12
TECHNOLOGICAL CHANGE, TECHNICAL AND ALLOCATIVE EFFICIENCY IN CHINESE AGRICULTURE: THE CASE OF RICE PRODUCTION IN JIANGSU SHENGGEN FAN* International Food Policy Research Institute (IFPRI), Washington, DC, USA Abstract: This paper develops a frontier shadow cost function approach to estimate empirically the eects of technological change, technical and allocative eciency improvement in Chinese agriculture during the reform period (1980–93). The results reveal that the first phase rural reforms (1979–84) which focused on the decentralization of the production system have had significant impact on technical eciency but not allocative eciency. However, during the second phase reforms which was supposed to focus on the liberalization of rural markets, technical eciency improved very little and allocative eciency has increased only slightly. Copyright # 2000 John Wiley & Sons, Ltd. 1 INTRODUCTION Institutional changes and market reforms initiated in 1979 have had great impacts on China’s agricultural production and productivity growth. Numerous studies have examined the eects of these reforms (Lin, 1992; Fan, 1991; Wen, 1993), but all used the production function approach. Production functions are easy to estimate, and can be used to identify the eects on production growth of technological change derived from the use of new technologies and technical eciency improvement due to institu- tional and market reforms. But the production function approach cannot measure the impact of improvement in allocative eciency due to these changes and reforms. This has become an increasingly serious problem because allocative eciency improvement may have been a more important component of overall eciency improvement in Chinese agriculture since 1984. The government did not begin to focus on the reform CCC 0954–1748/2000/010001–12$17.50 Copyright # 2000 John Wiley & Sons, Ltd. Journal of International Development J. Int. Dev. 12, 1–12 (2000) * Correspondence to: Mr Shenggen Fan, International Food Policy Research Institute, 2033 K Street NW, Washington, DC 20006, USA.

Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

Embed Size (px)

Citation preview

Page 1: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

TECHNOLOGICAL CHANGE,TECHNICAL AND ALLOCATIVE

EFFICIENCY IN CHINESEAGRICULTURE: THE CASE OF

RICE PRODUCTION IN JIANGSU

SHENGGEN FAN*

International Food Policy Research Institute (IFPRI), Washington, DC, USA

Abstract: This paper develops a frontier shadow cost function approach to estimate

empirically the e�ects of technological change, technical and allocative e�ciency

improvement in Chinese agriculture during the reform period (1980±93). The results

reveal that the ®rst phase rural reforms (1979±84) which focused on the decentralization

of the production system have had signi®cant impact on technical e�ciency but not

allocative e�ciency. However, during the second phase reforms which was supposed to

focus on the liberalization of rural markets, technical e�ciency improved very little

and allocative e�ciency has increased only slightly. Copyright # 2000 John Wiley &

Sons, Ltd.

1 INTRODUCTION

Institutional changes and market reforms initiated in 1979 have had great impactson China's agricultural production and productivity growth. Numerous studies haveexamined the e�ects of these reforms (Lin, 1992; Fan, 1991; Wen, 1993), but all usedthe production function approach. Production functions are easy to estimate, and canbe used to identify the e�ects on production growth of technological change derivedfrom the use of new technologies and technical e�ciency improvement due to institu-tional and market reforms. But the production function approach cannot measure theimpact of improvement in allocative e�ciency due to these changes and reforms. Thishas become an increasingly serious problem because allocative e�ciency improvementmay have been a more important component of overall e�ciency improvement inChinese agriculture since 1984. The government did not begin to focus on the reform

CCC 0954±1748/2000/010001±12$17.50Copyright # 2000 John Wiley & Sons, Ltd.

Journal of International DevelopmentJ. Int. Dev. 12, 1±12 (2000)

* Correspondence to: Mr Shenggen Fan, International Food Policy Research Institute, 2033 K Street NW,Washington, DC 20006, USA.

Page 2: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

of the rural input and output market system (the so-called second-phase reforms) until1985.1 It is anticipated that the second-phase reforms may have improved farmers'allocative e�ciency signi®cantly since 1985, but it is not known if or to what extentfarmers have improved their allocative e�ciency, in addition to their improvement intechnology and technical e�ciency. The objective of this study is to ®ll this gap in ourknowledge.

This study di�ers from the previous studies in several aspects. First, the studyempirically estimates a stochastic frontier shadow cost function for Chinese agri-culture using a ¯exible functional form. One of the problems in estimating the costfunction for a centrally planned economy such as China is government distortion inboth input and output markets. By using this new approach, the e�ects of thesegovernment distortions on allocative e�ciency can be estimated. Second, techno-logical change and technical and allocative e�ciency improvement are estimatedsimultaneously from the same cost function. Traditionally, technological change andtechnical e�ciency are estimated based on the assumption of allocative e�ciency, butthis assumption may not be realistic, and may result in biased estimates. Third, thestudy covers a longer period (1980±93), making it possible to identify the e�ects ofdi�erent phases of the reforms on e�ciency improvement.

The paper is organized as follows: in the next section, a framework to measureboth technical and allocative e�ciency plus technological change is developed, usingthe cost function framework. Section 3 describes the model speci®cation, whileSection 4 describes the data and estimation procedures. In Section 5, estimated resultsand measures of technological change and e�ciency improvement are presented.Conclusions are drawn in the ®nal section.

2 CONCEPTUAL FRAMEWORK

Technological change and e�ciency improvement are important sources of produc-tion growth in any economy. Technological change is de®ned as a shift in the frontierproduction function. E�ciency improvement can be further decomposed into tech-nical and allocative e�ciency. The concept of technical e�ciency is based on inputand output relationships. Technical ine�ciency arises when actual or observed out-put from a given input mix is less than the maximum possible. Technical e�ciencyimprovement by a farmer is measured by a reduction in the di�erence between hisoutput and the maximum output possible, given the input mix. Allocative ine�ciencyarises when the input mix is not consistent with cost minimization. Allocative ine�-ciency occurs when farmers do not equalize marginal returns with true factor marketprices due to government intervention in factor markets.

Di�erent concepts of e�ciency and technological change can be illustrated usingFigure 1, with two inputs, X1 and X2 , and a single product, Y. The two isoquants F1

and F2 represent production frontiers for the same physical output at time 1 and time 2,respectively. They are the best-practised technologies by farmers, however, a certainproducer may not reach the frontier because of his technical ine�ciency. Therefore,A1 , B1 ,A2 and B2 are technically e�cient, but C1 andC2 are not. The price of input X1

relative to X2 is represented by P1 and P2 for the two time periods, respectively.

1 For more information on input and out market reforms, see Lin, 1989; 1991.

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

2 S. Fan

Page 3: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

Allocative e�ciency occurs if the inputs are combined so that their marginal productsare in the same ratios as their relative prices. Thus, A1 and A2 are allocatively e�cient,but B1 and B2 are not. A1 and A2 are both technically and allocatively e�cient becausethey are on the production frontiers and are located where the ratios of marginalproducts are the same as the ratios of relative prices.

In the cost function framework, technological change can be measured asÿ�C�Xa2

1 ; Xa22 � ÿ C�Xa1

1 ; Xa12 �=C�Xa1

1 ; Xa12 �; technical e�ciency can be measured as

C�Xb11 ; Xb1

2 �=C�Xc11 ; Xc1

2 � and time 1, and C�Xb21 ; Xb2

2 �=C�Xc21 ; Xc2

2 � at time 2,respectively; and allocative e�ciency can be measured as C�Xa1

1 ; Xa12 �=C�Xb1

1 ; Xb12 �

at time 1, and C�Xa21 ; Xa2

2 �=C�Xb21 ; Xb2

2 � at time 2, respectively. Economic e�ciency isthe product of technical e�ciency and allocative e�ciency. Therefore, economice�ciency is C�Xa1

1 ; Xa12 �=C�Xc1

1 ; Xc12 � at time 1, and C�Xa2

1 ; Xa22 =C�Xc2

1 ;Xc22 � at time 2,

respectively. The model speci®ed in the next section is based on this conceptualframework.

3 MODEL SPECIFICATION

Economic theory argues that reducing government control on the factor market is ane�cient policy because the free market best determines prices according to the relativescarcity of resources. Fan (1991) developed a frontier production function approachto measure both technological change and technical e�ciency improvement onChina's agricultural production growth from 1965 to 1985. However, the productionfunction approach cannot model the e�ect of allocative e�ciency improvement onproduction growth. Wang et al. (1996) developed a frontier shadow pro®t function tostudy both technical and allocative e�ciency using cross-sectional household surveydata. But changes of e�ciency and technology over time were impossible to modeldue to unavailability of panel data. Furthermore, di�erent functions are used inestimating technical and allocative e�ciency in Wang et al.'s study. In this paper,a stochastic frontier shadow cost function approach is developed and applied to

Figure 1.

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

Technological Change in Chinese Agriculture 3

Page 4: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

Chinese agriculture to measure three di�erent e�ects on production: technologicalchange, technical and allocative e�ciency improvement.

It is assumed that farmers minimize cost subject to their technology and govern-ment policy constraints. Due to government regulations and input supply shortages,there exist shadow prices for inputs. Farmers make their decisions on shadow pricesof factor inputs in order to minimize shadow costs. In a free-market situation, theshadow price is equal to the market price. This condition, however, does not hold inChinese agriculture because of government distortions on factor markets, such asprice controls, rationing of certain inputs, and control of labour migration. As thegovernment gradually eliminates controls on the input markets, the di�erencebetween the shadow price and the observed price should diminish, therefore, farmers'allocative e�ciency should improve as a result.

To model farmers' decision making, it is assumed that the decision makers mini-mize a well-de®ned but unobserved shadow cost function subject to unobservedshadow input prices:

Cs � C

s�Y;Ps�; �1�where Y is output, and Ps is a vector of shadow input prices. If the farmer is e�cient inchoosing the cost-minimizing levels of the inputs, then

@f�X�=@Xi

@f�X�=@Xj

� P si

P sj

�2�

where f(X) is neoclassical production function common to all ®rms. We de®ne xi asthe ratio of the shadow price to the observed price for input i,

xi � Psi =Pi: �3�

If there exists a well-developed input market without government distortions, thenxi � 1. But if xi di�ers from 1, then factor input markets are distorted by policies,causing ine�cient use of resources.

Allocative e�ciency can be measured from a shadow pro®t or a shadowcost function. The shadow pro®t function framework was developed by Lau andYotopoulous (1971), and then extended by Atkinson and Halvorsen (1980) andLovell and Sickles (1983). The shadow cost function approach was ®rst introduced byToda (1976) and later applied by Atkinson and Halvorsen (1984). Recently, someeconomists have used the shadow cost function approach to measure both technicaland allocative e�ciency. Atkinson and Cornwell (1994) developed a shadow costfunction approach to identify both technical e�ciency and allocative e�ciency of themajor U.S. airlines. Parker (1995) used a similar approach in his study on China'sconstruction industry. But the cost functions they estimated are deterministic.2 One ofthe primary criticisms of the deterministic estimators is that no account is taken of thepossible in¯uence of measurement errors and other noise upon the shape andpositioning of the estimated frontier, since all observed deviations from the estimatedfrontier are assumed to be the results of technical ine�ciency (Coelli, 1995; Bauer,1990).

2 Atkinson and Cornwell (1994) normalize e�ciency measure to 1 for the most e�cient ®rm.

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

4 S. Fan

Page 5: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

To avoid this problem, a stochastic frontier shadow cost function is developed andused in this study. It is assumed that decision makers minimize a well-de®ned butunobserved shadow cost function subject to unobserved shadow input prices. Insteadof all inputs, we assume that fertilizer, machinery, labour and pesticides are variable.3

The shadow cost function C s is assumed to be a function of shadow prices Psi

(i � 1,2,3, and 4), and is speci®ed as the translog functional form which providesa convenient second-order approximation to any arbitrary continuously twice-di�erentiable cost function.

lnCs � A0 � atT � 1=2attT

2 � ay ln Y � 1=2ayy ln Y ln Y

� aytT � ln Y �X4i�1

ai ln Psi �

X4i�1

aiy ln Y ln Psi

� 1=2X4i�1

X4j�1

aij lnPsi ln P

sj �

X4i�1

aitT � ln Psi � e �4�

where A0 is a set of regional speci®c dummies; and T is time trend. The disturbanceterm, e � V � U, is assumed to be consistent with the frontier concept. We assume Vto be normally distributed to re¯ect the random factors such as weather, and use aone-sided disturbance U to represent the technical ine�ciency component,4 i.e. V isnormal with mean zero and variance s2v , j u j is truncated normal with variance s2u,and E(uitvit) � 0.

Since cost functions must be linearly homogenous and concave in factor prices, thefollowing restrictions are imposed:

X4i�1

ai � 1;X4i�1

aij � 0 8j;

X4i�1

ait � 0;X4i�1

aiy � 0: �5�

Symmetry is also imposed on the aij parameters. By Shephard's Lemma, the shadowshare of the ith input is the ®rst partial derivative of lnCs with respect to lnPs

i ,

Ssi �

@ lnCs

@ ln Psi

� ai �X4j�1

aij ln Psj � aiy ln Y � aitT � ei: �6�

The error terms ei in the shadow share equation represent statistical noises, such thatE(ei) � 0, and E(eiei) � sijI. It is also assumed that e and ei are independent of eachother.

The shadow cost function in equation (4) cannot be directly estimated, however,since neither shadow prices nor shadow costs are directly observed. Instead, inputprices, Pi, are observed, and it is assumed that, the ratio of shadow price to observed

3 The cost function is estimated on a per mu basis, assuming constant return to scale. One mu is equal to1/15 hectare.4 Note that only the technical ine�ciency component rather than overall cost ine�ciency is captured in thisterm, because the shadow cost function is estimated as the one for the most allocatively e�cient ®rm.

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

Technological Change in Chinese Agriculture 5

Page 6: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

price, xi , is an estimable function of other variables. We assume xi is a function ofland-to-labour ratio (ALR), and the ratio of government procurement prices tomarket prices of rice output (GMR). The land-to-labour ratio variable captures thee�ect on e�ciency of relaxation of migration control policy and labour movement dueto the rapid development of rural industry during the reform period, while the ratio ofgovernment to market prices capture the e�ect of changes in government pricepolicies. A regional dummy PD (�1 for southern prefectures of Jiangsu, and 0 fornorthern prefectures) is also added to capture the regional di�erence in shadowprices.5

ln xi � xi0 � xi1ln ALR � xi2 ln GMR � xi3PD: �7�From equations (4) and (6), the observed cost function C can be derived as a functionof the shadow (minimum) cost function and the shadow shares,6 such that

ln C � ln CS � ln

X4i�1

SSi

xi

!� e �8�

and the resulting observed share equations Si are:

Si �@ ln C

@ ln Pi

� SSi

xi

X4k�1

SSj

xk

!ÿ1�ei: �9�

Equations (8) and (9) can be estimated if the error terms of these two equations areassumed to be seemingly unrelated.7 Because the four observed shares sum to unity byde®nition, the last share equation Sf is dropped. The results are invariant to the shareequation dropped.

E�ciency indexes and the rate of technological change can be calculated from theestimated stochastic frontier shadow cost function. Allocative e�ciency, the ratio ofshadow cost to actual cost, is measured as:

ZA ��CS

�C�10�

This index is unity if xi � 1 8i.Based on the conditional distribution of U, given the distribution of V � U, the

technical e�ciency can be measured as:8

ZT � 1 ÿ E�U j e� � 1 ÿ sUsVs

f�el=s�1 ÿ F�el=s� ÿ

els

� ��11�

5 Due to linear homogeneity in prices, only relative price ine�ciency can be estimated; therefore, xf( fertilizer) is arbitrarily set to unity, and xl (labour), xp (pesticide), and xm (machinery) are relative shadowprice ratios.6 For derivation of equations (8) and (9), see Atkinson and Cornwell (1994) among others.7 Wang et al. (1996) and Parker (1995) have also proved that equations (8) and (9) can be estimated.8 For detailed derivation of (12), see Jondrow et al. (1982); Ali et al. (1996) also used this measure in a costfunction framework.

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

6 S. Fan

Page 7: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

where l � sU/sV and s2 � s2U � s2V; and f and F are the standard normal densityfunction and the standard normal distribution function, respectively evaluated at(el/s).

Overall economic e�ciency is the product of both technical and allocativee�ciency.

Z � ZAZT �12�Using the estimated stochastic frontier shadow cost function, the rate of technologicalchange can be derived as:

ÿ @ ln Cs

@t� ÿ�at � att*T �

X4i�1

ait ln PSi � ayt ln Y� �13�

4 DATA DESCRIPTION AND ESTIMATION PROCEDURES

Time-series cross-prefecture data are used in this study. They come from the Cost andProduction Survey conducted by the Jiangsu Provincial Price Bureau. JiangsuProvince is one of the most developed areas in terms of both agriculture and non-agriculture. The survey covers most of the major crops in the province and in 1993more than 900 households were surveyed. The well-trained survey team members atthe village, township, county, prefectural and provincial levels recorded quantitiesand costs of major inputs, and production for each commodity periodically everyyear. Only rice is chosen for the purpose of this study because of its signi®cance andrepresentativeness in the province's agricultural production. For this analysis,aggregated cell means of the prefectural households were used rather than individualhousehold observations. Fourteen years, 1980 to 1993, and 11 prefectures (Nanjing,Wuxi, Suzhou, Changzhou, Xuzhou, Lianyungang, Huaiyin, Yianchang, Nanton,Yangzhou and Zhengjiang) were included (a total of 154 observations). Input priceswere taken from the China's Price Yearbooks and Jiangsu Provincial Price Bureau.

Since the ®nal objective is to estimate the stochastic frontier shadow cost function(equation (4)) an iteration procedure is used in the estimation:9

(i) Both equations (8) and (9) are estimated as a system using the nonlinear seem-ingly unrelated regression technique (SURE). The error terms in these equationsare assumed to be seemingly unrelated with e assumed as traditional normaldistribution with zero mean and variance of s2e .

(ii) The second step involves the estimation of equation (4) with the error termsconsistent with the frontier framework. Both predicted shadow costs and pricesare used as dependent and independent variables. The predicted value of lnCs

is calculated from equation (8) using estimated xis from step 1. Similarly, thepredicted value of Ps

t is calculated from equation (3).

9 Ideally, we can estimate equations (8) and (9) with (8) speci®ed as a stochastic frontier function and (9) asseemingly unrelated. But there are no econometric packages available for estimating a frontier function in asystem.

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

Technological Change in Chinese Agriculture 7

Page 8: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

(iii) Using estimated parameters as from equation (4) to reestimate share equations(9) in order to obtain new estimates of xis.

(iv) Using newly estimated xis from step 3, re-estimate equation (4), and reiterate thisprocess until as converge.

5 RESULTS

The estimates of the convergent cost function for Jiangsu rice production is presentedin Table 1. Most of the estimates are statistically signi®cant at the 5 per cent level. TheWald test rejected that xi � 1 for labour, machinery and pesticide inputs at the oneper cent signi®cance level, indicating that the shadow prices of these inputs aresigni®cantly di�erent from the observed prices. The value of l is 1.772, which impliesthat one sided error term U dominates the symmetric error V.

In Table 2, mean statistics of the estimated stochastic frontier shadow cost functionare shown for two periods, 1980±84 and 1985±93, and for the whole period, 1980±93.These statistics are calculated for each observation and reported at their mean. Thesestatistics include estimated shadow factor price ratios, and observed cost shares as wellas estimated shadow shares. The changes in the ratios of shadow prices to observedprices indicate that the observed price ratios of labour, machinery and pesticides havemoved closer towards unity. Comparing the shadow and actual cost shares of inputsreveals that farmers have overused the labour input by more than 87 per cent duringthe ®rst phase reforms. Even during the second phase reforms from 1985 to 1993, theactual labour cost share was still more than 68 per cent higher than the shadow costshare. This con®rms that there still exists a large surplus of labour in rural China. The

Table 1. Estimates of stochastic frontier shadow cost function.

Parameter Coe�cient t-ratio Parameter Coe�cient t-ratio

at ÿ0.0353 1.989* x10 ÿ0.8355 ÿ2.873*att ÿ0.0018 ÿ0.483 x11 0.4451 4.096*

ay ÿ0.6163 ÿ1.889* x12 ÿ0.0233 ÿ0.308ayy 0.3571 0.345 x13 ÿ0.3431 ÿ3.285*ayt 0.0334 1.066 xm0 1.3686 4.448*

al 0.3324 6.841* xm1 ÿ0.3938 ÿ3.416*am 0.0672 2.894* xm2 ÿ0.1982 ÿ3.581ap 0.1169 2.824* xm3 ÿ1.9990 ÿ5.116*ayl 0.0451 1.894* xp0 ÿ1.6672 ÿ2.823*aym ÿ0.0253 ÿ1.315 xp1 ÿ0.3031 ÿ2.481*ayp 0.0032 0.487 xp2 ÿ0.1759 ÿ2.362*atl ÿ0.0041 ÿ1.723 xp3 1.8990 5.599*

atm 0.0047 3.508* l 1.7721 6.226*

atp ÿ0.0014 ÿ1.805* s2U 0.0201 2.967*

all 0.0135 1.876* s2V 0.0042 8.401*

alm ÿ0.0108 ÿ2.343*alp 0.0021 0.917

amm 0.0616 3.751*

amp ÿ0.0483 ÿ3.567*app 0.0173 1.736

Notes: The subscripts l, m, p, t, and y stand for labour, machinery, pesticide, the time trend, and output.Asterisk indicates signi®cant at the 5 per cent level. Regional dummies are not reported.

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

8 S. Fan

Page 9: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

shadow cost share of fertilizer was more than 60 per cent of the observed fertilizershare, implying that farmers have underused fertilizer relative to labour. However, thediscrepancy between the shadow and actual shares of pesticide and machinery hasbeen very small.

Measures of technical and allocative e�ciency, and the rate of technological changeare presented in Table 3. Technical e�ciency was relatively low at the beginning of thereforms. But as reforms progressed, it improved signi®cantly, by 8.5 per cent per yearfrom 1980 to 1984. But after 1985, technical e�ciency improved very little. Theseresults con®rmed many speculations that the e�ect of household responsibility systemhas been largely exhausted after 1985. During the ®rst phase of reforms, allocativee�ciency improved very little. Even during the second phase of reforms, it increased

Table 2. Mean statistics of the stochastic frontier shadow cost function estimation.

1980±84 1985±93 1980±93

xf 1.00 1.00 1.00

xm 2.11 2.02 2.06

xp 0.62 0.63 0.63

xl 0.46 0.47 0.60

Sf 0.31 0.32 0.32

Ssf 0.48 0.54 0.52

Sm 0.04 0.06 0.05

Ssm 0.13 0.08 0.10

Sp 0.06 0.07 0.07

Ssp 0.08 0.09 0.08

Sl 0.59 0.49 0.52

Ssl 0.31 0.29 0.30

Notes: Sf , Sm, Sp , and Sl are actual factor shares of fertilizer, machinery, pesticides, and labour, whileSsf ; Ss

m; Ssp; and Ss

l are shadow factor shares for these inputs.

Table 3. Measures of e�ciency improvement and technological change.

Year Allocativee�ciency

Technicale�ciency

Economice�ciency

Technologicalchange rate

1980 0.688 0.611 0.420 0.026

1981 0.689 0.612 0.421 0.032

1982 0.700 0.696 0.488 0.037

1983 0.700 0.796 0.557 0.040

1984 0.704 0.848 0.597 0.043

1985 0.695 0.819 0.568 0.043

1986 0.715 0.852 0.609 0.045

1987 0.719 0.839 0.603 0.046

1988 0.707 0.839 0.593 0.047

1989 0.702 0.878 0.616 0.051

1990 0.701 0.864 0.606 0.054

1991 0.731 0.859 0.628 0.058

1992 0.725 0.854 0.619 0.059

1993 0.745 0.909 0.677 0.062

Annual Growth Rate (%)

1980±84 0.593 8.533 9.177 13.275

1985±93 0.779 1.170 1.958 4.235

1980±93 0.614 3.102 3.734 6.980

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

Technological Change in Chinese Agriculture 9

Page 10: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

only slightly. Therefore, we can not con®rm many scholars' belief that the secondphase of reforms have improved allocative e�ciency substantially. This may indicatethat the input market in Jiangsu agriculture is still distorted by various policies inmany aspects, despite the e�orts that the government has made to liberalize themarket. Economic e�ciency, the combined e�ects of both technical and allocativee�ciency, improved sharply from 1980 to 1984, but began to stagnate after 1984, dueto the lack of improvement in both technical and allocative e�ciency.

The rate of technological change accelerated during the ®rst phase reform, with anannual growth of 13.3 per cent per annum. This is a result of long-term investment inagricultural research, irrigation and other technologies in Jiangsu agriculture. In fact,the province has one of the most productive research systems among all provinces.More than 80 per cent of total crop areas are sown with high-yielding varietiesdeveloped by the provincial as well as prefectural research institutes. But in recentyears, the rate of technological change has slowed down to 4.3 per cent per annum.

Allocative e�ciency varies greatly among regions (Table 4). The northern part ofthe province (Xuzhou, Nanton, Yangcheng and Huaiyin) has higher allocativee�ciency. The lower allocative e�ciency in southern Jiangsu may be due to the heavysubsidies of rural industry to rice production there, these subsidies may have severedmisallocation of inputs. The higher allocative e�ciency in the relatively poor Northalso con®rms T.W. Schultz's hypothesis that small farmers are e�cient in allocatingtheir resources although they are poor. In contrast to allocative e�ciency, technicale�ciency has relatively small regional variations, ranging from 0.76 in Nanton to0.85 in Lianuanggang. Therefore, it is not surprising that the variation in economice�ciency largely comes from the di�erence in allocative e�ciency.

6 CONCLUSIONS

Rural reforms since 1979 have been very successful, according to a number of recentstudies. But they all failed to measure the changes in allocative e�ciency due to thesereforms. This study developed and used a stochastic frontier shadow cost functionapproach to estimate the improvement of both technical and allocative e�ciency. The

Table 4. Regional di�erence in e�ciency level.

Allocative e�ciency Technical e�ciency Economic e�ciency

Nanjing 0.64 0.78 0.49

Wuxi 0.65 0.80 0.52

Xuzhou 0.78 0.80 0.63

Changzhou 0.64 0.84 0.54

Suzhou 0.68 0.78 0.53

Nanton 0.80 0.76 0.61

Lianyungang 0.72 0.85 0.61

Huaiyin 0.76 0.78 0.59

Yancheng 0.78 0.78 0.61

Yangzhou 0.69 0.79 0.54

Zhengjiang 0.66 0.84 0.55

Average 0.71 0.80 0.57

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

10 S. Fan

Page 11: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

analysis con®rms the ®ndings of previous studies on the important growth-promotinge�ects of e�ciency improvement of the rural reforms. More importantly, the import-ance of e�ciency improvement in promoting production growth from the reformsvaried markedly over time and across regions. Technical e�ciency improved substan-tially in the beginning of the reforms, while improvement of both technical andallocative e�ciency stagnated during the second phase of reforms. Overall economice�ciency has improved substantially, but the rate has declined after 1984.

These results have important policy implications for the government to increasefurther production and productivity growth. The large regional variation in allocativee�ciency among regions implies that China still has great potential to promoteproduction growth by reducing regional di�erences in allocative e�ciency. Thestagnation in technical e�ciency after 1984 may be a result of deterioration of theextension service after the reforms. Therefore, the extension system also needs to bestrengthened in order to gain further e�ciency in production.

ACKNOWLEDGEMENTS

The author is grateful for the support from IFPRI, the University of Arkansas,Nanjing Agricultural University, and the Chinese Academy of Agricultural Sciences.Partial funding for this study from the Australian Centre for International Agri-cultural Research is also acknowledged.

REFERENCES

Ali, F., Parihk, A. and Shah, M. (1996). `Measurement of economic e�ciency using the

behavioral and stochastic cost frontier approach', Journal of Policy Modeling, 18, 271±287.

Atkinson, S. and Cornwell, C. (1994). `Parametric estimation of technical and allocative

ine�ciency with panel data', International Economic Review, 35, 231±243.

Atkinson, S. and Halvorsen, R. (1980). `A test of relative and absolute price e�ciency in

regulated utilities', Review of Economics and Statistics, 62, 81±88.

Atkinson, S. andHalvorsen, R. (1984). `Parametric e�ciency tests, economies of scale, and input

demand in U.S. electric power generation', International Economics Review, 25, 647±662.

Bauer, P. W. (1990). `Recent developments in the econometric estimation of frontiers', Journal

of Econometrics, 46, 39±56.

Coelli, T. J. (1995). `Recent developments in frontier modelling and e�ciency measurement',

Australia Journal of Agricultural Economics, 39, 219±245.

Editorial Committee of China's Agricultural Yearbooks (1980±94). China's Agricultural

Yearbook (various issues). Beijing: China's Agricultural Press (in Chinese).

Fan, S. and Ruttan, V. (1992). `Technical change in centrally planned economies', Agricultural

Economics, 6, 301±314.

Fan, S. (1991). `E�ects of technological change and institutional reform on production growth

in Chinese agriculture', American Journal of Agricultural Economics, 73, 266±275.

Farrell, M. J. (1957). `The measurement of productive e�ciency', Journal of Royal Statistics

Society, Ser. A., 120, 253±281.

Jondrow, J., Lovell, C., Materov, I. S. and Schmidt, P. (1982). `On estimation of technical

ine�ciency in the stochastic frontier production function model', Journal of Econometrics,

19, 233±238.

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

Technological Change in Chinese Agriculture 11

Page 12: Technological change, technical and allocative efficiency in Chinese agriculture: the case of rice production in Jiangsu

Lau, L. J. and Yotopoulos, P. A. (1971). `A test for relative e�ciency and an application to

Indian agriculture', American Economic Review, 61, 94±109.

Lin, J. (1989). `Rural factor markets in China after the household responsibility reform'.

In Reynolds, B. (ed.), Chinese Economic Policy. New York: Wiley.

Lin, J. (1991). `Prohibition of factor market exchanges and technological choice in Chinese

agriculture', The Journal of Development Studies, 27, 1±15.

Lin, J. (1992). `Rural reforms and agricultural growth in China', American Economic Review,

82, 34±51.

Lovell, C. A. K. and Sickles, R. (1983). `Testing e�ciency hypotheses in joint production: a

parametric approach', Review of Statistics and Economics, 65, 51±58.

McMillan, J., Whalley, J. and Zhu, L. (1989). `The impact of China's economic reforms on

agricultural productivity growth', Journal of Political Economy 97, 781±807.

Parker, E. (1995). `Shadow factor price convergence and the response of Chinese state-owned

construction enterprises to reform', Journal of Comparative Economics, 21, 54±81.

Sicular, T. (1988). `Plan and market in China's agricultural commerce', Journal of Political

Economy, 92, 283±307.

Toda, Y. (1976). `Estimation of a cost function when the cost is not minimum: the case of

Soviet manufacturing industries, 1958±71', Review of Economics and Statistics, 58, 259±268.

Wang, J., Wailes, E. and Cramer, G. (1996). `A shadow-price frontier measurement of pro®t

e�ciency in Chinese agriculture', American Journal of Agricultural Economics, 78, 146±156.

Wen, G. J. (1993). `Total factor productivity change in China's farming sector: 1952±1989',

Economic Development and Cultural Change, 42, 1±41.

Copyright # 2000 John Wiley & Sons, Ltd. J. Int. Dev. 12, 1±12 (2000)

12 S. Fan