Mental health penalties of having a child: findings from the China family panel studies

In recent years, the birth rate in China has rapidly declined. While much research has been done on the penalties in earnings that women incur when they fall behind men in the labor market due to childbirth, there has been little to no research on the mental health effects. This study addresses the gap in current literature by examining the mental health penalties that women experience after having a child in comparison to men. We applied econometric modeling to data collected from China Family Panel Studies (CFPS) and found that women experienced a significant, immediate, and long-run decline (4.3%) in life satisfaction after their first child, while men were unaffected. We also found that women experienced a significant increase in depression after their first child. This suggests mental health penalties since the mental health risk proxied by these two measurements is only significant for women. This is likely related to child penalties in labor market performance and childbirth-related physical health issues. When countries adopt multiple tools to stimulate the birth rate for economic growth, they must consider the implicit burden on women—especially the long-term negative effects on mental health.

Each dot represents a regression coefficient and the corresponding vertical line (error bar) represents the 95% confidence interval. Schooling years, age-fixed effects, city-fixed effects, year-fixed effects, province-year-fixed effects, the yea-of-survey-fixed-effects, and the month-of-survey-fixed effects are included. When the 0 from y axis is not covered by the vertical line, it signifies that the coefficient is significant and different from zero. The dashed gray line in event 0 signifies the year of arrival of the first child.

Supplementary Table 1. Impact of childbirth on life satisfaction for women with additional controls
This table shows the results of event study for women. The dependent variable is life satisfaction. The independent variables are event time dummies that are identical to those in Table 2. In additional to the controls in Table 2, column 1 further controls for the number of children and the number of sons, column 2 further controls for whether they have borne at least two children and their age when the last child was born. In column 3, we restrict the sample to those who have borne at least two children and control for their age when the second child was born. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01. (1) (2)  Table 2, column 1 further controls for the number of children and the number of sons, column 2 further controls for whether they have borne at least two children and their age when the last child was born. In column 3, we restrict the sample to those who have borne at least two children and control for their age when the second child was born. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01.  (1) and (2), we excluded those who were only surveyed once in 5 waves. In columns (3) and (4), we only include those who gave birth to the first child after 2010. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 was omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, year-fixed effects, province-yearfixed effects, the year-of-survey-fixed effects, and the month-of-survey-fixed effects were included. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01. This table shows the results of event study with the dependent variable being depression. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 was omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, year-fixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the month-of-survey-fixed effects were included. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01.

Supplementary Table 5. Impact of childbirth on life satisfaction and depression for women and men (weights adjusted)
(1) This table shows the results of event study with sampling weights adjusted for women (columns 1-2) and men (columns 3-4) respectively. The dependent variables include life satisfaction and depression. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 was omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, year-fixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the-month-of-survey-fixed effects were included. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01.

Supplementary Table 6. Impact of childbirth on chronic disease for women and men
(1)  (1) and (2) show the impact of having a child on chronic disease for men and women, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, yearfixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the month-of-surveyfixed effects are included. To test whether the coefficients by subgroups are significantly different, we use the test drawing on seemingly unrelated regressions with the command suest in Stata. *p < 0.1, **p < 0.05, and ***p < 0.01, 95% confidence intervals are in brackets.  (1) and (2) show the impact of having a child on self-reported physical health status (improved health status is equal to 1) for men and women, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, year-fixed effects, province-year-fixed effects, the year-of-surveyfixed effects, and the month-of-survey-fixed effects are included. To test whether the coefficients by subgroups are significantly different, we use the test drawing on seemingly unrelated regressions with the command suest in Stata. *p < 0.1, **p < 0.05, and ***p < 0.01, 95% confidence intervals are in brackets.

Supplementary Table 8. Impact of childbirth on hospitalization for women and men
(1)  (1) and (2) show the impact of having a child on hospitalization (equals to 1 if hospitalized last year) for men and women, respectively. The independent variables are event time dummies, indexed such that u_bt mean t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. Cityfixed effects, year-fixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the month-of-survey-fixed effects are included. To test whether the coefficients by subgroups are significantly different, we use the test drawing on seemingly unrelated regressions with the command suest in Stata. *p < 0.1, **p < 0.05, and ***p < 0.01, 95% confidence intervals are in brackets.

Supplementary Table 9. Impact of childbirth on employment for women and men
(1)  (1) and (2) show the impact of having a child on employment (equals to 1 if employed) for men and women, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, year-fixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the monthof-survey-fixed effects are included. To test whether the coefficients by subgroups are significantly different, we use the test drawing on seemingly unrelated regressions with the command suest in Stata. *p < 0.1, **p < 0.05, and ***p < 0.01, 95% confidence intervals are in brackets.  (1) and (2) show the impact of having a child on hours worked per week for men and women, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, yearfixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the month-of-surveyfixed effects are included. To test whether the coefficients by subgroups are significantly different, we use the test drawing on seemingly unrelated regressions with the command suest in Stata. *p < 0.1, **p < 0.05, and ***p < 0.01, 95% confidence intervals are in brackets.

Supplementary Table 11. Impact of childbirth on income for women and men
(1)  (1) and (2) show the impact of having a child on income (10,000 Yuan) men and women, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, yearfixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the month-of-surveyfixed effects are included. To test whether the coefficients by subgroups are significantly different, we use the test drawing on seemingly unrelated regressions with the command suest in Stata. *p < 0.1, **p < 0.05, and ***p < 0.01, 95% confidence intervals are in brackets.  (1) and (2) show the results of event study for men and women, respectively. The dependent variables in both columns are hourly wage. The independent variables are event time dummies, indexed such that u_bt mean t years before the year of arrival of first child, and u_at mean t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 was omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. Cityfixed effects, year-fixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the month-of-survey-fixed effects are included. To test whether the coefficients by subgroups are significantly different, we use the test drawing on seemingly unrelated regressions with the command suest in Stata. *p < 0.1, **p < 0.05, and ***p < 0.01, 95% confidence intervals are in brackets.  (1) and (2) show the results of event study for women in urban and rural areas, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, year-fixed effects, province-yearfixed effects, the year-of-survey-fixed effects, and the month-of-survey-fixed effects are included. *p < 0.1, **p < 0.05, and ***p < 0.01, 95% confidence intervals are in brackets.  (1) and (2) show the results of event study for women in high-income and low-income groups, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, yearfixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the month-of-surveyfixed effects are included. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01.  (1) and (2) show the results of event study for women with high educational attainment and low educational attainment, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. Cityfixed effects, year-fixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the month-of-survey-fixed effects are included. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01.  (1) and (2) show the results of event study for women in younger and older groups, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, yearfixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the month-of-surveyfixed effects are included. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01.  (1) and (2) show the results of event study for women whose children were primarily looked after by themselves and those not, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City-fixed effects, year-fixed effects, province-year-fixed effects, the year-of-survey-fixed effects, and the month-of-survey-fixed effects are included. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01.  (1) and (2) show the results of event study for women who have the first child before the age of 24 years and after, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City fixed effects, year fixed effects, province year fixed effects, the year of survey fixed effects, and the month of survey fixed effects are included. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01. This table shows the results of event study for women without and with sons, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City fixed effects, year fixed effects, province year fixed effects, the year of survey fixed effects, and the month of survey fixed effects are included. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01. results of event study for women with one and multiple children, respectively. The independent variables are event time dummies, indexed such that u_bt means t years before the year of arrival of first child, and u_at means t years after the year of arrival of the first child, respectively. The event-time dummy at t = -1 is omitted, implying that the event-time coefficients measure the impact of children relative to the year just before the first childbirth. Individual controls included years of schooling and age fixed effects to control for life cycle trends. City fixed effects, year fixed effects, province year fixed effects, the year of survey fixed effects, and the month of survey fixed effects are included. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01. the results show the impact of childbirth on life satisfaction and mediators for women. In column 2, the dependent variable is financial support, which takes 1 when get economic support from parents, otherwise, it takes 0. The dependent variable in column 3 is personal income. In column 4, the coresidence with parents takes 1 when live with parents, otherwise, it takes 0. The dependent variable in column 5 is self-reported social status ranging from 1 to 5, with the higher the score representing higher social status. The dependent variable in column 6 is improved physical health, which takes 1 when having a better health, otherwise, it is 0. The independent variable of interest "after" is a time dummy. It takes 1 when the time is after the arrival of first child; 0, otherwise. The control variables are the same as in Table 2. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01.

Supplementary Table 22. Impact of childbirth on life satisfaction with mediators included
(1) (2) The variable financial support is equal to 1 if women receive financial supports from parents, 0, otherwise. The variable co-residence is equal to 1 if women live with parents, 0, otherwise. The variable self-reported social status, ranges from 1 to 5, with the higher the score representing the higher social status. The variable improved physical health is equal to 1 when having a better health, 0, otherwise. The independent variable of interest "after" is a time dummy, which is equal to 1 when the time is after the arrival of first child; 0, otherwise. The control variables are the same as in Table 2. 95% confidence intervals are in brackets, with *p < 0.1, **p < 0.05, and ***p < 0.01.