Subsidising inequality: Economic reforms, fiscal transfers and convergence across Chinese provinces

This article was downloaded by: [Universite De Paris 1]On: 04 September 2013, At: 00:46Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number:1072954 Registered office: Mortimer House, 37-41 Mortimer Street,London W1T 3JH, UK

The Journal of DevelopmentStudiesPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/fjds20

Subsidising inequality:Economic reforms, fiscaltransfers and convergenceacross Chinese provincesMartin Raiser aa Chief Economists Office, European Bank forReconstruction and Development, LondonPublished online: 23 Nov 2007.

To cite this article: Martin Raiser (1998) Subsidising inequality: Economicreforms, fiscal transfers and convergence across Chinese provinces, The Journalof Development Studies, 34:3, 1-26, DOI: 10.1080/00220389808422518

To link to this article: http://dx.doi.org/10.1080/00220389808422518

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of allthe information (the “Content”) contained in the publications on ourplatform. However, Taylor & Francis, our agents, and our licensorsmake no representations or warranties whatsoever as to the accuracy,completeness, or suitability for any purpose of the Content. Anyopinions and views expressed in this publication are the opinionsand views of the authors, and are not the views of or endorsed byTaylor & Francis. The accuracy of the Content should not be reliedupon and should be independently verified with primary sources ofinformation. Taylor and Francis shall not be liable for any losses, actions,

http://www.tandfonline.com/loi/fjds20

http://www.tandfonline.com/action/showCitFormats?doi=10.1080/00220389808422518

http://dx.doi.org/10.1080/00220389808422518

claims, proceedings, demands, costs, expenses, damages, and otherliabilities whatsoever or howsoever caused arising directly or indirectlyin connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private studypurposes. Any substantial or systematic reproduction, redistribution,reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of accessand use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/page/terms-and-conditions

Subsidising Inequality:Economic Reforms, Fiscal Transfers andConvergence Across Chinese Provinces

MARTIN RAISER

The article investigates per capita income convergence acrossChinese provinces over the 1978-92 period. It confirms previousstudies which find a reduction in inter-regional income inequalityover the course of economic reforms. However, the rate ofconvergence has declined since 1985 as a result of two factors.First, the shift from rural to industrial reforms hasdisproportionately benefited the relatively wealthier coastalprovinces. Second, the system of inter-provincial fiscal transfershas prevented convergence among interior provinces, as transfershave gone to the richer among them. Further fiscaldecentralisation and an acceleration of reforms in the interiorprovinces is thus unlikely to increase regional income inequality.

I. INTRODUCTION

In a vast country such as China, the regional impact of economic reformsnaturally commands high policy interest. As regions may differ accordingto initial conditions such as economic structure, resource endowments, andlevel of development, the outcome of China's transformation processtowards the market is expected to be regionally differentiated. Moreover,one element of the reform process in China has been the reduction of inter-provincial fiscal transfers from the coastal area to the interior. Indeed, in

Martin Raiser, Chief Economists Office, European Bank for Reconstruction and Development,London. This article was written while the author was a research fellow at the Kiel Institute ofWorld Economics, Germany. The views expressed are the sole responsibility of the author and inno way reflect the opinion of either the Kiel Institute or the EBRD. The author thanks ErichGundlach, Peter Nunnenkamp and Rolf J. Langhammer for extensive discussion and twoanonymous referees for comments on an earlier draft. Financial support from the VolkswagenFoundation for the Project 'Decentralisation and Enterprise Reform in China' is gratefullyacknowledged.

The Journal of Development Studies, Vol.34, No.3, February 1998, pp.1-26PUBLISHED BY FRANK CASS, LONDON

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

2 THE JOURNAL OF DEVELOPMENT STUDIES

recent years, there has been a growing concern that a widening income gapbetween the prosperous coast and the laggard interior might eventuallycause the break up of the Chinese national state. Such concerns (typicallyassociated with a call for more inter-provincial redistribution) run counterto the theoretical expectation that regional income convergence shouldaccelerate in a country implementing economic reforms as factor marketsimprove and capital starts flowing from richer to poorer areas.

This article presents results that overall confirm the theoreticalexpectation that reforms have led to the convergence of per capita incomesacross Chinese provinces. However, this process has been uneven bothacross time and across regions of China. The article shows that convergencewas more rapid during the early reform phase than after the mid 1980s. Thisrecent slow-down in convergence is due to a combination of two factors; awidening average income gap between the coast and the interior and thepartial reversal of convergence among interior provinces after 1985. Whilefaster growth along the Chinese coast can be explained as a result offavourable structural conditions and rising investment rates, the lack ofincome convergence among interior provinces after 1985 may at least partlybe attributed to the perverse effects of inter-provincial transfers, causingcapital to flow to rich rather than poor provinces. While data imperfectionslimit the policy conclusions that may be drawn from the present analysis,the unexpected impact of fiscal transfers on convergence suggests thatfurther fiscal decentralisation is unlikely to harm the poorest interiorprovinces.

The results of this article are related to a growing body of empiricalliterature on provincial economic development in China during the reformperiod. Hsueh, Rawski and Tsui [J994] and Wu [1995] look at regionalefficiency improvements in state industry, and state industry only, ruralindustry and agriculture respectively. The first source finds a convergenceof efficiency levels in state industry across provinces, indicating thattechnologically backward provinces have moved closer to the productionfrontier with economic reforms. The results in Wu [1995] are moreambiguous, revealing that the convergence of efficiency levels in stateindustry has been confined to the coastal area. In rural industry (which islargely free of state interference) efficiency levels have tended to convergethroughout China, while in agriculture they have diverged. The averageweighted increase in Total Factor Productivity (TFP) has been higher alongthe coast than in the interior, a result that will be supported below. Jian,Sachs and Warner [1996] and Gundlach [1996a] study the convergence ofGross Domestic Product (GDP) per worker across Chinese provinces overthe 1952-1993 and 1978-89 periods respectively. Both find a convergencerate of around two per cent per annum for the whole reform period. Jian,

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

ECONOMIC REFORMS . . . AND CONVERGENCE: CHINA 3

Sachs and Warner also show that there was no convergence before the startof reforms in 1978 and that the trend towards income convergence duringthe reform period had stopped by the early 1990s. These results are inaccordance with the present study, although the interpretation differs fromthe one given in this paper, as discussed below.

The article is structured as follows. Section II presents the evidence onregional income convergence in China since 1978. It identifies a slow-downin the rate of convergence for the 1985-92 period and discusses twotheoretical possibilities that could account for this finding. Section IIIexplains the better post 1985 growth performance along the coastalprovinces of China in terms of an increase in investment and higher TFPgrowth. It then estimates convergence rates conditional on investment ratesand structural conditions that might be related to higher TFP growth andfinds evidence for conditional convergence after 1985. Section IV turns tofiscal transfers and their impact on capital flows and investment rates acrossChinese provinces to account for the lack of income convergence amonginterior provinces. Section V concludes and draws some policyimplications.

II. CONVERGENCE IN CHINA SINCE 1978

(1) Measures of Convergence

There are two ways to measure the convergence of per capita incomes. Thefirst measure computes the coefficient of variation of per capita or perworker incomes. Thereby, GDP is typically taken as a measure of income,abstracting from transfers from abroad. If the coefficient of variation of percapita or per worker incomes falls over time in a given cross-section ofprovinces or countries, sigma convergence is said to obtain [Sala-i-Martin,1995].' The second measure of convergence is obtained by regressing thegrowth rate of per worker incomes against a constant and its initial level asdefined by equation (1). This yields an estimate of beta convergence.

\nYt - In7o = A - (l - (1)

Yt - income per worker at time tYO - initial income per workerA - regression constant equal to ( l -

with Y* = steady incomeP - convergence rate per annum/ — time index.

Equation (1) is derived from a standard neo-classical growth model. In

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


the recent reformulation of this model by Mankiw, Romer and Weil [1992],the level of income is described by a Cobb Douglas production function ofthe form Y=F(K,HK,L,t), where K, HK, L stand for physical capital, humancapital and labour inputs respectively and t represents time. Constantproportions of income Y (sk and sh) are saved and invested in physical andhuman capital and a constant rate of depreciation applies to both types ofcapital. Under these assumptions, the ratio of capital and human capital perworker and with it real income per worker grow at decreasing rates, until theeconomy reaches the so-called 'steady state'. In steady state, the ratios ofphysical and human capital to labour are constant and real income grows atthe (exogenous) rate of technological progress, g. The level of real incomeper worker in steady state depends positively on the savings rates inphysical and human capital and is negatively related to the rate ofdepreciation and the rate of population growth.2 The adjustment path of realincome per worker towards the steady state level is obtained by a linearTaylor approximation of steady state income and leads to the relationship in

(1) [Mankiw, Romer and Weil, 1992]?The sigma and beta measures of convergence are related to each other in

that the existence of beta convergence is a necessary condition for sigmaconvergence. However, it is not a sufficient condition. Because the cross-country evidence for sigma convergence is weak, recent contributions to theneo-classical growth literature have introduced the concept of conditionalconvergence, whereby the estimate of beta in (1) is conditioned by variablesthat control for differences in steady state incomes across countries oreconomic regions. Conditional beta convergence may obtain even in theabsence of sigma convergence [Sala-i-Martin, 1995]. In other words, whilesigma convergence measures the absolute dispersion of incomes, betaconvergence conveys information on the catch-up potential of pooreconomies.

(2) The Data

Researchers working on regional economic development in China in principlecan draw on both national and a set of provincial statistics, giving more or lesscomplete coverage of all provinces and autonomous regions from around1978 until 1995. However, complete volumes of regional statistical yearbooksare not easy to obtain outside China. Moreover, data collection methodologiesmay differ and issues of comparability and consistency across provinces arise[Herrmann-Pillath, 1995; Jian, Sachs and Warner, 1996]. The shortcomingsof the original sources have prompted a number of projects aimed at thecompilation of easily accessible and reasonably consistent cross-provincialdata sets. In computing the two convergence measures introduced above, Idraw on two of these secondary sources in particular.4

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

ECONOMIC REFORMS . . . A N D CONVERGENCE: CHINA 5

Hsueh, Li and Liu [1993] have provided a comprehensive set ofprovincial statistics drawn from the provincial statistical yearbooks andcovering the period 1949-89. The authors have attempted to use aconsistent methodology in compiling these data, thereby overcoming someof the potential problems of comparability. The disadvantage of this firstsource (referred to as Data 1 below) is that the time period starting with thereforms in 1978 is thereby shortened to only 11 years and more recentdevelopments are not captured. Moreover, as Jian, Sachs and Warner [1996]point out, the data in Hsueh et al. [1993] become increasingly unreliabletowards the end of the 1980s. These authors thus complement the firstsource with data from national sources (the China Statistical Yearbook andthe China Price Statistics Yearbook) in addition to more recent provincialstatistics for the second half of the 1980s and the early 1990s.

While I have not been able to obtain Jian, Sachs and Warner's data-set,Herrmann-Pillath [1995] offers a recent compilation of basic nationalincome and production statistics, retail prices, labour force and populationdata for 1978-92 drawn from the China Statistical Yearbook (SSB, variousissues). National sources also offer the advantage of methodologicalconsistency across provinces. Additionally, accordingly to Herrmann-Pillath [1995], SBB has attempted to correct labour force and populationdata at the province level to take the substantial increase in migration duringthe 1980s into account. When a large number of rural residents left theinterior to work as labourers in the booming coastal cities, such movementswere often not recorded by provincial employment records. However,national records still fail to account for the substantial 'floating population'of temporary migrants that live in the shanty towns of China's metropolitanareas [Herrmann-Pillath, 1995].5

Another potential problem with national data are the reported provincialprice deflators [Jian, 1997]. Herrmann-Pillath has a detailed discussionabout the accuracy of the SSB price data. At least up until the late 1980s,the differences between the two secondary sources are not large. Potentialmismeasurment resulting from unrecorded migration or from over or underdeflation remains a problem with official provincial level data in Chinawhich should be borne in mind when reading the results that follow.

As Gross Domestic Product (GDP) and labour force data in the secondsource (as indeed in the China Statistical Yearbook) are only available from1987 and 1985 respectively onwards, I splice this series with the data fromHsueh, Li and Liu [1993] using the implicit growth rates and goingbackwards from the 1987 and 1985 figures in Herrmann-Pillath [1995]. Inthis way, I obtain a second data set (Data 2) for the entire 1978-92 period.In spite of its obvious limitations (particularly due to the splicing of labourforce data), this data-set yields very similar results than the Jian Sachs and

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


Warner data set which was compiled with considerable care using originalChinese sources.

Both Data 1 and Data 2 contain 29 provinces. There are two missingobservations (Qinghai and Guangxi), however, for the years 1979 and1981-85. The series are filled up for reasons of comparability betweenyears using a constant growth rate between the last and the next yearsavailable.

(3) Convergence Results

Table 1 presents coefficients of variation for real GDP per capita and perworker for 1978-92. As can be seen, the differences between Data 1 andData 2 are small. There is a continuous decline in the coefficient of variationof GDP per capita from 1978 through 1992 and of GDP per worker from1978 through 1990 after which no further decline in this series can beobserved [also Man, Sachs and Warner, 1996]. Hence, Chinese provinceshave converged in the sense of sigma. For the real GDP per worker series,however, the decline in dispersion is larger during the early part of the1980s, a result that relates to the estimates of beta convergence for differenttime periods below. Note that GDP per worker dispersion is lower thanGDP per capita dispersion as labour force participation rates are positivelyrelated to the level of development. The fact that this difference hasnarrowed since 1978 might indicate growing labour mobility acrossprovinces [Herrmann-Pillath, 1995].6

While Table 1 overall offers an optimistic assessment of cross-provincialincome dynamics since the start of reforms it allows no firm conclusions onchanges in the rate of convergence over time which have received muchattention in the recent literature. Table 2 presents estimates of theconvergence coefficient beta obtained from Non-Linear-Least-Squaresestimation of equation (1) for various time periods. The sub-periods arechosen to be of equal length. It is not possible in the present context toidentify a specific year in which a break in the trend might have occurred.Non-linear estimation techniques have to be used because the coefficientbeta to be estimated appears as a higher order term in equation I).7 The pointestimates for the 1978-89 and 1978-92 periods (Data 1 and Data 2) are 2.24per cent and 2.55 per cent respectively. The results for Data 1 repeatGundlach's [1996a] results based on the same data set. The point estimatesof beta are in line with studies of income convergence across regions of theUnited States, Japan, the United Kingdom, West Germany, France andSpain [Sala-i-Martin, 1995] which produce estimates of beta around twoper cent per annum.

There are notable variations in the point estimates between periods, withthe late 1970s and early 1980s displaying a higher rate of convergence than

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

ECONOMIC REFORMS . . . A N D CONVERGENCE: CHINA

TABLE 1

VARIATION COEFFICIENTS OF GROSS DOMESTIC PRODUCT PER WORKERAND GROSS DOMESTIC PRODUCT PER CAPITA

Data set

Year:197819791980198119821983198419851986198719881989199019911992

GDP per worker1

0.710.660.660.630.600.600.590.580.570.550.550.55

__-

GDP per worker2

0.710.660.670.630.610.600.600.560.540.540.530.520.500.500.51

GDP per capita1

0.980.920.910.880.840.810.780.750.720.680.670.65

__-

GDP per capita2

0.980.93 -,0.910.880.840.810.780.720.690.660.630.610.570.570.56

Source: Own calculations; Hsueh, Li and Liu [1993]; Herrmann-Pillath (ed.) [1995].

TABLE 2

ESTIMATES OF BETA CONVERGENCE COEFFICIENTS, VARIOUS PERIODS'1

-ESTIMATED REGRESSION: lnYIt-\nYI0 = constant - ( 1 -e"P0 lnY0

Data SetTime period

11978-89 1978-83 1984-89 1978-92

21978-85 1985-92

Beta 0.0224*** 0.0307*** 0.0175*S.E. (0.0074) (0.0103) (0.0097)R2 0.305 0.294 0.124

a Standard errors (S.E.) are in parentheses; significance levels: *** = 1 per cent, ** = 5 percent, * = 10 per cent.

0.0255***(0.0093)0.287

0.0344***(0.0103)0.347

0.0082(0.0101)0.025

Source: Hsueh, Li and Liu [1993]; Hermann-Pillath (ed.) [1995]; own calculations.

the late 1980s and early 1990s. I construct a 95 per cent confidence intervalaround the point estimate for the first sub-periods (1978-83 and 1978-85respectively) to test for significant differences in the point estimates for the1984-89 and 1985-92 sub-periods respectively. The null hypothesis ofequal coefficients is rejected for the 1978-85 and 1985-92 sub-periods butnot for 1978-83 and 1984-89. In line with the results for sigma convergence

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


in Table 1, beta convergence across Chinese provinces slowed after the mid-1980s and may have stopped completely by the 1990s [Man, Sachs andWarner, 1996]. The remainder of the paper focuses on explaining the slowdown of convergence in the 1985-92 sub-period. All further calculationsare thus based on Data 2.

(4) Making Sense of the Convergence Slow-Down

The theory of growth from which equation (1) is derived provides twopossibilities to account for a change in beta convergence rates over time.The first possibility is that some relatively richer provinces haveexperienced a shift in their steady state incomes, either because of apermanent increase in investment or as a result of institutional ortechnological change [Islam, 1995]. Restricting all provinces to the samesteady state as in equation (1) could therefore produce a biased estimate ofbeta. To identify the 'true' rate of convergence equation (1) would have tobe appropriately conditioned.

The other possibility why beta convergence might fall is that capitalmobility across China as a whole has declined. In a neo-classical growthmodel, physical capital mobility accelerates convergence because of theassumption of diminishing returns to each factor. In a non-distortedenvironment, capital tends to flow from richer to poorer countries helpingthe latter to accumulate capital in excess of the increase in domestic savingsand hence close the gap to the richest economy more rapidly. In the extremecase of complete capital mobility, convergence is instantaneous; withlimited mobility of at least one type of capital (typically human capital), theconvergence rate beta will depend on the degree of physical capital mobility[Barro, Mankiw and Sala-i-Martin 1995]. Specifically, the convergencerates for open and closed economies are given by:

physical capital mobility: 3 = (n + A + o)\\ _ a^ (2a)

no physical capital mobility: P = (n + d + g)(l - a,\ - CC2 ) (2b)

a! = production elasticity of physical capitala2 = production elasticity of human capitaln = rate of population growthd = rate of depreciationg = rate of technical progress.

The two formulae are directly derived from the steady state expressions

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

ECONOMIC REFORMS ...AND CONVERGENCE: CHINA 9

of real income per worker in a neo-classical growth model with physical andhuman capital. The difference in the convergence rates for open and closedeconomies is the larger the higher the share of physical capital inproduction. Specifically, the ratio of beta(closed) to beta(open) is (l-0Ci)[also Gundlach, 1996b]? At a more general level, the second argumentrelates a change in the convergence rate over time to changes in theallocation of capital among provinces.

Both possibilities have some support in the literature. However, theunderlying theoretical reasoning is rarely made explicit. Jian, Sachs andWarner [1996] suggest that the slow down in convergence is due to a growingincome gap between the coast and the interior. The two major advantages thecoast has, according to these authors, is the access to international capitalmarkets, and the existence of a pool of cheap agricultural labour located closeto the main export centres. While the first factor would suggest that theconvergence rate among coastal provinces is faster because they are openeconomies, the second factor could imply that the coast may have increasedinvestment in response to the profit opportunities provided by cheap labourand export demand and thus has experienced a shift in the steady state.However, once the authors control for the two factors, the evidence forconvergence surprisingly disappears altogether. This makes an interpretationof their results difficult in the light of the foregoing theoretical analysis.

A number of other studies have dealt with capital mobility acrossChinese provinces [Ma, 1994; World Bank, 1994]. These studies argue thatcapital mobility in China as a whole has declined over the course of reformas provincial governments were given increasing authority over resourcemobilisation. One suggestion made in this context is that the reduction ofinter-provincial fiscal transfers might have greatly contributed to thisdecline and hence fiscal decentralisation could be one reason for the lack ofincome convergence in the second half of the 1980s.

The latter argument is built on a dubious assumption, however. Thus, ifcapital mobility declined after the mid-1980s as a result of fiscaldecentralisation, this implies that the system of inter-provincial fiscaltransfers before that was beneficial to income convergence. This wouldsingle out China as the only socialist economy that effectively emulated themarket, at least in the area of inter-regional capital flows. Reassuringly, thesystem of fiscal transfers was in place long before 1978 and for the entire1952-78 period there is hardly any evidence for per capita incomeconvergence [Cannon, 1990; Jian, Sachs and Warner, 1996]. While thisevidence would argue against an overall decline in capital mobility in Chinaas a cause of the convergence slow-down, it does not exclude the possibilitythat changes in capital allocation within regions of China may have slowedthe convergence process.

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


The above discussion suggests that in order to verify the plausibility ofthe two theoretical propositions, a regionally differentiated analysis is inorder. Thereby, I start from the observation in Jian, Sachs and Warner[1996] that cross-provincial income dynamics are characterised by agrowing income gap between the coastal and interior regions of China.

Figure 1 displays the log of GDP per worker over the 1978-92 periodfor all provinces, the coast, the coast without the three centrallyadministered municipalities ('cities' which were the richest provinces overthe whole reform period), the three cities only, and the interior. The growinggap in GDP per worker between the coastal provinces and the interior isclearly borne out. Another fact emerging from Figure 1 is the rapidconvergence of non-city coastal provinces to the level of income in the threecities. This trend is not interrupted by the mid 1980s, so that convergencewithin the coast would seem to continue at the same rate. Figure 1 does notallow conclusions on convergence among interior provinces. This isprovided in Table 3, which displays coefficients of variation for GDP perworker in three sub-samples. The striking result is that the coefficient ofvariation in the interior declines only up to 1985 and thereafter increasesalmost up to its 1978 value. Apparently, the convergence slow-down hasbeen the result of a combination of growing income disparities betweenregions and a partial reversal of convergence within the interior region.

TABLE 3

VARIATION COEFFICIENTS OF GROSS DOMESTIC PRODUCTPER WORKER BY REGIONS

Year 78 79 80 81 82 93 84

Interior 0.38 0.36 0.36 0.34 0.33 0.33 0.31Coast 0.78 0.74 0.73 0.69 0.66 0.66 0.64(without cities) (0.44) (0.40) (0.35) (0.33) (0.29) (0.31) (0.31)

Year 85 86 87 88 89 90 91 92

Interior 0.27 0.27 0.28 0.31 0.32 0.32 0.34 0.35Coast 0.61 0.58 0.56 0.55 0.53 0.51 0.48 0.46(without cities) (0.29) (0.30) (0.30) (0.31) (0.30) (0.28) (0.28) (0.27)

Source: Hsueh, Li and Liu [1993]; Hermann-Pillath (ed.) [1995].

Table 4 incorporates these observations into estimates of betaconvergence rates for the two time periods 1978-85 and 1985-92.' In theleft column of each panel, I show estimates of beta convergence conditional

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

ECONOMIC REFORMS . . .AND CONVERGENCE: CHINA

FIGURE 1

LOG OF PER CAPITA INCOME, VARIOUS REGIONS OF CHINA, 1978-92a

11

-1 -

1,2"

1,4-

1,6"

1,8 -

-2 -

2,2-

2,4-

2.6 -

y'

1978 1980 1982 1984 1986 1988 1990 1992

All Provinces Interior Coast Coast without Cities Cities

a GDP per worker is measured in Rmb. 10,000. Since no province had reached a per capitaincome above Rmb. 10,000 by 1992, the scale of the vertical axis is negative. (The log of anynumber smaller than 1 is a negative number.)

Source: Data 2.

on allowing for a higher average growth rate in the coastal provinces, duefor instance to a shift in their steady state income level. In the right column,I further allow the coastal provinces to converge to their steady state at adifferent rate than the interior provinces. The results may be summarised asfollows. First, the coastal provinces have grown significantly faster than theinterior. Second, conditioning for this faster growth, there is conditionalconvergence in the 1985-92 period, but the average convergence rate is stilllower than in the 1978-85 period. Third, the coast converges at a roughlyconstant rate (around three to four per cent). The decline in the averageconvergence rate once differences in steady states are taken into account isalmost entirely due to a partial reversal of convergence within the interior inthe 1985-92 period (note the negative implied beta in the last column ofTable 4).

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


TABLE 4

ESTIMATES OF BETA CONVERGENCE COEFFICIENTS INCLUDINGLEVEL AND SLOPE DUMMIES FOR COASTAL PROVINCES

- DEPENDENT VARIABLE: lnW, - ln}7ga

Time Period

Constant

Coast(level dummy)

lnK0

lnYl0* Coast(slope dummy)

ImpliedPImpliedP Coast

R2

1978-92

-0.363(0.217)

0.267***(0.085)

-0.382***(0.083)

-

0.0344***0.0096)

-

0.482

-0.419(0.399)

0.343(0.462)

-0.404***(0.156)

0.031(0.185)

0.0369*(0.0186)

0.0345***(0.0105)

0.484

1978-85

-0.213(0.150)

(0.101)*(0.059)

-0.245***(0.057)

-

0.0402***(0.0109)

-

0.414

a Standard errors in parentheses; * = 10 per cent, ** =level.

-0.453(0.270)

0.429(0.313)

-0.340***(0.105)

0.134(0.126)

0.0593**(0.0228)

0.0413***(0.0128)

0.440

5 percent, ***

1985-92

-0.090(0.141)

0.168**"(0.055)

-0.135**(0.064)

-

0.0207*(0.0106)

-

0.283

0.259(0.253)

-0.293(0.288)

0.029(0.118)

0.227*(0.139)

-0.0041](0.0164)

0.0326**(0.0146)

0.352

= 1 per cent significance

Source: Own calculations from Data Set 2.

In the remainder of the article, I argue that favourable structuralcharacteristics, a rise in investment and the change in reform emphasis fromagriculture to industry have increased the steady state income along thecoast while an increasingly perverse system of inter-provincial fiscaltransfers rather than a decline in capital mobility has led to the break downof convergence in the interior.

III. INDUSTRIAL REFORMS, STRUCTURAL CHARACTERISTICSAND CONDITIONAL CONVERGENCE

This section tries to explain why the coast has grown faster than the interiorsince the mid-1980s. My story is similar to the accounts in Yussuf [1994] andJian, Sachs and Warner [1996] in that it stresses rural industrialisation andaccess to export markets. In contrast to the latter source, however, I do findconditional convergence across all provinces once I control for a possibleshift in the coastal steady state. Thereby, the main factor behind the growingincome gap between the coast and the interior is higher TFP growth incoastal provinces [also Perkins et al, 1993; Raiser, 1995; Wu, 1995].

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


Economic reforms in China started in agriculture. The shift from thecommune system to the household responsibility system and the concurrentincrease in agricultural procurement prices led to a dramatic increase inagricultural productivity with growth of output in agriculture outstrippingindustrial output growth from 1978 to 1983. As the share of agriculture inGDP across Chinese provinces in 1978 was highly negatively correlatedwith initial per capita income, it is no wonder that the early reform periodhas exhibited rapid convergence.10 In the second half of the 1980s, reformefforts shifted to industry. In particular price liberalisation, the acceleratedentry of non-state (private and collective) enterprises (NSEs), and a numberof profit oriented incentive schemes in state industry have led to rapidindustrial output and substantial TFP growth. As I will show, this hasallowed the coast to grow more rapidly than the interior. By contrast,agriculture was no longer a growth motor on its own. The productivity gainsof the early 1980s were soon exhausted and the initial improvement inagriculture's terms of trade was partially reversed due to rising input pricesand remaining price controls on food. Because the beneficial impact ofindustrial reforms was largest in the coastal provinces, and the latter wererelatively rich by the mid-1980s (Figure 1), the 'side effect' of industrialreforms was to weaken the evidence for convergence.

The change in reform emphasis to industry has benefited the coast fortwo reasons. First, the coast has traditionally hosted most of China's lightindustry. This branch has experienced an improvement in its terms of tradeduring the two-track price liberalisation and has been far more exposed toforeign competition than the traditional heavy industry concentrated in theNorthern regions of China or mineral extraction in the West of the country.This has encouraged investment and has led to an above average rate of TFPgrowth in light industry." Second, the coast has started industrial reforms in1985 with a relatively well developed non-state sector. Particularly in ruralareas, this has helped the mobilisation of savings and created domesticcompetition for state-owned enterprises (SOEs). Again, one would expectan increase in investment and possibly higher TFP growth as a result. Notethat there is nothing irreversible about the structural advantages of the coast.However, as I will show in section IV, poorer interior provinces have so farfailed to attract capital for a similar pattern of rural, non-state sector basedindustrialisation process.

Table 5 presents average scores for the share of agriculture in GDP, theshare of non-state firms (NSEs measured as 1 minus the share of SOEs inindustrial production) and of light industry in industrial production in theyear 1985, and investment rates in physical and human capital for the1978-85 and 1985-92 period. The number of books published per capita, avariable proxy ing for the stock of human capital (average for 1985-89) is

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


also provided (see Gundlach [1996a] for a discussion of the appropriatenessof this variable). The table provides support to the contention that the coasthas enjoyed structural characteristics that allowed it to benefitdisportionately from industrial reforms. The average share of light industryand of NSEs in industrial production is significantly higher along the coast.Moreover, the coast has experienced a significant increase in investmentrates between the 1978-85 and 1985-92 periods, as expected. Note,however, that investment rates both in physical and human capital have notdiffered significantly between the coast and the interior in both timeperiods. While the increase in investment should have accelerated growthalong the coast, it is insufficient as a cause for the growing average incomegap to the interior. Finally, the share of agriculture has differed between thetwo regions, but only on account of the three cities. This relates well to thecontention that agriculture on its own was relatively unimportant for growthafter 1985. It receives no further consideration in the present analysis.

TABLE 5

: PROVINCIAL CHARACTERISTICS, REGIONAL AVERAGES,SELECTED YEARS OR PERIODS'*

Coast

Coast without cities

Interior

Coast

Coast without cities

Interior

Share ofagriculture inGDP (1985)

0.266*(0.042)0.334

(0.031)0.351

(0.019

Investment/GDP78/85 85/92

0.241 0.315(0.013) (0.013)0.232 0.304

(0.010) (0.015)0.262 0.292

(0.020) (0.016)

Share of non-stateenterprises in

industrialproduction (1985)

0.399*(0.042)0.449*

(0.032)0.256

(0.016)

Secondary school/Enrolment ratio

(SCHOOL)78/85 85/92

0.054 0.049(0.004) (0.002)0.052 0.047

(0.004) (0.002)0.053 0.050

(0.005) (0.005)

Share of lightindustry inindustrial

production (1985)

0.481*(0.024)0.480*

(0.032)0.389

(0.017)

No. of bookspublishedper head85/89

0.005*(0.002)0.002

(0.0003)0.002

(0.0002)

a A star indicates a significant difference in the sample mean compared to the interior.

Source: Hsueh, Li and Liu [1993]; SSB, various issues.

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


I now investigate the impact of the identified regional differences on percapita income convergence. For this purpose, I estimate convergenceregressions conditional on structural characteristics and on investment ratesin physical and human capital. I obtain strong evidence for conditionalconvergence among Chinese provinces (Table 6).12 The point estimate ofbeta rises from 0.008 in the unconditional case to 0.036 in the conditionedregression. The latter estimate is statistically the same as the unconditionedestimate for the 1978-85 period. The structural characteristics have theexpected sign. However, only the share of light industry is significant.Because investment rates are controlled for, the additional impact of thisvariable on growth is largely through TFP growth, as was posited above.NSEs have no impact on growth through a higher rate of increase in TFP, aresult that contradicts a number of studies on enterprise performanceincluding by the present author [Jefferson, Rawski and Zheng, 1992; Raiser,1995].

TABLE 6

CROSS-PROVINCIAL GROWTH CONDITIONAL ON INVESTMENT RATES ANDSTRUCTURAL CHARACTERISTICS, IMPLIED CONVERGENCE RATE 1985-92a

- DEPENDENT VARIABLE: \nYl,-\nYl0

Constant

lnI/GDP-ln(n + d + g)h

\n{SCHOOL) - ln(n + d + g)

Non-state enterprises /industrial production

Light industry /industrial production

Implied p

R2

Standard errors in brackets, * = 10 per cent, **"n = rate of labour force growth in each Drocin

-0.996***(0.345)

0.357**(0.157)

0.010(0.040)

0.078(0.219)

0.893***(0.302)

-0.224**(0.087)

0.036**(0.016)

0.459

= 5 per cent, *** = 1ce. d + e = 0.07 bv a:

1996a).

Source: Own calculations.

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


Most of these studies do not control for the possibility of selection biasdue to a different distribution of NSEs and SOEs across industries, whichmight explain the apparent contradiction to the present results.13 Finally,because investment rates during the 1985-92 period do not significantlydiffer between the coast and the interior, the evidence in Table 6 allows theconclusion that higher average growth along the coast is largely based onhigher TFP growth. Yet the significance of investment rates in physicalcapital suggests that variations in investment rates have had a substantialimpact on provincial growth rates.14 The next section demonstrates thatapart from higher TFP groy/th along the coast, the nature of capital flowsamong interior provinces and the impact of inter-provincial fiscal transferson investment are the principle reasons behind the convergence slow-down.

IV. CAPITAL FLOWS, FISCAL REDISTRIBUTION AND PERSISTENTINCOME GAPS IN THE INTERIOR

In neo-classical growth theory, capital is expected to flow from rich to poorareas and thereby accelerate the process of convergence. However, politicalinterventions may cause capital flows to go in the opposite direction.Correspondingly, such interventions are likely to reduce convergence.

In China, as in all socialist economies, the fiscal system has traditionallyplayed an overarching role in the allocation of investment. However, sincethe mid 1970s, China has begun to decentralise its fiscal system and allowprovincial governments to retain an increasing share of the revenue fromlocal economic activity. Generally, this system may be said to haveincreased incentives for revenue collection and efficiency improvements inthe local economy and thus to have been an important element of the overallsuccess of Chinese market oriented reforms up to date [Qian and Roland,1994; Raiser, 1995]. While this incentive effect is frequently mentioned, sofar little research has been done on the direction of the still existing fiscaltransfers. Moreover, while fiscal redistribution has diminished, the centralgovernment has repeatedly used the credit plan to fill the gaps. The aim ofthis section is to show that fiscal redistribution has retained a considerableimpact on overall capital flows across Chinese provinces and that thedirection of such flows has tended to follow less and less criteria of inter-regional equity.

Table 7 is reproduced from Ma [1994] and shows how fiscal transfershave diminished during the reform period (more precisely between 1983and 1991). Fiscal transfers are measured as the difference between locallycollected revenues and local government expenditures, a measure also usedby Ferdinand [1989] to characterise the process of fiscal decentralisation.The figures from Ma [1994] put this difference in relation to total revenue.

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

ECONOMIC REFORMS . . . AND CONVERGENCE: CHINA

TABLE 7

FISCAL TRANSFERS ACROSS CHINESE PROVINCES, 1983 AND 1991a

17

Province

Year

Coast:FujianGuangdongGuangxiHebeiJiangsuLiaoningShandongZhejiang

Cities:BeijingShanghaiTianjin

Interior:AnhuiGansuGuizhouHeilongjiangHenanHubeiHunanInner MongoliaJiangxiJilinNingxiaQinghaiShaanxiShanxiSichuanXinjiangYunnan

Fiscal Balance/Revenue in

per cent1983

-42-4

-362257503747

518847

9-42-88-42183014

-227-2837

-290-380

-291

11-231

40

1991

-125

-232

1161

20

104718

-57-28-22-16-3-5-6

-69-28-27

-107-167

-24-6

-10-98-11

a Hainan and Tibet excluded.

Fiscal Balance/

1983

-4.2-0.4-4.52.99.69.74.47.9

11.138.314.8

-8.7-5.1-8.3-3.42.04.61.5

-16.2-2.6-3.6

-25.9-26.9

-3.50.11.2

-16.7-5.7

GDP inper cent

I 1991

-1.50.5

-3.20.21.00.90.12.4

1.510.53.1

-5.1-4.4-3.6-2.1-0.4-0.5-0.5-8.5-3.1-3.9

-12.9-12.9

-3.1-1.0-1.1-8.41-2.6

GDP per capitain Yuan

1983

995.5990.1648.7986.7

1259.11895.91162.71613.4

2899.24212.42588.4

832.1979.1611.1

1993.6894.1

1091.3830.6

1102.3856.5

1512.91082.61255.7847.3

1244.7698.5374.5675.5

1991

1644.12087.6

874.91510.21771.82491.31881.81597.9

3509.84513.73328.6

972.51126.8795.3

2123.71199.91653.4996.4

1591.21151.81559.91484.11549.31195.21482.7968.1

2685.01030.4

Source: Ma [1994]; Data Set 2; own calculations.

I have added two columns giving the ratio of the fiscal balance to GDP andtwo further columns showing the level of GDP per capita in 1983 and 1991.The dependence of local governments on fiscal transfers in some provinces,most notably Xinjiang, Qinghai, Ningxia and Inner Mongolia has been verysubstantial indeed. Moreover, with resource shifts of up to 13 per cent ofprovincial GDP in some provinces as late as 1991, fiscal redistribution was

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


still a powerful tool for regional economic policy. Its significance as a taxburden for the richer provinces has, however, diminished very sharply since1983 and overall fiscal transfers are now small in relation to total ChineseGDP.

Table 7 also clearly reveals the largely unidirectional nature of fiscaltransfers from the rich coastal provinces to some selected interior provinces.The poor provinces (such as Guangxi and Fujian along the coast andGuizhou, Sichuan and Yunnan in the interior) have all received somesubsidies but much less than Xinjiang, Qinghai, Ningxia and InnerMongolia, all with income levels exceeding those of the poorest provincesby a ratio close to two. This already indicates that fiscal transfers followother considerations than regional equity. All of the large subsidy receiversare engaged in some form of mineral or energy production, two of the fewsectors still subject to central administrative price controls. Xinjiang,moreover has strategic value as China's outpost in Central Asia. Itspredominantly non-Han population, as in the cases of Qinghai and Tibet(the latter being excluded from the present analysis), may have triggeredfiscal transfers in order to appease ethnic or religious independencemovements. Fiscal transfers are thus largely politically motivated tocompensate for remaining price controls and to sponsor regionaldevelopment in areas of special interest.15

What has been the impact of such fiscal redistribution on overall capitalflows? Has fiscal decentralisation led to a reduction in capital mobility? Toanswer these questions, I make use of data on aggregate production andaggregate demand by provinces provided by Herrmann-Pillath [1995] forthe 1978-91 period.16 Subtracting the former from the latter, I get a value fortotal net capital flows into a province (positive for capital inflows, negativefor outflows). Subtracting this value from the total value of investment oneobtains provincial savings as a residual.

I first test the hypothesis that capital mobility has declined overall as aresult of fiscal decentralisation. For this, I run a regression of the averagerate of investment in GDP against the average rate of savings in GDP forthe two time periods. Low capital mobility implies a high positivecorrelation [Feldstein andHorioka, 1980]. The coefficients are -0.11 (0.09)and 0.04 (0.11) for 1978-85 and 1985-92 respectively (standard errors inparentheses), suggesting that capital mobility across Chinese provinces washigh in both periods [also Gundlach, 1996a].

In the next step, I regress the average ratio of net capital flows to GDPagainst average per capita GDP. In a second regression, I further add theratio of the fiscal balance to GDP (named SUB) as an explanatory variable.17

To account for regional differences, I allow for a slope dummy on GDP percapita for the coastal provinces. I also present results for a sub-sample

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


excluding the three cities, as they stand out as by far the biggest net payersin the system of fiscal redistribution. The results are presented in Table 8.During 1978-85, net capital inflows are negatively related to per capitaGDP, although significantly so only in the coast. In the 1985-92 period, thecorrelation turns positive for the interior while remaining significantlynegative for the coast, even once the three cities are excluded. Hence, in theinterior, there seems to have been a change in the pattern of capital flowsbetween the two periods in the direction of an increasing miss-allocation ofcapital from the point of view of income convergence.18 Once the impact offiscal transfers is controlled for, the negative correlation between per capitaincomes and net capital flows disappears for the coast as well. Capital flowsare dominated by the impact of fiscal redistribution. Indeed, the explanatorypower of the regression increases by close to 50 percentage points in thesecond period when SUB is included. Moreover, the coefficient on SUBincreases by a factor of two between the two periods. This suggests thatrather than declining, the effect of politically motivated inter-regionalredistribution on capital flows across China has been increasing.

TABLE 8

DETERMINANTS OF CAPITAL FLOWS ACROSS CHINESE PROVINCES, 1978-85,1985-92a - DEPENDENT VARIABLE: (Investment - savings) / GDP

Constant

GDP p.c.

GDP p.c.CoastSUB

R2

1978-1985

0.092(0.059)-0.093(0.617)-1.141**(0.459)-

0.512

0.007(0.033)-0.234(0.323)0.244

(0.293)-1.376***(0.168)0.872

0.051(0.055)0.395

(0.427)-0.656**(0.289)-

0.239

a Standard errors in brackets, * = 10 per cent, ** = 5

1985-92

-0.042(0.034)0.196

(0.256)0.125

(0.211)-2.573***(0.366)0.765

percent, *** = 1

without three cities

0.059(0.079)0.344

(0.582)-0.761**(0.309)_

0.189

0.007(0.044)-0.172(0.338)0.075

(0.216)-2.652***(0.383)0.771

per cent significance.

Source: Herrmann-Pillath (ed.) [1995] for Capital Flows; Ma [1994] for SUB.

Table 8 does not allow any firm conclusions on the determinants of othercapital flows across Chinese provinces. One may conjecture, however, thatbecause the ratio of SUB to GDP has declined overall while the magnitudeof its impact on overall capital flows has increased, other forms of capitalflows have tended to reinforce the pattern established by fiscalredistribution. One possible reason for this is the politically motivatedallocation of policy loans through the banking system." The central result in

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


Tables 7 and 8 is that fiscal transfers have increased the resources availableto relatively richer interior provinces above the level that would besustained by private capital flows in an undistorted environment. This hasallowed such provinces to invest more and this has slowed convergence.

An empirical corroboration of the above argument is given in Table 9.Here, I regress the rate of investment in physical capital to GDP in the1985-92 period against a constant, a level dummy for the coast and SUB.The results confirm that fiscal transfers have had an important impact oninvestment rates. The coefficient of -0.62 suggests that over 60 per cent ofall fiscal transfers have been used to reallocate investment resources fromnet payers to net receivers of inter-provincial taxes. Moreover, once thiseffect is accounted for the coast does have a significantly higher investmentrate than the interior. It is only because of the system of interprovincialtransfers that the beneficial impact of industrial reforms on economicperformance along the coast has not materialised in higher averageinvestment rates than in the interior (see Table 5). The upshot is that whilefiscal redistribution may have prevented the gap between the coast and theinterior to widen further, this has not helped convergence, because themoney has flown to rich rather than poor interior provinces.

TABLE 9

DETERMINANTS OF INVESTMENT RATES 1985-92a

- DEPENDENT VARIABLE: 1/GDP

Constant 0.252***(0.018)

Coast 0.058**(0.09)

SUB -0.624**(0.276)

R2 0.200

a Standard errors in brackets, * = 10 per cent, ** = 5 per cent, *** = 1 per cent significance.


While the above result of a strong positive impact of fiscal redistributionon investment rates arguably provides one important clue to understandingthe convergence slow-down in China, it is rather surprising in a differentcontext. Thus, it might be argued that the access to subsidies might lowerdomestic savings efforts, thereby neutralising the positive impact ongrowth. The high significance of SUB in Table 9 indicates that this is not thecase in China. Another possibility is that the dependence on subsidieslowers the efficiency of investment, such as suggested for instance byKornai's [1980] soft budget constraint theory. By the same token, a province

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

ECONOMIC REFORMS . . .AND CONVERGENCE: CHINA 21

that faces a high rate of central taxation on its tax income may reduce itsreform efforts and thus experience a similar decline in efficiency [Raiser,1995].

I thus return once more to conditional growth regressions and testwhether the system of fiscal redistribution has a negative effect on growthby lowering TFP, even once its positive contribution to investment inrecipient provinces is accounted for. Because the negative incentive effectis argued to effect net payers and receivers in the same direction, the squareof SUB rather than its original value is used as the variable reflecting theimpact of fiscal redistribution. Table 10 presents results for a regression ofprovincial growth rates against investment rates, the share of light industry,SUB2 and initial income. Compared with Table 6 above, I exclude the shareof the non-state sector in industrial production in order to save degrees offreedom, as it had no additional explanatory power. As expected, thecoefficient on SUB2 is negative, albeit only significant at the 11 per centlevel. This gives some support to the contention that a too heavydependence on subsidies or an excessive tax burden might lower aprovince's reform efforts and hence its rate of growth. All other coefficientsare not significantly different from their values in Table 6.

TABLE 10

CROSS-PROVINCIAL GROWTH CONDITIONAL ON INVESTMENT RATES,INITIAL CONDITIONS AND FISCAL TRANSFERS, IMPLIED

CONVERGENCE RATE 1985-92a

Dependent Variable:

Constant

lnI/GDP-\n(n+d + g)

\n{SCHOOL) - ln(n + d + g)

Light industry /industrial production

SUB2

Implied p

lnI7 , - lnW 0

-1.141(0.337)

0.489***(0.167)

0.004(0.038)

0.849***(0.264)

-9.418(5.697)

-0.253**

0.042***(0.015)

R2 0.516

a Standard errors in brackets, * = 10 per cent, ** = 5 per cent, *** = 1 per cent significance.


Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


V. CONCLUSIONS

This article has analysed per capita income convergence across 29 Chineseprovinces over the 1978-92 period. It finds that income levels haveconverged since the beginning of the reform period, in contrast to results forearlier years. However, it also confirms the evidence in other studies that theprocess of income convergence experienced a significant slow-down in thesecond half of the 1980s. I suggest two complementary reasons why thiswas so.

First, the average income gap between the relatively prosperous coastand the more backward interior has widened since 1985. This was due to theshift from rural to industrial reforms and the coast's better starting positionin this process with a high share of light industry and a correspondinglyhigher degree of internal and external competition.

Second, convergence among interior provinces has been reversed tosome extent since 1985. The system of inter-provincial fiscal redistributionstill in place after 15 years of substantial fiscal decentralisation is at leastpartly to blame for this. Particularly since the mid 1980s, fiscal transfershave tended to flow towards the relatively richer interior provinces tocompensate for remaining price distortions and for political reasons. As aresult, investment rates in relatively richer interior provinces have beenpushed up above levels that would be financed by private capital flows. Thishas not only maintained income gaps in the interior but probably alsoreduced the efficiency of investment in China overall. Controlling fordifferences in structural characteristics at the start of industrial reforms andinvestment rates, the paper finds strong evidence for conditionalconvergence for China as a whole.

Before drawing policy conclusions based on the present analysis, twopoints bear emphasis. First, the limitations of the data have to be kept inmind when interpreting the results. More research, extending the timeperiod further into the 1990s and further checks on the consistency ofsources across provinces are needed. Second, while regional convergence inincomes per capita is a highly topical political issue in China today, this veryfact also means that more than mere considerations of economic efficiencyare likely to determine the further evolution of central regional fiscalrelations. This article does not suggest that fiscal transfers to selectedinterior provinces should be discontinued at all costs. It does howeverexpose the fact that such costs exist in terms of cementing or evenincreasing regional income disparities in China, among interior provinces inparticular.

Maybe the clearest conclusion from the article is based on the positiveevidence for income convergence along the coast. It seems that by

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


specialising in light industry and exploiting to the full the local autonomyafforded by fiscal decentralisation, the poorer coastal provinces were able torapidly catch up with their richer neighbours. One important policy factor inthis process has been the ability to encourage enterprise entry throughfavourable tax treatment by provincial governments. The unification ofprofit tax rates with the 1994 tax reform in fact may limit the potential forcatch-up in the future [Man, 1997]. It remains an open question whether theslow-down in per worker income convergence among Chinese provincesidentified in this article is a temporary phenomenon or the beginning of anew trend.

final version received June 1997

NOTES

1. The coefficient of variation is a crude measure of income disparity. Other inequalitymeasures could be computed that take into account population weights (such as the GINIcoefficient) or a simple MinMax measure of distance [Song, 1995]. The results of this articleon convergence across Chinese provinces are actually strengthened by the use of apopulation weighted measure such as the GINI (see below).

2. The level of real income will also depend on the level of production efficiency in theeconomy. Neo-classical growth models generally assume the level of efficiency to be similaracross countries. Production technologies only differ because of variations in factorendowments (i.e. capital and human capital intensities). However, Islam [1995] has recentlyamended the Mankiw, Romer and Weil [1992] model by allowing production efficiency todiffer independently of factor endowments. This possibility plays some role in theestimations below, where I include variables that are hypothesised to influence changes intotal factor productivity across provinces in conditional convergence regressions.

3. Equation (1) is defined in income per worker terms because of its relationship to theunderlying production function. The labour force is counted as an input rather than thepopulation.

4. I should be frank enough to admit that even had I gained access to all original provincialyearbooks needed for my present purposes, my lack of command over the Chinese languagewould have precluded my working with these data. Jian [1997] is the only study I knowwhich directly uses provincial statistical yearbooks. It is worth noting, that his results arelargely congruent with my estimate of beta convergence over the entire 1978-92 period andconfirm the finding of a slow down in convergence. Although a word of caution regardingsecondary data sources is in order, the evidence presented here seems relatively robust to theuse of different sources.

5. In principle it could be argued that unrecorded labour mobility should strengthenconvergence in per capita incomes as labour has tended to move from the poorer interiorprovinces to the rich coastal cities. While this may explain part of the more rapid growthalong the coast and hence the slow down in cross provincial convergence by the late 1980s,it is not clear that migration within the interior can account for the lack of convergenceamong interior provinces which is one important result of this paper. Note that due tomigration, labour force and population data in the two secondary sources I use may differ bya considerable margin in selected provinces.

6. Note that the computations in Table 1 treat each province as a unit of observation with asimilar weight. If one were to compare the standard of living of individual Chinese acrossthe whole country, a population weighted measure such as the GINI coefficient would be

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013


more appropriate. I have also computed GINI coefficients for the same data set. They tendto greatly accentuate the decline in income convergence by the mid-1980s. Indeed, the GINIcoefficients increase after 1985 to regain almost their 1978 level by 1992. For similar results,see Song [1995]. This evidence might explain the large interest in issues of spatial andpersonal income distribution in the Chinese reform debate [see Zhao and Griffin, 1993].

7. In practice, this involves running an OLS regression of cross-provincial growth rates againsta constant and initial income per capita and using the estimated OLS coefficient on initialincome per capita to extract a point estimate for beta based on the non-linear restriction: beta= ln(a-1)/t, where a is the (negative) regression coefficient against initial income. Only theimplied estimates for beta are reported in Table 2.

8. In the present case of a decline in beta from 0.034 to 0.008, the implied value of alpha 1would be 0.77. This is in line with the high average share of physical capital in nationalincome in developing countries as shown by Gundlach [1996b] but contradicts the sameauthor's production function estimates for China [Gundlach, 1996a]. The present article isunable to resolve this issue. I will show below that an overall decline in capital mobility maysafely be excluded as a reason for the slow-down in convergence. Thus what value alpha 1actually has is inconsequential in this context.

9. Table 4 reports the coefficients of the original OLS regression of growth in GDP per workeragainst initial income, a constant and both a level and a slope dummy in addition to theimplied point estimates for beta, which are extracted from the coefficients on initial incomeand (for beta in the coast) the slope dummy. The OLS estimates appear in the upper tier ofthe table, the implied beta estimates are reported in the lower tier.

10. This fact also helps to explain why Jian, Sachs and Warner [1996] do not find anyconvergence once they control for the share of agriculture in their growth regressions oninitial income. In order to explain the slow-down in convergence it is more useful to controlfor variables that are at least not negatively related to initial income and whose positiveimpact on growth might thus overshadow the underlying trend towards convergence.

11. For micro-economic evidence see Hay et al. [1994] and Perkins [1996].12. The regression equation estimated in Table 6 is basically the same as the conditional

convergence equation derived by Mankiw, Romer and Weil [1992]. It involves substitutingthe expression for steady state income into equation (1). I assume constant returns to scale.Hence, the coefficient on the sum of population growth, technological progress and thedepreciation rate is assumed to be equal to minus the sum of the coefficients on investmentsin physical and human capital and ln(n+d+g) is subtracted from ln(I/GDP) andln(SCHOOL). In Table 6, I have additionally included two variables that are expected toinfluence the change in technical efficiency across provinces. The constant gives the averagegrowth rate once all of these factors have been taken into account. The implied beta is againextracted from the coefficient on initial income per worker.

13. Jian [1997] also uses only township and village enterprise (TVE) growth to account for bettercoastal growth performance. In Jian's paper, TVEs have a large positive impact on growth,although light industry is again not included as an explanatory variable. My NSE variableadditionally includes private enterprises, joint ventures and urban collectives, the lattergrowing considerably less vigorously than the TVEs.

14. The coefficient on human capital investment is insignificantly different from zero. Acceptingthis result, the results in Table 6 would imply a value of alpha 1 around 0.6. However, thereare reasons to believe that the insignificance of the implied alpha 2 results from measurementerror [Gundlach, 1996a]. The conditional convergence results are unaffected if once restrictsthe model to different values for alpha 1 and alpha 2.

15. SOE losses are another potential reason for fiscal subsidies from the central government. Forinstance, Jilin and Heilongjiang provinces have received fiscal transfers throughout the1980s, in spite of belonging to the traditionally rich north-eastern industrial heartland. In1993 both had a share of total SOE losses exceeding their contribution to total industrialoutput by a ratio of two [Broadman, 1995].

16. Gundlach [1996a] uses similar data provided by Hsueh, Li and Liu [1993] stretching only to1989. Amending these data with observations from the Chinese Statistical Yearbook leads toseries for provincial savings and investment, which are very close to the data used here. The

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

ECONOMIC REFORMS . . . AND CONVERGENCE: CHINA 25

results are unaffected by the use of the data set.17. The 1983 value is taken as representative for 1978-85 while the 1991 value is used for the

1985-92 period (Table 7).18. One referee has noted that the results in Table 8 should not be over interpreted, as they may

be driven by the role of fiscal transfers in a few interior provinces which account only forfractions of the Chinese population and total GDP. However, excluding these provinces(Xinjiang, Qinghai, Inner Mongolia, Ningxia) from the regressions is difficult in a sample ofonly 29 observations. While Table 8 certainly fails to capture a full picture of capital flowsacross China, it sill demonstrates a surprisingly important role for fiscal transfer

19. One further caveat applies to this result. To the extent that national accounts data areinaccurate, my measure of capital flows would be flawed. There is considerable anecdotalevidence that there are large cross-provincial capital flows through the banking system thatremain unrecorded. Some of these capital flows would effectively offset the official transferof resources to interior provinces, as local governments invest their revenues in moreprofitable ventures along the coast. The impact on overall convergence would be similar,however, as again the relatively more prosperous coastal areas would tend to benefit.

REFERENCES

Barro, R., Mankin, G. and X. Sala-i-Martin, 1995, 'Capital Mobility in Neo-classical Models ofEconomic Growth', American Economic Review, Vol.85, No.l, pp.103-15.

Broadman, H.G., 1995, Meeting the Challenges of Chinese Enterprise Reform, Discussion Papers283, World Bank, Washington, DC.

Cannon, T., 1990, Regions, Inequality and Spatial Policy in China, Working Paper 89, Instituteof Social Studies, The Hague.

Feldstein, M., and C. Horioka, 1980, 'Domestic Saving and International Capital Flows',Economic Journal, Vol.90, No.l, pp.314-29.

Ferdinand, P., 1989, 'The Economic and Financial Dimension', in D. Goodman (ed.), China'sRegional Development, London: Routledge, pp.27-45.

Groves, T, Yong, Y., McMillan, J. and B. Naughton, 1995, 'China's Evolving Managerial LaborMarket', Journal of Political Economy, Vol.103, No.4, pp.873-92.

Gundlach, E, 1996a, Solow Meets Market Socialism: Regional Convergence of Output perWorker in China, Working Papers 726, Institute of World Economics, Kiel.

Gundlach, E., 1996b, Openness and Economic Growth in Developing Countries, Working Papers749, Institute of World Economics, Kiel.

Hay, D., Morris, D., Liu, G. and S. Yao, 1994, Economic Reform and State-owned Enterprises inChina: 1979-1987, Oxford: Clarendon Press.

Herrmann-Pillath (ed.), 1995, Wirtschaftliche Entwicklung in Chinas Provinzen und Regionen,1978-1992, Baden-Baden: Nomos.

Hsueh, T.-T., Li, Q. and S. Liu, 1993, China's Provincial Statistics, Boulder, CO: Westview Press.Hsueh, T.-T., Rawski, T.G. and K. Tsui, 1994, The Impact of Reforms of China's State Industry:

A Regional Perspective, China Research Paper 21, World Bank, Transition and MacroAdjustment Division, Washington, DC.

Islam, N., 1995, 'Growth Empirics: A Panel Data Approach', Quarterly Journal of Economics,Vol.110, No.4, pp. 1127-70.

Jefferson, G.H., Rawski, T.G. and Y. Zheng, 1992, 'Growth, Efficiency, and Convergence inChina's State and Collective Industry', Economic Development and Cultural Change,Vol.40, No.2, pp.239-66.

Jian, T, Sachs, J. and A. Warner, 1996, Trends in Regional Inequality in China, Working Paper5412, NBER, Cambridge, MA.

Jian, T. (1997), Regional Economic Development, TVEs and Tax Refoms in China, DevelopmentDiscussion Paper No.572, Harvard University, Cambridge, MA.

Kornai, J., 1980, The Economics of Shortage, Amsterdam: North-Holland.Ma, Y, 1994, Macro-economic Management and Intergovernmental Relations in China, Policy

Research Working Papers 1408, World Bank, Washington, DC.

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

2 6 THE JOURNAL OF DEVELOPMENT STUDIES

Mankiw, N.G., Romer, D. and D.N. Weil, 1992, 'A Contribution to the Empirics of EconomicGrowth', Quarterly Journal of Economics, Vol.107, No.5, pp.407-37.

Perkins, F.C., 1996, 'Productivity, Performance and Priorities for the Reform of China's State-owned Enterprises', Journal of Development Studies, Vol.32, No.3, pp.414-14.

Perkins, F.C., Zheng, Y. and C. Yong, 1993, 'The Impact of Economic Reform on ProductivityGrowth in Chinese Industry: A Case Study of Xiamen Special Economic Zone', AsianEconomic Journal, Vol.7, No.2, pp.107-46.

Qian, Y. and G. Roland, 1994, Regional Decentralisation and the Soft Budget Constraint: TheCase of China, Discussion Papers 1013, Centre for Economic Policy Research, London.

Raiser, M., 1995, Decentralisation, Autonomy and Efficiency: Inconsistent Reforms andEnterprise Performance in China, Working Papers 689, Institute of World Economics, Kiel.

Sala-i-Martin, X., 1995, The Classical Approach to Convergence Analysis, Discussion Paper1254, CEPR, London.

Solow, R., 1956, 'A Contribution to the Theory of Economic Growth', Quarterly Journal ofEconomics, Vol.71, No.1, pp.65-94.

Song, X., 1995, Regionale Wirtschaftsentwicklung in China, 1978-1992, DuisburgerArbeitspapiere zur Ostasienwirtschaft 18, Gerhard Mercator Universität Duisburg.

World Bank, 1994, China: Internal Market Development and Regulation, A World Bank CountryStudy, Washington, DC.

Wu, Y., 1995, 'Productivity Growth, Technological Progress, and Technical Efficiency Change inChina: A Three Sector Analysis', Journal of Comparative Economics, Vol.21, No.2,pp.207-29.

Yusuf, S., 1994, 'China's Collective and Private Enterprises: Growth and Its Financing', in G.Caprio (ed.), Building Sound Finance in Transition Economies, IMF, Washington, DC,pp.YY.

Zhao, R. and K. Griffin (eds.), 1993, The Distribution of Income in China, New York: St. Martin'sPress.

Dow

nloa

ded

by [

Uni

vers

ite D

e Pa

ris

1] a

t 00:

46 0

4 Se

ptem

ber

2013

Documents

Subsidising inequality: Economic reforms, fiscal transfers and convergence across Chinese provinces