1.24, 58 df, prob. = 0.11.
e/ t-statistic comparing means for classes A and B vs. C
equals 1.9, 48 df, prob. = 0.03.
f/ Chi-square = 7.02, 4 df, prob. = 0.134.
Significance levels: *** = .01 level; ** = .05 level; *
= .10 level; + = .20 level
Variable
Sample size
Weights
2.43
2.92
0.93
29.8
Fig. 10: Change in Seed Rate per Ha
by Farmer Class
70
Cooperators Non-Cooperators Same
60- t l
Control Decrease
SIncrease
40,
+L-
C
a) 30
A B C
Farmer Class
Fig. 11: Cumulative Adoption
Carbofuran on Maize, by Farmer
of
Class
-U-
Cooperators
Non-Cooperators
Control
p.
I.
I.
I.
40-
n^<"
20-
10-
0-
82 83 84 85 86 87 88 89 90
Year
r1 r~
L~_LUIL__LLLIIIL*I-~ II I I 1IIII1~U~1~ iii II I ii i i I II UI1 ( i
I J I __ I _
n m n
Fig. 12: Shootfly Control Method,
by Variety
90
80 C
70 C
ca 60 C
c 504
c 40
i i
Hybrid Arjuna Local
Variety
m
)verplant
:arbofuran
theirr Insecticide
Continuous carbofuran use on maize1 and its use on other crops is concentrat-
ed in OFR farmer cooperators (Fig. 13). In general, farmers agree that carbofuran is
readily available in local markets (Table 6).
3.5 Adoption of improved germplasm
MARIF researchers in the maize OFR program have long believed that maize
productivity can be increased through farmer adoption of improved germplasm.
Indeed, the original intent of the OFR program was to ascertain why farmer adoption
of improved varieties had not proceeded more quickly2
Much of the evidence on the adoption of improved germplasm for maize has
been presented in previous sections. The use of unimproved local varieties has fallen
off dramatically while the use of Arjuna has increased among all farmer classes (Fig.
14, also Fig. 4). In addition, the use of hybrids has begun to pick up, especially among
OFR farmer cooperators, and especially since 1986. Unfortunately, the adoption
survey did not ascertain the year of adoption of Arjuna among recent adopters.
Consequently, it is not known to what extent the "Arjuna" reported by farmers is from
seed purchased many years ago, which by now will have lost its varietal purity.
Finally, it should be noted that farmer adoption of improved germplasm can
be traced to factors other than the OFR program. Widespread extension campaigns
1. This is a worrisome trend. The use of carbofuran should properly be restricted to shootfly hotspots,
especially for late plantings in the second season. Its widespread prophylatic use is likely to have
undesirable ecological effects.
2. A secondary objective, from the very beginning, was to ascertain why improved germplasm appeared
to perform more poorly in farmers' fields than in research trials.
Fig. 13: Use of Carbofuran
by Farmer Class
(Use on maize every season, or use on other crops)
Cooperators I 5- -_1 7I
Control
B
Farmer Class
I [ Maize every season Other crops
80-
70-
60-
50-
40-
30-
20-
10-
0-
I Irvi I-vvuyil arvi 3 1
-
-
1
I
Fig. 14: Adoption of Improved Maize
Germplasm (1984 vs. 1990)
-4
1984
1990
Year
50-
40
30
20
10
0
Table 6
Shootfly problem and control, by farmer class
Farmer class: a/
ALL
Farmers
Variable A B C (1990) b/
---------------------------------.----.--------------.-
Sample size 12 30 18 60
Weight 0.01 0.13 0.86
Report shootfly as a
common problem on
maize: c/
Method of shootfly control
this season:
Overplanting
Carbofuran
Other insecticides
Use carbofuran on maize
every season: *** d/
Use carbofuran
on other crops + e/
42% 71% 82% 80%
66% 19% 23% 23%
75% 61% 41% 44%
Feel carbofuran is readily available:
a/ Group A = Farmers that directly cooperated with the
OFR program; Group B = Farmers that did not directly
cooperate with the OFR program, but who lived in the
same village where the program was operating; Group C =
Farmers living in villages with no direct contact with
the OFR program.
b/ Weighted by the proportion of each class in the
population (row Labeled "weights")
c/ Chi-square = 5.6, 2 df, prob. = 0.06.
d/ Chi-square = 9.7, 2 df, prob. = 0.01.
e/ Chi-square = 3.5, 2 df, prob. = 0.17.
Significance levels: *** = .01 level; ** = .05 level;
* = .10 level; + = .20 level.
and private sector promotions have been conducted in the study area over an ex-
tended period of time. Private companies have participated in the production and
marketing of Arjuna seed as well as hybrid seed. The OFR program can probably
take only limited credit for the shift among farmers away from less productive maize
varieties.
3.6 Interactions among components sequential adoption
When taking the three major components of the recommendations -- im-
proved germplasm, chemical shootfly control and reduced planted density -- it is
interesting to note that more than half of the farmers in the study area have adopted
all three components. It is difficult at this stage to determine if farmers adopted these
components in a sequence or as a package. Of course, it probably doesn't matter --
sequences tend to blur into packages when adoption is swift.
Those farmers not adopting all three components of the recommendations
have tended to take up the germplasm and chemical shootfly control practices, while
leaving planted density unchanged (Table 7). Additional statistical analysis suggests
that these are farmers that initially used reasonable planted densities. A variable
measuring "seeds per hill used by farmers before carbofuran was adopted" was found
to be highly and positively correlated with another variable measuring "the change in
seeds per hill attributable to the adoption of carbofuran" (r = 0.74). A simple linear
regression (not shown) suggests that farmers using 10% fewer seeds per hill than
other farmers before carbofuran adoption reduced their seed rates 8.7% less than
other farmers after adoption. That is, farmers with initially reasonable densities had
less reason to reduce those densities after adoption.
Table 7
Adoption of three technical components in the second maize crop,
by farmer class
Farmer class: a/
All
Farmers Using b/ A B C Farmers c/
.....% . ...
Sample size 12 30 18
Weights 0.01 0.13 0.86
Local variety,
overplanting
Improved variety,
overplanting
Local variety,
chemical shootfly control
Improved variety,
chemical shootfly control,
no density reduction
Improved variety,
chemical shootfly control,
reduced density
8 13 6 7
0 13 6 7
0 6 6 6
42 36 24 26
50 32 59 55
a/ Group A = Farmers that directly cooperated with the OFR
program; Group B = Farmers that did not directly cooperate
with the OFR program, but who lived in the same village
where the program was operating; Group C = Farmers living
in villages with no direct contact with the OFR program.
b/ Farmer class and adoption behavior appear to be inde-
pendent, according to the chi-square test
c/ Weighted by the proportion of each class in the popula-
tion (row labeled "weights").
4 Logit analysis
The analyses presented above are not entirely satisfactory in the sense that
they are, for the most part, restricted to examining the relationships between two
variables at a time. In this section, analyses are presented that assess farmer adoption
of new technology in terms of several independent variables simultaneously. Adop-
tion decisions are typically represented as a binary variable (e.g., adoption vs. non-
adoption)1, so logit analysis is used.
4.1 Logit analysis: A review
Logit analysis examines the relationship between a binary variable (a variable
that takes on only two values) and one or more explanatory variables. The binary
dependent variable is usually coded 0 or 1. When dealing with a single independent
variable, logit uses a model of the form:
Li = In {Pi/(l-Pi)} = B1 + B2Xi
where Li (called the logit) is the log of the odds ratio Pi/(1-Pi) of event i occurring,
i.e., where Pi represents the probability of event i occurring (e.g., farmer adoption of
improved germplasm) and 1-Pi represents the probability of event i not occurring. Li
takes on values of plus to minus infinity while Pi takes on values of 0 to 1. Li, then,
can be explained in terms of a set of independent variables, e.g., Xi (Gujarati, 1988).
1. Adoption decisions are not always best modeled in terms of "yes" vs. "no". Adoption can be reversi-
ble and partial adoption (e.g., use of fertilizer, but at lower rates than recommended) is sometimes a
possibility. However, the adoption decisions of interest in this analysis -- shootfly control by over-
planting vs. use of chemical control, or use of improved vs. local germplasm -- can be represented
fairly readily as yes/ no decisions.
Numerous software packages can readily estimate logit models. Most of these
packages (including the one used in this analysis -- Statpac Gold) use maximum likeli-
hood techniques. The chi-square statistic is used to measure the overall significance
of the equation in explaining the dependent variable (equivalent to the F-test in
ordinary least squares multiple regression).
4.2 Current method of shootfly control
A model was estimated in which the current use of chemical shootfly control
(1990 post-rainy season) on maize was explained in terms of a number of independent
variables measured for that same crop cycle, including variety, planting date, cropping
pattern, frequency of extension contact, farm size, farm class (OFR cooperators vs.
others), and seed rate. Discrete variables (all variables except for farm size, extension
contact and seed rate) were introduced into the equation as dummy variables.
The equation did well in significantly explaining the variability in the depend-
ent variable (chi-square significant at the 0.01 level) (Table 8). As expected, variety
and cropping pattern were found to be significantly related to the use of chemical
shootfly control. Farmers using improved germplasm, or using a maize-maize pat-
tern, were more likely to use chemical shootfly control.1 Other, seemingly reasonable
factors such as farmer class, farm size or the frequency of extension contact were not
found to be significantly related to the adoption decision. This does not imply that
these factors were unimportant in the past in determining the rate and incidence of
adoption -- it only suggests that these factors have become relatively unimportant in
distinguishing adopters from non-adopters, now that adoption of chemical shootfly
control has become so widespread.
1. Seed rate was found to be related to the dependent variable at a lesser level of significance (0.15
level), but with an incorrect sign. Adoption of chemical shootfly control was found associated with
(slightly) higher seed rates.
Table 8
Model: Current use of chemical shootfty control
Logistic Regression Coefficients
Var. a/ Coefficient Std.Error T-Ratio
Constant
Variety
Pt. Date
Cr. Pattern
Freq. Ext.
Farm Size
Farmer Coop.
Seed Rate
-9.86958
5.44891
0.73443
2.81595
-0.25469
0.10227
1.19563
0.22032
5.35054
1.81909
1.01957
1.70436
0.33554
0.22559
1.60484
0.14957
-1.84459
2.99541
0.72033
1.65221
-0.75904
0.45336
0.74501
1.47307
Prob.
0.06511
0.00275
0.47133
0.09849
0.44784
0.65031
0.45627
0.14072
Log of likelihood function = -16.090219
Chi-square statistic for significance of equation = 22.41293
Degrees of freedom for chi-square statistic
Significance level for chi-square statistic
= 7
= 0.0022
Variety: improved germplasm (Arjuna or hybrid) = 1, otherwise 0;
PL. Date: February or March = 1, otherwise 0;
Cr. Pattern: maize-maize = 1, otherwise 0;
Farmer Coop: farmer cooperator = 1, otherwise 0;
Seed rate, frequency of extension and farm size were introduced
as continuous variables.
4.3 Timing of adoption of chemical shootfly control
Another model was estimated to explain the timing of adoption of chemical
shootfly control in terms of four independent variables: variety, farm size, frequency
of extension contact, and farmer class. Farmer class was subdivided into two varia-
bles: OFR cooperators vs. non-cooperators, and OFR villages vs. control villages.
These two variables, along with variety, were represented in the model by dummy
variables. Farmers were divided into early (before 1987) and late (1987 or later)
adopters. Strictly speaking, this model aims to explain past adoption decisions in
terms of present management practices and farmers' circumstances. This is useful
only if past and present practices and circumstances are highly correlated.
The model did not perform well in explaining the variability found in the
dependent variable (chi-square statistic significant at the 0.34 level). The only inde-
pendent variable found to be significantly related to the dependent variable was the
dummy representing OFR cooperator vs. non-cooperator (Table 9). When the equa-
tion was reestimated with only this single independent variable, the significance of the
chi-square statistic improved to the 0.05 level (Table 10).
This result statistically confirms the differential adoption behavior illustrated
in Fig. 11, in which adoption by farmer cooperators precedes that by other farmer
classes. It also indicates, however, that other, seemingly reasonable factors had little
effect in determining the timing of adoption of chemical shootfly control.
4.4 Current use of improved germplasm
A model was estimated to examine factors associated with farmer use of
improved maize germplasm during the 1990 post-rainy season. Independent variables
included in the model were: frequency of extension contact, farmer class (cooperator
vs. other farmers), farm size, village (OFR village vs. control villages), cropping
pattern, planting date (February and March, vs. April plantings), seed rate, and
Table 9
Model: Timing of adoption of chemical shootfly control
(full model)
Logistic Regression Coefficients
Var. a/ Coefficient Std.Error T-Ratio Prob.
Constant 0.99427 0.72417 1.37297 0.16974
Farmer Coop. -1.42094 0.81100 -1.75207 0.07977
Village -0.25294 0.74015 -0.34175 0.73255
Farm Size 0.07298 0.12316 0.59256 0.55350
Freq. Ext. -0.23781 0.26436 -0.89960 0.36833
Log of likelihood function = -30.644715
Chi-square statistic for significance of equation = 4.500733
Degrees of freedom for chi-square statistic = 4
Significance level for chi-square statistic = 0.3425
a/
Farmer Coop: farmer cooperator = 1, otherwise 0;
Village: OFR village = 1, otherwise 0;
Frequency of extension and farm size were used as continuous
variables.
Table 10
Model: Timing of adoption of chemical shootfly control
(restricted model)
Logistic Regression Coefficients
Var. a/ Coefficient Std.Error T-Ratio Prob.
Constant 0.53900 0.33630 1.60274 0.10899
Farmer Coop. -1.38629 0.76765 -1.80590 0.07094
Log of Likelihood function = -31.116823
Chi-square statistic for significance of equation = 3.556519
Degrees of freedom for chi-square statistic = 1
Significance Level for chi-square statistic = 0.0593
a/
Farmer Coop.: Farmer cooperator = 1; otherwise 0.
method of shootfly control. Seed rate, frequency of extension contact and farm size
were entered as continuous variables; others were entered as dummy variables.
The model did well in significantly explaining the variability in the dependent
variable (chi-square significant at the 0.01 level) (Table 11). Seed rate and shootfly
control method were found (not surprisingly) to be associated with the use of im-
proved germplasm: farmers using lower seed rates and chemical shootfly control were
more likely to use improved maize germplasm. This is additional confirmation of the
adoption behavior described in Table 7.
Table 11
Model: Current use of improved germplasm
Logistic Regression Coefficients
Var. a/ Coefficient Std.Error T-Ratio Prob.
...............................................................
Constant 18.96174 8.59029 2.20735 0.02731
Seed Rate -0.63999 0.28977 -2.20863 0.02722
Cr. Pattern -1.23168 1.58373 -0.77771 0.43675
Freq. Ext. 0.16021 0.67292 0.23808 0.81181
Farmer Coop. 1.89450 4.25740 0.44499 0.65635
Farm Size 0.22590 0.85861 0.26310 0.79247
PI. Date -1.77178 1.64250 -1.07870 0.28070
Village 0.67283 1.59625 0.42151 0.67340
Method 4.91906 2.08172 2.36298 0.01815
Log of likelihood function = -7.777954
Chi-square statistic for significance of equation = 29.7483
Degrees of freedom for chi-square statistic = 8
Significance level for chi-square statistic = 0.0002
a/
Cr. Pattern: maize-maize = 1, otherwise 0;
Farmer Coop: farmer cooperator = 1, otherwise 0;
PI. Date: February or March = 1, otherwise 0;
Village: OFR village = 1; otherwise 0;
Method: Chemical shootfly control = 1, otherwise 0.
Seed rate, frequency of extension and farm size were intro-
duced as continuous variables.
5 Conclusions
The MARIF OFR program in the young volcanic soils study area of Malang
District, East Java, has developed recommendations for shootfly control, plant stand
establishment, soil fertility management1 and variety for maize cultivation. Results of
an adoption survey conducted during the post-rainy season of 1990 indicate that some
of these recommendations have been widely adopted by farmers. Can a substantial
portion of this adoption be credited to the activities of the OFR program? If so, what
level of benefits have been earned as a result of investing in OFR? Finally, what
further activities might be used to expand or accelerate the distribution of these
benefits, or to follow-up on issues raised by the OFR program?
5.1 Linking OFR activities with farmer adoption
There have been relatively few studies that have quantified the economic
returns to investment in OFR; most studies on impacts of agricultural research
concentrate on returns to investment in germplasm development. Studies focusing on
returns earned by OFR (or crop management research in general) include Traxler
(1990) and Martinez and Sain (1983).
Four specific steps for assessing the economic impact of OFR are described by
Traxler (1990). The first three steps describe conditions necessary for a given re-
search activity to be logically linked to an increase in productivity, production and
eventually producers' surplus. Since these are asserted to be necessary conditions,
1. Little has been said about farmer adoption of improved soil fertility management techniques.
While it appears that there have been some changes in farmers' practices in this regard, data from the
adoption survey on farmers' fertilizer use practices are not available due to the timing of the survey (it
was carried out too early in the crop season to obtain information on fertilizer use).
they can be considered sequentially. The failure of any one of the conditions results
in an economic value of zero being assigned to the corresponding research invest-
ment.
The conditions are:
1) Research must have led to the discovery of an improved management prac-
tice which was embodied in a new recommendation issued to farmers.
2) Producers must have modified their management practice in a manner
consistent with the change in recommendation.
3) There must have been evidence of causality between the change in practice
and the change in recommendation. Simple heuristic criteria are suggested for assess-
ing causality:
a) the recommended practice and producer practice should have
changed in the same manner during the study period;
b) dissemination of the recommendation must have preceded the
change in farmers' practice;
c) it was "unlikely" that producers would have developed the change in
practice without the benefit of formal research.
4) The fourth and final step is to calculate the increase in economic surplus
generated by each recommendation that successfully meets all prior conditions.1
In the case of the MARIF OFR program, the first condition is met by the
information provided in Table 2 (above). Recommendations were developed and
disseminated to farmers (though tenuous research-extension links slowed the dissem-
1. Increases in economic surplus should not be estimated for recommendations failing to meet all
prior conditions. These recommendations were either not accepted by farmers, or farmer use of new
technology cannot be logically linked to the research program.
nation process, see section 2.3). The second condition -- farmer adoption -- has been
the main topic of this paper. Evidence was shown on farmer adoption of three
recommendations: chemical shootfly control; reduced planted densities; and use of
improved maize germplasm. (Information on farmer adoption of improved fertilizer
use practices was not collected in the adoption survey.)
Conclusions on causality are summarized in Table 12. In general, farmers did
adopt practices similar to recommendations, and recommendations did precede (for
the most part) farmer adoption. (The case of varietal change is less clear because
recommendations favoring new varieties had been in effect for more than a decade
before the OFR program even commenced.)
Finally, it is judged that farmers would probably not have developed chemical
shootfly control measures on their own. Farmers were accustomed to using carbofu-
ran on rice but had not tried using it on maize. Chemical shootfly control was not part
of extension recommendations disseminated prior to the OFR program. Moreover, it
was shown that OFR farmer cooperators adopted chemical shootfly control before
other farmers in the study area. Reduced planted density was shown to be strongly
linked to shootfly control, so the same conclusions apply to this component as to
chemical shootfly control.
However, the evidence regarding varietal adoption is less compelling. Numer-
ous other campaigns and programs, involving on-station researchers, extension
workers and the private sector, were active in fostering varietal change. It seems
likely that farmers would have adopted improved varieties even in the absence of the
OFR program.
Table 12
Evidence on causal links between OFR program and farmer
adoption
Recommendation
Chemical Reduced Improved
Shootfly Planted Maize
Criteria Control Density Germplasm
Recommendation and
farmer practice changed in
the same way Yes Yes Yes
Dissemination of the
recommendation preceded the
change in farmer's practice Yes Yes Some
Farmers unlikely to have
developed improved practice
without benefit of
formal research Probably Probably No
................................................................
5.2 Benefits earned from OFR
Measurement of the benefits earned through OFR involves estimating the
economic surplus generated by each recommendation that successfully meets all prior
conditions. The size and distribution of the surplus is determined by the magnitude of
the induced shift in the supply curve and by the elasticities of supply and demand. If
the area being studied produces a small share of the commodity consumed in the
country (or, as in Indonesia, if increased production substitutes for imports), changes
in consumers' surplus need not be considered and changes in returns above variable
cost (RAVC) are an exact measure of producer surplus (Traxler, 1990).
The annual total value of an improved practice then, is the product of the per
hectare impact times the area over which the information or innovation is employed
(equation 1) (Traxler, Renkow and Harrington, 1990).
QRit = Pi kit At (1)
Where QRit is the total quasi-rent1 generated by innovation i in year t, pi is the per
hectare change in RAVC due to innovation i, kit is the percent of enterprise harvest-
ed area employing innovation i in year t and At is the total harvested area in year t.
Farm survey and on-farm trial data can be used to estimate the values of pi
and kit, while secondary data can be used to identify At. The estimate of the total
change in producer surplus can then be compared with research costs, Ct, to calculate
the internal rate of return (IRR). The IRR is the discount rate (i), which satisfies
equation 2.
(Sum)t (QRt Ct) / (l+i)t = 0 (2)
In the case of the MARIF OFR program, economic benefits were estimated
for impacts of chemical shootfly control and reduced planted density only. Estimates
for yield response, RAVC per ha (an estimate of research impact), proportion of the
area covered by the new technology, total post-rainy season maize harvested area for
Malang District, and estimated quasi-rents are shown in Table 13, for the years 1984-
1990. These benefits total over US$4.3 million. The net present value of these ben-
fits (discounted back to 1984) is about US$2.6 million.
1. Quasi-refit is defined as "the payment to any input in temporarily fixed supply" (Mansfield, 1970).
In this context it is the total increment in profits to producers, as measured by the RAVC criterion.
Table 13
Economic benefits from OFR
(shootfly control and reduced planted density)
Proportion Total
Yield RAVC area maize Quasi-
Change change covered area rent
Year (kg/ha) a/ ($/ha) b/ (%) c/ (ha) d/ (S/ha)
1984 600 $50 0% 30000 $0
1985 600 $50 5% 30000 $75,000
1986 600 $50 28% 30000 $420,000
1987 600 $50 44% 30000 $660,000
1988 600 $50 67% 30000 $1,005,000
1989 600 $50 72% 30000 $1,080,000
1990 600 $50 72% 30000 $1,080,000
a/ Krisdiana, et at., 1991, Table 7.4. Estimated yield change
calculated from weighted average of yield responses measured
in farmer-managed verification trials in which chemical shootfly
control was superimposed on the unmodified farmers' practice.
b/ Maize price estimated at Rp. 200/kg or about $100/ton. Note
that farm-level maize price increases over time were almost en-
tirely off-set by exchange rate devaluations of a similar magni-
tude (data not shown). The increment in costs that vary estimated
at about $11 per ha per cycle, but may be considerably less. Note
that part of the pesticide cost is offset by reduced seed costs;
similarly, the pesticide is used at a very low dose.
c/ Figure 11
d/ Krisdiana, et at., 1991, p. 145. Includes only the post-rainy
crop seasons.
These estimates are very conservative. There is reason to believe that adop-
tion of shootfly control and reduced planted density has proceeded beyond the bor-
ders of Malang District. Moreover, benefits from the OFR program will continue to
be earned after 1990, the final year shown in the analytical model. In addition, the
OFR program can claim some credit for farmer adoption of improved varieties and
fertilizer management practices. Finally, benefits are only claimed for the post-rainy
season, not for the rainy season. None of these additional benefits are shown,
however, in Table 13.
For example, consider the effect of the following two assumptions (neither of
them unreasonable): 1. Farmer adoption of shootfly control and reduced planted
density covers twice the area indicated in Table 13 (because some farmers use the
technology in the rainy season, or because farmers in neighboring Districts operating
similar farming systems have picked up these practices). 2. Farmer use of these
techniques continues through 1995. In this case, total net benefits (quasi-rents) sum
to almost US$20 million, with a discounted value in 1984 of over US$9 million
(Table 14).
The costs of the OFR program have not been estimated in detail. They are
probably small, however, in comparison to the benefits earned, even when all OFR-
related expenses (salaries, vehicle depreciation and interest charges, operational
expenses, per diem, etc.) are included for the YVS study area. As specific estimates
of these costs are made, it will be possible to estimate the internal rate of return to
investment in OFR, as noted above.
5.3 Future activities
Evidence summarized in this paper suggest that the MARIF OFR program
has had a substantial impact, measured in millions of dollars of benefits per year for
maize producers. A number of questions remain, however. Follow-up activities to
address these might include the following:
Table 14
Sensitivity analysis on economic benefits from OFR
(shootfly control and reduced planted density)
Effect of expanded area of adoption, and additional years
Proportion Total
Yield RAVC area maize Quasi-
Change change covered area rent
Year (kg/ha) a/ ($/ha) b/ (%) c/ (ha) d/ ($/ha)
1984 600 $50 0% 60000 $0
1985 600 $50 5% 60000 $150,000
1986 600 $50 28% 60000 $840,000
1987 600 $50 44% 60000 $1,320,000
1988 600 $50 67% 60000 $2,010,000
1989 600 $50 72% 60000 $2,160,000
1990 600 $50 72% 60000 $2,160,000
1991 600 $50 72% 60000 $2,160.000
1992 600 $50 72% 60000 $2,160,000
1993 600 $50 72% 60000 $2,160,000
1994 600 $50 72% 60000 $2,160,000
1995 600 $50 72% 60000 $2,160,000
a/ Krisdiana, et al., 1991, Table 7.4. Estimated yield change
calculated from weighted average of yield responses measured
in farmer-managed verification trials in which chemical shootfly
control was superimposed on the unmodified farmers' practice.
b/ Maize price estimated at Rp. 200/kg or about $100/ton. Note
that farm-level maize price increases over time were almost en-
tirely off-set by exchange rate devaluations of a similar magni-
tude (data not shown). The increment in costs that vary estimated
at about $11 per ha per cycle, but may be considerably less. Note
that part of the pesticide cost is offset by reduced seed costs;
similarly, the pesticide is used at a very low dose.
c/ Figure 11, extrapolated to 1995
d/ Area assumed to be twice that shown in Table 13.
A second adoption survey, with a larger sample size, and covering Malang
district and neighboring Districts. This would allow more precise estimation of the
variables discussed in this paper;
A study to compute the costs of the OFR program over time, so that the internal
rate of return to OFR investment can be estimated.
Special purpose studies to ascertain extent and incidence of adoption of
improved soil fertility management practices, and the role that the MARIF OFR
program might have played in fostering this adoption;
Continued research on alternatives to carbofuran for shootfly control, given
the ecological danges associated with this chemical.
Incorporation of a sustainability perspective in future OFR activities, e.g.,
measurement of soil erosion or land degradation associated with farmers' practices vs.
improved practices, through long-term trials or farmer monitoring programs.
In addition, other programs underway at MARIF that aim to develop im-
proved crop management practices, or improved cropping patterns, in a farming
systems perspective, should be encouraged to monitor adoption and estimate benefits
earned from research, as this paper has attempted to do.
References
Anderson, J., 1990. "FSRE Impact Inquisition: Some Investor Issues". Presented at
the 1990 Asia Farming Systems Research and Extension Symposium, AIT,
Bangkok, Thailand, November 18-23, 1990.
Byerlee, D., L. Harrington and D. Winkelmann, 1982. "Farming Systems Research:
Issues in Research Strategy and Technology Design". American Journal of
Agricultural Economics. Vol. 64, No. 5, December 1982.
CIMMYT, 1990. 1989/90 CIMMYT World Maize Facts and Trends: Realizing the
Potential of Maize in Sub-Saharan Africa. Mexico, D. F.: CIMMYT.
Dahlan, M., Heriyanto, Sunarsedyono, Sri Wahyuni, C. van Santen, J. van Staveren,
and L. Harrington, 1987. Maize On-Farm Research in the District of Malang.
Malang Research Institute for Food Crops.
Gujarati, D., 1988. Basic Econometrics. New York: McGraw Hill.
Krisdiana, R., M. Dahlan, Herianto, C. van Santen, and L. Harrington, 1991. "From
Diagnosis to Farmer Adoption: The Story of MARIF's Maize On-Farm Reserch
Program in East Java, Indonesia", in Tripp (ed.), Planned Change in Farming
Systems: Progress in On-Farm Research. New York: Wiley.
Manwan, I. and M. Oka, 1990. "Research and Development for Sustainable Farming
System in Indonesia". Presented at the 1990 Asian Farming Systems Research
and Extension Symposium, AIT, Bangkok, Thailand, November 1990.
MARIF, 1985. "Report on the Maize Survey -- December, 1984". MARIF Working
Paper No. 8.
Martinez, J. and G. Sain, 1983. "The Economic Returns to Institutional Innovations
in National Agricultural Research: On-Farm Research in IDIAP, Panama".
CIMMYT Economics Working Paper 04/83.
Sumberg, J., and C. Okali, 1988. "Farmers, On-Farm Research and the Development
of New Technology". Experimental Agriculture. Volume 24, p. 333-342.
Timmer, P., (ed.), 1987. The Corn Economy of Indonesia. Ithaca: Cornell University
Press.
Traxler, G., 1990. The Economics of Crop Management Research. Unpublished
PhD dissertation, Iowa State University.
Traxler, G., M. Renkow and L. Harrington, 1990. "Assessing the Impact of New
Technology: Three Levels of Analysis". Presented at the 1990 Asian Farming
Systems Research and Extension Symposium, AIT, Bangkok, Thailand, 19-22
November, 1990.
Tripp, R., P. Anandajayasekeram, D. Byerlee and L. Harrington, 1990. "FSR:
Achievements, Deficiencies and Challenges for the 1990's". Presented at the 1990
Asian Farming Systems Research and Extension Symposium, AIT, Bangkok,
Thailand, November 1990.
|
