ARTICLE IN PRESS

Social Science Research xxx (2009) xxx–xxx

Contents lists available at ScienceDirect

Social Science Research

journal homepage: www.elsevier.com/locate/ssresearch

Voluntary and involuntary job mobility and earnings inequality in urban

China, 1993–2000 q
Xiaogang Wu *
Division of Social Science, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong

a r t i c l e i n f o a b s t r a c t

Article history: This paper proposes a model of selective mobility of workers from the state sector to the
Available online xxxx market sector to illustrate how the market transition has led to earnings inequality in for-
mer state socialist countries. Analysis of the survey data collected in 2000 from 10 Chinese
Keywords: cities reveals that recent entrants into the market are driven by two different institutional
China processes—some are self-selected for higher economic returns (voluntary entrants) and
Earnings inequality some are pushed into the market through layoffs (involuntary entrants), resulting in a
Labor markets
more heterogeneous body of workers in the market sector than before. Linear regression
Layoffs
Market transition
results show that the commonly observed higher earnings in the market sector are limited
Propensity-score matching methods only to a subgroup of later entrants who enter the sector voluntarily. Propensity-score
matching analyses further demonstrate that the effect of a late market entry on earnings
is negatively associated with the propensity of making such a transition. Those who would
otherwise do well in the state sector and therefore have a lower propensity for entering the
market benefit more from the entry.
Ó 2009 Elsevier Inc. All rights reserved.

1. Introduction

Empirical analysis of stratification deals primarily with the stratification outcomes, namely, the inequalities in education,
occupation and income among different social groups. Studies of the underlying process through which such inequalities are
produced, however, are yet to be emphasized in the literature (Blalock, 1991). Therefore, when sociologists attempt to make
inferences about the causal mechanisms of inequality, they tend to rely on many overly simplistic and untested assumptions,
often leading to substantial bias and untenable conclusions.
The existing literature on social inequality and stratification in former state socialist countries serves as a good example
to illustrate this problem. The institutional transition to market economies in China and East European countries has re-stim-
ulated scholars’ theoretical interest in how macro-level social transformation has reshaped the structure of inequality (Nee,
1989; Szelényi, 1978). Early empirical research in this field mainly relied on the investigation of income inequality, partic-
ularly on returns to human capital and political power, to infer the changes in the mechanism of social stratification (e.g.,
Bian and Logan, 1996; Gerber and Hout, 1998; Nee, 1989, 1996; Parish and Michelson, 1996; Róna-Tas, 1994; Xie and Han-

q
An earlier version of this paper was presented in the American Sociological Association Annual Meeting, Philadelphia, PA, August 12–15th, 2005, a
conference of the ISA Research Committee on Social Stratification and Mobility (RC28), University of California, Los Angeles, August 19–21st, 2005, the
Populations Association of America Annual Meeting, Los Angeles, March 30 to April 1st, 2006. The author thanks Zheng Hangsheng of Renmin University of
China for making the data available, Yu Xie of University of Michigan, Ann Arbor, several anonymous reviewers of the Social Science Research for comments,
suggestions. This research was supported by two research grants: ‘‘Educational Inequalities in Reform-era China, 1990–2000” (HKUST6424/05H), ‘‘Social
Mobility, Stratification Dynamics in China Since mid-1990s” (GRF 644208), both from the Hong Kong Research Grants Council.
* Fax: +852 23350014.
E-mail address: sowu@ust.hk

0049-089X/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.ssresearch.2009.11.003

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
2 X. Wu / Social Science Research xxx (2009) xxx–xxx

num, 1996). Such an ‘‘intellectual leap” from the observed income inequality to the underlying causal mechanisms has ini-
tiated an unresolved debate among sociologists on the social consequences of the market transition in the 1990s (Cao and
Nee, 2000; Zhou, 2000a,b).
Recent studies have shifted the theoretical paradigm by specifying concrete institutional settings or intermediate processes
through which income inequality is generated (Gerber, 2002; Walder, 2002, 2003; Wu, 2002, 2006; Wu and Xie, 2003; Zhou,
2000a,b). In analyzing income determination, scholars have emphasized structural changes in the emerging labor markets and
workers’ mobility in the process of producing new inequalities. For instance, rapid economic expansions in rural China during
the 1980s created many opportunities for wage employment other than farming and entrepreneurship. As a result, ordinary
peasants’ income from these sources substantially altered the income distribution among different social groups (Walder,
2002). In post-Soviet Russia, ‘‘shock therapy” led to sweeping structural changes in labor market opportunities, facilitating indi-
viduals’ labor market transitions (Gerber, 2002) and thereby affecting inequalities (Gerber and Hout, 1998).
Emphasis on the effect of structural changes in labor markets on social stratification has moved one step further towards
understanding the process of how diverse stratification outcomes are produced in the context of macro social changes in
post-socialist countries. This approach links the studies of social stratification in transition economies to a broad literature
on labor market processes in developed capitalist economies (e.g., DiPrete, 1993; DiPrete and Lynn Nonnemaker, 1997).
However, this is only one-sided story. Unless we are willing to make the fundamental assumption that either individual
behaviors are nearly completely determined by structural factors or, if not, individual differences are simply cancelled
out in some sort of aggregation process, it is difficult to account for structural changes without examining the more micro-
or individual-level process that undergirds them (Blalock, 1991; Blau, 1977; Hannan, 1991).
From this perspective, Wu and Xie (2003) argue that workers are dynamic social actors, who are not simply affected by
the market transition, but rather respond to it by actively situating themselves in the labor markets. Based on the history
of individual mobility in urban China, Wu and Xie (2003) characterized four types of workers: (1) those who are in the
state sector and continue to stay there (‘‘stayers”); (2) those who enter the market sector early and stay there (‘‘early
birds”); (3) those who begin in the state sector but later transfer to the market sector (‘‘later entrants”); and (4) those
who are initially in the market sector, but later retreat to the state sector (‘‘market losers”). They examined differential
earnings returns to education among early birds and later entrants in the market sector in comparison with stayers in
the state sector.1
Wu and Xie’s (2003) contribution was primarily conceptual and their empirical analyses hinged on a temporal argument:
the returns to the market sector employment and the returns to education within the market sector depended on when Chi-
nese workers entered the sector. Those who entered early on (early birds) experienced neither higher earnings nor higher
returns to education than those who remained in the state sector, but later entrants did experience an earnings premium
and higher returns to education. They interpreted this temporal variation as evidence that selection mechanisms differed
between early birds and later entrants. However, they had no direct evidence to support this claim. Moreover, the advanta-
ges that later entrants enjoyed over stayers were not convincingly explained (Jann, 2005; Xie and Wu, 2005). As Wu and Xie
(2003) speculated (pp. 438–439),
Two different institutional processes in urban China may have contributed to this phenomenon. On one hand, a growing
number of qualified workers voluntarily gave up their career opportunities in the state sector and entered the market, or
‘jumped into the sea’ (xiahai). On the other hand, an increasing number of workers were laid off by ailing state enterprises
and thus ‘pushed’ into the market (xiagang). Due to the different selection mechanisms, the two subgroups could be very
different from each other in observed and unobserved characteristics. . . .. . .. Pooling them would yield higher returns to
education for later market entrants as a single group.

Because the data analyzed by Wu and Xie (2003) did not contain a direct measurement of respondents’ motivations/rea-
sons for making labor market transitions (i.e., whether they ‘‘jumped” or were ‘‘pushed” into the market), they were unable
to test the distinction between voluntary and involuntary transitions and their associated consequences.
In this paper I advance the inquiry above in two respects. First, based on the survey data collected in 10 cities in 2000, I
investigate the patterns of entry into the market sector in urban China and directly test the claim on the differential mobility
patterns between early birds and later entrants, with a special attention to the distinction between involuntary and volun-
tary later entrants. Second, using both multiple linear regression and the propensity-score matching methods, I examine
how differential patterns of entry into the market sector affect earnings inequality among later entrants compared to those
who stayed in the state sector.

2. Individual labor market transitions, group heterogeneity and earnings inequality

For the past few decades or so, economic reforms in former state socialist countries have led to the emergence of a market
sector in the redistributive economy. The market sector has offered ‘‘a new window of opportunity” for social mobility, thus
yielding important implications for changes in social stratification (Nee, 1989). In the dual opportunity structure, ‘‘one could
climb the rank order of the bureaucratic hierarchy, or one could try the market” (Szelényi, 1988: 65). To a great extent, the

1
Only a few cases of market losers were documented. They were therefore excluded from Wu and Xie’s (2003) analysis.

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
X. Wu / Social Science Research xxx (2009) xxx–xxx 3

question under debate—who wins and who loses in the course of market transition—is contingent upon who stays in the
hierarchy and who switches to the market sector (Szelényi and Kostello, 1996).
Workers’ labor market transition is an integral part of the transition from state socialism to market capitalism. For exam-
ple, the rate of job mobility in China, particularly from the redistributive state sector to the market sector, has dramatically
increased since the economic reform. In 1978, only 150,000 Chinese workers in the entire country were employed in the
market sector; the employed workers in the market sector climbed to 62,410,000 in 1999, with an increase of 416 times
within two decades (National Bureau of Statistics, 2000).
Such dramatic changes in labor mobility were by no means a random process, but rather reflected individual selective
mobility in response to the changing opportunity structures during the market reform (Wu, 2006). There have been some
anecdotal and contradictory accounts of how different social groups associate themselves with new market opportunities.
In his early study of Hungarian rural entrepreneurship, Szelényi (1988) found that cadres are less likely to participate in mar-
ket-oriented businesses, and that the new economic elite are more likely to emerge from less privileged groups. Various sur-
veys conducted in urban China in the mid-1980s show that most private entrepreneurs and individual business owners
(getihu) were migrant peasants, unemployed youth, dismissed workers, former criminals released from prisons and retirees,
who were unable to secure employment in the state sector (Davis, 1999; Gold, 1990; Li, 1993). Zhou et al., 1997, based on the
data from a representative survey of urban China, reported minimal mobility from the state sector to the market sector in
urban China as of 1993, especially among those with high human and political capital (also see Gold, 1990; Li, 1993).
However, in post-1989 Hungary after the collapse of the communist regime, the communist elites became more likely
than ordinary workers to convert themselves into corporate entrepreneurs and to maintain their economic advantages
(Róna-Tas, 1994). In China, with the Communist Party still firmly holding onto political power, the market economy gained
full legitimacy since 1992, when the paramount leader Deng Xiaoping made his famous trip to southern China and called for
further economic reforms. Cadres and professionals increasingly started giving up their career opportunities within the state
sector in search for new advantages in the market sector. A new Chinese phrase, xia hai (‘‘jumping into the sea”), was coined
to refer to the new phenomenon of workers’ moving into the market sector (Wu and Xie, 2003; Wu, 2006).
Szelényi and Kostello (1996) brought together different and somewhat contradictory observations of entry into the mar-
ket sector by relating them to the process of marketization. According to them, in the early stages of economic reforms when
participation in the market was highly risky and required little skill, entrants into the market sector tended to be those in the
low tiers of the social hierarchy, who were not at risk of losing privileges like those enjoyed by workers in the state sector.
However, as marketization proceeded and the risks in the market were further reduced, workers with more marketable skills
started to seek the new opportunities available in the market. Communist cadres also learned to embrace the market to cash
in on their political and social capital (Wu, 2006). In the face of competition from these groups, ‘‘with more to lose but also
more to gain”, the early market pioneers were marginalized under certain circumstances (Szelényi and Kostello, 1996:1089).
Hence, workers are fluid in the labor markets and their mobility into the market sector is unlikely to be exogenous to the
process of marketization. It is this underlying social process that is directly responsible for the different stratification out-
comes that have been observed by many scholars involved in the debate.
Not all workers are rational actors who can control their own fate and maximize their benefits in the course of market-
ization. Market transitions have brought not only new opportunities for people to take advantage of, but also job losses and
downward mobility for those who suffer from, particularly in the late reform period. Gerber and Hout (1998) presented a
transition scenario that was clearly at odds with the claims in Nee’s market transition theory (Nee, 1989). After the sudden
collapse of the Soviet Union, the sweeping transition to capitalism did not increase returns to human capital and to profes-
sionals in post-socialist Russia. Since the mid-1990s, the growth of the private economy in China has gained a new momen-
tum, which, on one hand, attracted talented workers from the state sector, but on the other hand, pushed some state-owned
enterprises into bankruptcy because of market competition. Many workers were pushed into the market for a living (Cai,
2002; Lee, 2000; Solinger, 2002).
Hence, the fact that workers recently entered the market sector through two qualitatively different mechanisms—through
layoffs that pushed them into the market involuntarily or through self-selection that allowed them to voluntarily ‘‘jump”
into the market sector for new opportunities—further enhances the heterogeneity of workers in the market sector and com-
plicates our investigation of the impact of the labor market transition on stratification outcomes. Compared with those who
stayed in the state sector, the early market entrants came mostly from disadvantaged backgrounds and had little chance to
do well in the state sector. Among the recent market entrants, those who were forced to leave the state sector might possess
less human and political capital or other unobserved characteristics negatively associated with potential earnings, whereas
those who were self-selected into the market sector might possess certain observable or unobservable characteristics that
are positively associated with potential earnings. That the workers, who entered the market through these two mechanisms,
fared differently has been well documented by some ethnographic researchers (e.g., Solinger, 2002; Wank, 1999).
These differential sorting processes of workers into the market sector are endogenous to the pace of marketization and
therefore should be incorporated into the analysis of the change in earnings inequality in the context of market transitions.
Previous research in this field has largely ignored such varying processes and has relied on comparisons among highly het-
erogeneous groups to make causal inferences on the effects of the political or market institutions (mechanisms) on income
inequality. Without knowing how individuals are sorted/placed into different social groups/sectors, the estimates of group
differences in earnings based on aggregate data are likely to be biased, and the interpretation of the group differences as
caused by group membership is unwarranted (Gerber, 2000; Wu and Xie, 2003).

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
4 X. Wu / Social Science Research xxx (2009) xxx–xxx

3. Research design and methodology: bringing the process back in

The discussion above suggests that the group heterogeneity among later entrants is crucial to understanding causal
mechanisms of the stratification outcomes (earnings inequality) in Chinese labor markets. To advance this inquiry, I first dis-
aggregate later entrants into involuntary and voluntary entrants based on whether they had layoff (xiagang) experience in
the state sector and examine how these two subgroups fare differently in earnings return after entering the market sector. I
then re-conceptualize the substantive problem with an explicit counterfactual framework (Holland, 1986; Manski, 1995;
Winship and Morgan, 1999) and employ a propensity-score matching method to assess the causal effect of an individual’s
labor market transition on earnings in 2000, the year when the analyzed survey data were collected.
On the first step, later entrants are divided into two subgroups based on their previous layoff (xiagang) experience in the state
sector. Here, xiagang in Chinese literally means step-down-from-the-post. For a long time since the Maoist era, the Chinese
party-state had committed to providing urban citizens life-time employment and most social services via the unique work unit
(danwei) system, to which overstaffing was almost an inherent problem (Wu, 2002). With the competition from the emerging
market sector since the 1990s, Chinese state-owned enterprises (SOEs), especially in traditional manufacturing industries, have
increasingly failed to make profits (Xie and Wu, 2008). To make the state-owned enterprises more efficient and sustainable, the
government had to allow them to streamline superfluous laborers, though in a moderate way. The term xiagang initially applied
only to workers from bankrupted, merged, closed-down enterprises but gradually included those whose salary was cut or sus-
pended (Lee, 2000). These off-duty workers continued to maintain their employment relationship with their original work units
to receive living stipends and claim social security benefits (Lee, 2000).2 In this sense, xiagang arrangement does not apply to
employees in the private sector but instead serves as a transient step for the state work unit to streamline redundant work forces.
While some xiagang workers may still be able to find a new job within the state sector, more tend to enter the market for
survival, because the living stipends paid to xiagang workers would gradually fade away and the job opportunities in the
state sector had been shrinking since the 1990s. Earlier studies have consistently reported that xiagang workers in the state
sector tend to be less educated, in middle age, women, concentrated in certain industries and regions (Solinger, 2001; Xie,
2006). The state mass layoff thus had forced these workers to depart for the market sector in the end. Thus, those who had
experienced a layoff in the state sector and ended up in the market sector can be defined as ‘‘involuntary entrants”.
The substantive interest of this paper is to examine the causal impact of the transition to the market sector on subsequent
earnings. This question can be re-conceptualized in a counterfactual framework, in which a late entry into the market sector
is considered as a ‘‘treatment” whereas the ‘‘control group” consists of stayers. The causal effect of a late market entry for an
individual can be defined as the difference between his (her) observed earnings in the market sector and counterfactual
earnings had the person stayed in the state sector.3
The propensity score stratum also provides a continuous measure of voluntary and involuntary movers from the state sec-
tor to the market sector. Those who have the lower propensity for market entry, but indeed have made the transition, are as-
sumed to have more control over the change (more voluntary) than those who have higher propensity for entry. To answer
whether the process of how individuals move from the state sector to the market sector affects their current earnings, I show
the heterogeneous treatment effect on earnings across different propensity score strata (see also Xie and Wu, 2005).

4. Data and variables

4.1. Data

The empirical analyses are based on the data collected from ‘‘Social Changes in Urban China since the Reform”, a survey
conducted in 2000 in 10 cities of five provinces (Jilin, Shaanxi, Henan, Hunan and Guangdong) and a province-level munic-
ipality (Tianjin). These provinces represent three different geographic regions and also different levels of economic develop-
ment in China. The capital city of each province and Tianjin, one of the four municipalities directly under the central
government’s jurisdiction, were chosen as representative of large cities. In addition, a medium-size city was randomly se-
lected in four of the five provinces (except Shaanxi). The procedure resulted in the selection of 10 cities: Changchun, Chang-
sha, Guangzhou, Jilin, Kaifeng, Tianjin, Xi’an, Xiangtan, Zhengzhou and Zhongshan (see Fig. 1). In each city, the multi-stage
stratified probability sampling method was used to select individuals aged from 18 to 65 years old. As a result, 4307 indi-
viduals were selected, of which 2631 held a job in 2000.4 The survey contains the historical information on the respondents’
job mobility concerning work unit, occupation, industry, and income in 1993 and 2000.

2
According to Chinese government statistics, there were 12.7 million laid-off (xiagang) workers at the end of 1997, 8.77 million at the end of 1998, 9.37
million at the end of 1999, and 9.11 million at the end of 2000. These figures include only those who were still looking for jobs at the end of the respective year;
indeed, more workers have suffered from the experience of xiagang (Xie, 2006, p. 23).
3
To render the matching estimates unbiased for the true average treatment, the observable characteristics are assumed to be exhaustive, that is, no other
important covariates affect both the propensity of receiving the treatment and the earnings. This is known as the ignorability assumption (Rosenbaum and
Rubin, 1984).
4
According to the author’s own calculation based on a sample of the 2000 Population Census of China, the non-agricultural employment rate for men and
women aged between 18 and 65 was 61.8 percent in all urban areas and 57.3 percent in the 10 selected cities (ranging from 50 percent in Jilin to 71.7 percent in
Zhongshan).

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
X. Wu / Social Science Research xxx (2009) xxx–xxx 5

Fig. 1. Selected cities in the 2000 Urban Social Change Survey.

This study investigates the process of labor market transitions in the period between 1993 and 2000. It chooses 1993 as
the benchmark year for both substantive and methodological reasons. First, 1992 marked a milestone in the history of Chi-
na’s economic reforms, when Deng Xiaoping made his famous political tour to southern China to push for further market
reforms after the Tiananmen Incident in 1989. A one-year lag is used to gauge the impact of the new policy shift on workers’
transitions in labor markets. Second, the retrospective information from 1993 is subject to fewer recall errors than would be
information recalled from earlier years. Finally, focusing on a relatively short period may avoid a high sample attrition rate
and provides more sample cases in the labor force in both 1993 and 2000.5
Young people yet to enter the labor force in 1993 are excluded in the analysis for two reasons. First, the paper’s main
concern is the process of labor market transition from 1993 to 2000, applicable only to those who already held a job as
of 1993; second, as the state job allocation was ended around 1993, more and more workers started their first job directly
in the market sector.
Hence, workers’ status in 1993 and 2000 are used to characterize four types of workers in China’s urban labor markets: (1)
those in the state sector in both 1993 and 2000 are defined as ‘‘stayers”; (2) those in the market sector in both 1993 and 2000
are defined as ‘‘early birds”; (3) and those in the state sector in 1993, but in the market sector in 2000, are defined as ‘‘late
entrants”. After eliminating those not active in the labor force and those who held a job but without income information in
either 1993 or 2000, I obtain 1669 cases for analysis.

4.2. Variables

Distinguishing the market sector from the state sector is crucial to the typology of workers. Two criteria are employed to
make this distinction: respondent’s affiliated work organizations and respondent’s occupation. Concerning the first criterion,
workers in the newly emerging types of work organizations, such as ‘‘domestic private enterprises”, ‘‘joint ventures/foreign-
invested firms” (sanzi qiye), ‘‘individual-owned businesses” (geti hu), and ‘‘other self-employed businesses”, are coded as
being in the market sector, and workers in ‘‘government agencies”, ‘‘state institutions”, ‘‘state-enterprises”, and ‘‘collective
enterprises” are coded as being in the state (public) sector. Concerning the second criterion, workers whose occupations
are ‘‘individual business-owners” (geti hu), ‘‘private enterprise owners” (siying qiye zhu), and ‘‘other self-employed occupa-
tions” are coded as being in the market (private) sector.

5
Indeed, sample attributions from 1993 to 2000 are partly due to retirement. In the survey data, 880 respondents who had a job in 1993 but no long worked
in 2000, and 208 of them have passed the retirement age in 2000 (60 years old for men and 55 years old for women in China). The rest withdrew from the labor
market before retirement: majority indeed had worked in the state sector in 1993, and they tended to be less educated and older, more likely to be women and
non-party members, and to report to have xiagang experience since 1993. In other words, they are similar to involuntary later entrants, and income advantages
associated with voluntary later entrants over stayers presumably would not be affected by the omission of this group in the analysis.

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
6 X. Wu / Social Science Research xxx (2009) xxx–xxx

Table 1
Un-weighted summary statistics of variables for different types of workers in 10 selected Chinese cities, 2000.

Variables Overalla Stayers Early birds Later entrants

Combined Voluntary Involuntary
Monthly earnings 2000 (RMB) 1066.1 (1475.5)b 943.8 (1028.9) 1592.8 (2555.0) 1425.3 (2380.4) 1808.9 (2960.4) 1041.6 (1530.8)
Monthly earnings 1993 (RMB) 613.9 (2000.6) 526.0 (709.9) 1294.5 (5500.8) 447.5 (558.5) 516.7 (483.6) 378.3 (619.9)
Monthly earnings 2000 (logged) 6.643 (0.733) 6.595 (0.666) 6.888 (0.918) 6.721 (0.933) 6.992 (0.988) 6.521 (0.833)
Monthly earnings 1993 (logged) 5.966 (0.837) 5.915 (0.785) 6.476 (0.892) 5.722 (0.858) 5.882 (0.937) 5.563 (0.743)
Years of schooling 11.48 (3.395) 11.77 (3.368) 9.995 (3.361) 10.76 (3.146) 10.92 (3.572) 10.91 (2.690)
Party member (1993) 0.191 0.220 0.057 0.108 0.139 0.076
Male 0.558 0.556 0.593 0.519 0.620 0.418
Age 41.51 (9.232) 42.30 (9.341) 38.5 (8.928) 38.92 (8.523) 39.23 (10.25) 38.61 (6.402)
Layoff since 1993 0.100 0.037 – 0.500 – –
Manufacturing (1993) 0.367 0.362 0.258 0.544 0.494 0.595
Professionals/cadres (1993) 0.330 0.375 0.129 0.209 0.215 0.203
Province
Guangdong 0.182 0.165 0.289 0.177 0.291 0.063
Hu’nan 0.178 0.194 0.088 0.152 0.051 0.253
He’nan 0.143 0.140 0.134 0.177 0.165 0.190
Tianjin 0.147 0.155 0.113 0.127 0.152 0.101
Jilin 0.177 0.178 0.180 0.184 0.203 0.165
Shaanxi 0.173 0.167 0.196 0.184 0.139 0.228
Number of cases 1669 1303 194 158 79 79
a
Including 14 workers who were in the market sector in 1993 but moved to the state sector in 2000 (market losers).
b
Figures in parentheses are standard deviations for continuous variables.

The two criteria are then combined to define the number of workers in the market sector and in the state sector. Workers
are coded as being in the market sector if they meet any of the two criteria.6 This corresponds to the broad measure employed
by Wu and Xie (2003). As the bottom line of Table 1 shows, among 1,669 respondents included in the analysis, 194 worked in
the market sector before 1993 and stayed in 2000 (‘‘early birds”). During the same period, 158 who were in the state sector in
1993 had moved into the market sector by 2000 (‘‘later entrants), with the rest continuing to stay in the state sector (‘‘stayers”).
The later entrants and early birds, together with those who entered the market after 1993, constituted workers in the market
sector, accounting for 26 percent of all urban workers in 2000.7
The survey asked the respondents whether they had experienced a layoff (xiagang), and if so, when the most recent layoff
occurred. Based on the information, a dichotomous distinction is made between voluntary and involuntary later entrants
based on whether or not they had experienced layoff since 1993.
Earnings are the major indicator of social stratification outcomes.8 The survey collected information on the total monthly
income earned in both 1993 and 2000, measured in RMB yuan (where 1 RMB yuan 0.12 USD in 2000). Individual character-
istics commonly used in the studies of earnings determination include education, party membership, age and gender. Education,
which denotes human capital, is a continuous variable measured in years of schooling. Whether the respondent is a party mem-
ber or not as of 1993 measures his/her political capital, which is coded as a dummy variable (yes = 1). Age is measured as a
continuous variable, and gender is coded as a dummy variable, with male as 1 and female as 0.
Industry and occupation plays important roles in linking individuals’ labor market transitions and institutional changes
(Zhou et al., 1997). Traditional manufacturing industries in the state sector (e.g., machinery, mining, chemicals, textile,
clothes and foods), have lost their competitiveness in the market transition, and employment downsizing and layoffs took
place mainly in these industries (Solinger, 2001; Xie, 2006). Industry in 1993 is coded as a dummy variable, with traditional
manufacturing industry as 1 and otherwise as 0. A respondent’ occupation in 1993, measured by the 3-digit code of the Chi-
nese Standard Classification of Occupation (CSCO), is also coded as a dummy variable (professionals/leading cadres = 1).
Finally, the large regional variations in economic structures and the pace of the reforms have created varying opportuni-
ties for entering the market sector and consequently for income inequality (Hauser and Xie, 2005; Xie and Hannum, 1996).
For example, Guangdong accounts for 28.9 percent of the early birds and also 29.1 percent of voluntary later entrants,

6
The two criteria are indeed overlapped: both can be used to identify the self-employed and private entrepreneurs. However, the first criterion additionally
captures wage employees in the market sector.
7
Given the structural change in employment opportunities, those who entered the labor force after 1993 had a higher propensity of entering the market
sector directly than their older counterparts. Among 451 of these workers, 42 percent were in the market sector in 2000. These young later entrants were 27.7
years old and, on average, received 11.6 years of schooling and earned 1121 RMB yuan per month. Together with the later entrants who had been in the labor
force before, they contributed to the rapid employment growth in the market in the 1990s and characterizing the market sector as of 2000. Because this paper is
mainly concerned with individuals’ job shifts from the state sector to the market sector, those who entered the labor force after 1993 are excluded from the
analysis.
8
It should be noted that using earnings only to characterize the distinction between the state and market sectors would substantially underestimate the
economic returns for those in the state sector, which also cover fringe benefits such as health care, housing, and pension, etc.

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
X. Wu / Social Science Research xxx (2009) xxx–xxx 7

whereas involuntary later entrants are mainly concentrated in Hunan (25.3 percent) and Shaanxi (22.8 percent). Five dummy
variables are included in the model to represent the provinces from which the samples were drawn.
Table 1 presents descriptive statistics for the entire sample and the comparisons among three/four types of workers. For
instance, in 2000, early birds on average earned 1593 RMB yuan per month and later entrants earned 1425 RMB yuan per
month, compared to stayers who earned 944 RMB yuan per month but might receive more benefits. In 1993, their respective
monthly earnings were 1295 RMB yuan, 448 RMB yuan and 526 RMB yuan. Therefore, it seems that, by entering the market
sector, later entrants gained in terms of a higher income from 1993 to 2000.
Pertaining to the two variables—education and the Chinese Communist Party membership (1993)—which are of partic-
ular interest to some scholars, stayers have the highest years of schooling, whereas early birds have the least. Of the stayers,
22 percent were party members in 1993, but only 5.7 percent of early birds and 10.8 percent of later entrants were party
members in 1993. Such observations are consistent with the findings reported elsewhere: as the reforms proceeded, while
the redistributive state in China continued to be the dominant provider of career opportunities for both the professional and
political elite (Walder et al., 2000), the market started offering opportunities that increasingly attracted them (Wu, 2006). As
Table 1 shows, only 12.9 percent of early birds and 20.9 percent of later entrants were professionals/leading cadres in 1993,
in contrast to 37.5 percent of those who stayed in the state sector.9
Of all the surveyed workers in 2000, 10 percent had experienced a layoff (xiagang) between 1993 and 2000. Most of the
laid off workers were highly concentrated among later entrants in the market sector, as this had become new phenomenon
only since the mid-1990s. In fact, 50 percent of later entrants were former xiagang workers in the state sector, whereas few
laid off workers could have found another job within the state sector; they constituted only 3.7 percent of all current
stayers.10
The above observation can be further buttressed by the structural transformation of workers across industries since 1993.
Indeed, 54.4 percent of later entrants were from the traditional manufacturing industry in 1993, and the share among invol-
untary later entrants (presumably due to layoff) was even higher. As of 2000, over 80 percent of later entrants and early birds
are in the service sector.
The following empirical analyses consist of two parts: First, I examine the determinants of transition to the market sector
via multinomial logistic regressions; I then link the patterns to earnings inequality among different types of workers via lin-
ear regression methods, paying particular attention to voluntary and involuntary later entrants. Second, I employ the pro-
pensity-score matching methods to directly estimate the causal effect of a late market entry on earnings gains in 2000.
Because the sample was clustered within 52 street communities (jiedao) in each city, an adjustment of standard errors is
needed in regression analyses. In all model estimations, data are weighted to represent the general population of the selected
cities and robust standard errors are reported.

5. Patterns of entry into the market sector and earnings determination

Table 2 presents estimates for multinomial logit models that predict the likelihood of workers moving from the state to
the market sector under different scenarios. Respondents’ years of schooling, party membership as of 1993, age, age2, gender,
industry in 1993, and province are included as the covariates.11 First, the models among stayers, early birds and later entrants
are estimated, and then with later entrants divided into voluntary and involuntary later entrants, models are re-estimated and
stayers are used as the reference group. The following discussion is focused on the interpretation of the effects of schooling,
party membership, and industry, omitting those of other control variables.
As the models indicate, years of schooling are negatively associated with the likelihood of early entry into the market sec-
tor, and party members are also less likely than are non-party members to be in the market sector. This is because people
with more human and political capital would have more career opportunities in the state sector and also enjoy more fringe
benefits (e.g., housing, health care, and pension) other than cash income. However, as the marketization proceeds, such
deterring effects tend to fade away (Wu, 2006), as suggested by the difference in coefficients between the equations for early
birds and later entrants. More specifically, an additional year of schooling decreases the net odds of entering the market sec-
tor by 18.1 percent (e0.200  1) (p < .001) in an early period (in or before 1993), but only by 9.1 percent (e0.096  1) in a late
period (after 1993) (p < .01). Similarly, controlling for other variables, a party member’s odds of entry into the market are
only 34.2 percent (e1.072) (p < .05) of those for a non-party member in an early period, but the odds increase to 79.1 percent
(e0.235) and become statistically insignificant in a late period. A Wald test shows that the change in coefficients are statis-

9
Early birds’ major occupations include ‘‘individual business-owners” (geti hu), ‘‘private enterprise owners” (siying qiye zhu), and ‘‘salesmen/saleswomen”,
which could not attract those with human capital and political capital in early years; and ‘‘private enterprise owners” earned much higher than others and
thereby lifted the average earnings for early birds as a group.
10
Because xiagang arrangement does not apply to employees in the private sector and there is no information on the year when the respondent entered the
market sector, it should be noted that xiagang would not immediately lead to move into in the market sector, and that involuntary later entrants’ current job
may not necessarily be the result of the state layoff. Even for those who were first laid off from the state sector and landed a job in the market sector but
changed jobs within the market sector, they are still classified as involuntary later entrants, because the main concerns of this paper is to the impact of job
mobility across sectors on earnings inequality.
11
The inclusion of occupation as an independent variable in the model is complicated by the fact that it is also employed to define the type of workers moving
across sectors in the labor market.

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
8 X. Wu / Social Science Research xxx (2009) xxx–xxx

Table 2
Estimated coefficients from multinomial logit models of entry into the market sector in Chinese cities, 2000.

Variables Early birds vs Later entrants vs. Voluntary later entrants vs. Involuntary later entrants vs.
stayers stayers stayers stayers
Years of schooling 0.200*** (0.038) 0.096** (0.035) 0.105*(0.048) 0.080  (0.044)
Party membership (1993) 1.072* (0.484) 0.235 (0.298) 0.275 (0.369) 0.935* (0.452)
Age 0.197** (0.067) 0.085 (0.085) 0.373*** (0.091) 0.548** (0.171)
Age2*10 0.015 (0.008) 0.003 (0.010) 0.036*** (0.010) 0.076** (0.024)
Male 0.330 (0.179) 0.156 (0.188) 0.211 (0.328) 0.500* (0.199)
Manufacturing (1993) 0.724** (0.241) 0.653*** (0.202) 0.391 (0.345) 0.974*** (0.270)
Province [Guangdong omitted]
Hu’nan 0.601* (0.297) 0.085 (0.401) 0.763 (0.670) 0.556 (0.419)
He’nan 0.018 (0.330) 0.738  (0.387) 0.584 (0.513) 0.883* (0.441)
Tianjin 0.195 (0.320) 0.009 (0.426) 0.294 (0.580) 0.384 (0.552)
Jilin 0.086 (0.276) 0.392 (0.452) 0.536 (0.579) 0.219 (0.580)
Shaanxi 0.234 (0.251) 0.493 (0.440) 0.207 (0.582) 0.680 (0.492)
Constant 5.847*** (1.304) 1.404 (1.797) 6.506** (2.206) 11.74*** (3.095)
Pseudo R2 0.092 0.104
Wald v2 227.5 592.4
Degree of freedom 22 33

Notes: Data are weighted, numbers in parentheses are robust standard errors adjusted for clustering on community (jiedao).
 
p < .10.
*
p < .05.
**
p < .01.
***
p < .001.

tically significant (v2[2] = 6.8, p < .05), suggesting that early birds and later entrants indeed are different in terms of their
possession of human capital and political capital, although they both differ from stayers at the same time.
As previously pointed out, later entrants may be more heterogeneous because they consist of two subgroups of people
who entered the market sector via different institutional channels. Hence, later entrants are further divided into voluntary
and involuntary entrants based on whether they had layoff (xiagang) experience or not since 1993, and the multinomial logit
models are re-estimated among stayers, early birds, voluntary later entrants, and involuntary later entrants. Results in the
right two columns of Table 2 show that, while schooling seems to play a similar role in deterring one from entering the mar-
ket, both voluntarily and involuntarily (v2[1] = 0.17, p = 0.68), party membership plays an opposite role: party members are
less likely than non-party members to become involuntary later entrants (p < .05), but they are more likely to become vol-
untary later entrants, as opposed to staying in the state sector. As expected, those from traditional manufacturing industry in
1993 were much less likely to become early birds, but they are more likely to become later entrants (specifically, involuntary
later entrants).
The analyses from multinomial logistic regression above have revealed that early birds, later entrants (both voluntary and
involuntary), and stayers differ in certain characteristics. In other words, the patterns of entry into the market sector are con-
tingent upon when and how people moved from the state sector into the market sector. Such variant patterns have impor-
tant implications for the stratification outcomes (earnings), which are examined next.
The logarithm of monthly earnings is used as the dependent variable in the multiple linear regression analysis. Worker
types, the logarithm of monthly earnings in 1993, schooling, party membership in 1993, age, age2, gender, industry in 1993,
and province are included as independent variables. The analysis is focused on mean difference in earnings among three/four
groups of workers, adjusting for the effects of other covariates. Table 3 presents the estimated results.
In Model 1, all workers are divided into three groups: stayers, early birds, and later entrants, and included in the model as
two dummies, with stayers as the reference group. Because the model controls the earnings prior to 1993, the net effect of an
independent variable on the logged earnings in 2000 can be interpreted as that on earnings change. Similar to the results
reported by Bian and Logan (1996), education and party membership both are significant predictors of earnings growth from
1993 to 2000. Men enjoy significantly higher earnings gains than did women in the same period (p < .01), and those who
worked in the traditional manufacturing industry in 1993 continued to be disadvantaged in 2000 (p < .05). The effect of
age, nonetheless, is insignificant.
Controlling for other factors, Model 1 suggests that only those who have shifted from the state sector to the market sector
after 1993 enjoyed significantly higher earnings growth than did those who remained in the state sector, by 26.4 percent
(e0.234  1) (p < .05). This result provides some support to Wu and Xie’s (2003) argument that people with more human cap-
ital and political capital were self-selected into the market in the late reform stage, who tended to do better economically
(also see Szelényi and Kostello, 1996; Xie and Wu, 2005).
Not all later entrants were self-selected. In Model 2 of Table 3, later entrants are further divided into voluntary and invol-
untary later entrants, and enter the models as two dummy variables. The patterns become even more evident: neither invol-
untary entrants nor early birds differ from stayers in earnings; higher earnings are only limited to those who voluntarily
entered the market after 1993. Other things being equal, voluntary later entrants enjoy 47.1 percent (e0.386  1) higher earn-
ings change than stayers, and the difference is statistically significant (p < .01).

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
X. Wu / Social Science Research xxx (2009) xxx–xxx 9

Table 3
Estimated coefficients from the multiple linear regression of earnings in selected Chinese cities, 1993–2000.

Variables Model 1 Model 2

Worker type (stayers omitted)
Early bird 0.075 (0.057) 0.079 (0.056)
Later entrants 0.234* (0.110) —
Voluntary — 0.386** (0.134)
Involuntary (via layoff) — 0.093 (0.111)
Logged earnings (1993) 0.444*** (0.031) 0.444*** (0.031)
Years of schooling 0.039*** (0.005) 0.040*** (0.005)
Party member (1993) 0.102** (0.037) 0.097* (0.037)
Age 0.009 (0.018) 0.005 (0.017)
Age2*10 0.001 (0.002) 0.001 (0.002)
Male 0.133*** (0.032) 0.130*** (0.033)
Manufacturing (1993) 0.089* (0.042) 0.086* (0.042)
Province [Guangdong omitted]
Hu’nan 0.149* (0.071) 0.144* (0.071)
He’nan 0.310*** (0.082) 0.310*** (0.083)
Tianjin 0.282*** (0.062) 0.288*** (0.063)
Jilin 0.236*** (0.069) 0.240*** (0.068)
Shaanxi 0.317*** (0.050) 0.317*** (0.050)
Constant 3.900*** (0.435) 3.837*** (0.433)
R2 0.419 0.423

Notes: Data are weighted, numbers in parentheses are robust standard errors adjusted for clustering on community (jiedao).
*
p < .05.
**
p<.01.
***
p < .001.

It should be noted that ‘‘voluntary” here refers to those who did not experience a layoff (xiagang) since 1993. Such a sim-
ple dichotomous distinction between involuntary and voluntary later entrants cannot fully capture the group heterogeneity
this paper intends to highlight, because people may be forced to depart due to reasons other than xiagang. In fact, the results
from multinomial logistic regression show that people with higher education are still less likely to become ‘‘voluntary” later
entrants, but once they do, they enjoy higher income growth on average than those stayers. A possible explanation is that the
income advantages of later entrants may be due to the self-selection effects, which cannot be demonstrated with the sep-
arate analyses of multinomial logistic regression and linear regression. To establish such a causal link between labor market
transitions on earnings growth, I employ a propensity-score matching method.

6. Estimating the causal effect of market entry with the propensity-score matching method

In the causal inference framework, the matching method allows researchers to summarize all the differences between the
treatment and control groups with propensity scores, which are estimated from binary logit models. The propensity scores
are then stratified to balance both propensity scores and observed covariates, and the average treatment effect on earnings
within each stratum can be computed. A large body of literature has shown that the propensity score stratification method
can remove great amounts of biases in causal inference with observational data (e.g., Becker and Ichino, 2002; Dehejia and
Wahba, 1999; Winship and Morgan, 1999), affording us an easy way to examine differences in observed covariates and com-
pare the groups flexibly and non-parametrically (for recent applications in sociological research, see Brand and Halaby, 2005;
Harding, 2003; Morgan, 2001; Smith, 1997; Xie and Wu, 2005).
For the research question in this paper, the focus is explicitly on the between-group differences in earnings, whereas
group differences in observed covariates that affect earnings are encompassed by the propensity scores. An individual’s pro-
pensity for entering the market vs. staying in the state sector since 1993 can be estimated on a number of selected variables:
the logged income in 1993, years of schooling, party membership in 1993, age, age2, gender, xiagang experience, job satis-
faction in 1993, and province.12 Based on estimated propensity scores, respondents are divided into different strata to balance
both propensity score and covariates between the treatment and comparison groups (p < .001) (Becker and Ichino, 2002). As a
result, eight strata are obtained. Fig. 2 presents the number of cases in each stratum, separated by treatment and control
groups.13

12
Job satisfaction affected workers’ tendency for mobility. The more the workers were satisfied with their job in the state sector, the less likely they were to
change jobs and enter the market sector. Job satisfaction in 1993 was measured on a scale from 1 to 4 and treated as a continuous variable in the analysis. Other
variables such as occupation, industry, and welfare benefits from work units are also explored in the binary logit models. Their effects are not statistically
significant and the results from propensity-score matching analysis remain largely the same.
13
The pscore command in STATA was used to implement the stratification matching (Becker and Ichino, 2002). The values on the X-axis are the inferior
boundary of the eight propensity score strata (subsequently numbered as 1–8). The 11 workers in the control group whose estimated propensity score is less
than the minimum estimated propensity score for the treatment group are discarded, because they cannot be matched to any individuals in the treatment
group.

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
10 X. Wu / Social Science Research xxx (2009) xxx–xxx

1020 154
100
treatment(laterentrants) control(stayers)
90

80

70

Frequency
60

48
50
42

40

30
24
21 21
19
20 15
17
13 13
11
8
10 7
5

0
0.031 0.098 0.204 0.359 0.499 0.641 0.788 0.893
Propensity Score Stratum

Fig. 2. Histogram of the estimated propensity score for later entrants (treatment group) and stayers (control group).

The figure displays vividly how later entrants differ from other workers in observed covariates. Among workers who did
not make the transition, most of them have very low propensities for making such a transition. In contrast, workers who did
make the transition have relatively higher predicted propensity scores. Without the achieved balance through propensity
score stratification, the two subgroups would be incomparable. The non-comparability of cases is usually masked in esti-
mates based on the traditional linear regression models.
To answer the question of whether or not the transition to the market has caused an increase in earnings, detailed anal-
yses of the treatment effect of a market entry are conducted in three steps. First, the treatment effects specific to the pro-
pensity score strata were estimated. Because there are no systematic differences between the treatment group (entrants
to the market sector at a given time) and the control group (stayers in the state sector at a given time), the average earnings
differences within a propensity score stratum can thus be interpreted as the average causal effect of the market entry for that
stratum (Rosenbaum and Rubin, 1984). Second, the results across strata were pooled under the assumption of a homoge-
neous treatment effect, which was the weighted average of stratum-specific treatment effect.14 Finally, via a hierarchical-lin-
ear model (HLM) (Raudenbush and Bryk, 2002), I show how the treatment effects vary by different propensity strata.15
Fig. 3 presents the summary findings from the analysis of the treatment effects of a late transition. The diamond-shape
dots are point estimates of stratum-specific treatment effects, with corresponding t values given nearby for the null hypoth-
esis of the treatment effect is zero. As the figure shows, the treatment effect of making a late transition is relatively large and
significantly different from zero for the three lowest propensity strata. If the different strata are pooled together for an over-
all treatment effect under the homogeneous effect assumption, the estimate is 685 RMB yuan with a standard error of 195,
resulting in a highly significant t value of 3.5.
However, the assumption of the homogeneous treatment effect does not seem to hold true, as indicated clearly by the
downward trend in Fig. 3. The results from the HLM model reveals that the magnitude of the treatment effect is strongly
and negatively contingent on the propensity score, with a unit change in stratum rank (i.e., crossing a propensity score stra-
tum) associated with a reduction of 97.1 RMB yuan in the treatment effect (a significant relationship with t = 3.52 and

14
The atts command in STATA was used to estimate the average treatment effect on the treated using stratification matching of propensity scores (Becker and
Ichino, 2002).
15
The two-level models can be specified as follows:

Level 1 : EARNINGij ¼ b0j þ b1j TREATEDij þ ij ð1Þ

Level 2 : b0j ¼ c00 þ c01 STRATUM RANK j þ m0j ð2bÞ

b1j ¼ c10 þ c11 STRATUM RANK j þ m1j ð2cÞ

where in Eq. (1), EARNINGij stands for current monthly earnings for ith in jth propensity score stratum; TREATEDij is a dummy variable indicating whether or
not the ith in jth stratum is indeed in the market sector. b0j is the intercept and the coefficient for the dummy variable,b1j , denotes the mean difference in
earnings between the treatment group and control group within the jth stratum, i.e. the causal effect of treatment within the stratum; ij , is the error term
at individual level. In Level 2 Model, Eq. (2b) allows the intercept of level-1 b0j also to vary with different strata (STRATUM RANK j ), and (2c) also allows the
treatment effect b1j to vary with different strata (STRATUM RANK j ), in which the integer-score rank of a propensity score stratum is used as the predictor of
the intercept b0j and treatment effect b1j , c01 and c11 are respective coefficients, m0j and m1j are error terms at the stratum level. In the following analysis, I
focus on the coefficient c11 in (2c), which indicates how the treatment effect varies by propensity score stratum.

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
X. Wu / Social Science Research xxx (2009) xxx–xxx 11

3.17 observed linear fit

1200

Market Effect on Earnings (RMB Yuan)

Y=1067.1-97.1*STRATUM_RANK
1000
1.5
4.60
800

0.85

600
2.26
0.77

400

200 1.38
0.43

0
1 2 3 4 5 6 7 8
Propensity Score Stratum

Fig. 3. Market treatment effect on earnings by propensity strata: later entrants vs. stayers. Notes: (a) Numbers in the scatter plot are t values for earnings
comparison between later entrants (treatment group) and stayers (control group). A t value less than 1.96 indicates that there is no significant difference in
earnings between the treatment and control groups within a propensity score stratum. (b) The linear plot is based on the hierarchical-linear model (HLM)
estimates (level-2 model with slopes from level-1 model as outcomes regressed on propensity stratum rank. For model specification, see (2c) in footnote
15). The effect of the propensity stratum rank is statistically significant (std = 28.2, t = 3.45, and p = 0.017).

p < .001). That is, the benefit of a late transition into the market sector is the greatest among those who were least likely to
make such a transition and diminishes with the propensity for making the transition.16

7. Summary and discussion

To summarize, the analyses of this paper revealed that the pattern of entry into the market sector in urban China has
changed since 1993. While early entrants to the market tend to come from the lower tiers of the social hierarchy, with less
human capital and political capital, the disadvantages of later entrants have been reduced. Compared to early birds, for later
entrants, while education continues to play a significant role in deterring one from entry into the market, such deterring ef-
fect is weakened; the effect of party membership becomes positive and statistically insignificant.
Later entrants are recruited through two distinct processes. On the one hand, as the market reform proceeded, state lay-
offs took place and workers lacking educational credentials and political connections were thrown into the market for sur-
vival. On the other hand, the growth of the market economy also increasingly attracted from the state sector more capable
workers who had been able to cash their human capital and political capital for even greater economic advantages. Regres-
sion analysis reveals that those later entrants who entered the market involuntarily following a state layoff have no earnings
advantages compared to both stayers and earlier birds, and that the higher average earnings growth in the market sector is
limited to a subgroup of workers who entered the market sector voluntarily.
Based on an alternative operationalization of the distinction between involuntary and voluntary later entrants, it is con-
firmed that a late transition to the market sector is associated with higher earnings. Furthermore, among those later entrants,
the benefit of working in the market sector sharply decreases with their propensity for having made the transition. Hence,
the market premium is only limited to those later entrants who otherwise have a low propensity for making such a transition
to the market sector. These workers could have been doing quite well in the state sector but indeed ended up in the market
sector, and who are labeled as the voluntary later entrants in contrast to involuntary ones.
The propensity-score matching method is by no means perfect, because it takes into account of the effects of only obser-
vable characteristics and is unable to address the selection of unobservable characteristics (also see footnote 3). This limi-
tation is alleviated to a large extent given the fact that respondents’ pre-treatment earnings (earnings in 1993) are
included in estimating propensity scores, hence the effects of individuals’ unobserved characteristics on earnings are fixed
in this setup. The results using the matching method confirm the patterns observed in regression analysis based on the
dichotomous distinction between voluntary and involuntary entrants and therefore are robust.

16
While one might worry about whether or not the small number of level-2 units (8 strata) can afford estimation of a hierarchical–linear model, the number
of strata created is not fixed, dependent upon the data available. The key point here, however, is to parametrically show a varying treatment effect across
propensity score strata.

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
12 X. Wu / Social Science Research xxx (2009) xxx–xxx

Voluntary Entrants Layoffs

100

90

80

70

Percentage 60

50

40

30

20

10

0
1 2 3 4 5 6 7 8
Propensity Score Stratum

Fig. 4. The composition of two types of later entrants by propensity score stratum.

Indeed, the two approaches to operationalizing the distinction between voluntary vs. involuntary later entrants are quite
consistent. As Fig. 4 shows, in the three strata with the lowest propensity scores, most workers had not experienced layoff,
whereas in the three strata with the highest propensity scores, all workers experienced layoff. Workers who do well in the
state sector and thus are unlikely to lose their jobs have a low propensity for entering the market. For them, the attraction of
the market sector needs to be large enough to overcompensate for the advantages they already enjoyed in the state sector,
including fringe benefits that are less available in the market sector. Higher earnings of this subgroup of later entrants lifted
the average earnings of the workers in the market sector as a single group.
Hence, selectivity plays a key role in determining higher returns to the market sector employment compared to the state
sector employment. Without attending to the process through which individuals are selectively allocated or sorted into a
group, relying on the comparison of group means to address the causal effect of being in that group may lead to wrong
conclusions.
Why do some later entrants perform so well? The propensity-score matching analysis has illustrated a classical case of
the endogeneity problem: individuals select their ‘‘treatment” based on the anticipated outcome, which is not homogeneous
across workers. This suggests a strong self-selection mechanism at work: when workers with a low (latent) propensity of
making a transition indeed did make such a late transition to the market sector, they benefited the most from the transition.
In other words, it is a story about what drives workers to move into the market sector, and our research should be devoted to
understanding this micro-process and its economic consequences.
To be certain, the self-selection and sorting process cannot be fully understood without appreciating the changing
employment opportunities in the labor markets. Since the 1990s, more corporate jobs became available in the private sector
and foreign companies, given the development of labor markets and capital markets in China. The market started attracting
more and more talents, resulting in a more heterogeneous body of workers in the sector, and high earnings returns to the
employment in the market sector.17
How to interpret the apparent returns to market employment among those who choose it? We believe that individuals’
unobserved heterogeneity plays an important role in explaining the advantages, that is, those who have low observed pro-
pensity to make the transition and actually do would have higher earnings in either sector; what appears to be a return to
market employment for them actually reflects their higher expected earnings and non-monetary benefits in the state sector,
which raises the reservation wage at which they would take employment in the market sector.

8. Conclusions

This paper posited an existing intellectual gap in the literature on modern social stratification in general and on post-so-
cialist stratification in particular. Scholars often jump from the observed structure of inequality to the causal mechanisms on
the theory of social stratification, without solid knowledge of the underlying social processes that generate the stratification

17
Such a pattern is by no means unique in China’s transition economy. Studies of Poland, another former socialist country, revealed a similar story: wage
premium in the private sector is particularly pronounced for highly educated workers; and the positive selection effects into the private sector can be identified
for both men and women (Adamchik and Bedi, 2000).

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
X. Wu / Social Science Research xxx (2009) xxx–xxx 13

outcomes. Commenting on this gap, the late Hubert M. Blalock (1991) advises that researchers should delimit their theories
to what he refers to as allocation processes, through which individuals are assigned or sorted into positions by a series of
micro-level decisions. ‘‘Without benefit of more micro analyses, such macro theories are likely to require so many untested
assumptions, and to ignore such huge data gaps, our intellectual and ideological biases are likely to predominate, resulting in
unanswerable theoretical disputes that merely hamper the process of arriving at a cumulative body of knowledge” (Blalock,
1991: 27).
This paper proposed a micro perspective to shed new lights on the relationship between the institutional transition to a
market economy and the changes in earnings inequality in China. It emphasized the differential sorting processes of workers
in the emerging labor markets in understanding how earnings inequalities are produced in the course of China’s economic
transition. Using both the linear regression models and the propensity-score matching method, the paper found that work-
ers’ performance in the market sector is contingent upon how they entered the market, and only those voluntary entrants in
the late reform period enjoyed earnings advantages over those who stayed in the state sector; moreover, among those later
entrants, the causal effect of a market entry on earnings is negatively associated with the propensity for making such a tran-
sition. Those who would otherwise do well in the state sector and therefore have lower propensity of entering the market
have benefited more from the entry.
While the paper put the emphasis on the impact of workers’ self-selection processes on earnings inequality, such self-
selection processes were contingent upon the institutional transitions at the macro-level and interacting with the change
in opportunity structures in the labor markets (Wu, 2006). Future data collection and research should link the two processes
and specifically examine the career trajectory among those self-selected later entrants.
The findings from this paper suggest that researchers should be cautioned about the danger in reliance on group compar-
isons to make theoretical causal inferences without fully acknowledging the process of how individuals are sorted into a
group and the resulting heterogeneity. This methodological advice is equally applicable to the study of inequality between
party members and non-party members (e.g. Gerber, 2000), between cadres and non-cadres (Nee, 1996; Walder, 2002), and
among different types of work organizations (Zhou, 2000a,b). Hence, a simplistic causal explanations of post-socialism earn-
ings inequality in terms of the state or the market mechanisms is rejected, and further research is to be devoted to uncover
the underlying micro-process of individual mobility between social positions (constrained and influenced by macro-level
institutional transitions), and the aggregate outcomes of these processes (Blau, 1977; Hannan, 1991). Identifying these pro-
cesses is an intellectually challenging task but a necessary step towards constructing a solid theory on the changing strat-
ification mechanism in post-socialist transitional societies.

References

Adamchik, Vera A., Bedi, Arjun S., 2000. Wage differentials between the public and the private sectors: evidence from an economy in transition. Labor
Economics 7, 203–224.
Becker, Sascha O., Ichino, Andrea, 2002. Estimation of average treatment effects based on propensity scores. The Stata Journal 2 (4), 358–377.
Bian, Yanjie, Logan, John, 1996. Market transition and persistence of power: the changing stratification system in urban China. American Sociological
Review 61, 739–758.
Blalock, Hubert.M., 1991. Understanding Social Inequality. Sage, Newbury Park, CA.
Blau, Peter M., 1977. Inequality and Heterogeneity. Free Press, New York, NY.
Brand, Jennie E., Halaby, Charles, 2005. Regression and matching estimates of the effects of elite college attendance on educational and career achievement.
Social Science Research 35, 749–770.
Cai, Yongshun, 2002. The resistance of chinese laid-off workers in the reform period. The China Quarterly 170, 327–344.
Cao, Yang, Nee, Victor, 2000. Comment: Controversies and Evidence in the Market Transition Debate. American Journal of Sociology 105, 1175–1189.
Davis, Deborah, 1999. Self-employment in Shanghai: a research note. China Quarterly 157, 22–43.
Dehejia, H. Rajeev, Wahba, Sadek, 1999. Causal effects in non-experimental studies: reevaluating the evaluation of training programs. Journal of American
Statistical Association 94, 1053–1062.
DiPrete, Thomas A., 1993. Industrial restructuring and the mobility response of American workers in the 1980s. American Sociological Review 58, 74–96.
DiPrete, Thomas A., Lynn Nonnemaker, K., 1997. Structural change, labor market turbulence, and labor market outcomes. American Sociological Review 62,
386–404.
Gerber, Theodore, 2000. Membership benefits or selection effects? Why former communist party members do better in post-soviet Russia. Social Science
Research 29, 25–50.
Gerber, Theodore P., Hout, Michael, 1998. More shock than therapy: market transition, employment, and income in Russia, 1991–1995. American Journal of
Sociology 104, 1–50.
Gerber, Theodore, 2002. Structural change and post-socialist stratification: labor market transitions in contemporary Russia. American Sociological Review
67, 629–659.
Gold, Thomas, 1990. Urban private business and social change, In: Davis, D., Vogel, E.F., (Eds.), Chinese Society on the Eve of Tiananmen: the Impact of
Reform. Council on East Asian Studies, Harvard University, Cambridge, MA, pp. 157–180.
Harding, David, 2003. Counterfactual model of neighborhood effects: the effect of neighborhood poverty on high school dropout and teenage pregnancy.
American Journal of Sociology 109 (3), 676–719.
Hannan, Michael, 1991. Aggregation and Disaggregation in the Social Sciences. Lexington Books, Lexington, MA.
Hauser, Seth, Xie, Yu, 2005. Temporal and regional variation in earnings inequality: urban China in transition between 1988 and 1995. Social Science
Research 34, 44–79.
Holland, Paul W., 1986. Statistics and causal inference (with discussion). Journal of American Statistical Association 81, 945–970.
Jann, Ben, 2005. Earnings returns to education in urban China: a note on testing difference among groups. American Sociological Review 70, 860–865.
Lee, Hong Yung, 2000. Xiagang, the Chinese style of laying off workers. Asian Survey 40 (6), 917–937.
Li, Qiang, 1993. Dangdai Zhongguo de Shehui Fenceng Yu Liudong (Social Stratification and Mobility in Contemporary China). China Economy Press, Beijing
(in Chinese).
Manski, Charles, 1995. Identification problems in the social sciences. Harvard University Press, Boston, MA.
Morgan, Stephen L., 2001. Counterfactuals, causal effect heterogeneity, and the catholic school effect on learning. Sociology of Education 74, 341–374.
National Bureau of Statistics, PRC, 2000. China Statistical Yearbook. China Statistics Press, Beijing, China.

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003
 ARTICLE IN PRESS
14 X. Wu / Social Science Research xxx (2009) xxx–xxx

Nee, Victor, 1989. A theory of market transition: from redistribution to markets in state socialism. American Sociological Review 54, 663–681.
Nee, Victor., 1996. The emergence of a market society: changing mechanisms of stratification in China. American Journal of Sociology 101, 908–949.
Parish, William.L., Michelson, Ethan., 1996. Politics and markets: dual transformations. American Journal of Sociology 101, 1042–1059.
Raudenbush, Steve W., Bryk, Anthony S., 2002. Hierarchical Linear Models, second ed. Sage Publication, Thousand Oaks.
Róna-Tas, Akos, 1994. The first shall be last? Entrepreneurship and communist cadre in the transition from socialism. American Journal of Sociology 100,
40–69.
Rosenbaum, Paul R., Rubin, Donald B., 1984. Reducing bias in observational studies using sub-classification on the propensity score. Journal of American
Statistical Association 79, 516–524.
Smith, Herbert.L., 1997. Matching with multiple controls to estimate treatment effect in observational studies. Sociological Methodology 27, 325–343.
Solinger, Dorothy, 2001. Why we cannot count the ‘unemployed’. The China Quarterly 167, 671–88.
Solinger, Dorothy., 2002. Labor market reform and plight of the laid-off proletariat. The China Quarterly 170, 304–326.
Szelenyi, Ivan, 1978. Social inequalities in state socialist redistributive economies. International Journal of Comparative Sociology 19, 63–87.
Szelényi, Ivan, 1988. Socialist Entrepreneurs. University of Wisconsin Press, Madison, WI.
Szelényi, Ivan., Kostello, Eric., 1996. The market transition debate: toward a synthesis. American Journal of Sociology 101, 1082–1096.
Walder, Andrew.G., 2002. Markets and income inequality in rural China: political advantage in an expanding economy. American Sociological Review 67,
231–253.
Walder, Andrew.G., 2003. Elite opportunity in transitional economies. American Sociological Review 68, 899–916.
Walder, Andrew, Li, Bobai, Treiman, Donald, 2000. Politics and life chances in state socialist regime: dual career paths into the urban Chinese elite, 1949–
1996. American Sociological Review 65, 191–209.
Wank, David., 1999. Commodifying Communism: Business, Trust, and Politics in a Chinese City. Cambridge University Press.
Winship, Christopher., Morgan, Stephen.L., 1999. The estimation of causal effects from observational data. Annual Review of Sociology 25, 659–707.
Wu, Xiaogang., 2002. Work units and income inequality: the effect of market transition in urban China. Social Forces 80 (3), 1069–1099.
Wu, Xiaogang, 2006. Communist cadres and market opportunities: entry into self-employment in urban and rural China. Social Forces 85, 389–411.
Wu, Xiaogang., Xie, Yu., 2003. Does the market pay off? Earnings returns to education in urban China. American Sociological Review 68, 425–442.
Xie, Guihua, 2006. Shichang Zhuanxin yu Xiagang Gongren (Market transition and laid-off workers. Shehuixue Yanjiu 121 (1), 22–58 (in Chinese).
Xie, Yu., Hannum, Emily., 1996. Regional variation in earnings inequality in reform-era urban China. American Journal of Sociology 102, 950–992.
Xie, Yu, Wu, Xiaogang, 2005. Market premium, social process, and statisticism. American Sociological Review 70, 865–870.
Xie, Yu, Wu, Xiaogang, 2008. Danwei profitability and earnings inequality in urban China. The China Quarterly 195 (September), 558–581.
Zhou, Xueguang., 2000a. Economic transformation and income inequality in urban China. American Journal of Sociology 105, 1135–1174.
Zhou, Xueguang., 2000b. Reply: beyond the debate and toward substantive institutional analysis. American Journal of Sociology 105, 1190–1195.
Zhou, Xueguang., Tuma, Nancy.Brandon., Moen, Phyllis., 1997. Institutional change and job-shift patterns in urban China, 1949–1994. American Sociological
Review 62, 339–365.

Please cite this article in press as: Wu, X. Voluntary and involuntary job mobility and earnings inequality in urban China, 1993–2000. So-
cial Sci. Res. (2009), doi:10.1016/j.ssresearch.2009.11.003