Double Disadvantages: A Study of Ethnic and Hukou Effects on Class Mobility in China (1996–2014)

This paper examines the ethnic and household registration system ( hukou ) effects on intergenerational social mobility for men in China. Using national representative surveys covering almost two decades (1996–2014), we assess both absolute and relative rates of mobility by ethnicity and hukou origin. With regard to absolute mobility, we find that minority men had significantly lower rates of total and upward mobility than Han men, and those from rural hukou origins faced more unfavourable chances. With regard to relative mobility, we find men of rural ethnic origins significantly less likely to inherit their parental positions. Even with parental and own educational qualifications and party memberships controlled for, we still find ethnic minority men of rural hukou origins behind others in access to professional-managerial positions. Overall, our findings suggest that the preferential policies have largely removed the ethnic differences in the urban sector but ethnic minority men from rural hukou origins are faced with double disadvantages: in addition to the inequality of opportunity rooted in the institutional divide which they share with the majority group from similar backgrounds, they face much greater inequalities in conditions, namely, in having poorer socio-economic and cultural resources.


Introduction
The aim of this paper, drawing on national representative sample surveys in China, is to examine the social inequality in intergenerational class mobility in terms of ethnic and structural relations.We address ethnic and structural inequalities simultaneously by exploring the mobility differences between the ethnic minority and the majority (Han), and for people from urban and rural origins.The ethnic dimension is easy to understand but unlike the situation in western countries, ethnicity has added complexity in China due to the scale, diversity, ge-ography and political importance of the ethnic minority groups.The structural differences refer to the inequalities embedded in the household registration (hukou) system which was implemented in the 1950s and which has limited the upward mobility chances of people from rural origins for over half a century.Any mobility research on China would be incomplete without taking into account the hukou effects.Quite a few studies have been conducted examining the hukou effects on mobility but to the best of our knowledge, no research has been conducted on intergenerational social mobility by ethnicity in China.To this end, this paper seeks to understand the ethno-hukou intersection in contemporary China by providing evidence on both the absolute and the relative rates of social class mobility.
Social mobility research and ethnic studies are, as Li and Heath (2016) note, both concerned with the same underlying issue, that of social equality and social justice.Social inequality, be it manifested in terms of family background, ethnicity/race, gender, age, disability, sexuality or other ascribed characteristics, is economically inefficient, morally indefensible and politically illegitimate.In the western capitalist countries, ethnic issues are mainly concerned with the difficulties faced by immigrants and their descendants in education, employment and career advancement, and are thus a matter of inter-ethnic relations and of socio-economic integration.In China, ethnicity has, in addition to socio-economic equality, extra considerations such as political stability, national unity and border safety as most of the ethnic minorities live in the western part of the country which tends to be mountainous, under-developed and bordering other countries.The Chinese government attaches great importance to ethnic unity and equality and has, especially in the last few decades, instigated numerous preferential policies to try to help the minorities in their socio-economic development.Quite a few studies have been made to evaluate the effects of such policies but owing to the lack of appropriate data on parental class position, researchers have thus far been unable to assess whether ethnic minorities have similar mobility chances to those by the majority, the extent of difference, and the possible reasons that may underlie such differences.The present analysis wishes to provide evidence to these questions for the first time.
This paper is structured as follows.In the next section, we give a brief review of the socio-economic-demographic profile of China's ethnic minorities, the preferential policies adopted by the Chinese Government towards the minorities to improve their life chances, and the research findings about the education, employment and income situations of the ethnic groups.A brief discussion will also be made of the main findings from the social mobility research perspective for the urban and the rural populations in China.After that, we explain the data and methods to be used in the paper, followed by presentation of our findings on both the absolute and the relative mobility rates along ethnic and hukou dimensions.In the final section, we summarize our findings with some discussion.

Ethnic Minorities in China
China is a vast country with a very large population.It is also a multi-ethnic and multicultural country.There are currently 55 officially-designated ethnic minority groups in addition to the Han majority.According to the 2010 Census of the Population of the People's Republic of China (PRC), 114 million people were of ethnic minority heritages, accounting for 8.49% of the overall population (National Bureau of Statistics of the People's Republic of China, 2011).
Whilst ethnic minorities take up a relatively small proportion in the overall population in China, they occupy a strategically important position as they live in areas comprising around two thirds of the nation's territory.The Chinese Communist Party (CCP) enacted pro-ethnic policies in its controlled areas even during the war years, promising equal rights for all ethnic minorities (Sautman, 1999).After the PRC was founded in 1949, the central government made numerous efforts to protect minority interests and promote their socio-economic development such as by giving them official legal status, creating written language for some of the groups, guaranteeing equal rights in the 1982 Constitution, promoting equal opportunities in the 1984 Law of Regional Autonomy and many subsequent laws and regulations at the central and provincial levels (Gladney, 1996;Lin, 1997).
Owing to the low population density and the vast areas of territory the ethnic minorities populate, the central government gave areas with relatively high ethnic concentrations a special political and administrative status called 'ethnic autonomous regions'.Five such autonomous regions were created at the provincial level (Inner Mongolia, Guangxi, Tibet, Ningxia, and Xinjiang), 30 autonomous areas at the prefectural level and 120 autonomous administrative units at the county level (National Bureau of Statistics of the People's Republic of China, 2015).Being an autonomous region does not necessarily mean that ethnic minorities comprise the overwhelming majority of the residents in the area.As a matter of fact, they only consist of 45% of the population in the five regions.But the overall socio-economic levels in the areas are below those in the rest of the country, especially the big cities and the coastal areas which have witnessed rapid development since the reform programme was launched in 1978.
China has a tradition of treating all ethnic groups as members of a big family in the nationhood building.In fact, in the Chinese language, the word 'country' (guo jia) literally means 'nation family'.To ensure equal rights amongst the family members (the different ethnic groups in this case), the central government and the CCP have adopted many measures to enhance the political, social and economic wellbeing of the ethnic minority groups.Several social movements were launched in the 1950s and 1960s to combat Han chauvinism in minority areas, to show full respect to the cultural traditions of the ethnic minority groups, to avoid condescendence in attitudes, behaviours and practices when working with people of ethnic minority heritages, and to fight against discrimination of one kind or another, especially in recruitment and promotion (Chao, 1994;McMillen, 1979, p. 114).At the same time, tens of thousands of educated youths and technical personnel were sent from the majority population to the autonomous areas to help with the socioeconomic development, and to strengthen the national security of the border areas (Attané & Courbage, 2000).
The ethnic minorities, like people in the rest of the country, suffered during the chaotic period of the Cultural Revolution (1966-1976), with minority cultural customs labelled as primitive, minority cadres exposed to ridicule, and pro-ethnic policies regarded as unnecessary (Dreyer, 1968;He, 2005, p. 69;Lin, 1997).Since the early 1980s, a wide range of policy measures have been adopted to mitigate the adverse influence of the Cultural Revolution and to support the socio-economic development of minorities including, most importantly, the 1982 State Constitution of China in which equal rights were enshrined in law for all ethnicities in the country.The previous autonomous administration system was reinstated at the provincial and lower levels, and numerous preferential policies were enacted and implemented from the central to the local governments to help the minorities, including exemptions from the one-child family-planning policy, lower marks for university admissions, economic assistance and tax reduction, and more favourable opportunities for recruitment and promotion.The overall aim of the policies was not only for the minorities to have better opportunities, but also to achieve the ultimate aim of equal outcomes among all ethnic groups (Chen, 1991).
As compared with 'affirmative action' or 'positive discrimination' programmes in some other countries, China's preferential policies are, as Sautman notes, 'broader (encompassing a wider array), deeper (affecting almost all minority people) and more variegated (due to decentralization) ' (1998, p. 86).Perhaps the most notable example of such policies pertains to the minorities' exemption from the strict regulations on family planning.The one-child policy was implemented nationwide in the 1980s, but most of the minority groups were allowed to have two children in the urban areas and three or four in the rural areas (Attané & Courbage, 2000;Bulte, Heerink, & Zhang, 2011;Sautman, 1999).Although the policy varied to some extent over time and across regions, the principle of showing respect to minority cultures by allowing them to have more children remained unchanged.Yet, as we are going to see, this preferential treatment in the form of 'greater opportunities' may have yielded an unintended consequence, in the sense of aggravating the already existing inequality of condition and creating additional barriers for mobility.We shall come to this later when we discuss the findings.Here we focus on policies that might have had direct and positive impacts on the minorities' life chances, including those on education, employment, and economic support.
With regard to education, China's Compulsory Education Law (1992) covers all children for nine years, from primary to lower secondary schools.Yet full coverage was not achieved in the remote areas, most of which were in minority autonomous regions.Many people in these areas were in dire poverty and could not even afford the basic teaching materials.In order to improve the coverage and ensure the basic standard of education in the minority areas, the central government allocated special resources called the Ethnic Minorities Education Aid Special Fund.In 1994, the state raised the funding to one billion Chinese Yuan per year, a fivefold increase as compared with the previous years.With the assistance of the fund, hundreds of boarding schools were built in the remote areas.Students were exempted from fees for tuition and books and, in many cases, were guaranteed free food, clothing, lodging, and supplementary study materials (Sautman, 1998, p. 91).Ethnic autonomous regions were also encouraged to develop their own curriculum and use their own language in the local schools.These programmes catered for the special needs on language, culture and historical traditions of the local ethnic communities, and the expenses in teacher training, curriculum designing and extra printing were partly funded by the government budget (Postiglione, 2009, p. 504).Yet the special treatment for teaching ethnic languages and cultures also added to the students' difficulty in learning, making them proficient in neither the local nor the national language and lowering their prospects for advancing into higher levels of education.
As for higher education, ethnic minority students enjoy a range of preferential policies.For university admission, ethnic students were given extra points on the basis of their original marks in the national entrance examination or were offered a lower threshold for admission.For example, Sautman (1998) reported that in Xinjiang in 1987, the Han (majority) students needed an average score of 470 points in science and 445 points in liberal arts for access to key universities, while the minorities only needed an average score of 313 and 269 points respectively.In addition, the government has established more than 20 universities for ethnic minorities, mostly located within the minority areas, so that the minority students could have access to higher education near their hometown (Huang, 2000).These universities mainly enrol students from minority backgrounds and offer special subjects based on the needs of different ethnic regions.In order to enhance the education quality of minority students, the state has initiated a programme for top national universities to form partnerships with those in minority autonomous regions for cooperation and assistance, such as teacher training, facility donation, library resource sharing, and various other forms of exchange (Yang & Wu, 2009).
The preferential treatment does not stop at education.Ethnic minorities in China also enjoy favourable treatment in recruitment and promotion, especially in the autonomous regions.The 1993 Regulations on Work with the Urban National Minorities called on enterprises to recruit more minority workers in return for lower taxes or tax breaks.As in the case of university admission, when examinations were required for job seekers, minority applicants would have bonus points.For instance, in the employment examination for an oil and chemical factory in Urumqi, the capital city of Xinjiang autonomous region, applicants from the Han majority needed 250 points whereas the minimum for minorities was 120.In some cases, the minority status is necessary as employers need to meet the job quotas to reflect the ethnic composition of the population in the local area (Sautman, 1998).
The preferential policies give even greater advantages to ethnic minorities for access to cadre positions or promotions within government agencies.According to the 1982 State Constitution of China, the head of ethnic autonomous regions must be a member of the ethnic group that comprises the majority of the population in that particular location, and the preference was then extended to such an extent as to prioritise minority candidates in the recruitment of lower-level government personnel as well (Gustafsson & Li, 2003).As a result, many places have seen an upsurge of hiring and promoting minority cadres since the 1980s.Of particular note in this context is the fact that China has a dual leadership system, consisting of CCP leaders and administrative officials at all levels of government agencies and state-run enterprises.The party leaders in the minority areas do not have to be members from the ethnic minorities and thus the minorities are more represented in the administrative than in the party organisations.
The Chinese government has not only provided financial and other kinds of help to individuals of ethnic minority heritages in education and employment but also enacted favourable economic policies for the autonomous regions where minority ethnics tend to reside.In the more recent times, the government has initiated two grand projects, The Development of the Western Region ( 2000) and The Silk Road Economic Belt ( 2013), in which all the five autonomous regions were included as key areas for national and international investment.In addition, various preferential policies were introduced to support economic growth in minority autonomous areas.As instigated in the 1982 Constitution, the minority autonomous areas are allowed lower tax rates and greater fiscal controls over their local revenues (Zang & Li, 2001).At the meso level, programmes such as governmentsubsidized microcredit have also been set in motion prioritising poor households in minority areas, and direct subsidies and tax breaks have been granted to these areas and to businesses run by minorities.Given the vast array of economic and other favours enjoyed by ethnic minorities, having an ethnic minority status is, or is perceived as being, advantageous, which can be seen from the fact that millions of people changed their ethnic identity from Han to minority status in the 1980s and the 1990s to benefit from the preferential policies conferred upon the minorities (Sautman, 1999, p. 286), and those of mixed ethnic backgrounds have also been found to prefer a minority identity, which have contributed to the increase in the proportion of the minorities in the overall population (Hoddie, 1998).A study using a more recent census data arrived at a similar conclusion, namely, that children from inter-ethnic families are more inclined to be registered as minority ethnic rather than as majority Han (Guo & Li, 2008).
Given the breadth and depth of preferential policies implemented in China in the last few decades, one might ask whether ethnic minorities are still disadvantaged or whether they have caught up with or even surpassed the majority group in socio-economic life.Existing studies have explored this question from different perspectives and we here give a brief summary of the effects in education, income, and occupational attainment.
With regard to educational attainment, studies have shown that, after decades of government assistance, ethnic differences in primary schooling have been largely removed, but disparities remain at the transition to lower and higher secondary schools (Hannum, 2002;Hong, 2010).As for college education, although the number of minority students has been increasing rapidly, the percentage is still lower than expected.The ethnic minorities accounted for 8.41% and 9.44% of the national populations in 2000 and 2005 respectively,1 but only 5.71% and 6.10% of the university students came from minority backgrounds in the two years (Zhu, 2010).Moreover, one-fourth of these students were enrolled in ethnic minority colleges, which means the percentages of students studying at universities outside the autonomous regions were even lower (Qin, 2004).
Why are ethnic minorities underrepresented in higher levels of education?Scholars believe that regional economic differentiations and household socioeconomic disparities were the most important factors.The ethnic disadvantages tended to increase after the market transition as the local economic conditions have been playing an increasingly important role in local school resources and educational quality (Hannum, 2002).A lower level of proficiency in Mandarin has also been reported as a crucial contributing factor to the ethnic minority children's poorer school performance.Minority students with a family background using Mandarin were around twice as likely to be enrolled in high school as those who used minority languages at home (Hong, 2010).Bilingual education using Mandarin plus a minority language was an attempt to preserve minority cultures, but it added burdens to minority students from poor homes who were already struggling with their studies (Lin, 1997), incurring dissatisfaction among ethnic parents (Ma, 2007).On the other hand, the preferential policies enjoyed by the minority students also caused dissatisfaction from among the majority students and their parents who lived alongside the minorities in the minority regions (Ma, 2008).
Turning to income, findings are mixed.Some studies have shown that minorities are economically worse off than are the Han majority.For example, Gao and Teng (2006) showed that the ethnic regions in northwest China had an underdeveloped industrial economy which accounted for the ethnic poverty in those regions.Gustafsson and Sai (2009) found that the minorities had a poverty risk almost twice as high as that for the majority in rural China, but they attributed the difference to location.When location and household characteristics are both taken into account, they conclude, ethnicity does not 'have much of an independent effect on poverty status' and the most promising policies for narrowing the inter-ethnic poverty gaps should, according to the authors, be those 'promoting growth in low-income villages (that are concentrated to western China) irrespective of the ethnicity of the inhabitant' (Gustafsson & Sai, 2009, p. 604; see also Gustafsson & Li, 2003).In other words, the authors believe that, in fighting against poverty-related inequality, it is location, rather than ethnicity, that matters.
Yet ethnicity is found to matter in terms of occupational attainment.Hannum and Xie (1998) examined occupational differences between the Han majority and the ethnic minorities in Xinjiang.Using census data from 1982 and 1990, they found a rising occupational dissimilarity between Han and the minorities.For example, the overall percentage of the labour force in agricultural occupations declined from 58.9% to 49.6% between the two-time points, but the percentage for the minorities increased from 69.4% to 76.7%.The rising concentration of the minorities in agricultural work was accompanied by a falling representation in almost all other occupational categories, and educational inequality was held as a key factor for the increasing dissimilarity, especially in the attainment of high-status occupations.In a similar vein, Sautman (1999) found that whilst ethnic minorities had an overrepresentation of cadres as benefiting from the preferential policies in cadre appointment, they were underrepresented among technical and professional occupations.As both studies were based on Xinjiang, it is not possible to know whether the findings would be generalizable to the ethnic occupational difference in the overall population.
These and other studies which we have not been able to include here due to space limit of the paper have much enhanced our understanding of the ethnic minority disadvantages in the different domains of socio-economic life in China.Yet, there is one common feature, or rather a common limitation, in all such analyses, namely, the lack of information on parental social position which has prevented researchers from exploring the family origin effects on the ethnic minorities' mobility chances.In other words, much work has been conducted comparing the inequalities of opportunity but little work has been carried out that examines the inequalities of condition and the consequences of such inequalities on life chances between the minority and the majority groups.
With regard to studies that do explore the effects of family class, another deficiency is shown.Parallel to what Li and Heath (2016) observed for Britain, ethnic and mobility research scholars in China have also been travelling on separate tracks with the former ignoring family class and the latter ignoring ethnic effects.China's situation is more complicated than that in Britain though, due to additional layers of stratification.The most significant of such layers is that of the household registration (hukou) system.As earlier noted, this system was implemented in 1955 requiring all new-born babies to be registered following mother's status, forcing the great majority of the Chinese people at that time and in most of the subsequent periods to have rural hukou, unable to enjoy the benefits provided by the government to urbanites in education, employment, housing, medical insurance and a host of other areas.For decades, the hukou system has served as the most important institutional divide preventing rural citizens from upward mobility.Given this, almost all important mobility analyses in China have taken hukou into consideration.
Within the mobility domain, one of the most important studies is that by Wu and Treiman (2007).Using a national representative sample called Life History and Social Change (conducted in 1996), the authors conducted a thorough analysis of intergenerational social mobility for men in China and found a large-scale downward mobility into agriculture, a feature not elsewhere found in international mobility research.Yet this apparently surprising finding is exactly what would be expected from the hukou system: rural sons cannot follow in their fathers' footsteps if the fathers were professionals or managers working in the countryside such as teachers, doctors or cadres but their mothers were agricultural workers (peasants).Only a small proportion, 'the best and the brightest', of those from rural origins could gain entry tickets to urban life, mainly by gaining admission to university.On the other hand, given the huge size of China's countryside, even a small proportion from the rural backgrounds who managed to achieve hukou transition could make a big presence in the cities and, what is more, the exceptionally gifted, highly educated, new urbanites tend to take up elite positions.The hukou system had a hugely negative effect on rural people's mobility chances, which is also reported by Chen ( 2013 2010) for Britain to compare the mobility profiles in the two countries.They found that social inequality in China was much greater than that in Britain, especially for Chinese women.This, the authors showed, was chiefly due to the hukou system.Within each hukou sector, social inequality in China was little different from that in Britain.
Yet, in spite of the advances in mobility research in China, none of the available studies have looked at the ethnic differences in mobility chances.Thus, sadly, ethnic research has failed to look at the family origin effects just as mobility research has neglected the ethnic dimension in social stratification.Given this, it is imper-ative to know whether social reproduction mechanisms operate in a similar way among the ethnic minorities to that among the majority in China, or whether the ethnic minorities face additional barriers over and above those caused by class origin and hukou status.

Data and Variables
In the present study, we use the Life History and Social Change (LHSC, 1996), China General Social Survey (CGSS, 2005(CGSS, , 2006(CGSS, , 2008(CGSS, , 2010(CGSS, , 2011(CGSS, , 2012(CGSS, , 2013) ) and China Labour-force Dynamics Survey (CLDS, 2014).Although conducted by different institutions, all three survey series employed multistage probability sampling methods to draw national representative samples for mainland China.While there are some differences in geographical coverage between individual surveys, overall, all 31 provinces, autonomous regions and municipalities in mainland China are covered in the surveys. 2 The CGSS series also includes data for 2003 but that survey covers urban residents only and is therefore not used in the present analysis.
In this study, we use 'ethnic minority' as an overall category without further differentiating between the 55 minority groups.This by no means implies that they are homogenous.The Korean minority, for instance, have been found to have the best educational attainment in China (Lee, 1986).Thus, different ethnic groups do have considerable demographic, socioeconomic, and cultural differences including fertility, mortality, education, occupation, and use of Mandarin (Ma, 2008(Ma, , 2010;;Poston & Shu, 1987;Yusuf & Byrnes, 1994).We group them together based on two considerations: first, all ethnic minorities enjoy similar preferential policies despite their specific historical and cultural characteristics, which sets them apart from the Han majority.Secondly, as the minorities account for less than 10% in the population, further differentiations would lead to very small sample sizes for the groupings, making it untenable in statistical analysis.
With regard to father's and respondent's class, we use the EGP class schema (Erikson, Goldthorpe, & Portocarero, 1979), which is the most widely used class schema for intergenerational social mobility research adopted by practitioners in many countries including China (Chen, 2013;Li et al., 2015;Hannum & Xie, 1998;Wu & Treiman, 2007).We acknowledge that other markers of institutional division such as workplace unit (danwei), hukou status or political power might add complexity to the social stratification of the Chinese society, and mobility students on China may apply class-, statusor power-based schemas.In order to take account of the unique characteristics of the Chinese society and to make the results comparable to mobility findings from other countries, we adopt the EGP schema.In our further analysis, we also, where appropriate, control for other attributes relevant to mobility such as age, education and CCP membership.We take the hukou-ethnicity combination in much of our analysis as a key explanatory variable on people's mobility trajectory alongside father's class.The EGP class schema is based on employment status and occupational position (current or last main job).All information with regard to respondent's and father's employment status and occupational position was collected in the surveys and was coded in a consistent way into own and parental class variables.In the present study, we differentiate, for fathers and respondents alike, five classes: (1) professional and managerial salariat, (2) routine non-manual (clerical), (3) self-employed (ownaccount), (4) routine manual, and (5) agricultural workers (also called 'peasant' in China).For hukou status, we use the respondent's origin hukou, namely, the status in the adolescent years (around age 14-18).We confine our analysis to men aged 16 to 65 in this analysis, and exclude those who have never worked such as students, which gives us a sample size of 35,147 for analysis.Gender and workplace affiliation (danwei) are also very important dimensions of social stratification in China (Li et al., 2015;Lin & Bian, 1991) but space limitations do not allow us to explore these in this paper.

Analysis
Following standard approaches in the literature we address both absolute and relative rates of mobility by ethnicity and hukou status.To assess absolute mobility rates, we start with the class distributions by the various ethnohukou groups, supplemented by an analysis of gross and net amounts of mobility.The former is measured by using the dissimilarity index (DI) and the latter by using Lieberson's (1975) net difference index (NDI).The DI indicates the percentages of cases that would have to be reallocated to make the two distributions identical and thus measures the overall difference between any two distributions.As the DI is insensitive to the local concentrations in or the direction of differences between, father's and son's classes, we use the NDI to show the net class decline or advancement of the son's relative to the father's class. 3Proceeding from this, we analyse the total, 3 The NDI is defined as NDxy = pr(X > Y) − pr(Y > X) and further defined as where, in our case, X indicates father's class and Y that of the respondent.It is noted here that we reversed the class order in calculating the NDI with 1 referring to agricultural workers and 5 to the salariat.
upward and downward mobility rates and examine each by ethno-hukou differentiations.To address relative mobility, we use loglinear and log-multiplicative layer effect (also called uniform difference, UNIDIFF) models.These are then supplemented with logit models on son's access to the professional-managerial salariat controlling for ethno-hukou status, father's class, and a range of socio-demographic attributes.

Overall Distributions and Absolute Mobility
We begin with the overall class distributions of fathers and sons, distinguishing the Han majority and the ethnic minorities and, for each, within the urban and the rural sectors.The data are shown in Table 1 where we also show the DI and the NDI.Looking first at the urban sector, we can see that there is not much difference between the Han majority and the ethnic minorities in terms of father's or son's class distributions.Minority fathers have, if anything, a somewhat higher percentage in salariat positions (27.3%) than Han fathers (25.2%).Minority sons are less likely to find themselves in salariat or clerical, and more likely to find themselves in own-account, positions relative to their majority peers.When we look at the rural sector, we see a clear manifestation of ethnic divide.
As expected, the overwhelming majority of fathers from both groups were agricultural workers, 80% for Han and 84% for minority fathers, and there is little difference in the distributions to the other class categories.Yet, this overall similarity in fathers' distributions is not shown for sons.Sons from the Han majority group were clearly more likely to occupy higher social positions, from the salariat to the manual working class and are therefore much less likely to find themselves in agricultural positions (44.6%) as compared with minority men (64.8%).
The DI and the NDI also show greater gross mobility and net class advancement by Han than by minority men, and that in both urban and rural sectors.The data in Table 1 thus show that men from ethnic minority backgrounds were disadvantaged.This, of course, is an overall profile, without taking into account the effects of father's class on son's class.In Table 2, we show the class distribution of sons by father's class.For simplicity, we do it by ethnicity but in the 'summary statistics', we also report findings from further analyses by hukou status and by the ethno-hukou combination.
The ethnic differences are clearly shown in the two panels.For those from salariat families, we see that 38.2% of Han but only 28.7% of minority sons were able to get salariat positions and the risks in downward mobility into agriculture was three times as high: 34.5% versus 10.5%, or a difference of 24 percentage points.The difference in intergenerational stability in agricultural jobs is at a similar level: 68.5% for minority sons and 49.3% for Han sons.And comparing the figures under columns 1 and 5, we can see that the majority (Han) sons from each class backgrounds were more likely to find themselves in salariat, and less likely to find themselves in agricultural, jobs.
The data in the two panels of Table 2 are colour-coded and presented in row percentages.Had we conducted the analysis in cell percentages, we could have easily obtained the total immobility and mobility rates, with the latter further differentiated into upward and downward mobility.The details are shown in the lower part of the table called 'summary statistics' where, in addition to ethnic rates, we also present the rates by hukou and, furthermore, by ethno-hukou combinations.Also shown are results of statistical tests with Han, Urban, and Han urban as reference groups in the three domain respectively.
At the overall level, over half (53.5%) of the Chinese men were in different class positions to their fathers, with 41.8% having better jobs than their fathers and only 11.7% faring worse.Yet the ethnic differences were selfevident, with the minority men having significantly lower rates in both total and upward mobility rates than their majority peers.The hukou differences were less clearcut, though.Men of rural origins had lower total mobility rates than their urban peers but their upward mobility rates were actually significantly higher, and downward mobility rates significantly lower, than those of their ur-  , 1996), China General Social Survey (CGSS, 2005(CGSS, , 2006(CGSS, , 2008(CGSS, , 2010(CGSS, , 2011(CGSS, , 2012(CGSS, , 2013) ) and China Labour-force Dynamics Survey (CLDS, 2014).
Social Inclusion, 2017, Volume 5, Issue 1, Pages 5-19 ban peers.This is most likely due to the occupational upgrading witnessed in the last thirty years, with 260 million migrant workers moving to cities to find jobs not all of which were routine manual in nature.Further analysis shows that of those from rural origins and currently still holding rural hukou status but not engaged in agricultural work, 13% were doing professional or managerial work, and 9% were doing clerical work.The rural sector has been held at a much lower place since the 1950s and it is thus easier for rural residents to move up and more difficult for them to move further down relative to the urbanites.And looking at the data on the ethnohukou combination, we find what we have suspected: minority men of rural origins have benefited least from the socio-economic development of the country-they have significantly lower rates in both total and upward mobility than do the urban majority peers.4

Relative Social Mobility
We now turn to a different way of looking at class reproduction, namely, relative mobility.Our research questions can be expressed as follows: do we find a weaker association between class origins and destinations among the minority than among the majority group?Do we find a 'perverse openness' where minority status tends to 'trump' class so that minority men are held back irrespective of their fathers' class situation?Relative mobility (often termed 'fluidity' or 'openness') refers to the competition between people from different origins in obtaining advantaged and avoiding disadvantaged destinations, and is expressed in odds ratios.The closer the odds ratio is to 1, the weaker the class reproduction (and the greater the level of social fluidity or openness).Our interest, then, is whether the odds ratios for the minority group are closer to unity than they are for the majority group.Standard techniques for testing the relationships are to fit three models: the conditional independence model which serves as the baseline; the common social fluidity model (CmSF), which allows for an association between origin and destination but not the three-way interactions; and the log-multiplicative layer effect (also called uniform difference, or UNIDIFF) model which provides an assessment of the extent to which the minority group differs from the majority in the magnitude of the origin/destination odds ratios.This third model provides us with a general test of differences in fluidity, testing whether there is a uniform pattern for the odds ratios to be closer to (or further away from) 1 in a particular layer of the table.To save space, we do not provide the statistical formulae of the three models (interested readers could see Li & Heath, 2016).For the same reason and as the model fit statistics are quite abstract, we are not going to show the statistic details here (results available on request).Basically, none of the three models fit the data by ethnicity, hukou, and ethnohukou combination to an acceptable degree but the UNI-DIFF models do provide a significant improvement in fit over the CmSF models.Yet, it is of interest to present the UNIDIFF parameters for the ethno-hukou combination (Figure 1), which shows that the association between father's and son's class positions is significantly weaker for the minority men of rural origins than for the majority of urban origins.
The UNIDIFF model is good at capturing the magnitude of difference in the origins-destinations association between one layer and another, but it is not good at pinpointing the aspect in which the odds ratios defining the origins-destinations association is particularly strong or weak.As Li and Devine (2011) and Li and Heath (2016) show in the British context, the relative mobility can be much more salient in some aspects of the class competition than in others.To see this more clearly in the Chinese context, we show the symmetrical odds ratios.espaco The data in Table 3 show the relative social advantages and disadvantages associated with different classes and along the ethnicity, hukou and ethno-hukou dimensions.Take the top left-hand cell of the table for example.The figure (1.93) means that for men from the majority group (Han), they are nearly twice as likely to find themselves in salariat rather than clerical positions if they are from salariat families as compared with men from clerical families in the same kind of competition (for salariat rather than clerical jobs).The figure below it (2.45)refers to the odds ratio for the ethnic minority men.The minority figure is slightly higher than that for the majority but the difference is not significant.Now if we turn our gaze to the top-right cell, we see 16.99 and 8.26 as odds ratios for the competitions between Classes 1 and 5 for the majority and the minority men respectively.Thus, among the Han, as compared with a peasant's son, a salariat's son is 17 times as likely to find himself in a professional or a managerial position rather than as an agricultural worker.Compared with the majority, the minority is less unequal, with the odds ratio being half the magnitude, and the differences between the two sets of odds ratios are significant at the 0.001 level.The lesser inequality as shown here implies a much greater disadvantage faced by the ethnic minorities in being less able to inherit their father's superior class positions.There are, as expected, pronounced hukou effects on mobility chances, particularly surrounding Class 5 (agricultural workers).Thus, as we see in Panel 2, men of urban hukou origins tend to have more favourable chances, at three times or higher, than do men of rural origins in the class competitions, namely, in competing to avoid agricultural positions.For instance, in the Classes 1 versus 5 competitions, the odds ratios are 18.3 and 6.4 for urban and rural sectors respectively, and 16.3 and 5.8 for the routine manual versus peasant sons, both pairs being significantly different at the 0.001 levels.
There are thus both ethnic and hukou differences which are most concentrated in the class competition in trying to arrive at higher, and to avoid agricultural, jobs.The combined effects, as shown in Panel 3, bring the disadvantages of minority rural men to the sharpest relief.It is true that among these men, coming from salariat families still have an advantage of around 3.3 times being themselves found in salariat rather than peasant jobs as compared with their peasant peers in the same kind of competition, but this is less than half than for their Han peers (7.2), and well below that for the urban co-ethnics (26.3) or for the urban Han (16.5).Further analysis, using minority rural men as the reference group, shows that all three groups are significantly more advantaged (p.0.001, 0.016 and 0.009 respectively).
We have, in the preceding analysis, looked at the absolute and the relative mobility rates associated with ethnicity and hukou origins.Family class, ethnicity and hukou statuses are shown as being very important markers of social inequality in China.There are, however, other factors which we could not consider due to constraints in descriptive and loglinear modelling.For instance, minorities tend to reside in underdeveloped areas and face more structural barriers.In our data, only 22% of minority men as against 35% of the Han have highschool education or above.The former are also less likely than the latter to be CCP members (12% and 15% respectively).In the remaining part of this section, we turn to logit analysis on access to professional-managerial salariat to try to control for these and other factors.We construct five models.Model 1 includes father's class and respondent's own age; model 2 adds father's education and CCP membership as well as respondent's ethnohukou status (we use rural Han as the reference group for modelling stability as this is the largest group in our data); models 3 and 4 add respondent's education and CCP membership respectively (as indicators of cultural and political capital); and model 5 further adds an interaction term for ethno-hukou and education (level of education is used as a continuous variable here following Li & O'Leary, 2007).
With regard to data in Table 4, we find, in model 1, strong effects of father's class, and with clear gradients.Even with age differences taken into account, coming from salariat families is highly conducive to one's career life, increasing the odds of salariat access by around five-fold (e 1.604 = 4.97).When father's education, CCP membership and respondent's ethno-hukou status are entered in Model 2, we find that these have highly significant effects and that the effect of father's class, albeit reduced, is still highly significant.In Model 3, respondent's own education is taken into consideration which is shown as being of chief importance in explaining salariat access while the effects of father's education and CCP membership as well as respondent's own ethno-hukou status are reduced to a general insignificance, suggesting that these effects are mediated via respondents' education.Surprisingly, we see that father's class is still highly significant.In Model 4, we further control for respondent's own CCP membership which is found to be highly conducive to salariat access (in our data, 37% of the salariat, 23% of the clerical workers but only 8% of the agricultural workers are CCP members: joining the CCP is usually a first step in one's career advancement in China).Finally, in Model 5, we control for the interaction effects of ethno-hukou status and education.The pattern of the coefficients for the other variables in the model are little changed as compared with those in Model 4 but we do find that the returns to the majority men of urban hukou status are negative, echoing Wu and Treiman's (2007) finding that given the institutional divide, only the 'best and brightest' from rural origins could gain their passport to urban life via higher education and those who do manage to do so tend to occupy the elite positions.Also noticeable is the fact that the coefficients for the minority rural men remain strongly negative and persistently significant from Models 2 to 5.
Finally, we present a bird's eye view of mobility chances conditional on father's class and ethno-hukou status (Figure 2).The data are obtained from the predicted probabilities based on Model 5 of Table 4, with the effects of the covariates taken at their mean values.Hence the data pertain to the net effects of father's class and ethno-hukou statuses.The stratification order is clearly shown.Other things being equal, a minority man of rural hukou origin would still have a smaller chance of becoming a professional or manager than his Han compatriot with urban hukou even if the former comes from a salariat family and the latter from agricultural origin (typically working on state farms): 21.5% versus 22.6%.And the disparity runs seven times as high between urban Han men from salariat families and rural minority men from peasant families in the competition for salariat destinations (44.9% and 6.6%).

Discussion and Conclusion
We have, in this paper, made what we believe as the first attempt to study intergenerational social mobility by ethnicity in contemporary China.Drawing on the best available data containing the key variables for the analysis (father's and own class, ethno-hukou statuses and a range of other factors), we conducted the analysis of both absolute and relative rates of mobility for men from minority and majority ethnic groups and of rural and urban hukou statuses.Our main findings can be summarised as follows.
In terms of gross mobility, we found that minority and majority men from the urban sector were not much different from each other but those from rural origins had greater mobility, especially those from the majority group who had the greatest net class advancement.This is most likely due to the massive internal migration in the reform period where an estimated 260 million people of rural status went to cities to find jobs in the burgeoning and rapidly developing industries and commercial enterprises.Rural men of minority backgrounds also made use of the opportunities but to a lesser extent.Coming from a much lower starting point, ethnic minority men from rural origins not only lower total mobility, but their upward mobility rates were also the lowest as compared with the other groups.They had fewer family resources to start with and even those from similar backgrounds were less able to inherit their family resources, as shown in the 'symmetrical odds ratios', resulting in the significantly weaker association between origins and destinations.Even with parental and own cultural and political capital taken into account, we still found that ethnic minority men from rural origins were significantly behind other groups in access to the salariat.The net mobility profiles, as we saw in Figure 2 on predicted probabilities, show a clear stratification order in China with three main features.Firstly, family class has a very important influence for all ethno-hukou groups but more so for the majority men from urban origins than for others.Secondly, those of minority rural statuses faced the gravest disadvantages, much more so than their majority peers from similar (rural) origins.Thirdly, ethnicity is also shown to have some influence for the urban sector but the effects are generally weaker than for the rural sector.
How do we explain the findings?A great deal has been written about the household registration (hukou) system in creating the barriers against the rural citizens in upward social mobility (Bian, 2008;Chen, 2013;Li et al., 2015;Lin & Bian, 1991;Wu & Treiman, 2007) and the disadvantages faced by rural citizens are expected.We have also seen above that a whole range of preferential policies, especially in the last thirty years or so, have been adopted by governments at all levels to create more favourable opportunities for the ethnic minorities, and such policies cover a host of areas from education, employment to economic development.Given such favourable opportunities, why do ethnic minority men from rural origins still lag behind?
There are of course many reasons accounting for our findings the most important of which would be the longstanding poverty due to the relative underdevelopment of the local economy in the minority regions, especially in rural areas.The government assistance, in the forms of preferential policies, direct funding and investment, has no doubt helped the minority populations to ameliorate the situation.Without such generous support, the ethnic gaps in mobility would have been much larger.Yet some of the policies, well-intended as they were in respecting and preserving minority cultures such as the exemption from the family-planning policies, might have had the unintended consequences of perpetuating their poverty.For rural ethnic families already struggling in poverty and with rather limited upward mobility chances available in the local economy, having three, four or more children would only reduce their chances for improvement as compared with urban families with smaller family sizes or even with the majority peers in the rural, more developed, area where the family-planning policies were generally reinforced.
The bilingual education adopted for some ethnic schools may also have had the unintended consequences of adding to the learning difficulties of ethnic children, making it harder for them to compete with Mandarin speakers (Lin, 1997;Ma, 2007).This might not be a serious problem in the developed areas with high quality of education, such as areas where ethnic Koreans are located, but it may have aggravated the educational disadvantage for most of the other ethnic minority groups.Furthermore, the lower marks required of ethnic students for college admissions, whilst a clear boost to ethnic equality, may also have lowered their competitive edge as compared with that of the majority students studying in national universities.In an increasingly market-oriented economy, demonstrated competence, rather than a formal diploma or certificate, is becoming the order of the day in China.
There are many other reasons which we cannot cover here.Overall, we would say that ethnic minorities may have enjoyed equal or better opportunities (as entailed in the preferential policies) relative to the majority group but it is the inequality of condition that sets the two groups apart.Thus, the government should do more, rather than less, to help the minorities, especially those in the rural, poverty-stricken, regions.
Finally, we need to draw the readers' caution that, even though we have used what are the best data sources currently available for ethnic mobility analysis, there are limitations in our study.Firstly, due to the lack of data between 1996 and 2005, we could not explore mobility trajectories in the intervening years.On the other hand, given the pattern as shown in Footnote Figure 1, there is little reason to believe that the mobility trajectories in those years would have diverged greatly from what is revealed in the analysis had data for the years been available.Changes in the mechanisms of social reproduction usually take place at a rather slow pace, as the voluminous mobility research in the world over half a century has shown.Secondly, although our pooled data cover all 31 provinces, autonomous regions and municipalities, Tibet, Qinghai and Ningxia were underrepresented as they were not covered in the earlier years.Thus, the minority groups in those areas were less represented than in other areas.While this is less than desirable, we need to bear in mind that the minority groups in these provinces are mainly Zang and Hui ethnic minorities who are also found in large numbers in other provinces and cities such as Sichuan, Chongqing, Shaanxi, Gansu, Henan and Shanxi which were well represented in the data used in the analysis, and this would give us fairly good reasons to believe that the mobility chances of their co-ethnics who were surveyed in the other provinces will have been generally representative of their own trajectories.This having been said, the lower growth rates of the local economies in the three provinces in the last few decades might have had negative effects on the upward mobility chances of the ethnic minorities, resulting in an underestimation of the ethnic gaps in the provinces concerned.This needs to be further examined in future studies if and when appropriate data become available.Thirdly, as the minority groups constitute only a very small proportion in the overall population, we have not been able to make further differentiations amongst the ethnic groups in the present study.Fourthly and for similar reasons, we have not been able to examine the spatial effects on ethnic stratification.Our findings of overall similarity between the minority and the majority groups among the urban sector suggest, as Gustafsson and Li (2003) and Gustafsson and Sai (2009) showed in their poverty studies, that there might be strong spatial effects.Maybe ethnic mobility chances are similar in the more developed, coastal, provinces.Overall, while this is the first systematic study of social mobility by ethnic groups in contemporary China based on the best data currently available, there are many other domains of social mobility that we have not been able to explore.Our analysis has also opened up some new channels of exploration which future analyses could undertake, with more and better-quality sample surveys that are being collected in China.
) using the 2005 and 2006 China General Social Survey (CGSS).More recently, Li, Zhang and Kong (2015) used more data from the CGSS (2005, 2006, 2008, 2010) for China, and the General Household Survey (2005), the British Household Panel Survey (2006) and the UK Household Longitudinal Survey (
Notes: The odds ratios in each cell are for Han and minority in Panel 1; for urban and rural hukou origins in Panel 2, and for Han urban, minority urban, Han rural, and minority rural in Panel 3, respectively; For cell counts below 5, the odds ratios are not shown (as indicated by -); Significance tests are conducted with Han in Panel 1, urban in Panel 2 and Han urban in Panel 3 as reference groups.* p < 0.05; ** p < 0.01; *** p < 0.001.Social Inclusion, 2017, Volume 5, Issue 1, Pages 5-19

Table 1 .
Fathers' (F) and Respondents' (R) class distribution by ethnicity and hukou status.
Source: The Life History and Social Change survey (LHSC

Table 2 .
Class distribution by class of origin (percentage by row).
Note: Significance tests are conducted on the summary statistics.* p < 0.05; ** p < 0.01; *** p < 0.001.Minority rural men have significantly lower levels of total and upward mobility than Han rural men at the 0.001 level in both respects.

Table 3 .
Symmetrical odds ratios for mobility tables by ethnicity, hukou and ethno-hukou combination.