Son Preference, Parental Satisfaction, and Sex Ratio Transition Junji - PDF document

Son Preference, Parental Satisfaction, and Sex Ratio Transition Junji Kageyama a , Department of Economics, Meikai University Risa Hagiwara , Department of Economics, Meikai University Kazuma Sato , Department of Economics, Takushoku University Eriko Teramura , Department of Economics, Meikai University April, 2017 Abstract This study aims to understand the sources of son preference using satisfaction data in various domains of life. To do this, we use Korean panel data and apply regression analyses. The results show that sons better satisfy parents in the domains of household income, relations with relatives, and social relations at the timing of birth, while no advantage is found for daughters. These results are consistent with the idea that parents expect sons to contribute to the family in earning income, financially support aged parents, and represent the family in relative and social networks. Namely, we can argue that such expectation causes the parent of a newborn boy to be more satisfied in related domains of life and is manifested as a preference for sons. These results are also consistent with the idea that socioeconomic changes occurring in the demographic transition eventually lower the sex ratio even in a country with son preference. Socioeconomic changes, such as the introduction of social security system, the trend toward the nuclear family, more equal gender roles, and more working opportunity for females, cause the traditionally expected roles of sons to be less valuable and, therefore, the son preference to be weaker. This lowers the sex ratio at birth without any change in fertility. a Corresponding author. kagejun@me.com. This research is supported by Grant-in-Aid for Scientific Research from JPSS in Japan (263880243, 17KT0037). 1

1. Introduction This study aims to understand the sources of son preference using satisfaction data in various domains of life (See Van Praag et al. 2003; Easterlin 2006 and references therein for the domain-of-life approach). In particular, we examine whether the gender of children affects life satisfaction and domain-specific satisfaction in South Korea where culturally sons are preferred to daughters (See Das Gupta et al. 2003; Guilmoto 2009, 2012; Hesketh and Xing 2006 for reviews). The sex ratio at birth is naturally around 105 (105 boys per 100 girls), but, in South Korea, it hit 116.5 in 1991 and remained high in the 1990s and in the 2000s (Statistics Korea 2017). The effect of child gender on subjective well-being has been previously studied by Lee et. al. (2013) and Margolis and Myrskyla (2016) as far as we know. Lee et. al. (2013) used survey data conducted for the elderly (65 or older) in Kangwha County in South Korea, a rural island 50km from Seoul, and demonstrated that life satisfaction is highest when the parent has both sons and daughters. Margolis and Myrskyla (2016), on the other hand, used British and German longitudinal data and showed that the gender mix of children has little impact on parental satisfaction in European countries. However, no study has yet to employ panel data to examine the impact of the gender of children on parental satisfaction in a country with son preference, nor turn to domain-specific satisfaction. The reason for using domain-specific satisfaction data is that these data “reveal” parents’ feelings. For example, if having a boy, and not a girl, raises parental satisfaction in the financial domain, we can infer that parents feel positively about having a boy in the financial domain and can consider it as a source for son preference. This study is the first study to use domain-specific satisfaction to analyze the sources of son preference. By identifying sources of son preference, we can also provide evidence for explaining the sex ratio transition (Guilmoto 2009). Even in countries where sons are traditionally preferred, the sex ratio does not stay high for long and eventually falls toward the natural level (See Allendorf 2012; Angrist 2002; Bhattacharjya et al. 2008; Bongaarts 2013; Chung and Das Gupta 2007; Das Gupta et al. 2003; Diamond-Smith and Bishai 2015; Ebenstein and Leung 2010; Edlund and Lee 2013; Lin 2009). We apply the obtained results to explain this phenomenon. The reminder of the paper is organized as follows. The next section explains data and methods. Section 3 presents results. Our main findings are that sons better satisfy 2

parents in the domains of household income, relations with relatives, and social relations at the timing of birth, while no advantage is found for daughters. These results are consistent with the idea that parents expect sons to contribute to the family in earning income, financially support aged parents, and represent the family in relative and social networks. Section 4 concludes. 2. Data and Methods The data are taken form Korean Labor and Income Panel Survey, Waves 1-17 (Korean Labor Institute 2016). The survey was first conducted in 1998 for 13,321 individuals in 5,000 households. The survey tracks not only original households but also branch households, and added 1,415 households in Wave 12. As a result, Wave 17 covers 13,163 individuals in 6,738 households. The data set contains demographic and socio-economic characteristics, including the numbers of female and male children. The data set also includes overall life satisfaction and satisfaction in domains of household income, family relations, leisure activities, housing environment, relations with relatives (from Wave 3 onwards), and social relations (from Wave 3 onwards), all of which are coded between 1 (very dissatisfied) and 5 (very satisfied). This study employs Waves 2-17 for which the necessary data are available. We focus on married individuals since it is rare to have children without getting married in South Korea. In total, our data set contains 8,821 individuals and 74,523 observations. The descriptive statistics for key variables in the most recent wave appear in Table 1. [Table 1 around here] With this data set, we regress three models to examine the following relationships: (1) the gender of existing children and the probability of progressing to the next parity, (2) children and parental satisfaction, and, most importantly, (3) the gender of children and parental satisfaction. The aims of these models are: (1) to make sure that son preference translates into actual behavior, (2) to compare the impact of having children in South Korea with other countries, and (3) to test if son preference translates into parental satisfaction. For the second model, we include both the number of children and birth dummy 3

(having a child in the last year, including adoption) as explanatory variables. We use the birth dummy to separate the impact of birth (becoming a parent of a new child) from having children (being a parent). By doing this, we can distinguish the instantaneous and continuous impact of having children. However, to test the validity of the birth dummy, we regress the model with and without the birth dummy. We use the same strategy for the third model and employ the numbers of female/male children and female/male birth dummies. In particular, we expect that female/male birth dummies reveal parents’ feelings about having a female/male child. With respect to the model specification, we employ a pooled logit model for the first model. For the second and third models, we follow previous studies that treat satisfaction levels as cardinal and that apply the ordinary least squared (OLS) method with fixed individual effects (Ferrer-i-Carbonell and Frijters 2004; de Ree and Alessie 2011; Van Landeghem 2011; Frijters and Beatton 2012; Kassenboehmer and Haisken-DeNew 2012; Wunder et al. 2013). As Ferrer-i-Carbonell and Frijters (2004) has observed, “assuming ordinality or cardinality of happiness scores makes little difference, whilst allowing for fixed-effects does change results substantially.” For all regression models, we control for age, age-squared, household-size adjusted real income, standard income for measuring relative income (the average of household-size adjusted real income in the same year, in the same sex, in the same age cohort (5-year classification), and in the same district), job, ownership of the house, living district (15 districts), and the survey year, all of which are commonly used in the subjective well-being literature. 3. Results 3.1. Probability of Progressing to the Next Parity For this model, we focus on parents with either only daughters or only sons. We also exclude parents at age fifty and above since the chance of having a new child becomes very low. Table 2 presents the results. Equation (1-1) includes all the parents regardless of the number of existing children, and equations (1-2) and (1-3) are respectively restricted to parents with one existing child and parents with two existing children. [Table 2 around here] 4