Political institutions, economic institutions, and income inequality in urban China

Transcription

1 Political institutions, economic institutions, and income inequality in urban China HAI ZHONG School of Public Finance and Public Policy Central University of Finance and Economics 39 South College Road, Beijing, China, Tel: ; Fax: Abstract Income inequality is often increased significantly in the process of economic transition. The distribution of losses and gains generated in this process among different social groups has far-reaching implications. Further political and economic reforms might be constrained when the ruling elites responsible for institutional changes have vested interests. In this paper, we find that political rent has emerged as a more and more significant source of income inequality in China. The unconditional decomposition results show that 36.5% of the increase of income inequality from 1997 to 2006 can be attributed to political rent. Even after we control for various household characteristics, political rent still contributes 19.3% to the increase. It seems that the ruling elites have taken great advantages from the restructuring of the SOEs and government reforms occurred in this period. Further economic reforms may need political institutional reforms that create effective constraints on power-holders, and ensure relatively few rents to be captured by them. JEL classification: D31; O43; P21; P26 Keywords: Inequality; Institutions; Economic transition; Political economy; China 1

2 1. Introduction Income inequality is often increased significantly in the process of transition from a planned to a market oriented economy. This process generates major losses and gains, and the distribution of these losses and gains among different social groups has far-reaching influence on further reform of political and economic institutions. Economic institutions determine the incentives of and the constraints on all actors in an economy, and consequently become the fundamental determinants of economic development (Acemoglu et al. 2005). There is a conflict on the choices of economic institutions since different social groups benefit from different economic institutions. The social decisions on economic institutions are generally in favor of groups with greater political and economic power. The chosen economic institutions in turn determine the future distribution of economic resources and de facto political power, which has important influence on further reform of political and economic institutions. The role of the ruling elites in the income distribution is particular important. This social group consists of the high officials who have political power to determine economic institutions at the beginning of the transition process and the rank-and-file officials who implement government policies. Many have argued that whether a country can have a successful transition depends on whether this social group has continuous incentives to adopt and support economic institutions conducive to economic growth (e.g. Che and Qian, 1998; Frye and Shleifer, 1997; Hellman, 1998; Morduch and Sicular, 2000). If the ruling elites benefit at early stages of the transition process and further institutional changes potentially threaten their vested interests, they may even resist reforms that are conducive to economic growth. One example for this is the stop of reforms in Russia in the 1990s (Frye and Shleifer, 1997; Glaeser et al. 2003). Drawing on experiences of China in the early 1990s, Che and Qian (1998) 2

3 suggest the existence of an alternative scenario. When the ruling elites enjoy political rents that are positively associated with the pace of reform, and those rents have little impact on overall inequality, a politically and economically self-reinforcing reform becomes feasible. Using a data set for a representative rural county in northern China, Morduch and Sicular (2000) find a situation that is consistent with a politically sustainable reform process. Their analysis reveals the existence of substantial political rents associated with village cadres and officials. However, village cadres and officials do not benefit at the expense of others over time. During the four-year examining period, average household incomes in the county have grown over time and the impact of political rents on the overall level of measured inequality is minor. When everyone gains in the reform process, a population will be more willing to accept political rents. This situation is therefore equivalent to the existence of a coordinated (implicit) incentive contract that promotes ongoing reform. The political rents provide incentives for village cadres and officials to support market oriented reforms. There are several concerns. First, while the county examined in Morduch and Sicular (2000) is in many regards typical, China is a diverse country and so can their results be extended to other regions? Secondly, given the very different nature between urban and rural areas in China, do the mechanisms observed from a rural county also exist in the urban settings? Finally, over fifteen years has past, whether this mechanism has been maintained in the subsequent structural transformation in China? In this paper, we explore these empirical questions using the China Economic, Population, Nutrition, and Health Survey (CHNS) data sets. The CHNS data covers nine provinces that vary substantially in economic development, geography, and public resources. It allows us to examine those questions in a much larger and more 3

4 representative context. There are seven cycles of the CHNS data available: 1989, 1991, 1993, 1997, 2000, 2004, and The early cycles allow us to compare our results to those in Morduch and Sicular (2000), and the latest several cycles allow us to examine whether the mechanisms have been changed in the subsequent structural transformation. The process of reform in China has begun from the end of 1970s. Before the early 1990s, reform was mainly concentrated in rural areas. Urban reforms proceeded very slowly until the middle and late 1990s. To capture the consequences of acceleration of reforms in urban areas in the past decade, we focus our analysis on the urban samples of the CHNS data. The empirical strategy employed in this paper is slightly different from that in Morduch and Sicular (2000). In addition to the regression analysis on the existence of political rents, we adopt the regression-based inequality decomposition method developed by Wan (2004). This approach allows us to quantify the contributions of various income determinants to the evolution of income inequality in urban China from 1989 to Reduction of income inequality is an important policy objective for the Chinese government over the next few years. The national campaign of western development and the government s commitment to building a harmonious society exemplifies their recognition of this problem. For the purpose of setting policy priority, it is necessary to identify the causes of inequality. In the existing studies that try to decompose income inequality in China (e.g. Morduch and Sicular, 2002; Wan, 2004; Wan and Zhou, 2005; Wan et al., 2007), the authors mainly focus on variables that are related to specific economic policy, economic environments or human capital, few of them consider the institutional causes of inequality. 1 Our analysis explicitly considers the contributions of political rents to overall urban 1 Morduch and Sicular (2002) is an exception. Their study shows that the impacts of political status on overall inequality in rural China in the early 1990s are very small. 4

5 income inequalities from 1989 to The ruling elites can benefit from the reforms at the expense of others or along with others. If we find empirical evidences that the ruling elites grab a larger and larger share of the fruit of reforms over time (the impact of political rents on overall inequality is growing over time), it is consistent with a stalled reform process described above. Alternatively, if we find that the ruling elites enjoy a relatively fixed share of the fruit of reforms over time (the impact of political rents on the over time change in overall inequality is small), it may provide evidences for the existence of the positive scenario a politically and economically selfreinforcing reform process. We find similar mechanisms as described in Morduch and Sicular (2000) for the first half of our examining period: political rents existed, but average urban household income has grown over time and the impact of political rents on the overall level of measured inequality was minor. However, acceleration of reforms in urban areas since the late 1990s has caused fundamental changes: political rents and their impacts on overall income inequalities have become more and more significant. The other contribution of our study is that it provides more updated information on the evolution of urban income inequality in China. Due to data limitation, this kind of information for the period after 2002 is lacking. Based on newly available cycles of the CHNS data, we find that the level of income inequality has risen sharply between 2000 and 2006, and a significant portion of this increase can be attributed to the political rents. The paper is organized as follows. In the next section, we briefly introduce the background of our analysis. Section 3 provides a descriptive analysis of the data. In Section 4, we present the results of regression analysis and inequality decompositions. The last section is discussion and conclusion. 5

6 2. Background China has introduced sweeping reforms in the structure of its economy since 1979, and experienced rapid economic growth in the following three decades. During the period , per capita real GDP increased by 9.8 percent per year. Urban household real income increased by 7.2 percent for the periods (China State Statistical Bureau, 2008). Before the middle 1990s, economic reform was mainly concentrated in rural areas. Privatization of agriculture and encouragement for the establishment of rural collective enterprises greatly improved work incentives, leading to rapid productivity and income growth in rural areas. During this period, urban reform proceeded very slowly. Little had been changed in urban areas in terms of job security, wage determination, and sectoral composition of employment until the early 1990s (Meng, 2004).The most salient feature of the economic planning system instituted in the mid-1950s is a statedominated urban economy. Under the planning system, unemployment was rare and lifetime employment was guaranteed, but little labor mobility was permitted, either geographically or across occupations. By the end of 1980s, about three quarters of all workers are employed in the state sector. The collective sector employed the majority of the remaining workers. The total share of other ownership categories (e.g. private enterprises, self-employed individuals, foreign funded enterprises, and share-holding corporations) in the workforce is less than 1% (Zhang et al. 2005). State sector can be further divided into three sub-sectors: government units, state services/institutes, and state-owned enterprises (SOEs). Before the mid-1980s, all the state sector employees were paid according to a centrally determined wage grade system, which consisted of eight levels for blue-collar workers and 25 to 28 levels for administrative and managerial personnel. Slight adjustments were made for 6

7 regional differences in the cost of living. These wage scales were determined primarily on the basis of a worker s seniority rather than productivity. Although the wage scale permitted differentials by level of education, these wage differentials were very small. As a result, the wage structures of state sector were extremely compressed, both within and between working unit. Non-wage benefits to workers and their families, such as housing, child care, medical insurance, and pensions are provided by their working units. The wage and welfare bill of each working unit in the state sector was in fact paid by the state, and is not directly linked to profitability even for the SOEs. Wages in urban collective enterprises were set approximately equal to wages in the state sector. During this period, income distribution was remarkably egalitarian and changed very slightly. Ravallion and Chen (2007) report that the urban Gini coefficients of household income range from to between 1981 and The first stage of urban reform occurred in the mid-1980s. In October 1984, the Chinese government passed the Resolution on Economic Institutional Reform, by which the SOEs were allowed to keep part of their profits to improve workers welfare benefits or pay bonuses to more productive workers. Moreover, they were granted autonomy to hire contract workers. However, the overall impact of those reforms on economic structures is limited. The second stage of urban reform occurred in the 1990s, which is a decade marked by the profound shifts in industrial and enterprise structure in urban China. Since the early 1990s, non-state enterprises (including private, foreign, joint-venture, and mixed ownership) have emerged as important players in the Chinese economy. The share of labor force employed in the state sector has fallen continuously. Meanwhile, the SOEs have been performed poorly due to soft budget constraints and many other institutional problems. Radical 7

8 reforms for the SOEs were introduced in the middle and late 1990s. To vitalize the Chinese economy, a wave of ownership restructuring for the SOEs occurred in which many small and medium sized loss making enterprises were bankrupted, merged with other enterprises, transformed into joint-stock companies, or privatized. As a result, a large amount of workers in the SOEs were laid off. The reforms for the large SOEs were much more cautious. The policies on this issue were described as control the large and let go of the small in the 15 th Congress of the Chinese Communist Party held in late The government tried to convert the large SOEs into shareholding companies but the shares are mostly held by government organizations to present can be considered as the third stage of reform on the SOEs. During this period, privatization of most small and medium SOEs was completed. For the reforms on the large SOEs, one important policy in this period is to maintain and strengthen the control on strategically important industries such as oil, steel, energy, telecommunications, etc. Under this policy, although small fractions of the shares of some large SOEs are traded in Shanghai, Hong Kong or even New York stock market, management and governance of those enterprises do not change substantially. The State-owned Assets Supervision and Administration Commission of the State Council (SASAC) has been set up in 2003 to control those SOEs on behalf of the state. SASAC appoints and removes top executives, evaluates their performances, and grants them rewards or inflicts punishments under the supervision of the central government. Managers are often politically appointed. It is common to observe that those executives switch roles between government officials and firm managers periodically. 2 Competitions for these SOEs are weak, most of them have enjoyed monopoly power continuously. Moreover, it is much easier for them to get loans from 2 For example, almost one fourth provincial level officials that were born after 1960 have experiences as executives in the large SOEs. See the report at: (in Chinese, available at 20/6/2009). 8

9 state-owned banks and policy supports from various levels of government. Consequently, most large SOEs in China enjoy high profits and several of them have become most profitable firms in the world in the past several years. Board of directors in those enterprises has high degree of autonomy in the setting of wage, bonus, subsidy and welfare programs for their employees. Public discontent on the high remuneration of employees (especially senior executives) in the large SOEs can commonly be observed from reports in mass media. As a response to this public outrage, both the Ministry of Human Resources and Social Security and the Ministry of Finance drafted plans that try to regulate the remuneration system in the SOEs in the early of this year 3. In the latter plan, the annual salary of top executives in the state-owned financial companies will be capped within 2.8 million RMB. The ceiling was regarded too high by many since it is still about 140 times higher than the average employment income in 2008 in urban China. Another feature of the reforms in this stage can be described as guo jin min tui (the state advances, the private sector retreats). A number of private companies emerged in the late 1990s collapsed due to their own illegal activities in the process of privatization, gradually tightened supervisions from various levels of government, and lack of credit support form the state-owned banks. Many of them were taken over by the state sector companies in the process of restructuring. Another important aspect of urban reform is the reform in government units and state services/institutes. Government units at here refer to various levels of administrative organizations. State services/institutes at here refer to non-profit organizations that are fully financed by and accountable to the government. Examples include primary and middle schools, universities, research institutes, news agencies, 3 See the article China plans cap on SOE execs's payroll, by Hou Lei. Available at (20/2/2009): 9

10 health care institutions, etc. Wage structures and welfare programs in state services/institutes were set almost identical to those in the government units in the pre-reform era, and little has been changed on this aspect in the reform process. The size of government in People s Republic of China is historically large in that the government requires a large bureaucracy in managing many aspects of the lives of the Chinese people as never before (Chow, 2006). To improve efficiency and reduce the pressure on the state s budget, Premier Zhu Rongji introduced a program that tried to significantly cut the number of state cadres and reduce the number of administrative organizations during his term of office ( ). This program met strong resistance from central to local governments. The position cuts gave the real power holders incentives to protect their interests collectively. After two rounds of central demands in seven months, twenty-two provinces (out of total thirty-one) were still refusing to cut their staff. Nevertheless, by the end of 1999, four million bureaucrats had lost their posts, and many industrial ministries (bureaus) throughout China had been transformed into industrial associations (Zweig, 2001). This reform succeeded to some extent from the perspective of efficiency as it reduced the size of government considerably. However, its success was limited in term of reducing the state s budget pressure. Most redundant bureaucrats became employees of non-governmental associations (state services/institutes) that are still fully financed by the state s budget. Before the mid-1980s, all the state sector employees were paid according to the same centrally determined wage grade system. The payment scheme for the SOEs was separated from this system in For the government units and state services/institutes, there were eight rounds of adjustment or reform on the centrally determined wage grade system from 1985 to 2006, which occurred at May 1985, 1989, October 1993, July 1997, July 1999, January 2001, October 2001, July 2003, and July 10

11 2006. Meanwhile, local governments gradually have certain powers on the setting of wage and welfare programs for the employees in the government units and state services/institutes in their jurisdictions. 3. The CHNS data and descriptive analysis The China Economic, Population, Nutrition and Health Survey (CHNS) is an international collaborative project conducted by the Carolina Population Center at the University of North Carolina at Chapel Hill, the National Institute of Nutrition and Food Safety at the Chinese Center for Disease Control and Prevention. It is an ongoing survey and seven cycles of the data are currently available: 1989, 1991, 1993, 1997, 2000, 2004 and The CHNS covers nine provinces in China: Guangxi, Guizhou, Heilongjiang, Henan, Hubei, Hunan, Jiangsu, Liaoning and Shandong. The covered provinces vary substantially in geography, economic development and public resources. The survey follows a multistage, random cluster sampling design. Consequently, the data contain much of the social, economic and institutional variation that present in China. Although the CHNS was designed to examine the effects of health, nutrition, and family planning policies and programs implemented by national and local governments, it contains detailed information on the household and individual economic, demographic and social factors that are necessary for our analysis. Currently, there are about 4,400 households in the overall survey which cover some 19,000 individuals; about one third observations are drawn from urban areas. In our study, we will focus on the urban sample of each cycle of the CHNS data. [Table 1] Table 1 provides some basic statistics on the evolution of per capita household income from 1989 to We divide the whole sample into two groups: state sector 11

12 and private sector. State sector refers to those households that contain at least one member employed in the SOEs, government units, or state services/institutes. Private sector includes all the other households. Questions on income and time allocation in the CHNS probe for any possible activities that a person might have engaged in during the previous year, both in and out of the formal market. Inclusion of nonmonetary subsidies (e.g. state-subsidized housing, welfare benefits, and food coupons, etc.) is an especially important advance. Total income from market and non-market activities can be imputed. Household income in our study is defined as the sum of all market and non-market earnings and various subsidies across the household. Total household incomes are deflated by using region/year specific deflators. The regional deflators are not provided in 2006 data, so we use the 2004 deflators as proxies. The second column of Table 1 shows that share of households in the state sector had fallen continuously from 1989 to The shrinkage coincides with the process of reform in urban China. The sharp fall of state sector employment in the late 1990s and early 2000s reflects the restructuring of the SOEs during this period. Average per capita household income for our sample increased considerably from 1989 to 2006 but varied from period to period. The average annual growth rate is about 6.4%, which is slightly lower than but still close to the number in China State Statistical Bureau (2008). Variations of income growth between periods may reflect the hyperinflation in and the layoffs of millions of workers in the SOEs at the end of 1990s. The statistics in Table 1 indicate that households with state sector employees enjoyed both higher than average income and higher than average income growth rate. Calculation based on the fourth and sixth column shows that from 1989 to 2006, the average per capital household income for the state sector and private sector had increased at 8.9% and 5.5% respectively. While the income gap between 12

13 the two sectors was relatively small in 1989, the household income for the private sector was just about half of that for the state sector in Moreover, the state sector continuously enjoyed the higher income growth rate in recent years, and that is not the case for private sector. Between the last two time points in our examining period, we can observe that the increase of household income for the private sector is negligible, but the state sector s income still increased significantly. As we mentioned in last section, the wage setting mechanism for the SOEs, government units and state services/institutes has been separated since the end of 1980s. The wages in the latter two sub-sectors have still been determined by a state-controlled grade system. The last two cycles of the CHNS data contain more detailed information on the occupation of respondent, by which we can identify the average per capita household income for those households with members employed in government units and state services/institutes. The fifth and eighth column show that both the average income and income growth rate for this social group were even higher than that for the state sector as whole. [Table 2] Table 2 presents the results of several most commonly used income inequality measures for our sample from 1989 to It is clearly shown that household income inequality increased during this period. The trend of change in inequality from 1989 to 2000 is consistent to the findings in Ravallion and Chen (2007), which use the Urban Household Surveys of China s National Bureau of Statistics that cover all Chinese provinces. 4 Thus, it is fair to say that our result is sufficiently representative of the national trend. Due to data limitation, the information on the evolution of 4 This data is not public available. 13

14 income inequality in urban China after 2002 is lacking. Our results indicate that the level of income inequality has risen sharply between 2000 and The results in Table 1 preliminarily indicate that rents to the households in the state sector might be a source of the rapid increase of income inequality in recent years. However, we are more interested in the contribution of political rents to the ruling elites to this increase. State sector consists of the ruling elites of different rank and ordinary workers. Rents to the households in the state sector just partially reflect the rents to the ruling elites. In the process of economic transition, the ruling elites may arrange favorable economic institutions to the social group that they belong to. The consequence of such arrangements can be captured by the inequality between the state and private sector. At the same time, ruling elites of different rank may allocate resources according to their political powers within their social group. Part of the income inequality in the state sector can be attributed to this resource allocation process. The rule of rewards differs considerably between the state and private sector, with the latter being much more market-oriented than the former. Both Zhang et al. (2005) and Dong (2005) find that the allocation of labor and determination of wages have been mainly driven by market forces in the private sector since 1990s. If we assume the within-group inequality in the private sector (the inequality mainly driven by market forces) as fair inequality and use it as a benchmark, then the difference between the inequality in the state sector and that in the private sector can be considered as unfair inequality (the extra inequality caused by non-market forces). The contribution of political rents to the ruling elites to the overall inequality can be calculated as a weighted sum of the between-sector inequality and the unfair inequality. Of course, the two sectors may have many different characteristics that 14

15 lead to different level of fair inequalities. We will control those characteristics for each sector in the regression analysis in the next section. [Table 3] Table 3 presents the results of unconditional inequality decompositions between the state and private sector from 1989 to The inequality index employed is the Theil entropy measure, which is perfectly decomposable between sub-groups. 5 The third and fourth column show that both the between-group and within-group component of the overall inequality have increased continuously over time. However, the rate of increase of T is greater than that of T. The contribution b of the between-group component to the overall inequality was negligible in 1989, but it became a significant contributor to the overall inequality in This trend reflects that the rent of working in the state sector has increased over time. This rent might be related to the monopolistic profits of the SOEs and the fast wage increases for employees in government units and state services/institutes in recent years. The w decomposition weights used in the calculation of T w also reflect this trend. The weight is defined as the ratio between total income of a given sector and that of the whole population. While the share of households in the state sector was 67.6% in 1989, the income share for this social group was 70.9%. When the share of households in the state sector dropped to 28.4% in 2006, their income share only dropped to 48.7%. We also report the within-group inequalities for each sector. Before the end of 1990s, the compressed wage structures in the state sector lead to an income distribution that was remarkably egalitarian. However, after the radical 5 The overall inequality T can be decomposed as T Tb Tw Tb wt s s wptp, where T b measures the between-group component of the overall inequality, T w measures the within component of the overall inequality, T and T are measures of inequality in the state and private sector respectively, p s and w and p w are the decomposition weights attached to each sector. s 15

16 restructuring of the SOEs and intensified wage reforms in the government units and state services/institutes, the inequality in the state sector has increased dramatically in recent years. [Table 4] To have a better understanding on the over time evolution of inequality, we present a percentage decomposition on the increase of inequality for two selected time intervals in Table 4. The first time interval is , in which the urban reforms proceeded relatively slow. During this period, restructuring of the SOEs did not start and the wages for government units and state services/institutes had been adjusted only twice. In the second time interval, , radical reforms for the SOEs and government organizations had been completed and the wages for government units and state services/institutes had been adjusted six times. If we define the increase of inequality over a time interval as 100%, the third row of Table 4 shows that the contribution from the between-group inequality to the increase of inequality from 1989 to 1997 is only 6.9%. The increase of inequality in the first time period was mainly caused by the increase of the within-group inequality in the private sector. Contribution to the change in within-group component of the overall inequality from private sector (0.0718) is even greater than Tw (0.0684) itself. If we define T w as 100%, the contributions to T w from private and state sector are 105% and -5% respectively. During this period, non-state enterprises have emerged as important players in the Chinese economy. Allocation of labor and determination of wages have been mainly driven by market forces in the private sector. Zhang et al. (2005) show that returns to education and experience had increased steadily during this period, which certainly had driven up the income inequality. On the other hand, the reforms in the state sector were relatively mild, the still compressed wage structures therefore 16

17 generated equalizing effect on the overall inequality. Things changed radically in the second time period. Contribution from the between-group inequality to the increase of total inequality increased considerably to 14.3%. The major contributor to the withingroup component of the overall inequality became the within-group inequality in the state sector. Restructuring of the SOEs and the wages reforms for government units and state services/institutes had increased both the inequality between private and state sector and the inequality within the state sector. In the last section of Table 4, we report the policy (political rent) contribution to the change in income inequality. It is defined as follows. Income inequality in a given time point t can be written as: T t T t wt t t w t T t T t w t ( T t T t ) wt t t w t T t p t T t (1) p b s s p p b s s p s p p p The reward system in the state sector has been mainly determined by government policies rather than market forces during the whole examining period. Therefore, the between-sector income inequality t T b and the difference between the within-sector inequality in the state sector and that of the private sector can be considered as the results of policy distortions. We define this part of inequality ( p t T t w t ( T t T t )) b s s p as policy contribution to the overall income inequality. The policy contribution to the change in income inequality over time is defined as the difference in t p between two time points. Table 4 shows that the policy contribution to the change in income inequality in the first time period is negligible. However, it accounts more than one third of the increase of income inequality in the second time period. The above analysis provides preliminary evidences to support the conclusion that the ruling elites have extracted more and more political rents in the process of reforms. They have grabbed larger and larger share of the fruits of the reform. However, questions were naturally raised for the above analysis. As shown in Table 5 17

18 in the next section, households in the private and state sector have many different characteristics, and those characteristics had changed over time. If we control those characteristics, does the conclusion still hold? This is the question that we try to address in the next section. 4. Regression-based inequality decomposition analysis [Table 5] The inequality decomposition method employed in the last section is unable to control other factors that determine household income. Table 5 shows that households in the private and state sector have many different characteristics. Generally speaking, households in the state sector are younger and better educated. Moreover, they tend to have slightly larger household size, higher male proportion of adults, and fewer very young children. These differences may lead to a justifiable income gap between the state and private sector. One way to isolate the contribution of rents to the households in the state sector to the overall inequality is the regression-based inequality decomposition method. Researchers have used regression-based approaches to investigate income inequality since the early 1970s. However, most of the early studies focus on explaining the differences in income between distinct groups (e.g., Oaxaca, 1973; Blinder, 1973). Latterly, methods to quantify the contributions of specific factors to total inequality were developed by Fields and Yoo (2000), Morduch and Sicular (2002), and Wan (2004) 6. In our analysis, we adopt the Shapley value decomposition method (Wan, 2004) to identify the contribution of political rents to the overall inequality. The advantage of this method is that it can employ non-linear regression 6 For detailed comments on the advantages and shortcomings of those methods, please refer to Wan (2004). 18

19 equation and any inequality measures. For detailed description of this method, see Shorrocks (1999), Wan (2004), and Wan and Zhou (2005). The first step of the decomposition is to specify an income generation equation. We adopt a semi-log functional form by following most empirical studies in human capital theory. The dependent variable is log of per capita household income. Our interested variable is the number of household members employed in the state sector. Detailed definition of control variables and regression results are presented in Table 6. [Table 6] In view of space considerations, we just present the results for three selected time points: 1989, 1997 and As discussed in last section, we choose 1997 as the midpoint since it is a critical year in the process of the urban reforms. One key finding from the regression results is that education and skills play more and more important role over time in income determination. This result is consistent with many previous studies on the trends of changes in income in China (e.g. Zhang et al., 2005; Appleton et al., 2005). The link between the state determined wage grade system and factors of human capital was weak in the pre-reform era. Part of this policy distortion has been corrected in the reform process. Disparities in education and skills therefore become important contributors to the increase of income inequality. Secondly, we find that regional variations in income still exist but become less significant over time. Thirdly, some factors that are not commonly considered in previous studies have considerably influences on household income determination. For example, the negative effects of poor health and raising very young children on household income increase over time. Finally, the influence of our interested variable number of household members employed in the state sector on household income has steadily increased even after 19

20 we control various factors of income determination. This result indicates the existence of rents of working in the state sector. Based on the regression results, we decompose the overall inequality into its causes using the Shapley value decomposition method developed by Wan (2004). The results in value and in percentage are reported in Table 7 and Table 8 respectively. [Table 7] [Table 8] Since different inequality measures are associated with different social welfare functions and may lead to different decomposition results, we consider four most commonly used indices in our study for generality: Gini coefficient, Theil entropy measure, Theil mean log deviation, and Atkinson index. Independent variables in income regression equation are grouped into several categories. For example, the contribution of age to total inequality in Table 7 and Table 8 combines the contribution of average age of adults and that of squared average age of adults to total inequality. As expected, different inequality measures yield different decomposition results. Nevertheless, the rankings of contributors change little from one indicator to another. Generally speaking, we can observe three patterns from Table 7 and Table 8. First, disparities in education and skills (occupations) can explain a significant proportion of overall income inequality at all the three time points in our study. The combined effect of the two variables had increased over time. Secondly, consistent with most previous studies, the increase of income inequality in the 1980s and 1990s was associated with large regional variations in income. However, the unequal development among provinces was ceased to be a significant contributor to the sharp increase of income inequality in the past decade. Finally, the contribution of between-sector inequality to the overall inequality has increased 20

21 steadily over time. For the case of Theil entropy measure, the contribution of rent to workers in the state sector to the overall inequality in 2006 is 8.4%. Compared to the results in Table 3, that means even after we control other determinants of income, the rent to workers in the state sector is still an important source of total income inequality. As we argued in the last section, the between-sector inequality is just part of the rents to the ruling elites. The ruling elites of different rank can arrange economic rules that are favorable to themselves according to their relative political powers within the state sector. Since an individual s relative political power can be approximately measured by his or her rank in a sector, if we assume that the contribution of individuals rank to total income inequality in the private sector is mainly determined by market forces, then the difference in this term between the state and private sector can be considered as the result of policy distortions. That is the contribution of political rents to the ruling elites to the total inequality. Unfortunately, we don t have very detailed information on individual s rank in our data. However, we can use information on individual s occupation to construct proxy measures. The CHNS data contain the following occupation categories: senior professional, junior professional, administrator/executive/manager, office staff, skilled worker, nonskilled worker, army and police officer, ordinary soldier and policeman, driver, service worker, athlete/actor/musician, other. We group them into the following categories: administrator/executive/manager, senior professional, junior professional, office staff, skilled worker, non-skilled worker, other. 7 Each of those categories 7 Army and police officer are grouped into the category of administrator/executive/manager; driver is grouped into the category of skilled worker; ordinary soldier and policeman and service worker are grouped into the category of non-skilled worker; and athlete/actor/musician is grouped into senior professional. 21

22 certainly has different influence on the mechanism of wage setting. We therefore use them as proxy measures of individual s rank. We estimate the income generation equation for the state sector and private sector separately at the three time points: 1989, 1997 and The income generation equation includes those control variables specified in Table 6, which can take into account the different household characteristic in the two sectors. In view of space consideration, we don t report the regression results at here. Next we use the Shapley value decomposition method developed by Wan (2004) to isolate the contribution of individuals rank to the within-sector inequality for each sector. The differences in this term between the two sectors are reported in Table 9. [Table 9] Table 9 shows that in 1989 and 1997, the contributions of occupation rank to the inequality in the state sector are smaller than that in the private sector. However, it is clearly shown that occupation rank becomes a significant contributor to the inequality in the state sector in Its absolute contribution is much higher than that in the private sector, which is mainly determined by market forces. For the case t of Theil entropy measure, if we substitute those conditional terms ( T b in Table 7 and T t s T in Table 9) into the equation ( p t T t w t ( T t T t )) that we used in the last t p b s s p section to calculate the policy contribution to total income inequality, the results are (-3.8%), (1.8%), and (12.9%) for 1989, 1997 and 2006 respectively. To compare with the unconditional inequality decomposition, we present the controlled policy contribution to the change in income inequality over time in the last column in Table 4. The results show that even after we control various household characteristics, the impact of political rents to the ruling elites on income inequality still steadily increase over time. 22

23 Standard cross-sectional estimates may be biased in the presence of selection bias. While political rents appear to exist in the state sector, the positions in the state sector may be not allocated randomly. One may argue that households in the state sector have unobserved characteristics such as higher abilities. Therefore, the political rents could just reflect selection bias. Fixed effects estimates derived from panel data can control for unobserved, time-invariant household characteristics. We use the longitudinal components of the CHNS data to conduct a robust check for the above analysis. During this period, we observe a sharp increase of political rents. We estimate the income generation equation with pooled OLS, random-effect, and fixed-effect model. The results are presented in Table 10. [Table 10] Table 10 shows that the pooled OLS estimates are very similar to the random-effect estimates. There are little variations between the results in column 2 and column 4. Low over time variation of some variables increases the standard errors on the estimates in the fixed-effect model. However, the coefficients of the key variables are close in size to the results from the pooled OLS and random-effect regressions. Hausman test ( 2 2 x (15) 17.18, prob x ) indicates that we cannot reject the hypothesis that the random-effect coefficients do not differ systematically from coefficient estimated by the fixed-effect model. We therefore have reason to believe that the analysis in this section do not suffer from the selection bias problem. 5. Discussion and conclusion In this study, we examine the trend and causes of income inequality in urban China from 1989 to 2006 a period associated with radical economic restructuring and reforms in the urban areas. One key finding is that urban income inequality 23

24 increased dramatically in the first half of this decade. Maybe due to data limitation, studies that report this trend are few. Decomposition analysis shows that the main causes for the mild increase of income inequality from 1989 to 1997 are disparities in education and skills, and regional variations in income. This finding is consistent with the results in most previous studies on income inequality for this time period. While disparities in education and skills remain to be key contributors to the sharp increase of income inequality from 1997 to 2006, the influence of regional variations in income becomes weaker. The most important finding of our study is that political rent has emerged as a more and more significant source of income inequality. The unconditional decomposition results show that 36.5% of the increase of income inequality from 1997 to 2006 can be attributed to political rent. Even after we control for various determinants of household income, political rent still contributes 19.3% to the increase. 8 It seems that the ruling elites have taken great advantages from the restructuring of the SOEs and government reforms occurred in this period. The latter finding is important for two reasons. First, further reforms (both political and economic) might be constrained when the ruling elites responsible for institutional changes have vested interests. As argued in Chow (2006), one important hindrance to further reforms is the bureaucratic behavior and corruption of government officials. 9 The basic solution to this problem is to reduce the size and power of the government. In the existence of the vested interests as observed in our study, such reforms will be strongly resisted by all the employees in the public sector. The policy supports from government and the resulted monopoly profits for the large 8 The political rents attract more talented worker to work in the state sector. Implications of this distortion deserve some further studies. 9 The reported household income in our data may not be a good measure of economic well-being for the ruling elites. Corruption income, immeasurable job security and welfare benefits (e.g. better health insurance) in the state sector may have significant implication to economic inequality, which may need to be further studied. 24

25 SOEs discourage the incentives on technology innovation and encourage rent-seeking. High rewards to the managers in the SOEs make reforms on corporate governance to be difficult. Moreover, the high tax burden and cost burden associated with the large state sector significantly constrain the rigor of private enterprises in the economy. 10 Secondly, high level of income inequality generates widespread social discontent, especially when the poor believe that inequalities do not originate from productive efforts and market competition. The results in our study provide a possible explanation for why the discontent on income inequality was less severe in the 1990s than now. In the presence of unfair inequality, the poor have a much stronger moral justification for the demand of redistributive policies, which are related to more progressive income taxation, extensive regulation, and large social welfare programs. 11 The ruling elites may prefer to do so because it will provide more corruption and rent-seeking opportunities. More corruption and rent-seeking will further increase the level of income inequality. This may lead to a vicious cycle where large government size, high levels of government intervention, corruption and rentseeking, and high levels of income inequality are self-sustained in perpetuity (Alesina et al., 2005). To maintain the rapid economic growth in the past three decades, the Chinese government needs to conduct further reforms on economic institutions by enhancing market efficiency, eliminating corruption, maintaining the rigor of private enterprises, encouraging technology innovation, improving corporate governance, and enforcing 10 Prices of many inputs are controlled by the large SOEs, such as oil, electricity, etc. The monopolistic price increases firm s cost. 11 The government s commitment to building a harmonious society was emphasized during the two most important national conferences on the PRC s political calendar, the National Party s Congress and the National People s Congress (Held in October 2007 and March 2008). This might be a response to the social discontent from the central government. After the two conferences, the attempts on establishing a national health care insurance, enlarging old age security, expanding the social welfare program to the poor have begun. 25

26 property right protection. There is a conflict on the choices of economic institutions since different social groups benefit from different economic institutions. The social decisions on economic institutions are generally in favor of groups with greater political power. Therefore, further economic reforms may need political institutional reforms that create effective constraints on power-holders, and ensure relatively few rents to be captured by them. 26