The effect of the state sector on wage inequality in urban China: 1988–2007

This paper examines the effect of the public sector and state-owned enterprises (SOEs) on wage inequality in urban China using China Household Income Project data. It applies quantile regression analysis, the Machado and Mata decomposition to investigate how urban wage inequality was affected by the changes in wage structure and employment shares of the public sector and SOEs. We find that since the radical state sector reforms designed to reduce overstaffing and improve efficiency in the late 1990s, urban wage gaps were narrowed due to the reduction in the employment share of the state sector; the wage premium of the state sector in comparison with the non-state sector increased significantly; and changes in the wage structure of the labour market caused the rise in urban wage inequality.


Introduction
Although the Chinese economy has been transformed immensely, it is still dominated by the state bias towards the public service sector and state-owned enterprises. The bias is indicated by the share of state sector employment and related wage setting. In the past decade, there has been widespread concern that industrial wage gaps, particularly between the state and non-state sectors, have been increasing in China.
General equilibrium theory suggests that certain specific disequilibria might exist in some parts of a perfect market economy in a state of general equilibrium. For instance, there might be industrial profit and wage differentials. The reason for this is that profit and wages are normally higher in high-tech sunrise industries than in low/conventionaltech sunset industries ceteris paribus. Therefore, under such circumstances, even if the wage gaps are substantial, there is no need to directly restrict the pay in the highly paid industries although the government could tax high salaries and set a minimum wage, since the market will eventually equalize the wage differentials between different industries by the flows of investment and workers from the sunset industries to the sunrise ones. China is arguably different. The transition of the Chinese economy is characterized by an ongoing less well-functioning market system and strong government intervention. The state bias has been reinforced with a launch of a series of economic policies in China starting from mid-1990s and reinforced since 2002.
State-dominated policies have allowed the Chinese state to monopolize prices, retain monopolistic profit and set-up higher wage payments to state sector employees than their marginal product of labour or market clearing price. In fact, the central government of China from 2002 has established tight control of resource allocation, allowing state sector monopolization of key economic sectors (finance and banking, transportation and communication, and energy sector) and has created the taxation allocation system between central and local governments. This has strengthened central control over resource income (apart from local land), but left insufficient tax income for local governments. Those economic policies have safeguarded the political goals of relentless pursuit of GDP growth and this has continued up to the change of party leadership in 2012.
The aims of this paper are to examine whether state-biased policies have trigged the rise in urban wage inequality over time, and whether across the distribution, wage gaps are affected by the state sector domination. We begin by providing a brief overview of the change in China's state power. The power of the Chinese state sector has differed in strength since economic reform commenced. Power was decentralized in the first decade of the reforms, but then strengthened and made more centralized since the mid-1990s. The relationship between the central government and the state sectors affiliated to it during this time may have affected the payment of its employees.
We therefore hypothesize that the high pay in the public sector and monopolistic State-Owned-Enterprises (SOEs) is one of the forces responsible for the increase in the urban wage gap. To test this hypothesis, we employ the 1988, 1995, 2002 and 2007 China Household Income Project (CHIP) urban household survey data and Machado and Mata's (2005) parameterized counterfactual decomposition method based on multiple quantile regressions to identify the effect of the change in pay structure and employment shares in the public sector and SOEs on urban wage inequality. In this paper, the state sector consists of the SOE and public service sector which includes government offices and state-funded institutions such as schools, universities, hospitals, etc. while the non-state sector covers urban collective, private, foreign and other unidentifiable firms.
The rest of the paper is structured as follows. In Section 2, we introduce the data and econometric methods. Sections 3 and 4 give the regression results and decomposition of wage income inequality, respectively. Section 5 presents the summary and conclusions.

Data
We use 1988, 1995, 2002 and 2007 urban household survey data from the CHIP. The surveys were designed by a team of international scholars including the authors and researchers at the Institute of Economics of the Chinese Academy of Social Sciences. Subsamples were drawn from the larger annual national household income survey of the National Bureau of Statistics. The subsamples cover 10 out of 31 provinces in 1988, 11 in 1995, 12 in 2002 and 9 in 2007. The questionnaires designed for CHIP are more detailed than those in the official income surveys, particularly with respect to the measurement of income and labour issues. For the cross-sectional analysis, we construct a real wage variable that includes bonuses, price subsidies (which were important in 1988 before being largely withdrawn), regional allowances for working in Tibet or in mountainous areas, income in-kind and income from secondary jobs. 1 With respect to adjustment of the price level, we adjusted all wage income from all years to 2002 constant prices according to the urban consumer price indices published by China's National Statistical Bureau. Results from these surveys are in Griffin and Zhao (1993), Riskin, Zhao, and Li (2001), Li and Sato (2006), Gustafsson, Li, and Sicular (2008) and Li, Luo, and Sicular (2011).
These surveys cover only households with urban registration (hukou). Consequently, we exclude rural-urban migrant households because they are denied urban hukou status. However, estimating the wage functions of urban residents separately from those of migrants is appropriate because, as administrative controls make it extremely difficult for people of rural origin to acquire urban hukou, any sample selection bias is likely to be negligible. Confining the analysis to the subpopulation holding urban hukou allows us to examine what causes the increase in wage inequality for a specific group of people so that we may draw inferences about corresponding changes in economic wellbeing. Nonetheless, we are omitting an important dimension of the urban labour market by not being able to include migrants. Moreover, the potential importance of this omission has increased over time with the sharp increase in rural-urban migration during the sample period. Controls over rural-urban migration were relaxed significantly in 1988 when the government allowed farmers to conduct business in cities, as Linge and Forbes (1990) discussed. The rise in rural-urban migration is likely to have affected particular groups of urban workers differentially. Specifically, rural-urban migration is likely to have had a moderating impact on the wages of urban residents that have similar characteristics as, or working in similar sectors to, migrants. Hence, the effect is greater on urban workers with less education and those working in the service and commercial sectors. 2

Methods
We employ the quantile regression method to estimate the extended Mincerian earnings function (Mincer 1974). Let Q θ (w it |X it ) for θ ∈ (0,1) denote the θth quantile of (log) wages w of an individual i in year t for given explanatory variables, X. For each year, we model these conditional quantiles separately by: where β(θ) is a vector of quantile coefficients, and X is a vector of explanatory variables. The coefficients are estimated following Koenker and Bassett's (1978) quantile regression estimator. In practice, we run 19 quantile regressions (from quantile points 0.05, 0.10, 0.15, …, 0.95) for each of the four rounds of cross-sectional data. 3 Afterward, we plot a curve for the 19 coefficients on the dummy variable (with 1 indicating a worker employed by the state sector and 0 by the non-state sector) against the 19 quantile points of the wage distribution for each year. From these curves, we can observe the wage premium of the state sector vs. the non-state sector across the entire wage distribution over time.
One of our main purposes of using quantile regressions is to study the evolution of pay differentials between the state and non-state sectors for the period from 1988 to 2007. To this end, we control for the variables that apply across all four years of CHIP urban household survey data in the function (1). In detail, these variables are workers' schooling, experience and experience squared, dummy variables for sex, Communist Party membership, non-Han Chinese ethnicity, job characteristics in terms of occupations and industrial sectors, and provincial dummies. The variable of interest in this paper is a dummy variable identifying a worker's employer with 1 indicating the state sector and 0 the non-state sector.
The focus of this paper is on exploring the effect of the state sector's wage structure and employment share on urban wage inequality for the period from 1988 to 2007. For this reason, we employ Machado and Mata's (2005) method (MM method hereafter) to decompose changes in wage inequality into changes attributable to two sources. One is the change in wage structure in terms of the coefficients on the various explanatory variables. The other is the change in the distribution of explanatory variables, i.e. the change in workers' personal and productive characteristics, and in job characteristics. In detail, following Machado and Mata (2005), if α(.) are summary statistics for wagessuch as the Gini coefficientthen we can decompose the changes in α as follows: (2) where f (w(t)) denotes the estimator of the marginal density of w (the log wage) at t based on the observed sample {w i (t)}, f *(w(t)) denotes the estimator of density of w at t based on the generated sample {w i * (t)}, and t = 0, 1. The counterfactual densities are denoted by f *(w(1); X(0)), for the density that would result in t = 1 if all covariates had their t = 0 distributions, f *(w(1); X i (0)), for the wage density in t = 1 if only X i (part of the covariates) were distributed as in t = 0.
Furthermore, the contribution of an individual covariate x i to the total wage inequality could be measured by looking at indicators such as: Along the lines of MM method, we are able to counterfactually measure the contribution of an individual coefficient β i to the change in wage inequality by observing: denotes the estimator of the density of w with all covariates at period 0 and all coefficients, apart from β i (1), from period 0; β i (1) denotes the coefficient of x i from period 1. With formula (4), we then counterfactually analyse the change in the wage inequality and wage gap caused by specific changes in the pay structure, such as changes in the returns to education. In essence, Machado and Mata's counterfactual decomposition is an extension of Oaxaca's (1973) decomposition for quantile regressions. 4 A key exercise of the MM method is to obtain the generated sample {w i * (t)}. To obtain {w i * (t)}, one first needs to estimate n quantile regression coefficientsb t ðh i Þ (where θ i denotes the quantile point), and then generate a random sample of size n with replacement from the rows of X(t) denoted by fx Ã i g n t¼1 , and finally get w Ã i ðtÞ ¼ x Ã i ðtÞ 0b t ðh i Þ n t¼1 . 5 For details, the reader is referred to Machado and Mata (2005).
Finally, because China's economic transition has occurred gradually, the counterfactual decomposition is implemented period-by-period. In detail, these periods are 1988-1995, 1995-2002 and 2002-2007, based on the availability of the CHIP data. The same explanatory variables were applied in all the four years' earning functions.

Results from quantile regressions
In this paper, the state sector consists of two subsectors. One is the public sector, which covers civil servants, various state-funded institutions such as schools, universities, hospitals, etc. The other is the SOE sector. Therefore, before examining the wage gap between the state sector and non-state sector, it is necessary to investigate the pay differential between the public sector and SOE sector.
According to the CHIP urban household surveys, the public sector's employment share rose from 30% in 1995 to 32% in 2007, whereas the SOEs' share fell sharply from 51% in 1995 to 34% in 2002 and further to 18% in 2007 (see Table 1). The fall of employment for SOEs for the period from 1995 to 2002 was the result of the mass retrenchment of SOE workers in the 1990s, but why is there a further fall during the period from 2002 to 2007? The creation of a State Asset Supervision and Administration Commission (SASAC) in June of 2003 marked a significant change in the state administration of SOEs. One of the major changes is the total wages and salaries contract regime, by which in each year SASAC will negotiate with each of the centrallycontrolled SOEs over the amount of wages and salaries for next year. Soon this practice spread to the whole country. The total wages and salaries contract regime effectively restrains the unreasonable pay rise of SOE workers and increase of wage cost. However, the SOEs are reluctant to employ new workers because of the constraint of the total wages and salaries contract. This might be the reason that the employment share of SOEs had further been reduced in the period from 2002 to 2007.
As for the public sector, the 1993 reform of pay rules for civil servants allowed regions to set-up their own extra-pay scheme, which related the rise in salaries of regional civil servants to local economic growth. In other words, the reform sanctioned an arrangement under which provincial governments could set the salaries for their own civil servants according to their own budgets. Since then, pay differentials for civil servants have gradually been getting larger between provinces or between government agencies in the same region (Dai, Lou, and Wang 2005;Liu 2006). In the period from 1998 to 2002, the pay for civil servants and other public sector workers had been raised thrice (Zhu 2011, vol. 4, 482). Furthermore, the pay for civil servants is higher than that in SOEs, collective enterprises and private firms (Dai, Lou, and Wang 2005). To determine whether there is any pay differential between the public sector and SOEs, we employ CHIP urban household survey data for 1995, 2002 and 2007 in which the public sector and SOEs can be identified by exploring the pay gap between these two subsectors. 6 The OLS regression results reveal that the pay in the public sector was 8% higher than that in SOEs in 1995; however, this pay gap drops to 5% in 2002, and to -2% in 2007 although it is statistically insignificant in the latter. 7 The multiple quantile regression results (see Figure 1) demonstrate that in both 1995 and 2002, the pay gap between the public sector and SOEs steadily decreases as the wage level increases, and the curves for the two years are broadly indistinguishable, except for the top quartile of the wage distribution. For the top quartile, the pay gap is insignificant for 1995; however, for 2002, it declines continuously to -11% as the wage level increases. At 2007, the pay in the public sector is no longer higher than that in SOEs except for the top quintile indicating that the SOE workers are better paid than the public sector workers for quintiles other than the top one. Civil Servants-SOEs wage gap 3.1. Evolution of the wage premium of state sector vs. non-state sector Figure 2 presents the evolution of the wage premium curves of the state sector vs. non-state sector. For 1988, the wage premium of the state sector vs. non-state sector decreases as the wage level increases, with the largest premium of 24% occurring at the fifth percentile of the wage distribution, 14% at the median and 8% at the 95th percentile. In 1995, this trend is further strengthened, in particular, the wage premiums of the state sector vs. non-state sector at the fifth, median and 95th percentiles of the wage distribution increased to 34, 24 and 17%, respectively. This raises several questions. What caused the pay of the state sector to be significantly higher than that of non-state sector in 1988 and 1995? Why was this wage premium further raised in 1995 compared with 1988? And finally, why did the wage premium fall as the wage level increased?
The history of SOE reform might shed light on the above questions. Prior to the economic reform, wages were institutionally determined according to a national system of grades, scales and seniority, with education and skill receiving little reward (Knight and Song 1993, 221-239;Zhang et al. 2005). From 1981, and in particular after 1986, the Chinese government allowed SOEs to implement flexible pay schemes according to their profitability (Dai 1994). Because of the soft budget constraints and no accountability for business failure, the objective of SOE managers was not maximization of profits but maximization of the welfare of SOE workers in terms of wages and bonuses; sometimes bank loans were used to pay large bonuses when the SOE made a loss (Walder 1987(Walder , 1989. In fact, the SOE managers were also incentivized to raise the pay for their workers. The reason for this is that the Chinese government restricted the pay gap between the SOE managers and SOE workers. For example, the government decreed that the pay gap between the SOE managers and SOE workers be not more than two-to three-fold even for those profit-making SOEs in the period from 1986 to 1992 (Chen, Chen, and Wan 2005). Based on 1985-1992 firm data, Meng (2000, 107) found that retained profits were the main determinant of wages in SOEs, whereas in private firms, it was the productivity of workers that determined pay. Therefore, with the exception of frequent increases in bonuses and subsidies of all types, 8 the wage-setting mechanism of SOEs was broadly unchanged until the mid-1990s when the radical SOE reform was implemented. This problem, which resulted from poor institutional design, meant that the pay in SOEs was higher than that in private firms and that the magnitude of this pay differential increased in the period from 1988 to 1995. Although, initially the floating pay or bonus, which had been permitted since the early 1980s, was not allowed to be more than 5% of an SOE's total wage expenditure, this limit was gradually abolished (Meng 2000, 83). The SOE reform of 1986 allowed larger pay differentials within each SOE, but it was difficult to monitor workers' productivity; therefore, floating wages or bonuses were distributed equally among workers (Meng 2000, 83). By contrast, the wage of a worker in a private firm was determined by his/her marginal product (Meng 2000, 107). Therefore, we expect that the wage gap will be much larger in private firms than in SOEs, which this study confirms. Taken together, the pay in SOEs is more or less equally distributed among workers, whereas the wage gap in private firms is large, which accounts for the wage premium of the state sector being larger than that of the non-state sector in the lower part of the wage distribution and smaller in the upper part. Chamberlin (1994) found that the wage premium of trade union members is higher in the lower part of the wage distribution. This suggests that the SOEs in China played a similar role to that of trade unions in the US in terms of wage negotiations; that is the SOEs are more effective in protecting low-wage workers.
After Deng Xiaoping's southern tour in 1992, China's reform and marketization accelerated. A series of reform measures were carried out, such as abolishing the double-tract price system, 9 unifying the tax regime and tightening credit. The most radical reform was the mass retrenchment of superfluous workers and retaining larger SOEs while disbanding smaller ones. There had been serious over-manning and efficiency problems in the SOE sector; were it not for the bank loans and financial subsidies from the state budget, many SOEs would have long been bankrupt. 10 The implementation of the new reform measures exacerbated the difficulties encountered by those loss-making SOEs; for example SOEs had to pay market prices for key resources, faced the same value-added taxes as non-state firms, and could not rely on bank loans to pay bonuses and other subsidies to their workers. Confronted with the ever-increasing number of loss-making SOEs and an ever-rising heavy budget burden caused by subsidizing SOEs, the Chinese government was determined to reform the 120 thousand strong SOEs. The SOE reform was trialed in 1994 and finally implemented in 1997. A consequence of this was that the number of retrenched SOE workers reached 28 million, accounting for half of the SOE workforce (News Office of State Council 2004), and the number of SOEs was reduced (by 74%) from 120 thousand in the mid-1990s to 32 thousand in 2004 (Naughton 2007, 313). A key part of China's SOE reform was to transform the remaining SOEs, which are profitable and monopolistic with strategic importance for the country's economy, into modern market-oriented enterprises. The reform and restructuring of SOEs was followed by soaring wages and bonuses in the SOE sector. 11 Even in the period of mass retrenchment, there was a pay rise for those still-employed SOE workers (Appleton, Song, and Xia 2005;Bai, Lu, and Tao 2006). NCDR (2007) reported that the pay gap between the monopolistic SOEs and other enterprises has been increased largely in the period of 1990-2005.
What effect did these reforms have on the wage-setting mechanism of the state sector for the period from 1995 to 2002? Figure 2 shows that in 2002, the wage premium of the state sector vs. the non-state sector declined to 1988 levels (in the first quintile and the fourth quartile, the premium was even less than that in 1988), and hence was much lower than that in 1995. What factors caused the 2002 wage premium to be so far below the 1995 level? After the Asian financial crisis in the late 1990s, the Chinese economy experienced deflation until around 2002. Additionally, the urban labour market was heavily flooded by the 30 million retrenched SOE workers and the almost 100 million rural-urban migrants. 12 Therefore, there existed no wage inflation pressure in the non-state sector. 13 Any pay rises in the private sector would have been a consequence of productivity change that might be resulted from the unprecedented large-scale inflow of FDI 14 and the improved efficiency of the domestic non-state sector. In summary, the fall of the wage premium of the state sector vs. the non-state sector in 2002 relative to 1995 could only be the result of significant pay rises in the non-state sector during this period.
In 2007, the wage premium of the state sector was still less than that of 1988 over the first quintile of the wage distribution. However, for the second to fifth quintiles, the wage premium of the state sector for 2007 was higher than that for both 1988 and 2002, but still lower than that for 1995. Furthermore, the wage premium of the state sector for 2007 is close to constant for the middle 50% of the wage distribution. In spite of this, over the fourth quartile, the wage premium of the state sector for 2007 decreases as the wage level increases. In general, the wage premium of the state sector increased significantly during the period from 2002 to 2007. What factors could have led to this result? The SOEs' situation changed dramatically after 2002 with the completion of the radical SOE reforms. The remaining SOEs were profitable and monopolistic with strategic importance to the country, such as banks, financial firms, telecoms, and aviation, railway and energy enterprises. According to the National Statistical Bureau, the growth rates of wages in the SOE sector and non-state sector were 14.12% and 12.76%, respectively, during the period from 2002 to 2009, and the SOEs' wage premium relative to non-state firms increased from 0.30% in 2002 to 10.36% in 2009. 15 Gu and Feng (2008), Yue, Li, and Sicular (2010) and Jia (2011), among others, also found that the pay gap between monopolistic SOEs and other enterprises has been expanding. Therefore, Wu (2006) speculated that the expanding wage inequality was caused by the monopolistic SOEs and corruption.

Explained and unexplained components of the wage premium of the state sector vs. non-state sector
Following Yue, Li, and Sicular (2010), we also explore the extent to which the wage premium of the state sector vs. non-state sector is explained. We estimate an Oaxaca-Blinder decomposition 16 on the dummy variable of being employed in the state sector in the extended Mincerian earnings function of this paper for the four rounds of the CHIP urban household survey. The purpose of this is to measure explained and unexplained components of the wage premium of the state sector vs. non-state sector.
The Oaxaca-Blinder decomposition based on OLS regressions shows that the unexplained part of the wage premium of the state sector vs. non-state sector was in the range of 43-44% for the years 1988, 1995 and 2002, but that it climbed to 81% in 2007. 17 Put it differently, the unexplained part of the wage premium of the state sector had been kept at roughly the same level for the period from 1988 to 2002 but shot up sharply in 2007. Yue, Li, and Sicular (2010) found that the unexplained part of the wage premium of the monopolistic SOEs vs. non-monopolistic firms is as high as 60%. 18 Recall that the wage premium of the public sector vs. SOEs decreased as the wage level increased in 1995 and 2002; whereas in 2007, the opposite occurs. As mentioned above, after the reforms, the remaining SOEs were profitable and monopolistic. These monopolistic SOEs could obtain favourable treatment from various government agencies and banks (such as easy credit, lower taxes, right to control scarce resources), set monopolistic prices, harvest monopolistic profits and hence pay their workers' wages well above market prices. This could be the main factor that boosted the unexplained part of the wage premium of the state sector vs. non-state sector.
4. Counterfactual analysis: effect of the wage structure and employment share of the state sector on urban wage inequality Having examined the wage gap between the state sector and non-state sector and its evolution during the period from 1988 to 2007, we now examine how urban wage inequality was affected by the change in the wage structure and employment share of the state sector (encompassing the public sector and SOEs); in so doing, we test our hypothesis that the high pay in the public sector and SOEs caused the increase in urban wage inequality. As stated in Section 2, the change in wage income inequality can be counterfactually decomposed by the change in the wage structure (the change in regression coefficients of the earnings function) and the change in workers' characteristics (the explanatory variables of the earnings function). Through this counterfactual decomposition, we can observe how the wage inequality was influenced by change in any part of the wage structure (any coefficient or any group of coefficients of the earnings function) and in any explanatory variables. In this paper, we focus on how the urban wage inequality was modified by the change in the wage premium of state sector vs. non-state sector (the regression coefficient of the dummy variable 'state sector') and employment share of the state sector, respectively.
In practice, we employ the change in the Gini coefficient and various percentile ratios of the wage distribution to describe how urban wage inequality was affected by change in the wage premium of state sector or the regression coefficient of the state sector (all other regression coefficients and all explanatory variables remain unchanged) and by the change in employment share of the state sector (all other explanatory variables and all regression coefficients remain unchanged) (see Table 3). Percentile ratios of the wage distribution include 90/10, 75/25, 90/50 and 50/10. We carried out 10 rounds of counterfactual simulation to examine the effect of change in the wage premium of the state sector and the effect of change in the employment share of the state sector on urban wage inequality, respectively, and then averaged the wage inequality and gap indicators of the 10 rounds of counterfactual simulation. The purpose of this is to avoid bias from any single simulation result. In each round of the counterfactual simulation, we randomly select 999 observations of the explanatory variables from the data for any particular year.
Effect of the wage premium of state sector vs. non-state sector on urban wage inequality is different in each transitional period of the Chinese economy. Compared with 1988, the sharp rise in the wage premium of the state sector vs. non-state sector over the entire wage distribution in 1995 resulted in an increase in urban wage inequality (see Figure 2, Table 3). For example, the Gini coefficient increased by 0.007. The wage gap indicators such as 90/10, 75/25 and 50/10 showed significant signs of increases in the wage gap. However, the increase in the wage gap is asymmetric because the 50/10 indicator increased while the 90/50 indicator remains largely unchanged. This implies that the wage gap for workers with wages below the median level increased whereas the gap for workers above the median wage level was largely unaffected. Taken together, the sharp rise in the wage premium of the state sector vs. non-state sector for the period from 1988 to 1995 resulted in an increase in urban wage Table 3. MM counterfactual decomposition of the effect of wage structure and employment share of state sector vs. non-state sector on urban wage inequality based on the beginning year's wage structure.

Change of coefficients
Change of covariates 1988-1995 1995-2002 2002-2007 1988-1995 1995-2002 2002-2007 Gini  1995, 1995 to 2002 and 2002 to 2007. inequality and a rise in wage gap for those low wage workers whose wages were below the median level of the wage distribution. The mass retrenchment of SOE workers commenced in 1995 and was complete by around 2002. Compared with 1995, the fall in the wage premium of the state sector vs. non-state sector in 2002 resulted in a fall in urban wage inequality. For instance, the Gini coefficient fell by 0.012. The wage gap (in terms of 90/10, 75/25 and 90/50) also declined. Nevertheless, the fall in the wage gap for workers in the upper half of the wage distribution is larger than that for workers in the lower half. Overall, the fall in the wage premium of the state sector vs. non-state sector for the period from 1995 to 2002 brought about a fall in urban wage inequality and a reduction of wage gap for those high-wage earners whose wages were above the median level of the wage distribution.
During the period from 2002 to 2007, the remaining large monopolistic SOEs seized opportunities and achieved rapid growth, large monopolistic profits and steady pay rises for their employees, particularly their managerial staff. Liao, Zhang, and Liu (2006) and Bai (2008) reports that the implementation of an annual salary system for managerial staff in the SOEs has resulted in the enlargement of pay gap within SOEs, for instance, the annual compensation for SOE managers is 9-10 times higher than the average annual wages of SOE workers. Chen et al. (2005) found that the positionrelated consumption of SOE managers was more than 10 times their annual salary. Consequently, the clear rise in the wage premium of state sector vs. non-state sector in the upper half of the wage distribution in this period caused an increase in urban wage inequality. The Gini coefficient was increased by 0.003. The wage gap indicators of 90/ 10, 75/25, 90/50 and 50/10 all increase, although high-wage earners received larger pay rises than low-wage earners. As a whole, the rise in the wage premium of the state sector vs. non-state sector for the period from 2002 to 2007 led to an increase in the urban wage inequality particularly for the upper class whose earnings were above the median of the wage distribution.
For the period from 1988 to 1995, the employment share of the state sector remained largely unchanged so that the inconsistency is not apparent. However, the employment share of the state sector declined from 79% in 1995 to 65% in 2002 and further to 49% in 2007. The fall in the employment share of the state sector during the two periods from 1995 to 2002 and 2002 to 2007 results in a reduction in urban wage inequality using the Gini coefficient and urban wage gap in terms of 90/10 and 50/10.

Summary
This paper examined the effect of change in the state sector's wage structure and employment share on urban wage inequality by employing the 1988, 1995, 2002 and 2007 CHIP urban household survey data. Regarding our methodology, we employed multiple quantile regressions and MM counterfactual decompositions. The results of the multiple quantile regressions and MM counterfactual decomposition revealed that the sharp fall in the state sector's employment share, which was caused by the radical SOE reform in the second half of the 1990s, resulted in falls of urban wage inequality especially for the below median wage workers for the periods from 1995 to 2002 and 2002 to 2007, and the fall in the wage premium of the state sector vs. non-state sector for the period from 1995 to 2002 also gave rise to a fall in urban wage inequality; however, the rise in the wage premium of state sector vs. non-state sector before and after the SOE reform led to increases in urban wage inequality for the periods from 1988 to 1995 and 2002 to 2007. During the SOE reform, the wage premium of the state sector was reduced temporarily, leading to a fall in urban wage inequality for the period from 1995 to 2002. However, the SOEs that survived the reform are large and monopolistic ones, characterized by opportunistic monopolistic profit and payment of high salaries. Therefore, since 2002, the rate of pay rises for the SOE workers has not only been faster than that of the public sector but has also been faster than that of the private sector, which led to an increase in urban wage inequality. The Oaxaca-Blinder decomposition showed that the unexplained part of the wage premium of the state sector vs. non-state sector remained at about 44% until 2002, but then increased sharply to 81% in 2007. The monopolistic SOEs were able to set monopolistic prices, earn monopolistic profits and pay their employees wages that were higher than market prices. These might be the chief reasons that the wages in SOEs increased faster than did those in the public sector and private sector.
More than 30 years has passed since China's economic reforms were initiated in 1978. During these decades, the most spectacular and influential reforms were the abolishment of the collective agricultural regime in the early 1980s and the radical SOE reform in the 1990s. The rural reform, which returned the collectivized land to rural households, resulted in a large rise in agricultural output and a significant fall in rural poverty. Therefore, it was a pure Pareto improvementthere were no losers during the reform. However, the radical SOE reform led to the retrenchment of half of the SOE workforce and a massive reduction in the number of SOEs. The SOE reforms cast off the burden of subsidizing loss-making SOEs which used up the majority of the state fiscal income, and hence laid a solid financial foundation for the new Hu-Wen deal that focused on improving human development conditions of the country. 19 China's transitional market economy is still not perfect. The large monopolistic SOEs still play a dominant role and pay their employees very high salaries and welfare payments, which have led to the rise in urban wage inequality. For this reason, the decision-makers of China should put more effort into monitoring and regulating the monopolistic SOEs.

Funding
The key research project of Philosophy and Social Sciences of Ministry of Education 'a study of increasing the proportion of residential income in the national income distribution' (Grant No. 11JZD015) (2012年教育部哲学社会科学研究重大课题攻关项目《提高居民收入在国民收入 分配中的比重研究》), and ESRC (RES-238-25-0035). Notes 1. Our wage variable, although fairly comprehensive, does exclude some non-monetary benefits such as pension accruals, health insurance and housing. The contributions of these variables may vary under differing forms of ownership and over time. Nominal wages were converted into real wages by deflating by regional urban CPIs. Besides, there are grey incomes of many forms. Due to the limit of information, we are unable to take the grey incomes into account. 2. In the 1999 survey, the more settled migrants were surveyed and so we can compare their characteristics with those of workers with urban hukou (see Table 1 of Appleton et al. 2004). Over half the migrants were self-employed and so may not be directly competing for jobs with urban residents (only around 1% of whom were self-employed). Migrants tended to be