The World Bank Economic Review, 37(2), 2023, 257–282 https://doi.org10.1093/wber/lhac032 Article The Labor-Supply Consequences of Having a Child Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 in China Shing-Yi Wang Abstract Combining eight years of panel data with an event study approach, this study shows that rural Chinese women’s labor supply falls following the birth of a child. In contrast, men’s labor supply does not fall after birth. Fur- thermore, a woman’s labor supply falls more following the birth of a son than a daughter. Following the birth of a son relative to a daughter, household cigarette consumption declines, and a mother’s leisure time, her prob- ability of school enrollment, and her participation in decision-making increase. There are no increases in other investments in boys complementary to mothers’ time, such as food expenditures, breastfeeding, or immuniza- tions. These results are consistent with the idea that mothers are rewarded more for having a son, leading them to have more leisure and work less. JEL classification: J13, J16, J22, O12, O53 Keywords: birth, female labor supply, gender, China 1. Introduction In developing countries, theory suggests a U-shaped relationship between economic growth and female labor-force participation (Goldin 1994; Mammen and Paxson 2000). At the initial stage, economic activity is dominated by household enterprises, mainly farming, and men and women both participate in this work. Economic development corresponds to a shift from family enterprises into formal manufacturing jobs; this corresponds to an improvement in labor-market opportunities for men, as well as overall positive income effects for households. As a result, female labor-force participation declines. Then the next stage of development entails a shift towards service and white-collar jobs, which draws women back into the labor force. Many of the existing papers in economics focus on policies and developments that increase female labor-force participation (Goldin and Katz 2000, 2002; Bailey 2006; Jensen 2012; Heath and Mobarak 2015), but there is less focus on declines in female labor-force participation.1 As shown in Panel A of fig. 1, Shing-Yi Wang is an associate professor at the Wharton School of the University of Pennsylvania; her email is was@wharton.upenn.edu. This paper has benefitted from feedback from Santosh Anagol, S. Anukriti, Rachel Heath, Melanie Khamis, Annemie Maertens, Laura Schecter, Thomas Fujiwara, and seminar participants at Duke, Northwestern, Oxford, and UCL. Sunny Lee, Hu Liu, Imran Idzandqar, Adi Jahic, and Adam Streff provided excellent research assistance. All errors are my own. A supplementary online appendix is available with this article at The World Bank Economic Review website. 1 Some exceptions include recent working papers studying declines in female labor-force participation in India (Fletcher, Pande, and Moore 2017; Afridi, Dinkelman, and Mahajan 2018). © The Author(s) 2023. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com 258 Wang Figure 1. Trends in Labor Status by Gender and Location Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Source: These trends are calculated using data from the China Health and Nutrition Survey over a sample of individuals aged 20 to 40. Note: Panel A shows labor-force participation rates by gender and location (urban versus rural). Panel B shows the share of individuals who report being a housewife. The sample size is 41,417 for labor-force participation and 35,540 for housewife status. labor-force participation rates in China for both women and men generally declined between 1990 and 2015. Part of the reason for these overall trends is the shift away from the Communist labor-market sys- tem that both expected everyone to work and facilitated the provision of jobs.2 However, the decline for rural women is much steeper, providing some suggestive evidence that the economic shift away from agri- culture to other jobs may drive these trends as predicted by theory of the U-shaped relationship between development and women’s work.3 This decline in labor-force participation rates among rural women corresponds to a sharp increase in rural women reporting that they are housewives (Panel B of fig. 1). 2 As shown in supplementary online appendix fig. S1.2, the number of individuals reporting that they are unemployed (and searching for work) is generally increasing over this period. 3 Figure 1 used data from a household survey called the China Health and Nutrition Survey. As shown in fig. S1.3 with Ministry of Agriculture data, these trends are also evident in a much larger survey of rural households. The World Bank Economic Review 259 The paper examines how the labor-force participation of rural women and men in China responds to the birth of a child. The paper begins by exploiting a panel data set collected by the Chinese Ministry of Agriculture called the National Fixed Point Survey over an eight-year period that has a large enough sample to see a substantial number of births. This data set allows the paper to take an event study approach that looks at the labor-market outcomes of the same individuals for the periods leading up to birth, as well as the years following birth. To address some of the drawbacks of the main data set, the analysis also Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 supplements the main results with a second panel data set, the China Health and Nutrition Survey, with 10 waves of data collection between 1989 and 2015. The paper documents a large reduction in the labor-force participation of women of around 10 per- centage points following the birth of a child. In contrast, men’s labor supply does not change following the birth of a child. This paper contributes to an emerging literature that uses an event study approach to look at the consequences of having children on labor-force participation rates and hours worked.4 Kleven, Landais, and Søgaard (2019) show that the arrival of children in Denmark corresponds to a 20 percent drop in women’s earnings relative to men’s, driven by drops in participation rates, hours worked, and wage rates.5 Kuziemko et al. (2018) show similar employment declines for women in the United States and United Kingdom following birth, and argue that this is driven by an information shock to households about how hard parenting is. Lebedinski, Perugini, and Vladisavljevic (2022) find even larger declines in the employment of women in Russia. This paper takes a similar empirical approach to these prior papers in using an event study to look at women’s labor-force participation. However, the context in rural China studied in this paper is quite different, with many households engaging in agricultural production on fam- ily farms. Thus, the paper’s finding of large drops in women’s labor supply following birth in this context suggests that the results are driven by household preferences rather than by rigidities in the formal labor market.6 Another way in which this paper is unique compared to the prior literature on event studies and impact of birth is that it also examines the implications of the fall in women’s labor supply on total household income and consumption. The results show no significant change in total household income or in total household expenditures. However, the composition of expenditures changes, with a large drop in cigarette consumption and increases in milk and meat consumption. Finally, the prior literature looking at event studies around birth and labor-market outcomes in high- income countries has not documented differences in outcomes by the gender of the child. It is quite likely that in countries like the United States, United Kingdom, and Denmark, there are no differences by the gender of the child, whereas this is particularly relevant in a country like China, where the population exhibits considerable son preference. This paper documents substantial heterogeneity in outcomes fol- lowing the birth of a girl versus a boy in rural China. In particular, the results show women reduce their labor-force participation rates and amount of days worked much more for boys than for girls. For fathers, there are no significant differences in whether they work or the amount that they work for boys versus girls. However, men are more likely to migrate and earn more from migration following the birth of a son relative to a daughter. For the heterogeneity results that focus on the differences for the birth of a boy versus a girl, the results may reflect either causal effects of the gender of the child or selection of the types of households who have 4 Some other papers in this literature focus on earnings effects (e.g. Angelov, Johansson, and Lindahl 2016; Chung et al. 2017; Cortés and Pan 2020). However, given the context in which many rural workers in China work on household farms and enterprises, it is not possible to examine the impact on individual earnings here. 5 Kleven et al. (2019) show similar declines in five other countries. 6 Examples of rigidities include inflexible work arrangements, such as full-time jobs that do not offer fewer or flexible hours as options or jobs that require people to work on site with no options for working from home. This can cause combining parenting with work to be more difficult than in a context like working on a family farm, where hours can be more flexible and women could potentially work near their children (facilitating breastfeeding for example). 260 Wang a boy versus a girl. One of the key advantages of our analysis over the existing literature that compares the impacts of sons versus daughters is that the panel data set in this analysis allows for the removal of any time-invariant characteristics or preferences of households.7 Four sets of results are presented to address this concern. First, the paper shows that there are very few significant differences in the observable characteristics of individuals or households prior to birth of a son versus a daughter. Second, the paper exploits the panel data to show that there are no differences in the pre-birth trends in the outcomes of Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 interest. Third, the paper looks at the subsample of first-born children, for whom the sex ratio at birth is less skewed. Fourth, the paper implements a bounding exercise where some male births are reassigned to female births in order to achieve a natural sex ratio at birth and the reassignments in the bounding exercise are chosen to maximize the impact of the sex selection on the outcomes. The paper explores the mechanisms that can explain why the results differ by the gender of the child. One mechanism is about economic differences related to boys and girls. If there are higher returns to investments in boys than girls, households may want to invest more in boys (and have mothers provide inputs into sons rather than work). This hypothesis is tested by examining whether there is evidence of other types of investments in boys that would be complementary to mothers’ effort. Two changes are observed that may be complementary investments: a decrease in household cigarette consumption and an increase in mothers’ schooling. However, there are no other significant changes in investments in households in which boys are born, including expenditures on meat and milk, breastfeeding, and immu- nizations.8 Moreover, while women work less following the birth of a son, there is no evidence that they spend more time on child care for sons than daughters. Overall, the evidence does not strongly support the investment mechanism. A related economic mechanism is a pure wealth effect associated with a son (i.e. households need to save less for retirement) and even if they invest equally in sons and daughters, the wealth effect means that women can work less and consume more leisure. This hypothesis is tested by examining whether consumption increases as expected with a positive wealth effect. There is no increase in total consumption following the birth of a boy, suggesting that the labor-supply effects do not reflect a wealth effect. Another possibility is that the mechanism is through total fertility and the gender of the birth provides information to the household about their total fertility. Given that the gender of the child does not predict subsequent household size in this data set on rural China, this also cannot explain the differences in the labor-market responses to boys and girls. Second, the paper considers whether the results can be explained by household preferences where households reward mothers for producing a son. To our knowledge, this is among the first papers to consider and show empirical evidence for this idea. These rewards can correspond to the observed changes for women following a son’s birth: less agricultural work, more investment in mothers’ education, and increased participation of women in household decision-making (over the purchase of durable goods). Given that women in China dislike smoking, the decrease in smoking following the birth of a son can also be explained by this mechanism. Finally, mothers do not provide more child care for sons than daughters but they spend more time in adult leisure activities after having a son relative to a daughter. There is a literature that examines the consequences of having boys versus girls on parents’ outcomes. Ichino, Lindström, and Viviano (2014) find that women work less after a first-born son than a daughter in the United States, United Kingdom, Italy, and Sweden, possibly due to the positive impact of a son on total fertility. Lundberg and Rose (2002) find that fatherhood increases men’s wages and labor supply 7 However, there may be time-varying preferences or characteristics of households that change at the time of the birth for reasons not driven by the birth of the child itself. For example, if households choose to have a boy in response to receiving a positive economic shock but have girls at any time, then this could lead to selection driving the interpretation of the results. 8 The measures of meat and milk are in quantity rather than expenditures, so it is not possible to capture whether families increase the quality of food purchased for boys but not the quantity. The World Bank Economic Review 261 in the United States more for boys than girls. Having a son increases women’s decision-making power in China (Li and Wu 2011) but not in India or Bangladesh (Zimmermann 2012; Heath and Tan 2018). Having a son (rather than a daughter) leads to cleaner fuel use in India (Kishore and Spears 2014), heavier mothers during children’s adolescence in the United States (Pham-Kanter 2010), less criminal activities of fathers (Dustmann and Landersø 2018), and changes marriage outcomes (Bedard and Deschenes 2005; Ananat and Michaels 2008; Dahl and Moretti 2008; Anukriti, Kwon, and Prakash 2022). Because of Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 small sample sizes and often the lack of panel data, none of the prior papers in this literature are able to employ the event study approach utilized in this paper. In fact, none of them use any pre-birth data; they regress the outcomes of interest on the presence of a girl or boy. The pre-birth data allow for new evidence that there are no important trends in outcomes leading up to the birth event. The research question in this paper is particularly important for policy given the recent policy changes in China related to fertility. Implemented in 1979, the One Child Policy aimed to limit population growth by introducing a system of carrots and sticks to encourage households to reduce their fertility. One unin- tended consequence of this policy was an increase in the male-to-female sex ratio at birth. Indeed, given that our paper documents different labor-market responses to the birth of a son versus a daughter, changes in the number of births of boys versus girls can have unintended effects on labor-market outcomes. The rest of the paper proceeds as follows. The next section discusses the data sets and presents sum- mary statistics. After outlining the empirical specification the paper presents the conceptual framework. The final sections presents the main empirical results on labor-market outcomes, other outcomes and heterogeneity in outcomes. Finally, the paper concludes. 2. Data 2.1. Ministry of Agriculture National Fixed Point Survey Data It is relatively rare to find a panel data set of households in developing countries where the survey is conducted on an annual basis and contains a large enough sample to have a substantial number of births. One such data set is the National Fixed Point (NFP) Survey collected by the Research Center of Rural Economy (RCRE) of the Chinese Ministry of Agriculture. The villages in the sample were selected for representativeness of rural, agricultural households based on region, income, cropping pattern, popula- tion, and non-farm activities. Within each village chosen, a random sample of households was drawn to be included in the survey. In rare cases in which the entire household moves permanently, the household attrites from the survey and is replaced by another household. While the survey first began in 1986, the analysis in this paper focuses on the annual waves between 2003 and 2010, because the structure of the survey changed substantially in 2003. The 2003 wave is the first period in which there are some ques- tions, including on employment and health, that are asked at the individual level.9 The individual-level questions are key to looking at the separate effects of a birth on the labor supply of women versus men. At the individual level, the survey asks a relatively small set of questions including age, gender, edu- cation and training, relation to the household head, self-reported health status, the number of days the individual worked, occupation, industry, and whether the person is currently enrolled in school. At the household level, in addition to detailed questions about agricultural inputs and outputs, the survey also asks about total household income and several categories of consumption. There are several limitations to the data set. The main goal of the survey is to ask agricultural house- holds about farm inputs and outputs. Thus, the data lack details on non-agricultural decision-making of households. Another important limitation of the data is that it does not have specific questions about 9 Prior to 2003, all of the questions were asked at the household level. In other words, it is possible to observe how many days of work the total household supplied but it is not possible to separate out whether a woman or man worked those days if the household contained both a woman and man. 262 Wang birth, so the timing of birth must be inferred from the arrival of a zero- or one-year-old baby into the panel. For a child who arrives in the household and is reported to be age one, the prior calendar year is assigned as the child’s birth year. However, this may lead to our assignment of birth year actually cap- turing the year prior to birth in many cases.10 This is because there seems to be a substantial amount of rounding up of an infant’s age to one year. In the data set, there are 668 reports of a child of age zero and 7,141 reports of a child of age one. Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Because the parentage of each child is not specifically asked, the father and mother of a child are assigned based on a question about the relationship to the household head. If a child of the household head is born, the household head and spouse are assigned as its father and mother. In rural China, households often include three generations, so many babies born in households are the grandchild of the head of household. In this scenario, the head’s child and the spouse of the head’s child are only assigned as the parents if there is only one child of the head and corresponding spouse residing in the household. In other words, if two adult children of the household head reside in the household, it is not possible to determine which sibling is the parent of the baby and these births are excluded from the analysis. Given that a woman may give birth to more than one child in the eight-year panel, the analysis focuses on the first birth that occurs within the sample. The analysis also excludes the birth of twins (or higher- order multiple births) from the sample.11 Finally, the analysis excludes the rare cases where the assignment of a child to a parent corresponds to a parent who is under age 11. For the analysis on women’s outcomes, this yields 2,084 births. Summary statistics are presented for individual-level variables in the year prior to the birth of a child in table 1. The first two columns show statistics for women and the last two columns show the same statistics for men.12 The work indicator is defined based on a question in the survey on the number of days that the indi- vidual worked in the past year. The question refers to both days spent working on the family farm and working outside the home. About three-quarters of women were working in the year prior to the arrival of a child. The number is higher for men: 91 percent of men were working prior to the arrival of a child. Women worked 161 days per year while men worked over 240 days on average. Health status is self-reported on a scale of 1 to 5, where 1 corresponds to the best health and 5 the poorest health. Men report being slightly healthier than women prior to the birth of a child. The average age prior to birth is 26.8 years for women and 27.9 years for men. Women have slightly more than 8 years of education whereas men in the sample have an average of 8.5 years of schooling. There is also a question on whether the individual is currently enrolled in school. Just over 1 percent of women are enrolled in school prior to the arrival of an infant, and the corresponding number for men is slightly lower. Temporary migration is common in the data. Women are living at home about 281 days of the year prior to the birth of a child. Men are away from home slightly more than women: their average number of days at home is about 239. Corresponding to being away from home more, men also earn more away from home. Men earn 5,038 yuan per year in work away from home while women earn a little less than one-third of that amount.13 The survey does not ask about total individual earnings while at home; this is because many of the survey respondents are engaged in household agricultural production and it would be hard to assign joint agricultural profits to individuals. As a measure of more permanent migration, the 10 While the coefficient magnitudes change a bit, the broad economic results in the paper are robust to shifting the birth year of children who appear in the data as age one to assume instead that they are under one. 11 One issue with twins is that it is not possible to assign gender properly in the regression equation. 12 The sample size is slightly larger for men because women are less likely to have pre-birth data. This is because women are more likely to join their husbands’ families than vice versa. If a woman joins the NFP household at marriage and gives birth in the same year, then any pre-birth periods are not observed for her. 13 This amount is converted into real 2002 RMB using a consumer price index from the Regional Economy Database. The World Bank Economic Review 263 Table 1. Individual-Level Summary Statistics (NFP) Women Men Pre-child N Pre-child N Work indicator .7581 2,084 .9133 2,435 Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 (.4283) (.2813) Days of work 161.0 2,084 240.3 2,435 (132.8) (110.1) Health status 1.419 2,068 1.349 2,417 (.5349) (.5030) Age 26.83 2,084 27.93 2,435 (4.759) (4.646) Education 8.066 2,036 8.546 2,407 (2.338) (2.305) In school .0126 2,062 .0111 2,413 (.1116) (.1052) Days at home 281.3 2,036 238.7 2,356 (126.7) (142.1) Earnings away 1811.6 2,084 5038.9 2,435 (4542.1) (8792.7) Attrite .0366 2,044 .0279 2,395 (.1880) (.1649) Agriculture industry .4805 2,029 .2845 2,379 (.4997) (.4513) Family occupation .5878 2,043 .4474 2,398 (.4923) (.4973) Source: The means are calculated using data from the Chinese Ministry of Agriculture’s National Fixed Point (NFP) Survey. Note: The first two columns show summary statistics for women and the last two for men. The mean of each variable is shown for the year immediately prior to the birth of a boy or a girl with standard deviations below in parentheses. Health status is self-reported on a scale from 1 to 5 where 1 is the best. Education is measured in years. Family occupation is an indicator for whether the person’s occupation is in a family-based enterprise, including agriculture. analysis examines whether the individual attrites from the survey (i.e. is not surveyed in the subsequent year).14 Individual-level attrition is fairly low at 3.7 percent for women and 2.8 percent for men. The survey asks questions about the primary industry of each individual. About half of women re- port agriculture as their primary industry, while the corresponding number for men is much lower at 28 percent. There is also information about whether the primary occupation of the individual is in a family- based enterprise, including agriculture. About 59 percent of women and 45 percent of men work in an occupation in a family enterprise. Next, table 2 shows household-level variables in the NFP data.15 Prior to the birth of a child, house- holds report an average total income of RMB 35,000. Total expenditures are slightly lower than in- come, implying that the average household is saving.16 Several categories of consumption are presented: cigarettes, alcohol, milk, and meat. Cigarettes are measured as expenditures while alcohol, milk, and meat are measured in kilograms. Given that the survey questions on cigarette and alcohol consumption changed in 2009, the analysis on cigarette and alcohol consumption is limited to the waves prior to 2009.17 The 14 This is only measured in years 2003 to 2009 as it is unknown whether they will attrite in the subsequent year for the last period (2010) for which the data is available. 15 Most of the questions are household-level questions about detailed inputs and outputs into agricultural production broken down at the crop level. These are not useful for the purposes of this paper. 16 This is not surprising given the literature demonstrating a high savings rate in China. 17 In other words, the analysis drops 2009 and 2010 because it is not possible to make those values comparable to the prior years. 264 Wang Table 2. Household-Level Summary Statistics (NFP) Pre-child Pre-boy Pre-girl N p-value Total income 35075 33465 37175 2,408 .2679 (81416) (58682) (10383) Total expenditures 31055 29124 33576 2,393 .1803 Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 (80565) (56151) (10413) Cigarettes 628.7 619.0 641.4 2,412 .7190 (1521.6) (1456.2) (1603.5) Alcohol 12.24 13.96 10.00 2,412 .0925 (57.38) (69.88) (34.81) Milk 7.214 7.081 7.388 2,412 .7975 (29.13) (28.64) (29.77) Meat 155.2 154.9 155.6 2,431 .9189 (163.1) (156.2) (171.7) Others’ days of work 297.7 293.0 303.8 2,431 .2626 (235.7) (231.1) (241.5) Household size 4.222 4.263 4.170 2,431 .0419 (1.119) (1.123) (1.113) Grandparents .8728 .8740 .8713 2,431 .8397 (.3331) (.3318) (.3349) Minority .1197 .1278 .1091 2,354 .1652 (.3247) (.3341) (.3119) Source: The means are calculated using data from the Chinese Ministry of Agriculture’s National Fixed Point (NFP) Survey. Note: The mean of each variable is shown for the year immediately prior to the birth of a child, a boy, or a girl, with standard deviations below in parentheses. The p-value refers to whether the pre-boy mean is statistically different from the pre-girl mean. The units for cigarettes are RMB. The units for alcohol, milk, and meat are kilograms. number of days of work for all other people in the household (excluding the parents of the baby) is aggregated together. In total, people other than the parents work an average of 298 days. The average household size prior to the arrival of a baby is 4.2. The vast majority of households include the people who will become grandparents in the subsequent year. About 12 percent of the households are minorities (non-Han ethnicity). Table S1.1 and columns 2 and 3 in table 2 show the statistics separately by gender of the child. For women in table S1.1, there are no significant differences in any of these variables prior to the birth of a son versus a daughter. For men, out of the 11 variables, only the probability of working is statistically different (at the 5 percent level) prior to the arrival of a son versus a daughter.18 In table 2, alcohol consumption is statistically different at the 10 percent level and household size is statistically different at the 5 percent level. The magnitude of the difference in household size (4.222 versus 4.263 people) is essentially zero. Overall, this provides some assurance for the identification strategy that compares individuals before and after the arrival of a daughter with individuals before and after a son is born. 2.2. China Health and Nutrition Survey To address the fact that the variables in the NFP data are quite limited, the main analysis is supplemented with a panel survey of households in China called the China Health and Nutrition Survey (CHNS). There are 10 rounds of surveys, in the years 1989, 1991, 1993, 1997, 2000, 2004, 2006, 2009, 2011, and 2015.19 The sampling entailed a multistage, random cluster design where counties were stratified into three levels 18 Note that in the pre-birth trends, the results show no significant differences in the trends in men’s probability of working prior to a boy versus a girl. 19 The provinces include Beijing, Chongqing, Guangxi, Guizhou, Heilongjiang, Henan, Hubei, Hunan, Jiangsu, Liaoning, Shaanxi, Shandong, Shanghai, Yunnan, and Zhejiang. The World Bank Economic Review 265 Table 3. Summary Statistics for Rural Women (CHNS) Pre-child Pre-boy Pre-girl N p-value Work indicator .8178 .8155 .8203 604 .8786 (.3862) (.3884) (.3845) Age 26.36 26.57 26.13 613 .2710 Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 (5.047) (5.119) (4.967) Education 7.317 7.318 7.315 536 .9929 (3.563) (3.588) (3.543) Household smoking .6704 .6630 .6782 531 .7085 (.4704) (.4735) (.4680) Child care (hours/week) 18.79 18.50 19.11 370 .8529 (31.76) (31.07) (32.58) Leisure (hours/week) 18.77 18.77 18.76 177 .9930 (12.29) (12.37) (12.28) Chores (hours/day) 2.052 2.166 1.933 579 .1246 (1.821) (1.965) (1.651) Breastfeeding indicator .1460 .1479 .1441 664 .8912 (.3534) (.3555) (.3518) Household immunizations .5300 .5809 .4754 549 .0699 (.6818) (.7111) (.6457) Male share in household decisions .2157 .2124 .2187 219 .9024 (.3784) (.3882) (.3707) Source: The means are calculated using data from the China Health and Nutrition Survey (CHNS). Note: The mean of each variable is shown for the first year prior to the birth of a child, a boy, or a girl with standard deviations below in parentheses. The p-value refers to whether the pre-boy mean is statistically different from the pre-girl mean. of income, and a weighted sampling technique selected four counties in each province plus the provincial capital and one low-income city. The sample size included about 3,800 households in 1989 increasing to about 5,900 in 2015. While the CHNS data contain both urban and rural households, the analyses in this paper are limited to the rural households to correspond best to the NFP sample. Unlike the NFP data, parents in the CHNS are linked to children using direct birth history questions. There are also questions on time use, including time spent on child care and leisure activities. While there are several advantages of this data set over the NFP data, the smaller number of households means that there are relatively fewer births happening in the sample period. Thus, there will be issues with power in some of these analyses. As shown in table 3, about 82 percent of women are working prior to the birth of a child.20 This is slightly higher than in the NFP data. Women in the CHNS are similar in age to women in the NFP data prior to birth at 26.4 years old. They also have a little less education, averaging over 7.3 years of schooling. About 70 percent of households in the CHNS have a smoker. Individuals report the total number of child-care hours that they provided in the last week. The question is specific in that this should include times where they are simultaneously watching their child and doing something else, like cooking. It is asked in every wave except 1989. Women are watching children for 18.8 hours per week prior to the birth.21 Questions about time spent on leisure activities are asked in terms of minutes per day, separately for weekdays and weekends; total leisure time is calculated by aggregating across categories into to- tal hours per week. These activities include watching TV, watching movies, playing video games, read- ing/writing/drawing, and playing with games/toys. These questions are available only in the waves 2004, 20 This is defined using a question on the amount worked in the last year, to be in line with the way it is defined in the Ministry of Agriculture (MOA) data. It is filled in with whether the person is currently working when that is missing. 21 This can include care for older siblings or the children of others. 266 Wang 2006, 2009, 2011, and 2015. Prior to the arrival of an infant, women spend about 19 hours per week on leisure activities. Another set of time-use questions asks about the time the individual spends on chores in an average day. In this analysis, their responses to three questions about the amount of time spent on buying food, preparing food, and washing and ironing clothing are summed. These three questions are asked in all waves. Women spend an average of 2 hours per day on food purchases and preparation and cleaning Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 clothing prior to the birth of a child. In every round of the survey, the CHNS also asks about the breastfeeding status of each child in the birth history questions to women. The variable is constructed to equal 1 if the mother reports currently or ever having breastfed the child. Given the spacing of the survey of the CHNS of two- to four-year intervals, the fact that a fraction (14.6 percent) of women were breastfeeding in the survey wave prior to the birth is likely to be driven by a prior child. The analysis also make use of a question in the CHNS about whether each child had any immunization shots in the past year. This question is asked of all children under the age of 12 in the seven waves prior to the 2009 round. A household-level variable is created for the total number of immunizations received by children under the age of 12 in the past year. Finally, a variable is constructed for whether the husband alone makes decisions regarding the purchase of durable goods. This is based on questions that first ask the household about whether they own the good.22 Most goods are asked about in each of the four waves between 1989 and 1997, but some are only asked in a subset of the waves. Conditional on the household owning the good, they are asked who made the decision to purchase it, where the options are husband, wife, husband and wife together, and other. An index is generated representing the share of goods for which the husband alone makes the purchasing decision.23 Men make purchasing decisions alone for about 21 percent of durable goods purchases prior to the birth of a child. Across the 10 variables in table 3, 9 of them are not statistically different for women prior to the birth of a boy as compared to a girl. Household immunizations is the only variable that is statistically different (at the 10 percent level) in the survey wave prior to the birth of a son versus a daughter for CHNS households. Households who will subsequently have a son have a higher rate of immunizations (58 percent versus 48 percent). Overall, households look fairly similar prior to the birth of either a son or a daughter. 3. Empirical Specification For the event study analysis around the birth of a child, the following equation is estimated: 7 yit = α + β j Birthit , j + θa Agea it + τt + γi + it , (1) j=−3 a where i denotes individual (or household for household-level outcomes) and t denotes calendar time. The individual-level regressions are estimated separately for men and women and include individual fixed effects and indicators for age (in years). The age indicators control flexibly for any life-cycle trends. The variables denoted by Birthit, j are indicator variables that equal 1 in the period j relative to the birth of a child. For example, Birthit,−2 indicates two years prior to the birth year, while Birthit,2 indicates two years after the birth year. Thus, the coefficients β j provide information on how the outcome moves around the 22 The goods that the CHNS asks about are a stereo, VCR, black and white TV, color TV, washer, refrigerator, air condi- tioning, sewing machine, fan, clock, camera, microwave, electric pot, pressure cooker, and cooking tools. 23 In other words, among the goods that the household reports owning, the numbers where the response is husband alone are summed and then divided by the total number of goods owned. The World Bank Economic Review 267 birth of a child. The regression is restricted to the three periods prior to and seven periods after birth.24 The omitted category is the year prior to birth (j = −1). The regression includes a constant term and fixed effects for calendar year. The analysis also exploits the panel nature of the data and includes individual fixed effects to absorb any time-invariant characteristics of the individual or household. The standard errors are clustered at the individual level. The estimates of the years prior to the birth allow for an examination of whether households who Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 will have a child are systematically different from households who will not have a child in that year. The estimates of the years after birth allow for an examination of whether any outcomes that change after birth are temporary or long lasting. Given the large number of coefficients associated with the event study equation, the paper also present some estimates using the following more parsimonious difference-in-difference equation: yit = α + β1 PostBirthit , j + δ1 BirthYearit , j + θa Agea it + τt + γi + it , (2) a where PostBirthit, j indicates all years after (and not including) the year of birth and BirthYearit, j equals 1 in the year of birth. As discussed in describing the data, for many births, BirthYearit, j may actually indicate the year prior to birth, so it is important to treat it separately from the periods that are certainly after the arrival of the child. To look at the impacts of giving birth to a son relative to daughter, the following equations are esti- mated: 7 7 yit = α + β j Birthit , j × Soni + δ j Birthit , j + θa Agea it + τt + γi + it (3) j=−3 j=−3 a and yit = α + β1 PostBirthit , j × Soni + β2 BirthYearit , j × Soni + θa Agea it + δ1 PostBirthit , j + δ2 BirthYearit , j a +τt + γi + it , (4) where Soni is an indicator for whether the birth is male. In these equations, the coefficients δ j provide information on how the outcome moves around the birth of a girl, while the coefficients β j provide information on whether the effects of the birth of a male are different from the effects of the birth of a female. 4. Conceptual Framework for Differences in Women’s Labor Supply for Boys versus Girls There are many reasons why women’s labor-market outcomes may respond differently to the birth of a son versus a daughter. First, the paper considers two mechanisms related to differences in the economic returns to having boys and girls. Finally, there are three mechanisms related to fertility and conscious or unconscious gender preferences by households and firms. 24 Women often marry into an existing household (and enter into a household in the data) and give birth to their first child shortly after marriage, so women are usually observed in the data for more periods after the birth than prior to the birth. In 41.3 percent of households with a birth, at least three years of data for the mother prior to birth are observed. For two-thirds of households, outcomes for the mother are observed at least one year prior to the birth. As shown in fig. S1.4, the event study estimates are very similar if the analysis is restricted to the sample of mothers for whom at least one year of labor-market outcomes prior to birth are observed. 268 Wang 4.1. Economic Differences One possible mechanism is that women take more time out of the labor force for boys because the eco- nomic returns to their time and effort spent with boys are higher. This may be driven by the fact that men earn more than women, or even if their earnings were equal, it is possible that men provide more financial support for elderly parents than women. Consistent with this idea, parents are much more likely to core- side with an adult son than with an adult daughter in China. However, the literature on whether daughter Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 or sons provide more financial transfers in China is mixed (Zhu 2016; Gruijters 2018). To test this, the paper examines investments made in children, including child-care hours, breastfeeding, expenditures on child-specific services (immunizations), and cigarette consumption.25 A related mechanism is that boys will contribute more to the lifetime earnings of the household, and the reduction in female labor supply is not a direct investment in boys but a response to a pure wealth effect associated with a boy. Women work less because households do not need to save as much (i.e. for retirement) following the birth of a boy than a girl. To test the idea of the arrival of a boy being treated as a positive wealth shock on the household, the paper examines consumption expenditures. Note that prior research suggests the opposite. There is the potential for a negative wealth effect of having a son in China; when the sex ratio is skewed and brides are scarce, households have to save more for a boy in order to provide their son with a better match (Wei and Zhang 2011).26 4.2. Other Explanations 4.2.1. Fertility The paper also considers the idea that the arrival of a son versus a daughter provides information to the household about their total fertility and any differences in work behavior are driven by total fertility. For example, an existing literature suggests that son preference in India may lead households to target a desired number of sons (e.g. Gupta 1987; Jensen 2003; Jayachandran and Kuziemko 2011; Rosenblum 2013). Under such stopping rules, households continue to have children until they reach their desired number of sons. The key implication of this behavior is that households with boys will be smaller on average. If having a boy leads to fewer total children, women may be more likely to stop working if the household anticipates needing fewer resources for children in the future. On the other hand, women with boys may be less likely to stop working if fewer total (anticipated) children decreases the probability that mothers stay at home to care for their children. The analysis examines whether there are differences in fertility following the arrival of a boy versus a girl. 4.2.2. Labor-Market Discrimination Another explanation is that the labor market treats women who have boys differently from women who have girls. In this scenario, the results are not driven by the preferences of women or the households that they are in, but that firms treat women who give birth to boys differently from women who give birth to girls. This hypothesis is tested by examining whether the results are different for women who work in family enterprises, including agriculture, versus women who work for others. 4.2.3. Rewarding Women A final explanation is that households reward women with more leisure (and less work) for having a boy. There is anecdotal evidence to support this idea. In one newspaper story, Linlin describes how her in-laws had a lot of conflict with her after her marriage to their son, but this reversed immediately after she had a boy; after the birth of the son, the in-laws started doing a lot of cooking and housework for her (Fan 25 Cigarette consumption has negative effects on children’s health through second-hand smoke. 26 Brideprice is the cultural norm in China, where the groom’s family pays the bride’s family at the time of marriage. The World Bank Economic Review 269 Figure 2. Effects of the Birth on Mothers’ and Fathers’ Labor Outcomes Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Source: The estimates use data from the Chinese Ministry of Agriculture’s National Fixed Point Survey. Note: Each panel corresponds to one regression. The dependent variable in the first row is an indicator for work over the year and in the second row is the inverse hyperbolic sine of days of work over the year. The dots give the coefficient estimates for years around the birth. The line denotes the 95 percent confidence interval, where the standard errors are clustered at the household level. The regressions also include a constant term and fixed effects for household, age, and year. The sample size is 18,144. and Qing 2014).27 This idea is tested by looking at the amount of leisure time of women following a boy versus a girl. However, an increase in leisure following the birth of a boy may also be consistent with a model of returns to investment in which mothers exert more effort in caring for boys, and hence need extra recovery time to maintain a higher level of effort in her interactions and care of a son. Another outcome to test this mechanism is female participation in household decision-making. 5. Results on Labor-Supply Outcomes The event study estimates of equation (1), where the outcomes are parents’ labor supply, are presented in fig. 2. The first row shows the coefficient estimates, where the dependent variable is the labor-force participation. The figure on the left shows the estimates of β j around the birth for women, while the figure on the right shows the estimates for men. There are no significant trends in labor-force participation of women prior to the birth of a child. In the year immediately following birth (t = 1), there is a 10.1 percentage point decline in the probability of work among mothers, and this estimate is significant at the 27 There are also stories of poor treatment of Chinese women following the birth of a daughter. For example, one article describes a mother-in-law beginning to physically abuse her daughter-in-law after the birth of a daughter (Yangtse News 2016). 270 Wang Table 4. Effects of Birth on Individual Labor Outcomes Indicator for work IHS days of work Full Balanced First-birth Full Balanced First-birth sample sample sample sample sample sample (1) (2) (3) (4) (5) (6) Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Panel A: Mothers Post-birth −0.112*** −0.131*** −0.150*** −0.827*** −0.869*** −1.091*** (0.0136) (0.0455) (0.0216) (0.0797) (0.264) (0.127) Birth year −0.0874*** −0.0800*** −0.125*** −0.643*** −0.531*** −0.885*** (0.0121) (0.0300) (0.0197) (0.0719) (0.175) (0.117) Observations 19,928 2,946 10,538 19,928 2,946 10,538 Panel B: Fathers Post-birth 0.00375 0.0416* 0.0252** 0.00319 0.167 0.142** (0.00840) (0.0243) (0.0116) (0.0525) (0.151) (0.0718) Birth year −0.00130 0.0131 0.0100 −0.0360 0.0261 0.0210 (0.00743) (0.0160) (0.0104) (0.0459) (0.0981) (0.0649) Observations 22,414 4,944 12,650 22,414 4,944 12,650 Source: The estimates use data from the Chinese Ministry of Agriculture’s National Fixed Point Survey. Note: The dependent variable in the first three columns is an indicator for work status over the year, and in the last three columns is the inverse hyperbolic sine (IHS) function of the days of work over the year. Columns 1 and 4 include the full sample. Columns 2 and 5 restrict the sample to individuals who appear in every wave for six waves around the birth. Columns 3 and 6 restrict the sample to the first birth. The regressions all include fixed effects for age, household, and year, and a constant term. Standard errors are clustered at the household level. *, **, *** denote significance at the 10 percent, 5 percent, and 1 percent levels, respectively. 1 percent level. The labor-force participation rates remain below the pre-birth levels for at least seven years, though there is a positive gradient over the first years of the child’s life. For men (in Panel B), the pattern is very different than for women. There are no significant changes in men’s probability of working after the birth of a child. The bottom row of fig. 2 shows the same estimates for the inverse hyperbolic sine (IHS) function of the number of days that the mother or father spent working in the past year.28 This captures both the extensive margin given by the indicator for work, as well as the intensive margin, where women who work can change the amount that they work. The patterns of results are similar to labor-force participation for both women and men. In the year following birth, women work 74 percent fewer days than they did in the year prior to birth (significant at the 1 percent level). Seven years after birth, women are still working 39 percent fewer days (significant at the 5 percent level). For men, there is no significant change in the number of days worked in any years after birth. Next, the paper presents the parsimonious estimates of equation (2) using the NFP data in table 4.29 The estimates for women are shown in Panel A and men in Panel B. The estimates that use the full sample are presented in columns 1 and 4, where the dependent variables are an indicator for working in the last year and the IHS function of the days of work, respectively. In the full sample, the estimate indicates that women are 8.7 percentage points less likely to work in the year of birth of a child and 11.2 percentage points less likely to work afterwards. These estimates are significant at the 1 percent level. To address concerns about sample selection, the paper shows estimates where the sample is limited to individuals who appear for each of the six periods around the birth. For example, it is possible that households in which women marry in and have a child immediately are different from households in which women marry in and wait a few years prior to having a child. This restriction reduces the sample 28 The IHS function is similar to the logarithmic function but is well defined for zero values. 29 Column 1 of table 7 also shows the estimates using the CHNS. In this data set, after birth, women are 8.38 percentage points less likely to work and this is significant at the 5 percent level. The World Bank Economic Review 271 substantially, as there are some years that either the household or an individual is not present in data. The key estimate of the post-birth effects of a child for labor-force participation remains significant at the 5 percent level and the magnitude increases to 13.1 percentage points (column 2). Similarly, the estimates on the number of days worked are also slightly larger in magnitude (column 5). In order to address the potential concern that sex-selective abortion affects the results, the paper draws upon the prior research that demonstrates that the sex ratio in China is not very skewed for the first birth Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 but the distortion increases for higher-order births (Ebenstein 2010). Thus, the analysis includes a sample where the estimates are limited to the first birth.30 The estimates on labor-force participation are similar to those using the full sample but slightly larger in magnitude, and the estimates are significant at the standard level. Overall, the declines observed in labor-force participation for rural, Chinese women are similar to patterns shown in the literature for the impact of a birth on women’s labor-force participation and amount worked in rich countries, though the impact on labor-force participation is slightly smaller in magnitude (Kuziemko et al. 2018; Kleven, Landais, and Søgaard 2019). In the rural Chinese context, however, the negative impact of birth on women’s labor-force participation rates and days worked is less persistent over time than in the context of richer countries. The results for rural Chinese men show no decline in work, and in fact in some samples show a significant increase in labor supply, but this result is not robust across samples and specifications. 5.1. Permanent Control Group Robustness Check Our main estimates include a sample of households that give birth to a child in the sample period. The analysis also considers whether the event study results are robust to the inclusion of a “permanent” control that does not experience any births in the sample period. This control group solves a potential under- identification problem by helping to identify the time effects in the event study regression (Borusyak and Jaravel 2022). More specifically, the permanent control group is restricted to women (men) in households where there are no births happening in the sample period and whose age is restricted to the range 12 to 50 to be comparable to the women (men) who give birth (and the 10 years around birth). The estimates are presented in fig. S1.5 and show that the magnitudes of the coefficients and their significance remain very similar to the estimates without this additional control group. 5.2. Results by Gender of the Child Figure 3 shows the estimates corresponding to equation (3).31 For the birth of a daughter in Panel A, the impacts are not statistically different from zero. In Panel B, there are significant differences in the labor-force participation of a mother for the birth of a daughter versus a son. These large negative effects for boys persist for the following four years, where mothers of boys are less likely to work than mothers of daughters, and these four estimates are significant at the 5 percent level or higher. The parsimonious differences-in-differences estimates are shown in table S1.2. These estimates indicate that women are much more likely to leave the labor force following the birth of a son as compared to daughter. Panels C and D in fig. 3 show the corresponding estimates for fathers. As seen with mothers, prior to birth, there are no significant trends in anticipation of the birth of a daughter or a son. Unlike for women though, there are no significant differences in the labor-supply response of fathers of boys as compared with daughters. 30 The analysis is limited to the first birth in the sample where the parents have no older children residing in the household. If the parents have older children who are not residing in the household, it would not be possible to identify that the birth in the sample is not their first child because the NFP survey does not ask for a full birth history. In the data in this analysis, unlike in Ebenstein (2010), the skewed sex ratio in the NFP data is more muted but does still exist among first births (54 percent). 31 The corresponding estimates for days of work can be seen in Online Appendix fig. S1.1. 272 Wang Figure 3. Effects of the Birth on Parents’ Labor-Force Participation by Gender of Child Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Source: The estimates use data from the Chinese Ministry of Agriculture’s National Fixed Point Survey. Note: Each row corresponds to one regression. The dependent variable in the first row is an indicator for work over the year and in the second row is the inverse hyperbolic sine of days of work over the year. The dots give the coefficient estimates for years around the birth (δ j ) in the left-hand column and for the interaction between the years around the birth and the birth of a son (β j ) in the right-hand column. The line denotes the 95 percent confidence interval, where the standard errors are clustered at the household level. The regressions also include a constant term and fixed effects for household, age, and year. The sample size is 18,144 for women and 20,131 for men. While prior research on this topic by Ichino, Lindström, and Viviano (2014) in the United States, United Kingdom, Italy, and Sweden also finds stronger declines in female labor supply following the birth of a son relative to a daughter, the effects observed here are much larger. In addition to examining households in rural China that are much poorer and a context with stronger son preference, our analysis is limited to younger children than Ichino, Lindström, and Viviano (2014), who include children up to age 15. 5.3. Bounding Exercise for Sex Selection The sex ratio in China is quite skewed, suggesting that households are choosing to have boys rather than girls. Consistent with aggregate statistics for rural areas in China, in the NFP data used in this analysis, 57 percent of births in the sample period are males. The main concern for the identification strategy in the estimates that separate the impact of a boy versus a girl is that households with strong son preferences are also households that prefer women to reduce their labor supply after any child. In other words, the impact of having a boy on female labor supply might be overestimated if households who actively choose to have a boy rather than a girl would have reduced the amount that a mother works substantially even if they had a girl. Tables S1.1 and 2 show that the observable characteristics of individuals and households prior to the birth of a son or a daughter are similar, and fig. 3 indicates that the pre-birth trends in outcomes are similar. The World Bank Economic Review 273 Table 5. Bounding Effects of Sex Selection on Mothers’ Labor Outcomes Work indicator IHS work days (1) (2) Son imputed × post-birth −0.0296* −0.160 (0.0173) (0.102) Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Son imputed × birth year −0.0394* −0.179 (0.0214) (0.126) Post-birth −0.0952*** −0.739*** (0.0169) (0.0986) Birth year −0.0661*** −0.546*** (0.0166) (0.0985) Observations 19,928 19,928 Source: The estimates use data from the Chinese Ministry of Agriculture’s National Fixed Point Survey. Note: The indicator for the birth of a son is imputed to bound the effect of sex selection on the estimates. The regressions include fixed effects for year, age, and household, and a constant term. Standard errors are clustered at the household level. *, **, *** denote significance at the 10 percent, 5 percent, and 1 percent levels, respectively. IHS in column 2 refers to the inverse hyperbolic sine function. In addition, the results have been shown to be robust in the sample limited to first-born children, where the gender bias at birth is less pronounced. To further support the idea that selection in gender is not driving the results, a bounding exercise is done, where some male births are reassigned to female births such that the gender ratio at birth that is considered the natural sex ratio is achieved. Furthermore, the births selected to be reassigned maximize the reduction in the estimated impacts on labor-force participation.32 More specifically, to test for the idea that preferences for women working post-baby are correlated with sex-selective behavior, the pool of households is identified for which (a) the women gave birth to sons, (b) were working prior to the birth, and (c) stop working after birth. This is the pool of households for which assuming that they engaged in sex-selective abortion of girls and switching their births from sons to daughters gives the most conservative estimates of having a boy on labor supply. Thus, the impact of sex-selective abortion on the estimates of labor-supply effects is maximized, and the analysis considers whether this amount of sex-selective abortion can drive the estimates to zero. Among those households, one-third are selected to reassign those births from boys to girls.33 The results are presented in table 5. Mechanically, it must be the case that the estimates with this bounding are smaller differences in the effects of having a son as compared with having a daughter. However, the magnitudes are not trivial; women are 9.52 percentage points less likely to work after the birth of a daughter and 12.48 percentage points less likely to work after the birth of a son. The difference, which indicates how much more likely a woman is to stop working for a son relative to a daughter, is 2.96 percentage points and is significant at the 10 percent level. In terms of the number of work days, women work 73.9 percent fewer days after the birth of a daughter and 89.9 percent fewer days after the birth of a son. However, the difference between having a boy or a girl on the number of work days (−0.16) is only significant at the 11.5 percent level. Overall, the key results of declines in women’s labor supply survive this very conservative bounding exercise. 32 In other words, the coefficient is driven as far down towards zero as possible in reassigning the gender of the births. 33 The amount one-third is the number needed to switch in order to achieve the natural gender ratio at birth in the sample for the bounding exercise. Note that when the outcome is for labor-force participation, the random selection of the one-third of households does not matter that much because these are all households for which the woman is changed from working equals 1 to working equals zero. 274 Wang Table 6. Effects of Birth on Household Income and Consumption Total Total Cigarettes Alcohol Milk Meat income expenses (yuan) (kg) (kg) (kg) (1) (2) (3) (4) (5) (6) Panel A: Birth of a child Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Post-birth −0.0527*** −0.0262 −0.180** −0.0204 0.396*** 0.0465* (0.0174) (0.0209) (0.0836) (0.0238) (0.0542) (0.0241) Birth year −0.0199 0.0215 −0.0267 −0.0258 0.0882** 0.0748*** (0.0149) (0.0199) (0.0726) (0.0182) (0.0442) (0.0204) Observations 19,754 19,657 14,589 12,330 19,809 19,928 Panel B: By gender of the child Son × post-birth −0.0216 −0.0258 −0.340** −0.00256 −0.000737 0.0175 (0.0266) (0.0323) (0.133) (0.0368) (0.0826) (0.0346) Son × birth year 0.0176 −0.00833 −0.416*** −0.0198 −0.0606 0.0426 (0.0272) (0.0365) (0.140) (0.0327) (0.0836) (0.0396) Post-birth −0.0401* −0.0111 0.0163 −0.0187 0.396*** 0.0362 (0.0226) (0.0274) (0.117) (0.0289) (0.0737) (0.0324) Birth year −0.0298 0.0266 0.216* −0.0142 0.123* 0.0499 (0.0216) (0.0286) (0.113) (0.0244) (0.0694) (0.0318) Observations 19,754 19,657 14,589 12,330 19,809 19,928 Source: The estimates use data from the Chinese Ministry of Agriculture’s National Fixed Point Survey. Note: The regressions include fixed effects for year, age, and household, and a constant term. The dependent variables are transformed using the inverse hyperbolic sine function. Cigarette expenditures are transformed into real 2002 yuan. To maintain consistency in the survey question, the regressions for cigarettes and alcohol exclude the waves 2009 and 2010, and the regression for alcohol further excludes 2003. Standard errors are clustered at the household level. *, **, *** denote significance at the 10 percent, 5 percent, and 1 percent levels, respectively. 6. Impacts on Other Outcomes 6.1. Consumption In order to better understand the impact of the labor-supply response observed for mothers, the paper examines other outcomes of interest. As shown in column 1 in Panel A of table 6, total household income falls by 5.27 percent and this estimate is significant at the 1 percent level. The impact on total expendi- tures is not statistically different from zero at the standard levels, so it is possible that households reduce their savings rate in order to maintain consumption after having children. However, given the sizable standard errors on the estimate, it is also not possible to reject that the decline in expenditures has the same magnitude as the decline in income. The composition of household consumption changes. For example, households reduce their spending on cigarettes by 18 percent on average and increase their consumption of milk by 39.6 percent and meat by 4.65 percent following the birth of a child. These three estimates are significant at the 10 percent level or higher. Panel B of table 6 shows heterogeneity in the impacts by the gender of the child. While women are less likely to work following the birth of a son relative to a daughter, interestingly there is no signifi- cantly larger drop in household income or total expenditures following the birth of a son. The result on total expenditures provides some suggestive evidence that the birth of a son (rather than a daughter) is not interpreted by the household as a positive wealth shock. Similarly, the results do not show positive changes in the consumption of cigarettes, alcohol, milk, or meat following the birth of a boy relative to a girl. Furthermore, if milk and meat are investments in children, there is no support for the idea of more investment in boys than girls in these consumption outcomes; household consumption of milk and meat increases for all children, regardless of their gender, and not more for boys. The World Bank Economic Review 275 Table 7. Effects of a Birth on Mother’s and Household Outcomes (CHNS) Work Household Child-care Leisure Chores Breastfeed Household Male indicator smoking hours hours hours indicator immun. decision (1) (2) (3) (4) (5) (6) (7) (8) Panel A: Birth of a child Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Post-birth −0.0838** −0.0258 34.91*** −4.624** 0.259** 0.628*** 0.960*** −0.0200 (0.0370) (0.0462) (3.664) (2.213) (0.123) (0.0349) (0.0716) (0.122) Birth year −0.0866* 0.0850 37.77*** −0.797 0.139 0.688*** 0.465*** −0.0664 (0.0513) (0.0672) (5.526) (3.155) (0.328) (0.0401) (0.0984) (0.154) Observations 1,798 1,578 1,292 1,011 1,806 1,987 1,269 432 Panel B: By gender of the child Son × post-birth 0.0338 −0.114* −2.505 5.339* 0.110 0.0379 0.00613 −0.368** (0.0411) (0.0608) (3.901) (3.194) (0.137) (0.0561) (0.0862) (0.153) Son × birth year 0.0257 −0.288** −3.513 5.391 0.586 −0.0195 −0.0863 −0.228 (0.0957) (0.120) (10.72) (6.670) (0.547) (0.0791) (0.177) (0.203) Post-birth −0.101** 0.0468 36.23*** −7.106*** 0.200 0.609*** 0.958*** 0.250* (0.0431) (0.0545) (4.246) (2.396) (0.137) (0.0448) (0.0849) (0.144) Birth year −0.101 0.275*** 39.89*** −3.325 −0.226 0.702*** 0.518*** 0.0965 (0.0717) (0.0894) (8.290) (5.532) (0.252) (0.0621) (0.145) (0.144) Observations 1,798 1,588 1,292 1,011 1,806 1,987 1,269 432 Source: The estimates use data from the China Health and Nutrition Survey. Note: The sample is restricted to rural women who have data both before and after a birth. The regressions include fixed effects for year, age, and individual, and a constant term. The different sample sizes result from the questions being asked in different survey waves or subsets of households (such as those with children under age 12). Section 2 outlines all of the sample restrictions associated with each variable. Standard errors are clustered at the household level. *, **, *** denote significance at the 10 percent, 5 percent, and 1 percent levels, respectively. The only significant estimate for the interaction terms are negative effects on cigarette consumption, where the arrival of a son (relative to a daughter) corresponds to a large 34 percent decline in cigarette expenditure.34 This is significant at the 5 percent level. Similar to the NFP results, in the CHNS data in column 2 of table 7, there is a significant 11.4 percentage point decline in the probability that someone in the household smokes following the birth of a son relative to a daughter. The decline in cigarette consumption is consistent with a couple of mechanisms. First, given the negative health consequences to children of second-hand smoke, the result is consistent with the investment story where the households invest more in the health of sons than daughters, because they expect higher returns to investments in boys than in girls. Second, if mothers dislike people in the household smoking, it is also potentially consistent with the mechanism where the household rewards the mother for producing a son. Smoking is a male- favored good in China; over half of Chinese men in 2010 were current smokers with only 3.4 percent of Chinese women reporting ever smoking (Liu et al. 2017). Moreover, ethnographic studies with Chinese women suggest that they strongly dislike male smoking for a variety of reasons, including the health consequences and the associated dirtiness that falls on women to clean (Mao, Bristow, and Robinson 2012). 6.2. Migration The paper next considers the impact of the birth of a child on migration outcomes. In columns 1 and 2 of table 8, the outcome variable is the IHS function of the number of days that the person spent at home in the past year. In Panel A, the birth of a child corresponds to a significant increase in the number of days at 34 Figure S1.6 shows the event study estimates around birth. The relative decline in cigarette consumption for boys (given by the interaction term) persists for about four years. 276 Wang Table 8. Effects of Birth on Individual Migration and Earnings Away Days at home Attrition Earnings away Mothers Fathers Mothers Fathers Mothers Fathers (1) (2) (3) (4) (5) (6) Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Panel A: Birth of a child Post-birth 0.349*** 0.198*** 0.00613 0.0150*** −1202.9*** 521.5 (0.0332) (0.0302) (0.00712) (0.00577) (183.8) (572.5) Birth year 0.266*** 0.137*** −0.0285*** −0.0254*** −775.4*** 263.6 (0.0306) (0.0278) (0.00502) (0.00345) (143.3) (340.1) Observations 19,295 21,528 17,367 19,812 19,928 22,413 Panel B: By gender of the child Son × post-birth −0.0399 −0.122*** 0.0112 0.000125 −62.42 1616.4* (0.0539) (0.0475) (0.0108) (0.00876) (248.8) (898.3) Son × birth year −0.0560 −0.0771 0.00459 0.00536 −86.25 789.3 (0.0577) (0.0538) (0.00805) (0.00551) (242.2) (560.1) Post-birth 0.373*** 0.268*** -0.000329 0.0149** −1166.4*** −404.9 (0.0451) (0.0422) (0.00952) (0.00747) (208.0) (340.4) Birth year 0.298*** 0.182*** −0.0313*** −0.0285*** −724.7*** −204.6 (0.0432) (0.0426) (0.00684) (0.00515) (197.8) (308.2) Observations 19,295 21,528 17,367 19,812 19,928 22,413 Source: The estimates use data from the Chinese Ministry of Agriculture’s National Fixed Point Survey. Note: The regressions include fixed effects for year, age, and household, and a constant term. Days at home is transformed using the inverse hyperbolic sine function. Earnings away refers to earnings outside of the county of residence and are transformed into real 2002 yuan. Standard errors are clustered at the household level. *, **, *** denote significance at the 10 percent, 5 percent, and 1 percent levels, respectively. home for both mothers and fathers. These estimates are significant at the 1 percent level. Turning to the impacts for sons versus daughters in Panel B, there are no significant differences for mothers. In contrast, the number of days that the father is at home is significantly different for a son versus a daughter. More specifically, fathers are more likely to be away from home following the birth of a son than a daughter. The paper examines whether an individual attrites from the survey as a measure that would capture permanent migration. In contrast, the number of days at home versus away captures seasonal or temporary migration of household members. Panel A shows a significant 1.5 percent increase in attrition of fathers following the birth of a child but no change for mothers. There are no differences for sons versus daughters for these outcomes (in Panel B). Women earn significantly less money away from home following the birth of a child, while for men the coefficient is positive but not statistically different from zero at the standard levels (Panel A). Correspond- ing to the increase in temporary migration, there is an increase in the amount that fathers earn away from home (by RMB 1616) following the birth of their sons relative to daughters. This estimate is significant at the 10 percent level. This helps to explain how the results for sons and daughters can tie together: following the arrival of a son, mothers work less and fathers are more likely to migrate temporarily for work. Household consumption does not fall differentially because the composition of men’s work has changed.35 This change reflects a shift in household preferences following the birth of a son.36 35 As shown in column 1 of table 9, the labor supply of other people in the household (not including the parents) does not change significantly following the birth of a boy relative to a girl in the household. 36 This suggests that men could have earned more prior to the birth of a son, but migration lowers their utility perhaps because they must work longer hours in urban jobs or suffer from being apart from their family and friends. The World Bank Economic Review 277 6.3. Time Use The paper next turns to time-use outcomes available in the CHNS, including time spent on child care, shown in column 3 of table 7. Not surprisingly, the number of hours that women spend on child care increases dramatically (and significantly) post-birth by 34.9 hours per week. However, there is no signif- icant difference in the time spent on caring for a male or a female child. In fact, the coefficient estimate on the interaction between a son and post-birth is negative (−2 hours). Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 In terms of time spent on leisure (column 4), there is a drop of 4.6 hours following the birth of a child. Interestingly, women’s total leisure actually increases following the birth of a son relative to a daughter by 5.3 hours per week after birth. The difference between the effect of a girl versus a boy on leisure is statistically significant at the 10 percent level. After the birth of a child, a woman spends about 16 minutes more per day on food preparation and washing clothes. This estimate on the impact on these chores is significant at the 1 percent level. However, there is no significant difference in the time she spends on chores by the gender of the child. As expected, the results show a 68.8 percentage point increase in the probability of breastfeeding in the year that an infant is born and a 62.8 percentage point increase after the birth year (in column 6). These estimates are significant at the 1 percent level. However, there are no significant differences in breastfeed- ing by gender. These results are in contrast to papers that show differences in rates of breastfeeding by gender in India (e.g. Jayachandran and Kuziemko 2011; Barcellos, Carvalho, and Lleras-Muney 2014). However, they are consistent with evidence presented by Anderson and Ray (2010) that finds evidence of discrimination against girls after birth in India but not in China.37 The paper also consider whether households spend time and money investing in the health of their children. In column 7, the number of immunizations that children under 12 receive in the household increases by 0.47 in the year of birth and by 0.96 in the year following the birth. Both of the estimates are significant at the 1 percent level. However, as with breastfeeding, there are no significant differences by the gender of the child.38 Finally, the CHNS allows for consideration of whether women participate in household decisions to purchase durable goods or whether men make them without any input from women. In column 8, there is no effect of the birth of a child on decision-making by men, but this differs by the gender of the child. There is a 36.8 percent decrease in the probability of men making household decisions without input from their wives following the birth of a son relative to a daughter. This estimate is significant at the 5 percent level. Overall, the results from the CHNS suggest that the decline in labor supply of women corresponds to women shifting that time primarily into child care, with a small increase in time spent on household chores and no change in time for leisure. The results by the gender of child are also most consistent with the idea that women are being rewarded for having a boy, with less work, more leisure, and more power in household decision-making. Despite working less following sons than daughters, they are not spending more time taking care of their sons or breastfeeding them more. While the results on leisure are also potentially consistent with the investment mechanism where women exert more effort (but not time) with their sons than their daughters, under the investment story the additional effort is exhausting, so women may need more leisure to be able to maintain this high level of interaction and investment for sons. However, the results on breastfeeding, immunizations, and male decision-making are not supportive of the idea that households are generally making more investments in boys than girls. Unfortunately, there are no questions about child care, leisure, or female participation in household decision-making in the NFP data. However, there is some limited information about time use in a variable 37 Anderson and Ray (2010) find evidence that there is a skewed sex ratio at the time of birth in China, but not thereafter. 38 This is aggregated to the household level in order to run the same estimates as the outcomes in the CHNS. However, in estimates of the impact of having a boy versus a girl on whether that specific child is immunized or breastfed in a regression using only post-birth data, there are also no significant gender differences. 278 Wang Table 9. Effects of a Birth on Others’ Work, School Enrollment of Parents, and Household Size Mother Father Number of Household Others’ days in school in school children size of work (1) (2) (3) (4) (5) Panel A: Birth of a child Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Post-birth −0.0104*** −0.00118 0.854*** 0.623*** −0.0111 (0.00399) (0.00399) (0.0175) (0.0325) (0.0708) Birth year −0.00753** −0.00281 0.0737*** 0.000375 0.0119 (0.00352) (0.00326) (0.0164) (0.0308) (0.0633) Observations 19,568 21,956 19,928 19,928 19,928 Panel B: By gender of the child Son × post-birth 0.0133* −0.0117* −0.0186 −0.0803 −0.0643 (0.00804) (0.00706) (0.0289) (0.0534) (0.109) Son × birth year 0.00378 −0.00989 0.0466 −0.0530 −0.0439 (0.00855) (0.00820) (0.0312) (0.0568) (0.116) Post-birth −0.0181*** 0.00548 0.864*** 0.670*** −0.141 (0.00678) (0.00549) (0.0237) (0.0453) (0.0920) Birth year −0.00986 0.00292 0.0470** 0.0320 −0.0312 (0.00643) (0.00568) (0.0235) (0.0444) (0.0932) Observations 19,568 21,956 19,928 19,928 19,928 Source: The estimates use data from the Chinese Ministry of Agriculture’s National Fixed Point Survey. Note: The dependent variable is an indicator for whether the mother and father are currently enrolled in school in columns 1 and 2, respectively. The dependent variable is the number of children in column 3 and total household size in column 4. It is an inverse hyperbolic sine function of the number of days worked by people in the household excluding the mother and father in column 5. The regressions include fixed effects for year, age, and household, and a constant term. Standard errors are clustered at the household level. *, **, *** denote significance at the 10 percent, 5 percent, and 1 percent levels, respectively. in the NFP data about current enrollment in school. Table 9 shows a drop in women’s school enrollment of 1 percentage point (significant at the 1 percent level) following the birth of a child. Given that only 1.26 percent of women were enrolled in school prior to birth, this means that essentially all women who were in school dropped out following the birth of their daughters. However, the coefficient on the interaction between birth and a son is 1.3 percentage points, indicating that women are more likely to be in school following the birth of a son than a daughter. Consistent with the child-care results from the CHNS, the results on school enrollment using the NFP data show that the shift away from work following the birth of a son is not entirely about shifting into child care. Mothers of sons shift into other activities beyond child care, including remaining in school and leisure. Thus, these results are consistent with the mechanism whereby women are rewarded for producing a son. However, the results are also consistent with the investment mechanism if households believe that having educated mothers represents an investment in the children.39 6.4. Fertility, Household Size, and Other Adults’ Labor Supply Columns 3 and 4 of table 9 examine the impacts on the number of children and household size. It is interesting to note that the number of children increases significantly by 0.85 after birth, but the total household size only increases by 0.62; this indicates that some other adults are likely to leave the household after the arrival of a child.40 However, there is no significant change in the number of days of labor supplied 39 There is a large literature in economics linking mothers’ education outcomes with the outcomes of children. See for example Thomas, Strauss, and Henriques (1991) and Currie and Moretti (2003). 40 In other words, the arrival of a child drives other members out rather than attracting the coresidence of grandparents. The World Bank Economic Review 279 Table 10. Heterogeneity in the Impact of a Birth on Labor-Force Participation Agric Family Grandpar Young High Skewed Var: ind occ present mom edu Minority sex ratio (1) (2) (3) (4) (5) (6) (7) Post-birth × var 0.0588** 0.0368 −0.0512** −0.0226 0.00992 0.0351 −0.0247 Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 (0.0240) (0.0263) (0.0248) (0.0234) (0.0249) (0.0277) (0.0208) Birth year × var 0.104*** 0.0882*** −0.0881*** −0.0727*** −0.0231 0.0253 0.00330 (0.0242) (0.0262) (0.0248) (0.0238) (0.0254) (0.0279) (0.0233) Post-birth −0.122*** −0.114*** −0.0679*** −0.104*** −0.0979*** −0.117*** −0.0998*** (0.0230) (0.0253) (0.0242) (0.0164) (0.0198) (0.0142) (0.0159) Birth year −0.133*** −0.135*** −0.0124 −0.0526*** −0.0663*** −0.0909*** −0.0889*** (0.0198) (0.0224) (0.0214) (0.0151) (0.0165) (0.0131) (0.0158) Observations 10,324 10,374 15,993 19,799 10,585 19,304 19,928 Source: The estimates use data from the Chinese Ministry of Agriculture’s National Fixed Point Survey. Note: The dependent variable is female labor-force participation. Each column is a regression, where the column label indicates the variable that is interacted with the regressors. The regressions include fixed effects for year, age, and household, and a constant term. Standard errors are clustered at the household level. *, **, *** denote significance at the 10 percent, 5 percent, and 1 percent levels, respectively. by other adults in the household following the birth of a child. None of these estimates are significantly different following the birth of a son versus a daughter. 7. Heterogeneity The paper examines the type of work that the woman was engaged in prior to birth in order to consider whether discrimination by firms or inflexible work arrangements in the formal labor market can explain the decline in female labor-force participation following the birth of a child. It examines heterogeneity by whether the women’s primary occupation prior to birth was in a family enterprise and whether her primary industry was agriculture. The results are presented in columns 1 and 2 of table 10, where the outcome is the indicator for labor-force participation. In the year of birth (which includes several months of pregnancy for many women), the negative impact on women’s labor-force participation is mostly offset for women working in agriculture and in a family enterprise. However, in the years following birth, there are no significant differences in the impact of a child on women’s likelihood of working by whether the women’s occupation is defined as in family enterprises (column 2). In column 1, the negative impacts of birth are mitigated slightly for women in agriculture and offset about half of the total decline following the birth of a child. Next, the paper looks at heterogeneity by the presence of a coresident grandparent prior to the birth. A grandparent may be able to provide care for young children and allow women to continue to work following birth.41 In column 3, there is a larger decline in female labor-force participation in the year of birth, as well as the years following birth, in households with a grandparent present. These estimates are significant at the 1 percent and 5 percent levels, respectively. Columns 4 through 6 show heterogeneity along demographic characteristics of the mother. More specifically, they show whether the mother’s age at her child’s birth was above average, whether the woman had above average education, and whether the household is an ethnic minority (not ethnic Han Chinese). In the years after birth, there are no significant differences in the impact of having a child for younger mothers, more educated mothers, or ethnic minorities. 41 Alternatively, a coresident grandparent may be able to substitute into women’s labor activities in agriculture or a family enterprise, which may then work in the opposite direction. 280 Wang Finally, the paper considers whether mothers’ labor-supply response is stronger in areas where there is a stronger preference for sons. It examines heterogeneity by the provincial sex ratio. If the sex ratio is less skewed than average, the variable equals 0 and if the sex ratio in the province is more skewed than average, then the variable equals 1.42 There are no significant differences in the participation response of mothers to children by the provincial sex ratio. Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 8. Conclusion This paper provides new evidence on the impact of a birth on the labor supply of rural women in China. Women reduce their probability of working and the number of work days following the birth of a child. In contrast, men’s labor supply increases following the birth of a child. While men’s labor supply increases, it is not sufficient to totally offset the losses associated with the decline in women’s labor and total household income falls by about 5 percent. The composition of household consumption changes with shifts towards milk and meat and away from cigarettes. The magnitude of the effects on women documented in this paper of an 11 percent decline in female labor-force participation are somewhat smaller than the estimates seen in countries like the United States, United Kingdom, and Denmark. Furthermore, the results in rural China dissipate more over time than in the higher-income countries. One possible explanation for the differences in magnitudes and persistence is that the kind of formal sector jobs that employ women in high-income countries are too rigid and do not provide the kind of flexibility that mothers of small children need. Another possibility is that employ- ers discriminate against working mothers. It is important to note, however, that the magnitudes in China are about half of what is observed in high-income countries. In the rural Chinese context, in which most women are engaged in family agricultural enterprises, changes in the labor supply of women in the con- text where family agricultural work is the primary source of work is informative about the preferences of households. Thus, household preferences, and not rigid labor-market arrangements or labor-market dis- crimination, explain a substantial amount of impact of parenthood on women’s labor-supply decisions. Policies that encourage companies to offer parents more flexible work arrangements or reduce discrimina- tion against mothers are then unlikely to fully eliminate the impact of children on women’s labor-market outcomes. The paper also documents strong differences in mothers’ labor-supply response to the arrival of a son versus a daughter, but no difference by the gender of the child for fathers’ labor supply. In our data, there is no strong support for the results being driven by different wealth effects or fertility effects associated with boys and girls. Given the large increase in leisure of mothers following a son’s birth, the increased participation in decision-making, and the lack of increase in time inputs into boys relative to girls, the results are most consistent with the idea that mothers are being rewarded by households with less work and more leisure for producing sons. Another leading mechanism is that households are simply investing more in sons than daughters because the returns to investment in boys are higher. Under the investment mechanism, mothers spend more effort per unit of time on their sons and this effort is tiring and leads them to need more leisure. However, there is no evidence that there are other increases in investment in boys over girls in terms of immunizations, breastfeeding, and consumption of milk and meat. The results are consistent with Anderson and Ray (2010) who show that there is a skewed sex ratio at birth in China, but no evidence of household discrimination against girls and in favor of boys after birth. The microlevel estimates in this paper suggest a potential linkage in the macrolevel trends in falling female labor-force participation rates and the rise in sex ratios over time in China. Given that the results show that rural women work less following the birth of sons relative to daughters, this may explain part of the overall fall in women working in China. Furthermore, this suggests that policies that affect the 42 The provincial sex ratio used here is calculated within the NFP sample. The World Bank Economic Review 281 sex ratio, including changes to the One Child Policy, may then have unintended consequences on female labor-force participation. Data Availability Statement Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 There are two main data sets used in this analysis. The China Health and Nutrition Survey data are available in a public repository at https://www.cpc.unc.edu/projects/china. For the Chinese Ministry of Agriculture data, researchers must apply for access to the data. See http://www.rcre.moa.gov.cn/. We will provide the Stata do files that correspond to the analysis presented in the paper. References Afridi, F., T. Dinkelman, and K. Mahajan. 2018. “Why Are Fewer Married Women Joining the Work Force in Rural India? A Decomposition Analysis over Two Decades.” Journal of Population Economics 31(3): 783–818. Ananat, E. O., and G. Michaels. 2008. “The Effect of Marital Breakup on the Income Distribution of Women with Children.” Journal of Human Resources 43(3): 611–29. Anderson, S., and D. Ray. 2010. “Missing Women: Age and Disease.” Review of Economic Studies 77(4): 1262– 300. Angelov, N., P. Johansson, and E. Lindahl 2016. “Parenthood and the Gender Gap in Pay.” Journal of Labor Economics 34(3): 545–79. Anukriti, S., S. Kwon, and N. Prakash. 2022. “Saving for Dowry: Evidence from Rural India.” Journal of Development Economics 154: 1027–50. Bailey, M. J. 2006. “More Power to the Pill: The Impact of Contraceptive Freedom on Women’s Life Cycle Labor Supply.” Quarterly Journal of Economics 121(1): 289–320. Barcellos, S. H., L. S. Carvalho, and A. Lleras-Muney. 2014. “Child Gender and Parental Investments in India: Are Boys and Girls Treated Differently?” American Economic Journal: Applied Economics 6(1): 157–89. Bedard, K., and O. Deschenes. 2005. “Sex Preferences, Marital Dissolution, and the Economic Status of Women.” Journal of Human Resources 40(2): 411–34. Borusyak, K., and X. Jaravel. 2022. “Revisiting Event Study Designs.” SSRN Working Paper 2826228. Chung, Y., B. Downs, D. H. Sandler, and R. Sienkiewicz et al. 2017. “The Parental Gender Earnings Gap in the United States.” Working Papers 17–68, Center for Economic Studies, U.S. Census Bureau. Cortés, P., and J. Pan. 2020. “Children and the Remaining Gender Gaps in the Labor Market.” Technical report, National Bureau of Economic Research. Currie, J., and E. Moretti. 2003. “Mother’s Education and the Intergenerational Transmission of Human Capital: Evidence from College Openings.” Quarterly Journal of Economics 118(4): 1495–532. Dahl, G. B., and E. Moretti. 2008. “The Demand for Sons.” Review of Economic Studies 75(4): 1085–120. Dustmann, C., and R. Landersø. 2018. “Child’s Gender, Young Fathers’ Crime, and Spillover Effects in Criminal Behavior.” Technical report, Centre for Research and Analysis of Migration (CReAM). Ebenstein, A. 2010. “The ‘Missing Girls’ of China and the Unintended Consequences of the One Child Policy.” Journal of Human Resources 45(1): 87–115. Fan, M., and Q. Qing. 2014. “After Having a Son, I Really Started to Understand My In-Laws You erzi hou, wo zhenzheng lijiele gongpo.” Qingdao Morning Post (Qingdao Zao Bao). Fletcher, E., R. Pande, and C. M. T. Moore. 2017. “Women and Work in India: Descriptive Evidence and a Review of Potential Policies.” HKS Working Paper No. RWP18–004. Goldin, C. 1994. “The U-Shaped Female Labor Force Function in Economic Development and Economic History.” Technical report, National Bureau of Economic Research. Goldin, C., and L. F. Katz. 2000. “Career and Marriage in the Age of the Pill.” American Economic Review 90(2): 461–65. ———. 2002. “The Power of the Pill: Oral Contraceptives and Women’s Career and Marriage Decisions.” Journal of Political Economy 110(4): 730–70. 282 Wang Gruijters, R. J. 2018. “Daughters’ and Sons’ Remittances in Rural China: Findings from a National Survey.” Journal of Family Issues 39(11): 2911–34. Gupta, M. D. 1987. “Selective Discrimination against Female Children in Rural Punjab, India.” Population and De- velopment Review 13(1): 77–100. Heath, R., and A. M. Mobarak. 2015. “Manufacturing Growth and the Lives of Bangladeshi Women.” Journal of Development Economics 115: 1–15. Downloaded from https://academic.oup.com/wber/article/37/2/257/7056488 by Joint Bank-Fund library user on 04 September 2023 Heath, R., and X. Tan. 2018. “Worth Fighting For: Daughters Improve their Mothers’ Autonomy in South Asia.” Journal of Development Economics 135: 255–71. Ichino, A., E.-A. Lindström, and E. Viviano. 2014. “Hidden Consequences of a First-Born Boy for Mothers.” Eco- nomics Letters 123(3): 274–78. Jayachandran, S., and I. Kuziemko. 2011. “Why Do Mothers Breastfeed Girls Less than Boys? Evidence and Implica- tions for Child Health in India.” Quarterly Journal of Economics 126(3): 1485–538. Jensen, R. 2003. “Equal Treatment, Unequal Outcomes?: Generating Sex Inequality through Fertility Behaviour.” Citeseer. ———. 2012. “Do Labor Market Opportunities Affect Young Women’s Work and Family Decisions? Experimental Evidence from India.” Quarterly Journal of Economics 127(2): 753–92. Kishore, A., and D. Spears. 2014. “Having a Son Promotes Clean Cooking Fuel Use in Urban India: Women’s Status and Son Preference.” Economic Development and Cultural Change 62(4): 673–99. Kleven, H., C. Landais, J. Posch, A. Steinhauer, and J. Zweimuller. 2019. “Child Penalties across Countries: Evidence and Explanations.” AEA Papers and Proceedings, 109: 122–26. Kleven, H., C. Landais, and J. E. Søgaard. 2019. “Children and Gender Inequality: Evidence from Denmark.” American Economic Journal: Applied Economics 11(4): 181–209. Kuziemko, I., J. Pan, J. Shen, and E. Washington. 2018. “The Mommy Effect: Do Women Anticipate the Employment Effects of Motherhood?” Technical report, National Bureau of Economic Research. Lebedinski, L., C. Perugini, and M. Vladisavljevic. 2022. “Child Penalty in Russia: Evidence from an Event Study.” Review of Economics of the Household. Li, L., and X. Wu. 2011. “Gender of Children, Bargaining Power, and Intrahousehold Resource Allocation in China.” Journal of Human Resources 46(2): 295–316. Liu, S., M. Zhang, L. Yang, Y. Li, L. Wang, Z. Huang, L. Wang, Z. Chen, and M. Zhou. 2017. “Prevalence and Patterns of Tobacco Smoking Among Chinese Adult Men and Women: Findings of the 2010 National Smoking Survey.” J. Epidemiol. Community Health 71(2): 154–61. Lundberg, S., and E. Rose. 2002. “The Effects of Sons and Daughters on Men’s Labor Supply and Wages.” Review of Economics and Statistics 84(2): 251–68. Mammen, K., and C. Paxson 2000. “Women’s Work and Economic Development.” Journal of Economic Perspectives 14(4): 141–64. Mao, A., K. Bristow, and J. Robinson. 2012. “Caught in a Dilemma: Why Do Non-smoking Women in China Support the Smoking Behaviors of Men in Their Families?” Health Education Research 28(1): 153–64. Pham-Kanter, G. 2010. “The Gender Weight Gap, Sons, Daughters, and Maternal Weight.” University Library of Munich MPRA Paper No. 28997. Rosenblum, D. 2013. “The Effect of Fertility Decisions on Excess Female Mortality in India.” Journal of Population Economics 26(1): 147–80. Thomas, D., J. Strauss, and M.-H. Henriques. 1991. “How Does Mother’s Education Affect Child Height?” Journal of Human Resources 26(2): 183. Wei, S.-J., and X. Zhang. 2011. “The Competitive Saving Motive: Evidence from Rising Sex Ratios and Savings Rates in China.” Journal of Political Economy 119(3): 511–64. Yangtse News. 2016. “The Fierce Mother-in-Law Regularly Takes the Train to Abuse Her Daughter-in-Law for Not Giving Birth to a Son. (Erxi wei Nengsheng xia Nanhai Xionghan Popo Dingqi zuo Gaotie Shsngmen Ouda.).” Zhu, H. 2016. “Adult Children’s Characteristics and Intergenerational Financial Transfers in Urban China.” Chinese Journal of Sociology 2(1): 75–94. Zimmermann, L. 2012. “Reconsidering Gender Bias in Intrahousehold Allocation in India.” Journal of Development Studies 48(1): 151–63.