3 Provincial Growth Dynamics

Wanda Tseng, and Markus Rodlauer
Published Date:
February 2003
  • ShareShare
Show Summary Details
Jahangir Aziz and Christoph Duenwald 

The rapid pace of growth of China’s economy masks substantial differences in growth rates and levels of income per capita across different provinces in China. The issue of uneven regional development has recently moved to the top of the policy agenda in an effort to share the benefits of economic reform more broadly.

This chapter seeks answers to the following three questions:

  • Are the relatively poor provinces in China catching up with the rich ones?

  • If they are, what are the characteristics of the catch-up process?

  • On current trends, what will the provincial income distribution look like in the future?

In the empirical growth literature, these questions have been broadly classified as relating to the study of convergence of income levels among economic units—countries or provinces within a particular country. However, “convergence” has been used in the literature to refer to several different phenomena. In this chapter, convergence is defined as the phenomenon of income levels in poorer provinces catching up in relative terms with those in the rich provinces.

The subject of income disparities within China has of late been the focus of several studies. Generally speaking, three main results emerge from these studies:

  • If one looks only at the incomes per capita in China’s provinces, there is little evidence that, since 1978, the initially poorer provinces have on average grown faster than their richer counterparts: what the growth literature calls absolute (or unconditional) beta-convergence. However, if provincial incomes per capita are adjusted for differences in economic structure, economic policies, demographics, and geography, there is evidence that, on average, the initially poorer provinces have indeed grown faster: what the literature calls conditional beta-convergence.

  • Forces of conditional convergence were stronger in the pre-1990 era—the early reform period—and weaker in the 1990s.

  • The dispersion in relative incomes per capita of the provinces fell in the early reform period and rose in the 1990s. A decline in this dispersion is referred to as sigma-convergence.

However, virtually all these studies provide only a partial view of the convergence process. They tend to focus exclusively on the average (in the case of unconditional and conditional beta-convergence) or the standard deviation (in the case of sigma-convergence) of the relative income distribution of provinces and, based on the behavior of these two statistics, draw inferences about whether relative incomes in China’s provinces are converging or not. Although these two statistics provide valuable insights into the convergence process, as shown in many studies both theoretical and empirical, inferences based solely on the behavior of these two statistics are incomplete. In particular, the answer to whether or not the poor provinces are catching up with the rich ones depends on how the shape of the entire provincial relative income distribution has changed over time, and not simply on the behavior of two statistics of the distribution.

The approach taken in this chapter is to exploit more fully the information contained in the shape of the relative income distribution and how it has changed over time. To do this, in the spirit of Quah (1997), kernel estimates of the relative income distribution of China’s provinces are computed and their intertemporal properties characterized. The results from this exercise suggest the following:

  • Incomes per capita in the initially poor provinces are catching up with those in the initially richer provinces.

  • However, this overall tendency masks significant differences across provinces. On the one hand, the coastal provinces are growing relatively faster than the rest, including the initially richer provinces. On the other hand, many of the initially poorer provinces, after improving their relative rankings in the 1980s, fell behind in the 1990s. The initially richer provinces have been losing their standing in the relative income ladder quite rapidly.

  • As a result, the relative income distribution seems to be stratifying into a bimodal distribution, with the coastal provinces gravitating toward one mode and the remaining provinces toward the other.

  • Provinces’ economic structure and policies, in particular the concentration of state-owned enterprises (SOEs) and the openness of the province to external trade, have been significant factors in these distributional dynamics.

Recent Evidence on Convergence

This section summarizes existing evidence on convergence among China’s provinces, first by reviewing the record of economic performance across provinces and then by distilling the results of past studies of convergence in China.

Economic Performance of China’s Provinces

Economic performance has varied widely across China’s provinces.1 Between 1978 and 1997, incomes per capita grew in all Chinese provinces in absolute terms, and both initially rich and initially poor provinces experienced a significant increase in living standards (Figure 3.1). However, the extent of improvement in living standards differed substantially from province to province. Although real GDP per capita has grown by at least 5 percent on average during the reform period, some provinces—notably those along the coast—have grown more than twice that fast.2 Indeed, the variation in economic performance has displayed some distinct geographical patterns. Coastal provinces have tended to outperform the central provinces, which in turn have surpassed the western provinces. For example, coastal provinces such as Guangdong, Fujian, and Zhejiang each grew at an average annual rate of around 12 percent in real per capita terms during 1978–97, while central provinces like Hubei, Henan, and Jiangxi grew by about 9 percent a year. Western provinces such as Gansu, Qinghai, and Ningxia grew at rates between 5 and 7 percent.

Figure 3.1.Income per Capita by Province

(Shanghai, 1997=100)

Sources: National Bureau of Statistics; and IMF staff estimates.

Significant variation in relative economic performance over time is also a feature of the provincial data (Table 3.1). Shanghai and Guizhou were China’s richest and poorest provinces, respectively, both in 1978 and in 1997. However, there were considerable changes in the rankings (in terms of the level of real GDP per capita) of the other provinces between those two years. For instance, the coastal province of Zhejiang moved from 15th in the rankings in 1978 to 5th in 1997; the western province of Qinghai dropped from 6th to 20th over the same period. More broadly, in 1978 the most affluent provinces tended to be in the northeast, but by 1989 some of the coastal provinces had joined the northeastern region among the most affluent in the country. By 1997 the coastal provinces were clearly the most affluent (apart from the metropolitan areas of Shanghai, Beijing, and Tianjin), followed by the northeastern region.

Table 3.1.Real GDP per Capita Relative to Shanghai, by Province

(Index, Shanghai = 1.00)1

Hubei0.19Hebei0.23Inner Mongolia0.18
Fujian0.18Inner Mongolia0.22Guangxi0.17
Inner Mongolia0.18Shaanxi0.20Hunan0.16
Source: National Bureau of Statistics.

Hainan and Tibet Autonomous Region were excluded because income per capita data are not available before 1985. Data for Chongqing, which became a municipality in 1997, are included in the data for Sichuan.

Source: National Bureau of Statistics.

Hainan and Tibet Autonomous Region were excluded because income per capita data are not available before 1985. Data for Chongqing, which became a municipality in 1997, are included in the data for Sichuan.

Studies of Convergence in Growth Regressions

One of the key predictions of the neoclassical growth model is the convergence hypothesis: the tendency of poor countries, or regions within a country, to catch up with richer countries or regions.3 In the literature, the most common approach to identifying such a phenomenon is to conduct a beta-convergence exercise, which amounts to verifying whether the neoclassical (standard Solow or augmented endogenous growth) model is a good description of a country’s development experience.4 In this context, as already noted, a distinction between absolute and conditional beta-convergence is typically made. If economies vary in their saving rates and initial capital stocks, then the neoclassical model predicts conditional convergence, a situation in which incomes per capita converge, conditional on each economy’s steady state.5 That is, among economies that are similar in preferences, technologies, saving rates, and other structural characteristics, the lower the initial level of output per capita, the higher the growth rate.

Testing the hypothesis of beta-convergence commonly involves regressing growth in output per capita during a given time interval on a constant, initial income per capita, and a set of conditioning variables. Empirical studies differ in the conditioning variables included, but investment ratios, educational characteristics, population growth, and (in the case of China) dummy variables for coastal effects have typically been used. The conditional convergence hypothesis predicts a statistically significant negative coefficient on initial income (holding the conditioning variables constant).

In the case of China, several studies have shown that, in general, the relative dispersion of provincial incomes per capita fell in the 1980s but rose subsequently. This observation suggests that the evolution of China’s regional income dynamics can be roughly divided into two time periods: 1978–89 and 1990–97, with the former period characterized by convergence and the latter by divergence, as measured by sigma-convergence.

A number of studies have also tested the beta-convergence hypothesis using Chinese provincial data. Table 3.2. provides a selected survey of these studies, whose main findings are the following:

  • Jian, Sachs, and Warner (1996) find that China’s regional disparities narrowed between 1978 and 1990 but that, subsequently, the coastal and interior regions of China began to diverge, reflecting in part the special privileges given to the coastal regions. These authors explain the divergence in regional incomes by an increase in the variance between the coastal provinces and the interior provinces, rather than by an increase in the variance within either the coastal region or the interior. They conclude that China is on a dual track, with a prosperous coastal region that is growing rapidly and a poor interior that is growing more slowly.

  • Chen and Fleisher (1996), using an augmented Solow growth model, find evidence that convergence was conditional on coastal location. Their result suggests that convergence is occurring within the coastal and inland regions but not between these regions.

  • Raiser and Nunnenkamp (1997) find evidence in support of conditional income convergence among China’s provinces; however, they also show that the rate of convergence in the 1985–92 period was markedly slower than that for 1978–85.

  • Li, Liu, and Rebelo (1998), using the augmented Solow-Swan model, find support for the conditional convergence hypothesis and estimate the convergence rate to be a relatively high 4¾ percent a year.6 However, they also note an increase in income inequality after 1990, and they point out that although economic reforms in China have facilitated convergence of each province toward its steady state, they have also widened the gap between the steady states of different provinces. The authors also present evidence in support of unconditional convergence during the sample period, so that regional economies converge even though they have dissimilar steady states.

  • Dayal-Gulati and Husain (2000) show that regions are converging, but to different steady-state levels of income. They find that the pattern of foreign direct investment flows, as well as structural characteristics of the regions—including total investment, the concentration of SOEs, and bank loan-deposit ratios—are important factors determining growth and convergence.

Table 3.2.Selected Previously Reported Tests of Convergence for China’s Provinces
Study and PeriodType of Convergence Found
Jian, Sachs, and Warner (1996)
Chen and Fleisher (1996)
Raiser and Nunnenkamp (1997)
Li, Liu, and Rebelo (1998)
Dayal-Gulati and Husain (2000)
Source: Literature cited.
Source: Literature cited.

Using a representative set of the same conditioning variables used in the above studies, similar beta-convergence exercises were performed for this chapter for the period 1978–97 and the subperiods 1978–89 and 1990–97. The results, summarized in Table 3.3, generally confirm those from the earlier studies.

Table 3.3.Results of Provincial Growth Regressions
Initial income per capita (in logarithms)–1.24–3.57**–2.49***–5.62***2.77*–1.56
Population growth
(in percent)–0.52–2.010.3
Domestic investment–to–
GDP ratio0.0270.07*0.02
Foreign direct investment–
to–GDP ratio0.140.27***0.05
Government revenue–to–
expenditure ratio–0.0010.003–0.004
M2–to–GDP ratio–0.007–0.001–0.014
Share of SOEs in industrial output–0.045***–0.02–0.09***
Coastal dummy variable1.83***1.3*1.97
Summary statistics
Adjusted R20.030.790.090.610.020.84
Standard error of the regression3.030.81.480.972.481.01
Source: IMF staff estimates.

The dependent variable is the growth rate of real GDP per capita.

Indicates statistical significance at the 1 percent level.

Indicates statistical significance at the 5 percent level.

Indicates statistical significance at the 10 percent level.

Source: IMF staff estimates.

The dependent variable is the growth rate of real GDP per capita.

Indicates statistical significance at the 1 percent level.

Indicates statistical significance at the 5 percent level.

Indicates statistical significance at the 10 percent level.

Dynamics of Convergence

The results summarized in the previous section, although useful, do not provide a complete picture about the shape of the relative income distribution or how it has evolved over the years. To obtain such a picture, the kernels of the actual relative provincial incomes in different time periods are estimated so that their shapes and intertemporal dynamics can be studied. A kernel estimator of a set of observations—in this case the relative rankings of provincial income per capita—is an estimate of the distribution function from which the observations are likely to have been drawn (for details, see Silverman, 1986). Mathematically, the kernel estimator f(x) is defined as

where Xj are the individual observations, N is the number of data points, h is a window width or smoothing parameter,7 and K is a kernel or weighting function (assumed here to be the normal distribution).

Provincial Income Distribution

Figure 3.2 displays the kernels of provincial relative incomes in 1978, 1989, and 1997. In each panel the kernel was estimated in the following three steps:

  • In each year the real income per capita of each of China’s 28 provinces was rescaled as a fraction of Shanghai’s income per capita,8 such that the range of the distribution is restricted to lie between 0 and 1. Since, by construction, Shanghai’s relative income is always 1, it, too, was excluded from the sample.

  • For a suitably large number of points spanning the interval from 0 to 1,9 the relative frequency (that is, the unconditional probability) with which each of these values could have occurred was estimated. The probability of each point was computed as the weighted average of the distance of that point from the given relative incomes of all the 27 provinces, with the weights drawn from a normal or Gaussian distribution centered at that point.

  • The relative frequencies of these points were filtered for noise using the procedure in Silverman (1986). The collection of the filtered relative frequencies formed the kernel of the relative provincial incomes in that year. The area of the distribution was normalized to 100.

Figure 3.2.Gaussian Kernels of Provincial Relative Income Distribution

Relative frequency (percent)

Sources: National Bureau of Statistics; and IMF staff estimates.

One interpretation of the kernel estimators is that, based on the actual growth experience of China’s provinces, they tell us how likely it is that a province’s income per capita, on average, was a certain fraction of Shanghai’s income per capita in a particular year. For example, in the top panel of Figure 3.2, the unconditional probability that a province’s income per capita in 1978 was one-fifth of Shanghai’s was 14 percent. This probability declined to 10 percent in 1989 and then rose to 12 percent in 1997 (middle and bottom panels).

An examination of the provincial income distributions over the period 1978–97 reveals the following stylized facts:10

  • Most of the mass of the income distributions for all three periods remained below two-fifths of Shanghai’s income per capita, indicating that, throughout the last two decades of reform, on average, a province’s income per capita was most likely to have been less than 40 percent of Shanghai’s.

  • In the early reform period there was some decline in the mass of provinces in the first quintile of the distribution, but this was partly offset by an increase in the second quintile.

  • In the late reform period, the shift to the second quintile was reversed, as the proportion of provinces in the first quintile rose.

These stylized facts are consistent with some of the results of the convergence exercises. It shows that there was a tendency in the early reform period toward unconditional beta-convergence, which disappears in the later reform period. It is also consistent with a decrease in the standard deviation of provincial relative incomes in the 1980s (sigma-convergence), followed by a rising trend in the 1990s.

However, the relative income distributions tell us little about whether the poor provinces became richer or poorer in relative terms in the early and late reform periods. Figure 3.3 provides various examples of movements of provinces over time that preserve the overall shape of the distributions in Figure 3.2 but reflect dramatically different growth dynamics. In particular, it is possible that a province at point A in 1978 (at one-fifth of Shanghai’s income per capita) moved to point B in 1989 and then to point C in 1997. Alternatively, the province at point A could have moved to point D in 1989 and then fallen back to point E in 1997. These examples suggest that an answer to the question requires carefully tracking the positions of each province in the relative income distributions in 1978, 1989, and 1997.

Figure 3.3.Examples of Distribution-Preserving Movements Within Gaussian Kernels

Relative frequency (percent)

Sources: National Bureau of Statistics; and IMF staff estimates.

Growth Dynamics in China’s Provinces

The intradistributional dynamics among the Chinese provinces is displayed in Figure 3.4. In each panel the kernel of the joint distribution of relative incomes in the initial and terminal years is shown.

Figure 3.4.Distribution Dynamics Across the Reform Period and in Subperiods1

Sources: National Bureau of Statistics; and IMF staff estimates.

1The figures depict Gaussian kernels of the joint distribution of provincial relative income. The area under the entire distribution is normalized to be 100. Horizontal axes measure provincial income per capita as a fraction of Shanghai’s, and the vertical axis the relative frequency in percent. Movement to the right of the north-south diagonal indicates improvement in the relative ranking, and movement to the left a worsening, between the initial and the terminal years.

The horizontal axes measure the relative incomes in the initial and terminal years, and the vertical axis measures the frequency with which a particular growth experience occurred between the two periods. Points of the distribution that lie along the north-south diagonal represent unchanged relative incomes, whereas points to the right of the diagonal represent a rise in relative incomes between the two periods plotted, and points to the left represent a decline.

In the first panel of Figure 3.4, the kernel shows that the dominant experience among China’s provinces was that relative incomes were between one-fifth and two-fifths of Shanghai’s income in 1978 and remained in that interval in 1989; for a small number of provinces relative income was higher in 1978, but they remained around the same levels 11 years later. Put differently, based on China’s actual provincial growth experience during 1978–89, the probability that a province in the interval between one-fifth and two-fifths of Shanghai’s income in 1978 remained in that interval at the end of the period was fairly high, and that of reaching, say, three-fifths or more of Shanghai’s income was virtually zero.

This picture of apparent immobility, however, is not entirely correct. Along the north-south diagonal of the panel, the entire distribution is skewed to the right. This implies that although most provinces remained in the second quintile between 1978 and 1989, many shifted closer to the upper end of the interval during the period. The same was true for those in the fourth quintile in 1978; they moved up into the fifth quintile in 1989. Consequently, in the early reform period, not only was there considerable intradistributional mobility across provinces, but the poor provinces did in fact become richer in relative terms.

This trend seems to have been somewhat reversed in the later reform period (second panel of Figure 3.4). The kernel of the joint distribution between 1989 and 1997 shows a distinct leftward skew along the north-south diagonal at both the lower and the upper end of the range. This implies that although the provinces in the second quintile shifted toward the lower end of the interval, so did those in the upper quintiles. More interestingly, however, a set of provinces in the second quintile in 1989 broke out of that range to move into the third and fourth quintiles in subsequent years. Some of the relatively better-off provinces whose relative positions worsened were Heilongjiang, Liaoning, Qinghai, Tianjin, and Beijing, while provinces such as Guizhou, Yunnan, Gansu, and Shaanxi, which were already relatively poor, became relatively poorer. The provinces that gained the most in relative terms were those in the coastal region: Guangdong, Jiangsu, Zhejiang, and Fujian. Thus, while the coastal provinces gained in rank, some of the relatively rich provinces and many of the low-income provinces fell behind.

What do the intradistributional dynamics for the entire 1978–97 period look like? The third panel of Figure 3.4 shows the kernel of the joint relative income distribution between 1978 and 1997. The distribution has two distinct features: although the bulk of the distribution is shifted to the left of the north-south diagonal, there is also a significant mass skewed to the right. It would seem as if the coastal provinces are gravitating rightward and forming their own cluster, while the remaining regions—both the relatively rich and the relatively poor—are gravitating to the left to form a separate cluster. In other words, there is an emerging tendency for the distribution to be stratified into a bimodal distribution.11

These results provide somewhat firmer ground on which to answer the questions raised in the introduction:

  • The relatively poor provinces are catching up with the rich ones, but this is occurring in a somewhat complex manner. Some of the coastal provinces, which were relatively poor at the beginning of the reform period, have been growing at a considerably faster pace than the erstwhile rich provinces of the rust belt, especially in the 1990s. As a result, the gap between the coastal provinces and the initially rich provinces is closing in relative terms. On the other hand, the other initially poor provinces are falling behind in relative terms, such that the dispersion among these initially poor and initially rich provinces is also declining—the provinces seem to be clustering toward two separate relative income clubs.

  • Furthermore, there has been considerable mobility in the relative rankings. This raises a whole new set of questions, which are explored in the next section. If labor and capital were relatively immobile in China, as conventional wisdom suggests, what explains the intradistributional churning? Is it due to the structure of these economies, or to differences in the policies adopted by these provinces? The next section provides partial answers to these questions.

  • To answer the question about the future shape of the regional income distribution, the relative income rankings based on the relative growth differentials among the provinces in the 1990s were projected to 2010. Figure 3.5 shows the kernel of the joint distribution of relative incomes in 1978 and that projected for 2010. The stratification into two peaks has become more pronounced, underscoring the earlier conclusion about emerging club convergence.

Figure 3.5.Projected Distribution Dynamics on Current Trends1

Sources: National Bureau of Statistics; and IMF staff estimates.

1The figures depict Gaussian kernels of the joint distribution of provincial relative income. The area under the entire distribution is normalized to be 100. Horizontal axes measure provincial income per capita as a fraction of Shanghai’s, and the vertical axis the relative frequency in percent. Movement to the right of the north-south diagonal indicates improvement in the relative ranking, and movement to the left a worsening, between the initial and the terminal years.

Explaining Provincial Growth Dynamics

The analysis so far has not made a distinction of the kind commonly made between conditional and unconditional beta-convergence. Recall that this distinction was based on the notion that different provinces could be converging to different steady states, depending on the specific features of their economic structure and economic policies, and thus converging at different rates. This section focuses on whether the growth dynamics observed and inferences made in the previous section, which were based only on the growth experiences of the provinces and thus unconditional in nature, are affected by economic structure and policies.

The methodology used was the following:

  • For each province, “conditioned” growth rates for the periods 1979–89 and 1990–97 were constructed using the fraction of the growth rate explained by the variables in regression equations (4) and (6) in Table 3.3: the population growth rate, the foreign direct investment-to-GDP ratio, the domestic investment-to-GDP ratio, the government revenue-to-expenditure ratio, the M2-to-GDP ratio, the share of SOEs in industrial production, and a dummy variable for the coastal provinces.12

  • Residual growth rates were then computed by subtracting the conditioned growth rates from the realized growth rates. Using these residual growth rates, conditioned relative incomes for 1989 and 1997 were computed.

  • Finally, based on actual relative incomes in 1978 and the conditioned relative incomes in 1989 and 1997, kernels of the conditioned joint distribution of relative incomes between 1978–89, 1989–97, and 1978–97 were computed.

The noticeable feature of the conditioned distributions for 1978–89, 1989–97, and 1978–97 (Figure 3.6) is that they are all skewed to the right of the north-south diagonal. By comparing the conditioned joint distributions with the unconditioned ones (Figure 3.4), the following inferences can be drawn:

  • For the period 1978–89, the skew in the unconditioned joint distribution is not very different from that in the conditioned distribution; thus, economic structure and policies had little influence on interregional convergence.

  • However, for the period 1989–97, the conditioned distribution shows a more marked skew toward the right of the north-south diagonal than does the unconditioned distribution. This underscores the strong presence of convergence forces after the influence of policies and economic structure has been filtered out. Importantly, the coastal regions no longer show the previous strong shift to higher rankings. The absence of the marked shift was largely due to the strong explanatory power of the coastal dummy in the growth regression for the later reform period. As noted previously, the coastal dummy acts as a proxy for the external openness of the provinces, albeit a very weak one. This suggests that external trade was a strong factor behind the growth of the coastal provinces in the 1990s.

Figure 3.6.Conditioned Distribution Dynamics Across the Reform Period and in Subperiods1

Sources: National Bureau of Statistics; and IMF staff estimates.

1The figures depict Gaussian kernels of the joint distribution of provincial relative income. The area under the entire distribution is normalized to be 100. Horizontal axes measure provincial income per capita as a fraction of Shanghai’s, and the vertical axis the relative frequency in percent. Movement to the right of the north-south diagonal indicates improvement in the relative ranking, and movement to the left a worsening, between the initial and the terminal years.

  • For the period as a whole, the conditioned distribution shows an almost uniform tendency toward convergence. In contrast, the unconditioned distribution showed more complex dynamics, with an emerging tendency toward twin-peakedness. Consequently, it would appear that the economic structure and policies of the provinces have played an important role in increasing stratification.13

The variation in economic performance across provinces has been a function of the timing, sequencing, and targeting of economic reforms and the associated structural shifts. China’s pre-1978 development strategy had emphasized balance and equity, but this gave way to the pursuit, in the 1980s, of the objective of rapid growth based on gradual and incremental reforms.14 This new strategy shifted the focus of state investment from the interior to the coastal regions and granted the latter preferential treatment. Beginning in 1980, the coastal region was opened to foreign investment through the creation of special economic zones and open cities. Combined with this region’s other advantages, such as a relative abundance of human capital and favorable geographical location, this resulted in a sharp pickup in growth along China’s southern coast. The resulting narrowing gap between the southern coastal provinces and the eastern coastal provinces served to reduce regional disparities during the 1980s. As the government’s pro-coastal policy orientation strengthened further in the late 1980s, the interior provinces became increasingly cut off, and income disparities started to rise. This led the government to adjust its policies during the Eighth Five-Year Plan (1991–95) and to increase support to the less developed regions in the central and western parts of China in the Ninth Five-Year Plan (1996–2000). More recently the government has redoubled its efforts to reduce regional disparities.

The northeastern provinces, including Liaoning and Heilongjiang, were initially the most affluent (except for the largely metropolitan areas of Shanghai, Beijing, and Tianjin), reflecting large-scale investment in the industrial sector under central planning. As a result, on the eve of the reforms, provinces with relatively large industrial sectors tended to be the most affluent. However, the associated high concentration of SOEs in the eastern provinces subsequently became a drag on growth, with most of these provinces recording relatively low growth rates.

Apart from the selective opening during the 1980s, the other major area of reform in the post-1978 period was in the agricultural sector. A shift in agricultural production from the commune system to the household responsibility system led to a rapid increase in agricultural productivity, raising incomes per capita in those central and western provinces where agriculture was a relatively large share of provincial GDP. In addition, rural areas benefited from the emergence of township and village enterprises, which operated outside of the central plan and absorbed excess labor from agriculture. These reforms, which took place for the most part between 1979 and 1985, resulted in relatively strong growth in income per capita in those provinces where agriculture was the dominant sector and where there was a high concentration of township and village enterprises, contributing to a reduction in regional income inequalities during the 1980s.


Conventional convergence studies, which use regression techniques to determine the existence of convergence, do not provide a complete answer to whether poor provinces in China are catching up with richer provinces. A better approach is to examine the entire distribution of relative provincial incomes, because this sheds light on the intertemporal dynamics that lie behind China’s development and growth experience. The distributional dynamics suggest that China’s poor provinces are catching up with the rich ones. This catch-up, however, is occurring in a complex manner, with the gap between the coastal provinces and the initially rich provinces closing in relative terms. Moreover, the initially poor provinces, as well as the previously richer provinces, are both falling behind in relative terms (the latter at a faster pace). Thus provinces in China appear to be clustering into relative income clubs of their own, which is causing the distribution of relative incomes to become stratified into a bimodal distribution. An extrapolation of current growth trends to 2010 reveals an even more pronounced bimodal stratification. Finally, this chapter finds that economic structure and policies have played an important role in bringing about this stratification.

China’s entry into the World Trade Organization could accelerate the process of stratification. The coastal provinces and, to a lesser extent, those provinces’ immediate hinterlands, could benefit from the expected growth in international trade. On the other hand, the relatively poor provinces, particularly those dependent on agriculture, could see a further slowdown in the growth of incomes per capita as prices for agricultural products fall. Moreover, the provinces with a heavy concentration of SOEs may see their former affluence erode even further as pressures mount to restructure the state enterprises.

As mentioned earlier, the Ninth Five-Year Plan (1996–2000) addressed regional disparities, and since then the Chinese authorities have intensified their efforts to narrow the income gaps among China’s provinces. Indeed, about two-thirds of expenditure under the fiscal stimulus packages in 1998 and 1999 was targeted at the central and western provinces. Developing the central and western regions is also a cornerstone of the Tenth Five-Year Plan (2001–05), and a “Develop the West” initiative has recently been launched. Under this plan, efforts are focused on infrastructural investment, technological upgrading, and training and education. In addition, efforts are being made to equalize preferential tax policies between coastal and inland areas. With regard to the fiscal system more generally, although discretionary fiscal transfers have been used in the past to help offset large social safety net needs in poorer regions, concerns about regional disparities may necessitate a review of the entire system of intergovernmental fiscal relations at some point in the future.


    BarroRobert J. and XavierSala-i-Martin1995Economic Growth (New York: McGraw-Hill).

    BaumolWilliam1986“Productivity Growth, Convergence, and Welfare,”American Economic Review 76 (December) pp. 107285.

    CashinPaul and RatnaSahay1996“Internal Migration, Center-State Grants, and Economic Growth in the States of India,”Staff Papers International Monetary FundVol. 43 (March) pp. 12371.

    ChenJian and BeltonM. Fleisher1996“Regional Income Inequality and Economic Growth in China,”Journal of Comparative EconomicsVol. 22No. 2 pp. 14164.

    Dayal-GulatiAnuradha and Aasim M.Husain2000“Centripetal Forces in China’s Economic Take-off,”IMF Working Paper 00/86 (Washington: International Monetary Fund).

    HusainAasim M.1998“Economic Performance and Business Cycles in China’s Provinces and Regions” in People’s Republic of China—Selected Issues (Washington: International Monetary Fund, SM/98/184).

    JianTianlunJeffreyD. Sachs and AndrewM. Warner1996“Trends in Regional Inequality in China,”NBER Working Paper 5412 (Cambridge, Massachusetts: National Bureau of Economic Research).

    Li HongZinan Liu and IvoniaRebelo1998“Testing the Neoclassical Theory of Economic Growth: Evidence from Chinese Provinces,”Economics of PlanningVol. 31No. 2–3 pp. 11732.

    MankiwN. GregoryDavidRomer and DavidN. Weil1992“A Contribution to the Empirics of Economic Growth,”Quarterly Journal of EconomicsVol. 107No. 2 (May) pp. 40737.

    QuahDanny1997“Empirics for Growth and Distribution: Polarization, Stratification, and Convergence Clubs,”Journal of Economic GrowthVol. 2 (March) pp. 2759.

    RaiserMartin and PeterNunnenkamp1997“Die andere Seite Chinas—Strukturprobleme, Reformdefizite und verzögerte Aufholprozesse im chinesischen Binnenland,”Kiel Working Papers 794 (Kiel, Germany: Kiel Institute of World Economics).

    SilvermanBernard W.1986Density Estimation for Statistics and Data Analysis (New York: Chapman and Hall).

    WangShaoguang and AngangHu1999The Political Economy of Uneven Development—The Case of China (New York: M.E. Sharpe).

As noted in Chapter 2, some observers believe that China’s GDP growth rates are overstated by as much as 2 percentage points. This issue is not relevant to the present study, which focuses on relative provincial GDP levels and growth rates, under the assumption that there is no systematic tendency for growth in some provinces to be overstated more than in others.

More precisely, and as argued by Barro and Sala-i-Martin (1995), the neoclassical model leads to conditional rather than absolute convergence as these terms were defined in the introduction to this chapter.

Convergence in this case arises from the assumption of diminishing returns to capital. Since the rate of return on capital is lower in economies with more capital per worker, there are incentives for capital to flow from rich to poor economies, boosting growth in the latter relative to the former, and thus causing convergence.

By comparison, estimates of the rate of convergence for other economies center around 2 percent a year. See Mankiw, Romer, and Weil (1992), for example.

The window width was chosen following the suggestion in Silverman (1986) that it be given by 0.9 AN–⅕, where A= min(standard deviation, interquartile range/1.34).

The choice of Shanghai as the numeraire is arbitrary and has little impact on the analysis. Hainan and the Tibet Autonomous Region were excluded from the sample because data on income per capita are not available before 1985. Data for Chongqing, which became a municipality in 1997, are included in the data for Sichuan.

For these exercises, the interval from 0 to 1 was divided into equally spaced 50 subintervals.

In the remainder of this chapter the terms “kernel” and “distribution” will be used interchangeably. In the figures the income distribution is referred to as the Gaussian kernel, because the weights used were drawn from a Gaussian distribution. Weights drawn from an Epanechnikov distribution, the other frequently used weighting method, did not seem to make any material difference to the shape of the estimated kernels.

Quah (1997) termed this bimodality “twin peaks” in the context of cross-country growth experience. In that study the twin peaks lay along the north-south diagonal, implying little mobility in relative rankings. In the case of China’s provincial growth, the emergent twin peaks would lie across the north-south diagonal, implying significant mobility in the rankings. This is a specific example of what Baumol (1986) termed “club convergence.”

Except for population growth, all the other variables were chosen on the basis of results in Dayal-Gulati and Husain (2000).

The major factors, however, are mainly structural, because the coefficients on the structural variables in regressions (4) and (6) in Table 3.3 are statistically significant whereas those on the policy variables are not.

During the Sixth Five-Year Plan (1981–85), a pro-coastal policy program was adopted, and this orientation became even more pronounced in the Seventh Five-Year Plan (1986–90). The idea was that reforms should be conducted on an experimental basis in certain regions first. If these experiments were successful, their influence would eventually spread to other regions (see Wang and Hu, 1999).

    Other Resources Citing This Publication