Uploaded by Nguyễn Long

document

advertisement
Articles
The Adoption of Reduced Tillage: The Role
of Human Capital and Other Variables
Michael R. Rahm and Wallace E. Huffman
Key words: a d o p t i o n , a g r i c u l t u r e , c o r n , e d u c a t i o n , h u m a n capital, p r o b i t , r e d u c e d
tillage.
Although Schultz stresses the self-interest and formation, and experience affect these decilarge stake that farm operators have in their sions?
soil resources, few soil conservation studies
Numerous agricultural studies support the
develop and empirically test models of the hypothesis that investments in formal educamicroeconomic decision to adopt a soil con- tion and extension enhance allocative skills.
servation practice such as reduced tillage, t Studies by Huffman and Petzel show that inRecent papers by Kramer, McSweeney, and vestments in schooling and extension improve
Stavros; and Lee and Stewart are exceptions. farmers' response to economic disequilibria.
A number of published studies are either de- Fane and Khaldi show that farmers who have
scriptive or problematic. For example, Soth invested in more years of formal schooling
warns of additional soil loss caused by expand- are better cost minimizers. However, few
ing U.S. agricultural exports, and Heady and studies of the adoption and diffusion of new
Short attempt to quantify this relationship. technologies investigate the effect of human
Most of the studies conclude (or imply) that capital investments on adoption behavior, z
the lack of mass adoption of soil conserva- Earlier studies by Griliches, Dixon, Mansfield,
tion practices such as reduced tillage indi- Globerman, and Romeo place primary emcates either market failure or managerial in- phasis on profitability in explaining differential
efficiency. Why do some farm operators rates of adoption. Other studies have emutilize reduced tillage, while others do not? phasized risk and uncertainty, tenure type,
What are the key variables determining the and credit constraints. (See Feder, Just, and
economic feasibility of such practices; and Zilberman for a recent survey of the literature
how do investments in education, health, in- on adoption of agricultural innovations.) Previous agricultural studies of adoption and allocative efficiency generally have employed
The authors are an assistant professor of economics and business,
aggregate data. Microdata may yield new inMacalester College, a n d a professor of economics, Iowa State
formation
about adoption behavior because
University, respectively.
farm-level observations eliminate potential
Paper No. J-11442 of the lowa Agriculture and Home Economics
Experiment Station, Project 2314.
problems with aggregation bias and because
The authors wish to thank John Miranowski, T. W. Schuitz, and
effects of a wider set of human capital varitwo anonymous referees for helpful comments on earlier drafts of
ables (e.g., experience, health, and private
the paper.
Review was coordinated by Peter Berck, associate editor.
and public information) can be investigated.
~ Schultz writes: " W e proclaim to the world that U.S. farmers
are second to none in their agricultural achievements. When it
comes to soil erosion, the prevailing implicit assumption is that
farmers have no perception of the value of their soil resources and
that they are indifferent to soil Ioss'" (p. 18).
2 Some of the rural socioiogy literature ~Rogers and Shoemaker)
has suggested that adoption depends on the decision makers'
education and information.
Copyright 1984 American Agricultural Economics Association
Downloaded from http://ajae.oxfordjournals.org/ at Penn State University (Paterno Lib) on September 20, 2016
This paper presents a model of adoption behavior and explains differences
econometricaily in farmers' decisions to adopt reduced-tillage practices and in the
efficiency of farmers' adoption decisions. The empirical results, obtained from
microdata, show that the probability of adopting reduced tillage in corn enterprises
differs widely across farms and depends on soil characteristics, cropping systems, and
size of farming operation. The results also show that farmers' schooling enhances the
efficiency of the adoption decision.
406
N o v e m b e r 1984
Amer. J. Agr. Econ.
A Model of Adoption Behavior
Farmers face outcomes from the adoption of
production technologies that are uncertain. In
this model, farmers are assumed to make
adoption decisions based upon an objective of
utility maximization. Denote a technology
index by t, where t is equal to 1 for the old
technology and 2 for a new or different technology; and denote a utility function that ranks
the ith firm's preference for these technologies
b y U ( R t i , A , ) . Utility depends on a vector R t
of moments that describe the distribution of
net returns for technology t, including adoption cost, a n d a vector A t of other attributes
associated with the technology. 3 The variables
Rti and A t i a r e unobserved and unavailable,
but a linear relationship is postulated for the
ith firm between the utility derived from the
tth technology a n d a vector of observed firmspecific characteristics Xi (e.g., soil type,
cropping system, farm size) a n d a zero mean
random disturbance term et:
(1)
Uti = Xio~t + e,, t = 1, 2;
i= 1,...,n.
Farm operators are assumed to choose the
technology that gives them the largest utility.
The ith firm adopts the new technology if Uzi
3 Higher moments and other attributes may be important factors
affecting the decision to adopt reduced tillage. For example, the
impact of reduced tillage on the variance of yields and aesthetic
preferences for neatly tilled and clean-looking fields may inhibit
some farm operators from adopting reduced tillage.
exceeds Uli, and the qualitative variable Di
indexes the adoption decision:
(2)
Di =
I if Uli < Uzi, new technology is
adopted and
replaces old
technology,
IL
O if Uli >- U2i, old technology is
continued.
The probability that Di is equal to one can be
expressed a s a function of firm-specific characteristics:
(3)
Pi = P r ( D i = 1) = P r ( U l i < U2i)
= P,.(Xioq + eli < Xiol2 -~- e2i)
= Pr[eli -
e2i <~ Xi(og2 -
0~1)]
= P , . ( ~ < Xifl) = F ( X i f l ) ,
where P~(.) is a probability function, /~ =
eli - e2i is a random disturbance term, /3 =
a2 - ~~ is a coefficient vector, and F(Xi/3) is
the cumulative distribution function for
evaluated at Xi/3. Thus, the probability of the
ith firm adopting the new technology is the
probability that the utility of the old technology is less than the utility of the new technology or the cumulative distribution function F
evaluated at X~q The exact distribution for F
depends on the distribution of the random
term ~ = e~i - e2i. If p~ is normal, then F is a
cumulative normal; and if/.~ is uniform, then F
is triangular. 4
The marginal effect of a variable Xi on the
probability of adopting the new technology is
aPi/axij = f(x,/3) 9
where f(.) is the marginal probability density
function of ~ . Clearly, the direction of the
marginal effect is determined by the sign of/3j,
but q represents coefficient differences %j oqj. Thus /3j is expected to be positive (negatire, zero) if o~zj is positive and greater than
(less than, equal to) cq~.
Human capital variables are not included in
X, because adopting the new technology is not
always economically feasible. For a given soil
type and cropping system, increasing a farmer's schooling is not expected always to increase the probability of adopting reduced
tillage technology. Human capital variables,
however, are expected to increase the proba4 See Takeshi Amemiya for a broad survey of qualitative response models and economic applications. Jamison and Lau also
employ a similar model for explaining the adoption of chemical
inputs by Thai farmers.
Downloaded from http://ajae.oxfordjournals.org/ at Penn State University (Paterno Lib) on September 20, 2016
The objectives of this paper are (a) to present a model of adoption behavior and (b) to
explain differences econometrically in farmer's decisions to adopt reduced-tillage practices and the efficiency of farmers' adoption
decisions. The empirical results, obtained by
fitting the model to microdata for Iowa farms,
show that the probability of adopting reduced
tillage in Iowa corn enterprises differs widely
across farms and depends on soil characteristics, cropping system, and size of farming operation. In addition, these results show that
investments in farmers' formal schooling and
continuing education enhance the efficiency of
the adoption decision.
The paper is organized as follows: section 1
presents the adoption model; section 2 presents the empirical analyses of the tillage decisions of Iowa farm operators; and the final
section presents conclusions of the analysis.
Rahm and Huffman
firm,
(4) Ei - I D , - P~I = g(Z~~, + e4), g' > O.
Efficient adoption decisions imply adopting
new technology (Di = 1) when Pi is large or
nonadoption (Di = 0) when P, is small. In both
situations, the indicator of efficiency Ei is relatively small. 6 Grossly inefficient adoption decisions imply adoption when the predicted
probability is small or nonadoption when the
predicted probability is large. For both of
these situations, the indicator of adoption
efficiency is relatively large. Thus, larger values of E, are assumed to signal greater inefficiency in the adoption decision.
Efficiency of the adoption decision is hypothesized to be related to characteristics that
indicate allocative skills of farm operators and
a zero mean random disturbance ~. Following
human capital theory, allocative skills are assumed to be acquired or learned rather than
innate. In particular, farmers' investments in
schooling, experience, information, and health
are expected to enhance allocative skills and
to increase the efficiency of adoption decisions. If this proposition is correct, farm
operators who have invested in these activ-
5 In this adoption model, the actual net return for technology t is
uncertain. Utility of technology t is stochastic and depends upon
moments of the net-return distribution and attributes of the technology. Furthermore, decision makcrs arr assumed to behave as if
they know with certainty the parameters of the utility function.
Although some decision makers are actually misinformed, the
bulk of decision makers are assumed to be informed about the q
after the early adoption phase. Thus, on average, they make
correct optimizing decisions by adopting new technology when
Uu < U2~. Because Pi = F(X~~)is a function of parameters of the
utility functions and they (B's) are estimated consistently, the size
of the predicted probability P~ = F(X~B) is information about
correct behavior, conditional on X~. For ah alternative optimizing
model that is applied to adoption decisions, see Jensen.
6 An absolute-value function is only one of several symmetric
functions that could be employed.
407
ities will be better informed about the existence and general performance of different
technologies, will make more accurate assessments of differences in farm-level performance, and will make more efficient adoption
decisions.
Empirical Analysis
The empirical specification of the adoption
model is employed to investigate the corn tillage decisions of Iowa farmers. Corn is the
leading row crop in terms of acreage planted
nationally and in Iowa. The reason for concentration on Iowa, and not on other major cornproducing states, is the availability of a high
quality microdata set that was coUected in
1977.
Evaluation of Reduced Tillage
Historically, farmers prepared soils for row
crops with the aid of a moldboard plow. This
implement digs deeply and "turns over" the
soil, burying old crop biomass and exposing
the subsoil. This tŸ
practice facilitated the
preparation of a fine, firm seedbed and the
mechanical control of weeds, but ir increased
the erosion susceptibility of the soil. In recent
years, reduced-tillage technology has been
replacing the moldboard plow. Full-width
reduced-tillage technology refers to tillage by
a set of nonmoldboard plow implements, such
as chisel plows, primary tillage discs, and field
cultivators. Such irnplements stir up and level
the soil and may cut up old crop biomass.
However, old crop biomass remains on the
soil surface. A second type of reduced tillage
technology is a ridge-plant, or till-plant, systeta. Other reduced-tillage technologies include no-till, where the seed is knifed into the
soil without any general seedbed preparation.
Reduced tillage may affect the mean, variance, and higher moments of the yield distribution as well as the optimal factor proportions used in crop production. Because reduced tillage retains crop residues on the soil
surface, the technologies have the potential to
decrease soil losses from wind and water erosion and to increase soil moisture by improving water infiltration and decreasing evaporation. While decreases in soil and moisture
losses may improve yields, the moist residue
Downloaded from http://ajae.oxfordjournals.org/ at Penn State University (Paterno Lib) on September 20, 2016
bility of farmers making the economically
" c o r r e c t " adoption decision.
Human capital variables may enhance the
efficiency of adoption decisions. The actual
outcome of the adoption decision is indexed
by D~; and the predicted probability of adopting new technology, conditional on Xi, can be
obtained from equation (3) by replacing/3 with
a consistent estimator /3, Pi = F(Xi[3). This
predicted probability measures the typical or
normal effect on P~ of variables that have
homogenous effects on the adoption decision. 5
This study proposes to use the absolute difference between D~ and P~ as an indicator of
efficiency of the adoption decision for the ith
Adoption of Redued Tillage
408
November 1984
Amer. J. Agr. Econ.
cover also affect root mass, root size, and rooting patterns. Roots
concentrate near the soil surface with reduced-tillage technologies. Finally, reduced-tillage systems increase both soil aggregation and soil density. Highly aggregated and compacted soils restrict root growth and water movement. Minoru Amemiya and
Griffith et al. discuss potential production problems.
8 This definition of reduced tillage permits farmers to use a
moldboard plow on some of their farm's c o m acres, for example,
to bury biomass that accumulates with no-till technology and a
crop rotation including meadow.
Downloaded from http://ajae.oxfordjournals.org/ at Penn State University (Paterno Lib) on September 20, 2016
cover and shallow tiUage may cause other of the farm operators had adopted the techproduction problems. 7
nology to reduce soil loss, while 24% of the
Most research indicates that the net impact farm operators had adopted the technology to'
on yield is related directly to the soil charac- decrease field-time requirements at critical peteristics, annual rainfall, cropping system, and riods.
other management practices. Griffith et al. report results of tillage experiments in the east- The Data and Empirical Specification
ern Com Belt. The research indicates that re- of the Model
duced tillage likely will increase corn yield on
well-drained and poorly structured soils, on The model is fitted to a sample of Iowa farms.
poorly drained soils when corn follows any- These farms are a random sample of all Iowa
thing but corn, and in areas with longer grow- farms in 1976 having farm sales or value of
ing seasons. Minoru Amemiya reports results production greater than or equal to $2,500
of tillage experiments in the western Corn (Hoiberg and Huffman). The farm operator is
Belt. Iowa experiments indicate that reduced identified as, and defined to be, the individual
tillage significantly increased yield during who was the primary farm-business decision
water-deficient years. Under favorable weath- maker in 1976. The statistics laboratory at
er conditions, yield differences among tillage Iowa State University designed the survey and
systems were insignificant.
interviewers collected detailed information
Reduced-tillage systems require less labor about the farm business and household.
and machinery hours per acre, especially at
The operational definition of reduced tillage
critical planting time; but most systems re- is the nonuse of the moldboard plow for prequire greater quantities of herbicides and fun- paring farmland for planting a row crop.
gicides for weed and pest conrol. The reduc- Farmers are classified as adopters of reducedtion in the cost of producing a given level of tillage corn technology if they did not use a
output resulting from a change in optimal fac- moldboard plow to prepare all of their corn
tor proportions depends upon input prices and acreage for planting in 1976. 8 Employing this
change in input usage. This factor proportion definition, 58% of the sample farms utilized
or allocative impact, however, seems to be reduced tillage.
less important than the technical impacts on
The empirical definition of farm-specific
the mean and variance of yields.
variables that are expected to influence utility,
Iowa farmers used reduced tillage (non- and thus determine the probability of adopting
moldboard plow technologies) on less than reduced-tillage corn technology, are:
500,000 acres in 1968, but by 1977 they utilized
reduced tillage on 9.1 million acres, or more
(5) Pi = P r ( D i = 1)= F /3o +
/~~ª ,
than one-third of the state's crop acreage (Erj--1
bach, Lovely, and Ayres). Soil maintenance is
the primary reason given by Iowa farm where Xi is ACRES, acres of corn planted in
operators for the adoption of reduced tillage. 1976; X2 is ACRESQ., (ACRES)2; X3 is
Results from a 1976 U.S. Department of Ag- SBRATIO, ratio of soybean acreage to corn
riculture survey of Iowa farm managers who acreage in 1976; X4 is SBRSQ, (SBRATIO)2;
had adopted reduced tillage indicated that 52% X5 is RAINFALL, normal annual rainfall (20year average annual precipitation for U.S.
Weather Bureau District closest to the farm);
7 The moist residue cover provides an excellent germination
X6
is SEASON, length of normal growing seaenvironment for weeds, insects, and other pests; may impede
herbicide and pesticide effectiveness; and may decrease soil temson (average number of growing-degree days
peratures 3~ to 6~ in the early growing season. In addition, the
between spring and fall dates of less than 50%
cooler and wetter soils may reduce the availability of important
frost probability); X7 is TENURE, ratio of
nutrients such as copper, zinc, boron, and manganese, and it may
also lengthen the time required for seed germination (thus shortencropland acreage rented-in to total cropland
ing the effective growing season). Also, because reduced tillage
acreage;
and X8-X26 is SDk, soil association
systems do not incorporate fertilizers to the depths of convendummy variables (k = I, . . . , 19).
tional tillage systems, fertilizer utilization and soil acidity may be
affected by the type of tillage system. Tillage depth and residue
A discussion of the expected signs for the
Adoption of Redued Tillage
Rahm and Huffman
9 The cropping systena of a farm is assumed to be unaffected by
choice of tillage practices. This seems to be a reasonable assumption given the importance of other factors such as the expected
crop and livestock prices and past cropping patterns in determining the farm enterprise mix.
erosion susceptibility. Based on the discussion
of the soil and moisture-conserving properties
of reduced tillage, the probability of adopting
reduced tillage is expected to be greater for
operators who farm hillier, lighter, and
better-drained soils than for operators who
farm flatter, heavier, and poorly drained soils
(Griffith et al.).
Because reduced-tillage systems decrease
evaporation and increase infiltration, the
profitability, and thus the probability of adopting reduced tillage, is expected to be greater
on those farms that receive less than the state
average annual precipitation (32 inches per
year). In addition, because reduced-tillage
systems conserve on field work during the critical pre-plant and post-harvest periods, the
probability of adopting reduced-tillage technology is hypothesized to be larger on those
farms where the growing season is shorter, a~
Finally, the ratio of acres of cropland rented-in
to total acres of cropland is included to test the
hypothesis that tenure affects the probability
of adopting reduced tillage (Lee and Stewart).
Tillage choices for renters versus owners may
reflect a shorter planning horizon or greater
risk aversion, which might discourage them
from adopting new technology.
EŸ narrowly defined human capital variables are hypothesized to determine the
efficiency of adoption decisions: 11
8
(6)
In Ei = In [ D i - Pi[ = 3'0+ ~ yjZª + ~,
j=i
where ZI is EDUCATION, years of formal
schooling completed by farm operator; Z2 is
EDDUMI, dichotomous variable, 1 if farm
operator obtained vocational training in high
school and 0 otherwise; Z3 is EDDUM2,
dichotomous variable, 1 if farm operator completed an agricultural major in college (e.g.,
agronomy, animal science, agricultural business) and 0 otherwise; Z4 is CONTINUED,
dichotomous variable, 1 if farm operator or
spouse attended short courses, conferences,
and meetings on Iowa State University campus, and 0 otherwise; Z5 is EXTENSION,
dichotomous variable, 1 if farm operator attended (frequently or sometimes) meetings,
t o However, some of savings in time may be lost because additional surface biomass could also cause a colder seedbed and
slower seed germination.
1~ By defining the dependent variable in equation (6) as In Ei
rather than Et, the disturbance terna ei will be distributed between
+oo and - ~ . Furthermore, the loge transformation led to estimated
coefficients that had generally larger t-ratios than the level specification.
Downloaded from http://ajae.oxfordjournals.org/ at Penn State University (Paterno Lib) on September 20, 2016
coefficients of regressors ente¡
equation (5)
foUows. The change in mean net farm income
resulting from adopting reduced tiUage is the
product of the change in expected per acre net
returns and the com acreage of the farm. For a
given per acre net return, the total expected
return from adoption is proportional to the
size of the com enterprise. Farm firms with
larger com enterprises have a greater absolute
incentive to adopt and utilize more efficient
tillage technologies than farm firms with
smaller corn enterprises. Thus, the probability
of a farm operator adopting reduced tillage is
hypothesized to be positively related to the
number of com acres planted on his farm,
other things equal. The marginal effect of
ACRES on the probability of adopting reduced
tillage, however, is expected to decline as
ACRES increases because the probability of
adoption cannot exceed one.
Per acre profitability is expected to depend
on the cropping system, soil characteristics,
and weather. The cropping system variable is
the ratio of soybean-to-corn acreage. 9 Because soybean root systems break up the soil
naturally and because the threshed soybean
plant leaves a small amount of plant material
on the soil surface, farm operators can make
an excellent seedbed for com by applying reduced tillage to soybean stubble. The economic advantage of reduced tillage is less
when other crops precede com in a cropping
system. Assuming static cropping systems for
farms, a large ratio of soybean-to-corn acreage
indicates a potential economic advantage of
reduced tillage over moldboard tillage for preparing a corn seedbed. Thus, the probability of
adopting reduced tillage is hypothesized to be
greater for managers of farms with larger ratios of soybeans to com acreage than for managers of farms with smaller ratios. This variable is expected to have a diminishing marginal effect on the probability of adoption.
Soil association dummy variables categorize
soil characteristics. The state of Iowa is
classified into twenty-one geographical soil associations (Cooperative Extension Service).
The classification is based upon parent material (loess, glacial till, and alluvium) and topology. These characteristics determine the soil's
natural fertility, drainage characteristics, and
409
410 November 1984
The Empirical Results
The results from fitting the two-stage model of
reduced-tillage adoption are presented and assessed. Linear and probit probability specifications of the adoption equation are fitted to
data for 869 sample farms. If the error term
is uniformly distributed with a zero mean, the
linear specification is appropriate and the ort 2 if asset transfers to succeeding generations are important,
then additional expe¡
need not imply a shorter planning ho¡
zon.
Table 1. Descriptive Statistics for Variables of
Adoption Model, Iowa Farm Operators, 1976
Variable Name
Adoption equation (N = 869)
D (1-0)
ACRES
SBRATIO
R A I N F A L L (inches)
S E A S O N (growing degree days)
T E N U R E (1-0)
Adoption efficiency (n = 797)
E
E D U C A T I O N (years)
E X P E R I E N C E (years)
EDDUM1 (1-0)
EDDUM2 (1-0)
C O N T I N U E D (1-0)
E X T E N S I O N (1-0)
MGSERVICE (1-0)
H E A L T H (1-0)
Mean
Standard
Deviation
.58
145.1
.51
31.1
2,951.5
.58
136.2
.52
2.40
149.0
.38
11.3
22.9
.20
2.2
13.0
.25
.06
.1 i
.32
.12
.21
dinary least-squares estimator is unbiased and
inexpensive to obtain, but it is statistically inefficient. Alternatively, if/,q is normally distributed, as it will be if el~ and e2~ of equation
(3) are normally distributed, then probit--a
maximum likelihood estimation procedure-yields a consistent, efficient, and asymptotically normal estimator. 13 Table 2 presents ordinary least squares and probit estimates of
the adoption equation to facilitate comparisons.
The signs of the estimated coefficients are
consistent with expectations, t4 At sample
mean values of variables, the estimated coefficients of ACRES and ACRESQ imply that
the marginal effect of a farm' s corn acreage on
the probability of adopting reduced-tillage
technology is positive and diminishes as the
farm's corn acreage increases. Similarly, the
marginal effect of the ratio of a farm' s soybean
acreage-to-corn acreage on the probability of
adoption is positive and diminishes as
SBRATIO increases. The coefficients of
RAINFALL, SEASON, and TENURE have
expected signs, but they are not significantly
different from zero at conventional levels. The
soil association dummy variables have coefficients that are generally consistent with expectations. The Clarion-Nicolett-Webster soil
~3 A third alternative is a logit probability model. Takeshi
Amemiya and Pindyck and Rubinfeld note that empirical results
using probit and logit will generally be very similar.
~4 Because the dependent variable is dichotomous, the R2 of. 18
for the linear probability specification seems respectible (Takeshi
Amemiya, p. 1504).
Downloaded from http://ajae.oxfordjournals.org/ at Penn State University (Paterno Lib) on September 20, 2016
field days or demonstrations sponsored by the
extension service, and 0 otherwise; Z6 is
MGSERVICE, dichotomous variable, I iffarm
operator (frequently or sometimes) utilized
media sources of information published or
marketed by private information and management firms, and 0 otherwise; Z7 is EXPERIENCE, number of years since farm
operator first began farming independently;
and Z8 is HEALTH, dichotomous variable, 1 if
farm operator had a health condition limiting
the amount and type of work, and 0 otherwise.
The variables EDUCATION, EDDUM1,
and EDDUM2, represent the quantity and
type of schooling received by the farm
operator. The variables CONTINUE, EXTENSION, and MGSERVICE are indicators
of information sources currently utilized in
making decisions on adopting new farm technology in general. The sources, however, may
be heterogenous in information-quality respects and are not limited to reduced tillage
information. Each of the preceding six variables is expected to have negative coefficients
or to increase the efficiency of adoption decisions. Farming experience can have two opposing effects on adoption efficiency. First,
additional experience, holding the planning
horizon length constant, is expected to increase efficiency. Additional experience,
however, might be associated with a shortened planning horizon over which returns can
be captured from investing in new technology. 12 Thus, the net effect of EXPERIENCE
on adoption efficiency is a priori uncertain.
Because poor health may shorten the planning
horizon and impede normal decision making,
poor health of the operator is expected to decrease the efficiency of adoption decisions.
Table 1 presents descriptive statistics for the
empirical measures of variables entering equations (5) and (6).
Amer. J. Agr. Econ.
Rahm and Hufjman
Adoption o f Redued Tillage
Table 2. Estimates of Linear and Probit Probability Models, Adoption of Reduced Tillage
Practices by Iowa Farm Operators, 1976
Table 3. Frequency Distribution for Probability of Reduced Tillage Adoption and for Adoption Efficiency
Number of Farms
Probability Model
Linear a
Independent
Variable
lntercept
Probit b
Estimate
t-ratio
Estimate
t-ratio
Scale [Pi, Ei]
.5130
.0018
-.87E-6
.3785
- .0834
-.0141
.73E-5
- .0057
.2991
.1920
.1844
- .0862
- .0555
.2071
.1566
.0564
.1116
.0352
.2720
.0979
.1399
.1911
-.1329
.2220
.2546
.1867
.3240
-8.61
-4.81
6.97
- 5.06
- 1.34
.03
- . 14
1.91
1.65
1.73
- .65
- .30
1.86
1.84
.43
1.72
.48
2.36
.52
1.62
2.05
-.56
2.44
3.06
2.47
1.66
.0469
.0055
-.26E-5
1.0785
- .2460
-.0410
.35E-4
.0053
.8818
.5780
.5305
- .2255
-.2094
.6351
.4727
.2150
.3173
.1112
.7846
.2841
.4113
.5374
-.3139
.6524
.7394
.5633
.9657
.05
8.14
-4.64
6.43
- 4.17
- 1.34
.05
.05
1.89
1.66
1.69
- .57
- .33
1.90
1.89
.54
1.67
.52
2.28
.52
1.61
1.99
-.47
2.36
2.95
2.48
1.62
0-0.10
0.11-0.20
0.21-0.30
0.31-0.40
0.41-0.50
0.51-0.60
0.61-0.70
0.71-0.80
0.81-0.90
0.91-1.00
a n = 869, R ~ = .18, F = 6.90.
b n = 869, Iog likelihood: -507.4, chi square: 166.8.
association is the reference soil type that is
included in the intercept of the equation. This
soil association is located in north central
Iowa and is composed of flat, heavy, and
poorly drained soils. Thus, the positive and
frequently significant coefficients for the other
soil association variables, e.g., SD19 which
represents the Shelby- Sharp sburg- Macksburg
soils, is consistent with our hypothesis that the
probability of adopting is larger on rolling,
lighter, and better drained soils.
Column one of table 3 displays the predicted
probabilities of adopting reduced-tillage corn
technology for the 869 sample farms. The predicted probabilities are calculated from the
probit estimate of the adoption equation. The
frequency distribution is a distorted bell shape
and has a median of 0.62. The frequency distribution of Ei, the measure of adoption
efficiency [column (2) of table 4 ], obtained by
using probit predictions also has a distorted
bell shape and a median of 0.38.
Predicted
Probability
o f Adoption
Adoption
Efficiency
4
33
73
78
111
144
168
141
90
36
40
107
161
170
151
93
76
48
21
2
Equation (6), which explains tillage adoption efficiency, is fitted by OLS to data for 797
sample farms. Seventy-two observations were
lost because of missing data. Three regression
equations are reported in table 4. Regression
equation (1) includes all the human capital
variables, and regression equations (2) and (3)
contain a reduced set. Because the information variables might be jointly determined
with adoption efficiency, the variables CONTable 4. Estimate of Equations Explaining the
Efficiency of Reduced Tillage Adoption Decisions by Iowa Farm Operators, 1976
Independent
Variables
EDUCATION
EDDUM1
EDD UM2
(1)
-.035
(-2.72)
(2)
-.037
(-3.11)
(3)
-.044
(-3.78)
- .018
(-.31)
.062
(.55)
CONTINUED
- .212
( - 2.73)
EXTENSION
-.050
( - .97)
MGSER VICE
- . 230
(-3.15)
EXPERIENCE
-.003
( - 1.35)
-.002
( - 1.08)
-.003
( - 1.29)
HEALTH
.087
(1.39)
.088
(1.40)
.090
(1.43)
lntercept
- .567
( - 3.60)
- .616
( - 3.89)
- .550
( - 3.49)
R2
F
.05
4.61
- .221
( - 2.94)
.03
6.311
.02
5.49
Note: The dependent variable is In E~ = In ID~ - P~I; t-ratios are in
parentheses; sample size is 797.
Downloaded from http://ajae.oxfordjournals.org/ at Penn State University (Paterno Lib) on September 20, 2016
ACRES
ACRESQ
SBRATIO
SBRATIOSQ
RAINFALL
SEASON
TENURE
SD1
SD2
SD3
SD4
SD6
SD7
SD8
SD9
SDIO
SD12
SD13
SD14
SDI5
SD 16
SD17
SD18
SD19
SD20
SD21
411
412
November 1984
Conclusions
This paper has presented a model of adoption
and econometric evidence about determinants
of reduced tillage adoption and of adoption
efficiency. The empirical results show that soil
characteristics, the cropping system, and the
scale of operation significantly affect the probability of adopting reduced tillage in Iowa corn
enterprises. The predicted probability of
adopting differs widely across sample farms.
When adoption is not always economically
feasible, the results show that human capital
variables enhance the efficiency of the adoption decision. Other variables undoubtedly affect the efficiency of adoption decisions, but
they can be explored in future research.
HEALTH.
[Received February 1983; final revision
Although the R 2 for the regression equations
received April 1984. ]
in table 4 may seem small, this is expected.
Recall that Ei is approximately the absolute
value of the residual remaining after removing effects of variables expected to have
homogenous effects on the probability of References
adoption. It undoubtedly also includes some
measurement error because broad soil associ- Amemiya, Minoru. "Conservation Tillage in the Western
ation type classifications and regional weather
Corn Belt." Conservation Tillage: Problems •
Potentials. Ankeny lA: Soil Conservation Society of
data were employed rather than farm-level inAmerica, 1977.
formation on these variables. The small R 2
means that a small share of the variance in this Amemiya, Takeshi. "Qualitative Response Models: A
Survey." J. Econ. Lit. 29(1981):1483-1536.
measure of adoption efficiency is explained by
Cooperative Extension Service. "Highway Guide of Iowa
variation in the included human capital variSoil Association," Pm 389. Ames lA: Iowa State
ables. This does not mean, however, that
University, 1977.
these variables have no significant effect. A Dixon, R. "Hybrid C o m Revisited." Econometrica 48
test of the joint null hypothesis of no effect of
(1980): 1451-61.
human capital variables on efficiency of the Erbach, Donald R., Walter G. Lovely, and George E.
Ayres. "Conservation and Conventional Systems for
adoption decision can be rejected. Employing
Continuous Corn." Iowa State University Agr. Exp.
regression equation (2), table 4, the sample
Sta. Misc. Bull. 14, 1980.
value of the F-statistic is 6.3 and the critical
value of the tabled F with 4 and 790 degrees of Fane, George. "Education and the Managerial Efficiency
of Farmers." Rey. Econ. and Statist. 57(1975):452freedom at the one percent significance level is
61.
3.32. Thus, these results provide evidence that Feder,
G., R. E. Just, and D. Zilberman. "Adoption of
farmers' schooling and some other human capAgricultural Innovation in Developing Countries."
ital variables enhance the efficiency of farmWashington DC: World Bank Staff Work. Pap. No.
ers' tillage choices.
549, 1982.
~s Near multicollinea¡ is a potential problem in regression
equation (1), table 4. Although correlations between variables do
not indicate whether near multicollinearity is a serious problem,
the absolute value of no correlation coefficient exceeded .40.
The correlation between EDUCATION and EXPERIENCE,
EDDUMI, EDDUM2, and CONTINUED are -.396, .277, .348,
and .212, respectively. The correlation between EDDUM2 and
CONTINUED is .165 and, finaUy, the correlation between
EXPERIENCE and HEALTH is .218.
Globerman, Steven. "Technological Diffusion in the
Canadian Tool and Die lndustry." Rev. Econ. and
Statist. 57(1975):428-34.
Griffith, Donald R., et al. "Conservation Tillage in the
Eastern Corn Belt." Conservation Tillage: Problems
and Potentials. Ankeny lA: Soil Conservation Society of America, 1977.
Griliches, Zvi. "Hybrid Com. An Exploration in the Economics of Technological Change." Econometrica
25(1957):501-22.
Heady, Earl O., and Cameron Short. "Interrelationships
among Export Markets, Resource Conservation, and
Downloaded from http://ajae.oxfordjournals.org/ at Penn State University (Paterno Lib) on September 20, 2016
TINUE, EXTENSION, and MGSERVICE are
excluded from some specifications. A striking
feature of regression equation (1) is that signs
of all coefficients agree with expectations except for the coefficient of EDDUM2. Several
coefficients, however, have small t-ratios,
which suggests that they are not significantly
different from zero. ~s In regression equation
(2), the coefficients of EDUCATION and
CONTINUE are negative and significantly different from zero at better than the one percent
significance level, respectively. Also, additional experience tends to increase adoption
efficiency, and poor health tends to reduce it.
For comparison, regression equation (3) is
fitted with only three basic human capital variables, EDUCATION, EXPERIENCE, and
Amer. J. Agr. Econ.
Rahm and HuJJman
413
Mansfield, E. "Technical Change and the Rate of Imitation." Econometrica 29( 1961): 741- 66.
Petzel, Todd E. "The Role of Education in the Dynamics
of Supply," Amer. J. Agr. Econ. 60(1978):445-51.
Pindyck, Robert S., and Daniel L. Rubinfeld. Econometric Models and Economic Forecasts, 2nd ed. New
York: McGraw-Hill Book Co., 1981.
Rogers, Everett, and F. Floyd Shoemaker. Communication of Innovations, 2nd ed. New York: Free Press,
1971.
Romeo, A. A. "Interindustry and InterŸ Differences in
the Rate of Diffusion of an Innovation." Rev. Econ.
and Statist. 57( 1975): 31 I- 19.
Schultz, Theodore W. "The Dynamics of Soil Erosion in
the United States: A Critical View." Unpublished
paper. Agr. Econ. Pap. No. 82:8, University of
Chicago.
Soth, Lauren. "The Grain Export Boom: Should It Be
Tamed?" Foreign Affairs 59(1981):895-912.
U.S. Department of Agriculture, Iowa Crop and Livestock Reporting Service. Iowa Farm, Fuel, and
Equipment. Des Moines lA: Government Printing
Office, 1976.
Downloaded from http://ajae.oxfordjournals.org/ at Penn State University (Paterno Lib) on September 20, 2016
Agricultural Productivity." Amer. J. Agr. Econ.
63(1981):840-47.
Hoiberg, Eric O., and Wallace E. Huffman. "Profile of
Iowa Farms and Farm Families: 1978." Iowa State
University Coop. Extens. Serv., April 1978.
Huffman, Wallace E. "Allocative Efficiency: The Role of
Human Capital." Quart. J. Econ. 91(1977):59-79.
Kramer, R. A., W. T. McSweeny, and R. W. Stavros.
"Soil Conservation with Uncertain Revenues and
Input Supplies." Amer. J. Agr. Econ. 65(1983):694702.
Jamison, Dean, and L. J. Lau. Farmer Education and
Farm Efficiency. Baltimore MD: Johns Hopkins University Press, 1982.
Jensen, Richard. "Adoption and Diffusion of an Innovation of Uncertain Profitability." J. Econ. Theory
27( 1982): 182-93.
Khaldi, Nabil. "Education and Allocative Efficiency in
U.S. Agriculture." Amer. J. Agr. Econ. 57(1975):
650-57.
Lee, Linda K., and William H. Stewart. "Land Ownership and the Adoption of Minimum Tillage." Amer.
J. Agr. Econ. 65(1983):256-64.
Adoption of Redued Tillage
Download