Attitude toward gender inequality in China

The paper examines attitudes toward gender inequality in China within the context of a long-standing patrilineal tradition that historically prioritized male lineage and confined women to domestic roles. Although socialist-era policies promoted gender equality and women’s labor force participation, subsequent developments (e.g., the one-child policy) revived son preference, contributing to skewed sex ratios, trafficking, and overlooked forms of inequality such as domestic violence. The study’s purposes are: (1) to identify individual-level determinants of attitudes toward gender inequality, including whether patrilineal values are associated with more unequal attitudes; (2) among married households, to assess whether couples’ relative resources (age, education, income) correlate with attitudes; and (3) to compare urban–rural differences in attitudes and decompose the sources of these differences. It also investigates temporal change in attitudes using pooled cross-sectional data.

The paper reviews aggregate- and individual-level determinants of gender attitudes. At the aggregate level, higher national development is associated with more gender equality in education and longevity, partly because less physically intensive production reduces male advantage; cultural systems such as patrilineality correlate with male-skewed sex ratios and unequal resource allocation. At the individual level, prior research links education, income, religiosity, childhood experiences of equality, and parental attitudes to more egalitarian views; attitudes also relate to partnership stability and household divisions of labor. The authors highlight patrilineal values—obedience to paternal authority and preferential inheritance for eldest sons—as potential cultural roots of unequal gender attitudes within China.

Data: Five waves of CGSS (2010, 2012, 2013, 2015, 2017) are pooled; CGSS 2017 includes the EASS 2016 module. Over 58,000 respondents answered five attitudinal items related to gender roles: CAREER (men focus on career; women on family), ABILITY (men more capable), MARRIAGE (a woman’s good marriage is better than managing her career), LAYOFF (women should be laid off first in downturns), and CHORE (husband and wife should share housework equally). Responses are five-point ordinal scales from totally disagree to totally agree (plus small “don’t know/refuse”). For CAREER, ABILITY, MARRIAGE, LAYOFF higher agreement reflects more inequality against women; for CHORE higher agreement indicates more equality.

Measures and coding: For change-over-time figures, items are scored 1–5; for CHORE, higher points reflect agreement; for the other four, higher points reflect disagreement with inequality. For econometric models, the authors reverse signs as needed so positive coefficients uniformly indicate more egalitarian attitudes across items.

Empirical strategy:

• Ordered probit models for each item using pooled data, with controls: gender, age and age squared, ethnicity, religion, education levels, political party membership, household origin (rural/urban/special), log personal and household income, marital status, work status, province/municipality fixed effects, and year dummies (2010 as base).
• Patrilineal values (CGSS 2017 + EASS 2016): two proxies with 7-point responses collapsed to agree, disagree, and neutral (base): (i) obedience to father in all cases; (ii) eldest son’s right to the largest share of wealth. Estimated via ordered probit and Lewbel (2012) heteroskedasticity-based 2SLS to address endogeneity/reverse causality.
• Couples’ relative resources (married subsample): relative income (wife earns more vs base: husband earns more or equal), relative education (wife better educated; husband better educated; base: equal), and relative age (wife older; husband older; base: same age). Estimated via ordered probit; Lewbel 2SLS for potential endogeneity of wife’s higher income and wife’s higher education.
• Urban–rural decomposition: Oaxaca–Blinder decomposition of score differences (higher score = more egalitarian) to split urban–rural gaps into explained (covariates) and unexplained components, using the pooled CGSS sample; repeated with patrilineal variables using the CGSS 2017 + EASS 2016 subsample.

Samples: Main ordered probit observations per item are about 49k; married-subset models around 40k; patrilineal subsample around 3.5–3.6k. Province effects are included; standard errors reported.

• Descriptive attitudes (Table 1, pooled): For CAREER, 60.2% agreed that men focus on career and women on family (including totally agree); for CHORE, 70.7% agreed that couples should share housework equally. For ABILITY, disagreement (43.8%) and agreement (42.1%) were similar. For LAYOFF, 70.0% disagreed that women should be laid off first (12.4% agreed). Women tended to disagree with sexist statements more than men in CAREER, ABILITY, LAYOFF; both genders had sizable agreement with traditional views on some items.
• Trends over time: Average scores show small improvements in egalitarian attitudes for CAREER and MARRIAGE; ABILITY and LAYOFF improved overall with fluctuations; CHORE slightly declined by 2017. Percent positive (egalitarian) attitudes rose for several items in 2017; negative attitudes remained relatively stable.
• Ordered probit (Table 3): Year dummies for CAREER are positive and increasing (vs 2010), indicating rising disagreement with the traditional career/family divide. ABILITY and MARRIAGE show improvements in 2017. LAYOFF fluctuates (negative in 2013–2015, weakly positive in 2017). CHORE coefficients are negative across years, implying no increased expectation of men sharing housework. Women hold more egalitarian attitudes than men across items except MARRIAGE (women slightly more traditional there). Age has a concave relation (older less egalitarian, but diminishing). Education shows a strong gradient: below high school associates with less egalitarian attitudes; some college, university, and graduate degrees associate with more egalitarian attitudes. Higher personal income, retirement, and Communist Party membership correlate with more egalitarian views; rural household origin, being married or widowed, and Christianity or Muslim affiliation (on some items) correlate with less egalitarian views.
• Marginal effects of education (Figures 5–6): Compared to high school, higher education increases the probability of egalitarian responses by roughly 2–3 percentage points and lowers negative responses by about 3 points (varies by item). Upgrading from below high school to high school decreases negative attitudes by ≥4 points and increases positive attitudes by ≥4 points for most items.
• Patrilineal values (Table 4): Agreement with obedience to father and eldest son inheritance is significantly associated with less egalitarian attitudes (notably for CAREER, ABILITY, MARRIAGE). Disagreement with these patrilineal statements is associated with more egalitarian attitudes (e.g., for eldest-son inheritance, disagreement has strong positive associations across items, including LAYOFF and CHORE). Lewbel 2SLS results are consistent with a causal negative effect of patrilineal beliefs on gender equality attitudes and a positive effect of rejecting patrilineal beliefs.
• Couples’ relative resources (Table 5): In correlations, a wife earning more is associated with more egalitarian views for CAREER and MARRIAGE; however, 2SLS indicates no robust causal effect of wife’s higher income. In contrast, a wife being better educated is positively associated with egalitarian attitudes for all items, and 2SLS supports a causal positive effect. A husband being better educated correlates with more traditional views for CAREER and ABILITY. Relative age is not significantly related to attitudes.
• Urban–rural gaps (Table 6): Urban respondents have higher (more egalitarian) scores than rural respondents across items; differences (urban minus rural) are: CAREER 0.5017, ABILITY 0.3886, MARRIAGE 0.2587, LAYOFF 0.2585, CHORE 0.1003 (all p<0.01). Both explained and unexplained components are positive and significant; explained components account for a large share (e.g., CAREER explained 0.3155; unexplained 0.1862). Education levels (low schooling in rural areas vs higher in urban), Communist Party membership, retirement, and income are key contributors to the explained component; age slightly narrows the gap. CHORE’s gap is smaller and less well explained by observables.
• Patrilineal values and urban–rural gaps (Table 7): In the 2017+2016 subsample, urban–rural differences remain significant (e.g., CAREER difference 0.5844). Explained components are sizeable (e.g., CAREER 0.3661). Patrilineal indicators contribute positively to the explained component for CAREER and MARRIAGE (agreement and disagreement categories), and disagreement contributes for LAYOFF and CHORE, consistent with patrilineal beliefs being more prevalent in rural areas and linked to less egalitarian attitudes.
• Overall pattern: Attitudes appear “stuck” between modern and traditional views: women are expected to work more, without corresponding gains in job security or reductions in housework burden.

The findings address the research questions by showing: (1) Individual characteristics—especially education—strongly predict gender attitudes, with higher education linked to more egalitarian views. Women are more egalitarian than men on most items, and higher income and Communist Party membership also relate to egalitarianism. (2) Within married couples, relative resources matter: the wife’s higher education (causally) improves egalitarian attitudes across domains, whereas higher relative income of the wife does not show a robust causal effect. (3) Urban–rural differences in attitudes are statistically significant; a substantial portion is explained by differences in education and other demographics, and patrilineal values account for part of the gap, indicating cultural roots of inequality especially in rural areas. These results suggest policy levers—particularly education expansion and cultural change—can shift attitudes and reduce the attitudinal foundations of gender inequality. However, progress is uneven: expectations for women’s labor force participation increase without parallel change in housework divisions or stronger job security norms for women.

The study contributes by: (a) documenting modest temporal improvements in some egalitarian attitudes in China, especially regarding women’s participation in paid work; (b) identifying education as the most influential determinant of egalitarian gender attitudes at both individual and household (relative education) levels; (c) establishing that patrilineal values conflict with gender equality attitudes and help explain urban–rural gaps; and (d) quantifying urban–rural differences and the substantial role of educational composition in explaining them. Policy implications include expanding access to education, particularly for women and rural populations; leveraging technology (e.g., AI in education) to provide unbiased learning environments; and adopting gender-neutral procedures in selection and promotion to increase women’s participation in male-dominated fields. As economic development and mechanization reduce the premium on physical strength, urban–rural attitudinal gaps may narrow. The authors note that continued openness and exchange with countries promoting women’s rights can further support attitude change.

The study relies on self-reported survey data and is subject to social desirability bias; better-educated respondents or party members may underreport traditional attitudes. Attitudes can be time-specific and influenced by contemporaneous economic conditions (e.g., views on layoffs during downturns). Although Lewbel 2SLS addresses endogeneity for select variables, causal identification may remain imperfect. Measurement issues may affect interpretation of CHORE (e.g., differing understandings of “equal” vs “fair” division). Teachers and local institutions in rural areas may themselves hold traditional biases, potentially limiting the impact of schooling on attitudes; thus, the estimated role of education should not be overstated.