According to research conducted in Western societies, religious involvement is significantly and positively associated with subjective well-being (SWB). However, empirical evidence in Taiwan hardly shows significant differences in SWB between people with and without religious affiliation. Researchers argue this was because the analysis of SWB was biased by self-selection, so religious involvement should still have a positive influence on SWB. Nevertheless, due to limitations of research methods, existing research has not confirmed the causal order of religious involvement and SWB and cannot rule out the possible reverse explanation that “people with a higher SWB tend to attend religious activities.” Accordingly, this research uses survey data collected in the years 2011 and 2014 by the Taiwan Youth Project to examine whether individuals having religious affiliation in 2011 reported a significantly higher SWB in 2014. In order to deal with the potential selection bias on religious affiliation in 2011, the inverse probability of treatment weighting (IPTW) method is used to apply counterfactual inference of the religious function on SWB. Resultssuggest that, after applying IPTW, all the average treatment effects (ATE), average treatment effects for the treated (ATT), and average treatment effects for the controls (ATC) of religious affiliation on SWB become positively significant, implying religious affiliation has a positive influence on SWB among Taiwanese young adults when selection bias is controlled. Keywords: religious involvement, subjective well-being, self-selection bias, counterfactual inference

With the transformation of social structures and family norms in Taiwan, universal marriage has been gradually replaced by the postponement and abandonment of marriage. Marriage is no longer a necessary and inevitable choice for individuals. Under this trend, it is therefore important to ask whether married people are happier than their unmarried counterparts and who benefits from marriage? This paper aims to investigate whether marriage confers happiness among young adults and who benefits from marriage in Taiwan. Specifically, this project will estimate the heterogeneous treatment effects of marriage on happiness, gauging who benefits from marriage. This paper utilizes the 2011 and 2017 Taiwan Youth Project (TYP), a longitudinal dataset collected in the northern part of Taiwan. Propensity score weighting is applied to address selection on observables and to estimate heterogeneous treatment effects of marriage on happiness. The results show that marriage has a positive effect on happiness, and this positive effect is not explained by selection effects. With regards to heterogeneous treatment effects of marriage, there are no heterogeneous effects of marriage on happiness for men. Marriage confers happiness for men, regardless of their propensity to get married. However, not all women benefit from marriage. Marriage only confers happiness for women who have a high propensity to get married but has limited benefits for those who have low propensity to get married. Simply speaking, the results suggest that marriage remains an important social determinant of happiness for young adult men and women who have high propensity to get married, but less so for women with low propensity to get married. Official policies that encourage marriage should target the right population and avoid a one-sizefits- all solution. Keywords: marriage, social selection, propensity score weighting, heterogeneous treatment effects, happiness

In this study, we examined the effects of family disruption, including parental divorce, death, and remarriage, during the period of children’s childhood and adolescenthood on their later educational achievement by using sibling data from the 2000-2011 Taiwan Youth Project Phases I and II. By using the birth years of parents and their children, and the year that family disruption occurred, we construct the longitudinal status of family structure for each child from age 0-18. In this study, the educational achievement is assessed by taking into account the family unobserved heterogeneity and right censoring. Our results show that the three statuses of family structure: parental divorce, death, and remarriage, are all significantly and negatively associated with their children’s educational achievement. Parent’s divorce has the largest and long-lasting negative impacts on their children’s educational achievements. The parental divorce occurring during the childhood period has the largest accumulated impacts. The marginal effect, however, are largest during the adolescenthood of their children.Keywords: parental separation, educational achievement, random-effect Tobit model, sibling longitudinal data