Skip to main content
SpringerLink
Log in

Retirement Spillover Effects on Spousal Health in Urban China

  • Original Paper
  • Published:
Journal of Family and Economic Issues Aims and scope Submit manuscript

Abstract

This paper empirically studies the causal effect of retirement on spouses’ subjective health for the elderly in urban China. We find that women’s retirement positively affects their husbands, while husbands’ retirement tends to affect wives negatively. The difference in post-retirement healthy (and unhealthy) behaviors and emotions between men and women can explain gender asymmetry. Men tend to have a negative state of mind and unhealthy habits and behaviors more than women, which results in the negative spillover effect. We also estimate the marginal threshold treatment effect (MTTE), showing that a small delay of statutory retirement age is beneficial for improving overall subjective health, yet the conclusion would actually be the opposite if the spillover effect were to be ignored. These results provide useful references for the current discussion on retirement policy reform in China.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3

Similar content being viewed by others

Data Availability

The data that support the fndings of this study is available from the National Bureau of Statistics, but restrictions apply to the availability of these data. So the population census data used in this manuscript is not publicly available, while the scripts and tables generated by the authors are available upon reasonable request.

Notes

  1. World Bank Open Data: https://data.worldbank.org/indicator/SP.POP.DPND.OL?end=2019&locations=1W-CN-XP&start=1981&view=chart

  2. United Nations, Department of Economic and Social Affairs, Population Division (2019). World Population Prospects 2019, Online Edition. https://population.un.org/wpp/Download/Standard/Population

  3. The empirical research on the own effect of China’s retirement indeed consists of mixed findings. For example, Lei & Liu (2018) estimated the impacts of retirement and found that retirement has a significantly positive effect on the cognitive competence of the male but an adverse effect on the female. Che & Li (2018) concluded a positive impact of retirement on health and behavior, but Lei et al. (2014a), and Feng et al. (2020) found a negative impact of retirement on the health of the male.

  4. If the treatment effect is homogeneous near the cutoff, then this also represents an estimate of the average treatment effect at the cutoff.

  5. An incomplete list includes papers which focus on the impacts of retirement on personal health (Behncke, 2012; Che & Li, 2018; Coe & Zamarro, 2011; Eibich, 2015; Feng et al., 2020; Fitzpatrick & Moore, 2018; Hagen, 2018; Insler, 2014), on psychological health (Heller-Sahlgren, 2017; Kolodziej & García-Gómez, 2019), on cognitive competence Mazzonna & Peracchi (2012); Lei & Liu (2018), and family income or consumption (Battistin et al., 2009; Dong & Yang, 2017; Li et al., 2015), and on the utilization of health equipment (Zhang et al., 2018). These researches conclude positive, negative, or insignificant effects on various health-related outcomes. See also the survey paper of Cawley & Ruhm (2011).

  6. MS2018 briefly mentioned the effect by gender, but no formal results were presented in the paper. Please see Müller & Shaikh (2018, Sect. 5.4).

  7. Our results are therefore specific to males and females, respectively. As we demonstrate in Table 1, in our sample for studying the retirement effect of husbands on wives, both men and women tend to belong to an older cohort and have lower education than those in the sample for studying the wife’s retirement effect. This can partly contribute to the asymmetry of the husband’s and wives’ retirement spillover effect. Unfortunately, with only one-year cross-sectional data, it is difficult to identify the magnitude of it. Examining the dynamics of the spillover effect would be an exciting research topic when richer data is available.

  8. There are other researches which study the effect of retirement on varies of spousal outcomes (e.g. Szinovacz & Davey (2004), and Smith & Moen (2004)), where ordinary least square (OLS) estimation is used to estimate the spillover effect. As we discussed earlier, the OLS estimator is likely to be biased. Table A.3 shows that OLS estimators produce qualitatively and quantitatively different results from the FRD estimates in our case.

  9. Another difference is that we use the birth quarter to construct the running variable, whereas Lei et al. (2011) use birth year.

  10. More specifically, "1" means that the respondent is in good health and is qualified for daily work and can take care of himself or herself, "2" means that the respondent is in average health but can take care of his/her daily life, "3" means that the respondent is in poor health and cannot take care of his or her daily life, and "4" means that it is difficult for the respondent to tell which of the above three choices is most suitable. Therefore, the question measures self-reported or self-assessed health status.

  11. Although without additional information, it is not possible to exactly identify who are compliers in the LATE framework, some studies suggest that retirement compliers are more likely to have higher education and lower health condition than the averages of their peers (see discussions in Sect. 4.1 of Müller & Shaikh,2018). We conjecture it is also the case for our data: higher-educated individuals tend to be more obedient, and people with relatively weaker health conditions (but not bad enough for early retirement) have less incentive to work beyond their retirement age.

  12. Here we implicitly assume \(\delta (r,c)\) is well defined for any value of \(r\) and \(c\).

  13. In general, the distributional local continuity assumption is stronger than what one needs to identify the mean effect, but given \(\gamma\) is binary, it is not over-restrictive.

  14. For incorporating the covariates, please see Calonico et al. (2017) for details.

  15. Please see Arai et al. (2022) for the relationship between the testable implication (6)-(7) and the running variable density test. Indeed, they recommend empirical researchers to implement both tests.

  16. In the notation of Arai et al. (2022), we set ξ = 0.01 and the number of bootstrap sample 1000. We tried other tuning parameter values, and the results are consistent.

  17. The large p-value can possibly be due to the small sample size since most of their husbands have not yet retired when women retire.

  18. We also match husbands and wives in for the CHARLS data. The matched data set provides too few observations near the cutoff.

  19. Because the sample size of women who smoke is too small, we do not report results on how wives’ retirement affects their smoking quantity (number of cigarettes).

  20. One way of analyzing the mediation effect is to consider an approach in the spirit of Baron & Kenny (1986), where a system of equations is built with the mediator (e.g. drinking) appears as the dependent variable of one regression and appears as an independent variable in another long regression together with the explanatory variable of interest (retirement). The existence of mediation effects boils down to testing the product of specific regression coefficients being zero or not. However, as Imai et al. (2011) points out, even in linear models, such an approach requires additional assumptions, which can be hard to justify with either experimental data or observational data. Therefore we do not pursue this direction. To the best of our knowledge, there is no coherent statistical framework for point identification of the mediation effect or testing its existence in the regression discontinuity setup yet.

  21. Recall that husbands are on average two years older than wives, but their retirement age cutoff is 60 whereas wives’ is 50.

References

Download references

Acknowledgements

Shenglong Liu thanks the support of the National Natural Science Foundation of China (#72073078; #71991475) and the Tsinghua University Initiative Scientific Research Program#2021THZWYY03. Yuanyuan Wan thanks the support of the SSHRC Insight Grant #435190500. We also thank the editor of the journal and two anonymous referees for their valuable and constructive comments.

Funding

National Natural Science Foundation of China, 71991475, SHENGLONG LIU, 72073078, SHENGLONG LIU, Tsinghua University Initiative Scientific Research Program, 2021THZWYY03, Shenglong Liu, SSHRC Insight Grant, 435190500, Yuanyuan Wan.

Author information

Authors and Affiliations

  1. School of Public Policy and Management, Tsinghua University, Beijing, China

    Shenglong Liu & Xiaoming Zhang

  2. Department of Economics, University of Toronto, Toronto, Canada

    Yuanyuan Wan

Authors
  1. Shenglong Liu
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Yuanyuan Wan
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Xiaoming Zhang
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

The authors all contributed to the writing and editing of the manuscript. The data collection and data analysis were completed by all three authors.

Corresponding author

Correspondence to Xiaoming Zhang.

Ethics declarations

Conflict of Interest

Author Shenglong Liu declares that he/she has no conflict of interest. Author Yuanyuan Wan declares that he/she has no conflict of interest. Author Xiaoming Zhang declares that he/she has no conflict of interest.

Ethical Approval

As a secondary analysis of the data provided by 1% population survey in 2005 and China Health and Retirement Longitudinal Study, ethical approval was not required for this study.

Informed Consent

Informed consent was obtained from all individual participants involved in the survey.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendices

Additional Empirical Results

We collect additional tables and empirical results in this section.

Table A.1 (See Table 21)

Table 21 Effect of Retirement on Spousal Health: Female Workers
Full size table

Table A.2 (See Table 22)

Table 22 Effect of Retirement on Spousal Health: Normalized Results
Full size table

Table A.3 (See Table 23)

Table 23 Effect of Retirement on Spousal Health: OLS Estimation
Full size table

Table A.4 (See Table 24)

Table 24 Effect of Retirement on Spousal Health: Pooled Estimation
Full size table

Table A.5 (See Table 25)

Table 25 Effect of Retirement on Spousal Health: with more controls
Full size table

Table A.6 (See Table 26)

Table 26 Effect of Retirement on Spousal Health: with city-specific clusters
Full size table

Table A.7 (See Table 27)

Table 27 Placebo test for informal employment
Full size table

Table A.8 (See Table 28)

Table 28 Summary Statistics: CHARLS
Full size table

Table A.9 (See Table 29)

Table 29 First Stage Estimation of the CHARLS Data
Full size table

Table A.10 (See Table 30)

Table 30 Effect of Retirement on Own Smoking Behaviors
Full size table

Table A.11 (See Table 31)

Table 31 Effect of Retirement on Own Drinking Behaviors
Full size table

Table A.12 (See Table 32)

Table 32 Effect of Retirement on Own Social and Sports
Full size table

Table A.13 (See Table 33)

Table 33 Effect of Retirement on Own Emotions
Full size table

Figure A.1 (See Fig. 4)

Fig. 4
figure 4

Retirement Rate by Age. Note: In the mini-census, we do not directly observe at which year an individual retires, but only the retirement status at the survey year. This figure plots the differences in retirement status ratio between two consecutive age groups for males and females, respectively

Full size image

Details on Data Processing

The mini-census data used in this paper contains a total of 2,585,481 observations. In this section, we summarize the process of obtaining our final data for empirical analysis. It includes the following four steps:

  1. (1)

    First, we keep the sample with individuals who are either "household head" or "spouse of the household head", which contains 1,346,434 observations in total. This is the sample we use to match husbands and wives in later steps.

  2. (2)

    Second, we divide the above samples by gender, resulting in a total of 676,044 observations in the wife sample and 670,390 observations in the husband sample. On this basis, the husband’s characteristics regarding employment status, year of birth, and health level are matched with the wife’s sample by household information to obtain the household sample, which has a total of 437,611 observations.

  3. (3)

    In the third step, we construct a further subsample by (1) retaining males aged 50-70 years and females aged 40-60 years, (2) retaining observations with urban household registration and (3) excluding individuals who are studying in school and incapacitated. After this step, we have 49,002 male and 64,822 female observations.

  4. (4)

    Finally, the optimal bandwidths are estimated separately for the male and female samples, and the optimal bandwidths are found to be 11 quarters for both male and female samples. For the purpose of the subsequent robustness check, we limit the sample for the parameter and non-parametric estimation to observations who are within 20 quarters before and after their retirement, resulting in a total of 21,116 observations for males and 31,858 observations for females.

After the processing, our final sample retains only the urban household population so that both the rural population and the migrant worker group are excluded, thus also essentially excluding the urban informal employment, which is mostly taken by the migrant worker group.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Liu, S., Wan, Y. & Zhang, X. Retirement Spillover Effects on Spousal Health in Urban China. J Fam Econ Iss (2023). https://doi.org/10.1007/s10834-023-09935-7

Download citation

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1007/s10834-023-09935-7

Keywords

JEL Classification

Search

Navigation