Does Husband's Education Level Affect Wife's Employment
Participation? Mediation Effect Test based on CGSS
Yang Shi
Department of Accounting Shandong Youth University of Political Science Jinan, 250103, China
Keywords: Husband's Education Level, Gross Monthly Income, Wife's Employment Participation, Mediation Effect Test,
Big Data Statistics Analysis.
Abstract: In reality, women's employment participation is affected by various factors. Using the data of CGSS, through
the micro-econometric model, we use computer statistical software—Stata, to test the effect of the husband's
education level on the wife's employment participation through the mediating variable--personal gross
monthly income. The study shows that: the husband's education level has a significant positive impact on his
gross monthly income; only when the husband's education level is above college, the mediating effect is
established, that is, only the husband has received above college education, then can positively affect wife's
employment participation. In order to increase the employment participation of wives, the joint efforts of
many parties are needed.
1 INTRODUCTION
In reality, married women's employment participation
is affected by various factors (Xiang 2019). With the
gradual improvement of the education level of indi-
viduals in modern society, couples are also more
well-matched in education (Zheng 2020). This article
takes the husband's education level as the starting
point, focuses on the wife's employment participa-
tion, and uses micro-data and econometric methods to
explore the relationship between the husband's edu-
cation level and the wife's employment participation,
then tries to answer the following questions:
First, does the husband's education level affect his
personal income?
Second, does the husband's education level affect
the wife's employment participation through his per-
sonal income?
The labor force participation rate of the female
population in our country has always remained at a
relatively high level
(Zheng 2020), therefore, the em-
ployment participation of women, especially married
women, has an important impact on economic devel-
opment. Relying on big data, using computer statisti-
cal software to study the most basic micro-family de-
cision-making on employment participation of cou-
ples, has important practical significance for improv-
ing labor market policies and promoting true equality
between men and women.
2 VARIABLE SELECTION AND
MODEL BUILDING
2.1 Data Sources
The article selects the data of the Chinese General So-
cial Survey (CGSS) in 2015， which started in 2003
and is the earliest national, com-prehensive and con-
tinuous academic survey project in China, covering
28 provinces (municipalities, autonomous Region),
involving information at multiple levels such as soci-
ety, community, family, and individual. The survey
objects are adults over the age of 18. The total number
of samples in CGSS2015 is 10,968. According to the
research purpose of the article, married males were
selected as the sample, and the samples with missing
key variables in the survey year were excluded, then
finally 2171 samples were obtained.
Shi, Y.
Does Husbandâ
˘
A
´
Zs Education Level Affect Wifeâ
˘
A
´
Zs Employment Participation? Mediation Effect Test Based on CGSS.
DOI: 10.5220/0011897900003613
In Proceedings of the 2nd International Conference on New Media Development and Modernized Education (NMDME 2022), pages 51-56
ISBN: 978-989-758-630-9
Copyright
c
 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)
51

2.2 Variable Selection
2.2.1 Wife's Employment Participation
Wife's employment participation is the main ex-
plained variable. In CGSS2015, the question concern-
ing wife's employment participation is: “What was
the employment status of your spouse or common-
law partner in the last week?” There are four options.
Excluding the samples that selected items 2 and 3,
items 1 and 4 are reserved as the basis for assignment
of employment participation. If you select item 1 “not
engaged in any work for the purpose of obtaining eco-
nomic income”, the wife's employment participation
is assigned a value of 0; If you select the item 4 “en-
gaged in work for the purpose of obtaining economic
income (including joining the army)”, then the wife's
employment participation is assigned a value of 1,
that is, the explained variable is a binary variable.
2.2.2 The Education Level of the Husband
The core explanatory variable of the article is the ed-
ucation level of the husband. The highest level of ed-
ucation (including current students) of the sample was
selected as the education level of the husband, and it
was divided into seven levels, and the illiterate group
was used as the control group.
2.2.3 Husband’s Gross Monthly Income
According to the research purpose of the article, the
gross monthly income (GMI) of the sample was se-
lected as the mediating variable, which is obtained by
dividing the annual total income by 12, and the loga-
rithm of the GMI is smoothed as the mediating varia-
ble of the model.
2.2.4 Control Variables
In addition to the above, the article selects the sam-
ple's monthly working hours (weekly working
hours*4), age, nationality, household registration, po-
litical status, and work ownership as control varia-
bles.
2.3 Definition of Each Variable and
Descriptive Statistics
The descriptive statistics of each variable (except for
control variables) are shown in the following table 1.
Table 1. Descriptive statistics of variables
Variable Definition and Assignment Min Max Mean
Number of
Samples
Wife's
employ-
ment
participa-
tion
Engaged in work for the purpose of obtaining eco-
nomic income in the past week, assign a value of 1
0 1 0.59
1283
Not engaged in any work for the purpose of eco-
nomic income in the last week, assign a value of 0
888
Primary
school
Primary school is 1; non-primary school is 0 0 1 0.22 476
Junior
high
school
Junior high school is 1; non-junior high school is 0 0 1 0.35 753
High
school
High school is 1, including vocational high school,
ordinary high school, technical secondary school
and technical school; non-high school is 0
0 1 0.21 459
College
College is 1, including college (adult higher educa-
tion), college (formal higher education); non-col-
lege is 0
0 1 0.08 180
Under-
graduate
Undergraduate is 1, including undergraduate (adult
higher education), undergraduate (formal higher ed-
ucation); non-undergraduate is 0
0 1 0.08 170
Postgrad-
uate
Postgraduate or above is 1; non-postgraduate or
above is 0
0 1 0.01 22
GMI lo
g
Gross annual income/12 3.51 13.63 7.68 2171
Note: In the above table, except for the wife's employment participation in the first row, the rest of the variables are descriptive statistics of
the husband's sample.
NMDME 2022 - The International Conference on New Media Development and Modernized Education
52

According to Table 1, among the screening sam-
ples, the employment participation rate of wives is
only 59.1%, which is not particularly high. The pos-
sible reason for this situation is that the age of the
wives of the sample is not specially screened, that is,
a sample of all age groups that meet the conditions is
selected, including retirement. The reason for this
treatment is that although the current retirement age
in our country is 60 years old for men and 55 years
old for women, because the question is “spouse's
work situation in the past week”, even if the wife has
retired, The short-term employment situation of re-
turning part-time jobs is also common, so the age
screening of wives is not considered here, and a sam-
ple of all ages is used.
The highest educational background of the se-
lected sample is mainly junior high school and high
school, and only 372 have received college education
or above, accounting for only 17.13% of the whole
sample.
There are only 22 with a postgraduate degree
or above, accounting for only 1%. This shows that alt-
hough our country's university enrollment has been
expanding, the proportion of people who have re-
ceived college education or above is still relatively
small, especially the shortage of high-level talents
above postgraduate level.
Other control variables. The samples are mainly
Han nationality, and the age group is mainly young
and middle-aged; the agricultural household registra-
tion samples are mostly; the party members are few;
especially the samples engaged in “good jobs within
the system” in the traditional sense are few.
2.4 Model Building
According to the research purpose of the article and
the practice of Wen Zhonglin and Ye Baojuan (2014)
(Wen, Ye, 2014), the following model(1)(2)(3) are
constructed for verification:
𝑊𝑜𝑟𝑘=𝛼

+𝛽

𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛+𝛾

𝑋

+𝜇

(1)
𝑌=𝛼

+𝛽

𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛+𝛾

𝑋

+𝜇

(2)
𝑊𝑜𝑟𝑘=𝛼

+𝛽

𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛+𝛿𝑌 + 𝛾

𝑋

+𝜇

(3)
According to model (1)(2)(3), 𝑊𝑜𝑟𝑘 represents
the employment participation of the wife, which is the
explained variable, and is a binary variable; 𝑌 repre-
sents the mediating variable, the GMI of the husband
(logarithmic form), 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 represents the ex-
planatory variable--the education level of the hus-
band, which is a multi-category variable; 𝑋

repre-
sents the control variable. 𝛽

,𝛾

,𝛿 are parameters to
be estimated, 𝜇

is random errorterm, and 𝛼

is a
constant term.
Wen Zhonglin and Ye Baojuan (2014) (Wen, Ye,
2014) pointed out that the mediation effect test should
be tested in sequence first, that is, for the above
model, if at least one of 𝛽

and δ is found to be insig-
nificant, the Sobel test is required. If both 𝛽

and 𝛿
are significant, the result of the sequential test is
stronger than that of the Bootstrap method. Therefore,
the standards for judging the mediation effect of the
above model are: the explained variable 𝑊𝑜𝑟𝑘 has a
regression on the explanatory variable 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛,
and the coefficient 𝛽

is significant; the mediating
variable 𝑌 has a regression on the explanatory varia-
ble 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛, the explained variable 𝑊𝑜𝑟𝑘 has a
regression on the mediating variable 𝑌 and the ex-
planatory variable 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛, if the coefficients 𝛽

and 𝛿 are both significant, it means that there is a me-
diating effect.
3 METHODS: ECONOMETRIC
TEST OF THE MEDIATION
EFFECT
According to the above-mentioned mediation effect
test method, this paper intends to use Stata to test the
significance of the regression coefficients of the con-
structed models respectively. Stata is a set of com-
puter statistics software that relies on big data to ana-
lyze data, manage data, and draw professional charts
for users. Stata is powerful and a very convenient
computer software.
3.1 A Test of the Effect of Husband's
Education Level on Wife's
Employment Participation
Use binary logistic regression to test the significance
of the regression coefficients of the model (1). The
regression results are as follows (Table 2):
Does Husbandâ
˘
A
´
Zs Education Level Affect Wifeâ
˘
A
´
Zs Employment Participation? Mediation Effect Test Based on CGSS
53

Table 2. Logistic regression results of model (1)
Ex
p
lanator
y
Variable Coefficient Standard Erro
r
Ex
p
lanator
y
Variable Coefficient Standard Erro
r
Primary school
-0.367*
（2.827）
0.218 Age
-0.015***
（12.708）
0.004
Junior high school
-0.342
（2.446）
0.219 Nationality
-0.240
（2.045）
0.168
High school
-0.160
（0.465）
0.235
Household
registration
-0.032
（0.073）
0.118
College
0.494*
（2.945）
0.288 Political status
-0.187
(1.770)
0.141
Undergraduate
0.738**
（5.869）
0.304 Ownership
-0.303**
（4.749）
0.139
Postgraduate
1.419**
（4.431）
0.674 Constant
1.299***
（11.678）
0.380
Monthly working
hours
0.001**
（3.822）
0.001
Hosmer and
Lemeshow test
0.375
Note: The data in parentheses are wals values, ***, **, * indicate significance at the 1%, 5%, and 10% confidence levels, respectively.
For the Hosmer and Lemeshow test of the above
binary logistic regression, the P value is 0.375>0.05,
indicating that the model (1) has a good fit. The re-
gression results show that, for the education level of
the husband, primary school,college, undergraduate
and postgraduate all have a significant impact on the
wife's employment participation. The difference is
that the education level of the sample with primary
school has a significant negative impact on the wife's
employment participation; while the education level
of the sample of college, undergraduate and postgrad-
uate students has a significant positive impact on the
wife's employment participation, and the higher the
husband's education level, the greater the impact on
the wife's employment participation.
For the remaining control variables, the husband's
monthly working time has a significant positive im-
pact on the wife's employment participation.How-
ever, the increase of the husband's age will have a sig-
nificant negative impact on the wife's employment
participation. The ownership of the husband’s work
also has a negative impact on the wife’s employment
participation. Finally, the husband's nationality and
household registration have no significant effect on
the wife's employment participation.
3.2 A Test of the Effect of Husband's
Education Level on Husband's
GMI
Apply multiple regression analysis to test the signifi-
cance of the regression coefficients of the model (2).
The regression results are as follows (Table 3):
Table 3. Multiple regression results of model (2)
Explanatory Variable Coefficient Standard Error
Explanatory
Variable
Coefficient Standard Error
Primary School
0.151
（1.562）
0.097 Age
-0.024***
（-12.592）
0.002
Junior high school
0.557***
（5.750）
0.097 Nationality
0.486***
（6.705）
0.073
High school
0.786***
（7.570）
0.104
Household
registration
0.395***
（7.722）
0.051
College
1.054***
（8.536）
0.123 Political status
0.042
（0.688）
0.061
Undergraduate
1.317***
（10.286）
0.128 Ownership
0.034
（0.568）
0.060
Postgraduate
1.629***
（7.302）
0.223 Constant
7.402***
（44.509）
0.166
Monthly working hours
0.001***
（3.827）
0.000 Adjust R
2
0.352
Note: The data in parentheses are t values, ***, **, * indicate significance at the 1%, 5%, and 10% confidence levels, respectively.
NMDME 2022 - The International Conference on New Media Development and Modernized Education
54

The above regression results show that, except for
primary school, the husband's education level has a
significant positive impact on the husband's GMI, and
the higher the education level, the greater the impact
on the GMI. For example, compared with the illiterate
sample, the sample with a high school education has
a 78.6% increase in GMI, while the sample with a col-
lege education, the value is 105.4%, and at the post-
graduate level, the value rises to 162.9%, it can be
seen that higher education level especially above
graduate level has a great impact on GMI.
For other control variables, the effect of monthly
working hours on GMI is positive, The effect of age
on GMI is negative. Compared with the minorities,
the GMI of the Han nationality is higher; the sample
income of the non-agricultural household registration
is higher than that of the agricultural household reg-
istration, which is consistent with the actual situation.
Finally, The effect of political status and ownership
on GMI is not significant.
3.3 A Test of the Effect of Husband's
Education Level and GMI on
Wife's Employment Participation
Use binary logistic regression to test the significance
of the regression coefficients of the model (3). The
regression results are as follows (Table 4):
Table 4. Logistic regression results of model (3)
Explanatory Variable Coefficient Standard Error Explanatory Variable Coefficient Standard Error
Primary school
-0.354
（2.630）
0.218 Monthly working hours
0.001**
（4.418）
0.001
Junior high school
-0.294
（1.781）
0.220 Age
-0.017***
（15.335）
0.004
High school
-0.092
（0.150）
0.238 Nationality
-0.197
（1.342）
0.170
College
0.586**
（4.019）
0.292
Household
registration
0.004
（0.001）
0.120
Undergraduate
0.853**
（7.520）
0.311 Political status
-0.183
（1.689）
0.141
Postgraduate
1.561**
（5.297）
0.678 Ownership
-0.299**
（4.604）
0.139
GMI log
-0.089*
（3.270）
0.049 Constant
1.962***
（13.747）
0.529
Hosmer and
Lemeshow test
0.895
Note: The data in parentheses are wals values, ***, **, * indicate significance at the 1%, 5%, and 10% confidence levels, respectively.
It can be seen from the above table that the P value
of the Hosmer and Lemeshow test is 0.895>0.05, in-
dicating that the model (3) has a good fit. The core
explanatory variable--the husband's education level,
has a significant positive impact on the wife's em-
ployment participation above the college, and the
higher the education level, the greater the impact,
which is consistent with the regression results of
model (1). The core explanatory variable logarithm of
husband's GMI has a significant negative impact on
wife’s employment participation.
For other control variables, both husband's age
and work ownership have significant negative effects
on wife’s employment participation; while household
registration and political status have no significant ef-
fects on wife’s employment participation.
To sum up, if the education level of the husband
is in three stages: college, undergraduate and post-
graduate, the regression results of the above three
models are: the explanated variable 𝑊𝑜𝑟𝑘 has a re-
gression on the explanatory variable 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛, and
the coefficient 𝛽

is significant; the mediator varia-
ble 𝑌 has a regression on the explanatory varia-
ble 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 , the coefficient 𝛽

is significant; the
explanated variable 𝑊𝑜𝑟𝑘 is regressed on the medi-
ating variable 𝑌 and the explanatory variable
𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛, and 𝛿 is significant, so according to the
sequential test method of the mediating effect, it can
be seen that the mediating effect of the husband’s
GMI exists, that is, the husband's education level can
affect the wife's employment participation through
the mediating variable of the husband's GMI, and the
Does Husbandâ
˘
A
´
Zs Education Level Affect Wifeâ
˘
A
´
Zs Employment Participation? Mediation Effect Test Based on CGSS
55

husband's education level has a significant positive
impact on the wife's employment participation.
4 CONCLUSIONS AND
SUGGESTIONS
Relying on big data and using Stata computer statisti-
cal software to complete the mediation effect test, we
can draw the following conclusions: only when the
husband's education level is above college, the GMI
of the mediating variable can positively affect the
wife's employment participation.
Based on the further discussion above, in order to
increase the wife's employment participation, the fol-
lowing suggestions are put forward:
Increase investment in education. In modern soci-
ety, the employment participation of wives is not only
related to their own level of education, but also has a
significant positive correlation with their husbands’
education level. Therefore, in order to improve the
employment participation of women, the fundamental
is to improve the education level of the whole people,
especially to increase the investment in education at
the college level and above. Although our country has
expanded its higher education enrollment since 2000,
compared with developed countries, the proportion of
people with higher education in our country is still
small, and even fewer have received postgraduate ed-
ucation. Relevant data from the official website of the
Ministry of Education shows that in 2021, the number
of ordinary undergraduates and college graduates was
8.265 million. There are 10.013 million enrollments
and 34.961 million students in school
(Ministry of Ed-
ucation, 2021), and even fewer postgraduate students.
However, the total number of people in our country
in 2021 is about 1.41 billion. It can be seen that our
country's higher education investment needs to be
further improved.
Destroy gender social norms. In addition to edu-
cational level, women's employment participation is
effected by a variety of factors. One of the most im-
portant points is the social norms of gender. The tra-
ditional gender social norm advocates that “men take
charge of the outside, and females take care of the in-
side”, When roles conflict with the work, based on the
theory of gender dominance, women should give up
the work and return to their families (Cheng, He,
2017). However, with the progress of society, The
concept of equality between men and women is grad-
ually recognized. In the workplace, the government,
society and enterprises should try their best to elimi-
nate the “glass ceiling” for women's career advance-
ment, break down barriers to women's professional ti-
tle promotion, and create a good working space for
women; in the family, it is necessary to advocate the
sharing of housework by husbands and wives, free
women's productivity from housework, etc., then pro-
mote women's employment participation, especially
married women.
This research was supported by a project of social
science research of Shandong Youth University of
Political Science (Grant No.: SJYBXM202212)
REFERENCES
Ministry of Education. The main results of the national ed-
ucation statistics in 2021 [EB/OL].
http://www.moe.gov.cn/jyb_xwfb/gzdt_gzdt/s5987/20
2203/t20220301_603262.html, 2022-03-01.
Yixuan Cheng, Xiaobo He. First marriage age, female em-
ployment and family status: Evidence from Chinese
surveys (1990, 2000, 2010) [J]. Review of Industrial
Economics, 2017(06): 60-76.
Yunfan Xiang. Gender Employment Willingness Differ-
ences in China's Labor Market Influencing Factors and
Trends [J]. Wuhan University of Technology (Social
Science Edi-tion), 2019(05): 62-69.
Zhenzhen Zheng, Progression of Women’s Status in
Health, Education and Employment Reflected in Cen-
sus Data [J]. Journal of Shandong Women’s University,
2020(04): 1-7.
Zhonlin Wen, Baojuan Ye. Analyses of Mediating Effects:
The Development of Methods and Models [J]. Ad-
vances in Psychological Science, 2014(05): 731-745.
NMDME 2022 - The International Conference on New Media Development and Modernized Education
56