Developed and developing countries alike are experiencing demographic shifts toward an aging population. As evidenced by the seventh national census of China in 2021, 264 million people are aged 60 and above, representing 18.70% of the total population, a figure that has doubled since 2000. Remarkably, this number surpasses the entire population of the world’s fifth most populous nation. Projections indicate that this proportion will escalate to a staggering 40% by 2050. Within this context, health risks have emerged as a prominent concern given their upward trajectory across the life cycle.

This study examines the repercussions of co-resident parental health shocks on the labor supply of younger cohorts in China. Amid limited provision of market and public insurance, family has a vital role in risk-sharing in developing countries (Fafchamps and Lund, 2003, Liu, 2016). Through this mechanism, health risks stemming from population aging may extend their reach beyond the elderly and impact younger generations. Intrahousehold specialization can induce the unequal distribution of costs, particularly for females (Dercon and Krishnan, 2000). While existing studies examining developed countries have found no evidence of such spillover effects (Rellstab et al., 2020), do parental health shocks affect the labor market outcomes of the younger generation in developing countries? Do these effects disproportionately impact females? If so, to what extent does these impact persist? Empirical answers to these questions can help elucidate the costs of family risk-sharing and offer insights into the potential dividends of market and social insurance reform in the developing world.

We begin by emphasizing the dual shortage in formal care and health insurance faced in many developing countries, positing that parental health shocks can generate a substitution effect that elevates the demand for informal care and diminishes labor supply, and an income effect that arises from heightened medical costs and motivates labor participation. These effects have significance non only for theoretical considerations but also for empirical and econometric practices. Empirically, existing research has predominantly scrutinized the substitution effect, specifically gauging the link between informal care and female labor supply. However, it is essential to consider income effects for a comprehensive understanding of the overall repercussions of parental health shocks on adult children. Methodologically, these dual effects presents challenges to the conventional instrumental variable (IV) approach, rendering it unable to satisfy monotonicity and exclusion restriction assumptions.

In response, this study uses the staggered difference-in-differences (DiD) design to elicit empirical causal evidence. The ideal dataset would require detailed and longitudinal information on both parental health and adult children’s economic and demographic variables. The longitudinal dataset must also be mature with sufficient waves to capture any long-term effects and provide credible pre-trend tests. It must also be nationally representative to shed light on the global consequences of population aging. Our empirical analysis uses the China Family Panel Studies (CFPS) dataset spanning from 2012 to 2020, which uniquely meets these requirements. Following recent studies (García-Gómez et al., 2013, Dobkin et al., 2018, Rellstab et al., 2020), we define parental health shock as the initial hospitalization of either adult children’s parents or parents-in-law. This measure mitigates reporting bias inherent in subjective health variables. We also demonstrate that hospitalization, while capturing severe symptoms, exhibits prevalence, persistence, and a significant upward trajectory with age.

We investigate labor supply effects along the extensive margin, gauged through employment status, and the intensive margin, measured by weekly work hours conditional on employment. Our benchmark results reveal a noteworthy decline in female employment rates, with an average reduction of 3.7 percentage points subsequent to parental health shocks. Conversely, no significant change is found in male adult children’s employment. The results also find no significant changes in work hours. The shifts observed on the extensive margin are particularly likely to have scarring effects on future labor market performance. Therefore, an important concern is whether females affected by parental health shocks are capable of quickly returning to employment. Additionally, any negative impact can also be fueled by the persistent nature of health shocks at older ages. We then undertake a dynamic analysis using the event-study approach. Strikingly, our investigation demonstrates no evidence of recovery in female employment after at least 6 years following a shock, which constitutes the maximum span of our sample.

Subgroup analysis reveals strong gradients in employment effects based on individual income and family wealth. Notably, some subgroups exhibit an increase in employment following a parental shock, indicating strong income effects. Specifically, females with median income experience a 4.77 percentage point decline in employment following a parental health shock, whereas the effects turn positive for those with top decile income, and the male labor income gradient is not evident. Interestingly, a male employment gradient emerges along household wealth, wherein males from households with assets below the median witness a 1.90 percentage point surge in employment post-shock. Strikingly, this effect escalates to 6.02 percentage points for those from households with assets below the quintile threshold. However, no such gradients are found for females. These positive effects cannot be explained by the rising demand for informal care and are consistent with the presence of income effects. We also explore heterogeneous effects based on education, marital status, the distinction between parental and parent-in-law shocks, and urban and rural residents.

Our robustness analyses reveals significant employment impacts of adult children’s own hospitalization, supporting the validity of using hospitalization as a measure for health shocks. Remarkably, females recover more quickly from their own health shocks. We also follow the recent methodological advancements in DiD with treatment effect heterogeneity (Callaway and Sant’Anna, 2021) to rule out the “forbidden comparison”, which is particularly pertinent to our context due to our keen focus on dynamic effects (Sun and Abraham, 2021). We also restrict the sample to individuals with no prior hospitalization records to examine more unanticipated shocks. Additionally, our robustness analysis narrows the sample to adult children of prime working age, a cohort that is expected to demonstrate robust labor market attachment and resilience to adverse shocks. We also control for functions fully saturated in the age of each parent, which is a critical factor associated with parental health and children’s labor supply. The results remain robust across these various tests.

This study pioneers a causal exploration into the impact of parental health shocks on the labor supply of adult children within the context of developing countries. While research has demonstrated the sensitivity of female labor supply to spouses’ health shocks (Coile, 2004, Fadlon and Nielsen, 2021) or from young children (Gould, 2004, Eriksen et al., 2021, Breivik and Costa-Ramón, 2022), this study investigates a new source of risk, parental health shocks, with particular concern for rapidly aging populations.2 Previous studies examining developed countries have indicated no effects or minimal gender disparities. For instance, in the Netherlands, Rellstab et al. (2020) found no significant effect of parental hospitalization on children’s labor supply, attributing it to the country’s well-established formal care system. In Austria, Frimmel et al. (2023) determined that both female and male employment decreased following a parental stroke, with a modest gender difference.3 Interestingly, the authors found that the decline disappeared following the liberalization of the formal care market.

The consequences of parental health shocks loom large and unequal in developing countries, where societies have strong cultural family ties and insufficient formal insurance. Our findings reveal a persistent effect of approximately four percentage points that is specific to women, highlighting the nontrivial and uneven burden associated with population aging. A concurrent paper by Brito and Contreras (2024) provides complementary evidence from Chile, finding a negative and persistent impact of initial parental cancer hospitalization on female employment, and a weakly positive effect for males. Their estimates cover families regardless of residential arrangements, whereas our results are based on families that reside together.4

More broadly, this study contributes to the research regarding the adverse effects of health shocks on labor supply, such as Bound et al., 1999, García-Gómez, 2011, Cai et al., 2014 and Dobkin et al. (2018), among others. These studies primarily focused on the impact of shocks on individuals themselves. Notably, Dobkin et al. (2018) employed the event study approach and provide early evidence based on a quasi-experimental design. They found that individuals’ employment rate decreases by 8.9 percentage points one year following hospitalization and by 11.1 points three years later.

This study also contributes to the literature on the effect of informal care on female labor supply, such as Van Houtven et al., 2013, Crespo and Mira, 2014 and Schmitz and Westphal (2017). It is essential to note that informal care, which constitutes the substitution effect within our framework, represents just one of the mechanisms through which parental health shocks take effect. We aim to assess the comprehensive influence of parental health shocks across both genders, with the income effect emerging as a pivotal determinant of gender differences. Additionally, this study reveals the potential econometric issues of ignoring the income effect. Research on informal care typically adopts parental health as an IV for care provision (Bolin et al., 2008, Crespo and Mira, 2014), where the presence of income effects may invalidate the exclusion restriction and monotonicity assumption.

Another contribution of this study is revealing the enduring impacts using a quasi-experimental event-study design. Studies on the effects of informal care employing the IV method typically estimate a static effect. To our knowledge, Schmitz and Westphal (2017) are the only exception, which carefully selected control variables based on the conditional independence assumption. Our strategy relies on the parallel trend assumption and we provide credible tests prior to the event. The implied costs of population aging may be far greater and more uneven than those suggested by short-term effects, providing valuable insights on the potential benefits that policies on long-term care insurance could yield.

Finally, this study offers a new perspective to the literature on intrahousehold labor division and the gender gap, which typically attributed specialization to factors such as children, social norms, and marriage quality (Yamaguchi et al., 2014, Bertrand et al., 2015, Juhn and McCue, 2017). We highlight the role of parental health under population aging.

The next section introduces the study’s background. Section 3 describes the data and econometric methods, followed by empirical results and robustness tests in Sections 4 Empirical results, 5 Robustness tests. Section 6 concludes.