The 2009 ethnic unrest in Ürümchi and the large number of casualties involved call for an urgent study of ethnic inequality in the Xinjiang Uyghur Autonomous Region of the People's Republic of China (PRC). Many scholars and commentators have cited ethnic disparities as a main cause of intergroup tensions and conflicts in Xinjiang.Footnote ¹ Yet it is not clear who among Uyghurs have suffered most from income inequality by ethnicity in the region. Is it old people, illiterates, women or some other Uyghur group? Those who are most oppressed economically are more likely than other Uyghurs to challenge the status quo. This research goes one step further than making a distinction between Uyghurs and Han Chinese, focusing on the effect of age on the wage gap between Han Chinese and Uyghurs. It asks: do young Uyghurs do better than old Uyghurs in the labour market? Is the cost of being Uyghurs higher for the latter than the former? It addresses these questions with data from a survey of 2,947 people conducted in Ürümchi in 2005.

The cost of being Uyghurs refers to the extent of economic inequality in the earnings of Han Chinese and Uyghurs. Scholars have regarded the earnings gap between whites and African-Americans (not including educational and occupational differences) as the cost of being African-Americans in the US. The cost of ethnic membership in earnings provides a key indicator of one aspect of socioeconomic discrimination.Footnote ² This article follows this approach, comparing the earnings of three Han age groups with those of three Uyghur age groups in Ürümchi. The result shows interesting patterns of Han–Uyhgur inequality in income. It is necessary to point out that large groups of rural Uyghurs fare even worse than the urban Uyghurs I examine in this article.

The article first provides some background information about Uyghurs and rising ethnic income inequalities in Xinjiang since the 1990s. Next, it puts forward three hypotheses on the effect of age on Han–Uyghur earnings differentials. It then discusses the data and measures and conducts empirical analyses. Although young Uyghurs are better educated and earn more than old Uyghurs, they are more likely than old Uyghurs to suffer from being Uyghurs in Ürümchi. The article offers a plausible account to explain this important finding and examines its policy implications for the reduction of ethnic inequality in Xinjiang.

Historical Context

Ürümchi is the capital of Xinjiang, which is located in north-west China and occupies one-sixth of China's territory with 5,600 kilometres of international borders with Pakistan, India, Afghanistan, Kazakhstan, Kyrgyzstan, Tajikistan, Mongolia and Russia. Uyghurs are one of the 56 ethnic groups in China, which is dominated both politically and economically by the ethnic majority Han Chinese. Uyghurs are a Turkic people and the vast majority of them are Sunni Muslims. The 2000 census found nearly 8.35 million Uyghurs in Xinjiang.

Historically, Uyghurs had lived in north-western Mongolia before they migrated en masse to Xinjiang after the demise of the Uyghur Empire in 840. They reportedly practised Manichaeanism, Nestorian Christianity and shamanism before 932. Some Uyghurs later became Buddhist; others were converted to Islam before the Mongol conquest of the region known as Xinjiang today around 1200. The massive Uyghur conversion to Islam started after the Mongol conquest and was not completed until the mid-1400s; some scholars claim that the conversion was concluded in the 1600s. After 1759, part of the region, the Zhungarian Basin, was called Xinjiang, and “the name was later applied to the whole region.”Footnote ³ Xinjiang became a province of the Qing Empire in 1884. After the establishment of the Republic of China in 1911, it was ruled successively by Han warlords, before the Republic managed to place it under its direct control in 1944. The Chinese Communist Party (CCP) took over Xinjiang in 1949 and renamed it the Xinjiang Uyghur Autonomous Region on 1 October 1955.Footnote ⁴

Xinjiang experienced rapid social transformation after 1949. It is well known that there were anti-Chinese riots and rebellions for political and religious reasons in the region in the 1950s. Some 60,000 Uyghur and Kazak refugees fled northern Xinjiang into the Soviet Union to escape the Great Chinese Famine and political purges in 1962. Nevertheless, during the course of fieldwork in Ürümchi, I heard repeatedly that before the Cultural Revolution of 1966–76, many poor Uyghur regarded the CCP as their saviour. The Chinese government demolished Islamic establishments, carried out land reforms in rural Xinjiang, and raised a large group of Uyghur intellectuals and professionals in urban areas. The CCP also set up factories, schools and other institutions in the region and recruited many Uyghurs into the state workforce. The majority of Uyghurs benefited from upward mobility and greatly improved their living standards.Footnote ⁵ Of course, some Uyghur intellectuals complained about Han chauvinism, and they paid a high price for making such complaints during the Hundred Flowers Campaign. Others were victimized by the CCP's minority policies. During the Cultural Revolution, many Uyghur intellectuals and officials were persecuted despite their commitment to the PRC and enthusiastic participation in socialist construction in the pre-1966 era.Footnote ⁶

The socio-political situation in Xinjiang changed markedly after 1976. The Chinese government allowed a relatively tolerant environment for ethnic and religious expression in Xinjiang in the 1980s, and emphasized the importance of unity between Han Chinese and minority groups.Footnote ⁷ Both the PRC's Constitution and the Law on Regional Ethnic Autonomy stipulate that the government shall promote the common prosperity of all ethnic groups. The government carried out equal opportunity programmes in education and employment, which played a key role in enlarging and diversifying the Uyghur middle class in Xinjiang.Footnote ⁸ In addition, in 2000 it declared a national strategy to develop the Great Western Region (xibu da kaifa 西部大开发) and has carried out a batch of key construction projects in Xinjiang, including petroleum exploration, the West–East Natural Gas Transmission Project, the West–East Power Transmission Project and the “Project for Comprehensive Improvement of the Environment of the Tarim Basin.”Footnote ⁹ The Chinese government is hoping to develop the local economy to raise living standards among Uyghurs in Xinjiang.

The Chinese government has also promoted bilingual education in Xinjiang as a way to help Uyghurs compete in the labour market. However this policy has not been entirely successful since Uygurs have seen it as the promotion of monolingual Chinese education.Footnote ¹⁰ Furthermore, the government has supported large-scale Han migration into Xinjiang as a key measure to promote economic growth. Indeed, the Uyghur population in Xinjiang declined from over 75 per cent in 1949 to less than 46 per cent in 2004, whereas the Han population grew from under 10 per cent in 1949 to nearly 40 per cent in 2004.Footnote ¹¹ And Han migration has not reduced Uyghur–Han inequality in income. On the contrary, because of the intergroup differences in the first language spoken, schooling, capital, and access to bank loans and markets in China proper, Han Chinese are more likely than Uyghurs to take advantage of the development of the Great Western Region.Footnote ¹² A Uyghur woman cadre (informant 1–2) asserted that Han Chinese made a lot of money in Xinjiang: “Each year Han workers from Sichuan province send 40 billion yuan to their families.” A Uyghur housewife (informant 1–4) said that a Han couple slept on the streets in her neighbourhood when they arrived in Ürümchi. They moved to a rental property later. “Now they are the owner of a good house, which is better than mine. I cannot believe this.” Another Uyghur housewife (informant 1–10) told a similar story: “My husband's Chinese friend rented a room when he came here. He did odd jobs and saved some money with which he opened a grain shop. He has since made a lot of money and bought a nice house. He has improved his living standard, whereas we are where we were many years ago.” A Uyghur shop owner (informant 2–2) was surprised to find out that many Han workers became rich two or three years after they came to Xinjiang. Another Uyghur man (informant 1–9) claimed that Uyghurs were honest, whereas Han Chinese were crafty and would do anything to make money.

In addition, the Chinese government reduced the workforce in state enterprises to promote efficiency and profitability in the 1990s and early 2000s. Most managers in the state sector in Xinjiang are Han Chinese. They choose to sack minority workers first, then, if a state firm is profitable and needs to hire new workers, they prefer Han applicants over Uyghur ones because of Han chauvinism against Uyghur culture.Footnote ¹³ A Uyghur housewife (informant 6–7) recalled that when her son had a job interview, the Han boss asked him to go away since he did not want to hire minority workers. Not surprisingly, it is found that, overall, “Uyghurs are poorer than the Han.”Footnote ¹⁴ It is also found that “most of the modern sector in a very advanced city such as Ürümchi appears to be dominated by Han, not minorities.”Footnote ¹⁵ Thus “urban dissatisfaction stems from the fact that Uyghurs now have something to compare themselves with. It is socio-economic inequalities, there, which lies at the root of a rapidly strengthening Uyghur national identity.”Footnote ¹⁶

Age Groups and Han–Uyghur Income Inequality

While there is a consensus on intergroup income inequality in Xinjiang, it is not clear how well different Uyghur age groups have fared in the post-1978 era. Existing studies have found that young adults in the West tend to have higher rates of unemployment and lower wages than the general workforce. Poverty rates among young adults have increased in the past decade.Footnote ¹⁷ Yet in Xinjiang, young Uyghur workers may not be a vulnerable group. Post-1978 market reforms in China have been described as a transition from a centrally controlled economy to a market economy, which has raised the importance of human capital as a result of the emphasis on performance and efficiency. Uyghur youth are more capable of adapting in the new circumstances in the labour market than old Uyghurs. Ongoing market reforms have coincided with a global information technology revolution and affected the way the labour market values the different forms of human capital. Uyghur youth have received more schooling and a more modern education so it is relatively easy for them to meet the challenges of a market economy. Old Uyghurs have fewer opportunities in the labour market since their human capital, accumulated in the Maoist socialist planned economy, has lost its value in the new market circumstances. This leads to Hypothesis 1: the cost of being Uyghurs is higher for old Uyghurs than young Uyghurs, controlling for key background characteristics.

Hypothesis 1 cannot be taken for granted, however. This is partly because some scholars have argued that labour market discrimination is responsible for ethnic disparity in Xinjiang.Footnote ¹⁸ According to Amnesty International: “Many Uighurs complain that racial abuse and discrimination against ethnic minorities is common.”Footnote ¹⁹ Some Uyghurs maintain that attitudes of racial and cultural superiority became commonplace after the 1990s,Footnote ²⁰ while others “complained that the government has forsaken the preferential policy for national minorities.”Footnote ²¹ Still others have claimed that they encountered prejudice when seeking access to bank loans and administrative authorizations from government officials.Footnote ²² Although many urban Uyghur youths have a university degree, their human capital is not automatically converted into an earning capacity. It is difficult for them to find good jobs.Footnote ²³ Young Uyghurs may not do better than old Uyghurs in the labour market since Han prejudice does not discriminate between different Uyghur age groups. This leads to Hypothesis 2: there is little variation in the cost of being Uyghurs between old and young Uyghurs, holding everything constant.

Hypothesis 2 cannot be taken for granted either, however. This is partly because old Uyghur workers entered the labour force in the Mao era when equal opportunity programmes were implemented. They and their Han counterparts are not well educated. These factors suggest minor ethnic income inequality between these two groups: they accumulated similar work experience before market reforms, and the earnings gap between them may not have enlarged greatly after the reforms. For both old Han workers and old Uyghur workers, advanced age and the declining opportunities for them to retrain for better-paid jobs means that they are likely to stay where they are, and this immobility suggests that growing discrimination against ethnic minorities will not greatly affect their earnings. In comparison, Uyghur youth entered the labour market after 1980 and are more likely than old Uyghur workers to encounter Han discrimination, whereas Han youth are more likely than old Han Chinese to benefit from growing discrimination against ethnic minorities. There is likely to be major income inequality by ethnicity between young Han Chinese and young Uyghurs but not between old Han Chinese and old Uyghurs. Hence Hypothesis 3: the cost of being Uyghurs is higher for young Uyghurs than old Uyghurs, other things being equal.

Data, Variables and Measures

The above three hypotheses are examined with data from the survey conducted in Ürümchi in 2005. Ürümchi is located in northern Xinjiang. Although Uyghurs are the largest ethnic group in Xinjiang, they are a minority group in Ürümchi. The official statistics show that 76 per cent of Ürümchi residents are Han Chinese and 13 per cent are Uyghurs.Footnote ²⁴ Uyghur residents are not distributed evenly among the 52 urban districts in the city. The 2000 census showed that Tuanjielu had the highest percentage of Uyghur households (47.5 per cent) followed by Yan'anlu with 38.1 per cent. Xinhuananlu was ranked eighth with 21.1 per cent of Uyghur households. Given such a population composition and distribution of Uyghur residents in Ürümchi, a randomly selected sample would not be suitable.

A disproportionate sampling method was therefore used for the 2005 survey. The local Han collaborators first chose eight neighbourhoods with the highest percentages of Uyghur households in Ürümchi (the Uyghur clusters). They then randomly selected eight neighbourhoods from the remaining 44 (the Han clusters). They used a disproportionate sampling method to select 2,437 households from the 16 clusters. Households in the Uyghur clusters were given a disproportionately better chance of selection than those in the Han clusters to yield roughly similar numbers of Uyghur and Han respondents in the final sample. Survey data are weighted to adjust for the sampling bias for data analyses in the next section.

In all, 1,600 interviews were conducted, representing a completion rate of nearly 66 per cent (69.6 per cent for the Uyghurs and 62.1 per cent for the Han). There was no access to the other 34 per cent of the sampled households because of refusals, absence, poor access to gated residential buildings or other reasons. Of the respondents, 799 were Uyghur households and 801 were Han households. Information on both the household heads and their spouses was solicited, resulting in a sample of 2,947 cases (1,464 Han, 1,483 Uyghur) for data analysis.

The questionnaire for the 2005 Ürümchi survey was first written in English and then translated into a Chinese version, which was translated back into an English version by different translators to ensure consistency between the two versions. Next, the local Chinese collaborators hired professional translators to produce a Uyghur version, which was then compared with the Chinese version to ensure consistency. Interviewers were trained by experienced survey researchers. Some Han interviewers spoke Uyghur and many Uyghur interviewers spoke Mandarin, the language of the Han Chinese. In the 2005 survey, Uyghur interviewers were assigned to interview Uyghur respondents only and Han interviewers were assigned to interview Han respondents only. This was designed to enhance the degree of cooperation from the interviewee and effective communication between the interviewer and the interviewee. All respondents were asked to provide information on demographics, income and education. The survey started in March 2005 and ended in August 2005.

In addition, I commissioned the local collaborators to conduct 230 in-depth interviews with both Han and Uyghur informants between 2005 and 2008. I also conducted fieldwork in Ürümchi during this time. I relied on my bilingual Han and Uyghur contacts to interview local Uyghurs, but I also asked Mandarin-speaking Uyghur taxi drivers, pedestrians, hawkers and shop owners some short questions. Occasionally, these brief social encounters extended into lengthy discussions of Uyghur behaviour and culture. Interestingly, some Uyghurs saw me as a lone Han traveller and sometimes struck up a conversation with me. Finally, I conducted field observations in some mosques, buses, wet markets and shopping malls. This fieldwork greatly enhanced my understanding of intergroup inequality in Ürümchi.

I divided the sample (N = 2,947) into three subsamples to observe the changes in the cost of being Uyghurs in Ürümchi: Age Group 1 (N = 893) consists of the respondents who were 18 or older before 1950; Age Group 2 (N = 1,013) is the respondents who were 18 or older before 1979; and Age Group 3 (N = 1,041) is the respondents who were 18 or older in 2004. These divisions are based on three milestones in the history of the PRC. People in Age Group 1 were educated and entered the labour force before the CCP came to power in China and Xinjiang. They grew up during the chaotic years of the pre-1949 era, valued stability, and were at the late stages of their careers when economic reforms started in 1978. Much of their working lives was protected by equal opportunities programmes, and post-1990 Han discrimination has not affected their jobs and income very much. People in Age Group 2 were schooled and spent some part of their careers in the Maoist socialist planned economy. They enjoyed the government protection of ethnic minorities until the 1990s, when discrimination against ethnic minorities in the labour market increased and post-1978 reforms started to reduce the state labour force and intensify labour market competition. People in Age Group 3 were educated and found jobs during market reforms. Only those who are fortunate enough to have entered the state sector are protected by affirmative action. Others are at the mercy of market competition and growing discrimination against ethnic minorities in Xinjiang.Footnote ²⁵

In the analyses conducted below, a respondent's earnings are used as the dependent variable, which are operationally defined as his or her total monthly income (wage, bonuses and all other income) in 2004, the year prior to the data collection in 2005. Because of its wide range, income is log-transformed to establish normality and constant error variances to obtain better linearity in the regression function. Earnings have been widely used in research on intergroup inequality in income.Footnote ²⁶

The independent variables include control variables (age, gender, marital status, urban status), education, occupational attainment (state employment, skilled worker, professional), and Uyghur. Age is a continuous variable. Gender is a dummy variable with men coded as 1 and women as 0. Marital status is a dummy variable with married respondents coded as 1 and others as 0. Urban status is also a dummy variable with urban residents coded as 1 and rural migrants coded as 0. Educational attainment is measured by an ordinal variable with seven levels: (1) illiterate/semi-illiterate, (2) primary school, (3) junior high school, (4) senior high school, (5) vocational school, (6) community college and (7) university. Occupational attainment is measured by state employment (= 1), skilled worker (= 1) and professionals (= 1).Footnote ²⁷ Skilled worker includes such jobs as truck driver. Professional includes accountants, doctors and engineers. Finally, Uyghur is a dummy variable with Uyghurs coded as 1 and Han Chinese coded as 0.

Findings

Table 1 shows the key background characteristics of the Uyghur and Han respondents in the three age groups in Ürümchi. For Age Group 1, the mean age is 61.5 for the Uyghur respondents and 62.6 for the Han respondents. The corresponding figures for Age Groups 2 and 3 are 48.0 and 48.3, and 37.6 and 37.5, respectively. Also, for Age Group 1, 63.1 per cent of the Han respondents and 80.3 per cent of the Uyghur respondents are urban residents; for Age Group 2, the figures are 84.3 per cent and 85.7 per cent respectively; and for Age Group 3, 86.9 per cent and 81.2 per cent respectively. Table 1 also shows data on gender and marital status for the three age groups.

Table 1: Key Background Characteristics of the Ürümchi Respondents by Ethnicity and Age Groups

Note:

Figures in parentheses are the numbers of cases.

Table 1 also shows that for Age Group 1, the mean score for education among the Han respondents is 2.87, compared with 3.17 among the Uyghur respondents. The Han–Uyghur gap in schooling is –.3. For Age Group 2, the mean score for education among the Han respondents is 4.12, compared with 3.50 among the Uyghur respondents, so that the Han–Uyghur gap in schooling is .62. The corresponding figures for Age Group 3 are 4.65 and 3.92 respectively, giving a Han–Uyghur gap in schooling of .73. Uyghurs have improved their levels of schooling, yet their improvement is slower than that of Han Chinese. Thus, for Age Groups 2 and 3, there are large gaps in educational attainment between Han Chinese and Uyghurs.

In addition, Table 1 shows that for Age Group 1, 81.5 per cent of the Han respondents and 60.7 per cent of the Uyghur respondents are state workers. The Han–Uyghur difference in state employment is 20.8 per cent. The corresponding figures for Age Group 2 are 77.4 per cent and 55.9 per cent respectively, giving a Han–Uyghur difference of 17.5 per cent. For Age Group 3 the figures are 61.6 per cent and 51.3 per cent respectively, giving a Han–Uyghur difference of 10.3 per cent. Generally, the proportion of state workers is declining for each succeeding age group, which is consistent with the post-1978 market reforms that have reduced the state workforce. Nevertheless, the Han respondents are more likely than the Uyghur respondents to work in the state sector. Table 1 also shows data on the proportions of skilled workers and professionals among the Han and Uyghur respondents in each of the age groups. It is clear that the Han respondents are more likely than the Uyghur respondents to become skilled workers and professionals.

Finally, Table 1 shows that for Age Group 1, the monthly income for Han Chinese is 702.66 yuan, compared with 641.83 yuan among Uyghurs (the Han–Uyghur gap in income is 60.83 yuan per month). It is likely that many people in this age group have retired, and for them pensions are a major source of income. The corresponding figures for Age Group 2 are 1,178.90 yuan and 791.00 yuan respectively (the Han–Uyghur gap is 387.90 yuan per month). For Age Group 3 the figures are 1,142.03 yuan and 731.24 yuan respectively (the Han–Uyghur gap is 410.69 yuan per month). It is clear that young Uyghur workers earn more than old Uyghur workers. However, for Age Group 1, Uyghurs earn 91 per cent of what Han Chinese earn per month; for Age Group 2 they earn 67.1 per cent; and for Age Group 3, 64.0 per cent. In other words, although young Uyghur workers earn more than old Uyghur workers, they are more likely than old Uyghur workers to be subject to income inequality by ethnicity. These findings tentatively confirm Hypothesis 3 and do not support Hypotheses 1 and 2.

Table 2 shows the correlations between the dependent variable and the independent variables for three subsamples. The data are weighted to correct the sampling bias. As a result, the number of people is 895 in Age Group 1, 864 in Age Group 2 and 969 in Age Group 3. It can be seen that for all three age groups, income is positively related to men, urban status, education and state worker, and the relationships are statistically significant at .001 level. It can also be seen that Uyghur is not associated with income for Age Group 1, that is, there is no significant income inequality by ethnicity for this age group. For Age Group 2, however, Uyghur is correlated with income and the correlation is negative. There is also a negative correlation between Uyghur and income for Age Group 3. The negative Pearson correlation coefficient for Age Group 3 is greater than that for Age Group 2 (−.169 compared with −.150), consistent with the findings reported in Table 1. Again, these findings tentatively support Hypothesis 3 and do not support Hypotheses 1 and 2.

Table 2: Correlations

Notes:

*P < .05 **P < .01 ***P < .001. Figures in parentheses are Sig (2-tailed).

The above findings are derived from bivariate analyses. It is not clear if they are the outcomes of spurious relationships. Accordingly, I perform three OLS regression analyses to verify the relationship between Uyghur ethnicity and income inequality. Again, the data are weighted to correct the sampling bias before empirical analyses. It is necessary to mention that multicollinearity is not an issue as all independent variables in the multivariate analyses pass the tolerance tests (the tolerance levels range from .518 to .986).

Model 1 of Table 3 shows that for Age Group 1, men earn a lot more than women. Education is positively related to high income. So are state workers and professionals. There are close relationships between income and these independent variables. These findings make good sense and are consistent with existing research on income inequality in other parts of China.Footnote ²⁸ Models 3 and 5 of Table 3 show similar results regarding the relationship between income and the measures of socioeconomic status, holding other independent variables constant.

Table 3: Effects of Uyghur Ethnicity on Earnings

Notes:

*P < .05 **P < .01 ***P < .001. Figures in parentheses are t-ratios.

Model 1 of Table 3 also shows that the OLS regression coefficient for Uyghur is not statistically related to the dependent variable, that is, there is no statistically significant income inequality by ethnicity for Age Group 1. In other words, Uyghurs in this age group do not pay a penalty for being Uyghurs. However, Model 3 of Table 3 shows that for Age Group 2, the OLS regression coefficient for Uyghur is related to income and the relationship is negative and statistically significant at .05 level. In other words, Uyghurs in this age group pay a penalty for being Uyghurs. Model 5 of Table 3 shows that for Age Group 3, the OLS regression coefficient for Uyghur is related to earnings and the relationship is negative and statistically significant at .001 level. Taking the exponent of the coefficient for Uyghur in Model 3, it can be seen that the Uyghur respondents earn 35.6 per cent less than the Han respondents in Age Group 2. Taking the exponent of the coefficient for Uyghur in Model 5, it can be seen that the Uyghur respondents earn 51.2 per cent less than the Han respondents in Age Group 3. Clearly, the younger Uyghurs are, the higher the cost they have to pay for being a Uyghur. These findings confirm Hypothesis 3 and do not support Hypotheses 1 and 2.

Why do Uyghurs in Age Groups 2 and 3 pay a high cost for being Uyghurs in Ürümchi whereas their counterparts in Age Group 1 do not? This is puzzling since, as Table 1 shows, they are better educated than old Uyghurs in Age Group 1, and there are not sizeable differences in the probabilities of becoming skilled workers and professionals between the Uyghurs in Age Group 1 and those in Age Groups 2 and 3. Table 1 also shows that Uyghurs in Age Groups 2 and 3 earn more than the Uyghurs in Age Group 1. Since overall the Uyghurs in Age Groups 2 and 3 have more human capital than those in Age Group 1, the wage gap between Han Chinese and the Uyghurs in Age Groups 2 and 3 should not be greater than that in Age Group 1. Yet the above analysis shows that young Uyghurs are more likely than old Uyghurs to suffer from income inequality by ethnicity.

How can this puzzle be accounted for? A plausible explanation is the different returns to education and occupation for Han Chinese and Uyghurs in different age groups: in other words, the returns may be greater for Han Chinese than for Uyghurs in Age Groups 2 and 3, but similar for both groups of workers in Age Group 1. This in turn may explain the differences in the cost of being Uyghurs between Age Group 1 and Age Groups 2 and 3. The different returns to education and occupation for Han Chinese and Uyghurs in different age groups can be attributed to equal opportunities programmes before 1978 and rising discrimination in the post-1978 labour market. Accordingly, I conducted three OLS regression analyses to examine this possibility. I created four interaction variables for data analyses: Uyghur*education, Uyghur*state employment, Uyghur*skilled worker and Uyghur*professional.

Model 2 of Table 3 shows no differences in returns to education and occupational attainment between Han Chinese and Uyghurs in Age Group 1, confirming the above speculation. This finding may partly account for why, in Model 1 of Table 3, there is no income inequality by ethnicity for Age Group 1. People in Age Group 1 spent more working time in the Maoist socialist planned system with active affirmative action than those in Groups 2 and 3. Han workers and Uyghur workers received similar returns to education and occupational attainment when ethnic discrimination was kept to a minimum.

However, contrary to the above speculation that the returns may be greater for Han Chinese than for Uyghurs in Age Groups 2 and 3, Model 4 of Table 3 shows similar returns to education, skilled worker and professional for both Han Chinese and Uyghurs. Furthermore, the OLS regression coefficient for Uyghur*state employment is positive and statistically significant at .01 level. The findings from Model 6 of Table 3 are similar to those reported in Model 4, but the OLS regression coefficient for Uyghur*state employment is larger than that in Model 4 (1.680 compared to 1.292). These findings suggest that if Uyghur workers in these two age groups achieved similar socioeconomic status as Han workers, their earnings would be similar to those of their Han counterparts. Thus, income inequality by ethnicity in Age Groups 2 and 3 is not due to higher returns to education and occupation for Han Chinese than for Uyghurs.

Then, how can income inequality by ethnicity in these three age groups be accounted for? A plausible explanation can be derived from the regression coefficients for Uyghur*state employment for Age Groups 2 and 3, which suggest that Uyghurs benefit more than Han Chinese from state employment. For Age Group 2, Uyghur state workers earn 1,028.44 yuan per month whereas Uyghurs in other sectors earn 436.43 yuan per month. The corresponding figures for Han Chinese are 1,240.68 yuan and 966.94 yuan respectively. For Age Group 3, Uyghur state workers earn 985.46 yuan per month whereas Uyghurs in other sectors earn 463.15 yuan per month. The corresponding figures for Han Chinese are 1,243.21 yuan and 979.65 yuan respectively. The earnings ratios between Han and Uyghur state workers are 82.9 per cent for Age Group 2 and 79.3 per cent for Age Group 3. In comparison, the earnings ratios between Han workers and Uyghur workers in non-state sectors are 45.1 per cent for Age Group 2 and 47.3 per cent for Age Group 3. The ethnic differences in earnings ratios in the state and non-state sectors make sense since the state sector is legally required to protect minority rights whereas non-state sectors are not.

Given the ethnic differences in earnings ratios, it is possible to explain Han–Uyghur earnings differentials in Age Groups 2 and 3. Table 1 shows that the percentage of Uyghur state workers has decreased from 60.7 per cent in Age Group 1 to 51.3 per cent in Age Group 3, which contributes to income inequality by ethnicity: a lower percentage of Uyghur workers is protected by the state sector and a higher percentage of Uyghur workers is exposed to ethnic discrimination in the non-state sectors, earning much less than Han non-state workers. This in turn enhances the value of state employment for Uyghurs in earnings since Uyghur workers in the non-state sectors are poorly paid. Similarly, the percentage of Han state workers has decreased from 81.5 per cent in Age Group 1 to 61.6 per cent in Age Group 3, which enlarges the earnings differences between Han Chinese and Uyghurs since a higher percentage of Han workers are employed in and can benefit from discrimination against ethnic minorities in the non-state sectors. These explanations account for relative ethnic parity in income in Age Group 1, Han–Uyghur earnings differentials in Age Groups 2 and 3, and the positive regression coefficients for Uyghur*state employment in Table 3.