Effects of Technological Change and

Institutional Reform on Production

Growth in Chinese Agriculture
Shenggen Fan

Recent rapid agricultural production growth in Chinese agriculture could be attributed to

an increase in inputs, technological change, and institutional reform. An accounting
approach was used to separate the relative contribution of these three factors.
Institutional change, like the introduction of the household production responsibility
system, has contributed to past growth in production. However, technological change is
crucial to furthering production growth because of the limited potential for significant
increase in the use of conventional inputs, in particular land. Continued institutional
change must accompany corresponding technological changes.

Key words: agricultural production growth, China, frontier production function,

institutional change, regional growth, technological change, total factor productivity.

From 1949 to 1986 agricultural production grew rapid growth in agricultural production from 1980
4% a year in China (Fan). This growth was the to 1984 to the household production responsi-
most rapid among all the socialist countries bility system. He found that 20% of productivity
(Wong) and even more rapid than growth in most growth, or 60% of agricultural production
developing countries (Hayami and Ruttan). growth, was attributed to the institutional change.
Contributing to the rapid production growth was However, he ignored the effects of technologi-
a series of technological and institutional changes cal change on production and productivity
and rapid increase of modem inputs. Since 1979, growth. McMillan, Whalley, and Zhu used the
efforts have been made to improve incentives accounting approach to capture the effects of re-
and stimulate production by decentralizing au- forms in prices and incentive systems on total
thority and responsibility for production deci- productivity growth. Their results suggest that
sion to family units. Substantial improvement in 22% of the increase in productivity in China's
productive efficiency has resulted. agriculture between 1978 and 1984 resulted from
Using a traditional accounting approach ini- higher prices and 78% from change in the in-
tiated by Solow, Perkins and Yusuf, and Wiens centive system. They also ignored the effects of
measured the total factor productivity in Chinese technological change.
agriculture; however, the sources of productiv- The purposes of this study are to develop a
ity growth in their studies were not identified. new approach to capture the relative contribu-
Recently, some studies have measured the ef- tions of input growth, technological change, and
fects of institutional change on production and organizational reforms to growth of agricultural
productivity growth. Lin (1987) attributed the production and to apply the approach to the ma-
jor agricultural production regions of China.
During the 1950s, the Chinese government di-
At present, Shenggen Fan is jointly a research associate at the In- vided the country into six administrative re-
ternational Service for National Agricultural Research, the Hague, gions. This division is inappropriate for an anal-
Netherlands, and at the Center for International Food and Agri-
cultural Policy, University of Minnesota. ysis of agricultural productivity. However,
Published as Contribution No. 18172 of the series of the Min- formulating regions on differences in land use
nesota Agricultural Experiment Station based on research supported
by the station and the Rockefeller Foundation.
is not feasible because of data limitations.
The author is grateful to Vernon W. Ruttan for his guidance and Therefore, in this study the country is divided
insights in the development of this paper; to G. Edward Schuh, into seven regions that take into account the
Willis Peterson, and Karen Brooks for their comments; and to Syl-
via W. Rosen for editorial assistance. He also thanks two anony- availability of the agricultural data, the geo-
mous reviewers for their extensive help. graphical features, and the current social and

Copyright 1991 American Agricultural Economics Association

Fan Growth of Chinese Agriculture 267

cultural conditions. These regions adhere closely y

to the administrative division and are as follows: Frontier 2
(a) Northeast (NE): Heilongjiang, Liaoning, and
Jilin provinces. (b) North (N): municipalities of
Beijing and Tianjin; Hebei, Henan, Shandong,
Shanxi, Shaanxi, and Gansu provinces. (c)
Northwest (NW): autonomous regions of Nei
Monggol, Ningxia, Xinjiang, and Tibet; Qin-
ghai province. (d) Central (C): Jiangxi, Hunan,
.-~- Frontier 1
and Hubei provinces. (e) Southeast: (SE):
Shanghai municipality; Jiangsu, Zhejiang, and
Anhui provinces. if) Southwest (SW): Sichuan,
Guizhou, and Yunnan provinces. (g) South (S):
Guangxi autonomous region; Fujian and Guang-
dong provinces. I
-::..---------L.....-----Io----x
X1 X2
Effects of Input Growth, Technological
Change, and Efficiency Improvement Figure 1. Effects on production growth of
input increase, technological change, and ef-
In traditional productivity theory, total produc- ficiency improvement
tion growth consists of movements along the
production function (an increase of total inputs)
and shifts of the production function (techno- production frontiers I and 2, respectively. If
logical change), assuming that the firm is per- production were perfectly efficient, output would
fectly efficient in production. The growth rate be Tf at time 1 and Tf at time 2. However, the
of total factor productivity is the growth rate of producer's realized output is Y I at time I and Y2
total output minus the growth rate of total input; at time 2 owing to production inefficiency.
hence, technological change is considered the Technological change is measured by the dis-
unique source of productivity growth, and the tance between frontier 2 and frontier 1, i.e.,
effects of efficiency improvement on productiv- n - Tt . Inefficiency is measured as the dis-
ity growth are ignored. The assumption of per- tance between the frontier and the output real-
fect efficiency in production is unrealistic. Dif- ized by the producer, i.e., E I at time 1 and E 2
ferences among firms between realized output at time 2. Hence, the improvement of efficiency
and potential output are caused by differences over time is the difference between Eland E 2 •
in the capacity to use new technological knowl- The contribution of input change is measured as
edge and in the motivations of farmers. If this Z. Therefore, the total production growth can be
assumption is relaxed, total production growth decomposed to three effects: input growth, tech-
can be attributed to efficiency improvement as nological change, and efficiency improvement.
well as to increased inputs and technological
change. Different policy inferences may be drawn
Y2 - YI = Z + (Tr - Tn + (E I - E2) .
inasmuch as technological change and effi- Prior to the introduction of the household pro-
ciency improvement represent fundamentally duction responsibility system to Chinese agri-
different sources of growth in production. culture, production was organized by produc-
Therefore, a new approach will be developed to tion teams or state farms. A farmer's income
capture all three effects on production growth. was not closely related to his production effort.
In this study technological change is defined After the reform, when producers became re-
as a shift of the frontier production function. Ef- sponsible for their plots, they worked harder,
ficiency improvement is defined as the decrease allocated resources more efficiently, and pro-
in the distance between the firm's realized out- duced more output with the same input and
put and its potential output (or frontier). The technology. If only technological change is con-
different sources of production growth are shown sidered as the source of production and produc-
in figure 1. At times I and 2, the producer faces tivity growth, the effects of technological change
will be overestimated by ignoring institutional
change. Therefore, efficiency improvement is
1 Hainan was not separated from Guangdong province. used in this study to capture the effect of insti-
268 May 1991 Amer. J. Agr. Econ.

tutional change on production and productivity

growth. (2) { ( u., )}
E exp - - -
u.; + Vit

Frontier Production Function

= ex p[ _ (cr cr
u

a
v
) (/(.)
1 - F(.)
_ Bit
a 1- A
VA)],
The frontier production function approach, ini- where Bit = Vit + Uit, a is standard error of Bit>
= ~/lT , and 1(.) and F(.) are the values of
2
tiated by Farrell in 1957, has been expanded by A
various methods of measuring and computing the standard normal density function and stan-
production functions and efficiency (Lovell and dard normal distribution function evaluated at
Schmidt). The main approaches include pure
programming, modified programming, the de-
terministic statistical frontier, and the stochastic
frontier. Pitt and Lee indicated that the pro-
gramming approach and the deterministic fron- The next step of the specification is to choose
tier approach do not allow for random shocks in an appropriate functional form. Consider a pro-
the production process; as a result, a few ex- duction process that uses n inputs to produce one
treme observations can determine the frontier and output represented by the production function
exaggerate the maximum possible output. In this
(3) Y = I(Xl ... , ... Xn , T),
study, the stochastic frontier approach is em-
ployed to avoid this problem. where Y is output, Xi is ith input and T is used
Consider the following production function: to catch technical progress (time trend). The un-
r b) e"it e"it ,or restricted translog form can be used to represent
(1) Yit = fi\Xit, production function (3). However, the translog
1nYit = 1nfixit, b) + Vit + Uit form needs considerable data and has many
variables which may lead to multicollinearity
where i denotes the ith firm or region, and t de- problems. Consider a restriction that all inputs
notes time t; Y it is output; Xit is 1 x k rows of are separable from each other, but each input
inputs; I(Xit, b) is potential output; Vit is a sto- cannot be separated from technical progress:
chastic variable representing uncontrolled ran-
dom shocks in production; and u., is one-sided
distribution, U :::; 0, which represents technical The theoretical background of this form comes
inefficiency. In (l), I(xit, b )ev/t is the stochastic from the fact that every input changes over time
frontier, given that Vit consists of random factors while the effects among inputs are indirect
outside the firm's control. The nonpositive dis- through time. Then, the following production
turbance U indicates that output must lie on or function form can be used to represent (4):
Vit
below the frontier I(Xit, b )e because e"it has a
value between zero and one. It is assumed that
for t =1-= t', E(UirUit') = 0 for all i, and E(uituic')
= 0 for all i =1-= j. In this specification, the firm's
inefficiency may change over time by learning
+ L a.. 1n(xi) x t + atl.
from experience . We also assume U is truncated i
normal with variance ~, v is normal with mean
zero and variance cr?" and E(UitVit') = O. If all inputs and time are considered separa-
The efficiency for a firm or region i at time ble, the production function can be expressed as
t, then, is defined as

The Cobb-Douglas production function can

be used to represent (6)

Based on the conditional distribution of Uit'

given the distribution Vit + Uit, the efficiency of
(7) In(Y) = a; + L a, In(xJ + as.
a specific firm or region at a given time can be
measured as (Ka1irajan and Flinn) Owing to the serious multicollinearity prob-
Fan Growth of Chinese Agriculture 269

lem of the translog form and the constancies of fate." Machinery input is measured by total
production elasticities in the Cobb-Douglas form, horsepower at year end."
functional form (5) is used for the estimations. Manurial fertilizer, which always has been very
The Cobb-Douglas form and average production important in China, includes animal, human, and
functions are also estimated for comparison pur- crop wastes; green manures; and water plants.
poses.? In this study, manurial fertilizer is measured from
the agricultural population (i.e., human waste)
and numbers of domestic animals. 8 Draft ani-
Estimation of Production Functions and mals are measured at year end in units of heads
Efficiency that are used for agricultural activities and rural
transportation. They include water buffaloes,
Panel data from twenty-nine provinces, munic- cattle, horses, asses, mules, and camels." Irri-
ipalities, and autonomous regions in 1965, 1970, gation input is measured as irrigated areas. 10
1975, 1976, through 1986 are used in the esti- The results of production function estimation
mations. Gross agricultural production value for the different specifications are shown in ta-
serves as the aggregate total output using 1980 ble 1. The ordinary least squares technique is
constant prices. The subaggregates are (a) crop used for the average production function esti-
production, (b) forestry, (c) animal husbandry, mation and the maximum likelihood technique
(d) sideline industries, and (e) fisheries. Rural for the frontier production function. The Cobb-
industry at all levels (including town, village, Douglas form is used for regressions 1 and 2.
and teams) is excluded from agricultural pro- Time trend (T) measures neutral technological
duction. 3 change over time. Except for machinery and ir-
Labor input in agriculture is measured by the rigation, the coefficients of regressions 1 and 2
numbers of employed persons at year end." The are very significant considering the crudeness of
sum of sown areas and pasture is used to mea- the data. However, the negative coefficients of
sure land input because the arable land data are draft animals are unrealistic. The sum of pro-
inaccurate. Pasture areas are calculated in sown duction elasticities of traditional inputs (except
land area equivalence for output value, i.e., one for draft animals) is more than. 75, which im-
unit of pasture equals .0124 of a unit of sown plies that traditional inputs still dominate Chi-
land (in 1985). 5 Chemical fertilizer input is na's agricultural production. Chemical fertilizer
measured by pure nutrients, using the following input plays an important role in production. The
percentage: 20% for ammonium sulfate, 18.7% significant and positive time trend coefficient
for super phosphate, and 40% for potassium sul-

6 The data before 1980 are reported in National Agricultural Sta-

tistical Materials for 30 years (1949-1979). The data after 1980
2 The traditional estimation of a production function assumes that are taken from various issues of China's statistical yearbooks.
every firm is technically efficient, resulting in the average produc- 7 The horsepower of 1965 and 1970 is interpolated based on the
tion function, i.e., Yit = f(xit, bie:", where fit has normal distri- numbers of hand tractors and other tractors. The horsepower from
bution, N(O, if). 1970 to 1975 is taken from the National Agricultural Statistical
3 The time series of provincial monetary value of total production
Materials for 30 Years (1949-1979). The horsepower after 1980
(measured in 1980 constant prices) before 1985 is reported in Col- is taken from various issues of the statistical yearbooks.
lection of Statistical Materials in National Income, 1945-1985, State 8 The FAD estimated that one animal (horse unit) produces about
Statistical Bureau. The data after 1985 are reported in China's Sta- 4 tons of manure per year and a person produces .25 ton per year.
tistical Yearbooks, 1986, 1987, State Statistical Bureau. Manure contains 2.2% pure nutrient, and the manure availability
4 The provincial data of labor before 1980 are calculated from
is about 75% of total use. Therefore, manurial resources are esti-
the provincial agricultural population. mated as follows:

Lit = Pit x ri,80 x r n" , Annual manurial resources (tons)

r n,80 = ((.25 X rural population + 4
x numbers oflivestock) x 2.2%) x 75%.
where Lit denotes ith region's labor input in year t; Pit, ith region's
population in year t; r,,80, ith region's ratio of labor to population The results of this estimation are not significantly different from
in year of 1980; rn ,80 , national ratio of labor to population in year that of Stone (Tang and Stone).
1980. rn,t, national ratio of labor to population in year t. The data 9 The numbers of draft animals before 1980 are taken from the

for agricultural population before 1980 are taken from National Ag- National Agricultural Statistical Materialsfor 30 Years (1949-1979).
ricultural Statistical Materials for 30 Years (1949-1979), State The numbers after 1980 are taken from various issues of statistical
Statistical Bureau. The data of agricultural labor after 1980 are taken yearbooks after 1980.
from various issues of China's statistical yearbooks. 10 The data of irrigated areas before 1980 are reported in National
5 The data for sown areas and pasture are taken from National Agricultural Statistical Materials for 30 Years (1949-1979). Those
Agricultural Statistical Materials for 30 Years (1949-1979), State after 1980 are published in the various issues of statistical year-
Statistical Bureau. books.
270 May 1991 Amer. J. Agr. Econ.

Table 1. Estimates of Production Functions

Rl R2 R3 R4 R5 R6
Regression No: (Average) (Frontier) (Average) (Frontier) (Average) (Frontier)

Constant -2.81 -2.70 -2.81 -3.19 -2.92 -2.82

(10.72)" (-11.27) (-5.23) (-6.13) (-6.24) (-6.14)
LABOR .278*b .266* .420* .417* .438* .428*
(7.19) (6.14) (5.16) (4.66) (5.40) (4.94)
LAND .356* .379* .243* .331* .246* .261*
(7.88) (9.39) (2.40) (3.99) (2.78) (3.60)
C. FERT c .235* .236* .140* .089*** .132* .132*
(8.71) (9.29) (2.70) (1.66) (2.57) (2.61)
MACHINERY .055** .051** .078*** .123* .075*** .068***
(1.77) (1.82) (1.39) (2.52) (1.35) (1.30)
M. FERr .185* .178* .227* .266* .241* .241*
(5.30) (5.67) (2.99) (3.27) (4.18) (3.40)
ANIMALS -.132* -.133* .002 -.026
(-5.13) (-4.94) (.037) (- .301)
IRRIGATION .059** .055** .009 -.037
(1.81) (1.66) (.145) (- .537)
TC .0123* .0125* .0014 .0420 .0496 .0505***
(2.41) (2.17) (.364) (.980) (1.28) (1.33)
LABORr -.0097** -.0109** -.0111* -.0108**
(-1.822) (-1.79) (-2.07) (-1.83)
LANDr -.0024 -.0065 -.0073 -.0077***
(- .368) (-1.20) (-1.25) (-1.64)
C. FERTT c .0068** .0087* .0083* .0081*
(1.83) (2.41) (2.23) (2.30)
MACHINERyr .0080** .0083* .0092* .0098*
(1.93) (2.08) (2.33) (2.56)
M. FERTT c -.00006 -.0014 -.0050 -.0051
(- .013) (- .273) (-1.27) (-1.13)
ANIMALST c -.006 -.0041
(-1.51) (-.725)
IRRIGATIONr -.0003 .0006
(-.064) (.118)
T2 .00147* .0013* .0012** .0011***
(2.23) (2.30) (1.80) (1.58)
A .822* 1.278* .821***
(2.17) (3.23) (1.56)
(T .288* .266* .254*
(9.38) (10.99) (6.84)
Observations 406 406 406 406 406 406
R2 .940 .932 .957 .942 .954 .959
a Numbers in parentheses are r-test values.
b Single asterisk indicates significant at 5% level, double asterisk indicates significant at 10% level, and triple asterisk indicates significant
at 20% level.
C C. FERT is chemical fertilizer; M. FERT, manurial fertilizer; T, Time Trend, T = I for 1965, T = 6 for 1970, ... T = 22 for 1986;
LABORT. cross term of labor and time trend; LANDT, cross term of land and time trend ... ; IRRIGATIONT, cross term of irrigated
areas and time trends.

strongly suggests that total factor productivity in cross-term of each input and time trend captures
Chinese agriculture has increased through neu- the relative changes of each input in total input
tral technological change. over time. The greater significance of the coef-
Functional form (5) is used for regressions 3, ficients in regression 4 relative to those in
4, 5, and 6. Production elasticity for input i in regression 3 implies that the frontier production
this production functional form is alnY/ alnxi = function used for estimation improved the re-
a, + aitt. Thus, if ail > 0, production elasticity sults. Labor, land, draft animals, and manurial
of input i is increasing; if ail < 0, production fertilizer playa decreasing role in production,
elasticity of input i is decreasing. whereas production elasticities of chemical fer-
Regressions 3 and 4 use the same input vari- tilizer and machinery increase over time.
ables as regressions 1 and 2. In addition, the Because the coefficients of draft animals are
Fan Growth of Chinese Agriculture 271

negative and the irrigation coefficients are not labor, .223; land, .143; chemical fertilizer, .177;
significant in regressions 1 through 4, these two machinery, .122; and livestock, .233. Compar-
variables are omitted in regressions 5 and 6. Some ing those to the production elasticities in table
effects of draft animals on production are re- 2, we observe that the elasticities of land and
flected by manurial fertilizer. The improvement labor in China are greater than those in the so-
in irrigation in China mainly occurs through in- cialist countries, indicating that Chinese agri-
creased irrigation power rather than an expan- culture uses more traditional inputs than other
sion in the size of irrigated areas. Therefore, these socialist countries.
omissions do not greatly affect the estimation. The level and variability of technical effi-
Furthermore, these omissions avoid the collin- ciency for each region are calculated in table 3,
earity among draft animals, manurial fertilizer, using (2) and the results of the frontier produc-
and land input. Most of the estimators in regres- tion function from regression 6. During the 1960s
sions 5. and 6 are significant. The omissions of and 1970s, technical efficiency was about 70%.
draft animals and irrigation did not cause changes Efficiency has improved significantly since the
in other coefficients. Again, the frontier esti- institutional change in 1979. The institutional
mation is superior to the average estimation. change has three effects: (a) Farmers' incomes
Table 2 shows that production elasticities and efforts have been linked through improved
(calculated using regression 6) of traditional in- incentive systems. (b) Farmers may leave agri-
puts-land, labor, and manurial fertilizer-are culture to engage in nonagricultural activities
decreasing: labor by 3.6% per year; land, 4.6%; (mainly rural industry), thus improving the land/
and manurial fertilizer, 3. 1%. The annual rates labor ratio. (c) Farmers may allocate their time
of increase of production elasticities for modern and resources to produce high-profit crops, which
inputs-machinery, 6.5 %; chemical fertilizer, has improved allocative efficiency and the full
3.9%-are greater than the rates of decrease for use of regional comparative advantages.
traditional inputs. It is widely accepted that the introduction of
The results in table 2 can be compared to those the household production responsibility system
of other studies. For example, Ma, Calkins, and enlarged the differences in income among re-
Johnson estimated the production elasticities gions (Jiang and Luo). However, there is no
(using 1984 data) for Shuyang county, Jiangsu evidence that the differences in productive ef-
province. The ranges in value for their elastic- ficiency have increased-the coefficient of vari-
ities were as follows: labor, .25 to .36; land, .17 ation in productive efficiency has decreased since
to .20; chemical fertilizer, .17 to .23; manurial the reform (see the last column of table 3). The
fertilizer, .08 to . 11; and other inputs, .22 to disparity between the production efficiency im-
.29. The elasticities vary depending on crops. provement and income growth among regions
Wong's estimation of the production functions suggests that the substantial improvement in
(using 1960-80 data) for nine socialist countries production efficiency in poor regions owing to
resulted in the following production elasticities: the recent institutional reform did not result in
a corresponding increase in income. One reason
for this lack of response is the distorted prices
in agriculture. Despite the substantial increase
Table 2. Production Elasticities for Differ- in prices in the last ten years, the agricultural
ent Inputs, 1965-1985 product prices still are not reflected by supply
Chemical Manurial
and demand. Further reform in prices is needed
Labor Land Fertilizer Machinery Fertilizer to give farmers greater incentive to promote fur-
ther production growth. Another reason is the
1965 .417 .253 .140 .078 .235 uneven development of rural industry. The low
1970 .363 .215 .181 .127 .210
1975 .309 .176 .221 .176 .185
level of income per capita, especially in the
1976 .298 .168 .229 .186 .180 Southwest, is the result of the underdevelop-
1977 .287 .161 .237 .195 .174 ment of rural industry .
1978 .276 .153 .246 .205 .169
1979 .265 .145 .254 .215 .164
1980 .254 .138 .262 .225 .159
1981 .244 .130 .270 .234 .154 Accounting for Total Production Growth
1982 .233 .122 .278 .244 .149
1983 .222 .114 .286 .254 .144
1984 .211 .107 .294 .264 .139 In this part an empirical approach is developed
1985 .200 .099 .303 .274 .134
and used to separate the effects on production
272 May 1991 Amer. J. Agr. Econ.

Table 3. Level and Variability in Technical Efficiency of Seven Regions for Selected Years
National
NE N NW C SE SW S Average C.V."

1965 .868 .433 .698 .728 .679 .681 .644 .646 .191
1970 .853 .561 .844 .844 .847 .731 .846 .772 .138
1975 .887 .581 .808 .881 .866 .652 .812 .761 .127

Average 65-79 .892 .574 .758 .850 .817 .713 .789 .737
Rank 1 7 5 2 3 6 4
C.V.65-79 .033 .117 .103 .069 .084 .061 .087 .132
Rank 7 1 2 5 4 6 3

1980 .917 .625 .692 .826 .802 .781 .756 .753 .122
1981 .911 .630 .774 .858 .851 .791 .758 .768 .114
1982 .911 .645 .777 .885 .863 .851 .810 .788 .109
1983 .939 .681 .751 .863 .847 .858 .795 .791 .103
1984 .934 .726 .799 .908 .900 .894 .831 .831 .070
1985 .891 .725 .829 .909 .906 .891 .870 .843 .076

.1 70s-85 b .001 .151 .071 .059 .089 .178 .081 .106

Rank 7 2 5 6 3 1 4

.1 65-85 .023 .292 .131 .181 .227 .210 .226 .197

Rank 7 1 6 5 2 4 3
Average 65-85 .898 .616 .766 .863 .844 .771 .807 .772
Rank 1 7 6 2 3 5 4
C. V. 65-85 .033 .123 .081 .056 .073 .105 .081 .130
Rank 7 1 3 6 5 2 4

• C. V. is coefficient of variation
b J 70s-85 indicates the absolute improvement of technical efficiency between 1965-79 average and 1985.

growth of an increase in inputs, technological nological change. The second term captures the
change, and institutional reform. Using func- effect of input change on production growth; it
tional form (5), the production function can be is the sum of growth rates in inputs weighted by
expressed as the relevant production elasticities. The third term
measures the the effects of biased technological
(8)
change on production growth; if it is positive,
lnY(t) = a; + 2: a, lnr, (r) + 2: au (lnr, (t)) X t output has increased through biased technolog-
ical change (using abundant resources to sub-
stitute for scarce resources). The last term re-
flects the effect of institutional change (or
efficiency improvement) on production growth.
(9) = lnAo(t) + 2: a, (t)lox (r) + lo£(t),
i
Using (10), the accounting for the sources of
total production growth is presented in table 4.
Neutral and biased technological change are
where lnAit) = a o + att + at/ + v(t) , alt) = considered as total technological change in the
a, + aut, and E(t) = eu(t). accounting and treated as the residual. For the
Taking the first derivative of (9) with respect whole country, total production growth rate was
to time t, the growth rate of total production can 5.04% per year from 1965 to 1985; 57.7% of
be accounted for as the growth is explained by increased use of total
(10) alnY(t)/at = alnAit)/at + 2: a, (t) input and 42.3% by growth in total factor pro-
ductivity. About 63% of productivity change is
x alox (t) / at ~ 2: lnr, (t)
i
attributed to institutional change (or efficiency
i
improvement) and about 37%, to technological
x aalt)/at + alo£(t)/at. change. The increase of labor still explains about
7.7% of total production growth. The change of
The first term in (10) measures neutral tech- land input had the least effect because acreage
Fan Growth of Chinese Agriculture 273

Table 4. Accounting for Growth of Total Agricultural Production in Terms of Annual Growth
Rates, 1965-1985
NE N NW C SE sw S National

Total Production Growth

5.09 5.88 3.70 4.40 5.50 4.40 4.50 5.04
(100) (100) (100) (100) (100) (100) (100) (100)
Total Input Growth
3.10 3.10 2.72 2.71 2.80 3.66 2.55 2.91
(60.9) (52.7) (73.5) (61.6) (50.9) (83.2) (56.7) (57.7)
Labor .23 .24 .45 .43 .25 .67 .49 .39
(4.5) (4.1) (12.2) (9.8) (4.5) (15.2) (10.9) (7.7)
Land .04 -.05 -.07 -.01 .06 .11 0 .002
(.8) (-.9) (-1.9) (- .2) (1.1) (2.5) (0) (.04)
C. Fert. 1.73 1.61 1.51 1.22 1.29 1.45 .79 1.32
(34.0) (27.4) (40.8) (27.7) (23.5) (33.0) (17.3) (26.2)
M. Fert. .20 .35 .18 .13 .04 .36 .31 .25
(3.9) (6.0) (4.9) (3.0) (.73) (8.2) (6.9) (5.0)
Machinery .90 .95 .65 .94 1.16 1.07 .96 .95
(17.7) (16.2) (17.6) (21.4) (21.1) (24.3) (21.3) (18.8)
Total Productivity Growth
1.99 2.78 .98 1.69 2.70 .74 1.95 2.13
(39.1) (47.3) (26.5) (38.4) (49.1) (16.8) (43.3) (42.3)
Institutional Change
.13 2.61 .86 1.11 1.45 .82 1.52 1.34
(2.5) (44.4) (23.2) (25.2) (26.4) (18.6) (33.8) (26.6)
Technological Change
1.86 .17 .12 .58 1.25 -.08 .43 .79
(36.5) (2.9) (3.2) (13.2) (22.7) (-1.8) (9.6) (15.7)

Note: (10) is employed for the accounting.

used for agriculture remained nearly constant. change to production growth also has varied
Among all inputs, increased chemical fertilizer substantially among regions. Total factor pro-
input contributed most significantly to produc- ductivity growth in the Northeast is mainly ex-
tion growth (26.2%), while manurial fertilizer plained by technological change. Technological
explained 5% of total production growth. The change contributed more than 45% of the total
increase in machinery use is the second most factor productivity in the Southeast. However,
important factor in total production increase. technological change in the North, Northwest,
The differences ill sources of production growth and Southwest contributed little to total factor
among regions are substantial because of dif- productivity and total production growth.
ferences in the resource endowments and total
factor productivity growth. Growth in total ag-
ricultural production varied from 3.70% in the Concluding Comments
Northwest to 5.88% in the North region. The
contribution of total input growth to production The major findings of this study are summarized
growth varies from 50.9% in Southeast to 83.2% as follows: The estimates of the frontier pro-
in Southwest. The differences in modern input duction functions indicate that traditional inputs
(chemical fertilizer and machinery) growth ex- are still important to China's agriculture. How-
plains most of the differences in total input ever, the importance of the traditional inputs of
growth. Among modern inputs, chemical fertil- land, labor, and manurial fertilizer is decreasing
izer has the largest effects. The differences in rapidly. In contrast, the coefficients of modern
traditional input growth are small. inputs, e. g., chemical fertilizer and machinery
The differences of the effects of institutional inputs, were small in 1965 but have since in-
change on production growth explain the largest creased rapidly. By 1985, the modem inputs were
share of the differences in total production as important as the traditional inputs.
growth, ranging from 2.5% in Northeast to 44.4% Efficiency measurements indicate that the
in North. The contribution of technological household production responsibility system has
274 May 1991 Amer. J. Agr. Econ.

contributed significantly to production growth. Technological Change

However, the regional differences in perfor-
mance are large. In general, land-scarce regions The results of this study indicate that techno-
gained more from the reform. logical change accounts for 15.7% of total ag-
The accounting for production growth showed ricultural production growth in China. Com-
that a significant share of total production growth pared to other countries, this proportion is very
still can be attributed to increases in traditional small. In Japan, from 1960 to 1980, 47.4% of
inputs. Among all inputs, increased chemical total production growth stemmed from techno-
fertilizer use was the most important source of logical change, and technological change ac-
production growth. Increased machinery input counted for 84.2% of the growth in U.S. total
ranked second in importance. Total input growth output (Hayami and Ruttan).'! Underinvestment
explained 57.7% of total production growth. The in agriculture may explain the slow technolog-
residual, the proxy for technological change and ical change in China. In 1985, the agricultural
efficiency improvement, accounted for 42.3% sector produced 28.1 % of total national output
of total production growth. Institutional change and 41.1 % of national income, although the ag-
has had greater effects on productivity and pro- ricultural investment was only 3.4% of total in-
duction growth than has technological change. vestment (China's Statistical Yearbook, 1986).
These findings have important policy impli- The underinvestment in agriculture has resulted
cations in promoting further production growth in poor rural infrastructure and insufficient ag-
and smoothing regional inequalities. China's ricultural research. An increase in agricultural
population reached 1065.29 million in 1987. The investment, especially in research and devel-
population growth rate from 1949 to 1987 was opment, is needed to stimulate technological
1.84%, although it declined to 1.29% in last de- change.
cade. Further decreases in population growth will
not be easy in the next decade because the base
population is large and those born in the 1960s Institutional Change
are entering reproductive age. Thus, the demand
for food will continue to grow even apart from Recent institutional changes have improved ag-
income effects. The demand for cash crops is ricultural production efficiency greatly; 26.6%
increasing with the development of industriali- of production growth has been contributed by
zation. How to meet the future demand for rapid institutional change. The new strategy should
increases in food and in industrial materials is focus on greater regional specialization, based
an urgent problem. on comparative advantages. The self-suffi-
ciency policy both at the national level and local
levels should be discarded. Crops should be
Increased Input Use grown where soil and climate provide the most
favorable conditions. Although rural labor has
The quickest solution for China is to increase more opportunities to work outside agriculture,
the use of inputs, such as land, labor, chemical labor immobility will become a major source of
fertilizer, machinery, and others, However, the inefficiency. The pattern of land holdings (in
total land input is likely to decline in the future terms of size distribution of farm), land tenure
(Sun). Without an increase in land areas, an in- and other contractual arrangements in agricul-
crease in labor will have only a limited effect ture should be adjusted appropriately to gain more
on total production. Increased use of modem in- efficiency. The recent introduction of factor and
puts, especially chemical fertilizer, likely has the product markets in agriculture has contributed to
greatest potential for increasing total produc- more efficient allocation of resources; however,
tion. Although fertilizer input per unit of land instability of input and output prices and the in-
in China is higher than in most developing sufficient supplies of modem inputs will con-
countries, the output increase from greater fer- tinue constrain agricultural production.
tilizer use is still potentially large in some re-
gions (see table 4), particularly in the Northeast,
Northwest, North, and Southwest. Increased
11 See table 7.2, Hayami and Ruttan (1985 00.). Total output
machinery input will have little effect on pro- growth is 1.9% a year in Japan from 1960 to 1980; total produc-
duction unless it increases land productivity. tivity growth (the contribution of technical change to output growth),
Thus, a top priority in mechanization involves .9%. Thus, the relative contribution of technical change to total
output growth is 47.4%. Using the same calculation, the relative
increased land productivity (e.g., mechaniza- contribution of technical change to total output growth is 84.2% in
tion of irrigation). the United States.
Fan Growth of Chinese Agriculture 275

Smoothing Regional Inequalities Kalirajan. K. P., and 1. C. Flinn. "The Measurement of

Farm Specific Technical Efficiency. " Pakistan J. Appl.
Differential growth rates in agricultural devel- Econ. 2(1983): 167-80.
opment among regions of a country represent a Lin, Justin Yifu. "Household Farm, Cooperative Farm, and
Efficiency: Evidence from Rural Decollectivization in
persistent challenge to policy makers. Smooth-
China." Economic Growth Center Work Pap. No. 533,
ing the differences in technological and insti- Yale University, 1987.
tutional changes among regions is needed to re- - - . "The Household Responsibility System in China's
duce the differences in production and income Agricultural Reform: A Theoretical and Empirical
growth. A well-integrated and extensive physi- Study." Econ. Develop. and Cultur. Change
cal infrastructure and a strong regional agricul- 36(1988):200-224.
tural research capacity adapted to the needs of Lovell, C.A.K., and P. Schmidt. "A Comparison of Al-
the regional agricultural economy are important ternative Approaches to the Measurement of Produc-
in contributing to development of new compar- tive Efficiency." Application of Modern Production
ative advantages in technology in the regions Theory: Efficiency and Productivity, ed. Ali Dorgra-
maci and Rolf Fare. Boston: Kluwer Academic Pub-
disadvantaged by resource endowments and to
lishers, 1988.
stimulation of more even rates of technological Ma, Nathaniel J., Peter H. Calkins, and Stanley R. John-
change across regions. son. "The Household Responsibility System: Techni-
New agricultural policies and institutional cal and Allocative Efficiency vs. Equity." China's Ru-
changes should create more geographically even ral Development Miracle, with International
growth in agricultural production and income. Comparison, ed. John W. Longworth. University of
For example, crop prices should be raised in or- Queensland, Australia, 1989.
der to narrow the income differences between McMillan, John, John Whalley, and Lijing Zhu. "The Im-
the regions with an advantage of crop produc- pact of China's Economic Reforms on Agricultural Pro-
tion but with a disadvantage of rural industry, ductivity Growth." J. Polito Econ. 97(1989):781-807.
Perkins, Dwight, and Shahid Yusuf. Rural Development in
and the regions with well-developed rural in-
China. Baltimore MD: Johns Hopkins University Press,
dustries. 1984.
Pitt, M. M., and L. F. Lee. "The Measurement and Sources
[Received August 1989; final revision received of Technical Inefficiency in the Indonesian Weaving
May 1990.] Industry." J. Develop. Econ. 9(1981):43-64.
Solow, Robert. "Technical Change and the Aggregate Pro-
duction Function." Rev. Econ. and Statist. 39 (1957):
312-22.
References State Statistical Bureau. China Statistical Yearbook. Bei-
jing, China, various issues. Statistical Press, 1981, 1982,
Agricultural Yearbook Editing Committee. China Agricul- 1983, 1984, 1985, 1986, and 1987.
tural Yearbook. Beijing: Agricultural Publishing House, - - . Collection of Statistical Materials in National In-
1980,1981,1982,1983,1984,1985,1986, and 1987, come, 1945-1985. Beijing: China Statistical Press,
various issues. 1987.
Fan, Shenggen. "Regional Productivity Growth in China's - - . National Agricultural Statistical Materials for 30 Years
Agriculture." Ph.D thesis, University of Minnesota, (1949-1979). Unpublished data. Beijing, 1980.
1989. Sun, Han. "9.8 Million Square Kilometers of Territory."
FAD. China: Recycling of Organic Wastes in Agriculture. Feeding a Billion, ed. Wittwer Sylvan. East Lansing:
Rome: FAa Soils Bull. No. 40, 1977. Michigan State University Press, 1987.
Farrell, M. J. "The Measurement of Productive Effi- Tang, Anthony. An Analytical and Empirical Investigation
ciency." J. Royal Statist. Soc., Series A 120(1957):253- ofAgriculture in Mainland China, 1952-1980. Taibei:
81. Chung-Hua Institution for Economic Research, 1984.
Hayami, Yujiro, and Vernon W. Ruttan. Agricultural De- Tang, Anthony, and Bruce Stone. Food Production in PRC.
velopment: An International Perspective. Baltimore MD: Washington DC: International Food Policy Research
Johns Hopkins University Press, 1971, 1985. Institute, 1980.
Jiang, Jinyong, and Xiaopeng Luo. "Changes in the In- Wiens, B. Thomas. "Technical Change." The Chinese Ag-
come of Chinese Peasants since 1978." China's Rural ricultural Economy, ed. Randolph Barker, Radha Sinha
Development Miracle, with International Comparison, and Beth Rose. Boulder CO: Westview Press, 1982.
ed. John W. Longworth. University of Queensland, Wong, Lung-Fai. Agricultural Productivity in the Socialist
Australia, 1989. Countries. Boulder CO: Westview Press, 1986.