Labor Market Segmentation in Urumqi, Xinjiang - Wiley Online Library

11 downloads 0 Views 469KB Size Report
Xinjiang: Exposing Labor Market Segments and Testing the Relationship between. Migration and Segmentation. ANTHONY HOWELL. ABSTRACT Labor market ...
Growth and Change Vol. 42 No. 2 (June 2011), pp. 200–226

Labor Market Segmentation in Urumqi, Xinjiang: Exposing Labor Market Segments and Testing the Relationship between Migration and Segmentation ANTHONY HOWELL ABSTRACT Labor market segmentation and migration are two phenomena that are dramatically reshaping the spatial, economic, and social relationships of many urban cities in both developed and developing countries. To this point, the bulk of Chinese literature falls within the context of area studies, without much effort to link Chinese migration and emerging labor market outcomes to larger global trends and discourse. This research attempts to link the body of internal Chinese migration and emerging labor markets to labor market segmentation theory, primarily developed by urban economists and sociologists. My findings provide evidence that applying labor market segmentation theory to examine emerging markets in China offers fruitful results that help to identify the new urban stratification that exists in China. I employ a set of quantitative methods using employee-level field data that I collected in Urumqi in 2008 to identify distinct segments within Urumqi’s labor market and argue that migration is a major driver of labor market segmentation. Cluster analysis shows Uyghur minorities and women are found to be overwhelmingly concentrated in the lower sector, composed mostly of “bad” jobs. Discriminant analysis reveals that migrant status and ethnicity are the most important variables that deepen the gap among the labor market segments. The social inequality created as a result of market segmentation can partially explain Uyghur discontent in the region and the July 2009 riots, one of the worst riots in Xinjiang’s modern history. grow_550

200..226

Introduction

A

midst the interplay of national economic reforms and the development of local labor markets, China represents a rapidly changing region where the new winners and losers of economic reforms tend to be situated along ethnic, Anthony Howell is a doctoral student at the Department of Geography, UCLA, 1255 Bunche Hall Box 951524, Geography Department, UCLA Campus, Los Angeles, CA 90095. His e-mail address is: [email protected]. I would like to acknowledge the Department of Geography and the Graduate School and Asian Studies Center at Michigan State University, as well as the Department of Geography and the Academic Senate at UCLA, for providing funding which supported this research and also to thank the informants at Xinjiang University for their assistance in collecting the data.

Submitted August 2010; revised October 2010; accepted January 2011. © 2011 Wiley Periodicals, Inc

LABOR MARKET SEGMENTATION

201

gender, and migrant status divides. Over the past 30 years, economic reforms in China have led to enhanced labor mobility and the emergence of local labor markets (Xu, Tan, and Wang 2006). As a result, labor migration and emerging labor market segmentation (LMS) are two phenomena dramatically reshaping the spatial, economic, and social relationships in Chinese cities. LMS is important to understand since segmented markets are a major cause of economic inefficiency and source of growing inequality. For example, excluding a certain group of people from specific labor markets leads to wasteful utilization of human resources and incurs a reduction in the flexibility of the labor market to handle economic change (Xu, Tan, and Wang 2006). Migration is equally important to research within the context of studying LMS since scholars have shown that rural–urban migration plays an important role in reinforcing segmentation of the labor markets (Fan 2002, 2003; Gordon 1995; Li 1997). Although a tenuous relationship between migration and segmentation is established in the literature, few studies offer quantitative findings that asserts migration’s role in deepening segmentation (Li 1997). Of the quantitative-based studies, focus concentrates on migrant–native or male–female labor outcomes, thus excluding minority labor market placement in the analysis. This deficiency in the literature has created a large gap regarding our understanding of minority stratification and placement in the labor market, a gap that I attempt to fill in this research. Traditionally, Chinese research on emerging labor markets places greater emphasis on the roles of institutional forces than does traditional Segmented Labor Market (SLM) theory (Fan 2003; Solinger 1999; Zang 2002); whereas Western scholars use SLM theory to explain how minorities, women, and the working class become economically marginalized in society (Gordon 1995; Hudson 2007; Reich, Gordon, and Edwards 1973). In order to overcome this impasse, this research attempts to bridge the two approaches by remaining aware of institutional structures in China while also examining labor market outcomes for minorities, women, and migrants. The merits of this study are two-fold. First, my findings can be linked to theoretical discourses on LMS that have mainly been developed outside of China. There is a commonly held view by China scholars that recent social and economic changes occurring in China deviate from the expectations of conventional models and theoretical frameworks derived from an Anglo-American perspective. A few scholars, however, recognize the disconnect between Chinese studies and the larger bodies of discourse and call for a greater commitment by scholars to test Western theories in the developing world (Lin 2002; Murphy and O’Loughlin 2009). To accept this challenge made by Lin (2002) and others, and diminish the general view that assumes China’s “uniqueness” as given, I use the findings from

202

GROWTH AND CHANGE, JUNE 2011

my research to assert that minority and female placement in the Chinese labor market reflects a similar emergence of segmentation occurring in both developed and developing Western countries. Second, my empirical findings highlight the importance of incorporating ethnicity into the data collecting process. While much scholarly attention in China focuses only on gender and migrant status dichotomies, a comparatively sparse body of literature has focused on Chinese minority placement in the labor market. This is largely the result of empirical labor market research predominately carried out in the ethnically homogenous coastal provinces. I test LMS theory using employee-level field data that I collected in Urumqi in Summer 2008. I use a set of quantitative methods to define segments in the labor market and to reveal the linkages between migration and LMS. Besides, furthering our understanding about Urumqi using LMS theory, this paper also provides a creative and rigorous methodological approach to studying Urumqi’s labor market. The outline of this paper is as follows. First, I provide a summary of the literature in order to situate Urumqi within the context of LMS theory. Since LMS theory has primarily been applied and developed within a Western context, I incorporate a brief review of the literature on China’s internal labor markets to fill in the gaps left by the SLM literature and ensure its applicability in the Chinese context. The second part of the paper attempts to statistically identify whether labor markets exist in Urumqi’s service sector and to determine the role of migration and its impact on deepening the gap among labor market segments. I begin by providing basic information pertaining to Urumqi, the location of the survey site, as well as describe the data collecting process. Basic descriptive statistics are provided to describe the nature of the respondents captured in the data set. Principal components analysis (PCA) and cluster analysis are employed to delineate segments within the labor market and group respondents into the resulting clusters, or segments. Lastly, I highlight the implications of the findings on LMS theory and its applicability to non-Western contexts and elaborate on the nature of Urumqi’s SLM.

LMS Theory According to Hudson (2007), LMS theory supposes that there are clearly identifiable segments in the labor market; segments are clusters of jobs associated with certain combinations of occupation, industry, and respondent characteristics. Hudson (2007) identified three major labor market segments: primaryindependent (PI), primary-subordinate (PS), and secondary. The PI segment is characterized as free from elaborate rules and procedures, and places a premium on creativity and problem solving. Jobs in the PI segment are high-paying, offer

LABOR MARKET SEGMENTATION

203

prestige, and provide workers a high chance of internal advancement. Employment usually requires college education. Professions include doctors, professors, teachers, electricians, and so forth. The PS segment is composed of relatively high-paying jobs; however, jobs follow a routine, task-oriented schedule, and require more direct supervision compared with the PI sector. Job examples include managers of retail stores and restaurants, clerical work, police officer, and so forth. The secondary segment is composed of low-paying jobs with few benefits, little opportunity for job advancement, and generally work long hours in poor or unsafe working conditions. The jobs in this sector are unskilled, highly unstable, and require little formal education (Fichtenbaum et al. 1994). Examples include employees in retail, waiters/ cooks in restaurants, drivers, informal street vendors and repairmen, and so forth. In general, conventional SLM theory focuses on supply-side factors, which include workers’ personal endowments (education and work experience), constraints (language), and preferences (working environment) (Xu, Tan, and Wang 2006). Contrary to neoclassical labor economists who posit that wage differentials are primarily the result of differences in acquired human capital, SLM theorists argue that market segmentation is the result of institutional rules that differ across labor market segments and have thus replaced the market processes of supply and demand (Leontaridi 1998). SLM theory often describes a dual labor market; all jobs fall into one of two separate sectors, e.g., primary and secondary. The primary sector represents good jobs in the labor market, which are marked by high negotiated wages, fringe benefits, and high employment security, whereas the secondary sector represents bad jobs, marked by low skill requirement, low wage rates, and little or no access to career advancement opportunities (Zang 2002). Leontaridi (1998) argued that there are several key tenets central to segmentation theory: the labor market consists of a few clearly identifiable segments; mobility barriers exist and prevent individuals from obtaining jobs in other segments; each segment is subject to a different set of employment and wage setting mechanisms; and neoclassical theory for returns on human capital is not applicable for the lower segment of the labor market (Leontaridi 1998; Zang 2002). When vulnerable groups such as minorities and women become trapped in the lower segment of the market, differing wage setting mechanisms between the primary and secondary sectors are created. In the secondary sector, wages are based on supply and demand; however, the primary sector’s wages are insulated from supply and demand forces. The differing wage mechanisms are created, in part, by the segmentation gap between primary and secondary sectors where excess labor supply outside the primary sector is unable to move across occupational strata, and the concentration of labor supply becomes trapped in the lower sector.

204

GROWTH AND CHANGE, JUNE 2011

The effect of this segmentation gap, therefore, works to keep the primary sector’s labor supply low and also keeps wages artificially low in the secondary sector (Leontaridi 1998). Hence, some researchers argue that different sets of wage earning mechanisms exist for both the primary and secondary sectors. While human capital variables, such as education and work experience, are useful for explaining wage variations in the primary sector, they do not explain wage variation in the secondary sector (Zang 2002). Employers in labor markets play a large role in channeling job seekers into different segments of the market. LMS, in part, stems from employers’ desire to minimize costs and maximize efficiency, which promotes strict regulations and compartmentalization of skills and results in peasant migrants being funneled into a “narrow selection of gender-segregated jobs” (Fan 2003). Western scholars have also emphasized the role of employers in deepening LMS (Gordon 1995; Peck 1996). Employee applicants’ production potentials are not readily known to the employer, and their work performance depends on a combination of factors such as incentives and personal perceptions; therefore, employers are likely to rely on judgments regarding applicant’s personal responsibility, commitment, and development potential. Ultimately, employers must rely on stereotypes to a large degree in deciding whether or not to hire an applicant. Stereotypes are based on such indicators as age, gender, marital status, ethnicity, and work history. Historical evidence shows that disadvantaged groups, such as ethnic minorities, find it difficult to be hired into “good” jobs (Gordon 1995; Zang 2002). According to Peck (1996), the “social nature of labor” is a social phenomenon in which employer stereotypes are constructed outside the market and affect employment relationships to the point of contributing to LMS (Peck 1996). In many Western industrialized countries, including the U.S., a large body of labor market literature dealing with racial, gender, and migrant discrimination exists. Researchers have used empirical evidence to support the existence of wage inequality members of differing social groups who possess similar human capital characteristics; many experts conclude that minority groups are overrepresented in the secondary sector or the bottom tier of the labor market, face wage, and other forms of discrimination, and are oftentimes unable to achieve inter-sectoral mobility (Altonji and Blank 1999; Bauder 2001; Constant and Massey 2005; Gordon 1995; Hayter and Barnes 1992; Hiebert 1999; Hudson 2007; McLafferty and Preston 1992; Reich, Gordon, and Edwards 1973). Vulnerable groups become trapped in the lower segment of the labor market because of mobility barriers, e.g., place of residence, poor work histories, and discrimination, which reduces inter-sectoral job transfers while occupational

LABOR MARKET SEGMENTATION

205

stratification increases (Bauder 2001; Gordon 1995). LMS theory posits that even after controlling for human capital factors, minority workers will earn less money than members of the majority and are less likely to be hired for employment (Becker 1971). In the U.S., Hudson (2007) found that Hispanics of both sexes, Black and Native American men, were all at an increased risk of being employed in lower sectors of the economy compared with White people. Furthermore, a substantial proportion of Black people were found to remain in the secondary sector over the course of their working life (Hudson 2007). In Canada’s three largest metropolitan areas, female groups and minorities, particularly immigrants, were all found to be overrepresented in poorly paid, vulnerable jobs (Hiebert 1999). In addition to discrimination practices, McLafferty and Preston (1992) found that LMS is deepened due to job mismatching. Job mismatching includes a spatial component where minorities, including women, have a lack of access to the location of primary sector jobs; therefore, they are forced into low-paying “bad jobs” in the secondary sector. Although most of the SLM theory literature has been developed and applied in a Western context, more recently, scholars have applied and modified SLM theory to better capture social and economic dynamics in developing countries. In his research on developing countries, Fields (2008) concluded that LMS exists if one or both of the following conditions exist: jobs for individuals with the same skill level differ in terms of wages of other characteristics and access to good jobs is limited in that people who want better jobs are unable to obtain them (Fields 2008). According to Fields’ (2008) definition, LMS in developing countries reflects similarities to developed countries. Similar to the cases in developed countries, findings from developing countries show minorities, females, and in many cases, migrants have limited access to good jobs and are disproportionately placed in lower segments of society (Beck, Horan, and Tolbert 1980; Coverdill 1988; Hiebert 1999; McLafferty and Preston 1992; Reid and Rubin 2003; Stolzenberg 1990). In Brazil, Telles (1993) analyses showed that Black people and mixed-race workers earn less than White people and that they are overly represented in informal sector employment (Telles 1993). In the case of Urumqi, conventional SLM theory is used to posit that Uyghur minorities, females, and migrant groups are more likely to be placed in lower tiers of the market and earn fewer wages relative to Han workers. The following section will turn toward an analysis of China and review the structural changes that have occurred that have led to local emerging labor markets in cities across the country.

206

GROWTH AND CHANGE, JUNE 2011

Occupational Structural Change and Emerging Labor Markets in China The structural change framework, advocated by Syrquin and Chenery (1989), examines how a nation’s output economy changes over time (Pannell and Schmidt 2006). The sectoral labor shift model, as applied to industrialized nations, is outlined by the following phases: agricultural productivity increases and the industrial sector attract surplus rural labor; industrial output reaches its peak, coupled with technological advancements, which reduces demand for industrial laborers; and a tertiary sector emerges to absorb excess industrial labor supply (Pannell and Schmidt 2006). With regard to the newly emerging structures of change within China, economic reforms have led to the following: decreased role of the state in job provision, specifically, the state sector no longer assigns jobs to graduating students, rather young adults are now encouraged to seek employment in the labor markets; the state sector has experienced substantial decline in terms of labor employment and productive output; non-state sectors, such as collective enterprises, private firms, and joint ventures, play a more predominate role in society (Li 1997); and returns to political capital have declined while returns to human capital have increased (Zang 2002). Adverse effects associated with structural changes, induced by neoliberal economic reforms, have varied in scale and scope. For example, the remaining intact population controls, such as the Hukou system, have not allowed for wage convergence between natives and migrants; rather a multitiered labor market has emerged (Fan 2002; Solinger 1999; Xu, Tan, and Wang 2006). Moreover, despite official unemployment rates listed as low as 3.1 percent in 2001, several prominent Chinese scholars have estimated unemployment rates to be 7.5 percent in 1997 or even as high as 10.4 percent in 1998 (Wu and Perloff 2004). Fan (2003) gave a detailed account of the nature of gendered LMS that has emerged from rural–urban migration. Her research showed that as a result of sociocultural traditions and social networks, male and female migrants are channeled into distinct sectors of the labor market. Young women are selected for factory work while men are recruited into hard labor, such as construction. Furthermore, middle-aged women are often discriminated against on the basis of their age due to employers’ association between youth and productivity (Fan 2003). There has been little effort by local governments to prevent labor market discrimination based on age and gender. If left unmitigated, the rural–urban migration flows will continue to reinforce sociocultural traditions and increase the gaps between labor market sectors.

LABOR MARKET SEGMENTATION

207

In addition, Zang (2002) examined how economic reforms have led to structural changes in the social stratification hierarchy. Permanent migrants are at the top, followed by urban nonmigrants, and finally, temporary migrants at the bottom (Fan 2002; Zang 2002). In addition to rural migrants, other marginal groups have emerged in Chinese cities: laid-off state-owned enterprise employees who are unable to find a new job; people who have never entered into the work-unit system, e.g., disabled and the widowed elderly; and those who retired before reforms brought about new forms of in-kind welfare guarantees (Wu 2004). Scholars have shown that rural–urban migration in China has acted to reinforce the gap among labor market segments (Fan 2002, 2003; Gordon 1995; Li 1997). However, in the case of Xinjiang, Han migrants are expected to further deepen LMS due to the strong presence of minorities and likelihood of discrimination in the labor market. Although Han Chinese play a vital role in Xinjiang’s economic development, they are perceived by many Uyghurs as monopolizing local natural resources and accumulating wealth by exploiting minorities (Bhattacharji 2009; Bovingdon 2002, 2004; Mackerras 2001; Webber 2009). Many studies have shown that the Han–Uyghur income gap is large and that Uyghurs’ quality of life has risen at a slower rate than that of Hans’ (Becquelin 2004; Bovingdon 2002; Gladney 2004; Mackerras 2001; Pannell and Schmidt 2006; Iredale et al. 2001; Toops 2004; Yee 2003). According to scholars, the emergence of local labor markets in Xinjiang contributes to growing wage inequality, deepening occupational segmentation, and pervasive flows of Han migrants into the region (Gladney 2004; Pannell and Schmidt 2006; Iredale et al. 2001). Hans are viewed as taking good jobs away from Uyghurs, at the expense of the latter’s social mobility and earnings capacity (Beller-Hann 2002; Pannell and Schmidt 2006; Iredale et al. 2001). Hannum and Xie (1998) showed that minorities are underrepresented in Xinjiang’s high-skill service sector, such as technical, administrative, and professional jobs. Pannell and Schmidt (2006) observed that Uyghurs are typically excluded from the industrial job market and the energy service sector. They found that Uyghur migrants from the less developed southern Xinjiang to the more developed north, including Urumqi, tend to be concentrated in low-paying service jobs, including petty vendors and informal sector employment, thus widening further the income gap between Han and Uyghur groups.

Survey Site and Research Design Urumqi, the capital of Xinjiang, is one of the few cities within the autonomous region that attracts a large number of migrants, who are diverse in terms of

208

GROWTH AND CHANGE, JUNE 2011

TABLE 1. POPULATION OF URUMQI BY MAJOR ETHNICITY AND DISTRICT, 2009.

Urumqi City (million) Tianshan District (%) Shayibak District (%) Xinshi District (%) Shui mogou District (%) Tou Tunhe District (%) Da Bancheng District (%) Midong District (%) Urumqi County (%)

Total population

Han

Uyghur

Hui

Kazak

Other

2.41 22.8

1.75 20.2

0.31 40.1

0.24 15.4

0.07 23.4

0.04 29.0

21.8

23.0

22.2

15.0

11.9

24.6

21.5 10.9

24.0 12.6

16.4 8.5

13.3 3.9

7.6 4.1

23.9 5.4

5.7

5.6

6.3

6.4

2.5

5.4

1.9

1.3

0.82

5.8

9.7

0.55

11.6 3.8

10.9 2.4

3.5 2.18

29.3 6.9

6.7 34.1

7.3 3.85

Source: Statistical Bureau of Xinjiang Uyghur Autonomous Region (SBX 2010). ethnicity, intra- and inter-provincial origin, and gender. In contrast to cities in eastern China where the ethnic composition is highly homogeneous, Urumqi’s population composition as well as the ethnic diversity of its in-migrants makes it an ideal city to examine the relationship between migration and LMS. Urumqi is an attractive destination city for both Han and Uyghur migrants partly because it has an established and vibrant labor market, a direct result of state policy. Recently, the “Open up the West” campaign enacted in 2001 boosted Urumqi’s economic development and brought about a boom in its service sector, drawing migrants from both within the province and other provinces (Pannell and Schmidt 2006). In 2008, Urumqi’s population reached 2.36 million (SBX 2010). The city is composed of seven districts. Of those, we selected four districts—Tianshan, Shayibake, Xinshi, and Shuimogou—for the survey because they have the largest and most developed commercial areas in Urumqi. Together, the four districts represent 77 percent of Urumqi’s total population, 87 percent of its Uyghur population, and 80 percent of its Han population (Table 1). Given the limitations of existing data sources (Bhalla and Qiu 2006), I conducted my own survey in Urumqi during July and August of 2008 with the aim of

LABOR MARKET SEGMENTATION

209

gathering pertinent information relating occupational outcomes to income, migration to ethnicity, education, income, and gender. The questionnaire includes questions on earnings (monthly income), demographic characteristics (gender, age, marital status, and education), migration characteristics (reason for migration, place of origin, length of migration, and household registration status), occupation (type of employment, type of firm, and work experience), and social capital (how did you find employment). With the aid of local informants at Xinjiang University, we selected survey sites that have a concentration of service activities. The sectoral shift model for Urumqi indicates that minorities and inter-provincial migrants are representative in the service economy but are underrepresented in other sectors of the economy. For this reason, my survey focuses on the service sector to ensure that both Uyghur and Han respondents are adequately represented in the sample. Because my survey team interviewed service-related establishments only, the results are not generalizable to other sectors of Urumqi’s economy. In total, we selected 30 sites: 13 in Tianshan, 10 in Shayibake, four in Xinshi, and three in Shuimogou. At each of the 30 sites, we selected every 10th store/ establishment to administer the questionnaire. The stores/establishments surveyed include a range of service activities, such as indoor/outdoor shopping centers, specialty goods and cultural markets, electronics markets, hospitals, restaurants, banks, offices, telecommunications, karaoke centers, and so forth. Because Han and Uyghur establishments tend to be segregated, it is important that the survey includes both predominantly Han sites and predominantly Uyghur sites. Of the 30 selected sites, 25 are predominantly Han and five are predominately Uyghur. Because Tianshan and Shayibake have the most notable concentrations of Uyghur commercial activities and account for 64.5 percent of the Uyghur population in Urumqi, we selected three Uyghur sites from Tianshan and two from Shayibake. We aimed at interviewing 15 to 30 establishments at each site. The number of establishments actually surveyed depends on the size and accessibility of the site but is always within the 15–30 range. At each establishment, we aimed at interviewing the store owner or the manager. If neither was available, we interviewed one of the employees. In the event of failure—for example, our request for interview was refused—we would drop the establishment from the sample and select the next establishment instead. In total, we collected 595 questionnaires. Of the 595 respondents surveyed, 72 percent of the respondents are Han, 21 percent are Uyghur, and 7 percent are other minority groups such as Hui, Kazak, and Tajik (Table 2). Twenty-nine percent of the respondents are natives, 47 percent are intra-provincial migrants, and 24 percent are inter-provincial migrants. I define

210

GROWTH AND CHANGE, JUNE 2011

TABLE 2. VARIABLE DEFINITIONS AND SAMPLE MEANS. Variable Ethnicity Han Uyghur Other Migrant status Native Intra-provincial Inter-provincial Female Age Mandarin fluent Education Primary Secondary High Technical College Employment type Employee Self-employed Employer Industry Retail Services Professional Formal Job tenure Poverty Wages

Definition

1 for Han, 0 for others. 1 for Uyghur, 0 for others 1 for Other, 0 for others 1 for native, 0 for not native 1 for intra-provincial migrant, 0 for not 1 for inter-provincial migrant, 0 for not 1 for female, 0 for male Years 1 for fluent, 0 for not

Mean

0.72 0.21 0.07 0.29 0.47 0.24 0.48 33 0.81

1 for primary school, 0 for others 1 for secondary, 0 for others 1 for high school, 0 for others 1 for technical, 0 for others 1 for college, 0 for others

0.16 0.28 0.26 0.16 0.10

1 if employee, 0 for others 1 if self-employed, 0 for others 1 if employer, 0 for others

0.48 0.38 0.14

1 if in retail/wholesale, 0 for other 1 if in service job, 0 if others 1 if professional or business, 0 if other 1 for work informal establishment, 0 for formal 1 for ten or more years of job experience, 0 for others 1 if earn less than half of average income, 0 if not Monthly wages (Yuan)

0.49 0.37 0.14 0.63 0.15 0.18 1,552

LABOR MARKET SEGMENTATION

211

natives as individuals born in Urumqi and migrants as those born outside of Urumqi. This definition of migrants is analogous to that of the floating population, a stock measure that largely ignores the timing of migration. For the purpose of this research, a stock measure is appropriate because I am concerned with whether an individual is a native of Urumqi or not, rather than when he or she arrived in the city. Also, I selected place of birth rather than Hukou—a key dimension of the floating population definition—again because I am interested in whether an individual is a native rather than whether he or she has obtained Urumqi hukou. Fifty-two percent of the respondents are male. The average age of the respondents is 33 years, 29 percent of the respondents are not conversationally fluent in Mandarin, and 52 percent of the respondents have at least a high school education whereas 48 percent have less than a high school degree, 4 percent of which have no schooling. In terms of respondents’ employment, we inquired about their employment type and industry type. Employment type is composed of three categories: employers, self-employed, and employees: 48 percent of the respondents are categorized as employees, 38 percent are self-employed, and 14 percent are employers. In the survey, no employers hire more than eight people. In other words, most establishments are small. Self-employed are those who own the business but do not employ any workers. Among employees in our survey, the most common jobs are retail, transportation, and clerical. In general, employers are in the highest socio-economic strata and employees make up the lowest socio-economic strata. The self-employed tend to be in the middle, but they are a diverse group, including very small businesses such as street vendors who are associated with low socio-economic status. I identified three industry types: retail, services, and professional. Almost half of the respondents are employed in retail, 37 percent in services, and 14 percent in professional. Retail refers to the selling of a variety of items, including snacks and beverages, electronics, home appliances, clothes, home furniture, and minority specialty foods or goods (carpets, knives, etc.). Services jobs include telecommunications providers, clerks, small electronics repairers, security guards, construction workers, restaurant servers, barbers, taxi and truck drivers, and various informal jobs including fruit vendors, shoe shiners, key makers and bike repairers, etc. Professional jobs include professors, doctors, engineers, computer technicians, clerical workers, and businesspersons. The last group of variables includes whether the respondent is employed in formal verse informal establishments (formal), the length of work experience (job tenure), a measure of inequality (poverty), and monthly income (wages). Sixtythree percent of the respondents are employed in formal establishments. Of the 37

212

GROWTH AND CHANGE, JUNE 2011

percent of the respondents working in informal establishments, most are retailers and vendors selling fruits, drinks, handicrafts, used electronics, or they provide services such as shoe cleaning and repair, bike repair, and key making. They conduct their businesses off of the main streets, usually on side streets or back alleys behind major shopping centers. My field observations show that Han and Uyghur migrants who work in the informal sector appear to have different niches. Han migrants tend to be engaged in a larger range of work, including selling specialty foods, fruits, appliances, clothing, jewelry, and shoe vendors, key makers, truck drivers, painters, construction, bike and electronics repair, and shoe shiners. Uyghur migrants in the informal sector tend to concentrate in selling foods or goods of higher value in the streets. Examples of the first category are flat bread and roast lamb meat, and examples of the second category include jewelry, electronics such as cell phones and computers, and counterfeit Renminbi bills. Those different niches suggest that the labor market is segmented by ethnicity. Poverty provides a sense of relative inequality that is captured in the data set. This measure is computed by obtaining the percentage of respondents who earn less than half of the average income for all the respondents. In this case, 18 percent of the respondents earn less than half the average income, 1,552 yuan. Because this research attempts to identify the relationship between migration and labor market outcomes, I will provide a brief overview of the migrant population collected in the data. The majority of both the Han and Uyghur respondents indicated that they first migrated to Urumqi to find work. Of the 417 migrants included in the survey, 324 are Han, 75 are Uyghur, and 18 are of other ethnicities, including Hui, Kazik, and Tajik. The vast majority of Han migrants— 81.5 percent—are inter-provincial migrants. Figure 1 shows their origin provinces, with Henan (15.2 percent), Sichuan (12.7 percent), Gansu (12.1 percent), and Shanxi (9.2 percent) as the leading origins. It appears that distance and economic development are among the factors of their migration. For example, Sichuan and Gansu are among the provinces closest to Xinjiang whereas Henan is among the poor provinces in China. There is also a well-established migration link between Henan and Xinjiang (Fan 2008:37) developed by the state during the era of mass mobilization and state-led orchestration of people from the east to populate Xinjiang (Howell and Fan 2011). Recently, the persistent flow of non-state sponsored Henan migrants to Xinjiang most likely can be explained due to the significant role of native place ties and guanxi, or social networks, which remain important determinants of migration (Ma and Xiang 1998) All the Uyghur migrants in our survey are from within Xinjiang, that is, they are intra-provincial migrants. Figure 2 highlights the counties of origin in

LABOR MARKET SEGMENTATION

213

Heilongjiang

Jilin Xinjiang Gansu Shanxi

Hebei Shandong

Henan Anhui Sichuan

Hubei Zhejiang Hunan Fujian

Yunnan

Legend Receiving Province

Provincial Origins (%) None

0

430

860 Kilometers

10

FIGURE 1. ORIGIN PROVINCES OF HAN INTER-PROVINCIAL MIGRANTS. Xinjiang for both Uyghur and Han intra-provincial migrants. About 75 percent of Uyghur migrants are from the southern counties of the province, with Kashgar (41.5 percent), Akesu (13.5 percent), Atushi (9.5 percent), and Hetian (7.8 percent) as the leading counties of origin. It is not surprising that more than 40 percent of the Uyghur migrants are from Kashgar, which boasts the largest economic development relative to the other predominately Uyghur counties in southern Xinjiang (Howell and Fan 2011). The Uyghur migrants captured in this survey are most likely the beneficiaries of the economic development policies aimed at advancing Uyghur communities; hence, they tend to possess higher levels of education and are more easily able to migrate to Urumqi compared with other Uyghur counties, which remain extremely impoverished. Of the Han intra-provincial migrants (n = 77), 88 percent are from northern and central Xinjiang, with Kuerla (21.5 percent), Yili (14 percent), Shihezi (14

214

GROWTH AND CHANGE, JUNE 2011

FIGURE 2. ORIGIN COUNTIES OF HAN AND UYGHUR INTRA-PROVINCIAL MIGRANTS. percent), and Tulufan (11.5 percent) as the leading origins. Tulufan, Kuerla, and Yili are the only three origin countries that sent at least 5 percent of both Han and Uyghur migrants, but the proportions from these three counties are much higher for Han migrants than for Uyghur migrants. The aforementioned spatial pattern clearly reflects the segregation of the southern Uyghur communities from Han settlements in the central and northern parts of Xinjiang. It shows also that Uyghur migrants, on average, had traveled a greater distance than Han intra-provincial migrants (most of whom are from nearby central counties) to reach Urumqi.

LABOR MARKET SEGMENTATION

215

Exposing Labor Market Segments in Urumqi There is not so much a debate on whether LMS exists rather the real problem occurs when trying to delineate segments. Because of truncation bias, which occurs as a distortion of results due to the omission of values that fall outside a given range, sectors cannot be a prior defined by wage earnings or occupational types (Leontaridi 1998). Other alternative classification systems must be implemented in order to reduce such biases as much as possible. An interesting statistical procedure has been the use of hierarchical clustering, which organizes occupational types into separate groups according to a priori selection of variables. While this procedure is not without its bias, interesting results have been found by other researchers (Anderson, Butler, and Sloan 1987; Drago 1995). A similar process is carried out in the subsequent sections to place respondents into corresponding segments captured in the data set. To obtain an ideal demarcation among segments in society, an extensive list of key segment characteristics would focus on various aspects of workers, jobs and firms (Drago 1995). However, such an extensive data set does not exist, especially in the case of Urumqi. Refer to Table 2 to see the list of selected variables used in the PCA that pertain to job and employment type and respondent characteristics. Table 3 provides predictions for each variable’s expected relationship with a labor market sector. A “+” denotes that the variable should be most highly represented within the particular segment, a “–” indicates a negative predicted relationship, and a “0” indicates either an uncertain relationship or a projected intermediate ranking (Drago 1995). Employees are expected to be most pronounced in the secondary sector and least pronounced in the PI sector; a ranking that is reversed for both self-employed and employer. The respondents employed in retail and in services are expected to be fewer in the primary sector and the majority employed in the secondary sector. This is because the respondents in these jobs earn relatively low levels of income and they sell common goods that are usually inexpensive, indicative of secondary sector jobs. Professional jobs are expected to be most pronounced as primary jobs and very few will be considered secondary sector jobs. The remaining job types are difficult to predict because some of the respondents may sell/repair high-priced or low-priced items, and therefore, some may have good jobs, representative of the primary sector, and others may have bad jobs, representative of the secondary sector. Formal, job tenure, education, and Mandarin fluency are all human capital variables that positively correspond to the respondents who are more likely to be placed in the primary sector. Because of the embedded discrimination in the labor market, ethnic, gender, and migrant status

216

GROWTH AND CHANGE, JUNE 2011

TABLE 3. LABOR MARKET SEGMENT OUTCOMES USING PREDICTOR VARIABLES. Labor market segment

Ethnicity Han Uyghur Other Migrant status Native Intra-provincial Inter-provincial Female Age Mandarin fluent Education Employment type Employee Self-employed Employer Industry Retail Services Professional Formal Job tenure Poverty

Primary-independent

Primary-subordinate

Secondary

+ + 0

+ + 0

0

+ 0 + +

+ + + + 0 + +

+ + + 0 -

+ +

0 0 0

+ -

+ + + -

0 0 + + + -

+ + +

are used to predict the location of respondents in the labor market. For example, Han, males, and natives are expected to be disproportionately placed in the primary sector, compared with their Uyghur, female, and migrant counterparts who are expected to be concentrated in secondary sector jobs. PCA is used to delineate trends within the data set by creating statistically independent principal components (O’Sullivan and Unwin 2003). In addition, PCA diminishes the problem of multi-collinearity (Rummel 1967). Carrying out PCA before running the cluster analysis ensures that the cluster outputs are not affected by collinearity among the variables. The PCA output produces a score

LABOR MARKET SEGMENTATION

217

TABLE 4. PRINCIPAL COMPONENTS ANALYSIS RESULTS.

Ethnicity Han Uyghur Other Migrant status Native Intra-provincial Inter-provincial Female Age Mandarin fluent Education Employment type Employee Self-employed Employer Industry Retail Services Professional Formal Job tenure Poverty Total variance explained (%)

Employment type (factor 1)

Ethnicity (factor 2)

+ +

+* -* -

+ + + +

+ + + +

+* -* -

+ +

+ + +* 13.5

+ + 12.0

“+” indicates the variable loads positive with the factor, “-” indicates the variable loads negatively with the factor, “*” indicates a variable loading with absolute value of at least 0.70.

vector for each principal component extracted. The score vectors are then used in the cluster analysis in order to classify respondents into the group or market segment they most readily correspond. Twenty-four variables, including industry, occupational, and human capital variables, are included in the PCA (Table 4). Varimax rotation is used in order to

218

GROWTH AND CHANGE, JUNE 2011

maximize the sum of the variances of the squared coefficients within each eigenvector, and the rotated axes remain orthogonal (Rummel 1967). In effect, this will enable a more straightforward interpretation of the component loadings. A loading of 0.7 or greater is the threshold used to indicate whether a variable is highly correlated to its corresponding factor. To start, five components were chosen from the PCA output and the corresponding rotated component loadings were analyzed. This process was repeated until there were only two components produced by the PCA. When extracting 5, 4, and then 3 components, each time only the first two components explained more than 5 percent of the variation in the data set. Running PCA with only two components offered the best outcomes in terms of high loadings and interpretable factors. Loadings on employee, self-employed, and formal all have absolute value of 0.7 or higher. As such, factor 1 is labeled “employment type.” The interpretation of this factor is that employees (low skill) are at the negative end of the factor, while nonemployees (high skill) are found at the positive end of the factor. For factor 2, only Han and Uyghur have loadings of absolute value 0.7 or higher. Factor 2 therefore is labeled as “ethnicity.” Han is found to be at the positive end of the factor, while Uyghur is found at the negative end.

Clustering Techniques To determine how jobs are grouped within each sector of the labor market, hierarchical and K-means clustering techniques are carried out (Anderson, Butler, and Sloan 1987). Hierarchical clustering is first used to classify a set of observations that are similar to each other while relatively different from other observation sets. The hierarchical clustering process starts with linking similar respondents into small clusters and then each small cluster is placed into larger groupings higher up in the hierarchy (O’Sullivan and Unwin 2003). The multiple score vectors produced from the PCA are saved as a new variable. Using Ward’s minimum variance, I ran hierarchical cluster analysis. Figure 3 shows the cluster tree output that delineates three distinct clusters. K-means cluster analysis is next used to assign the respondents into one of the three segments. Figure 4 maps the clusters into component space, there is an obvious delineation by ethnicity and employment type. The resulting clusters are identified with their corresponding labor market segment based on the cluster means and characteristics of the grouped respondents. Table 5 provides cluster means, which help identify the clusters. Cluster 1 contains 167 cases and has a mean of -1.19 for factor 1 (“employment type”), and a mean of 0.522 for factor 2 (“ethnicity”). Cluster 1 therefore represents Han respondents who are either self-employed or employers. Cluster 2 has a mean of 0.68 for factor 1 and

LABOR MARKET SEGMENTATION

219

Cluster Tree

0

50 100 Distances

150

200

FIGURE 3. TREE PRODUCED BY HIERARCHICAL CLUSTER ANALYSIS.

FIGURE 4. SCORES MAPPED IN FACTOR COMPONENT SPACE BY CLUSTER. Factor 1 is labeled “employment type”; factor 2 is labeled “ethnicity.” a mean of 0.53 for factor 2. Cluster 2 therefore represents Han respondents who are employees. Cluster 3 has a mean of 0.133 for factor 1 and a mean of -1.52 for factor 2. Cluster 3 therefore represents Uyghur respondents who are employees. Based on careful inspection of each cluster’s characteristics and mean values, Cluster 3 most closely represents the secondary sector. Clusters 1 and 2 represent

220

GROWTH AND CHANGE, JUNE 2011

TABLE 5. CLUSTER MEANS BY FACTOR.

Factor 1: employment type Factor 2: ethnicity

Cluster 1 (mean)

Cluster 2 (mean)

Cluster 3 (mean)

-1.19 0.522

0.68 0.53

0.133 -1.52

the primary segments; although it is still unclear which primary segment. In order to distinguish which primary segment clusters 1 and 2 represent, individual means for variables within each cluster are examined. Cluster 2 has the greatest concentration of high income compared with cluster 1 and has the lowest proportion of respondents in poverty. Furthermore, cluster 2 has the highest average education level relative to cluster 1. Hence, these comparisons indicate that cluster 2 represents the PI sector and cluster 1 represents the PS sector. To verify cluster 3’s “secondary segment” label, individual variable means within the cluster are also examined. As expected, cluster 3 has the highest proportion of respondents who are in poverty, the highest concentration of females and Uyghur respondents, and the lowest average education level attainment (Table 6).

Discriminant Analysis Canonical discriminant analysis (DA) is used for three purposes: 1) to test whether the labor market segments created using PCA and clustering analysis are significantly different from one another, canonical signifies that the dependent variable is not binary, regular DA is not applicable; 2) to see how well the model accurately predicts the observed categories of the dependent variable; and 3) to assess the relative importance of the independent variables in classifying the dependent variable (labor market sectors) (Klecka 1980). The dependent variable consists of three groups, PI, PS, and secondary; therefore, two orthogonal functions are listed in the output. The first function provides the most overall discrimination among the three groups and the second function provides the second most overall discrimination among groups (Table 7). We obtain a large Wilks’ lambda F-statistic at 10.25, which indicates that the DA model does a good job at discriminating among the three labor market segment groups. The canonical correlation on the first discriminant function is 0.658, meaning that 65.8 percent of the variation in the dependent variable is discriminated by the set of independents. The canonical correlation for the second discriminant function explains 34.9 percent of the remaining unexplained variation in the dependent variable.

LABOR MARKET SEGMENTATION

221

TABLE 6. SELECTED VARIABLE MEANS FOR EACH CLUSTER. Primary-independent (cluster 1)

Primary-subordinate (cluster 2)

Secondary (cluster 3)

0.25 0.51 0.07 0.83 0.41 0.94 2.6 0.06 0.30 0.13 0.20 1,477

0.25 0.33 0.20 0.47 0.48 0.98 2.9 0.05 0.52 0.16 0.16 1,514

0.60 0.34 0.59 0.06 0.60 0.78 2.4 0.03 0.82 0.18 0.37 1,189

Uyghur Native Intra-provincial Inter-provincial Female Mandarin fluent Education Professional Informal Job tenure Poverty Average wage

TABLE 7. DISCRIMINANT ANALYSIS- LAMBDA TEST STATISTIC.

Wilks’ lambda Canonical correlations

Value

Approx. F-value

p-value

0.489

10.24

0.000

Function 1 0.658

Function 2 0.349

This overall high percentage shows that much of the variance in the discriminant scores can be attributed to group differences. Table 8 shows the standardized discriminant function coefficients listed in order to interpret the unique contribution of each variable to the discriminant function. In the first discriminant function, native and inter-provincial migrant have the highest coefficients. This result indicates that migrant status contributes the highest discriminating effect on placing respondents into one of the three identified clusters or segments. In the second discriminant function, Uyghur has the highest coefficient, followed by native and inter-provincial migrant. This second discriminant function indicates that ethnicity does play a strong discriminating role in dictating in which sector an individual will find employment. The

222

GROWTH AND CHANGE, JUNE 2011

TABLE 8. STANDARDIZED CANONICAL DISCRIMINANT FUNCTIONS.

Ethnicity Han Uyghur Migrant Status Native Intra-provincial Inter-provincial Female Age Education Mandarin fluent Employment type Employee Self-employed Employer Industry Retail Services Professional Formal Job tenure Poverty a

Function 1

Function 2

0.106 -0.277

0.400 0.975a

-1.518a -0.603 -1.441a 0.027 -0.009 0.177 0.084

0.709 0.349 0.708 -0.083 -0.057 -0.057 -0.189

0.511 -0.142 0.783

0.474 -0.126 0.252

-0.286 -0.353 -0.176 0.049 -0.047 -0.014

-0.127 -0.024 -0.126 -0.063 0.107 0.132

Indicates largest discerning variable(s).

results from the second function also further reinforce the importance of migrant status in discriminating which sector of the economy respondents enter. The classification matrix is used to assess the performance of the DA (Table 9). The rows indicate the observed categories of the dependent variable and the columns indicate the predicted categories of the dependents. The classification matrix shows that the DA model does the best job at predicting occurrences in the secondary sector, a percentage of 81 percent. The PI sector shows a 66 percent correct prediction and the PS sector shows the least correct prediction of only 57 percent. It is not surprising to see PI and PS sectors have lower prediction scores relative to the secondary sector, as many jobs in these two sectors tend to share

LABOR MARKET SEGMENTATION

223

TABLE 9. CANONICAL DA CLASSIFICATION MATRIX.

PI PS Secondary Total

PI

PS

Secondary

% Correct

107 77 2 186

47 147 26 220

9 34 116 159

66 57 81 65

PI, primary-independent; PS, primary-subordinate.

similar job characteristics. Other scholars have recognized similar findings (Fichtenbaum et al. 1994; Hudson 2007).

Summary and Conclusions In support of LMS theory and its applicability to the Chinese context, my findings clearly delineate segments within the labor market, and as LMS theory predicts, a disproportionate number of Uyghur, females, and migrants are found to be located within the lower segment of the labor market, relative to their respective Han, males, and native counterparts. While the data are used to inform, this investigation is not without shortcomings; nevertheless, the outcomes from the PCA and cluster analysis correspond to similar patterns that have emerged in Western countries, such as the U.S. and Brazil, where segmentation exists. The implications of these findings for Urumqi and other multiethnic Chinese cities suggest that returns to human capital indicators, such as education and work experience, will not be equal across segments. A summary of the findings conclude that there is a strong link between migration and segmentation. Based on the DA, of all the variables expected to influence labor market outcomes, migrant status is found to have the strongest impact in placing respondents in the labor market. This outcome supports the perceptions noted by scholars that Uyghur believe Han migrants take away the good jobs and make life more difficult in terms of social mobility, discrimination, and earnings capacity (Mackerras 2001; Pannell and Schmidt 2006). Furthermore, the high loading on Uyghur in the second discriminant function suggests that ethnicity also plays an important role in channeling respondents into a particular segment of the labor market. In addition to meticulously describing the nature of Urumqi’s labor market, these findings extend beyond the empirical location-based studies and contribute to the larger body of LMS theory. The outcomes confirm that conventional LMS

224

GROWTH AND CHANGE, JUNE 2011

theory can be applied to the Chinese case, particularly within Chinese cities that are comprised of ethnic minorities. LMS theory is not only applicable to the multiethnic Chinese city, but the results shed new knowledge on the nature of minority–majority relations and social stratification that has emerged as a result of China’s economic reforms. Given the regional sensitivities of social relations in Xinjiang, greater monitoring of Urumqi’s service sector need to be placed on Han in-migration, particularly as China continues to move away from a centrally planned economy to market-oriented one and population migration is expected to increase. While past policy measures have in fact had the opposite effect and have relied on Han migration to develop the region, new measures need to be enacted to protect Uyghur urban workers and limit the proportion of Han migrant workers entering the workforce. Studying LMS in Xinjiang is especially timely and important particularly since unintended consequences of market reforms have already emerged in other places in China and the world, e.g., market discrimination, occupation stratification, and a rising gap in social and income inequality (Wu and Perloff 2004). The results from this research suggest that the labor market in Xinjiang is already experiencing negative consequences similar to other places in China and elsewhere. If labor market outcomes continue to polarize and are left unchecked, the region will likely continue to experience sporadic uprisings and rioting as seen during the 1990s or even the recent July 2009 riot, further deepening the animosities between Han and Uyghur locals. REFERENCES Altonji, J., and R. Blank. 1999. Race and gender in the labor market in Orely Ashenfelter and David Card (eds). Handbook of Labor Economics 3(3): 3143–3259. Anderson, K.H., Butler, J.S., and F.A. Sloan. 1987. Labor market segmentation: A cluster analysis of job groupings and barriers to entry. Southern Economic Journal 53(3): 571–590. doi:10.2307/ 1058755. Bauder, H. 2001. Culture in the labor market: Segmentation theory and perspectives of place. Progress in Human Geography 25(1): 37–52. Beck, E.M., Horan, P.M., and C.M. Tolbert, II. 1980. Industrial segmentation and labor market discrimination. Social Problems 28(2): 113–130. Becker, G.S. 1971. The economics of discrimination. Chicago, IL: University of Chicago Press. Becquelin, N. 2004. Staged development in Xinjiang. The China Quarterly 178(1): 358–378. Beller-Hann, I. 2002. Temperamental neighbors: Uyghur-Han relations in Xinjiang, Northwest China. In Imagined differences: Hatred and construction of identity, ed. Gunther Schlee, 57–81. London: Palgrave. Bhalla, A.S., and S. Qiu. 2006. Poverty and inequality among Chinese minorities. London and New York: Routledge.

LABOR MARKET SEGMENTATION

225

Bhattacharji, P. 2009. Uyghurs and China’s Xinjiang region. Washington, DC: Council on Foreign Relations. Bovingdon, G. 2002. The not-so-silent majority: Uyghur resistance to Han rule in Xinjiang. Modern China 28(1): 39–78. ———. 2004. Autonomy in Xinjiang: Han nationalist imperatives and Uyghur discontent. Washington DC: East-West Center Washington. Constant, A., and D. Massey. 2005. Labor market segmentation and the earnings of German guestworkers. Population Research and Policy Review 24(5): 489–512. Coverdill, J.E. 1988. The dual economy and sex differences in earnings. Social Forces 66(4): 970–993. Drago, R. 1995. Divide and conquer in Australia: A study of labor segmentation. Review of Radical Political Economics 27(1): 25–70. Fan, C. 2002. The elite, the natives, and the outsiders: Migration and labor market segmentation in urban China. Annals of the Association of American Geographers 92(1): 103–124. ———. 2003. Rural-urban migration and gender division of labor in transitional China. International Journal of Urban and Regional Research 27(1): 24–47. ———. 2008. China on the move: Migration, the state and the household. New York: Routledge. Fichtenbaum, R., Gyimah-Brempong, K., Olson, P., and G. Rudy Fichtenbaum. 1994. New evidence on the labor market segmentation hypothesis. Review of Social Economy 52(1): 20–39. Fields, G. 2008. Segmented labor market models in developing countries. In The Oxford handbook of philosophy of economics, ed. H. Kincaid, and D. Ross, 476–510. Oxford: University Press USA. Gladney, D.C. 2004. Dislocating China: Reflections on Muslims, minorities, and other subaltern subjects. Chicago, IL: University of Chicago Press. Gordon, I. 1995. Migration in a segmented labour market. Transactions of the Institute of British Geographers 20(2): 139–155. Hannum, E., and Y. Xie. 1998. Ethnic stratification in Northwest China: Occupational differences between Han Chinese and national minorities in Xinjiang, 1982–1990. Demography 35(3): 323– 333. Hayter, R., and T.J. Barnes. 1992. Labour market segmentation, flexibility, and recession: A British Columbian case study. Environment and Planning C 10: 333–353. Hiebert, D. 1999. Local geographies of labor market segmentation: Montréal, Toronto, and Vancouver, 1991. Economic Geography 75(4): 339–369. Howell, A., and C.C. Fan. 2011. Migration and inequality in Xinjiang: A survey of Han and Uyghur migrants in Urumqi. Eurasian Geography and Economics 52(1): 119–139. Hudson, K. 2007. The new labor market segmentation: Labor market dualism in the new economy. Social Science Research 36(1): 286–312. Iredale, R., Bilik, N., Su, W., Guo, F., and C. Hoy. 2001. Contemporary minority migration, education, and ethnicity in China. Cheltenham, UK: Edward Elgar Publishing. Klecka, W.R. 1980. Discriminant analysis. London: SAGE. Leontaridi, M. 1998. Segmented labour markets: Theory and evidence. Journal of Economic Surveys 12(1): 103–109. Li, S.M. 1997. Population migration, regional economic growth and income determination: A comparative study of Dongguan and Meizhou, China. Urban Studies 34(7): 999–1026. Lin, G. 2002. Changing discourses in China geography: A narrative evaluation. Environment and Planning A 34(2): 1809–1831.

226

GROWTH AND CHANGE, JUNE 2011

Ma, L., and B. Xiang. 1998. Native place, migration and the emergence of peasant enclaves in Beijing. The China Quarterly 155: 546–581. Mackerras, C. 2001. Xinjiang at the turn of the century: The causes of separatism. Central Asian Survey 20(3): 289–303. McLafferty, S., and V. Preston. 1992. Spatial mismatch and labor market segmentation for AfricanAmerican and Latina women. Economic Geography 68(4): 406–431. Murphy, A., and J. O’Loughlin. 2009. New horizons for regional geography. Eurasian Geogrpahy and Economics 50(3): 241–251. O’Sullivan, D., and D.J. Unwin. 2003. Geographic information analysis. Hoboken, NJ: John Wiley. Pannell, C.W., and P. Schmidt. 2006. Structural change and regional disparities in Xinjiang, China. Eurasian Geography and Economics 47(3): 329–352. Peck, J. 1996. Work-place: The social regulation of labor markets. New York, NY: The Guilford Press. Reich, M., Gordon, D.M., and R.C. Edwards. 1973. A theory of labor market segmentation. The American Economic Review 63(2): 359–365. Reid, L.W., and B.A. Rubin. 2003. Integrating economic dualism and labor market segmentation: The effects of race, gender, and structural location on earnings, 1974–2000. Sociological Quarterly 44(3): 405–432. Rummel, R.J. 1967. Understanding factor analysis. Journal of Conflict Resolution 11(4): 444–480. SBX (Statistical Bureau of Xinjiang Uyghur Autonomous Region) 2010. Xinjiang Tongji Nianjian (Xinjiang statistical yearbook 2010). Beijing, China: China Statistics Press. Solinger, D. 1999. Contesting citizenship in urban China: Peasant migrants, the State, and the logic of the market. Berkeley, CA: University of California Press. Stolzenberg, R.M. 1990. Ethnicity, geography, and occupational achievement of Hispanic men in the United States. American Sociological Review 55(1): 143–154. Syrquin, M., and H. Chenery. 1989. Patterns of Development, 1950–1983. Washington, DC: The World Bank. Telles, E.E. 1993. Urban labor market segmentation and income in Brazil. Economic Development and Cultural Change 41(2): 231–249. Toops, S. 2004. The demography of Xinjiang. In Xinjiang: China’s Muslim Borderland, ed. S. Frederick Starr, 241–263. Armonk, NY: M.E. Sharpe. Webber, M. 2009. The places of primitive accumulation in rural China. Economic Geography 84(4): 395–421. Wu, F. 2004. Urban poverty and marginalization under market transition: the case of Chinese cities. International Journal of Urban and Regional Studies 28(2): 401–423. Wu, X. and J. Perloff. 2004. China’s Income Distribution Over Time: Reasons for Rising Inequality. SSRN Working paper. Xu, W., Tan, K., and G. Wang. 2006. Segmented local labor markets in post-reform China: Gender earnings inequality in the case of two towns in Zhejiang province. Environment and Planning A 38(1): 85–109. Yee, H.S. 2003. Ethnic relations in Xinjiang: A survey of Uygur–Han relations in Urumqi. Journal of Contemporary China 12(36): 431–452. Zang, X. 2002. Labor market segmentation and income inequality in urban China. Sociological Quarterly 43(1): 27–44.