Allocative Efficiency-The Role of Human Capital

Allocative Efficiency: The Role of Human Capital
Author(s): Wallace E. Huffman

Source: The Quarterly Journal of Economics, Vol. 91, No. 1 (Feb., 1977), pp. 59-79
Published by: Oxford University Press
Stable URL: https://www.jstor.org/stable/1883138
Accessed: 27-04-2020 11:32 UTC
REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/1883138?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access
to The Quarterly Journal of Economics
This content downloaded from 133.19.169.5 on Mon, 27 Apr 2020 11:32:06 UTC
All use subject to https://about.jstor.org/terms
ALLOCATIVE EFFICIENCY: THE ROLE OF HUMAN
CAPITAL*
WALLACE E. HUFFMAN
I. Introduction, 59.-II. The model, 62.-III. Empirical analysis, 66.-IV. Furt

implications, 74.-V. Conclusions, 77.-Appendix: data and sources, 77.
I. INTRODUCTION
In a growing economy we observe that many people are reallo-

cating their resources in response to changes in economic conditions.
How efficiently they perceive changes and respond to changes in
economic conditions is attributed to allocative ability. The human-
capital approach to allocative efficiency postulates that allocative
ability is largely an acquired, opposed to an innate, skill.
The objective of this paper is to present some econometric evi-
dence that investments in education and extension (information)
improve the economic performance of U. S. Corn Belt farmers by
increasing the alacrity with which they respond to changes in eco-
nomic conditions.1 The data are averages per county for counties of
Illinois, Indiana, Iowa, Minnesota, and Ohio, taken largely from the
1959 and 1964 Censuses of Agriculture. The changes in economic
conditions are the sizable decline in the real market price of nitrogen
fertilizer and the increase in the nitrogen responsiveness of new hy-
brid-corn varieties during the 1959 to 1964 period. The most im-
portant finding is that investments in education and extension in-
crease the rate that farmers reduce disequilibrium in nitrogen fer-
tilizer usage. A major implication is that perceiving and responding
* The author is indebted to T. W. Schultz, M. Nerlove, and D. G. Johnson for

providing valuable direction to this research project at an early stage; to R. Evenson,
H. G. Lewis, G. Ghez, and M. Boehlje for providing comments that stimulated new and
enlightening exploration; and to Robert Dorfman for generous editorial assistance on
an earlier draft of this paper. Also, M. Boehlje, G. Rausser, and an anonymous reviewer
provided helpful suggestions on an earlier draft of this paper. Journal Paper No. J-8417
of the Iowa Agriculture and Home Economics Experiment Station, Ames, Iowa. Project
No. 2078.
1. The same general topic is treated elsewhere by the author, "Decision Making:
The Role of Education," American Journal of Agricultural Economics, LVI (Feb.
1974), 85-97. In the current paper a different and larger sample and more rigorous
statistical methods are used, resulting in a revision of some of the earlier conclu-
sions.
Also, see George Fane, "Education and the Managerial Efficiency of Farmers,"
Review of Economics and Statistics, LVII (Nov. 1975), 452-461; and Finis Welch,
"Education in Production," Journal of Political Economy, LXXVIII (Jan.-Feb. 1970),
35-59.
?) 1977 by the President and Fellows of Harvard College. Published by John Wiley & Sons, Inc.
60 QUARTERLY JOURNAL OF ECONOMICS
correctly to changes in economic incentives require allocative ability

that is acquired by investing in education and information.
In Section I the role of human capital in the allocative perfor-
mance of heads of firms and households and the rationale for ana-
lyzing the allocative performance of U. S. farmers are developed.
Section II presents the form of the empirical model. The empirical
analysis, presented in Section III, describes the methodology and data,
and reports the results from fitting the model to U. S. county aggregate
data. Section IV presents implications from the empirical results and
suggests related research. Section V presents the conclusions.
Human Capital and Allocative Efficiency

In a growing economy we observe that many people-both heads
of firms and households2-are reallocating their resources in response
to changes in economic conditions. Reallocation requires (a) per-
ceiving that change has occurred, (b) collecting, retrieving, and ana-
lyzing useful information, (c) drawing valid conclusions from the
available information, and (d) acting quickly and decisively.3 Allo-
cative ability is the name given to this human ability to perceive
changes in economic conditions and to respond efficiently.4
The human-capital approach to allocative efficiency postulates
that allocative ability is an acquired skill. Allocative skill is (human)
capital in the sense that it is acquired at a cost and yields a valuable
stream of services over future periods. It is acquired in schooling, by
searching for useful information, and in experience from reallocating
2. In the productive household, heads face decisions that are similar to the deci-
sions faced by heads of firms. (G. S. Becker, "A Theory of the Allocation of Time,"
Economic Journal, LXXV (Sept. 1965), 493-517; and M. Nerlove, "Household and
Economy: Toward a New Theory of Population and Economic Growth," Journal of
Political Economy, LXXXII (Mar.-Apr. 1974), S200-18.)
3. The seminal paper on the economic value of information is G. J. Stigler's articl
"Economics of Information," Journal of Political Economy, LXIX (June 1961), 213-25.
Two important activities for acquiring information are search-the process whereby
an individual ferrets out information on price and productive characteristics before
purchasing input services-and experimentation-the process whereby an individual
inspects an input and monitors its performance in production by using it after pur-
chasing. For households, or consumers, P. Nelson, "Information and Consumer Be-
havior," Journal of Political Economy, LXXVIII (Mar.-Apr. 1970), 311-29, makes
a fundamental distinction between search and experience qualities of market goods.
A. A. Alchian and H. Demsetz, "Production, Information Costs, and Economic Orga-
nization," American Economic Review, LXII (Dec. 1972), 777-95, argue that firms
are, among other things, a specialized market institution for collecting, collating, and
selling information. The heads of firms (employers), as they monitor many inputs,
acquire special superior information about their productive qualities.
4. In neoclassical economic theory the simplifying assumptions of perfect infor-
mation and rationality preclude allocative ability from being a valuable skill to firms
and households.
THE ROLE OF HUMAN CAPITAL 61
resources.5 The services are allocated jointly with the total available
time of an individual. Thus, there are many opportunities to capture
returns on the investment. Observed interpersonal differences in the
quantity of allocative skill are attributed to perceived differences in
the expected rate of return from investment in this form of capital
compared to alternative investments. If marginal returns to the al-
locative skill were to rise (fall), as they would if the rate of economic
growth were to increase (decrease), then individuals would be induced
to increase (decrease) their stock of allocative skill.
U. S. Agriculture: A Case for Analysis

Why U. S. agriculture? First, changes in relative prices and in
available superior techniques have performed some interesting eco-
nomic experiments to analyze. Second, the organizational structure
of decision making is simple, compared with most other industries,
and the individual responsible for allocative decisions (the farmer)
is easily identifiable. Third, the data on U. S. agriculture are readily
accessible.
Large changes in relative factor prices and proportions and in
techniques have occurred in U. S. agriculture.6 In this study the
changes in economic conditions, which are singled out for analysis,
are (1) the decline in the real price7 of nitrogen fertilizer, the single
most important fertilizer nutrient in corn production, by 22-25 per-
cent8 from 1959 to 1964, and (2) the increase in nitrogen responsive-
ness of new hybrid-corn varieties, which were the result of past bio-
logical research and development. Furthermore, from 1959-1964, the
5. Under conditions of imperfect information modern decision theory has at-

tempted to provide a systematic approach to decision making. These approaches use
mathematics and statistics to build intuitively appealing models of a sequential decision
process; for example, see G. B. Wetherill, Sequential Methods in Statistics (New York:
Wiley & Sons, Inc., 1966). Also, advances in analytical techniques like linear pro-
gramming assist decision making by increasing the number of different activities that
heads of firms can successfully consider. But these approaches abstract from the role
of human ability in perceiving and responding efficiently to changing economic con-
ditions.
6. See Y. Hayami and V. Ruttan, "Factor Prices and Technical Change in Agri-
cultural Development: The United States and Japan, 1880-1960," Journal of Political
Economy, LXXVIII (Sept.-Oct. 1970), 115-41; Z. Griliches, "Agriculture: Productivity
and Technology," in International Encyclopedia of the Social Sciences (New York:
MacMillan, 1968), Vol. 1, 241-45; Z. Griliches, "Hybrid Corn: An Exploration in the
Economics of Technological Change," Econometrica, XXV (Oct. 1957), 501-22; and
Z. Griliches, "The Demand for Fertilizer: An Economic Interpretation of Technical
Change," Journal of Farm Economics, XL (Aug. 1958), 591-606.
7. The real price of nitrogen is the price per unit of the nutrient nitrogen in fer-
tilizer divided by the price of the final product corn.
8. The decline of the price of nitrogen fertilizer to farmers occurred because
farmers shifted to fertilizers with a higher concentration of nitrogen and because of
technical change in the fertilizer industry. For a discussion of the latter, see Gian Sahota,
Fertilizer in Economic Development (New York: Praeger, 1968).
quantity of nitrogen fertilizer applied annually to corn by farmers in

the five Corn Belt states of this study doubled. Thus, changes in
economic conditions and the associated response by Corn Belt farmers
have performed an interesting experiment to analyze.
II. THE MODEL
A model consisting of three equations summarizes the change

in annual nitrogen fertilizer usage as farmers perceived and responded
to changes in economic incentives. The model is
(1) t= D(P*, T),

(2) Nt-Nt-5 =O (N -Nt-A
(3) d3=B(Z), O < d < 1.

The demand function (1) for (optimal quantity of) nitrogen fertilizer
per acre per year N*, is a function of a vector of "permanent real factor
prices," Pt, the permanent component of factor prices divided by the
permanent component of the final product price, and a vector of en-
vironmental variables T, including education for the "worker effect."9
This specification of the demand function considers output and the
other factors of production to be determined simultaneously with the
use of nitrogen fertilizer; whereas, the permanent prices are considered
to be exogenous to farmers' decisions.
The adjustment (or reaction) equation (2) specifies that the ac-
tual change in nitrogen fertilizer applied between years t and t - 5
is proportional to the gap between the optimal quantity of nitrogen
in year t and the actual quantity in year t - 5. The adjustment coef-
ficient f summarizes the rate that farmers adjust the actual quantity
of nitrogen fertilizer to the optimal quantity. It is restricted to values
between zero and one. If : is unity, then adjustment is complete within
5 years-the actual and the optimal quantity are equal. If d is zero,
no adjustment occurs-the actual quantities of nitrogen fertilizer for
successive years are unchanged and are equal (Nt = Nt_5).
The novelty of the model is that the adjustment coefficient d
9. The assumption is that when prices fluctuate widely, the permanent changes
in prices signal changes to which farmers respond. M. Friedman, A Theory of the
Consumption Function (Princeton: Princeton University Press for National Bureau
of Economic Research, 1957), assumes that consumers behave as if consumption and
income have two parts, a permanent and a transitory part, and that the systematic
relationship is between permanent consumption and permanent income. M. Nerlove,
Distributed Lags and Demand Analysis for Agricultural and Other Commodities
(Washington, D. C.: Government Printing Office, 1958), presents distributed lag models
where observed prices are assumed to be decomposed into permanent and transitory
parts.
(equation (3)) is specified as a function of a

ables Z.10 The variable coefficient specifica
hypothesis that economic variables explain
adjustment, a measure of allocative perfor
components of changes in economic incenti
Specifically, a logistic type function specif
(4) f3= [1 + exp (bo + blE + b2EXTIH + b3E
X EXTIH + b4ACRES + b5DE)]-
with
(5) a9/0E = -(b, + b3EXTIH)f(l -O

(6) a 0/aEXTIH = -(b2 + b3E)f(l -),
(7) a8/laACRES = -b4f(1 - /),
(8) aO/aDE = -b5l( - ),

where
E = education level of farmers,

EXTIH = amount of contact between agricultural extension
agents and farmers,
ACRES = acres of corn per farm,
DE = the beginning period disequilibrium in nitrogen usage,
Nt5 -Nt_5,
and where equations (5) through (8) denote the respective pa

derivatives of the adjustment coefficient.
Information obtained from agencies specializing in processi
and selling information can be expected to affect allocative per-
formance. The agricultural extension service supplies decoded results
to farmers from basic and applied research conducted at state agri-
cultural experiment stations and at the USDA. These decoded results
are acquired by farmers from bulletins (mostly at zero direct cost) and
through personal contact, conferences, and field demonstrations. Also,
the agricultural extension agents serve as consultants to farmers on
technical production problems. Thus, the amount of contact between
10. E. Mansfield, "Technical Changes and the Rate of Imitation," Econometrica,
XXIX (Oct. 1961), 741-66, uses a variable rate of imitation model to explain the rate
at which firms in four industries came to use twelve important innovations. C. Phelps,
"Real and Monetary Determinants of State and Local Highway Investment, 1951-
1966," American Economic Review, LIX (Sept. 1969), 507-21, uses a stock adjustment
model with a variable adjustment coefficient for explaining state and local government
capital outlays 6n highways. Z. Griliches, "Hybrid Corn: An Exploration in the Eco-
nomics of Technical Change," explains variations in the rate of adoption of hybrid corn
by farmers.
agricultural extension agents and farmers is a plausible measure of

the useful information obtained from the agricultural extension ser-
vice about relevant production techniques and changes in prices.
Based upon the human-capital approach to allocative efficiency,
the hypothesis is that the size of the adjustment coefficient (I) is
positively related to the level of education of farmers and to the size
of the extension variable. No hypothesis is stated for the direction of
the interaction effect between education and extension because ex-
tension information may be either a substitute for or a complement
to farmers' education in determining allocative performance. The size
of the corn-growing activity, as measured by the acres of corn per farm,
serves as a measure of the economic incentive for farmers to be in-
formed about optimal nitrogen usage in corn production. Further-
more, economies of scale in information usage suggest the hypothesis
that : is positively related to the size of the corn-growing activity.11
The size of the beginning period disequilibrium gap in nitrogen fer-
tilizer usage is to be held constant while analyzing the effects of the
other variables on the size of the adjustment coefficient.
The demand function for nitrogen fertilizer is derived from a
generalized production function:
(9) Y = A II Xfi(X)eg(x) Y> 0X > X OA > 0,0 ' fi(X) < 1,

i~ 1
where fi (X) and g(X) are functions of a vector X of factors of pro-

duction. If g(X) = 0, then equation (9) reduces to a Cobb-Douglas
production function with variable elasticities of production.12 The
production function is
n
(10) Y = ANa1Ea2Wa3ei bh fi Xi a-; Y, N, E, W, Xi ' 0;

i=4
0 < a- < 1, A > 0;
where
Y = corn yield,
N = nitrogen fertilizer,
E = education of farmers,
W = an index of weather and natural endowment,
11. Stigler, op. cit.
12. See E. Ulveling and L. Fletcher, "A Cobb-Douglas Production Function with
Variable Returns to Scale," American Journal of Agricultural Economics, LII (May
1970), 322-26.
D1 = geographical dummy variables,

Xi= other factors of production.
The variable elasticity of output coefficient of N is specified as
(11) al = a + a2 (GDD X R2 X Yo);
and the index of weather and natural endowment is specified as
(12) W a3 = Rle1R2 2GDDY 3Y Y4,
where
GDD = length of corn-growing season,

R 1 = accumulated annual rainfall for the nonsummer months,
R2= accumulated rainfall for the summer months (June, July,
and August),
Yo = base year corn yield, a proxy for natural endowment of land
for corn production.
If
n
A' = A fi X, ai,
i=4
then equation (10) reduces to a Cobb-Douglas production function

with a variable elasticity of output for one input N. But does this
specification, which excludes labor and machinery services, ade-
quately represent the relationship? Potential corn yields are deter-
mined largely by a biological process depending on the quality of seed,
fertilizer, length of growing season, rainfall, and natural endowment
of the soils. Also choice of methods and timing are important, but their
determination along with choice of seed is part of the worker effect
of education. Inputs of labor and machinery services occur largely
during the preparation for planting and harvesting phases of pro-
duction. Thus, the specification captures factors primarily responsible
for corn-yield variation.13
By equating the marginal product of nitrogen to the price of ni-
trogen (Pp) deflated by the permanent price of corn (Pt) and solving
for N subject to a nonnegativity constraint, the derived demand
function for (the optimal quantity of) nitrogen fertilizer is
(13) N* = [alA1Ea12Wa3P~gP7 le jDj]ID]l/l-al

provided that N* > 0, and zero otherwise.
13. Data on the quantity of labor services and machinery services allocated to corn
production are not available. Also, the geographical dummy variables will capture,
among other things, regional differences in the levels of the excluded inputs.
III. EMPIRICAL ANALYSIS
In this section the methodology of the empirical test and the da

are described; the production function is fitted to U. S. Corn Belt
cross-sectional data pooled over 1959 and 1964; implied optimal ni-
trogen usage is estimated for each country from the demand for ni-
trogen function; and the adjustment function is fitted between 1959
and 1964 for cross-sectional data.
Methodology and the Data
The empirical specification of the model and results of this paper

are derived from a methodology of model selection and testing that
is different from and, hopefully, superior to that used in most em-
pirical analyses. In testing the hypothesis that human capital of
farmers enhances their allocative performance in adjusting to changes
in economic incentives, theoretical considerations and implications
from previously reported empirical research failed to solve the model
choice problem-the choice between alternative functional forms and
empirical definition of variables and of exact number of variables. A
commonly followed procedure by empirical researchers is to experi-
ment widely with model specification on a single data set14 and then
report the "best" results as the estimate of the model for drawing
inferences about behavioral relations. However, when a single data
set is used both for performing statistical tests to select a particular
specification of the model and for estimating parameters of the se-
lected specification (final form), the estimator of the final form does
not have standard statistical properties.15 Thus, when both steps have
been performed on a single data set, the empirical results should be
viewed with skepticism.
An alternative procedure is to divide the available data into two
nonoverlapping parts and to use one data set for model selection and
a second data set for obtaining estimates for reporting and drawing
14. G. E. P. Box and D. R. Cox, "An Analysis of Transformations," Journal of

Royal Statistical Society, Series B, XXVI (1964), 211-43, suggest procedures for de-
termining functional form. The Bayesians provide a procedure that can be applied to
the model selection problem-comparing the posterior odds pertaining to alternative
models. See A. Zellner and M. S. Geisel, "Analysis of Distributed Lag Models with
Application to Consumption Function Estimation," Econometrica, XXXVIII (Nov.
1970), 865-88.
15. For experimentation using nested hypotheses, the estimator of the final form
will in general be biased, and the effective significant level associated with tests of
hypotheses on the coefficients will exceed the nominal significance level (a) by an
unknown magnitude. See H. Theil, Principles of Econometrics (New York: Wiley &
Sons, 1971), 603-06; and W. T. Kennedy and T. A. Bancroft, "Model Building for
Prediction in Regression Based Upon Repeated Significance Tests," Annals of
Mathematical Statistics, XLII (1971), 1273-84. In experiments using nonnested hy-
potheses the properties of the estimator are not well established.
inferences about behavioral.relationships.16 When independent data

sets are used for the two steps, preliminary testing to select a par-
ticular model specification does not alter the standard statistical
properties of the estimator of the final form.17 In this paper the par-
ticular specification of the empirical model was determined by ex-
perimenting extensively with one data set, and the estimate of the
model is from a second nonoverlapping data set.
In testing the hypothesis that human capital of farmers enhances
their allocative performance in adjusting to changes in economic
conditions affecting optimal nitrogen fertilizer usage, a time series
of disaggregate cross-sectional data is required on output, input of
nitrogen, and human-capital characteristics of farmers. Although
individual firm data would be conceptually ideal for testing the model,
county aggregate data are pragmatically better and more readily
available. County averages have less variation than individual firm
data in the physical environment of production, allocative efficiency
of farm firms, and allocative skills of farmers. However, specification
errors from fitting an economic model, which at best captures the
effects of only the most important variables on individual firm be-
havior, to individual firm data may be larger than the aggregation bias
resulting from fitting the model to county aggregate data.'8 Fur-
thermore, county aggregates are small units, and county averages have
a relatively large amount of interunit variation in the physical envi-
ronment of production, allocative efficiency of farm firms, and allo-
cative skills of farmers, especially when compared with state or re-
gional averages.
The study area consists of 306 Corn Belt counties in Illinois,
Indiana, Iowa, Minnesota, and Ohio. The principal criterion for in-
cluding a county in the study area was that a large percentage of the
farmers engages in corn production. Hence, the study area covers the
major corn-growing counties of the U. S. Corn Belt. The 306 counties
were split into two parts. A random sample of 122 (40 percent)
counties was used for the experimentation that led to the particular
16. Theil (op. cit.) actually suggests a strategy of splitting the total available data
into three parts: one for choice of specification, a second for estimation, and a third
for making conditional predictions based upon the estimated equation from part two.
If the total number of observations is small relative to the number of unknown pa-
rameters of the model, then these procedures cannot be followed.
17. See Theil, op. cit.
18. See Y. Grunfeld and Z. Griliches, "Is Aggregation Necessarily Bad?" Review
of Economics and Statistics, XLII (Feb. 1960), 1-13; and D. J. Aigner and S. M.
Goldfeld, "Estimation and Prediction from Aggregate Data When Aggregates Are
Measured More Accurately Than Their Components," Econometrica, XLII (Jan. 1974),
113-34.
specification of the empirical model presented in this paper.19 The

nonoverlapping complementary random sample of the other 184
counties provides the independent data set for analysis in this
paper.20
Most of the empirical measures of the variables were derived from
data taken primarily fron the Census of Agriculture and USDA
publications. When the study was initiated, published county data
did not exist on the quantity of nitrogen fertilizer used in corn pro-
duction during 1959 and 1964; consequently, one of the first tasks was
to derive these data.21 The two human capital variables-education
and extension-were also derived. The 1964 Census of Agriculture
provided (for the first, and only time?) county data on the number
of farm operators in seven schooling completion classes: 0-4, 5-7, 8,
9-11, 12, 13-15, and 16 or more. A weighted average level of education
of farm operators was then constructed by assuming that schooling
quality was homogeneous across the study area and applying Welch's
income weight: 0.25,0.65, 1.00, 1.63,2.26,3.05, and 4.27 to the above
respective schooling completion classes.22 County data, collected
annually (but unpublished) by the Cooperative Extension Service,
on (1) the annual number of days allocated to crops by agricultural
extension agents and (2) the number of farmers and other individuals
assisted with any phase of grain crop production were used to derive
an index of the average amount of contact between farm operators
and agricultural extension agents.
Estimation of the Production Function and Optimal N

Production functions for 1959 and 1964 are estimated by fitting
equation (10) to pooled cross-sectional data for county aggregate
averages for 1959 and 1964.23 All coefficients of the production
function are unrestricted for the two years; except for a2, the coeffi-
19. Some results from the experimentation on this data set are reported in the
author's Ph.D. thesis ("The Contribution of Education and Extension to Differential
Rates of Change," Ph.D. thesis, University of Chicago, 1972) and the author's earlier
paper (op. cit.).
20. Using a random process for choosing observations yields complications; t
coefficients of the model are random. (See Theil, op. cit., 570.) However, after selecti
the observations for each sample, one can proceed by taking the samples as given
21. See Huffman, "The Contribution of Education and Extension to Different
Rates of Change," 113-26, for further details.
22. F. Welch, "The Determinants of the Return to Schooling in Rural Farm Areas
1959," Ph.D. thesis, University of Chicago, 1966.
23. The fitted function is
In Y = In A' + a iln N + a2(GDD X R2 X Y0) In N + a2 In E

+ -yj In R1 + -Y2 In R2 + -pY In GDD + _Y4 In Yo + lbjDj + ,
where e is a random disturbance.
cient of the interaction variable between nitrogen and the environ-

mental index; 'y, the coefficient of annual rainfall for the nonsummer
months; and 72, the coefficient of rainfall for the summer months.24
The geographical dummy variables, Dj's represent relatively homo-
geneous groups of counties, state-parts of agricultural subregions. The
coefficients of the regional dummy variables, which are to estimate
the regional logarithmic intercept terms, are free to take different
values for the two years.
The results from fitting the production function are reported in
Table I. The function was fitted to 368 pooled observations for 1959
and 1964 by using the classical least-squares estimation method. All
the coefficients have plausible signs; the t -ratios are generally large,
and the R2 is large, 0.81.25 The small t-values of the coefficients of the
education variables, however, lead one to conclude that this sample
contains no evidence that farmers' education increases corn yields
apart from its influence on the choice of the level of N26 (see
below).
A comparison of actual with implied optimal nitrogen usage for
1959 and 1964 confirms the existence of disequilibrium in the usage
of nitrogen. The implied optimal rate of nitrogen usage for each
county is predicted from equation (13). Estimates of the unknown
parameters for (13) are taken from Table I. Permanent prices for corn
and nitrogen are obtained by applying (arbitrary) weights of 0.69,0.23,
and 0.08 to the reported prices for the current and two preceding
years, respectively.27 (See Table II.) The state average optimal ni-
trogen fertilization rate and average deviation of actual from implied
optimal nitrogen usage are presented in Table III. Between 1959 and
1964, the average optimal nitrogen application increased by more than
24. In pretesting the model against data set one (the 40-percent sample), this
specification of the model for the pooled data performed the "best." The goodness-
of-performance criteria included variables with statistically significant coefficients,
reasonable predictions of N*, and large R2.
25. Estimation of the equation by applying classical least squares can yield biased
and inconsistent estimates of the coefficients because of simultaneous equation bias.
However, if farmers make input decisions based upon anticipated output as suggested
by I. Hoch, "Simultaneous Equation Bias in the Context of the Cobb-Douglas Pro-
duction Function," Econometrica, XXVI (Oct. 1958), 566-78, then simultaneous
equation bias is eliminated. Furthermore, applying the classical least-squares-esti-,
mation method to cross-sectional data pooled over two years into the multivariate re-
gression model may violate the assumptions of homoskedasticity and uncorrelated
disturbances. The size of the variance of the disturbances may differ contempora-
neously by state and(or) serially, and the disturbances may be serially and(or) con-
temporaneously correlated. When heteroskedasticity and correlated disturbances are
present, the classical least-squares estimator is unbiased but is not minimum vari-
ance.
26. This is the principal divergence from the conclusions of the author's earlier
paper, "Decision Making: The Role of Education."
27. The actual values of other relevant variables are used in calculating N* exce
for R1 and R2, which are the thirty-year average values.
TABLE I
ESTIMATES OF THE AGGREGATE CORN PRODUCTION FUNCTION:

ILLINOIS, INDIANA, IOWA, MINNESOTA, AND OHIO COUNTIES,
1959 AND 1964a
1959 1964
Independent variables Coefficients t-ratios* Coefficients t-ratios*
Nb 0.2907 6.988 0.3723 7.266

GDD x R2 x Y. x InN -0.00001269 -5.389 -0.00001269 -5.389
Eb 0.1468 1.560 0.0301 0.319
Rb -0.1125 -1.389 -0.1125 -1.389
R2 0 0.7870 6.180 0.7870 6.180
GDDb 0.5730 3.170 1.1001 5.005
y b 0.9590 7.009 1.4740 8.512
DiC -3.298 -2.978 -3.2698 -2.978

D1 64 -3.8764 -3.527
D2 -0.2368 -3.044 -4.2383 -3.953
Do S-0.3263 -5.735 -3.9295 -3.616
D4 -0.2296 -3.410 -3.9862 -3.649
D5 -0.1591 -2.179 -4.1433 -3.831
D6 --0.0842 -1.700 -4.0608 -3.690
D7 0.0021 0.039 -4.0595 -3.648
Ds -0.0831 -1.633 -4.2088 -3.830
D9 -0.1686 -2.984 -4.1522 -3.749
D1o 0.0278 0.385 -4.1086 -3.695
D11 0.0962 1.612 -4.0716 -3.645
DI2 -0.0180 -0.311 -4.0845 -3.661
DI 3 -0.1670 -2.338 -4.2512 -3.784
D, 4 -0.1674 -1.884 -4.5922 -4.097
D,5 -0.1490 -1.994 -4.2778 -3.825
D, 6 -0.1319 -1.670 -4.1743 -3.772
DI 7 -0.1801 -3.114 -4.2492 -3.843
Di 19 -0.0989 -1.683 -4.2297 -3.786
Di 9 -0.1377 -2.384 -4.3012 -3.892
D20 -0.1233 -1.771 -4.34e3 -3.904
D2, 0.0590 0.821 -4.0905 -3.691
R2 = 0.810. Degrees of freedom = 315.
*Ratio of coefficient to standard error.

a. Dependent variable in in Y. All coefficients are estimated from a single regres-
sion equation. See Appendix for data and sources.
b. Variable enters the regression in the loge form.
c. State agricultural subregion (SASR) 71 in Iowa for 1959 serves as the base of
reference for the intercept of the regression. The intercept for other regions, say re-
gion i, is estimated as the sum 61 + 6 i. The coefficient of D 64 estimate
the intercept for SASR 71 in Iowa between 1959 and 1964.
2.5 times; and, although farmers made sizable increases in the actual
rate of nitrogen consumption, there was a dramatic increase in the
size of disequilibrium, as measured by the average deviation of actual
from implied optimal N. These results suggest that long-term eco-
TABLE II
PERMANENT PRICES FOR CORN AND NITROGEN FERTILIZER,

SAMPLE STATES, 1959 AND 1964a
Real price
Permanent Permanent of nitrogen
price of corn price of nitrogenb (bu/lb)
($/bu) (U / lb) 1959 1964

Sample 1959 1964 1959 1964 (5)= (6)=
states (1) (2) (3) (4) (3)/(1) (4)/(2)
Illinois 1.08 1.14 8.9 7.6 0.082 0.067

Indiana 1.07 1.10 8.1 7.5 0.077 0.067
Iowa 1.02 1.16 8.9 7.8 0.087 0.068
Minnesota 0.95 1.08 9.5 9.0 0.101 0.083
Ohio 1.07 1.11 10.3 8.6 0.097 0.077
a. Permanent prices are a weighted average of actual prices over three years: Pt*
0.69 Pt + 0.23 Pt-, + 0.08 Pt-2.
b. From anhydrous ammonia.
Sources: U. S. Department of Agriculture, Agricultural Statistics (Washington:
Government Printing Office), 1956-1966 inclusive; U. S. Department of Agriculture,
Agricultural Prices (Washington: Government Printing Office), April issues and Annu-
al Summary, 1956-1966 inclusive.
nomic conditions were changing the optimal rate of N

farmers were adjusting their actual rate of N consumption.2
the exact magnitudes of the optimal rate and size of dise
should not be taken too seriously because their measurement is un-
doubtedly subject to a wide margin of error. They should be viewed
as being suggestive of the order of magnitude of the respective vari-
ables.
Estimation of Human Capital's Contribution in the Adjustment

Function
The coefficients of the adjustment function are estimated by

fitting a restatement of equation (2) to cross-sectional county average
28. Data taken from Agricultural Statistics (Washington, D. C.: Government

Printing Office, 1966 and 1971) show that farmers in these five states did not remain
idle in the face of such large disequilibrium gaps in nitrogen usage in 1964. Between
1964 and 1969, the average total consumption of N for all purposes per acre of corn in
the states of Illinois, Indiana, and Ohio increased by 51.7 percent (forty-four pounds
per acre) or an annual average of 10.3 percent, and in the states of Iowa and Minnesota,
which had the largest disequilibrium gaps in 1964, by 133.1 percent (seventy pounds
per acre) or an annual average of 26.4 percent. (Note that in 1964, 90 percent of total
N consumption was used as fertilizer on corn, and the percentage has risen.) Thus, these
data suggest that farmers in the five states were attempting to close the disequilibrium
gap that existed in 1964. Furthermore, both the percentage and absolute changes be-
tween 1964 and 1969 in the rate of N usage by Iowa and Minnesota farmers relative
to the change by Illinois, Indiana, and Ohio farmers are consistent with the finding that
Iowa and Minnesota farmers were much farther from optimal N usage in 1964.
TABLE III
IMPLIED OTIMAL N RATES AND DEVIATION OF ACTUAL FROM

OPTIMAL N RATES: STATE AVERAGES OF COUNTY DATA FOR SAMPLE
STATES, 1959-1964.
Average deviation
of actual from
Number Average optimal optimal N rates
of N rates (lb/ac/yr) (lb/ac/yr)

sample
State counties 1959 1964 1959 1964
Illinois 48 42.1 167.5 21.6 106.2

Indiana 34 48.6 162.6 18.2 104.6
Iowa 57 55.3 203.9 42.3 165.4
Minnesota 28 88.4 245.3 73.1 221.8
Ohio 17 59.6 159.5 29.0 93.8
data between two points in time, 1959 and 1964. By substituting

equation (4) into equation (2), rearranging equation (2), and taking
the natural logarithm, the restatement of equation (2) that is fitted
to the data is
(14) In [(N - Nt-5)/(Nt - Nt-5) - 1] = bo + b1E

+ b2EXTIH + b3E X EXTIH + b4ACRES + b5DE + w,
where t = 1964 (and t - 5 = 1959).29 After substituting the implied
values of N2 (same basic procedure as for Table III) for the unknown
optimal values, equation (14), which is linear in the unknown pa-
rameters of the adjustment function and a random disturbance, can
be estimated by the classical least-squares estimation method.
The results from fitting equation (14) to the data are reported
in Table IV. All coefficients have plausible signs, and the t-ratios for
all the coefficients of the variables, except ACRES, are large.30
Moreover, education and extension, the two human capital variables,
are individually and jointly significant explanatory variables. How-
ever, the important characteristic of the regression is the conformity
of the signs of the estimates of the partial derivatives of the adjust-
29. The left-hand side of equation (14) is In (1/# - 1). Arithmetic restrictions are
imposed on the values of the differences N2 - Nt5 and Nt -Nt-5 that appear in the
transformed dependent variable. The differences must be nonzero and of the same
sign; otherwise the natural logarithm of the transformed dependent variable is unde-
fined.
30. The R2 of this regression should be taken with a grain of salt because it mea-
sures the fraction of the variation of the transformed dependent variable In [Nt -
Nt-5)1(Nt - Nt5) -1] that is "explained."
TABLE IV
ESTIMATION OF THE ADJUSTMENT FUNCTION FOR NITROGEN

FERTILIZER USAGE BETWEEN 1959 AND 1964: ILLINOIS, INDIANA,
IOWA, MINNESOTA, AND OHIO COUNTIESa
Explanatory Sample Regression equation

variable mean coefficients t-ratios
E 1.769 -2.761 -3.792

EXTIH 1.077 -3.390 -3.575
E x EXTIH 1.918 1.764 3.319
ACRES 70.16 -0.001 -0.411
DE 35.90 0.022 12.094
INTERCEPT 6.094 4.640
R2 = 0.525 Degrees of freedom = 178.
a. Dependent variable is in [(Nt* - Nt5s)I(Nt - Nt-5)-1] = ln (1/1 - 1). See Ap-

pendix for data and sources.
TABLE V
EXPECTED SIGNS AND IMPLIED VALUES OF THE PARTIAL DERIVATIVES

OF THE ADJUSTMENT COEFFICIENT FOR A TYPICAL COUNTY
Expected
Partial derivatives sign Implied values
a3/aE + -[-2.761 + 1.764(1.077)] ((1 _O)a = 0.115

a3laEXTIH + -[,-3.390 + 1.764(1.769)] 3(1 - () = 0.036
a(3a01EXTIH)kaE ? -1.764 p3(1 - 1) = -0.236
a(3/aACRES + -(-0.001) 0(31 - (3) = 0.0001
ao/aDE ? -0.022 f(1- f) = -0.003
a. (= 0.160.
ment coefficient (equations (5) to (8)) with expectations. The expected

signs and implied values of the partial derivatives, evaluated at the
sample's arithmetic mean values for education and extension, are
presented in Table V.
The results support the hypothesis of the human-capital ap-
proach to allocative efficiency. The size of the adjustment coefficient
and of allocative efficiency of farmers is positively related to their level
of education and to information from extension. Furthermore, the
positive (and statistically significant) coefficient of the education-
extension interaction variable indicates that education and extension
are substitutes for enhancing the allocative performance of farmers.
Allocative performance, as measured by fi, is positively (but not sig-
nificantly) related to the size of the corn-growing activity. The in-
terpretation of the positive and statistically significant regression

coefficient of DE, the size of the beginning period disequilibrium gap,
is that factors that were inhibiting adjustment to equilibrium prior
to 1959 were inhibiting adjustment between 1959 and 1964.31 Thus,
large changes in actual N usage between 1959 and 1964 cannot in
general be attributed to a differential "catching-up effect."32
IV. FURTHER IMPLICATIONS
In this section estimates of marginal product of education and

extension from enhanced allocative efficiency are presented, and
implications for related research are suggested.
The Return to Allocative Ability
This study has shown that increasing education or extension

improves allocative efficiency and thereby raises farm profit. The
increase in profit is a measure of the marginal product of education
(extension) from this allocative gain. These marginal products are
evaluated at sample mean values of the variables and are calculated
as
(15) (VMPNt - PNt)8Nt/lV,

where 8Nt/8V = (N - Nt5)8i/8V, V = E, EXTIH.33 The term
aNt/aV shows the effect of a change of a human-capital variable on
the choice of Nt, given N* and Nt-5.
The marginal products of education and extension from en-
hanced allocative efficiency are sizable. First, increasing the level of
farmers' education by one unit (approximately one year) in 1959
implies an increase of 18.2 pounds per acre per year in N usage in 1964,
31. The results are robust to excluding DE from the regression. The sign of ACRES
switches, but the t-ratio remains small; the signs of the coefficients of the other variables
remain unchanged, and the t-ratios of these coefficients increase in size.
32. The regression equation reported in Table IV, evaluated at the sample mean
values of the explanatory variables, implies a f of 0.160. Given the five-year time interval
for adjustment to occur, this seems remarkably low. We must remember, however, that,
although the real price of nitrogen fertilizer did decline by 20-25 percent between 1959
and 1964 and the nitrogen responsiveness of hybrid corn varieties did increase over
this period, these changes did not all occur at the beginning of the 1960 crop year. They
were spread over the five-year period. Thus, farmers had less than a five-year time
interval to adjust to the full magnitude of these changes.
33. Predictions of the N* are made from equation (13). The parameters of the
demand function and adjustment function are taken from Tables I and IV, respec-
tively.
or a 10.7 percent reduction in the poundage gap.34 The implied size

of the disequilibrium gap in 1964, measured as the difference between
VMPN and PN, is $0.147 per pound of nitrogen per acre.35 Hence, the
implied marginal product in 1964 of an additional unit of education
in 1959 is 0.147 X 18.2 = $2.68 per acre of corn. However, given that
an average of 60.7 acres of corn was harvested for grain per farm in
1964, the marginal product of one year of education per farm operator
is $163 per year.
Although a marginal product of education of $163 per year is not
large,36 one must remember that many other allocative decisions are
associated with the complex farming activities of modern corn pro-
duction, and some of these decisions are promising sources for addi-
tional allocative gains from education. Furthermore, farming in the
U. S. Corn Belt generally consists of several different complex crop
and livestock activities that give rise to a multitude of activity com-
binations. Given that U. S. farmers operate in an economic environ-
ment calling for skill at separating long-term relative price changes
from transitory changes, distinguishing superior new techniques from
other techniques, and understanding the interrelationship between
changes in relative prices and techniques, these activities have the
potential for yielding additional allocative gains from education of
farm operators.
Second, increasing extension by approximately one minute per farm
(one-sixth of a unit) in 1959 implies an increase of one pound per acre
in N usage in 1964. The implied size of the difference between the
value of the marginal product of N and its price is $0.177 per pound
34. In calculating the increase in the rate of N usage, the extra education is as-
sumed to leave unchanged the optimal rate of N. This assumption is consistent with
the small t -ratios of the estimated coefficients of education in the production function.
This assumption probably leads to an underestimate of the implied change in the rate
of actual N usage between 1959 and 1964. The implied difference between optimal N
in 1964 and actual N in 1959 is 170.65 pounds per acre. Hence, the size of the increase
in the rate of N usage due to the extra education is not particularly large, given the
deviation of actual from implied optimal N of 152 pounds per year.
35. The disequilibrium gap is calculated as (bo + 61)/2, where 6o = VMPN - PN
at the actual rate of N usage in 1964 and 31 = VMPN - PN at the higher rate of N usage
after the implied increase in N consumption due to an extra year of education.
The order of magnitude of an alternative measure of the implied disequilibrium
gap, VMP/factor price, of 2.87 (at the margin VMPN = $0.227 and PN = $0.079 in this
study is consistent with the findings of Z. Griliches, "Research Expenditures, Education,
and the Aggregate Agricultural Production Function," American Economic Review,
LIV (Dec. 1964), 961-74; and J. Headley, "Estimating the Productivity of Agricultural
Pesticides," American Journal of Agricultural Economics, L (Feb. 1968), 13-23.
Griliches' results imply a disequilibrium gap for fertilizer (weighted primarily plant
nutrients) of 5 in 1949 and 2.7 in 1959. Headley's results imply a disequilibrium gap
for fertilizer of 3.95 to 4.91 in 1963.
36. The present value of $163 per year for thirty years at 8percent discount rate
is about $1,835-which is insufficient to justify the expenses on an additional year of
education unless reinforced by other considerations, as suggested below.
per acre. Hence, the marginal product in 1964 of additional extension

in 1959 is $0.177 per acre, and the marginal product of one minute per
farm operator is $11 ($0.177 X 60.7) or of one day allocated over all
farms of a county is $660. Clearly the implication is that the marginal
product of time spent by farmers with county extension agents has
a large payoff.37 To the extent, however, that the extension variable
is positively correlated with left-out variables that represent infor-
mation purchased by farmers from firms operating for profit, inter-
personal communication with neighbors and friends, and other
sources, the marginal product of extension will be biased upward.
Given the changes in economic incentives between 1959 and 1964,
the empirical results suggest that the marginal product of education
and extension from enhanced allocative efficiency is modest to very
large.
Related Research
The generality of the conclusion that investments in education

and information enhance allocative performance when economic
conditions change is subject to verification by examining other bodies
of data. Two potentially interesting cases for analysis are the intro-
duction and adoption of birth-control pills by married couples and
of numerical control of machine tools in the tool-and-die industry.38
These are cases where economic conditions changed markedly in a
short time period and where the organization of decision making is
simple.
First, the introduction and adoption of innovations in contra-
ceptive techniques during the 1960's appear to have revolutionized
the household production of children and sex. The most popular of
these new techniques is the "pill," and during the 1960's, it rose to
dominance among contraceptives used by U. S. married couples.
Furthermore, Ryder reports data for married couples that suggest
households with higher levels of education introduced the pill earlier
and adopted the pill at a faster rate than households with less edu-
cation.39
Second, the introduction and adoption of numerical control of
37. In an aggregate production function study, Griliches, "Research Expenditure,
Education, and the Aggregate Agricultural Production Function," finds a fantastically
large marginal product of $13 of output per dollar of expenditure on agricultural re-
search and extension per year.
38. Numerical control of machine tools is a way of operating them by means of
numerical instructions expressed in code. These instructions, prepared in advance of
tool operation and recorded on tapes and cards, control the sequence of machine op-
erations.
39. Norman Ryder, "Time Series of Pill and IUD Use: United States, 1961-1970,"
Studies in Family Planning, The Population Council, III (Oct. 1972), 233-40.
machine tools in the tool-and-die industry in the United States seem

to have revolutionized the making of consumer and industrial goods.
Numerical control of machine tools, regarded as one of the most im-
portant innovations in manufacturing during this century, was de-
veloped during the mid 1950's and introduced and adopted during
the 1960's and 1970's in the tool-and-die industry. Adoption of nu-
merical control seemed to be quite profitable; and when adoption
occurred, it replaced manual control of machine tools. Mansfield re-
ports some evidence that numerical control was introduced earlier
in firms where managers were college graduates than in other firms.40
But the diffusion process was incomplete when he initiated his
study.
We await the empirical investigation of these and other experi-
ments to see whether the proposition that investment in human
capital enhances allocative performance is verified.
V. CONCLUSIONS
In a growing economy many people are reallocating their re-

sources in response to changes in economic conditions. Allocative
ability is the name given to the human ability to perceive changes in
economic conditions and to respond efficiently. The human-capital
approach to allocative efficiency postulates that allocative ability is
an acquired skill.
The econometric evidence presented in this paper has shown that
investments in education and extension improve the allocative per-
formance of U. S. Corn Belt farmers by increasing the alacrity with
which they respond to changes in economic conditions. Furthermore,
there is no evidence that investment in farmers' education increases
production apart from its influence on the choice of the level of factor
quantities. The major implication is that perceiving and responding
efficiently to changes in economic conditions require allocative ability
that is acquired by investing in education and useful information.
APPENDIX: DATA AND SOURCES
Y-County average corn yield in 1959 and 1964 is derived as total

bushels of corn harvested divided by total acres of corn harvested
for grain from the U. S. Bureau of the Census, Census of Agri-
culture, Vol. I, 1959 and 1964, for respective years.
40. E. Mansfield, J. Rapaport, J. Schnee, S. Wagner, and M. Hamburger, Research

and Innovation in the Modern Corporation (New York: Norton & Co., Inc., 1971),
186-205.
N-County average quantity of nitrogen fertilizer per acre of corn

for 1964 and 1959 is calculated as total pounds of commercial
nitrogen fertilizer applied to corn divided by Census of Agri-
culture number of acres of corn for all purposes for respective
years. Tons of nitrogen applied to corn are estimated as the
Census of Agriculture number of tons of fertilizer applied to corn
multiplied by an estimate of the nitrogen content (proportion)
of commercial fertilizer for the state agricultural subregion where
the county is located. The proportions are from W. Huffman,
"The Contribution of Education and Extension to Differential
Rates of Change," Ph.D. thesis, University of Chicago, 1972.
E-Education refers to a county weighted average of the number
(Census of Agriculture, 1964) of farm operators in seven
schooling completion classes; hence the level of education of
farmers in 1964 also serves as a proxy for their average level of
education in 1959. The weights are from F. Welch, "The Deter-
minants of the Return to Schooling in Rural Farm Areas, 1959,"
Ph.D. thesis, University of Chicago, 1966.
Years of Schooling
0-4 5-7 8 9-11 12 13-15 16

Weights 0.25 0.65 1.00 1.63 2.26 3.05 4.27
R1 -Accumulated rainfall (inches) for

Weather Bureau District where coun
are from U. S. Weather Bureau, Clim
States, Annual Summary, 1959 and 1964.
R2-Accumulated rainfall (inches) for summer months (June, July,
and August) is for the Weather Bureau District where the county
is located. Data are from U. S. Weather Bureau, op. cit.
GDD Length of corn growing season is measured as a county's an-
nual normal-growing-degree days between 50 percent freeze
probability dates. Data are expressed in hundreds and are from
U. S. Department of Agriculture and U. S. Department of
Commerce, "Mean Growing Degree Days Accumulated Weekly
March 1 to Indicated Dates," Weekly Weather and Crop Bul-
letin, LVII (April 6, 1970), 12-16.
Yo-County average base year corn yield is calculated as aggregate
bushels of corn harvested divided by aggregate acres of corn
harvested for grain in 1949. Data are from U. S. Bureau of Census,
Census of Agriculture, 1949.
D1-Geographical dummy variables refer to state agricultural su-
bregions (SARS's), i.e., groups of counties: DI (D164)-Iowa
SASR 71, North Central; D2-Minnesota SASR 72, West Cen-
tral; D3-Minnesota SASR 71, Southwest; D4-Minnesota SASR
57, Central Southeast; D5-Minnesota SASR 56, East Central;
D6-Iowa SASR 70, Western; D7-Iowa SASR 58, East Central;
D8-Iowa SASR 57, Northeast; D9-Iowa SASR 59, South
Central; D10-Illinois SASR 57, Northwest, DI -Illinois SASR
58, Northwest Central; D12-Illinois SASR 51, Central, East

Central; D13-Jllinois SASR 59, Southwest Central; D14-Jllinois
SASR 60, Southwest; D15-Jllinois SASR 42, Southeast;
D16-Indiana SASR 52, Northwest; D17-Indiana SASR 39,
Northeast, North Central; D18-Indiana SASR 38, Central;
D19-Ohio SASR 39, Northwest; D20-Ohio SASR 38, West
Central; D21-Jllinois SASR 52, Northeast. State agricultural
subregion boundaries are from U. S. Department of Agriculture,
Statistical Bulletin No. 431, 1967.
PC Permanent price of corn per bushel is derived as a weighted
average of actual prices of corn for grain (U. S. Department of
Agriculture), Agricultural Statistics (Washington: Government
Printing Office), 1956-1966 inclusive) over three years: P =
0.69PCt + 0.23Pct-, + 0.08Pct-2.
P* Permanent price of nitrogen fertilizer per pound is derived as
a weighted average of the actual price of nitrogen from anhydrous
ammonia (U. S. Department of Agriculture, Agricultural Prices
(Washington: Government Printing Office) Annual Summary,
1956-1966) over three years P1t = 0.69PNt + 0.23PNt-1 +
0.O8PNt-2-
EXTIH-Extension, the county average amount of contact between
agricultural extension agents and farmers, is derived as (average
over 1958 and 1960 of the time in one-tenth hours devoted to
crops by agents doing primarily agricultural work divided by
average over 1958 and 1960 of the number of farmers and other
individuals assisted with any phase of production of grain
crops)112 X (average over 1958 and 1960 of the number of farme
and other individuals assisted with any phase of production of
grain crops divided by Census of Agriculture, 1959, number of
farm operators). Scaling was used in an attempt to remove sQme
of the reporting errors in the underlying data. The extension data
are from unpublished reports of Federal Extension Service.
ACRES-County average number of acres of corn per farm is derived
as the aggregate number of acres of corn for all purposes in 1964
divided by the aggregate number of farms in 1964 (Census of
Agriculture, 1964).
DE-The beginning period (1959) disequilibrium gap in quantity of
nitrogen fertilizer used in corn production is calculated as the
implied average optimal quantity of nitrogen in 1959 (from
equation (13)) minus the average actual quantity of nitrogen in
1959 (see N above).
IOWA STATE UNIVERSITY

Allocative Efficiency-The Role of Human Capital

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Allocative Efficiency-The Role of Human Capital

Enviado por

Direitos autorais:

Formatos disponíveis

Allocative Efficiency: The Role of Human Capital

Author(s): Wallace E. Huffman

I. Introduction, 59.-II. The model, 62.-III. Empirical analysis, 66.-IV. Furt

In a growing economy we observe that many people are reallo-

* The author is indebted to T. W. Schultz, M. Nerlove, and D. G. Johnson for

correctly to changes in economic incentives require allocative ability

Human Capital and Allocative Efficiency

U. S. Agriculture: A Case for Analysis

5. Under conditions of imperfect information modern decision theory has at-

quantity of nitrogen fertilizer applied annually to corn by farmers in

II. THE MODEL

A model consisting of three equations summarizes the change

(1) t= D(P*, T),

(3) d3=B(Z), O < d < 1.

(equation (3)) is specified as a function of a

(5) a9/0E = -(b, + b3EXTIH)f(l -O

(8) aO/aDE = -b5l( - ),

E = education level of farmers,

and where equations (5) through (8) denote the respective pa

agricultural extension agents and farmers is a plausible measure of

(9) Y = A II Xfi(X)eg(x) Y> 0X > X OA > 0,0 ' fi(X) < 1,

where fi (X) and g(X) are functions of a vector X of factors of pro-

(10) Y = ANa1Ea2Wa3ei bh fi Xi a-; Y, N, E, W, Xi ' 0;

0 < a- < 1, A > 0;

D1 = geographical dummy variables,

The variable elasticity of output coefficient of N is specified as

(11) al = a + a2 (GDD X R2 X Yo);

and the index of weather and natural endowment is specified as

(12) W a3 = Rle1R2 2GDDY 3Y Y4,

GDD = length of corn-growing season,

then equation (10) reduces to a Cobb-Douglas production function

(13) N* = [alA1Ea12Wa3P~gP7 le jDj]ID]l/l-al

III. EMPIRICAL ANALYSIS

In this section the methodology of the empirical test and the da

Methodology and the Data

The empirical specification of the model and results of this paper

14. G. E. P. Box and D. R. Cox, "An Analysis of Transformations," Journal of

inferences about behavioral.relationships.16 When independent data

specification of the empirical model presented in this paper.19 The

Estimation of the Production Function and Optimal N

In Y = In A' + a iln N + a2(GDD X R2 X Y0) In N + a2 In E

cient of the interaction variable between nitrogen and the environ-

ESTIMATES OF THE AGGREGATE CORN PRODUCTION FUNCTION:

Independent variables Coefficients t-ratios* Coefficients t-ratios*

Nb 0.2907 6.988 0.3723 7.266

DiC -3.298 -2.978 -3.2698 -2.978

R2 = 0.810. Degrees of freedom = 315.

*Ratio of coefficient to standard error.

PERMANENT PRICES FOR CORN AND NITROGEN FERTILIZER,

($/bu) (U / lb) 1959 1964

Illinois 1.08 1.14 8.9 7.6 0.082 0.067

nomic conditions were changing the optimal rate of N

Estimation of Human Capital's Contribution in the Adjustment

The coefficients of the adjustment function are estimated by

28. Data taken from Agricultural Statistics (Washington, D. C.: Government

IMPLIED OTIMAL N RATES AND DEVIATION OF ACTUAL FROM

of N rates (lb/ac/yr) (lb/ac/yr)

Illinois 48 42.1 167.5 21.6 106.2

data between two points in time, 1959 and 1964. By substituting

(14) In [(N - Nt-5)/(Nt - Nt-5) - 1] = bo + b1E

ESTIMATION OF THE ADJUSTMENT FUNCTION FOR NITROGEN

Explanatory Sample Regression equation

E 1.769 -2.761 -3.792

R2 = 0.525 Degrees of freedom = 178.