Shaping Gender Role Attitudes: Intergenerational Impacts of Parental Occupational Differences during Adolescence

Abstract

This paper examines how the relative occupational status of parents shapes individuals’ gender role attitudes. By leveraging the State-Owned Enterprise reform in China during the 1990s as a source of exogenous variation in occupational status, we employ an instrumental variable approach to identify the consequential impacts. Our findings reveal that improved occupational status of mothers within the family is associated with the development of more egalitarian gender role attitudes among their children. Moreover, for these children, their marriage matching patterns in terms of occupational status are strongly shaped, as indicated by the increased differences in occupational status between wives and husbands. We further provide evidence on both the instillation and internalization channels. Overall, this study underscores the importance of family socialization in fostering egalitarian gender role attitudes among the next generation.

Full text

Shaping Gender Role Attitudes: Intergenerational Impacts of Parental
Occupational Differences during Adolescence
*
Shu Cai, Wei Luo, and Zheng Zhong
†
Abstract
This paper examines how the relative occupational status of parents shapes individuals’ gender role
attitudes. By leveraging the State-Owned Enterprise reform in China during the 1990s as a source of
exogenous variation in occupational status, we employ an instrumental variable approach to identify
the consequential impacts. Our findings reveal that improved occupational status of mothers within the
family is associated with the development of more egalitarian gender role attitudes among their children.
Moreover, for these children, their marriage matching patterns in terms of occupational status are
strongly shaped, as indicated by the increased differences in occupational status between wives and
husbands. We further provide evidence on both the instillation and internalization channels. Overall,
this study underscores the importance of family socialization in fostering egalitarian gender role
attitudes among the next generation.
Keywords: Gender Role Attitudes, Occupational Status, Sate-owned Enterprises
JEL Classification: J16, J24, D10
*
We thank Gaojie Tang, Xin Meng, Xiaoyue Shan, Klaus F. Zimmermann, Sharon Xuejing Zuo, and the seminar participants
at Jinan University, the 7th Annual Meeting of China Labor Economics Forum, the 20th Seminar of Chinese Women
Economists, the 6th Renmin University of China & GLO Conference, and the 6th IESR-GLO Joint Workshop for many
constructive comments. Shu Cai acknowledges financial support from China Natural Science Foundation (Project No.
72173056) and the Fundamental Research Funds for the Central Universities (Project No. 23JNQMX29). All errors are our
own. We have no conflicts of interest to declare.
†
Shu Cai, Institute for Economic and Social Research, Jinan University, shucai.ccer@gmail.com; Wei Luo, School of
economics, Jinan University, weiluo.natalie@outlook.com; Zheng Zhong, Institute for Economic and Social Research, Jinan
University, zhongzheng_327@163.com.

1
1 Introduction
Gender inequality remains a pressing issue worldwide, with far-reaching impacts on women’s
opportunities and life-long welfare. The inequality manifests persistent gender disparities in educational
attainment, labor market outcomes, political representation, and decision-making within the family.
1
The hurdles to addressing these inequalities are particularly daunting for developing countries, where,
despite economic progress, signs of diminishing disparities remain limited. Scholars have recognized
the continued influence of social norms and gender roles in perpetuating gender inequality (Alesina et
al., 2013; Jayachandran, 2015, 2021). As social and cultural expectations regarding behavior based on
gender, gender roles significantly constrain women’s choices and prospects (Blau et al., 2011), further
exacerbating this persistent challenge.
The examination of how attitudes toward gender roles are shaped has garnered growing attention,
encompassing investigations into historical roots (Alesina et al., 2013; Grosjean and Khattar, 2019) and
the influence of secondary socialization agents such as communities, schools, and workplaces
(Beanman et al., 2009; Cavapozzi et al., 2021; Dhar et al., 2022). Recent studies have shed light on the
significant role of family (primary) socialization in shaping individuals’ values, beliefs, and behaviors
through interactions within their family environment (Bisin and Verdier, 2000, 2001). Adolescence
emerges as a crucial period for gender-related constructs, such as gender role attitudes, as both girls and
boys confront heightened societal pressures to adopt culturally sanctioned gender-role identities during
this developmental stage (Hill and Lynch, 1983; Olivetti et al., 2021).
2
Changes that occur in families
during adolescence can deeply affect individuals’ views of gender roles.
In this paper, we empirically assess the role of family socialization on the formation of more
egalitarian gender role attitudes, a pursuit of great significance given the persistent gender inequality
and traditional gender role attitudes in East Asian societies.
3
In particular, our focus centers on
understanding how individuals’ attitudes toward gender roles are influenced by the relative occupational
1
For instance, according to the International Labour Organization, as of 2022, the female labor force participation rate was
just under 47%. For men, it was 72%.
2
This phenomenon is identified as the “gender intensification hypothesis,” which was initially proposed by Hill and Lynch
(1983) and subsequently supported by studies in both psychology and economics (Eagly and Wood, 2012; John et al., 2017;
Dhar et al., 2022; Ullrich et al., 2022).
3
The latest data from the 2022 Global Gender Gap Index by the World Economic Forum reveals sobering statistics, with
China ranking 102nd, South Korea 99th, and Japan 116th among the 146 countries assessed (World Economic Forum,
2022).This ranking reflects the gender gap in various areas, including economic participation and opportunity, educational
attainment, health and survival, and political empowerment.

2
status of their parents during adolescence in China. The pivotal role of occupational status as a robust
indicator of intra-household bargaining power in the Chinese context is underscored by its reflection of
intricate interactions across economic, cultural, and social dimensions.
4
Additionally, our study
benefits from the unique context of China’s State-Owned Enterprise (SOE, henceforth) reform starting
from the mid-1990s. This reform led to large-scale negative impacts on local labor markets which
disproportionately affected female workers (Zuo, 2016; Xie et al., 2022). Capitalizing on this distinctive
backdrop of SOE reform, we leverage exogenous changes in relative occupational status between
parents based on their exposure to local SOE reform. Specifically, we construct an instrumental variable
(IV) for parental occupational differences by combining the information on cohort-specific local SOE
reform intensity and the age of parents.
Using data from the Chinese General Social Survey (CGSS), we first investigate the relationship
between the improved occupational status of the mother relative to the father within the family during
adolescence and the development of individuals’ gender role attitudes, particularly in the domain of
housework division. Our analyses consistently show that when there is a narrowed difference in
occupational status between mothers and fathers, their children tend to develop more egalitarian gender
role attitudes. We used both ordinary least squares (OLS) and two-stage least square (2SLS)
specifications, as well as various measures of occupational status, to arrive at this conclusion. Our
preferred 2SLS estimates indicate that a one standard deviation increase in parental occupational
difference would result in an increase in their children’s gender role attitudes pertaining to housework
division by 40.3% of the standard deviation.
We conduct a comprehensive set of sensitivity analyses to probe the robustness of our findings.
The findings persist even after accounting for parental differences in years of education, age, and
mother’s income share, imputed based on the 2005 population census data. In addition, we test the
validity of our instrumental variable and show that the results remain robust after controlling for parental
occupational segregation, divorce risk, and the presence of siblings. The inclusion of possible
confounding policies and shocks, such as higher education system expansion in 1999, the 1997 Asian
financial crisis, and China’s WTO accession, does not alter our main conclusions. Sensitivity checks on
migration, missing variables of parental age, and absolute parental occupational status also confirm the
4
A c omprehensive dis cus sio n o f occupational st atu s a nd its measureme nt is pro vided in Section 3.

3
robustness of our results. Furthermore, to strengthen the validity of our findings, we use another
independent dataset, the China Family Panel Studies (CFPS) data, with similar IV specifications. Our
consistent findings on gender role attitudes related to the distribution of household chores further
reinforce our baseline results.
We extend our analysis beyond gender role attitudes by investigating individuals’ behaviors in the
marriage market, with a focus on their marriage formation and marriage matching patterns. Gender role
attitudes may reflect individuals’ beliefs and expectations about the roles and responsibilities of men
and women within a relationship, which in turn shapes their behaviors in the context of marriage. Our
findings show that the improved occupational status of mothers relative to fathers within the family
plays a strong role in shaping the marriage matching patterns of their children, in terms of occupational
status. Specifically, we observe an increase in the differences in occupational status between wives and
husbands in the next generation, indicating an improvement in the occupational status of wives within
the family. The effect is particularly pronounced for their male children. We do not find any evidence
of the selection for marriage formation
To better understand the mechanisms through which egalitarian gender role attitudes are developed
within families, we explore two possible channels. First, such attitudes can be instilled by mothers who
have greater intra-household bargaining power due to higher occupational status within the family—a
phenomenon we refer to as the “instillation” channel. Second, children can learn and internalize
egalitarian gender role attitudes through their perception of the distribution of household chores
between their parents, a pathway we term the “internalization” channel. Analyzing the CFPS data, we
find compelling evidence that supports both channels.
Our paper aims to introduce a fresh perspective to the vast and continually expanding literature on
the formation of gender role attitudes. While multiple studies have explored the historical roots of
gender norms and attitudes, including ploughing (Alesina et al., 2013), irrigation (Fredriksson and
Gupta, 2023), and historical demographic shocks (Grosjean and Khattar, 2019; Teso, 2019), recent
research has explored short-term factors from such as schooling and community settings. For instance,
Dhar et al. (2022) and Beanman et al. (2009) have investigated the impact of classroom discussions and
exposure to community female leaders on gender norms, respectively. In the context of China, Du et al.
(2020) used Compulsory Education Law as an instrumental variable to demonstrate that increased
schooling leads to more egalitarian gender attitudes. However, Si (2022) found that college expansion

4
in China negatively impacted women’s progressive views on gender roles due to deteriorating labor
market opportunities. Acknowledging the significance of secondary socialization agents like schools,
neighborhoods, workplaces, and peers in shaping values (Cavapozzi et al., 2021; Olivetti et al., 2021;
Cortés et al., 2022), we present a novel perspective that highlights the critical role of the family
environment or primary socialization during adolescence in shaping gender role attitudes. We
demonstrate that when occupational status between fathers and mothers within the family during
adolescence is narrowed, it leads significantly to the development of more egalitarian gender role
attitudes among their children.
Our paper enriches the growing body of literature on the relative economic or social status between
women and men. In the pioneering study, Bertrand et al. (2015) found that an increase in females’
income relative to that of males lowers the likelihood of marriage and reduces the labor force
participation of females. Subsequent studies explored the demographic consequences of the shift in the
relative economic stature of young men compared to that of young women as a result of increasing
import competition (Autor et al., 2019), task-based demand shifts (Shenhav, 2021), and the rise of
robots (Anelli et al., 2021), with a particular focus on demographic outcomes (e.g., marriage formation,
matching, and fertility behaviors). We add to this literature by emphasizing the intergenerational
impacts of the relative occupational status between parents which serves as a strong indicator of intra-
household bargaining power in the Chinese context. Specifically, we find that when the occupational
status of mothers increases relative to fathers, their children are more likely to develop more egalitarian
gender role attitudes. This effect also carries over to the behaviors exhibited by these children in the
marriage market and shapes marriage matching patterns among couples in the next generation in terms
of their occupational status.
We also contribute to the emerging studies on the intergenerational transmission of culture,
preferences, and attitudes. Previous studies have shown that cultures, preferences, and attitudes can be
persistent across generations (Alan et al., 2017) and influence the labor market and demographic
behaviors of offspring (Fernández and Fogli, 2009; Bertrand et al., 2021). Early work has examined the
influence of a society’s macro environment on these transmissions (Andersen et al., 2013; Giuliano and
Nunn, 2021). More recent works explore the role of family socialization in shaping preferences and
attitudes (Bisin and Verdier, 2000, 2001), with a special highlight on parents’ influence (Doepke and
Zilibotti, 2008; Dohmen et al., 2013; Zumbuehl et al., 2021). For instance, Fernández et al. (2004) show

5
that the wives of men whose mothers worked are themselves significantly more likely to work, while
Kawaguchi and Miyazaki (2009) find that men raised by full-time working mothers are less likely to
support traditional gender roles. Building on these research, we provide supportive evidence of the
crucial role of family in shaping gender role attitudes. What sets us apart from previous studies is our
exploration of two potential channels. Specifically, the findings reveal that gender role attitudes can be
1) instilled by mothers who have higher occupational status within the family and possess greater intra-
household bargaining power; and 2) learned and internalized by children from their observations of
household division between parents.
The remainder of the paper is organized as follows. Section 2 introduces the institutional
background. Section 3 describes the data and measurement. Section 4 presents the empirical strategy.
Section 5 reports the results on gender role attitudes. Section 6 extends the analysis and examines the
impacts on the marital behaviors of the next generation. Section 7 discusses the potential mechanisms.
Section 8 concludes the paper.
2 Institutional Background
This section describes the institutional background of China’s SOE reform. China launched a
nationwide restructuring of the State-owned sector in the mid-1990s in response to the economic
downturn and to mitigate the substantial financial losses incurred by State-Owned Enterprises (SOEs).
In 1994, the Central Economic Work Conference proposed to deepen the reform of the economic system
targeting SOEs. However, due to the significant impact of SOEs on people’s lives and social stability,
the central government did not endorse the SOE reform until the 15th Communist Party Congress in
September 1997 (Frazier, 2006). At this Congress, the “Three-year Target” was set, aiming specifically
to merge, privatize, and close the majority of medium-to-small enterprises within three years. This
target aims to alleviate the financial burden and revitalize the majority of large and medium-sized SOEs
that were experiencing losses. In 2000, at the Third Session of the Ninth National People’s Congress,
the government announced that the “Three-year Target” was reached. The number of laid-off SOE
workers began to decrease gradually (Liu and Zhao, 2014). In 2005, the re-employment service center
closed, marking the end of the SOE reform. Figure A1 in the Appendix outlines the timing of China’s
SOE reform.

6
The SOE reform had a significant impact on the labor market. This reform initiative involved a
massive wave of layoffs, affecting 35 million SOE workers (World Bank, 2007). Appendix Figure A2
showcases the ratio of laid-off workers between 1998 and 2001, categorized by gender, using data from
the China Labor Statistics Yearbook. During this period, the laid-off ratio for females was 7.0%, 7.4%,
7.6%, and 5.9%, whereas for males, it stood at 4.6%, 5.2%, 5.2%, and 4.0%. It is worth highlighting
that the SOE reform impacted female workers significantly more than male workers, which is consistent
with the finding of Zuo (2016).
Appendix Figure A3 displays geographic variations in the average ratio of laid-off SOE workers
during the reform period (i.e., between 1998 and 2001) by gender. Provinces in north-eastern China,
including Heilong Jiang, Jilin, and Liaoning, as well as central provinces such as Hunan, Jiangxi, and
Hubei, were particularly impacted by the SOE reform. This is consistent with the fact that these areas
gathered a large number of heavy industry, which was usually dominated by SOEs.
The restructuring of SOEs had wide-ranging labor market consequences, not just confined to the
boundaries of SOEs themselves. The impact of the reform extended to local non-SOE firms, including
private enterprises and Township and Village Enterprises (TVEs). These firms experienced ripple
effects as they absorbed laid-off SOE workers and underwent labor reallocation (Lam et al., 2017).
Furthermore, firms situated within the supply chain of SOEs, whether upstream or downstream, adapted
their employment strategies in response to the industrial upgrading brought about by the reform (Du
and Yang, 2013). In sum, the SOE reform wielded a multifaceted influence on the local labor market,
which ultimately influenced the occupational status of individuals. Many displaced SOE workers
looked for new job opportunities and transitioned into different job roles (Fang et al., 2023). They also
encountered significant obstacles in getting these new positions due to factors such as limited
educational qualifications, skill sets, and age-related biases (Tian et al., 2022). This resulted in a
substantial decline in the occupational status of many former SOE employees. For employees of non-
SOE firms who were squeezed out, they also had to find new jobs and their occupational status may
also decline.
3 Data and Measurement
The main dataset used in our analysis is from the Chinese General Social Survey (CGSS), a part

7
of the International Social Survey Program, covering 31 provinces in mainland China. The survey,
comparable to the General Social Survey (GSS) in the United States, includes detailed information on
respondents’ demographics, family background, occupation, and perspectives on gender role attitudes
related questions. We focus on the wave of 2015 as it offers a sufficient sample size of individuals
whose parents experienced the SOE reform—an exogenous shift in their occupational status—during
the individuals’ adolescence, and these individuals had also reached adulthood when the survey was
conducted. Moreover, the CGSS data following the 2015 wave does not include geographic identifiers
at the sub-provincial level.
Gender role attitudes. Gender role attitudes refer to beliefs and expectations that people hold
about the roles and responsibilities assigned to men and women in a society, spanning areas such as
career choices, competence in certain fields, marriage decisions, the allocation of resources when faced
with scarcity, and the division of housework. In patriarchal societies such as China, women usually face
social disadvantages relative to men and are typically expected to take on more family obligations.
Despite advancements in women’s employment and education in China, significant gender gaps persist
in household work allocation (Yu and Yu, 2012).
5
We assess gender role attitudes with a commonly
used question related to the distribution of household chores (Dahl et al., 2022; Du et al., 2021; Si,
2022). In the CGSS, respondents were asked to what extent they agreed with the statement “Husbands
and wives should share household chores equally.” Answers to the question are recorded on a 5-level
Likert scale, ranging from 1 (totally disagree) to 5 (totally agree). A higher score indicates a more
egalitarian view of the distribution of household chores between husbands and wives.
Occupational Status Score. The CGSS data has the advantage of providing information on the
occupations of the individuals’ parents when the respondents were 14 years old, categorized using the
scheme of the 4-digit ISCO-88 coding.
6
It is important to note that information on parents’ income when
the respondents were in their adolescence is not available in the CGSS data. We exploit two widely
used indexes to measure occupational status: (1) The International Socioeconomic Index (ISEI),
developed by Ganzeboom et al. (1992), measures an occupation’s socioeconomic status based on its
5
Chinese women spent approximately three times more time on housework than men, surpassing gaps seen in countries like
the United State. See https://www.bloomberg.com/news/articles/2019-01-25/being-a-woman-in-china-means-working-a-
sixth-of-your-life-unpaid.
6
The International Standard Classification of Occupations (ISCO) is a tool for organizing jobs into a clearly defined set of
groups according to the tasks and duties undertaken in the job, which is adopted by the International Conference of Labour
Statisticians (ICLS). ISCO-88 was the third version of ISCO and was adopted by the Fourteenth ICLS in 1987.

8
educational requirements, income, and age (see Appendix Table A1 for the detail). Occupations are
assigned scores ranging from 16 (lowest socioeconomic status) to 90 (highest socioeconomic status).
For instance, occupations with low scores include forestry workers and loggers, while department
managers of large enterprises represent high scores (refer to Appendix Figure A4 for additional
examples). (2) The Standard International Occupational Prestige Scale (SIOPS), assessed by survey
respondents through a series of questions to evaluate occupations according to their social standing on
a scale of 0-100, with a higher score indicating a higher level of prestige (Treiman, 1977) (see Table
A1 for the detail). This index captures individuals’ subjective perceptions of an occupation’s prestige.
As a robustness check, we use the Chinese International Socioeconomic Index (CISEI), developed
by Li (2005). It resembles the ISEI measure but is tailored to the Chinese context, by considering three
additional factors as indicated in Appendix Table A1: the power factor (indicating managerial positions),
the department factor (government or public institutions), and the social discrimination factor
(occupying a socially discriminated position).
7
We provide a detailed description of the indices used to
create these measures in the Appendix A1. By merging the occupational information into the three sets
of indices mentioned above, we can measure the occupational status of parents when the children are
14 years old. The occupational status score was assigned as 0 for unemployed parents. We subsequently
calculated the difference in parental occupational status by subtracting the father’s score from the
mother’s score.
While differences among parents span various other dimensions, such as age, education, income,
and personality, we contend that occupational distinctions wield significant influence over intra-
household bargaining power. To discern the powerful determinants of intra-household bargaining
power, we employ a horse race regression using the CFPS data (detailed in the subsequent subsection).
Notably, housing assets, integral to Chinese household wealth (Xie and Jin, 2015), can reliably indicate
an individual’s bargaining power (Wang, 2014). Appendix Figure A5 indicates that the difference
between mother and father in occupational status is associated with a significant reduction in the
likelihood of husbands owning a house (see Panel A) and a slight increase in the likelihood of a wife
owning a house (see Panel B), signifying an increase in wives’ relative bargaining power for property
ownership. The difference in years of education also contribute significantly in predicting household
7
We manually matched our dataset with the CISEI, covering 161 occupations. As some data points lack CISEI information,
potentially reducing the sample size, we use CISEI data for supplementary analysis.

9
bargaining power, whereas the differences in age, income, and personality do not.
Other data set. To shed light on the potential mechanisms and cross-validate our findings, we
additionally use data from the China Family Panel Studies (CFPS), which is a nationally representative
sample. In the 2012 wave, the survey contains information on parents’ occupations when their children
were 14 years old, while the 2014 wave collects data on individuals’ gender role attitudes towards the
distribution of household work which is similar to the CGSS. Specifically, respondents were asked to
what extent they agreed with the statement “Men should do half of the housework.” The CFPS also
includes information on individual demographic characteristics and family background. A unique
feature of the CFPS is that it has personal identifiers for each household member, allowing for linking
a child with their parents and spouse, which facilitates subsequent analysis. However, one caveat for
using the CFPS dataset is that, since the information on occupation and gender role attitude was
collected in different waves, we have an attrition rate of around 20% when tracking respondents across
these two waves. Appendix Table A2 reports the summary statistics of the variables of interest in the
CFPS dataset.
Sample restriction and summary statistics. We restrict our sample to individuals whose parents
were alive and worked in non-agriculture-related jobs when the individual was 14 years old. The reason
for this restriction is that in rural areas, it is common for both parents to be farmers or engage in
agricultural-related work. Including these samples would result in a large proportion of zeros in the
difference in parental occupational status, as shown in Appendix Figure A6. Given China’s dualistic
urban-rural structure, we opt to exclude these samples, resulting in a final sample size of 2,058.
8
Table 1 presents the summary statistics of the main variables for the CGSS sample. The mean
values of the three indices of the differences in parental occupational status are all negative, indicating
that, on average, fathers held higher occupational status than mothers when the children were 14 years
old. The mean score for gender role attitudes is 3.864, indicating a relatively neutral stance on gender
roles related to the distribution of household chores. Since we focus on respondents whose parents were
non-farmers, the majority of the samples live in cities (83.7%), and both parents and children have
relatively high educational levels.
8
In our sample, approximately 20% of fathers’ age information and 18% of mothers’ age information is missing. To avoid
sample selection due to missing values in parental age, we grouped the children by their age and imputed the missing data by
calculating the mean parental age for each group. It is noteworthy that our results remain robust without the imputation of age.
See Section 5.3 below for the detail.

10
4 Estimation Strategy
To investigate how the difference in parental occupational status affects children’s gender role
attitudes, we estimate the following regression equation:
!""#"$%&!"# ' ($)(%%#*!"# )(&+!"# ),"# ) -!"#.
/
0
1
where
!""#"$%&!"#
represents the gender role attitudes of child
#
, who was born in year
"
and lives
in county
2
;
%#*!"#
denotes the difference between the mother’s and father’s occupational status when
the child was 14 years old;
+!"#
is a vector of control variables, which contains children’s
characteristics (i.e., gender, ethnicity, education level, and hukou status) and parents’ characteristics
(i.e., parental education level, parental age difference, and the family’s social class when the child was
14 years old);
,"#
is the county-by-birth-year fixed effects, netting off cohort-specific macro shocks
within a county. We cluster the standard error by sampling unit (i.e., county).
(%
, as the parameter of
interest, captures the impact of the difference in parental occupational status on children’s gender role
attitudes.
Estimating Equation (1) through ordinary least squares (OLS) regressions may encounter potential
endogeneity issues. First, parental occupational status might be correlated with omitted variables, such
as their own gender role attitudes, thereby influencing the gender role attitudes of their children (Dhar
et al., 2019; Farre and Vella, 2013). Second, measurement errors in parental occupational status may
lead to attenuation bias. To mitigate these concerns, we employ an instrumental variable strategy,
leveraging the exogenous variation introduced by the SOE reform in China during the 1990s.
Specifically, we employ three sources of variation in SOE reforms. (1) Cohort variation in parental
exposure to the SOE reform. We define children aged 12 to 14 years during the SOE reform as the
treated group, aligning with our main explanatory variable being the difference in parental occupation
when the child was 14 years old.
9
(2) Regional variations in gender-specific layoff intensity during the
reform, measured as the ratio of the number of female (or male) layoffs in province
3
at the end of
year
" ) 4
to the number of female (or male) SOE workers in 1993 (
356365"#67'(#)*
female/male
8
, collected
from the China Labor Statistics Yearbook.
10
(3) Variation in parental age when their children were 14
9
We us e o th er a ge c ri te ri a i n t he d ef in ition of treated children in Appendix Table A13. The results are insensitive to this
definition.
10
We c ho se 19 93 be ca us e i t is th e y ea r pr e ce di ng t h e Ce nt r al E co n om ic Wo rk Co nf er en ce , a t w hi ch t h e ce nt r al g ov e rn me nt
proposed a deepening of the economic system reform targeting State-Owned Enterprises.

11
years old. This is considered because layoff risks varied by age, particularly affecting middle-aged
workers during the reform (Appleton et al. 2002; He et al. 2018).
11
Formally, we define the intensity of female-specific or male-specific exposure to SOE reform for
parents of child
#
born in year
"
in province
3
(i.e.,
9:;!('(#
female/male) as follows:
9:;!('(#
female/male
'
<
=
%+
*,%&
356365"#67'(#)*
female/male
>?
@
" ) 4 A B-./
C
.
/
D
1
where
4
represents the age of child,
3
represents the province of residence, and the remaining
notations are the same as Equation (1).
B-./
is the time window of SOE reform (i.e., 1998 to 2001).
12
Intuitively,
9:;!('(#
female/male captures the share of affected years in the window of the SOE reform period
weighted by the intensity of the SOE reform, based on the province of residence and year of birth of
the child. Thus, it measures the extent to which the mother’s or father’s occupation suffered from the
SOE layoff shock when the children were 12 to 14 years old. To account for the differential exposure
to SOE layoffs based on parental age, we incorporate the interaction between the intensity of SOE
layoffs with the quadratic terms of parental age when the children were 14 years old. These variables
serve as instrumental variables for the difference in parental occupational status (i.e.,
%#*!"#
).
Consequently, we have the following first-stage regression equation:
%#*!"# ' E$)E%9:;!('(#
female
>4F&*,%+
mother+
E&
9:;!('(#
female
>G4F&*,%+
mother
8&
)E09:;!('(#
male
>4F&*,%+
father +
E+
9:;!('(#
male
>G4F&*,%+
father
8&)
!
+!"# ),"# )$!"#
/
H
1
where
4F&*,%+
mother/father is the parental age when their child was 14 years old, and the other notations are
the same as before. This approach exploits exogenous shifts in differences in parental occupational
status resulting from variations in exposure to SOE layoff shocks among children whose parents were
at different ages. In particular, by controlling for the county by birth year fixed effects in the
specification, we can account for the potential concern that the intensity of male- and female-specific
SOE laid-off shocks affects the local labor market differently, which in turn may have an impact on
children’s views on gender roles.
11
According to the China Labor Statistical Yearbook, workers aged 36 to 46 accounted for 42.4% of laid-off workers in 1998,
whereas workers younger than 36 and those older than 46 accounted for 35.9% and 21.7%, respectively.
12
This time window was selected for three main reasons. First, it corresponds to the implementation of the three-year target,
which represents the peak of laid-off and workforce restrictions. Second, the main source of data that we rely on to measure
SOE reform intensity, the China Labor Statistics Yearbook, only provides comprehensive information on the number of male
and female SOE workers laid off in each province between 1996 and 2001. Third, China’s accession to the World Trade
Organization in 2001 had a significant impact on the Chinese labor market in subsequent years (Feng et al., 2017). By selecting
this time period, our study can effectively avoid the confounding effects of global trade arising from China’s WTO accession.

12
5 Effects on Gender Role Attitudes
5.1 Baseline Results
We begin by offering visual evidence in Figure 1 of the relationship between parental occupational
status differences and children’s gender role attitudes. We observe that as the occupational status
differences between mothers and fathers increase, there is a corresponding trend towards more
egalitarian gender role attitudes among their children. These associations persist across various
measures of occupational status, implying the potential role of relative occupational status between
parents in shaping children’s gender role attitudes.
We proceed to our regression analysis of these associations in Table 2. Column (1) displays the
OLS results when we use the ISEI measure to capture the occupational differences between parents.
We observe that a higher relative occupational status of the mother within a family is associated with
more egalitarian gender role attitudes among their children (coef.=0.005, s. e.=0.002). Columns (2) and
(3) further reinforce our findings when we use the SIOPS and CISEI measures to calculate parental
occupational differences.
To address concerns about endogeneity in parental occupational status differences affecting their
children’s gender role attitudes, we employ an instrumental variable approach as outlined in Section 4.
Appendix Table A3 presents the first-stage results of our two-stage least squares (2SLS) estimation.
Columns (1) to (3) show a strongly association between the parental occupational status difference and
the instrumental variables, confirming our hypothesis that SOE reform induces a shift in the difference
in occupational status between parents. The first-stage F-statistics range from 18.12, 11.80, and 14.51,
indicating that weak IV is not a concern. In the remaining columns, we run regressions separately on
father’s and mother’s occupational status scores. The results show that the change in the difference in
parental occupational status score is primarily driven by the decline on the mothers’ occupational status,
which is consistent with evidence we presented earlier.
Turning to the second-stage results of the 2SLS estimation, Columns (4) to (6) of Table 2 show a
positive and statistically significant coefficient on parental occupational differences. Estimates in
Column (4) indicate that a one standard deviation increase in the occupational differences between
mothers and fathers (as measured by the ISEI index) raises the score of gender role attitudes by 0.40

13
(=0.016"25.311), accounting for 40.3% of the standard deviation of the dependent variable. Our 2SLS
estimates are larger in magnitude than the OLS counterparts. The difference between OLS and 2SLS
estimates suggests the potential correlation between parental occupational status difference and omitted
variables, such as occupational segregation, and susceptibility to measurement errors. In addition, the
2SLS estimates provide prediction of the impact among compliers, which may differ from those of the
non-compliers.
For comparison, Du et al. (2021) use the CGSS data and find that one additional year of schooling
induced by China’s Compulsory Education Law increases the probability of the endorsement of
egalitarian gender role attitudes related to housework by approximately 5 percentage points. In
Appendix Table A4, where we use the same indicator variables of egalitarian gender role attitudes as
Du et al. (2021), we find that a one standard deviation increase in the occupational differences between
mothers and fathers (as measured by the ISEI index) raises the probability of egalitarian gender role
attitudes by 12.6 percentage points (=0.005×25.311). That is, roughly speaking, the impact on
egalitarian gender role attitudes as a result of a one standard deviation increase in occupation differences
between mothers and fathers in adolescence is equivalent to the impact of 2.5 years of schooling.
5.2 Validity of IV
The validity of our IV estimation relies on the assumption that parental exposure to the SOE reform
does not affect children’s gender role attitude conditional on difference in parental occupation status,
the observable characteristics of parents and children, and the county-by-birth-year fixed effects.
One concern of the exclusion restriction is that, in regions with strong gender norms, women are
more likely to be laid off during the reform, and these norms may also shape children’s gender role
attitudes. To address this concern, we partition our sample based on the sex ratios of newborns in 1990
as a proxy for local gender norms. Panel A of Appendix Table A5 demonstrates that our first-stage
results are notably stronger in regions with low sex ratios. Meanwhile, our second-stage estimates in
Panel B remain robust and significant in these low sex ratio regions, while no such patterns are observed
in regions characterized by strong gender norms. The results suggest that our benchmark results are not
driven by regions with strong gender norms, which should alleviate the above concern.

14
In addition, recognizing that men and women often occupy distinct occupations, the SOE reform
may impact male’s and female’s laid-off disproportionately depending on the occupational segregation
by gender, which may directly influence gender roles transmitted to children. To account for this
possibility, we control for the average sex ratios of mothers’ and fathers’ occupations, calculated from
the 1990 population census, in Equations (1) and (3). Appendix Table A6 shows that the results are
quite similar to the benchmark estimates by controlling for occupational segregation by gender.
SOE reform may potentially affect the stability of the family (e.g., divorce of parents) through the
mass laid off, and, subsequently, influence children’s gender role attitudes (Kiecolt and Acock, 1988).
To gauge this concern, we include parents’ divorce status as an additional control variable. As shown
in Appendix Table A7, the results remain robust. Furthermore, SOE reform might impact mothers’
fertility decisions (Xie et al., 2022), and subsequently the sibling structure may influence individuals’
gender role attitudes (Bidner and Yang, 2022). In particular, the sibling effect on gender role attitudes
may depend on the gender of the children. To address the concern, we control for a dummy indicating
whether the children have a sibling in Columns (1) and (2) in Appendix Table A8, and its interaction
with the gender of the children in Columns (3) and (4) of the same table. As shown, our results are
robust even after considering the presence of siblings and its interaction with the gender of the
children.
13
Overall, the results from these analyses should further reduce the concerns about the
exclusion restriction.
Last but not least, we conduct two placebo tests to check the validity of our IV estimation. First,
we conduct a falsification test among farmer families in Appendix Table A9.
14
The SOE reform was
conducted in urban areas. For families living in rural areas and engaged in farming, their occupational
status was unlikely to be affected by the SOE reform. If our exclusion restriction is valid, we would
expect no impact of the exposure to SOE reform on gender role attitudes among those who grew up in
farmer families. In line with the conjecture, we find no statistically significant impact of parental
exposure to SOE reform on their occupational differences. The small first-stage F-statistics suggest a
lack of predictive power of SOE reform for parental occupational differences among farmer families.
In the reduced-reform results in Column (4), the gender role attitudes among respondents from farmer
13
The analyses of both Tables A7 and A8 are conducted using CFPS 2014 due to data availability. See Section 5.3 below for
cross-validation using the CFPS data.
14
We de fi ne a f a rm er fa mi l y a s o ne in w hi ch bo th p ar en ts wo rk i n a gr ic ul t ur e-related jobs when the respondents were 14
years old.

15
families show virtually no impact from parental exposure to SOE reform, reinforcing confidence in the
validity of our instrumental variable. Our second placebo test considers children in different age groups
at the time of SOE reform. Prior studies have shown that adolescence is a crucial period for the
development of gender role attitudes (Eagly and Wood, 2012; John et al., 2017; Dhar, Jain, and
Jayachandran, 2022; Ullrich, Becker, and Scharf, 2022). Thus, we expect for those who experienced
SOE reforms when they were not during their adolescence, their gender role attitudes were unlikely
being affected by the SOE reforms. The reduced form results in Appendix Table A10 show that parental
exposure to SOE reform significantly influences the gender role attitudes of cohorts aged 12-14 during
the reform (see Column (2)). Consistent with our conjecture, there are no significant effects for either
the younger or the older cohorts.
5.3 Robustness
Confounding policies and shocks
We assess the robustness of our baseline IV estimates by accounting for potential confounding
policies and shocks. Table 3 presents results from 2SLS regressions that consider three sets of
confounding policies and shocks, including the college expansion in 1999, the Asian financial crisis in
1997, and China’s entering WTO in 2001.
The Chinese government initiated expansion of college education in 1999. The college expansion
may impact gender roles through greater access to higher education or changing labor market conditions.
Moreover, the influence can be heterogeneous by gender (Si, 2022). To address the concern, we
introduce a triple interaction term involving provincial-level college expansion intensity, individual
exposure to college expansion (18 years old or younger in 1999), and gender. Provincial-level college
expansion intensity is measured by the proportion of university enrollment quota of a province relative
to the number of high school graduates in that province in 1998—the year prior to college expansion.
15
As shown in Panel A of Table 3, the results indicate that the main estimates remain similar and
statistically significant after accounting for the potential role of college expansion.
15
See Section B.2 in the Appendix for the detail about the measurement the confounding policies and shocks discussed in this
section.

16
We also consider the 1997 Asian financial crisis, which might have influenced parental
occupational status through a slowdown of economic growth and the restructuring of industries (Atinc
and Walton, 1998). To account for the impact of the Asian financial crisis, we introduce two triple
interaction terms that combine provincial-level intensity of exposure to the crisis, an indicator of
whether individuals were aged 12-14 during the period of SOE reform, and parental age when children
were 14 years old (separately for the father and the mother). We measure the intensity of a province’s
exposure to the Asian financial crisis by the ratio of the province’s FDI to GDP in 1996. In Panel B of
Table 3, we find that the inclusion of the additional variables does not change the main conclusion of
our results.
Another confounding policy shock is China’s WTO accession. Prior studies have documented that
China’s WTO accession, which was accompanied by increased import penetration, led to the retention
of more females in the workforce, reducing the gender employment gap (Wang, Kis-Katos, and Zhou,
2022). We control for parental exposure to China’s WTO accession in 2001. Specifically, we interact
the geographical distance between provincial capitals to the nearest port with an indicator of parental
exposure to the SOE reform when the children were 12 to 14 years old and parental age when the
children were 14 years old. Results in Panel C of Table 3 demonstrate consistent findings, reinforcing
the robustness of our results. In summary, the analyses in Table 3 enhance our confidence in the
robustness of our key findings.
Migration issue
In constructing the instrumental variable, we assume respondents’ province of residence during
their adolescence was the same as their current place of residence, as the data only provides information
on the current place of residence. While there is a possibility of individuals migrating after adolescence,
such behavior could be correlated with unobserved characteristics that might confound the formation
of gender role attitudes. It is worth noting that only 12.93% of respondents migrated across provinces
in our sample, indicating that migration is not a severe concern for our identification. To further alleviate
this concern, we restrict our analysis to non-migrants in Columns (1) to (3) in Appendix Table A11.
The results are robust and remain statistically significant when using this non-migrant sample.
16
16
We d ef in e th e n on -migrant sample as respondents who have always lived in the same county.

17
Additionally, we introduce flexible trends for migrants by interacting the migration dummy variable
with county-by-birth-year fixed effects in the last three columns of Table A11, and these adjustments
do not significantly alter our results. In sum, these results suggest that our main estimates are unlikely
to be severely biased due to possible measurement issues of the place of residence during adolescence.
Other measures of gender role attitude
In addition to the division of housework, the CGSS data also asks respondents about their
agreement on gender stereotypes associated with careers, natural features, marriage, and employment.
Appendix Figure A7 further explores the impact of parental relative occupational status on the other
dimensions related to individuals’ gender role attitudes in the CGSS data. The results reveal that higher
levels of parental occupational differences result in more egalitarian gender role attitudes associated
with all the other domains, although the estimates are statistically insignificant. We also construct a
composite index as the average of the standardized values of all dimensions of gender role attitudes.
Consistent with our main finding, the figure illustrates that improved occupational status of mothers
within the family causes more egalitarian gender role attitudes assessed by the composite index. The
estimates are statistically significant at the level of 10%.
Cross-validation using the CFPS data
To provide additional validation of our findings, we use another dataset, the China Family Panel
Studies (CFPS). In the 2014 wave of CFPS data, respondents were asked about their level of agreement
with the statement “Men should do half of the housework,” which is comparable to the one used in the
CGSS. The survey also contains questions related to individuals’ gender role attitudes across other
domains, including careers, marriage, and childbearing. By leveraging the CFPS data, we can examine
the robustness and generalizability of our baseline analysis. Appendix Figure A8 shows that increased
parental occupational differences lead to more egalitarian gender role attitudes in the domain of the
division of housework, which is consistent with our previous findings using the CGSS dataset.
Meanwhile, the figure also indicates that imposed parental occupational differences cause more
egalitarian gender role attitudes related to childbearing.
Other robustness checks

18
We provide additional exercises to guarantee the robustness of our findings. First, we assess the
sensitivity of our results to the imputation of parental age. The first-stage results in Columns (1), (3),
and (5) in Appendix Table A12 demonstrate strong F-statistics when we use different definitions of
occupational status. Furthermore, the second-stage results in Columns (2), (4), and (6) indicate that our
results are robust when not imputing parental age, with the magnitudes of the coefficients being very
similar to the baseline estimates. Second, we examine the consistency of our results across different
definitions of parental exposure to SOE reforms in Appendix Table A13. Specifically, we use
alternative specification of the affected age during the period of SOE reform. Remarkably, we find that
our 2SLS results remain unchanged. Moreover, the F-statistics of our first-stage results demonstrate
strong statistical power subject to these alternative definitions. Third, to address the concern about the
possible impacts of absolute levels of parental occupational status on the formation of gender role
attitudes on top of the impact of their relative status, we include the average score of parental
occupational status in the regressions. Appendix Table A14 shows that, once again, the inclusion of this
variable does not materially change our findings across various definitions of occupational status.
5.4 Heterogeneity by Gender
After establishing the robust relationship between parental occupational differences and gender
role attitudes, we extend the analysis by exploring the heterogeneous effects by demographic attributes.
Table A16 in the Appendix presents both the OLS and 2SLS estimates of the effects of occupational
differences on gender role attitudes for males and females, respectively. Interestingly, we observe a
more pronounced effect for male children compared to female children. This pattern can be partially
explained by the fact that males exhibit a lower level of egalitarian gender role attitudes than females
do. It also echoes prior studies which find that males are more likely to exhibit egalitarian gender role
attitudes if they have a working mother (Fernández et al., 2004; Chen and Ge, 2018).
6 Marital Behaviors of the Next Generations
Our preceding analysis has predominantly focused on discerning the impact of parental
occupational differences on the cultivation of more egalitarian gender role attitudes among their
offspring. Extending our inquiry, we delve into whether alterations in gender role attitudes precipitate

19
behavioral changes in the marriage market for their children, as gender role attitudes can influence
individuals’ behaviors in their choice of a mate by shaping their preferences, expectations, and criteria
for selecting a partner (Fernández et al., 2004; Trimarchi, 2022). We specifically scrutinize marriage
matching patterns concerning the occupational status of succeeding-generation couples, as in traditional
gender norms in China, men are typically expected to have higher occupational status than women.
Given that marriage matching is a conditional outcome based on marriage formation, our initial
exploration concentrates on discerning if parental occupational differences influence the selection of
their offspring into marriage.
Columns (1) and (2) of Panel A of Appendix Table A17 show that the difference in parental
occupational status (measured by the ISEI score) does not significantly affect the likelihood of being
married for both genders, using the OLS specification. Consistent results are observed in Columns (3)
and (4) when utilizing the SIOPS score to measure occupational differences. The 2SLS estimates in the
subsequent columns of Panel A corroborate the OLS findings. Panels B and C further reinforce these
results, indicating no statistically significant association between parental occupational differences and
the probability of being single or divorced, indicating that parental occupational differences do not play
a significant role in determining individuals’ marital status.
We proceed to explore the association between parental occupational status difference and the
marriage matching patterns of couples in the next generations, focusing specifically on their
occupational status. In Panel A of Table 4, we explore whether the disparity in occupational status
between parents influences the occupational status of their children. The 2SLS estimates reveal that a
mother’s higher relative occupational status in a family is associated with a higher occupational status
for female children while imposing no statistically significant impact on male children’s occupation.
Moving to Panel B of Table 4, it is evident that a mother’s higher relative occupational status in a family
reduces the occupational status of the spouse of their female children while increasing the spousal
occupational status of their male children.
Shifting attention to Panel C of Table 4, we concentrate on the influence of the relative
occupational status of the mother within the family on the occupational differences between couples in
the next generation. The results indicate that an improved relative occupational status of the mother
significantly amplifies the difference in occupational status between wives and husbands in the
subsequent generation. This is substantiated by both OLS and IV estimates. Our preferred IV estimates

20
with ISEI occupational measure in Column (5) of Panel C indicate that a one standard deviation increase
in parental occupational differences (mother-father) results in a 28.92 increase in the occupational status
score difference between wives and husbands for their female children, or about 1.18 times of the
standard deviation. Notably, Column (6) of Panel C indicates an even more pronounced impact on male
children. A one standard deviation increase in parental occupational differences results in an increase
of 55.23 units in the occupational status score difference between wives and husbands, corresponding
to 2.26 times of the standard deviation. These findings highlight the influential role of improved
maternal occupational status relative to fathers in shaping the marriage matching patterns of their
children in terms of occupational status. This potentially reflects the intergenerational transmission of
gender role attitudes.
7 Discussion on Potential Channels
The influence of the presence of mothers with higher occupational status within a family on the
formation of more egalitarian gender role attitudes can be explained by two possible channels. First,
mothers with higher occupational status may possess increased intra-household bargaining power,
which in turn allows them to exert influence in socializing their children and transmitting more
egalitarian gender role attitudes (the “instillation” channel). Second, the relative occupational status
between parents may influence the division of household chores, providing a context from which their
children can learn and form gender role attitudes related to housework. Observing a more equitable
distribution of household chores can shape children’s perceptions and expectations regarding gender
roles within the home (the “internalization” channel). We investigate the two channels in the analyses
below using data from CFPS.
7.1 Bargaining Power and Indoctrination
We initiate our analysis by examining whether an increase in the wife’s relative occupational status
can enhance her bargaining power within the household, which may indicate greater influential power
in instilling egalitarian gender role attitudes in their offspring. We focus on two sets of measures of
intra-household bargaining power using the CFPS data: (1) housing ownership, as housing assets are a

21
crucial component of household wealth in China (Wang, 2014; Xie and Jin, 2015);
17
(2) household
consumption patterns related to gender-specific items (Brown, 2009; Li and Wu, 2011). Figure 2
visualizes the estimated coefficients for each gender. Our findings reveal that a women’s higher
occupational status within the family is associated with a significant reduction in her husband’s
likelihood of owning a house and a marginal increase in her own likelihood of having a house. In
addition, Figure 2 demonstrates a negative association between the higher occupational status of women
within the family and the consumption of these male-preferred items, but no statistical associations with
female-preferred goods such as cosmetics and washing machines. Interestingly, we uncover a
statistically significant positive association of the higher occupational status of women within the family
and household education expenses, aligning with the findings that women with greater bargaining
power tend to allocate family resources toward education-related expenses (Duflo, 2003; Bobonis,
2009). Overall, the results in Figure 2 imply that women’s higher occupational status within the family
enhances their bargaining power relative to their husbands.
Subsequently, we investigate whether occupational differences shape the gender role attitudes of
wives and husbands. Column (1) of Table 5 shows that the difference in occupational status measured
by the ISEI score is significantly associated with more egalitarian gender role attitudes among women
(wives). This finding is further supported by consistent results obtained in Column (3) when using the
SIOPS score to measure the occupational difference. We do not find any substantial association of
occupational status difference and the gender role attitudes of men (husbands) in Columns (2) and (4)
in Table 5.
In sum, our findings in Figure 2 and Table 5 indicate that mothers with higher occupational status
exhibit more egalitarian gender role attitudes and greater bargaining power within the household. This
implies that they may exert a more influential role in fostering more egalitarianism in gender role
attitudes among their children.
7.2 Learning and Internalization
Gender norms can be transmitted through the process of learning and imitation (Miho et al., 2023).
In this subsection, we test whether the relative occupational status between parents may influence the
17
Xie and Jin (2015) pointed out that housing assets consisted of 74% of total household wealth in 2012, a significantly higher
proportion compared to other countries.

22
division of household chores, creating a context through which their children can learn and develop
more egalitarian gender role attitudes in the domain of household chores. The results, presented in Table
6 and disaggregated by gender (Panel A) and the difference between wives and husbands (Panel B),
reveal insightful patterns. In Panel A, Columns (1) and (2) demonstrate significant associations of
occupational status difference (measured by the ISEI score) and the division of housework on weekdays.
Specifically, the higher occupational status of women within the family is associated with less hours of
household chores for wives (coef.=-0.009, s.e.=0.002) and more hours for husbands (coef.=0.008,
s.e.=0.001). These findings hold consistently when employing the SIOPS score for occupational
differences in Columns (3) and (4). The remaining columns in Panel A exhibit parallel results for the
division of housework on weekends.
In Panel B of Table 6, we shift our focus to examine the effects on the difference in hours spent
on housework between wives and husbands. Here, both the ISEI and SIOPS scores reveal a significant
reduction in the inequitable division of household chores between wives and husbands in both weekdays
and weekends, associated with a women’s higher occupational status within the family. This more
equitable division of household responsibilities may contribute to developing egalitarian gender role
attitudes among their children, as they can observe and internalize these patterns.
8 Conclusion
Can family socialization impact a person’s gender role attitudes? This paper delves into how the
relative occupational status of parents during their children’s adolescence plays a role. Our main
findings demonstrate that one standard deviation increase in the occupational difference between
mothers and fathers leads to a 0.40 standard deviation increase on gender role attitudes related to
housework division, indicating a more egalitarian perspective. Additionally, our findings are reinforced
by evidence of individuals’ behaviors in the marriage market. Specifically, we document that mothers’
improved occupational status within the family plays a strong role in shaping the marriage matching
patterns of their children in the dimension of occupational status, but it does not alter the marriage
formation of the children. We highlight two possible channels through which egalitarian gender role
attitudes are cultivated within the family. We provide supportive evidence that egalitarian gender role

23
attitudes can be instilled by mothers with greater intra-household bargaining power, but can also be
learned and internalized by children from an equitable atmosphere within the family.
In many societies, there are certain expectations for how people should behave based on their
gender. These expectations can limit personal choices and contribute to unequal treatment between
genders. However, in recent years, more and more women have been breaking barriers and achieving
high-ranking positions in fields such as politics, business, and academia.
18
Studies have shown that
having powerful women in positions of influence can help reduce gender inequality and promote more
equal gender roles (Yao and You, 2018). Our findings highlight that improved women’s occupational
status relative to men within the family also has the potential to influence their offspring’s attitudes and
beliefs, fostering a more egalitarian gender role attitude in the next generation. As these children grow
up and enter society, they have the potential to contribute to the promotion of gender equality on a
broader scale. In this sense, empowering women can play a vital role in fostering societal change in
gender equality.
18
See http://www.chinatoday.com.cn/ctenglish/2018/zdtj/202009/t20200925_800221877.html for examples from China.

24
References
Alan, S., Baydar, N., Boneva, T., Crossley, T. F., & Ertac, S. (2017). Transmission of risk preferences
from mothers to daughters. Journal of Economic Behavior & Organization, 134, 60-77.
Alesina, A., Giuliano, P., & Nunn, N. (2013). On the origins of gender roles: Women and the plough. The
Quarterly Journal of Economics, 128(2), 469-530.
Anelli, M., Giuntella, O., & Stella, L. (2021). Robots, marriageable men, family, and fertility. Journal
of Human Resources, 1020-11223R1.
Andersen, S., Ertac, S., Gneezy, U., List, J. A., & Maximiano, S. (2013). Gender, competitiveness, and
socialization at a young age: Evidence from a matrilineal and a patriarchal society. Review of
Economics and Statistics, 95(4), 1438-1443.
Appleton, S., Song, L., & Xia, Q. (2005). Has China crossed the river? The evolution of wage structure
in urban China during reform and retrenchment. Journal of Comparative Economics, 33(4), 644-
663.
Atinc, T. M., & Walton, M. (1998). Social consequences of the East Asian financial crisis. World Bank
Group.
Autor, D., Dorn, D., & Hanson, G. (2019). When work disappears: Manufacturing decline and the
falling marriage market value of young men. American Economic Review: Insights, 1(2), 161-178.
Beaman, L., Chattopadhyay, R., Duflo, E., Pande, R., & Topalova, P. (2009). Powerful women: Does
exposure reduce bias? The Quarterly Journal of Economics, 124(4), 1497-1540.
Bertrand, M., Kamenica, E., & Pan, J. (2015). Gender identity and relative income within
households. The Quarterly Journal of Economics, 130(2), 571-614.
Bertrand, M., Cortes, P., Olivetti, C., & Pan, J. (2021). Social norms, labour market opportunities, and
the marriage gap between skilled and unskilled women. The Review of Economic Studies, 88(4),
1936-1978.
Bidner, C., & Yang, B. (2022). Becoming a woman: sibling structure and gender attitudes. Wo rk in g
paper.
Bisin, A., & Verdier, T. (2000). “Beyond the melting pot”: Cultural transmission, marriage, and the
evolution of ethnic and religious traits. The Quarterly Journal of Economics, 115(3), 955-988.
Bisin, A., & Verdier, T. (2001). The economics of cultural transmission and the dynamics of
preferences. Journal of Economic Theory, 97(2), 298-319.
Blau, F. D., Kahn, L. M., & Papps, K. L. (2011). Gender, source country characteristics, and labor
market assimilation among immigrants. The Review of Economics and Statistics, 93(1), 43-58.
Bobonis, G. J. (2009). Is the allocation of resources within the household efficient? New evidence from
a randomized experiment. Journal of Political Economy, 117(3), 453-503.
Brown, P. H. (2009). Dowry and intrahousehold bargaining: Evidence from China. Journal of Human
Resources, 44(1), 25-46.
Cavapozzi, D., Francesconi, M., & Nicoletti, C. (2021). The impact of gender role norms on mothers’
labor supply. Journal of Economic Behavior & Organization, 186: 113-134.
Cortés, P., Koşar, G., Pan, J., & Zafar, B. (2022). Should mothers work? How perceptions of the social
norm affect individual attitudes toward work in the US. National Bureau of Economic Research
Work in g Pa pe r (N o. w 30 60 6) .
Chen, X., & Ge, S. (2018). Social norms and female labor force participation in urban China. Journal
of Comparative Economics, 46(4), 966-987.
Doepke, M., & Zilibotti, F. (2008). Occupational choice and the spirit of capitalism. The Quarterly

25
Journal of Economics, 123(2), 747-793.
Dong, X. Y., & Pandey, M. (2012). Gender and labor retrenchment in Chinese state-owned enterprises:
Investigation using firm-level panel data. China Economic Review, 23(2), 385-395.
Dhar, D., Jain, T., & Jayachandran, S. (2019). Intergenerational transmission of gender attitudes:
Evidence from India. The Journal of Development Studies, 55(12), 2572-2592.
Dhar, D., Jain, T., & Jayachandran, S. (2022). Reshaping adolescents’ gender attitudes: Evidence from
a school-based experiment in India. American Economic Review, 112(3), 899-927.
Dohmen, T., Falk, A., Huffman, D., & Sunde, U. (2012). The intergenerational transmission of risk and
trust attitudes. The Review of Economic Studies, 79(2), 645-677.
Du, H., Xiao, Y., & Zhao, L. (2021). Education and gender role attitudes. Journal of Population
Economics, 34, 475-513.
Du, J., & Wang, Y. (2013). Reforming SOEs under China’s state capitalism. In Jun Zhang (ed.)
Unfinished Reforms in the Chinese Economy (pp. 1-38). New Jersey: World Scientific Publishing
Company Incorporated
Duflo, E. (2003). Grandmothers and granddaughters: Old-age pensions and intrahousehold allocation
in South Africa. The World Bank Economic Review, 17(1), 1-25.
Eagly, A. H., & Wood, W. (2012). Social role theory. Handbook of Theories of Social Psychology, 2.
Fang, H., Li, M., Wu, Z., & Zhang, Y. (2023). Reluctant Entrepreneurs: Evidence from China’s SOE
Reform. NBER Working Paper No. 31700.
Farré, L., & Vella, F. (2013). The intergenerational transmission of gender role attitudes and its
implications for female labour force participation. Economica, 80(318), 219-247.
Feng, S., Hu, Y., & Moffitt, R. (2017). Long run trends in unemployment and labor force participation
in urban China. Journal of Comparative Economics, 45(2), 304-324.
Fernández, R., Fogli, A., & Olivetti, C. (2004). Mothers and sons: Preference formation and female
labor force dynamics. The Quarterly Journal of Economics, 119(4), 1249-1299.
Fernández, R., & Fogli, A. (2009). Culture: An empirical investigation of beliefs, work, and
fertility. American Economic Journal: Macroeconomics, 1(1), 146-177.
Frazier, M. W. (2006). State-sector shrinkage and workforce reduction in China. European Journal of
Political Economy, 22(2), 435-451.
Fredriksson, P. G., & Gupta, S. K. (2023). Irrigation and gender roles. Journal of Development
Economics, 103076.
Ganzeboom, H. B., De Graaf, P. M., & Treiman, D. J. (1992). A standard international socio-economic
index of occupational status. Social Science Research, 21(1), 1-56.
Giuliano, P., and Nunn, N. (2021). Understanding cultural persistence and change. The Review of
Economic Studies, 88(4): 1541-1581.
Grosjean, P., & Khattar, R. (2019). It’s raining men! Hallelujah? The long-run consequences of male-
biased sex ratios. The Review of Economic Studies, 86(2), 723-754.
He, H., Huang, F., Liu, Z., & Zhu, D. (2018). Breaking the “iron rice bowl:” Evidence of precautionary
savings from the Chinese state-owned enterprises reform. Journal of Monetary Economics, 94, 94-
113.
Hill, J. P., & Lynch, M. E. (1983). The intensification of gender-related role expectations during early
adolescence. In Girls at Puberty: Biological and Psychosocial Perspectives (pp. 201-228). Boston,
MA: Springer US.
Jayachandran, S. (2015). The roots of gender inequality in developing countries. Economics, 7(1), 63-

26
88.
Jayachandran, S. (2021). Social norms as a barrier to women’s employment in developing
countries. IMF Economic Review, 69(3), 576-595.
John, N. A., Stoebenau, K., Ritter, S., Edmeades, J., Balvin, N., & Unicef. (2017). Gender socialization
during adolescence in low-and middle-income countries: Conceptualization, influences and
outcomes. Innocenti Discussion Papers, No. IDP_2017_01, UNICEF Office of Research –
Innocenti, Florence.
Kawaguchi, D., & Miyazaki, J. (2009). Working mothers and sons’ preferences regarding female labor
supply: direct evidence from stated preferences. Journal of Population Economics, 22, 115-130.
Kiecolt, K. J., & Acock, A. C. (1988). The long-term effects of family structure on gender-role
attitudes. Journal of Marriage and the Family, 709-717.
Lam, W. R., Rodlauer, M. M., & Schipke, M. A. (2017). Modernizing China: Investing in Soft
Infrastructure. International Monetary Fund.
Li, C. (2005). Prestige stratification in the contemporary China: Occupational prestige measures and
socio-economic index. Sociological Research, 2(74), 74-102.
Li, L., & Wu, X. (2011). Gender of children, bargaining power, and intrahousehold resource allocation
in China. Journal of Human Resources, 46(2), 295-316.
Liu, H., & Zhao, Z. (2014). Parental job loss and children's health: Ten years after the massive layoff of
the SOEs’ workers in China. China Economic Review, 31, 303-319.
Miho, A., Jarotschkin, A., & Zhuravskaya, E. (2023). Diffusion of gender norms: Evidence from Stalin’s
ethnic deportations. Journal of the European Economic Association.
Nadeem, F., & Khalid, R. (2018). The relationship of gender role attitudes with career aspirations and
career choices among young adults. Pakistan Journal of Psychological Research, 33(2), 455-471.
Olivetti, C., Patacchini, E., & Zenou, Y. (2020). Mothers, peers, and gender-role identity. Journal of the
European Economic Association, 18(1), 266-301.
Shenhav, N. A. (2021). Lowering standards to wed? Spouse quality, marriage, and labor market
responses to the gender wage gap. The Review of Economics and Statistics, 103(2), 265-279.
Si, W. (2022). Higher education expansion and gender norms: Evidence from China. Journal of
Population Economics, 35(4), 1821-1858.
Teso , E. (20 19). The lon g-term effect of demographic shocks on the evolution of gender roles: Evidence
from the transatlantic slave trade. Journal of the European Economic Association, 17(2), 497-534.
Tian, X., Gong, J., & Zhai, Z. (2022). The effect of job displacement on labor market outcomes:
Evidence from the Chinese state-owned enterprise reform. China Economic Review, 72, 101743.
Treiman, D. J. (1977) Problems of Method in Comparing Occupational Prestige Systems. In
Occupational Prestige in Comparative Perspective. Academic Press, New York.
Trimarchi, A. (2022). Gender-egalitarian attitudes and assortative mating by age and
education. European Journal of Population, 38(3), 429-456.
Ullrich, R., Becker, M., & Scharf, J. (2022). The development of gender role attitudes during
adolescence: Effects of sex, socioeconomic background, and cognitive abilities. Journal of Youth
and Adolescence, 51(11), 2114-2129.
Wang , S. Y. ( 20 14 ). P ro pe rt y ri gh ts a nd i nt ra -household bargaining. Journal of Development
Economics, 107, 192-201.
Wang , F. , Kis-Katos, K., & Zhou, M. (2022). Import competition and the gender employment gap in
China. Journal of Human Resources, 1220-11399R3.

27
Worl d Ba nk . (2 00 7) . China’s modernizing labor market: Trends and emerging challenges.
Xie, Y., & Jin, Y. (2015). Household wealth in China. Chinese Sociological Review, 47(3), 203-229.
Xie, J., Zhang, J., & Zhou, K. (2022). Stepping down from life-long posts: Layoffs, fertility, and
educational attainment in urban China. Work in g pa pe r.
Zumbuehl, M., Dohmen, T., & Pfann, G. (2021). Parental involvement and the intergenerational
transmission of economic preferences, attitudes and personality traits. The Economic
Journal, 131(638), 2642-2670.
Ya o, Y. , & You , W. ( 2018). Wo me n ’s p ol i ti c al participation and gender gaps of education in China:
1950–1990. Wo rl d D ev el o pm en t, 106, 220-237.
Yu , J . , & X i e , Y. ( 2 0 11 ) . T he varying display of “Gender Display” A comparative study of mainland
China and Taiwan. Chinese Sociological Review, 44(2), 5-30.
Zuo, S. X. (2016). Holding up half of the sky: The impact of state-owned enterprise reform on the
gender inequality in labor market outcomes in urban China, 1988-2007. Work in g pa pe r.

28
Figures and Tables
Figure 1. Parental occupation difference and children’s gender role attitude
Notes: Each panel of the figure shows the scatterplot of parental occupation difference on children’s gender role
attitudes. The points represent the mean values of difference in parental occupational status and children’s gender
role attitudes for every 5th percentile of the sample by the corresponding index of parental occupational status.
The line is the estimated best-fit line from a linear regression of parental occupation difference on children’s
gender role attitude.
Source: CGSS 2015

29
Figure 2. The occupational difference and bargaining power
Notes: The figure reports the point estimates and 90% confidence intervals of the coefficients of occupational
difference between wives and husbands on variables indicating their bargaining power from OLS regressions. The
ownership of a house, motorcycle and washing machines are binary variables. All the variables on expenses are
transformed using the IHS function. The observation units for regressions of house ownership are individuals and
the regressions control for couples’ education, hukou status, age difference between couples, number of children,
log of household income per capita, and county-by-cohort fixed effects. The observation units for the other
regressions are households, and the regressions control for the log of total family expenses besides the variables
mentioned above and control for the county fixed effects instead of county-by-cohort fixed effects. The standard
errors are clustered at the county level.
Source: CFPS 2010 for individual-level regressions and CFPS 2012 for household-level regressions

30
Table 1. Summary statistics
Source: CGSS 2015
NMean SD
Panel A: Key variables
Difference in ISEI score (mother - father) 2058 -17.799 25.311
Difference in SIOPS s core (moth er - father) 2058 -16.396 23.323
Difference in CISEI score (mother - fat her) 1953 -23.453 31.125
Ge n d e r r o le a t t i t u d es 2058 3.864 1.004
Panel B: Individual characteris tics
Male (yes=1) 2058 0.494 0.5
Age 2058 43.664 16.716
Et hn icit y (han =1) 2058 0.948 0.221
Ye a rs o f s c h o o l in g 2058 11.857 3.186
Ye a rs o f s c h o o l in g (f a t h er ) 2058 8.057 4.383
Ye a rs o f s c h o o l in g (mo t h e r ) 2058 6.369 4.752
Urba n hukou (yes=1) 2058 0.837 0.37
Paren tal age d ifference (moth er - father) 2058 -2.386 3.807
Family class 2058 3.926 1.928
CGSS
Va r ia b l e

31
Table 2. Impact of parental occupation difference on children’s gender role attitude
Notes: The dependent variables are children’s ge nd er role attitudes. Columns (1) to (3) report the OLS estimates,
and Columns (4) to (6) report the IV estimates. For parents with missing information on age, the parents’ ag es are
imputed using the average age of parents of children of the same age. The control variables include the child’s
gender, ethnicity, education level, hukou status, parental education level, parental age difference, and the family’s
social class when the child was 14 years old. Robust standard errors in parentheses are clustered at the county
level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
(1) (2) (3) (4) (5) (6)
Difference in ISEI sc ore (mother - fat her) 0.005** 0.016**
(0.002) (0.006)
Difference in SIOPS s co re (moth er - father) 0.005** 0.022**
(0.002) (0.009)
Difference in CISEI score (mother - father) 0.003* 0.012* *
(0.002) (0.006)
Obs ervations 2,058 2,058 1,953 2,058 2,058 1,953
Mean 3.864 3.864 3.868 3.864 3.864 3.868
First-st age F-s tatis tics 18.116 11.797 14.510
p-value of Han s en J-s tatis tic 0.780 0.854 0.717
Cont ro ls Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -co hort fixed effect s Ye s Ye s Ye s Ye s Ye s Ye s
OLS
IV
Hus ban ds and wives s hould equ ally share hou s eh old ch ores

32
Table 3. Robustness checks: control for confounding policies
Notes: The table reports the IV estimates of the impact of difference in parental occupational status on children’s
gender role attitude for the robustness checks by further controlling for the confounding policies. Panel A controls
for children’s exposure to college expansion in 1999. Panel B controls for parental exposure to the 1997 Asian
financial crisis. Panel C controls for parental exposure to China joining the WTO in 2001. The other control
variables include the child’s gender, ethnicity, education level, hukou status, parental education level, parental age
difference, and the family’s social class when the child was 14 years old. Robust standard errors in parentheses
are clustered at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
(1) (2) (3)
Panel A: Control for college expansion
Difference in ISEI score (mother - father) 0.014**
(0.007)
Difference in SIOPS s co re (moth er - father) 0.019**
(0.009)
Difference in CISEI score (mother - fat he r) 0.010*
(0.006)
Obs ervations 2,058 2,058 1,953
Mean 3.864 3.864 3.868
First-st age F-s tatis tics 15.089 10.216 16.909
p-value of Han s en J-s tatis tic 0.740 0.810 0.807
Cont ro ls Ye s Ye s Yes
Coun ty -co hort fixed effe ct s Ye s Yes Ye s
Panel B: Control for 1997 Asian financial cris is
Difference in ISEI score (mother - father) 0.013*
(0.007)
Difference in SIOPS s co re (moth er - father) 0.019**
(0.009)
Difference in CISEI score (mother - fat he r) 0.011
(0.007)
Obs ervations 2,058 2,058 1,953
Mean 3.864 3.864 3.868
First-st age F-s tatis tics 17.233 10.978 8.202
p-value of Han s en J-s tatis tic 0.365 0.503 0.435
Cont ro ls Ye s Ye s Yes
Coun ty -co hort fixed effe ct s Ye s Yes Ye s
Panel C: Control for China's joining WTO
Difference in ISEI score (mother - father) 0.011**
(0.005)
Difference in SIOPS s co re (moth er - father) 0.013**
(0.006)
Difference in CISEI score (mother - fat he r) 0.007*
(0.004)
Obs ervations 2,058 2,058 1,953
Mean 3.864 3.864 3.868
First-st age F-s tatis tics 21.037 13.849 21.012
p-value of Han s en J-s tatis tic 0.453 0.443 0.331
Cont ro ls Ye s Ye s Yes
Coun ty -co hort fixed effe ct s Ye s Yes Ye s
IV es timates
Hus ban ds and wives s hou ld equ ally share h ous eh old chores

33
Table 4. Intergenerational effects of occupation choices
Notes: In panel A, the dependent variables are the occupational status of children. In panel B, the dependent
variables are the occupational status of spouse of children. In panel C, the dependent variables are the difference
in occupation status of children’s household (wife versus husband). Columns (1) to (4) report the OLS estimates
by gender and Columns (5) to (8) report the IV estimates by gender. For parents with missing information on age,
the parents’ ag es are i mput ed using the average age of parents of children of the same age. The control variables
include the child’s gender, ethnicity, education level, hukou status, parental education level, parental age difference,
and the family’s social class when the child was 14 years old. Robust standard errors in parentheses are clustered
at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
Female Male Female Male Female Male Female Male
(1) (2) (3) (4) (5) (6) (7) (8)
Difference in ISEI score (mother - father) 0.129 -0.219
0.825***
-1.002
(0.149) (0.158) (0.211) (0.613)
Difference in SIOPS s core (moth er - father) 0.019 -0.220
0.593***
-0.250
(0.133) (0.149) (0.195) (0.344)
Mean 23.845 29.409 21.392 26.731 23.845 29.409 21.392 26.731
p-value of Han s en J-s tatis tic 0.794 0.537 0.612 0.297
Difference in ISEI score (mother - father) -0.161* 0.136
-0.294**
1.262**
(0.092) (0.112) (0.145) (0.556)
Difference in SIOPS s core (moth er - father) -0.184* 0.123
-0.377**
0.962**
(0.096) (0.117) (0.160) (0.449)
Mean 31.050 20.175 28.184 17.913 31.050 20.175 28.184 17.913
p-value of Han s en J-s tatis tic 0.932 0.580 0.568 0.430
Difference in ISEI score (mother - father) 0.290* 0.355* *
1.119***
2.264**
(0.158) (0.170) (0.282) (1.049)
Difference in SIOPS s core (moth er - father) 0.202 0.343*
0.970***
1.212*
(0.172) (0.179) (0.265) (0.639)
Mean -7.205 -9.234 -6.792 -8.819 -7.205 -9.234 -6.792 -8.819
p-value of tes ting coef[female]=co ef[male]
p-value of Han s en J-s tatis tic 0.937 0.956 0.845 0.492
Obs ervations 664 629 664 629 664 629 664 629
Cont ro ls Ye s Ye s Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -co hort fixed effe ct s Ye s Ye s Ye s Ye s Ye s Ye s Ye s Ye s
First-st age F-s ta tis tics 18.210 2.682 19.138 5.459
Panel A: Child's occupation status
Panel B: Spouse's occupation status
Panel C: Children's household occupation difference (wife - husband)
OLS
IV
0.48
0.13
0.083
0.036

34
Table 5. The difference in occupational status and gender role attitude
Notes: The dependent variables are the gender role attitudes. The odd columns are the results for wives, and the
even columns are the results for husbands. The control variables include education level, hukou status, and log of
household income per capita. Robust standard errors in parentheses are clustered at the county level. *** p<0.01,
** p<0.05, * p<0.1.
Source: CFPS 2014
Wife Husb an d Wife Husband
(1) (2) (3) (4)
Difference in ISEI score (wife - hus band) 0.0020* * -0.0001
(0.0010) (0.0011)
Difference in SIOPS s core (wife - husband ) 0.0023* * -0.0003
(0.0010) (0.0012)
Obs ervations 5,562 5,503 5,562 5,503
Mean 4.150 3.977 4.150 3.977
Cont ro l Ye s Yes Yes Yes
Coun ty -co hort fixed effect s Ye s Ye s Ye s Ye s
Hus ban ds and wiv es sh ou ld equally s hare hous eho ld chores
OLS es tima te s

35
Table 6. The difference in occupational status and time spent on household chores
Notes: The dependent variables in Panel A are the hours wives and husbands spent on household chores. The
dependent variables in Panel B are the differences in hours spent on household chores between wives and husbands.
The control variables include the education level, spouse’s education, the number of children, the number of
children aged less than 6, hukou status, health status, whether living with parents, age difference between spouses,
and wives’ share of income. Robust standard errors in parentheses are clustered at the county level. *** p<0.01,
** p<0.05, * p<0.1.
Source: CFPS 2010
Wife
Hus band
Wife
Hus band
Wife
Hus band
Wife
Hus band
(1) (2) (3) (4) (5) (6) (7) (8)
Panel A: By wife and husband
Difference in ISEI score (wife - hus ban d)
-0.009***
0.008*** -0.003** 0.007* **
(0.002) (0.001) (0.002) (0.001)
Difference in SIOPS score (wife - hu sband)
-0.009***
0.009*** -0.004** 0.007* **
(0.002) (0.001) (0.002) (0.002)
Obs ervation s 4,773 4,773 4,773 4,773 4,772 4,772 4,772 4,772
Mean 2.267 0.875 2.267 0.875 2.539 1.117 2.539 1.117
Cont ro l Ye s Ye s Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -coho rt fixed effect s Ye s Ye s Ye s Ye s Ye s Ye s Ye s Ye s
Panel B: Wife - Husband
Difference in ISEI score (wife - hus ban d)
Difference in SIOPS score (wife - hu sband)
Obs ervation s
Mean
Cont ro l
Coun ty fixed effec ts
Hou rs spent on ho usehold chores
Weekdays
Weekends
-0.016***
(0.001)
-0.016***
(0.001)
-0.009***
(0.001)
-0.009***
(0.001)
Wife - Husband
Wife - Husband
4,772
Ye s
Ye s
4,772
Ye s
Ye s
1.421
1.421
4,773
Ye s
Ye s
4,773
Ye s
Ye s
1.392
1.392
OLS es timates
OLS es timates
Hou rs spent on ho usehold chores
Weekdays
Weekends
Wife - Husband
Wife - Husband
(1)
(2)
(3)
(4)

36
Appendices Figures and Tables
Figure A1. Timeline of China’s SOE reform
Source: Own construction based on Zuo (2016).

37
Figure A2. Gender-specific ratio of laid-off SOE workers over time
Notes: This bar chart illustrates the gender-specific ratio of laid-off workers from State-Owned Enterprises in
China between 1998 and 2001.
Source: China Labor Statistics Yearbook

38
Figure A3. Spatial distribution of the ratio of laid-off SOE workers
Notes: Each panel in the figure displays the spatial distribution of the average ratio of laid-off workers from State-
Owned Enterprises between 1998 and 2001. Panel A shows the distribution of the ratio for female workers, while
panel B displays the distribution of the ratio for male workers.
Source: China Labor Statistics Yearbook

39
Figure A4. Occupational categories by decile of occupational scores
Notes: Each occupation represents a decile of occupational scores, starting with the 10th percentile.
Source: Own construction based on the international version of the Socioeconomic Status Indicator (ISEI)
developed by Ganzeboom et al. (1992).

40
Figure A5. The differences in parental characteristics and bargaining power
Notes: The figure reports the point estimates and 90% confidence intervals of the coefficients of parental
characteristics difference between wives and husbands on variables indicating their bargaining power from OLS
regressions. Occupation difference means wife’s ISEI score minus that of husband. Education difference means
the difference of the years of education attained between wives and husbands. Income share means the share of
wife’s income over couple’s total income. Age difference means the difference of age between couples. Personality
difference means the difference of big five personality scores between couples. All variables indicating bargaining
power and parental characteristics difference are standardized. The observation units for regressions of house
ownership are individuals and the regressions control for couple’s hukou status, number of children, whether
cohabiting with parents, and county-by-cohort fixed effects. The standard errors are clustered at the county level.
Source: CFPS 2010

41
Figure A6. Density of parental occupation difference before and after sample restriction
Notes: The figure compares the distribution of samples before and after restriction, by excluding children whose
mother or father is a farmer.
Source: CGSS 2015

42
Figure A7. IV estimates of parental occupation difference on children’s statements about
gender role attitude using CGSS data
Notes: The figure reports the point estimates and 90% confidence intervals of the coefficients of parental
occupation difference on children's statement about gender role attitude from IV regressions. The control variables
include the child’s gender, ethnicity, education level, hukou status, parental education level, parental age difference,
the family’s social class when the child was 14 years old, and the county-by-cohort fixed effects. The standard
errors are clustered at the county level.
Source: CGSS 2015

43
Figure A8. IV estimates of parental occupation difference on children’s statements about
gender role attitude using CFPS data
Notes: The figure reports the point estimates and 90% confidence intervals of the coefficients of parental
occupation difference on children’s statement about gender role attitude from IV regressions. The instrumental
variables are four triple interaction terms between index of parental exposure to SOE reform and parental age (or
age squared) and an indicator of whether the parent worked when the children were 14 years old. The control
variables include the child’s gender, ethnicity, education level, hukou status, parental education level, parental age
difference, and the county-by-cohort fixed effects. The standard errors are clustered at the county level.
Source: CFPS 2014

44
Table A1. Measures of occupational status
Notes: This table summarizes the methods and particular factors or questions asked in the construction of the three
types of occupational status index.
Source: Ganzeboom et al. (1992); Treiman (1977); Li (2005).
Ind ex Method Relat ed fa cto rs o r que st ion s
1. Age: s ubject ’s age when surveying
2. Ed uc a t i o n : years of s ch oo ling
3. Income: pers onal income
Select ed su rvey ques tions are as follows :
1. Modified paired comparis on; o ccupa tion al t itles
presented in 31 sets of 6 and within each set ranked
according to “whi c h o c c upa ti o n h as t he be tt e r s o c i al
standing.”
2. Occu pation s rate d on 5-po int s s cale accord ing t o “
your own personal opinion of the g eneral s tanding that
such a job has”
3. Occu pation s ranke d acco rding to th eir “ pr e s t i g e ”
Bes ides age, education and income, they involve:
1. Power factor: whether they are the top/mid dle-
lev el/gras sroo ts -level managers of the o rganization .
2. Department factor: whether they are employed in
government agencies/public institutions or private
enterprise.
3. Social discrimination factors: whether th ey are in a
discriminated occupation.
Combin e s ub ject’s age, inco me an d
education level to estimate
standardized socioeconomic index for
the occupation.
Cond uc t s urv eys to gat her in div iduals ’
opinions about a specific occupation.
Similar to ISEI, it combines some socio-
economic fa ctors of certain ind ividu als
to estimate an occupation index.
Int ernational
Socioeco no mic
Ind ex (ISEI)
Standard
Int ernational
Occup ational
Pres tige Scale
(SIOPS)
Ch ine s e
Int ernational
Socioeco no mic
Ind ex (CISEI)

45
Table A2. Summary statistics of CFPS data
Source: CFPS 2010, 2012, 2014
NMean SD
Panel A: Key variables
Difference in ISEI sc ore (mother - fat her) 5761 -10.134 25.147
Difference in SIOPS s co re (mother - fath er) 5761 -10.101 23.975
Ge n d e r ro le a t t it u d e 5760 4.049 1.083
Panel B: Individual characteristics
Male (yes=1) 5761 4.049 1.083
Age 5761 0.537 0.499
Et hn icit y (han =1) 5761 41.817 16.513
Ye a rs o f s c h o o l in g 5761 0.942 0.234
Ye a rs o f s c h o o l in g ( fa t h e r ) 5761 9.793 4.21
Ye a rs o f s c h o o l in g ( mo t h e r) 5761 5.969 4.74
Urba n hukou (yes=1) 5761 0.516 0.5
Paren tal age d ifference (moth er - father) 5761 -2.328 3.996
Panel C: Household chores (hours )
Wife - husband (weekdays) 4773 1.392 1.951
Wife (weekdays) 4773 2.267 1.689
Hus ban d (weekday s ) 4773 0.875 1.174
Wife - husband (weekends) 4772 1.421 2.002
Wife (weekends) 4772 2.539 1.705
Hus ban d (weekend s ) 4772 1.117 1.322
Panel D: Asset Ownership
Hou s e o wners hip (wife) 4783 0.103 0.304
Hou s eh old ownership (hu sband ) 4783 0.511 0.5
Motorcycle 4599 0.373 0.484
Washing machine 4599 0.798 0.401
Panel E: Household expenditure (yuan )
Cigare ttes and alc oh ol 4599 81.628 189.13
Cloth in g 4599 2340.368 4261.899
Educat io n 4599 3489.118 8308.269
Co s met ic 4599 300.457 1185.412
Food 4599 16928.778 14506.819
Medical 4599 4119.709 10265.792
CFPS
Va r ia b l e

46
Table A3. The impact of SOE reform on parental occupation scores
Notes: Columns (1) to (3) report the first-stage results and the IV estimation. The dependent variables are the difference in parental occupational status measured in ISEI, SIOPS,
and CISEI, respectively. For parents with missing information on age, the parents’ ages are imputed using the average age of parents of children of the same age. Columns (4)
to (9) present the OLS estimates of the direct impact of IVs on mothers’/fathers’ occupation scores. The control variables include the child’s gender, ethnicity, education level,
hukou status, parental education level, parental age difference, and the family’s social class when the child was 14 years old. Robust standard errors in parentheses are clustered
at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
ISEI SIOPS CISEI
Father Mother Father Mother Fath er Mother
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Mother's exposure to SOE reform × age -87.111*** -60.023*** -110.742*** -17.235 -104.346*** -20.050 -80.074*** 0.137 -110.605***
(16.891) (16.915) (20.908) (15.772) (14.902) (14.545) (13.237) (13.363) (15.313)
Mother's exposure to SOE reform × age squared 1.321** * 0.922* * * 1.878* ** 0.376 1.698* ** 0.386 1.308*** 0.039 1.917** *
(0.337) (0.346) (0.414) (0.347) (0.295) (0.329) (0.262) (0.293) (0.287)
Father's expos ure to SOE reform × age 194.054 135.374 336.534 15.695 209.749 -8.051 127.323 51.650 388.185*
(178.453) (155.110) (224.348) (82.621) (165.785) (81.488) (140.042) (82.194) (229.061)
Father's expos ure to SOE reform × age squared -2.387 -1.644 -4.358 -0.441 -2.828 -0.097 -1.741 -0.771 -5.128*
(2.153) (1.856) (2.753) (0.998) (2.030) (0.970) (1.714) (1.027) (2.843)
Obs ervations 2,058 2,058 1,953 2,058 2,058 2,058 2,058 1,953 1,953
Mean -17.799 -16.396 -23.453 41.495 23.696 38.328 21.931 58.398 34.945
F- statistics of joint F- test 18.116 11.797 14.510 0.585 19.783 0.765 17.415 1.223 17.040
p-value o f Han sen J-statistic 0.780 0.854 0.717
Cont ro ls Ye s Ye s Ye s Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -co hort fixed effect s Ye s Ye s Ye s Ye s Ye s Yes Ye s Ye s Ye s
ISEI
SIOPS
CISEI
Mother - Father
Paren tal occu pation scores

47
Table A4. Impact on the probability of egalitarian gender role attitudes
Notes: The table reports the IV estimates of regressions on an indicator of egalitarian gender role attitudes, which
equals 1 if response to the statement “Husbands and wives should share household chores equally” is “totally
agree,” “somewhat agree,” or “neutral”; and 0 otherwise. The control variables include the child’s gender, ethnicity,
education level, hukou status, parental education level, parental age difference, and the family’s social class when
the child was 14 years old. Robust standard errors in parentheses are clustered at the county level. *** p<0.01, **
p<0.05, * p<0.1.
Source: CGSS 2015
(1) (2) (3)
Difference in ISEI sc ore (mother - fat her) 0.005**
(0.003)
Difference in SIOPS s co re (moth er - father) 0.008**
(0.003)
Difference in CISEI score (mother - father) 0.004**
(0.002)
Obs ervations 2,058 2,058 1,953
Mean 0.880 0.880 0.880
First-st age F-s tatis tics 18.116 11.797 14.510
p-value of Han s en J-s tatis tic 0.035 0.035 0.012
Cont ro ls Ye s Ye s Ye s
Coun ty -co hort fixed effect s Ye s Ye s Ye s
IV es timates
Ind icat or of egalitarian gender role attitudes

48
Table A5. Heterogeneous effects by sex ratio at birth
Notes: The table reports the first-stage and second-stage results of the IV estimates of the impact of difference in parental occupational status on children’s gender role attitude
by high and low sex ratio at birth in 1990. The sample is equally divided into two categories based on the sex ratio at birth in 1990 in each province, which is calculated using
2000 population census data. The dependent variables in Panel A are the difference of parental occupational status. The dependent variables in Panel B are children’s gender
role attitudes. The control variables include the child’s gender, ethnicity, education level, hukou status, parental education level, parental age difference, and the family’s social
class when the child was 14 years old. Robust standard errors in parentheses are clustered at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015, Population Census 2000
Lo w s ex ra tio High sex rat io Low s ex rat io High s ex ratio Lo w s ex rat io High s ex ratio
(1) (2) (3) (4) (5) (6)
Panel A: Firs t-stage results
Mother's exposure to SOE reform × age -112.975*** -220.927 -85.128*** -193.779 -143.704*** -520.316**
(20.775) (209.450) (21.594) (182.816) (24.803) (244.920)
Mother's exposure to SOE reform × age squared 1.851* ** 2.792 1.439*** 2.363 2.483** * 6.818**
(0.373) (2.607) (0.404) (2.307) (0.424) (3.164)
Father's expos ure to SOE reform × age 324.328* 277.601 240.517 260.361 546.647** 631.683
(177.067) (409.488) (170.975) (352.382) (272.337) (476.789)
Father's expos ure to SOE reform × age squared -4.232* -3.172 -3.168 -2.918 -7.154** -7.654
(2.218) (4.882) (2.151) (4.220) (3.367) (5.785)
Mean -18.642 -16.937 -17.438 -15.331 -25.177 -21.667
First-st age F -st atist ics 20.043 1.192 13.617 0.759 12.707 2.698
p-value of Han s en J st atistic 0.392 0.403 0.347 0.388 0.436 0.346
Panel B: Second-stage results
Difference in ISEI score (mother - father) 0.015* 0.039
(0.007) (0.030)
Difference in SIOPS s co re (moth er - father) 0.017* 0.041
(0.010) (0.039)
Difference in CISEI score (mother - fat he r) 0.013* 0.019
(0.007) (0.012)
Mean 3.856 3.873 3.856 3.873 3.854 3.882
Obs ervations 1,040 1,018 1,040 1,018 994 959
Cont ro ls Ye s Ye s Yes Ye s Ye s Yes
Coun ty -co hort fixed effe ct s Ye s Ye s Ye s Ye s Ye s Ye s
IV es timation of the impact of p arental occupation difference on children's gender role attitu de

49
Table A6. Robustness checks: control for occupational sex ratio
Notes: The table reports the IV estimates of the impact of difference in parental occupational status on children’s
gender role attitude for the robustness checks by further controlling for provincial sex ratio in the same
occupational category of each parent in 1990. The occupational sex ratio is calculated using 1990 population
census data. The control variables include the child’s gender, ethnicity, education level, hukou status, parental
education level, parental age difference, and the family’s social class when the child was 14 years old. Robust
standard errors in parentheses are clustered at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
(1) (2) (3)
Difference in ISEI sc ore (mother - fat her) 0.015** *
(0.005)
Difference in SIOPS s co re (mother - fa th er) 0.019***
(0.007)
Difference in CISEI score (moth er - father) 0.013**
(0.006)
Occupation sex ratio (mother) -0.000 -0.000 -0.000
(0.000) (0.000) (0.000)
Occupation sex ratio (father) 0.000 0.000 0.000
(0.000) (0.000) (0.000)
Obs ervations 2,058 2,058 1,953
Mean 3.864 3.864 3.868
First-st age F-s tatis tics 27.664 24.448 14.076
p-value o f Han sen J-s tatistic 0.762 0.802 0.775
Cont ro ls Ye s Ye s Ye s
Coun ty -co hort fixed effect s Ye s Ye s Ye s
Hus ban ds and wives s hould equ ally share hou sehold ch ores
IV es timates

50
Table A7. Robustness checks: control for parents’ marital status
Notes: The table reports the IV estimates of the impact of difference in parental occupational status on children’s
gender role attitude for the robustness checks by further controlling for whether parents are divorced. The divorced
variable is a binary variable indicating whether parents are divorced or not. The control variables include the
child’s gender, ethnicity, education level, hukou status, parental education level, and parental age difference.
Robust standard errors in parentheses are clustered at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CFPS 2014
(1) (2) (3) (4)
Difference in ISEI sc ore (mother - fat her) 0.012** 0.012**
(0.006) (0.006)
Difference in SIOPS s co re (moth er - father) 0.012** 0.012**
(0.006) (0.006)
Divo rced fa mily 0.151 0.162
(0.216) (0.215)
Obs ervations 5,761 5,761 4,534 4,534
Mean 4.049 4.049 4.019 4.019
First-st age F-s tatis tics 16.993 17.004 16.917 16.767
p-value of Han s en J-s tatis tic 0.439 0.417 0.376 0.363
Cont ro ls Ye s Ye s Ye s Ye s
Coun ty -co hort fixed effect s Ye s Ye s Ye s Ye s
Bas e lin e
Cont rol pare nt's marit al s tatu s
Hus ban ds and wives s hould equ ally share hou s eh old ch ores
IV es timates

51
Table A8. Robustness checks: control for the presence of siblings
Notes: The table reports the IV estimates of regressions which further control for the presence of siblings. The
other control variables include the child’s gender, ethnicity, education level, hukou status, parental education level,
and parental age difference. Robust standard errors in parentheses are clustered at the county level. *** p<0.01,
** p<0.05, * p<0.1.
Source: CFPS 2014
(1) (2) (3) (4)
Difference in ISEI sc ore (mother - fat her) 0.011** 0.011**
(0.006) (0.006)
Difference in SIOPS s co re (mother - fa th er) 0.012** 0.012**
(0.006) (0.006)
Having s ibling (y es =1) -0.072 -0.081
(0.080) (0.080)
Female × having sibling (yes=1) -0.199* -0.215**
(0.105) (0.105)
Male × having sibling (yes=1) 0.012 0.008
(0.102) (0.101)
Obs ervations 5,485 5,485 5,485 5,485
Mean 4.047 4.047 4.047 4.047
First-st age F-s tatis tics 12.612 12.431 12.621 12.425
p-value o f Han sen J-s tatistic 0.449 0.421 0.495 0.473
Cont ro ls Ye s Ye s Ye s Ye s
Coun ty -co hort fixed effect s Ye s Ye s Ye s Ye s
Hus ban ds and wives s hould equ ally share hou sehold ch ores
IV es timates

52
Table A9. Falsification tests
Notes: The sample includes children whose parents were both farmers. Columns (1) to (3) report the first-stage
results of IV estimation, and the last column reports the reduced-form estimates on gender role attitudes measured
by the survey question “Husbands and wives should equally share household chores.” The control variables
include the child’s gender, ethnicity, education level, hukou status, parental education level, parental age difference,
and the family’s social class when the child was 14 years old. Robust standard errors in parentheses are clustered
at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
Red u ced -fo rm
ISEI SIOPS CISEI
(1) (2) (3) (4)
Mother's exposure to SOE reform × age -0.052 0.007 -0.071 -0.326
(0.067) (0.008) (0.096) (0.537)
Mother's exposure to SOE reform × age squared 0.001 -0.000 0.001 0.004
(0.001) (0.000) (0.002) (0.011)
Father's expos ure to SOE reform × age 0.216 0.027 0.328 2.404
(0.233) (0.041) (0.332) (2.667)
Father's expos ure to SOE reform × age squared -0.003 -0.000 -0.004 -0.032
(0.003) (0.001) (0.004) (0.032)
Obs ervations 5,180 5,180 5,180 5,180
Mean -0.018 0.009 -0.022 3.767
First-st age F-s tatis tics 0.241 0.633 0.314
p-value o f Han sen J-statistic 0.886 0.672 0.867
p-values of joint F-t es t 0.719
Cont ro ls Ye s Ye s Ye s Yes
Coun ty -birt h ye ar fixed effec ts Ye s Ye s Ye s Yes
First st age
Paren tal occu pation differen ce
(moth er - father)
Ge n d e r ro le
attitudes

53
Table A10. Placebo test: the impact of SOE reform on gender role attitude across cohorts
Notes: The table reports the reduced-form estimates of IVs on children’s gender role attitude across different
cohorts. The division of cohort is based on age during the time window of SOE reform. Columns (1) to (4) report
the reduced-form estimates for samples aged 3-10, 12-14, 18-30, and 31-50, respectively, during the time window
of SOE reform. The variables of parental exposure to SOE reform in Column (2) are constructed in the same way
as Equation (2) in the text, whereas the variables of parental exposure to SOE reform in Columns (1), (3), and (4)
are constructed similar to by Equation (2) by replacing the age range with the corresponding age range of each
cohort. The control variables include the child’s gender, ethnicity, education level, hukou status, parental education
level, parental age difference, and the family’s social class when the child was 14 years old. Robust standard errors
in parentheses are clustered at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
3-10 12-14 18-30 31-50
(1) (2) (3) (4)
Mother's exposure to SOE reform × age -0.513 1.083** 2.334 -0.029
(0.575) (0.432) (1.689) (0.487)
Mother's exposure to SOE reform × age squared 0.003 -0.017** -0.017 -0.000
(0.003) (0.007) (0.012) (0.003)
Father's expos ure to SOE reform × age 0.962 -4.969 -2.200 -0.715
(0.710) (7.433) (2.699) (0.911)
Father's expos ure to SOE reform × age squared -0.005 0.057 0.015 0.004
(0.004) (0.065) (0.018) (0.005)
Obs ervations 595 332 683 1,001
Mean 3.908 3.919 3.843 3.833
p-values of joint F-t es t 0.318 0.018 0.359 0.554
Cont ro ls Ye s Ye s Ye s Ye s
Coun ty -birth ye ar fixed effec ts Ye s Ye s Ye s Ye s
Husbands and wives should equally share household chores

54
Table A11. Robustness checks: accounting for migration
Notes: The dependent variables are children’s g ender role attitudes. Columns (1) to (3) report the IV estimates by
excluding the migrant sample, and Columns (4) to (6) report the IV estimates from regressions that further control
for county by cohort by mover fixed effects. For parents with missing information on age, the parents’ ag es a re
imputed using the average age of parents of children of the same age. The control variables include the child’s
gender, ethnicity, education level, hukou status, parental education level, parental age difference, and the family’s
social class when the child was 14 years old. Robust standard errors in parentheses are clustered at the county
level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
(1) (2) (3) (4) (5) (6)
Difference in ISEI score (mother - fat her) 0.027** * 0.015* *
(0.009) (0.007)
Difference in SIOPS s co re (moth er - father) 0.035* ** 0.019*
(0.012) (0.010)
Difference in CISEI score (mother - fathe r) 0.034*** 0.016**
(0.010) (0.008)
Obs ervations 1,258 1,258 1,204 2,058 2,058 1,953
Mean 3.860 3.860 3.865 3.864 3.864 3.868
First-st age F-s tatis tics 13.900 10.650 8.245 18.890 13.450 11.205
p-value of Han s en J-s tatis tic 0.915 0.941 0.973 0.736 0.643 0.653
Cont ro ls Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -co hort fixed effect s Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -co hort-mov er fixed effec ts No No No Ye s Ye s Ye s
Non -migrat ing s ample
Int eracted with move s tatu s
Hus ban ds and wives s hou ld equ ally share h ou s eh old chores
IV es timates

55
Table A12. Robustness checks: without imputing age
Notes: The table reports the first-stage and second-stage results of the IV estimation without imputing parental
age when the information is missing. Columns (1), (3), and (5) report the first-stage results, and the dependent
variables are the difference of parental occupational status. Columns (2), (4), and (6) report the second-stage
results, and the dependent variables are children’s gender role attitudes. The control variables include the child’s
gender, ethnicity, education level, hukou status, parental education level, parental age difference, and the family’s
social class when the child was 14 years old. Robust standard errors in parentheses are clustered at the county
level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
first
stage
second
stage
first
stage
second
stage
first
stage
second
stage
(1) (2) (3) (4) (5) (6)
Mother's exposure to SOE reform × age
-88.012***
-60.548***
-114.502***
(19.728) (18.556) (22.948)
Mother's exposure to SOE reform × age squared 1.388* ** 0.983*** 2.034** *
(0.369) (0.360) (0.417)
Father's expos ure to SOE reform × age
360.130**
281.831*
574.316***
(168.061) (153.824) (221.464)
Father's expos ure to SOE reform × age squared -4.526** -3.528*
-7.487***
(2.054) (1.866) (2.716)
Difference in ISEI sc ore (mother - fat her) 0.013*
(0.008)
Difference in SIOPS s co re (moth er - father) 0.018*
(0.011)
Difference in CISEI score (mother - father) 0.007
(0.007)
Obs ervations 1,612 1,612 1,612 1,612 1,528 1,528
Mean -16.583 3.894 -15.328 3.894 -21.547 3.901
First-st age F-s tatis tics 12.196 9.175 10.381
p-value of Han s en J-s tatis tic 0.319 0.392 0.293
Cont ro ls Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -co hort fixed effect s Ye s Ye s Ye s Ye s Ye s Ye s

56
Table A13. Robustness checks: alternative specification of affected ages
Notes: The dependent variables are parental gender role attitudes. Columns (1) to (3) report the results of
specification which defines children’s age of 11 to 14 to be the affected age during the period of SOE reform.
Columns (4) to (6) report the results of specification which defines the ages of children between 13 and 14 to be
the affected ages. The control variables include the child’s gender, ethnicity, education level, hukou status, parental
education level, parental age difference, and the family’s social class when the child was 14 years old. Robust
standard errors in parentheses are clustered at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
(1) (2) (3) (4) (5) (6)
Difference in ISEI score (mother - father) 0.018* * 0.019* *
(0.007) (0.009)
Difference in SIOPS s co re (moth er - father) 0.023** 0.026**
(0.010) (0.012)
Difference in CISEI score (mother - fat he r) 0.014** 0.011
(0.006) (0.007)
Obs ervations 2,058 2,058 1,953 2,058 2,058 1,953
Mean 3.864 3.864 3.868 3.864 3.864 3.868
First-st age F-s ta tis tics 13.990 8.062 10.798 5.803 4.240 7.654
p-value of Han s en J-s tatis tic 0.858 0.869 0.787 0.470 0.596 0.395
Cont ro ls Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -co hort fixed effe ct s Ye s Ye s Ye s Ye s Ye s Ye s
Hus ban ds and wives s hou ld e qu ally share h ous ehold cho res
Children ag ed 13-14 d uring the
period of SOE reform
Children ag ed 11-14 d uring the
period of SOE reform
IV es timates

57
Table A14. Robustness checks: control for average score of parental occupational status
Notes: The table reports the IV estimates of regressions which further control for the average score of parental
occupational status in the benchmark specification reported in Table 2 in the text. The other control variables
include the child’s gender, ethnicity, education level, hukou status, parental education level, parental age difference,
and the family’s social class when the child was 14 years old. Robust standard errors in parentheses are clustered
at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
(1) (2) (3)
Difference in ISEI sc ore (mother - fat her) 0.017* *
(0.008)
Difference in SIOPS s co re (moth er - father) 0.023**
(0.010)
Difference in CISEI score (mother - father) 0.017*
(0.009)
Obs ervations 2,058 2,058 1,953
Mean 3.864 3.864 3.868
First-st age F-s tatis tics 12.902 8.527 8.857
p-value of Han s en J-s tatis tic 0.798 0.872 0.819
Cont ro ls Ye s Ye s Ye s
Coun ty -co hort fixed effect s Yes Ye s Ye s
Hus ban ds and wives s hould equ ally share h ou s eh old ch ores
IV es timates

58
Table A15. Robustness checks: control for other parental characteristics difference
Notes: The table reports the IV estimates of the impact of difference in parental occupational status on children’s
gender role attitude for the robustness checks by further controlling for other parental characteristics difference.
Other parental characteristics differences include difference in education years obtained, age difference, and
mother’s income share. The parents’ income is imputed based on the average income within the same occupational
category in the city using the 2005 population census data. The control variables include the child’s gender,
ethnicity, education level, hukou status, and the family’s social class when the child was 14 years old. Robust
standard errors in parentheses are clustered at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
(1) (2) (3) (4) (5) (6)
Difference in CISEI score (mother - fathe r) 0.015* * 0.027*
(0.006) (0.014)
Difference in SIOPS s co re (moth er - father) 0.020** 0.048
(0.008) (0.033)
Difference in CISEI score (mother - fathe r) 0.011* 0.050*
(0.006) (0.029)
Difference in education years (mo ther - fat he r) -0.012 -0.017 -0.016 -0.033 -0.003 -0.027
(0.018) (0.023) (0.020) (0.036) (0.017) (0.032)
Difference in age (mot he r - father) 0.004 -0.001 0.005 -0.005 0.007 0.002
(0.014) (0.016) (0.014) (0.020) (0.015) (0.021)
Imput ed moth er's income s hare -1.120 -2.093 -4.248
(0.799) (1.701) (2.627)
Obs ervations 2,058 1,975 2,058 1,975 1,953 1,870
Mean 3.864 3.872 3.864 3.872 3.868 3.876
First-st age F-s tatis tics 22.277 7.107 14.756 3.068 15.739 3.104
p-value of Han s en J-s tatis tic 0.767 0.506 0.848 0.524 0.744 0.768
Cont ro ls Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -co hort fixed effect s Ye s Ye s Ye s Ye s Ye s Ye s
Husbands and wives should equally share household chores

59
Table A16. Heterogeneous effects by gender
Notes: The dependent variables are children’s g ender role attitudes. Columns (1) to (4) report the OLS estimates
by gender, and Columns (5) to (8) report the IV estimates by gender. For parents with missing information on age,
the parents’ ag es are i mput ed using the average age of parents of children of the same age. The control variables
include the child’s gender, ethnicity, education level, hukou status, parental education level, parental age difference,
and the family’s social class when the child was 14 years old. Robust standard errors in parentheses are clustered
at the county level. *** p<0.01, ** p<0.05, * p<0.1.
Source: CGSS 2015
Female Male Female Male Female Male Female Male
(1) (2) (3) (4) (5) (6) (7) (8)
Difference in ISEI sc ore (mother - fat her) 0.004
0.013***
-0.033 0.010
(0.005) (0.005) (0.036) (0.014)
Difference in SIOPS s co re (moth er - father) 0.005
0.013***
-0.022 0.013
(0.006) (0.005) (0.023) (0.014)
Obs ervations 1,041 1,017 1,041 1,017 1,041 1,017 1,041 1,017
Mean 4.026 3.699 4.026 3.699 4.026 3.699 4.026 3.699
First-st age F-s tatis tics 0.546 3.527 0.786 2.279
p-value of Han s en J-s tatis tic 0.403 0.361 0.159 0.379
Cont ro ls Ye s Ye s Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -co hort fixed effect s Ye s Ye s Ye s Ye s Ye s Ye s Ye s Ye s
OLS
IV
Hus ban ds and wives s hould equ ally share h ou s eh old ch ores

60
Table A17. Intergenerational effects of marriage outcomes
Notes: In panel A, the dependent variables are whether the children were married or not. In panel B, the dependent
variables are whether the children were single or not. In panel C, the dependent variables are whether the children
were divorced or not. Columns (1) to (4) report the OLS estimates by gender, and Columns (5) to (8) report the
IV estimates by gender. For parents with missing information on age, the parents’ ag es are im pute d using the
average age of parents of children of the same age. The control variables include the child’s gender, ethnicity,
education level, hukou status, parental education level, parental age difference, and the family’s social class when
the child was 14 years old. Robust standard errors in parentheses are clustered at the county level. *** p<0.01, **
p<0.05, * p<0.1.
Source: CGSS 2015
Female Male Female Male Female Male Female Male
(1) (2) (3) (4) (5) (6) (7) (8)
Difference in ISEI score (mother - father) 0.000 -0.002 -0.035 0.002
(0.002) (0.001) (0.030) (0.006)
Difference in SIOPS s co re (moth er - father) 0.001 -0.002 -0.026 0.002
(0.002) (0.002) (0.019) (0.006)
Mean 0.683 0.656 0.683 0.656 0.683 0.656 0.683 0.656
p-value of Han s en J-s tatis tic 0.540 0.726 0.265 0.696
Difference in ISEI score (mother - father) 0.001 0.001 0.031 -0.002
(0.002) (0.002) (0.025) (0.005)
Difference in SIOPS s co re (moth er - father) 0.001 0.001 0.023 -0.003
(0.002) (0.002) (0.015) (0.006)
Mean 0.233 0.312 0.233 0.312 0.233 0.312 0.233 0.312
p-value of Han s en J-s tatis tic 0.803 0.697 0.621 0.672
Difference in ISEI score (mother - father) -0.001 0.001 0.002 0.000
(0.001) (0.001) (0.004) (0.003)
Difference in SIOPS s co re (moth er - father) -0.001 0.001 0.001 -0.001
(0.001) (0.001) (0.002) (0.003)
Mean 0.035 0.041 0.035 0.041 0.035 0.041 0.035 0.041
p-value of Han s en J-s tatis tic 0.584 0.323 0.487 0.437
Obs ervations 1,045 1,022 1,045 1,022 1,045 1,022 1,045 1,022
Cont ro ls Ye s Ye s Ye s Ye s Ye s Ye s Ye s Ye s
Coun ty -co hort FE Ye s Ye s Ye s Ye s Ye s Ye s Ye s Ye s
First-st age F-s tatis tics 0.423 3.527 0.606 2.279
Panel A: Married/Unmarried
Panel B: Single/Non-single
Panel C: Divorced/Un-divorced
OLS
IV

61
Appendices Data and Measures
B.1 Occupational Status Score
The term occupational status is widely used in sociological and other social science studies. Three
widely recognized indices used in sociological studies are the International Socioeconomic Index (ISEI),
the Standard International Occupational Prestige Scale (SIOPS), and the Chinese International
Socioeconomic Index (CISEI).
As indicated by Appendix Table A1, the ISEI is an international version of the Socioeconomic
Status Indicator (SEI) originally developed by Ganzeboom et al. (1992). It measures an occupation’s
socioeconomic status based on its educational requirements, income levels, and age distribution. Each
occupation is assigned a score ranging from 16 (lowest socioeconomic status) to 90 (highest
socioeconomic status). Conceptually, the ISEI index measures how much an occupation’s
characteristics translate into a person’s income based on their education level.
To provide a practical perspective, we offer examples of occupational categories with different
scores of occupational status. Appendix Figure A4 displays the examples of occupations and their
corresponding status scores for each decile of the distribution of occupational scores measured by ISEI,
beginning with the 10th percentile from the bottom to up. Occupations located at the 10th percentile
score include forestry workers and loggers, with scores gradually increasing for occupations such as
handicraft workers, chemical-products machine operators, electrical-equipment assemblers, precision
instrument makers and repairers, electrical engineering and technicians, computer equipment operators,
and senior officials. Department managers of larger enterprises represent one example of occupations,
with scores among the 90th percentile.
The SIOPS is an internationally comparable measure of occupational prestige. It is assessed by
survey respondents through a series of questions to evaluate occupations according to their social
standing on a scale of 0-100, with a higher score indicating a higher level of prestige (Treiman, 1977).
These questions include inquiries such as “Which occupation do you believe holds a superior social
standing?” and “What is your personal opinion of the general standing of such a job?” (see Table A1
for details). This index captures individuals’ subjective perceptions of an occupation’s social
importance, value, contribution, respect, and admiration. Hence, the SIOPS index reflects more of the

62
subjective dimension about an occupation compared to ISEI which incorporates objective indicators of
occupational status.
To better capture occupational status in the Chinese context, we also use the CISEI developed by
Li (2005). This index is constructed using national survey data from China, akin to the ISEI measure,
but with a distinct focus on factors relevant to the Chinese context. Specifically, in addition to age,
education, and income, this measure considers three additional factors, as shown in Appendix Table A1:
the power factor (indicating whether individuals hold a managerial position), the department factor
(indicating whether they are employed by government or a public institution), and the social
discrimination factor (indicating whether they occupy a position subject to social discrimination). By
incorporating these unique elements, the CISEI provides a more appropriate and nuanced measure of
occupational status in the Chinese context. We manually merged the occupational information from our
dataset to the CISEI. However, since the CISEI covers only 161 occupations, some occupations in the
data have missing information on the CISEI index, which may result in a reduction in sample size.
Therefore, we use only the CISEI to provide complementary evidence for the baseline analyses.
B.2 Controls of Confounding Policies and Shocks
In this section, we describe the details of the construction of variables to measure the confounding
policies and shocks discussed in Section 5.2.
College Expansion. We include the following triple interaction term to capture the potential
influence of college expansion on individuals’ gender role attitudes:
𝑅𝐻𝐸!"#$$% ×𝑡𝑟𝑒𝑎𝑡&×𝑔𝑒𝑛𝑑𝑒𝑟&
where
IJ;'(%112
denotes the proportion of university enrollment quota relative to the number of high
school graduates in province
3
in 1998, one year before the start of college expansion. Regions with
higher ratios suggest a greater historical availability of higher education resources and, subsequently,
greater intensity of the college expansion. The variable
"5&4"!
is an indicator of individuals’ exposure
to the college expansion, which equals 1 if they were aged 18 or younger in 1999 and 0 otherwise. Si
(2022) documented gender heterogeneity in the impact of college expansion on the development of
gender role attitudes. To account for this, we also incorporate a dummy variable,
F&7%&5!
, in the
construction, where
F&7%&5!
takes a value of 1 for males, and 0 for females.

63
Asian Financial Crisis. To capture the potential influence of the Asian financial crisis, we
incorporate two triple interaction terms:
𝐹𝐷𝐼!×𝑆𝑂𝐸_𝑡𝑟𝑒𝑎𝑡"×𝑎𝑔𝑒#$%&
'#()*+
𝐹𝐷𝐼!×𝑆𝑂𝐸_𝑡𝑟𝑒𝑎𝑡"×𝑎𝑔𝑒#$%&
,-()*+
where
KL?'
represents the ratio of foreign direct investment (FDI) to GDP for province
3
in 1996,
the year preceding the onset of the Asian financial crisis. This ratio serves as an indicator of regional
disparities in exposure to the Asian financial crisis. Additionally,
9:;M"5&4"!
denotes whether
individuals were aged 12-14 during the period of SOE reform. It is worth noting that this construction
differs from how we establish our IV, and we do not consider the cumulative effect during the SOE
reform time window. The variables
4F&*,%+
3*#456789#456
refer to parental age when their children were
aged 14.
China’s accession to the WTO. We account for the potentially confounding effects of China’s
WTO accession by introducing the following two triple interaction terms, analogous to our approach
for examining the impact of the Asian financial crisis:
𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒!×𝑆𝑂𝐸_𝑡𝑟𝑒𝑎𝑡&×𝑎𝑔𝑒'(#)
*'+,-.
𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒!×𝑆𝑂𝐸_𝑡𝑟𝑒𝑎𝑡&×𝑎𝑔𝑒'(#)
/0+,-.
where
%#N"472&'
denotes the distance between each province p and the nearest port, which reflects
regional variations in exposure to WTO shocks. The other variables maintain the same definitions as
those employed for the analysis of the Asian financial crisis.