Examining Ethnicity: Patterns of Minority Identification Among Children of Interethnic Marriages in China

Abstract

Ethnic identity is a highly contested issue in China. Yet, the literature on the social construction of ethnoracial identity is dominated by research on the Americas. In this study, we investigate patterns of ethnic identification among children of interethnic parents in China using census data from 2000 and survey data from 2010–2018. We focus on children who are aged 20 or younger and have one parent identified as an ethnic minority and one parent identified as an ethnic Han. We find that the strongest predictor of a child's minority identification is the father's ethnicity. Minority identification is also associated with gender, birth year, mother's education, household income, migration status, parent's perception of the child's diligence, the geographic concentration of minorities, and eligibility for ethnicity-based bonus points on the college entrance examination. Taken together, the results suggest that children's ethnoracial identity is shaped by family demographic characteristics as well as by education policy.

Introduction

The tension between maintaining a distinct ethnic identity and a unified national identity underlies the history of Chinese ethnic minorities. After the People's Republic of China was established in 1949, the state announced plans to improve the economic welfare of ethnic minorities and integrate them into society. However, during the Cultural Revolution, from 1966 to 1977, ethnic differences were considered antithetical to national unity, and minorities were forced to assimilate. By the early 1980s, the state began to promote cultural diversity and develop programs and policies to reduce inequalities between majority and minority populations. Today, anxiety about diversity is rising again, as exemplified by the struggle over ethnic identity in Xinjiang.¹

Thus, ethnic identity has endured as a highly contested issue throughout the recent history of China. Yet, there is surprisingly little quantitative research about the construction of ethnoracial identity. In this article, we quantitatively investigate patterns of ethnic identification among children of interethnic parents in China using census data from 2000 and survey data from 2010–2018. Our study makes three contributions to the literature. First, it provides evidence on the construction of ethnoracial identification in Asia, going beyond the literature's usual focus on the Americas. Second, it offers a window into how public policy influences ethnicity in an extraordinary population that must choose between minority and majority ethnic affiliation to determine the allocation of education benefits. Third, the study advances research on affirmative action in China. To our knowledge, this is the first study to organize and analyze detailed information on ethnicity-based bonus points on the college entrance examination.

We leverage two data sources. First, we use census data, focusing on children who were aged 20 or younger in 2000 and had one parent identified as an ethnic minority and one parent identified as an ethnic Han, since only those with parents of different ethnicities have a choice regarding their state-registered ethnicity. Second, we use survey data from the China Family Panel Studies, focusing on children who were aged 20 or younger in 2010, had one parent identified as an ethnic minority and one parent identified as an ethnic Han, and appeared in any survey wave between 2010 and 2018. Our dependent variable is an indicator for whether a child is identified as a minority, which is consistent with the monoethnic classification system in China. Logistic regressions are employed to estimate the predictors of minority identification.

In summary, we find that the strongest predictor of a child's minority identification is the father's ethnicity. Minority identification is also associated with gender, birth year, mother's education, household income, migration status, parent's perception of child's diligence, and the geographic concentration of minorities. Notably, eligibility for bonus points in college admissions is positively associated with being identified as a minority. Taken together, the results suggest that children's ethnoracial identity is shaped by family demographic characteristics as well as by education policy, which treats majority and minority ethnic groups differently in some circumstances. Eligibility for bonus points on the college entrance examination increases the incentive to identify as a minority because intense competition for college admission makes bonus points extremely valuable.

Background and Previous Literature

The Social Construction of Ethnoracial Identity

Traditional conceptions of race and ethnicity as static and inherent have been challenged by social constructionist theories that illuminate their fluid and dynamic nature (e.g., Barth 1969; Loveman 2014; Marx 1998; Nagel 1994; Wimmer 2008). Constructionist theories are concerned with the determinants of ethnoracial boundaries defining who belongs to an ethnoracial group and who is excluded.

Two research areas are highly relevant to the present study. One has investigated patterns of identification among children with parents from different ethnoracial groups. Although most of these studies have focused on the United States (e.g., Bratter 2007; Bratter and Heard 2009; Brunsma 2005; Davenport 2016; Khanna 2004; Lichter and Qian 2018; Liebler 2010; Qian 2004; Roth 2005; Xie and Goyette 1997), some have examined other contexts, such as New Zealand (e.g., Kukutai 2007) and Brazil (e.g., Marteleto 2012; Schwartzman 2007). Such studies have regressed ethnoracial identification on individual, household, and community factors. Although most studies have relied on parents' classification, a few studies have used self-identification (e.g., Bratter and Heard 2009; Davenport 2016; Khanna 2004). These types of classification represent distinct aspects of race that may be influenced by different social processes.

Three factors emerged as significant predictors of identification across studies: the ethnoracial composition of the local population, parents' socioeconomic status (often proxied by education), and father's ethnoracial group. Specifically, the likelihood of minority identification increases as the local population share of the minority group increases (Bratter 2007; Brunsma 2005; Davenport 2016; Liebler 2010; Qian 2004; Roth 2005; Xie and Goyette 1997). The higher the parents’ education, the lower the likelihood of (monoracial) minority identification (Bratter 2007; Brunsma 2005; Davenport 2016; Lichter and Qian 2018; Roth 2005; Schwartzman 2007); an exception is Marteleto (2012), who found that Preto (Black or dark-skinned) identification rises with parental education in Brazil. Also, having a minority father tends to increase the likelihood of minority identification (Bratter and Heard 2009; Brunsma 2005; Qian 2004; Xie and Goyette 1997). Notably, Bratter and Heard (2009) confirmed this pattern for Americans with Black and White parents but not for Americans with Asian and White parents.

Another highly relevant area of research has investigated the impact of public policy on identification. Scholars have long recognized the state's role in race-making (see Loveman 2014; Marx 1998). Quantitative studies have demonstrated how specific policies influence ethnoracial identification (Antman and Duncan 2015; Bratter and O'Connell 2017; Francis and Tannuri-Pianto 2013; Francis-Tan and Tannuri-Pianto 2015). In the U.S. context, Antman and Duncan (2015) found that multiracial persons who stood to benefit from identifying as minorities became less likely to do so following state bans on affirmative action in college admissions and government hiring, while Bratter and O'Connell (2017) found that children of Black–White intermarriages were less likely to be classified as multiracial in states with historical bans on interracial marriage. In Brazil, Francis and Tannuri-Pianto (2013) and Francis-Tan and Tannuri-Pianto (2015) found that the implementation of quotas for Blacks at a prestigious university inspired shifts from non-Black to Black identity and from lighter to darker racial categories.

However, the literature is almost entirely focused on the Americas. Quantitative research on ethnoracial identification in Asia is rare. Studies on Chinese ethnic minorities have typically focused on their distinct family patterns (e.g., Mu and Lai 2016) or their socioeconomic disparities relative to the Han majority (e.g., Hannum and Xie 1998). Several features of the Chinese context make the quantitative study of ethnic identification both interesting and important.

Ethnic Minorities in China

China is a multiethnic country with 56 officially recognized ethnic groups (see Table A1 in the online appendix for a list of these groups and their population sizes). Although ethnic Han make up the vast majority of the population, the percentage of ethnic minorities has been rising. In 1982, the year in which the system of ethnic classification was finalized, the percentage of Chinese identifying as minorities was 6.7%. Ethnic minorities represented 8.0% of the population in 1990, compared with 8.8% in 2000 and 8.4% in 2010. Ethnic groups display significant heterogeneity in religion, language, and cultural practices (Zang 2015). While some minority groups are quite distinctive from the Han majority, others are almost indistinguishable.

The tension between maintaining a distinct ethnic identity and a unified national identity underlies the history of Chinese ethnic minorities (Gladney 1994). After the founding of the People's Republic of China, the state took action to recognize ethnic minorities, raise their economic status, and integrate them into society (Zhou 2009). In the early 1950s, the state commissioned party cadres and researchers to identify ethnic groups to be classified on the basis of religion, language, and culture (Gladney 2004), but only 10% of the groups that applied for recognition were listed in the 1953 census.

In 1966, Mao Zedong launched the Cultural Revolution, a brutal campaign to consolidate power and impose uniformity throughout China (Meisner 1999), significantly affecting many groups, including ethnic minorities. Ethnic diversity was considered antithetical to the notion of a unified Chinese people (Heberer 1989). Facing the pressure of the state's anti-tradition campaign, many ethnic minorities were forced to assimilate. This period, 1966–1977, was disastrous for ethnic communities because they had to reject their traditions and identities (Dillon 1994).

However, by the early 1980s, the situation improved dramatically when official policy became favorable toward ethnic minorities (Hannum and Xie 1998; Wu and Song 2014). The state began to promote cultural diversity and developed programs to reduce inequalities between the majority and minority populations. Thus, when the ethnic classification system was revised in 1982, many people wanted to officially register their minority identity, motivated by a desire to embrace their ethnic heritage, make use of ethnic policies, or both (Scharping 2003).

Ethnicity is reported on a person's identity card and household registration, or hukou.² Historically, minority identification was recognized for those who could provide evidence of their ethnic heritage. In some cases, this simply entailed proving one's village of birth, given that certain villages were associated with certain ethnic groups. The informality of the system made it easy to obtain minority status, but it also allowed some fraudulent claims (Scharping 2003). By the 1990s, a person could claim minority status only if one or both parents were registered as minorities. Therefore, only those with parents of different ethnicities can choose their ethnicity. Only monoethnic classification is permitted officially; the notion of multiethnic identity is not well established (Lu 2022).

Ethnic minorities have benefited from public policies regarding political representation, fertility, employment, and education (Zang 2015). Perhaps the most salient was the one-child policy (OCP). Most provinces allowed minority couples to have more than one child (Greenhalgh 2008). However, the importance of the OCP declined over time, even before its relaxation in 2013 and termination in 2015, as the level of economic development increased and the demand for children decreased (Zhang 2017). In contrast, the importance of education policies has risen over time because the returns to education have increased markedly since the 1980s (Yang and Wu 2009; Zhao 2010).

Many education initiatives have been enacted for ethnic minorities. At the primary and secondary levels, the government subsidizes the construction and operation of schools in minority areas and offers some Tibetan and Uyghur students the opportunity to study in special classes in Han-dominated schools (Leibold 2016). At the university level, the government supports 12 ethnic minority universities and organizes preparatory courses to help minority students transition to college (Leibold 2016; Sautman 1998; Zhu 2010). Nevertheless, the most widespread policy is the allotment of bonus points on the national college entrance examination, or gaokao (Sautman 1998; Wang 2007). Most provinces, though not all, award bonus points to minorities. We detail these policies in the Methods section.

The Social Construction of Ethnic Identity in China

Qualitative studies have long documented the endogenous boundaries between ethnic groups in China (e.g., Gladney 1994; Harrell 2001; Joniak-Lüthi 2016; Mullaney 2011; Stroup 2017; Wang 2010; Yuan et al. 2014; Zhu 2007). Recent studies have focused on the role of education. For example, Yang (2017, 2020) investigated schooling and the formation of ethnic identity among Tibetan students. Lu (2022) conducted in-depth interviews with 20 Hui–Han biethnic college students. They were asked why their parents registered them as Hui. Respondents mentioned the community's tradition of inheriting their father's ethnicity and their grandparents' desire to transmit religion. But they most frequently mentioned their parents' desire for them to receive bonus points in college admissions. Respondents noted that bonus points raised their self-awareness of being Hui and deepened their ethnic identity.

Little quantitative research investigates the social construction of ethnicity in China. Recent studies suggest that changes in minority populations across census years are explained not only by changes in fertility but also by changes in ethnic identification (Francis-Tan and Mu 2019; Wu and Ingram 2019). Perhaps the clearest evidence of this phenomenon is that all birth cohorts experienced a rise in minority identification across censuses from 1982 to 2000. For example, among those born in 1965, 6.8%, 7.8%, and 8.3% were identified as minorities in 1982, 1990, and 2000, respectively. Francis-Tan and Mu (2019) linked policy incentives to these trends, finding that the OCP inspired an increase in minority identification during the 1980s and 1990s, especially among lower status families.

Research has also explored the predictors of minority identification (Francis-Tan and Mu 2019; Wu and He 2018). Using data from 2005, Wu and He (2018) found that children of more educated, interethnic parents in autonomous provinces are more likely to be identified as minorities. Building on this study, Francis-Tan and Mu (2019) verified the positive association between parents' education and the child's minority identification. They also reported that this pattern has been growing over time, with estimates being two to three times higher for 2000 than for 1982 and 1990.

Data and Methods

Census

In this study, we use two data sets. The first is the 1% sample of the 2000 Chinese Census, made available through IPUMS International (Minnesota Population Center 2018). The main advantage of using this data set is its size, which is important in a study examining minority populations. We focus specifically on children aged 20 or younger in 2000 who have two parents in a first-time marriage, one identified as Han and the other identified as a minority. By age 20, the vast majority of children who lived with both parents at age 5 are unmarried and still living with their parents. Limiting the sample to first-time marriages minimizes the presence of children from blended families who are not explicitly identified as stepchildren in the census. The estimation sample, which is representative of children living with two biological parents of different ethnicities, is composed of 86,762 children living in 58,264 households.

Our dependent variable is an indicator for whether a child was identified as an ethnic minority. For the census, a parent is almost always the one who provides information on the household members. Information on ethnicity was obtained using a closed-ended question with 56 officially recognized ethnic groups as options. This variable is consistent with the monoethnic classification system adopted throughout China, and it mirrors the principal ethnoracial distinction embedded in regulations concerning bonus points in college admissions (i.e., majority vs. minority groups).

We also construct independent variables from the census data. We consider sociodemographic variables that are typically included in studies of China: gender, age, hukou status, and province. Hukou status is categorized as rural (rural hukou and residence), urban (urban hukou and residence), and migrant (hukou and residence differ). We also consider variables that are commonly included in research on children from multiracial families: an indicator for whether the father or mother is identified as a minority, the mother's education, and the percentage of ethnic minorities at the prefecture level (the smallest geographic unit provided in the census data).³ Additionally, we construct an indicator for whether the minority parent's ethnic group has historically been associated with Islam. Given that having a strong religious background provides a foundation for identity (Mu 2021; Van Niekerk and Verkuyten 2018), we would expect a positive association between Islamic background and minority status. In China, 10 ethnic groups (Hui, Uyghur, Kazakh, Dongxiang, Kyrgyz, Salar, Tajik, Uzbek, Bonan, and Tatars) are associated with Islam. Table 1 displays summary statistics for the census variables.

China Family Panel Studies

The second data set that we use is the China Family Panel Studies (CFPS; Institute of Social Science Survey 2015), whose main advantages are its wide range of variables and its recency. The CFPS, designed by researchers at Peking University, is China's first large-scale longitudinal survey focused on family and society. The baseline survey was conducted in 2010 and sampled 25 provinces/municipalities where approximately 95% of the population resides. Follow-up surveys were conducted in 2012, 2014, 2016, and 2018. We focus on children who were aged 20 or younger in 2010, were claimed as children of both parents, have one parent identified as Han and one parent identified as a minority, and appeared in any survey wave between 2010 and 2018. Note that we are unable to leverage the panel structure of the CFPS because, by design, ethnicity does not vary by wave.⁴ Thus, we collapse information across waves so that each child contributes one observation to the estimation sample. The estimation sample, which is representative of children living with two biological parents of different ethnicities, is composed of 699 children living in 487 households.

Our dependent variable is an indicator for whether a child was identified as an ethnic minority. In 83% of cases, a parent is the one who provides the CFPS with information on the child's ethnicity; in 17% of cases, the child provides information on their own ethnicity. The question on ethnicity is asked only the first time a person appears in the survey. In later waves, information on ethnicity is prepopulated. Again, the dependent variable is consistent with China's monoethnic classification system and mirrors the principal ethnoracial distinction embedded in regulations concerning bonus points in college admissions.

Independent variables are constructed from the CFPS data. We consider sociodemographic variables that are typically included in studies of China: gender, birth year, hukou status, and province. Hukou status and province are calculated with the earliest available information. We also consider variables that are commonly included in research on children from multiracial families: an indicator for whether the father or mother is identified as a minority, mother's education (median across waves), and the percentage of ethnic minorities at the county level in 2010. County (CFPS version of the variable) is a finer geographic division than prefecture (the census version of the variable). As we did with the census, we also include an indicator for whether the minority parent's ethnic group has historically been associated with Islam. Additionally, we consider variables that are unique to the CFPS: a proxy for household income, an indicator for whether the father was a member of the Communist Party in any wave, parent's perceived diligence of the child, and parent's educational aspirations for the child.⁵

To obtain a proxy for household income, we calculate the decile of net household income by wave for each child, calculate the mean of deciles across waves for each child, and categorize the average decile into quantiles. One might expect that money would “whiten” (Schwartzman 2007), such that higher household income would imply a greater likelihood that children are classified as Han. Yet, it is also possible that minority status would be more likely at higher levels of income if having greater financial resources fosters college readiness or enables families to embrace their ethnic identities. Communist Party membership is known to confer economic benefits, which may increase the chances of college attendance. We hypothesize that party membership is positively associated with identifying one's child as a minority because embracing minority status may be advantageous for gaining party membership.

Parents are asked about perceived diligence and educational aspirations only if their children are aged 15 or younger; these variables are the means across waves.⁶ Perceived diligence is a composite of seven responses (e.g., my child studies very hard), where 1 equals strongly disagree and 5 equals strongly agree. The literature has documented both negative and positive associations between a child's minority status and their perceived diligence (Blanchard and Muller 2015; Reyna 2000, 2008). It is possible that children who are more diligent are less likely to be seen as minorities or that children who are minorities are less likely to be perceived as diligent. However, we do not believe that parents would strongly exhibit such tendencies. Thus, we hypothesize a positive association between perceived diligence and minority identification, since diligent children are expected to be college bound and minority status may yield positive returns to children in college and employment. Educational aspirations are measured by a parent's response to “the highest level of education that you wish your child can obtain,” where the response units are years of schooling. Approximately 75% of parents indicated a desire for their children to obtain a college degree. We expect that educational aspirations will be positively associated with minority identification. Table 2 displays summary statistics for the CFPS variables.

Bonus Points on the Gaokao

We began compiling information about bonus points on the national college entrance examination from government and nongovernment websites in 2019. The information gathered reflects the policies circa 2018, at the end of our sample period. Ideally, we would have the time series of policies for the entire period, but this was not possible.

Nevertheless, it is useful to get a sense of the evolution of these benefits. In 1977, the gaokao was reinstated after being suspended during the Cultural Revolution. In 1980, the Ministry of Education issued regulations promoting the establishment of preferences for ethnic minorities in university admissions (Wang 2007). Thus, even the oldest cohorts in the data were not yet enrolled in school when the bonus point system was established. The policies were relatively stable over time. The most common type of change was to moderate the number of bonus points awarded to ethnic minorities.⁷

Therefore, even if the precise number of points is unknown, our information reflects which groups were eligible for bonus points during the study period. In any case, we do not know exactly when parents made decisions regarding their children's ethnicity. Parents are permitted to change their child's official ethnic designation until they reach adulthood.

Table 3 provides a summary of the policies; Table A2 in the online appendix provides detailed policy information. Among the provinces, 26 offer bonus points to (at least some) ethnic minorities, and five do not. Eligibility for bonus points is jointly determined by province, ethnic group, and geography. For example, Jilin gives 5 points to Mongol, Manchu, and Korean minorities living in autonomous counties; Jiangsu gives 3 points to all minorities; and Shandong gives 5 points to all minorities living in minority areas, which typically include autonomous prefectures, autonomous counties, ethnic townships, and other minority concentrated areas designated by the government.

On the basis of the information in Table 3, we calculate for each child in our estimation sample whether they would be eligible to receive more bonus points on the gaokao as a minority than as a Han. This binary variable measures the extensive margin. Bonus points are so valuable that whether there are bonus points is more relevant to decisions regarding minority identification than how many points there are. Overall, 56% of children in the census estimation sample and 50% of children in the CFPS estimation sample were eligible for bonus points on the gaokao.

To calculate eligibility, we assume that each province's policy applies only to persons whose hukou registration was in the province and that tiebreaker benefits are negligible.⁸ Most students attend universities in their province of hukou registration. Coding the ethnic group restrictions is straightforward, but coding the geographic restrictions is not. Using the census, we define autonomous counties and minority areas as prefectures where ethnic minorities account for a majority of the population. Using the CFPS, we define autonomous counties as counties designated as ethnic autonomous counties and minority areas as autonomous counties plus villages/urban neighborhoods reported to be ethnic minority areas (based on the CFPS community files).

Statistical Models

In the main analysis, we investigate the predictors of ethnic identification among children of interethnic marriages. Logistic regressions are employed to estimate the statistical association between minority identification and the set of independent variables. The following basic model is implemented for child i in geographic area j:

where Y is minority identification, X is a vector of individual-level predictors, E is eligibility for bonus points on the gaokao, and M is the percentage of ethnic minorities in the area. The index j corresponds to prefectures in the census sample and to counties in the CFPS sample. Adjusted odds ratios are reported in the tables. Standard errors are adjusted for clustering on household because some households have multiple children.⁹

Results

Evidence From the Census (2000)

Table 4 displays descriptive statistics from the census for children with two Han parents, one Han parent, and no Han parents. Approximately 3% of children living with two biological parents have interethnic parents. Relative to children with two Han parents, those with one Han parent are more likely to have parents with a low or high level of education, reside in an urban area, and live in a prefecture with an elevated concentration of minorities. Children with no Han parents are more likely to have a sibling, have parents with a low level of education, reside in a rural area, and live in a prefecture with an elevated concentration of minorities.

Table 5 displays multivariate logistic regressions of ethnic identification based on the census sample. The estimated coefficients are adjusted odds ratios. Province fixed effects are also included in Models 3 and 4. Unless otherwise noted, we use Model 4 to provide numerical estimates in the text.

The table shows that girls have a higher odds of minority identification, as the coefficients are significant in all four models. Specifically, girls have a 5.5% higher odds of being identified as a minority. Moreover, the odds rise uniformly with the child's age. Relative to children aged 0–3, children aged 8–11, 12–14, and 15–20 have a 10.7%, 21.1%, and 24.2% higher odds of minority identification, respectively. The strongest association, by far, is that with father's ethnicity. The odds of being identified as a minority are approximately 12 times higher when a child's father is an ethnic minority. Minority identification is also positively associated with Islam. Once province is controlled for in Models 3 and 4, a child whose parent belongs to an ethnic group historically associated with Islam has twice the odds of minority identification. Additionally, the odds increase with the mother's education level. For example, relative to a child whose mother completed no schooling, a child whose mother completed middle school has about 1.6 times the odds of being identified as a minority, whereas a child whose mother completed some college has about 2.9 times the odds.

Although urban resident children are slightly less likely than rural resident children to be identified as a minority, the difference is not statistically significant. Migrant children have 18.5% higher odds of minority identification. The coefficient on the percentage of ethnic minorities at the prefecture level is significant in both Model 3 and Model 4. As the concentration of minorities increases by 10 percentage points, the odds of minority identification increase by 9%. Finally, the results indicate that eligibility for bonus points on the gaokao increases the likelihood of minority identification. Children who are eligible for bonus points have nearly 2.1 times the odds of being identified as a minority.

Evidence From the CFPS (2010–2018)

Table 6 displays multivariate logistic regressions of ethnic identification based on the CFPS sample. As with the census results, the estimated coefficients are adjusted odds ratios, and province fixed effects are included in Models 3 and 4. However, the sample size is less than 1% of the census sample size. Unless otherwise noted, we use Model 4 to provide numerical estimates in the text.

The results indicate that minority identification is more likely for girls than for boys, although the difference is not statistically significant. The odds of minority identification increase with birth year—opposite the pattern found in the census data. Relative to children born during 1998–2002, children born during 2009–2012 and 2013–2018 have 2.4 and 1.6 times the odds of being identified as a minority, respectively. As with the census data, the strongest association is that with father's ethnicity. The odds of being identified as a minority are almost 6 times higher when a child's father is an ethnic minority. Minority identification is not significantly associated with Islam, but it is significantly associated with mother's education. Relative to a child whose mother completed primary school, a child whose mother completed high school has almost 2.0 times the odds of minority identification, and a child whose mother completed some college or more has about 3.5 times the odds.

Whether a child's father is a member of the Communist Party, a variable unique to the CFPS, is not significant. Another unique variable, household income, is significant. Children who belong to the poorest households have 2.1 times the odds of minority identification, and those who belong to the richest households have 4.0 times the odds, relative to children who belong to households in the fourth quantile. As with the census data, the coefficient for migrant children is greater than 1, but here the coefficient is not significant. The coefficient on the percentage of ethnic minorities at the county level is significant in Model 3, and its size is comparable to previous estimates. According to Model 3, as the concentration of minorities increases by 10 percentage points, the odds of being identified as a minority increase by 16%. Also consistent with the previous results, eligibility for bonus points on the gaokao is associated with a higher odds of minority identification. Those children who are eligible for bonus points have approximately 2.8 times the odds of minority identification.

Table 7 displays multivariate logistic regressions of ethnic identification, adding perceived diligence of the child and the parent's educational aspirations for the child as predictors. The sample size declines because these variables are missing for some children who were infants or were over age 15 during the sample period. The estimated coefficients are adjusted odds ratios, and province fixed effects are included in all models. Unless otherwise noted, we use Model 4 to provide numerical estimates in the text.

In every column of the table, the association between perceived diligence and minority identification is negative and significant. A 1-standard-deviation (0.51) increase in perceived diligence is associated with roughly a 36%¹⁰ decrease in the odds of being identified as a minority. The association between the parent's educational aspirations and minority identification is positive but not significant. Statistical variation in this variable is limited because most parents hope that their children will attend college.

Comparing Table 7 with Table 6 is illuminating. Girls are more likely than boys to be identified as minorities, and now the difference is at least marginally significant in three of four models. As in Table 6, minority identification is significantly associated with birth year, father's ethnicity, and mother's education. Children born during 2009–2012 have a higher odds of being identified as a minority than children born during 1998–2002. The odds are approximately 11 times higher when a child's father is an ethnic minority. The odds of minority identification are about 2.4 times higher when a child's mother completed high school and about 2.9 times higher when his or her mother completed some college or more, relative to children whose mother completed primary school. As in Table 6, the coefficients on Islam and Communist Party member are not significant. Household income is significant, and consistent with the previous results, children who belong to the poorest households and those who belong to the richest households have the highest odds of minority identification. Both migrant status and the percentage of minorities at the county level are positively associated with being identified as a minority, and their respective coefficients are significant in some of the four models. Notably, the association between minority identification and eligibility for bonus points on the gaokao is stronger in Table 7 than in Table 6. Eligible children have 6.7 times the odds of being identified as a minority when controls for perceived diligence and educational aspirations are included in Model 4.

Discussion

In quantitatively investigating patterns of ethnic identification among children of interethnic parents in China, this study makes several contributions to the literature. First, it adds to the evidence on the construction of ethnoracial identification in Asia, going beyond the quantitative literature's dominant focus on the Americas. Second, it offers a window into how public policy influences ethnicity in an extraordinary population that must choose between minority and majority ethnic affiliation to determine the allocation of education benefits. Third, the study advances research on affirmative action in China. To our knowledge, this is the first study to organize and analyze detailed information on bonus points on the college entrance examination.

Three variables commonly included in the study of China emerge as predictors: gender, birth year, and hukou status. Girls tend to have a higher odds of minority identification than boys. In analyses of American populations, gender is included in regressions but often yields insignificant or mixed results (e.g., Bratter 2007; Khanna 2004; Xie and Goyette 1997). One exception is Davenport (2016), who found that biracial women are much more likely than biracial men to identify as multiracial. The interactive model of identity, which emphasizes the complementary nature of marginalized identities (see Davenport 2016), may help to explain our results on gender. Women, who are marginalized because of preferences for male children in Chinese families, may be more likely to be assigned a marginalized ethnicity by their parents.

The only conflicting results across our data sets are with respect to birth year, which is negatively associated with minority identification in the census but positively associated in the CFPS. The literature is not informative regarding age or birth cohort, as few systematic patterns emerge. The census results, which may be more reliable given the larger sample size, may suggest that children are more likely to be identified as minorities as they approach college age. But our findings may be only an apparent contradiction and may suggest that the incentives to identify as a minority began to rise in the late 1990s and early 2000s, a period when returns to college education were increasing rapidly in China.

We also find that migrant children tend to have a higher odds of minority identification. The only other evidence on hukou comes from Wu and He (2018), who reported that urban hukou is positively associated with minority identity for children in autonomous regions. Perhaps minorities who tend to live in less privileged living conditions are more likely to migrate to places with better living conditions. Alternatively, perhaps rural migrant families in urban areas perceive themselves as marginalized and are more likely to identify their children as marginalized ethnically; that is, a sense of geographic dislocation may amplify feelings of ethnic dislocation.

Variables often included in research on children from multiracial families are also predictors: father's ethnicity, mother's education, and the geographic concentration of ethnic minorities. In our study, the strongest predictor of a child's minority identification is the father's minority identification, in line with previous research on China (Francis-Tan and Mu 2019; Wu and He 2018). This result is also consistent with U.S. studies finding that having a minority father increases the likelihood of minority identification (Brunsma 2005; Qian 2004; Xie and Goyette 1997).¹¹ Nevertheless, the association with father's ethnicity is much larger in the Chinese setting. Patriarchal norms remain prevalent in China (Li et al. 2011; Mu and Xie 2016). In her qualitative work, Lu (2022) reported that Hui–Han respondents partly explained their minority identification by appealing to the traditional practice of inheriting the father's ethnicity.

Minority identification is positively associated with the mother's level of education in both of our data sets. Even though this result affirms and extends work on China (Francis-Tan and Mu 2019; Wu and He 2018), it appears to run counter to most of the literature on the Americas, which has found that monoracial minority identification varies inversely with parents' level of education (Bratter 2007; Brunsma 2005; Davenport 2016; Lichter and Qian 2018; Roth 2005; Schwartzman 2007). However, the difference in results might be explained by the fact that respondents are unable to choose a multiracial category or multiple ethnic categories in the Chinese surveys. In the United States, some educated parents favor a multiracial classification to describe their children rather than a monoracial White or minority classification (Bratter 2007; Roth 2005). Consistent with our findings, Marteleto (2012) reported that Black identification varies positively with parental education in Brazil, which might be related to the adoption of race-based affirmative action in higher education. Analogously, our results for mother's education may reflect the desire to take advantage of education benefits among children from households with high socioeconomic status. Another possibility is that parents with more education may be more likely to embrace minority identities because they have been exposed to the discourse of multiculturalism and minority rights in school (see Howard 2001; Paschel 2016; Telles and Paschel 2014).

The percentage of ethnic minorities at the community level is positively correlated with the likelihood of minority identification. In the literature, the local population share of the minority group is perhaps the most common predictor of minority identification (Bratter 2007; Brunsma 2005; Davenport 2016; Liebler 2010; Qian 2004; Roth 2005; Xie and Goyette 1997). The binary indicator that Wu and He (2018) included for autonomous regions, which have a higher concentration of minorities, was likewise positive and significant. Across these cases, the underlying mechanism is undoubtedly related to the degree of contact with persons identifying as ethnic minorities. Minority identification tends to rise with exposure to minority culture, language, and traditions; social interactions with peers; and information about benefits for ethnic minorities.

Furthermore, variables that are relatively unique to our study emerge as predictors: ethnic group associated with Islam, household income, perceived diligence of the child, and eligibility for bonus points on the gaokao. In the census sample, having a parent whose ethnic group has historically been associated with Islam increases the odds of being identified as a minority. This variable is not significant in the CFPS, likely because only 25 children had such a parent. Although no studies have considered this variable, some studies have explored the heterogeneous effects of specific ethnoracial groups (Davenport 2016; Roth 2005; Xie and Goyette 1997). In our sample, more than 90% of parents belonging to an ethnic group associated with Islam identify as Hui, who are known to have distinct cultural practices and salient Islamic identities (Mu and Lai 2016). The Hui–Han respondents that Lu (2022) interviewed mentioned that their decision to identify as Hui was partly influenced by their grandparents' desire to transmit religion.

In the CFPS sample, we find that children in the poorest households and those in the richest households have the highest likelihood of minority identification. Most studies have not included household income as a predictor due to data unavailability. Nevertheless, Davenport (2016) found that biracial Americans living in wealthy households are more likely to identify as White. Perhaps biracial persons are able to cross racial boundaries as they obtain traits (here, income) associated with one group or another. Thus, low income may pose a barrier to boundary crossing and may explain why the poorest children in our sample are less likely to be classified as Han. However, boundary crossing cannot explain why the richest children in our sample are also less likely to be classified as Han. Similar to our argument concerning mother's education, the desire for children from higher status families to gain admission to college may drive these results. It is also possible that more financially established families can afford to fully embrace their ethnic identities.

Perceived diligence of the child is negatively associated with minority identification in the CFPS sample, which is congruent with several interpretations. It might be that parents are more likely to classify less diligent children as minorities, that children identified as minorities are perceived to be less diligent, or that a third variable is positively correlated with diligence and Han identification. By taking advantage of heterogeneity in minority status among households with multiple children, we can investigate these hypotheses further. Among siblings with discordant minority status, diligence is positively associated with minority identification. In most cases, the child identified as a minority was rated more diligent than the child identified as a Han. This evidence casts doubt on the first two interpretations. Therefore, what more likely explains the negative association in the full sample is a third variable (e.g., social capital) that is correlated with both diligence and minority status.

Finally, our analysis indicates that eligibility for bonus points on the gaokao is positively associated with identifying a child as a minority. These and other results highlight the role of education policy in shaping an individual's identity. Findings on eligibility for bonus points may be interpreted as direct effects. Jointly determined by one's province, area of residence, and ethnic group, eligibility substantially raises the incentives for minority identification because bonus points are extremely valuable given the intense competition for college admission and high returns to college attendance. Findings on mother's education and household income may be interpreted as indirect effects. Children for whom a college education is attainable—those from educated, high-income households—are more likely to be identified as ethnic minorities, despite their ability to identify as Han.

All in all, our results justify the qualitative literature's recent focus on the formation of ethnic identity in school settings (Lu 2022; Yang 2017, 2020). The issue of bonus points arose in research by Lu (2022) when her college informants mentioned bonus points as one of the main reasons their parents registered them as minorities. Moreover, our research sheds light on the history of ethnic policy in China. During the 1980s and 1990s, changes in ethnic identification were mostly attributable to the OCP (Francis-Tan and Mu 2019). However, during the 2000s and thereafter, changes in ethnic identification were increasingly attributable to education policy. The results also illustrate the process by which the state makes race (Loveman 2014; Marx 1998; Wimmer 2008), therefore complementing other quantitative studies (Antman and Duncan 2015; Bratter and O'Connell 2017; Francis and Tannuri-Pianto 2013; Francis-Tan and Tannuri-Pianto 2015). Similar to how racial quotas in college admissions inspire shifts from non-Black to Black identity in Brazil (Francis and Tannuri-Pianto 2013), bonus points in college admissions inspire shifts from Han to minority identity in China.

It is important to recognize the limitations of our research. Even if changes in ethnic designation on a survey reflect changes in official records, they are not necessarily accompanied by changes in behaviors or practices. Ethnographic research is needed to determine whether the ethnic fluidity exhibited among children of interethnic parents is instrumental or culturally meaningful. Although we utilize multiple data sources, the estimation samples are cross-sectional. Hence, our estimates are largely correlational. Another limitation is that both of our samples are only representative of children living with two biological parents. However, this household type is the most prevalent in China, as single-parent, cohabiting-parent, and blended households are relatively uncommon. Additionally, despite our efforts, many details about bonus points on the gaokao are unknown, and little is understood regarding how families interpret such policies at the micro level. Much remains to be learned about patterns of ethnoracial identity in Asia. Future research can explore causal effects, examine other types of policies, and investigate contexts beyond Mainland China. Understanding more about the determinants of identity is not only of academic value but also of practical value. In China and elsewhere, who gets to shape ethnoracial identity and how it gets shaped are highly contested issues.

Acknowledgments

We appreciate the insightful comments that we have received from the Editor and anonymous reviewers.

Notes

References