Minimum Wages and Employment in China

Since China promulgated new minimum wage regulations in 2004, the magnitude and frequency of changes in the minimum wage have been substantial, both over time and across jurisdictions. This paper uses county-level minimum wage panel data and a longitudinal household survey from 16 representative provinces to estimate the employment effects of minimum wage changes in China over the period of 2004 to 2009. In contrast to the mixed results of previous studies using provincial-level data, we present evidence that minimum wage changes have significant adverse effects on employment in the Eastern and Central regions of China, and result in disemployment for females, young adults, and low-skilled workers.


Introduction
Since China issued its new minimum wage regulations in 2004, minimum wages have sparked intense debate in the country. There is little doubt that employees generally welcome the minimum wage. However, there is considerably less agreement regarding whether the minimum wage is effective at attaining its goals. The issue, from the time of its introduction, has been highly controversial among scholars and policy-makers. 1 The contentious nature of the minimum wage policy in scholarly work does not allow for its impact to be easily understood. However, the initial evidence seems to show that the magnitude [ Figure 1 about here] Figure 1 shows the nominal and real minimum wage (monthly average) in China from 1995 to 2012 as well as those of the corresponding provinces that raised the minimum wage standards 1 In China, supporters of minimum wages advocate them as a way to assist individuals or families to achieve selfsufficiency, to protect workers in low-paid occupations (Zhang and Deng, 2005;Sun, 2006), to help reduce inequality and to serve as an important safety net by providing a wage floor (Zhang, 2007;Jia and Zhang, 2013). In addition, the higher labor cost may promote managerial efficiency and labor productivity, inducing employers to invest in productivity-improving technology (Cooke, 2005). Along these lines, many Chinese scholars have argued in favor of the more proactive increase of minimum wages (Du and Wang, 2008;Ding, 2009;Han and Wei, 2011). On the other hand, opponents argue that raising the minimum wage can undermine enterprises' dividend policies, reduce China's comparative advantage given the abundance of low-wage labor (Cheung, 2004(Cheung, , 2010, decrease the employment opportunities of low-wage workers, and lead to reduction in the compensation package (Xue, 2004;Ping, 2005;Gong, 2009). Furthermore, rural-urban migrant workers tend to have very low pay and may accept jobs which pay less than the current minimum wage, making it exist in name only (Chan, 2001;Ye, 2005).

2
for each year and its moving average over the same period. Between 1995 and 2003, the average nominal minimum wage increased steadily from 169 RMB to 301 RMB, amounting to a 78 percent growth in 9 years. However, since China issued the new minimum wage regulations in 2004, the nominal minimum wage has increased rapidly by more than 200 percent, reaching 944 RMB in 2012. 2 The real minimum wage grew at a slower pace before 2004 and began to rise thereafter. Furthermore, as shown by the moving average curve in Figure 1, there is an apparent rise in the number of provinces that raised the minimum wage standards in 2004, indicating that the minimum wage adjustment had become more frequent since that year.
How had this regulatory environment affected the labor market outcomes in China? More specifically, did changes in the minimum wages have any impact on employment in the Chinese labor market? Despite the enormous literature documenting numerous aspects of minimum wages and their role in the labor market, there is no consensus on the magnitude of an "average" effect of minimum wages on employment.
Empirically speaking, there are at least three challenges involved in measuring the effect.
First, because provinces and municipalities in China have considerable flexibility in setting their minimum wage according to local conditions, there are often at least 3 or 4 levels of minimum wage standards applicable to various counties in most provinces, meaning that county-or citylevel minimum wage data containing the relevant information on the dates and the extent of minimum wage increase are not readily available. 3 Second, omitted variables and endogeneity 2 The growth rates of average nominal wage are 155 and 194 percent for the periods of 1995-2003, respectively (National Bureau of Statistics of China, 2012. 3 The implementation date of a new minimum wage standard of a county can also differ across geographically contiguous neighbors within the same province. For example, Liaoning Province has the most complicated minimum wage scheme, in which 14 jurisdictions may enact their own standards on different dates. For instance, in 2007, Shenyang, Benxi, Dandong, and Panjin cities did not increase their minimum wages. In contrast, Dalian and Anshan cities increased their minimum wages from 600 RMB to 700 RMB on December 20th, on which day Jinzhou and Liaoyang cities increased their minimum wages from 480 RMB to 580 RMB and Chaoyang city increased its minimum wage from 35 0RMB to 530 RMB. Furthermore, the minimum wages of Fushun and 3 issues (such as the decision regarding the adjustment of minimum wage standards) make it difficult to separate causal effects from effects due to other unobserved confounding factors.
Third, it is difficult to find microdata that can be plausibly representative of the population and may be influenced by the minimum wage increases. Furthermore, some provinces, such as Beijing and Shanghai, do not include social security payments and housing provident funds as part of wages when calculating the minimum wage, making their "official" minimum wage virtually higher. 4 In the paper, we first assess whether and the extent to which minimum wage changes affected the Chinese labor market by measuring the average effect of the minimum wage on employment.
To do so, we begin by analyzing the labor market reaction to changes in minimum wage standards using panel data regressions. The most distinctive feature of our data-crucial for our research design-is the combination of a large county-level panel, which includes all counties in China and contains relevant information on minimum wages, with a longitudinal household survey of 16 representative provinces between 2002 and 2009. The use of county data allows a reasonable representation and more variation in detecting the effects of minimum wages on employment in China. 5 In particular, this feature allows us to directly evaluate the effects on subgroups of the population, especially those who are at risk of being affected by a minimum wage increase, such as young adults, female workers, low-skilled workers, and rural migrant workers.
Huludao cities increased from 400 RMB to 480 RMB on January 1st, whereas that of Yingkou city increased from 380 RMB to 480 RMB, that of Fuxin city increased from 350 RMB to 420 RMB, and that of Tieling city increased from 380 RMB to 420 RMB the following year. 4 In other words, with or without accounting for this issue, the difference can be substantial. For instance, the mean monthly minimum wages in Beijing and Shanghai were 651 RMB and767 RMB in 2004-2009; however, the average expenses of both social security payments and housing provident funds in Beijing and Shanghai are as high as 376 RMB and 452 RMB over the same period, amounting to 58 percent and 59 percent of the nominal minimum wages, respectively. We discuss how we address this issue in Section 3.1. 5 There are 31 administrative units at the provincial level in China, including 22 provinces, 5 autonomous regions, and 4 municipalities; as of 2012, there are 2,862 county-level administrative units.

4
Our panel data regressions reveal significant disemployment effects of minimum wages on young adults (age 15-29) between 2004 and 2009 over the country-a 10 percent increase in the current and previous year's minimum wages led to a statistically significant .88 percent and 1.36 to 1.56 percent reduction in employment, respectively. Furthermore, we find that the minimum wage has the largest lagged adverse effect on the employment of at-risk groups (defined as workers whose monthly wages are between the old and new minimum wage standards), showing that the elasticities are in the range of -.265 to -.340 for the entire sample over the same period.
To further substantiate our findings, we re-estimate the effects for three different time periods-pre-2004, 2004-2007, and 2008-2009  Several studies on the employment effects of minimum wages in China find mixed results, and the results for different regions are often opposite to one another. For example, Ni, Wang, and Yao (2011) focus on all employees and find some negative effects in the more prosperous and rapidly growing East and some positive effects in the developing Central and less developed Western regions over the 2000-2005 period. In contrast, Gunderson (2011) use 2000-2007 data on rural migrants and find no adverse effects and indeed a positive employment effect in state-owned enterprises in the East and negative effects in the Central and Western regions.

5
The discrepancies between these studies may be explained in part by the fact that the employment effects on different target groups tend to differ. Indeed, by examining the effects on several subgroups, our estimates seem to reconcile the results of previous studies-we find that, similar to Ni, Wang, and Yao (2011), the minimum wage has a significantly negative effect on all employees in the East and a lagged positive (though statistically insignificant) effect in the Western region in 2004-2009. In contrast, using rural migrants as the target group, we find that the minimum wage has an adverse and significant effect in the West and a positive (though statistically insignificant) effect in the East over the same period studied in Wang and Gunderson (2011).
Finally, we investigate the impact of the minimum wage on the employment of workers by skill level. In theory, low-skilled workers are relatively vulnerable when facing minimum wage increases. As anticipated, our panel data regression results show that the minimum wage has an adverse, though perhaps mild, effect on the employment of low-skilled workers (defined as high school graduates or below), a 10 percent increase in the current minimum wage results in statistically significant reductions in employment of .54 to .80 percent for the entire sample, .70 percent for the East, and .71 to .77 percent for the Central region. In contrast, we do not find a statistically significant effect for the Western region or for those workers who have at least a college degree.
The remainder of the paper is organized as follows. We provide a review of the development of minimum wages in China in Section 2. Section 3 provides details pertaining to the data and research design of the paper. In Section 4, we present and discuss the empirical results. Section 5 presents the paper's conclusions.

6
Prior to 1994, China had no minimum wage law. In 1984, the country simply acknowledged the 1928 "Minimum Wage Treaty" of the International Labour Organization (ILO) (Su, 1993;Gunderson, 2011, 2012). Due to the sluggish wage growth and high inflation in the late 1980s, Zhuhai of Guangdong Province first implemented its local minimum wage regulations, followed by Shenzhen, Guangzhou, and Jiangmen in 1989. It was not until the eruption of private enterprises in 1992 when labor disputes became frequent that the Chinese Central Government began to consider the minimum wage legislation (Yang, 2006). In 1993, China issued its first national minimum wage regulations, and in July 1994, they were written into China's new version of the Labor Law.
The 1994 legislation required that all employers comply with wages no less than the local minimum wages. All provincial, autonomous-region, and municipality governments should set their minimum wages according to five principles and report them to the State Council of the Central Government. More specifically, the five principles indicated that the setting and adjustment of the local minimum wage should synthetically consider the lowest living expenses of workers and the average number of dependents they support, local average wages, labor productivity, local employment, and levels of economic development among regions. These conditions provided considerable flexibility for provinces and municipalities in setting minimum wage standards, with the economic development principle giving them the flexibility to restrain minimum wages to attract foreign investment (Frost, 2002;Wang and Gunderson, 2011). By December 1994, 7 of 31 provinces and municipalities had set their own minimum wages. By the end of 1995, the number increased to 24. In addition, the new regulation required local governments to renew the minimum wage standards at least once every two years, and penalties for violation were increased from 20 percent to 100 percent of the owed wage to 100 percent to 500 percent of the owed wage.
Employers cannot include subsidies such as overtime pay or canteen and traveling supplements as part of the wage when calculating minimum wages. The new regulations were put into effect on March 1st, 2004 and led to substantial increases in minimum wages.

Data and Research Design
The data collection and research design were motivated by a desire to estimate the average effect of minimum wages on employment and to attempt to address some of the aforementioned challenges. In collecting the data, the goal was to obtain information on the minimum wage at 8 the county or city level over a long time span, with a panel structure allowing for the use of fixed time and county effects to eliminate omitted variable bias arising from unobserved variables that are constant over time and those that are constant across counties. The wage sample needed to be a longitudinal microdata sample to allow the distribution of minimum wage workers-in each geographic region, age cohort, skill level, and industry-to be estimated. For these reasons, and because the paper also aimed to examine how the 2008 global financial crisis influenced our results, we sought to collect information on provinces that were potentially affected over as many years as possible.

Data
Our study primarily uses two data sources: the annual Urban Household Survey (UHS) from [ Figure 2 about here] Figure 2 depicts the 16 representative provinces/municipalities used to study the impact of minimum wages on the Chinese labor market. We divide the 31 jurisdictions into three regions 9 as in Wang and Gunderson (2011): the more prosperous and rapidly growing East, the developing Central region and the less developed and more slowly growing West. As shown in Our primary objective was to thoroughly and accurately acquire relevant information on the minimum wage for each county. In China, provinces and municipalities have considerable flexibility in setting their minimum wage standards according to local economic conditions, resulting in several levels of standards across counties/cities within the same province.
Moreover, the adjustment date of a county's new minimum wage standard can also differ from its geographically contiguous neighbors within the same province, making the estimation of minimum wage effects more challenging. To effectively address this issue, we collected our minimum wage data from every local government website and carefully recorded the minimum wage information for approximately 2,000 counties every year from 1994 to 2012. As such, our data contain monthly minimum wages for full-time employees, hourly minimum wages for parttime employees, the effective dates of the minimum wage standards and the extent to which social security payments and/or housing provident funds were included as part of the minimum wage calculations.
[ Table 1 about here] We then merge the minimum wage data into the UHS, a 16-province panel dataset that contains individual/household socio-economic information over the 2002-2009 period. We present a brief summary of the minimum wage data used in our main analysis for the post new minimum wage regulations (2004) period in Table 1. For example, columns (1), (2), and (3) correspond to the mean of the monthly minimum wages, the standard deviation, and the number of counties for the three regions as well as the 16 provinces/municipalities in 2004, respectively. 7 When calculating the mean minimum wages, we use the time-weighted method, as suggested in Rama (2001), to address the issue of different adjustment dates among counties within a province within a year. The last row reports the mean of the minimum wages of all provinces, its standard deviation, and the total number of counties for each year. and 2009 for all counties as a whole. 8 Second, the East region has the highest minimum wage, with an average of 522 RMB per month in this period, followed by the West (436 RMB) and the Central region (424 RMB). Surprisingly, minimum wages of the three regions have similar annual growth rates of 13 percent. 9 Third, raising the minimum wage standards sometimes occurred more than once in a year. For example, Beijing increased its minimum wages in January and July of 2004, and Jiangsu raised its standards in April and July of 2008.
[ Table 2 about here] 7 Note that there was no minimum wage increase in 2009 because of the global financial crisis. The mean minimum wages have been adjusted for inflation and converted into 2005 RMB. 8 In fact, the average real minimum wage has also grown at a similar rate. 9 The average annual growth rate of the minimum wage is 12.7 percent in the Eastern region, 13.2 percent in the Central region, and 12.5 percent in the Western region over the 2004-2009 period. We restrict the analysis to working-age population between the ages of 15 and 64 who are employed in the civilian labor force, report positive annual earnings, are not self-employed, and not enrolled in school. Individuals who work in the agricultural production or services, farming, forestry, fishing, and ranching industries are also excluded (Neumark and Wascher, 1992). Table 2 presents summary statistics of the two key variables, minimum-to-average wage ratio and employment-to-population ratio, from 2004 to 2009. The second and third rows of the table show that male workers have approximately 10 percentage points lower minimum-to-averagewage ratios and 15 percentage points higher employment-to-population ratios than females, meaning that Chinese female workers are comparatively disadvantaged in the labor market relative to their male counterparts. As anticipated, this result shows that the more prosperous Eastern region has the lowest minimum-to-average-wage ratio (.276) and the highest employment-to-population ratio (.607) of the three regions. 10 Mounting evidence from minimum wage studies has consistently found that minimum wages have a greater impact on young and low-skilled workers, especially teenagers. Compared to their senior counterparts, young workers, who are often equipped with less human capital, are more likely to earn the minimum wage. Table 2 also shows the two key variables by age cohort and by educational attainment over the 2004-2009 period. Indeed, we find that young Chinese workers aged 15 to 29 have the highest minimum-to-average-wage ratio (.392), at least 10 percentage points higher than that of other age cohorts. For workers with different skills, the evidence demonstrates that as the skill level increases, the minimum-to-average-wage ratio 12 decreases quickly-dropping continuously from .593 for elementary school or below to .183 for college or above. Table 2 also presents the minimum-to-average-wage ratio by industry. The manufacturing sector contains the largest share (21.6 percent) of workers in our sample; the public service sector is the second largest (13.9 percent); and the third and the fourth sectors are wholesales and retail sales trade (9.9 percent) and housekeeping (9.6 percent), respectively. Looking at the minimum-to-average-wage ratios, unsurprisingly, we find that the housekeeping sector has the highest ratio (.509) among all industries, followed by the hotel and restaurant sector (.498) and wholesales and retail sales trade (.471).
[ Table 3 about here] We also provide a summary of the characteristics of workers who earn the minimum wage as well as less/more than the minimum wage over 2004-2009 in Table 3. The first row of Table 3 shows that approximately 5.62 percent of all workers earned less than the minimum wage and 3.28 percent earned just the minimum, meaning that a combined 8.90 percent of Chinese employees are minimum wage workers over the 2004-2009 period. Among those who earned the minimum wage exactly and less than the minimum wage, 63.84 percent and 61.52 percent are females, respectively. Furthermore, the minimum-to-average-wage ratio of workers receiving less than the minimum wage is 2.52, meaning that these disadvantaged workers earn a wage that is only approximately one-quarter of the official standard.
For different age cohorts, Table 3 shows that young adults (age 15-29) are more likely to be minimum wage workers. With increase age, the percentage decreases. Similarly, we find the same decreasing pattern in the skill panel. Looking at the characteristics of workers by industry, Table 3 shows that the housekeeping sector has the largest share of minimum wage workers: 13 approximately 20.21 percent of housekeepers earn less than or equal to the minimum wage.
Wholesales and retail sales as well as hotel and restaurant sectors also have 16.76 percent and 16.50 percent of workers earning below or equal to the minimum wage, respectively.

Research Design
Our objective is to assess the impact of minimum wages on the employment of potentially affected workers. As noted in Section 1, nearly all existing studies on minimum wages in China use pooled time-series/cross-section data at the provincial level and tend to find mixed results, implying that a "consensus" of employment effects remains to be established. Thus, our study attempts to reconcile the existing findings using more sophisticated minimum wage data, which permit the use of a panel structure analysis of minimum wage effects, exploiting the greater variation in relative minimum wages at the county level and avoiding the measurement error caused by using a uniform provincial minimum wage. Moreover, unlike previous studies that use aggregate published statistics, our study used household survey microdata, which allows us to calculate the dependent variable-the employment-to-population ratio-at the county level, which contains more variation and information on local conditions. Ideally, this feature should yield more reliable estimates of the employment effects of minimum wages in China.
Specifically, our panel data allow us to estimate a prespecified equation of the form proposed in Neumark (2001) and used in Campolieti, Gunderson, and Riddell (2006) and Wang and Gunderson (2011). Before the data analysis, the methodology involves precluding running alternative specifications until preferred results are obtained. Our estimation equation is wage ratio) of county i in year t and year t-1, respectively; X is a set of control variables to capture aggregate business cycle effects; t Y is a set of fixed year effects; and i C is a set of fixed county effects. The disturbance term ε is assumed to be serially uncorrelated and orthogonal to the independent variables.
To address the bias from the specification error and the potential endogeneity problem, we include several control variables in estimating the equation. First, the county GDP per capita and CPI (city level) capture aggregate business cycle effects and controls for the global financial crisis of 2008. Second, the county foreign direct investment (FDI) is used to control for that provinces may restrain the minimum wage to attract foreign investors (Frost, 2002). Because the decisions of whether to increase minimum wages are determined by government officials, who often must consider local economic conditions, we collectively include these controls to address this issue.

Minimum Wage Effects Across Regions
We first present the estimation results for young adults, at-risk groups, and the entire sample for the East, Central, West, and all regions in Table 4. In each region, we estimate Eq. (1)  [ Table 4 about here] The first and second columns of Table 4 report the estimates with cluster-robust standard errors in parentheses for young adults and at-risk groups across different regions using Eq.
(1), while in the third column, we report the estimates of the entire sample for comparison. The significance of our results is compelling: over the country, we find negative effects of the current and lagged minimum wages on employment. A 10 percent increase in the current and previous year's minimum wage led to a statistically significant .88 percent and 1.36 to 1.56 percent reduction in young adults' employment, respectively. A 10 percent increase in the current and previous year's minimum wage led to a statistically significant 2.13 percent and 2.65 to 3.40 percent reduction in at-risk groups' employment, respectively. For the entire sample, a 10 percent increase in the current and previous year's minimum wage led to a statistically significant .45 to .55 percent and .28 to .31 percent reduction in employment, respectively.
In the more developed and prosperous East China, which has a large population residing in large cities, such as Beijing, Shanghai, and Guangzhou, the minimum wage has been an important policy tool as China makes the critical transition into a market economy.
Consequently, the magnitude and frequency of minimum wage increases are relatively high in the regions in which the impact of minimum wages on employment could be evident. Indeed, consistent with the evidence in Table 4, our estimates indicate that minimum wage increases in the Eastern region have a statistically significant adverse impact on employment with elasticities ranging from -.154 to -.234 and a lagged adverse effect with an elasticity of -.100 for young adults. Furthermore, we find a large and negative lagged minimum wage effect on the employment of at-risk groups-a 10 percent increase in the minimum wage led to a statistically significant 3.10 to 3.22 percent reduction in employment. The current minimum wage effects are negative; however, they are not statistically significant.
In the developing Central region, we also find all lagged minimum wages to have a strong negative employment effect on young adults, at-risk groups, and the entire working population.
The minimum wage has an adverse lagged employment effect with an elasticity of -.216 for young adults and -.310 to -.336 for at-risk groups. For the entire working population in the Central, the elasticity is in the range of -.041 to -.042. The estimates of the current minimum wage variable are negative; however, they are not statistically significant.
Finally, in the less developed West, we do not find an effect of the minimum wage on employment. Nevertheless, without controlling for local economic conditions, our empirical results show positive (not statistically significant) coefficients for the current and the lagged minimum wages of young adults and at-risk groups. When economic conditions are controlled, we find positive but insignificant estimates for the current and the lagged minimum wages for atrisk groups. [ Table 5 about here]

Gender and Age Cohort
We present the estimates for all regions in panel A. The results show that the current minimum wage has an adverse effect on the employment of female young workers (age 15-29): a 10 percent increase in the minimum wage results in a statistically significant 1.48 percent reduction in employment and a minor lagged effect with an elasticity of -.061. Furthermore, we find that the negative effects on females decrease as the age cohort moves up, showing that the elasticity of the current effect is -.068 for females aged 30-39 and that of the lagged effect is -.040 for females aged 40-49. In contrast, we do not find a significant effect of minimum wages on females aged 50-64 or on male employment for any cohort over the country.
In other regions, minimum wages seem to have an adverse employment effect on young females in Eastern and Central regions, for whom a 10 percent increase in the current year's minimum wage led to a statistically significant 1.72 percent and 1.55 percent reduction in employment, respectively. We also find minor disemployment effects of minimum wages on males aged 30-39 in the Central region, with elasticities of -.052 for the current and -.072 for the lagged minimum wage variables.

Skill Level
In the literature, the preponderance of evidence supports the view that minimum wages reduce the employment of low-wage workers. Moreover, when researchers focus on the leastskilled groups, which are most likely to be directly affected by minimum wage increases, the evidence for disemployment effects seems to be especially strong (Neumark and Wascher, 2008).
We present the estimation results by three skill groups as measured by educational attainment in Table 6. In each group, we report the estimates using the fixed-effects model with both fixed year and county effects.
[ Table 6 about here] Our estimates reveal disemployment effects of minimum wages on low-skilled workers (high school graduates or below). For example, looking at panel A of Table 6, the results show that the current minimum wage has an adverse effect on the employment of workers who are high school graduates or below: the elasticities of -.054 and -.080 are statistically significant at the 5 percent level. Furthermore, we also find lagged negative effects of minimum wages on the employment of vocational school degree workers-a 10 percent increase in the previous year minimum wage results in a statistically significant .40 to .47 percent reduction in the current year employment.
In the East, we find that the current minimum wage has a negative employment effect on low-skilled workers, with an elasticity of -.070, but no effect on other workers with higher degrees. As shown in Panel C of Table 6, we find that the minimum wage has an adverse effect on low-skilled workers in the Central region, with elasticities of -.071 to -.077 for the current year and -.047 to -.052 for the previous year minimum wage variables. In addition, we also find a lagged disemployment effect on workers with vocational school degrees in the Central region, with elasticities in the range of -.083 to -.090. Finally, we examine the effect of minimum wages on workers with a college degree or above (including junior college) and do not find a significant effect in any region.

Minimum Wage Effects on Migrant Workers
The new minimum wage regulations of 2004 were designed in large part to protect rural migrant workers, who tend to work in non-state enterprises in which labor standards and wages are low (Cooke, 2005;Zhang and Deng, 2005;Wang and Gunderson, 2011). Minimum wages are expected to have a stronger effect on rural migrant workers because they tend to work in lowwage sectors and the higher wages will induce some enterprises to use more skilled workers or more capital to substitute for the now more expensive rural workers (Wang and Gunderson, 2011).
Using the micro-level UHS data, we are able to examine how the minimum wage affects the employment of rural migrant workers at the county level. Because very few rural migrants work in state-owned enterprises in our sample, we focus on non-state enterprises and report the results for all enterprises as well. Table 7 reports the results for Eastern, Central, and Western regions. (2011), we find that the minimum wage has negative employment effects on rural migrant workers in the less developed and more slowly growing Western regions: for all enterprises, a 10 percent increase in the lagged minimum wage results in a statistically significant 2.16 to 2.82 percent reduction in employment. In particular, for migrant workers in non-state enterprises, we find a larger disemployment effect of current minimum wages, with elasticities of -.408 and -.411. In contrast, the results show positive coefficients (though statistically insignificant) of the minimum wage variables in the East, which is consistent with the monopsonistic behavior found in Wang and Gunderson (2011) .

Endogeneity Issue
In China, the decisions of whether to increase minimum wages are determined by local In short, our results in Table 8 seem to support the pattern observed in Figure 1,  the mean of the minimum wage level in all geographically bordering provinces (Neumark and Wascher, 1992), real minimum wages (Feliciano, 1998), and the changes of province leader variable to capture the effect of Chinese political system. We instrument the current and previous year minimum wage variables and present the estimates in the estimation equations (1) and (2) of Table 9, respectively.
Viewing the results, along with considerations of the validity of the instruments, we find that minimum wages have negative effects on young adults-a 10 percent increase in the previous year's minimum wage led to a statistically significant .92 percent reduction in young adults' employment for the entire sample; in the Eastern and Central regions, we find a 10 percent increase in the current minimum wage led to a statistically significant 2.15 percent and 2.56 percent reduction in young adults' employment, respectively; in the Western region, the minimum wage has a larger negative effect on young adults with elasticities in the range of -.272 to -.845.
Furthermore, the results show that the employment elasticities are more strongly negative for at-risk groups: the elasticities are -1.678 (current minimum wage) and -.697 (lagged minimum wage) for the entire sample and -1.263 (current minimum wage) and -.663 (lagged minimum wage) for the Central region. The finding is consistent with several studies such as Neumark and Wascher (1992) and Feliciano (1998) that endogeneity bias generates a positive correlation between employment rates and minimum wages. In contrast, we do not find a statistically significant effect in the East and West of China.

Conclusions
We .088 and -.136 to -.156, respectively. Furthermore, we found that minimum wage changes over the country resulted in a large lagged disemployment effect for at-risk groups over the same period, with elasticities in the range of -.265 to -.340. In the Central region, we found that the minimum wage has lagged effects-a 10 percent increase in the minimum wage led to a statistically significant 2.16 percent and 3.10 to 3.36 percent reduction in employment for young adults and at-risk groups, respectively. In contrast, we did not find significant effects in the less prosperous and slower-growing Western region.
We then assessed the effect of minimum wages by gender and age cohort. Consistent with most studies in the literature, we found that the minimum wage has negative effects on female young workers (age 15-29)-the most disadvantaged and vulnerable groups in the labor market.
However, we did not find significant effects on the employment of their male (age 15-29) and senior counterparts (age 50-64) for the entire sample.
Our study seems to reconcile the mixed results reported by Ni, Wang, and Yao (2011) and Wang and Gunderson (2011). By examining the effects for several subgroups, we found that, similar to Ni, Wang, and Yao (2011), the minimum wage has a significantly negative effect on all employees in the East and a lagged positive effect in the Western region in 2004-2009; on the contrary, using rural migrants as the target group, we found that the minimum wage has an adverse and significant effect in the West and a positive effect (though statistically insignificant) in Eastern region over the same period, as found in Wang and Gunderson (2011).
Finally, we investigated whether the minimum wage affects the employment of low-skilled workers. Our results support that minimum wages reduce the employment of low-skilled workers, indicating that Chinese workers who are high school graduates or below or have vocational school degrees were directly and adversely affected by minimum wage increases.     Note: *** , ** , and * indicate that the estimate is statistically significant at the 1 percent, 5 percent, and 10 percent levels, respectively. Cluster-robust standard errors are in parentheses. All variables in the table are at the county level, except that CPI is at the city level. Note: *** , ** , and * indicate that the estimate is statistically significant at the 1 percent, 5 percent, and 10 percent levels, respectively. Cluster-robust standard errors are in parentheses. All variables in the table are at county level. All variables in the table are at the county level, except that CPI is at the city level. Note: *** , ** , and * indicate that the estimate is statistically significant at the 1 percent, 5 percent, and 10 percent levels, respectively. Cluster-robust standard errors are in parentheses. All variables in the table are at the county level, except that CPI is at the city level. The effects of migrant workers of state-owned enterprises cannot be estimated due to an insufficient number of observations.  Note: *** , ** , and * indicate that the estimate is statistically significant at the 1 percent, 5 percent, and 10 percent levels, respectively. Cluster-robust standard errors are in parentheses, except that of Hansen J statistics report Chisquare P-values. All variables in the table are at the county level, except that CPI is at the city level. Models (1) and (2) report the estimates of instrumenting the current and previous year minimum wages, respectively. Weak identification test reports the Cragg-Donald Wald F statistic in which ++ and + indicate that the statistics are greater than the 5 percent and 10 percent maximal IV relative bias of Stock-Yogo weak identification test critical values, respectively. Hansen J test reports the overidentification test of all instruments.