Balancing roles and well-being in later life: insights from Chinese older women’s life course Authors: Xinyi Chen, Wei-Jun Jean Yeung Published: 2025 ABSTRACT AbstractBackgroundResearch on productive aging and health has gained increasing attention, yet older women remain understudied—despite often balancing both family responsibilities and economic roles. This issue is particularly salient in China, where women retire unusually early. This research examines the impact of productive aging on the health of Chinese older women, considering opposing hypotheses from the role enhancement/accumulation and role strain perspectives.MethodsUsing longitudinal data from the China Family Panel Studies, this study uses hybrid models to examine how different types and intensities of productive activities influence functional capacities and self-rated health for women aged 55–84. Karlson–Holm–Breen (KHB) mediation analysis further examines the roles of instrumental and psychological pathways.ResultsFindings support the role enhancement/accumulation perspective, showing that multiple and intense productive roles improve health. Women receiving care from children and those with an urban hukou benefit more from productive engagement. Mediation analysis reveals that depressive symptoms and financial support from children partially mediate these effects.ConclusionsThe findings provide nuanced insight into productive aging and health for women in a non-western developing country, where old-age assistance is limited. While Chinese older women demonstrate resilience in balancing multiple roles, enhanced instrumental and emotional support from family may further amplify the health benefits of productive aging.Keywords: Productive aging, Gender, Role theory, Health, Chinese older women Background Productive aging has received increasing attention in East Asia where population aging is currently the most rapid. It benefits not only older adults’ health but also their families, communities, and society by easing labor force shortage and offsetting the fiscal burden of elderly care [1, 2]. However, women often experience fewer health benefits compared to men due to their greater likelihood to shoulder (unpaid) caregiving and domestic duties or the “double burden” of both paid employment and family obligation that may lead to health decline [3, 4]. This disparity worsens when gender intersects with socioeconomic status, rendering some older women overburdened with “productive” activities that may lead to health problems [5, 6]. Given these complexities, it is crucial to focus on older women, as they exhibit diverse and heterogeneous characteristics that influence their engagement in and benefits from productive aging, especially in non-western developing regions with limited old-age assistance. China presents a unique context for understanding the impact of older women’s productive activities on health. Chinese women face mandatory retirement age younger than most countries, in their 50 s depending on their occupations, about 5 to 10 years younger than Chinese men. They engage in extensive paid and unpaid productive activities throughout life, due to a feminized labor market and cultural caregiver expectations [7, 8]. However, many work in the informal sector, experience disrupted careers, and retire early [9, 10]. These conditions further limit access to pensions, leaving Chinese older women with much lower pension benefits than men and confining their productive engagement largely to unpaid labor in a developing economy with limited old age support [11]. How productive Chinese older women are and how these activities affect the health for different subgroups of women have not been well understood. This study addresses the research gap by focusing on Chinese older women to contextualize and examine the relationship between productive aging and health. We examine both the types and intensities of productive activities, adding a new focus on activity intensity to existing literature. We also explore key moderating factors and the mediating pathways relevant to the Chinese context. We start by conceptualizing productive aging and reviewing relevant literature, including role theory and its applicability in the Chinese context, to develop research hypotheses. These will then follow an introduction of the dataset, the sample, and the analytic strategy. After presenting the results, we highlight the contributions and limitations of this study and discuss policy implications. This paper will make both empirical and theoretical contributions to the productive aging literature by examining both types and intensity of the activities, mediators and moderators for the longitudinal productive aging-health link in a non-Western developing society. The relationship between productive aging and health While there is not a universal definition of productive aging [12], this study adopts Morrow-Howell’s widely cited definition: productive aging is a continuous process, along which people remain economically productive through various kinds of activities such as employment, caregiving, housework, volunteering, and informal social assistance. Productivity contingents on an individual’s basic physical, cognitive, and emotional functioning [13]. Longitudinal studies show that good health promotes labor force engagement, while poor health impedes it [14]. Ko & Yeung reveal that health and nutrition in early childhood can have long-term impact on one’s productive engagement in later years [15]. Productive activities have also been found in different social contexts including the U.S., South Korea and Chile to generally benefit older adults’ health compared to not engaging in any activities at all [12, 16–18]. Research shows that employment and volunteering usually lead to better physical functioning, cognitive status, self-rated health, and less frailty, although certain work conditions can negatively impact mental health [19, 20]. The effects of caregiving and family helping tend to be mixed [21]. Caregiving can be beneficial by reducing loneliness [22] but it can also be physically and mentally demanding, financially burdensome, and limit the time and opportunities for leisure activities and self-care of older adults [23, 24]. Choi based on the China Health and Retirement Longitudinal Study shows that rural older adults experienced increased depressive symptoms from grandparenting [25]. Lavela & Ather’s review paper regarding the U.S., UK, Netherlands, and Japan shows that older caregivers experienced cognitive functioning difficulties, stress, depression, and anxiety from caring for spouses [26]. Older women in many societies are found to engage more in caregiving and family helping than in formal employment [27, 28]. Similarly, in China, since the mandatory retirement age is 5 to 10 years earlier for women than for men, domestic responsibilities are often shifted onto women after they retire. Moreover, China has a distinctive urban-rural divide resulted by a household registration system (Hukou system) with most rural residents lacking pension and less adequate access to medical resources [29]. In addition, many young adults moved to cities leaving young children to be cared for by their aging parents at their rural homes. Therefore, rural older women are in an extremely disadvantaged state in terms of accessing the various type of productive activity, which is closely related to health outcomes of productive aging [30, 31]. Health outcomes can be measured in multiple ways, including self-rated health, psychological or mental well-being, physical health, cognitive functioning, morbidity, frailty, disability, mortality, etc [14, 19, 28, 32].,. This study employs two crucial indicators of older adults’ general well-being: functional capacities and self-rated health status [28]. Functional capacities are especially relevant for older adults as they encompass abilities necessary for independence and quality of life [33]. Changes in functional capacities often predict future health events such as hospitalization [34]. Similarly, self-rated health status, a widely used indicator of overall health status, captures older adults’ perception of their own health and predicts mortality [35, 36]. By including both measures, we provide a broader understanding of the relationship between older women’s productive aging and health. Theoretical explanations for productive aging and health The effects of productive aging on health are usually explained by the role theory in existing literature, which categorizes productive activities as taking up productive roles [13]. The directions of effects compose two competing perspectives: role accumulation/enhancement effect explains the positive health effects whereas role strain effect explains the negative health effects of roles [18, 37]. Most existing literature highlights the normative effects of roles on health. According to this view, adhering to role expectations results in positive health effects while deviating from expected roles results in negative health effects [37, 38]. In the context of Chinese older women, taking up caregiver and homemaker roles is associated with achieving good health because these roles align with the traditional expectation [24]. Conversely, older women who fail to shoulder family duties or who engage in formal employment that defies traditional role expectation may experience health decline [39]. In sum, social norms regarding roles navigate the relationship between productive activity types and health. On the other hand, the effect of multiple roles on health receives much less acknowledgment but is also significant for understanding the two competing role effects perspectives [40]. Role enhancement/accumulation perspective suggests that holding multiple roles boosts self-esteem, enhances social status, and thereby promoting health by allowing individuals to buffer stress across roles and transfer resources between them [37]. On the contrary, the role strain perspective posits that multiple roles can also be over-demanding or cause conflicts of time, place, or resources, which can exert negative influences on health [37, 41]. For women, the double burden of both family and formal work roles can cause role conflicts and role overload, which may subject them to deteriorating health [41, 42]. Multiple roles’ effects can be operationalized by examining how the combination of multiple productive activities and accumulation of productive intensity affects health. We propose four distinct pathways to understand the relationship between productive aging and health among Chinese older women, considering both the direction (positive vs. negative) and the mechanisms (role types and number of roles). Examining these four pathways allows for a better understanding of how various types and intensities of productive engagement impact older women’s health. Productive activities have enhancing effects through role types - namely, fulfilling socially expected roles leads to feelings of satisfaction and self-worth, hence positively influencing health. Productive activities have enhancing effects through role types - namely, fulfilling socially expected roles leads to feelings of satisfaction and self-worth, hence positively influencing health. Productive activities have enhancing effects through number of roles - namely, the benefits and resources accumulated by multiple roles to influence health positively. Productive activities have enhancing effects through number of roles - namely, the benefits and resources accumulated by multiple roles to influence health positively. Productive activities have straining effects through role types - namely, deviation from role expectations results in role conflicts, negatively influencing health. Productive activities have straining effects through role types - namely, deviation from role expectations results in role conflicts, negatively influencing health. Productive activities have straining effects through number of roles - namely, the demands of multiple roles may lead to role overload, negatively affecting health. Productive activities have straining effects through number of roles - namely, the demands of multiple roles may lead to role overload, negatively affecting health. Contextualizing productive aging in the Chinese context and research hypotheses Direct effects Contextually, China has a profound familism and a patriarchal gender role ideology rooted in Confucian culture, leading to older women shouldering a lion’s share of the care responsibilities for family members [43]. The current cohort of older women adheres to traditional gender roles and practice the traditional division of labor to be the primary caregiver and homemaker in the family; breadwinning is generally not expected to be shouldered by them [44]. Grandparenting, in particular, has long been a central domestic duty for older women and has become an even more prominent responsibility in recent decades, driven by broader societal changes—such as labor migration and the phenomenon of left-behind children—where economic pressures lead parents to seek work elsewhere, leaving children in the care of their grandparents [45]. We thus hypothesize that the normative effects of roles on older women’ health root in the traditional gender roles. Hypothesis 1a Engaging in family duties benefits older women’s health by enhancing their well-being through the fulfillment of culturally expected roles. Hypothesis 1b Engaging in work negatively influences older women’s health due to the strain caused by the demands of normative role expectations. On the other hand, intense productive roles and the “double burden” of both paid and unpaid roles may lead to significant stress and create incompatibility between role demands and personal capacity [46], ultimately harming older women’s health. Thus, we hypothesize that the role strain perspective applies to intense productive roles and the combination of work and family roles, as they are likely to cause role overload. Hypothesis 2 Intense or multiple forms of productive engagement are linked to worse health outcomes in older women, due to the strain from managing multiple roles simultaneously. Moderating factors The second research question explores the moderators that influence the relationship between productive aging and health. Family support can ease the dual burdens of work and caregiving, while fostering meaningful relationships that integrate older women more deeply into family life. This connection not only strengthens their social ties but also amplifies the health benefits of productive aging, contributing to improved overall well-being [47, 48]. Maintaining a mutual support relationship with children exerts beneficial influences on older adults’ health [49], particularly for older women [50]. Chinese older women often experience a sense of self-worth and satisfaction through reciprocal support with their children, in line with cultural values that emphasize the importance of harmonious family relationships. This exchange not only nurtures family bonds but may also have a positive impact on their health and well-being [28]. We thus have the following hypothesis: Hypothesis 3a Older women who receive care or financial support from children are less likely to experience health decline resulting from more intense or multiple productive engagement s than non-receivers . Similarly, Chinese older women with spouses are more likely to assume traditional roles as housewives, which contributes to their self-identity and better health outcomes compared to those without spouses. Given their shorter durations in the labor market and reliance on their spouses for financial security, having a spouse may also help alleviate work-related stress [14]. Additionally, spouses may share burdens from multiple family or work roles [51]. Therefore, our next hypothesis is as follows: Hypothesis 3b Older women with a spouse are less likely to experience health decline due to intense or multiple forms of productive engagement than those without a spouse. Another moderating factor is the Chinese household registration system (hukou), which assigns individuals to specific statuses linked to access to public assistance based on one’s birthplace. In rural areas, multigenerational living arrangements are more common, often leading older women to assume burdensome caregiving roles, especially when their adult children migrate to urban areas for work. This shift can place significant strain on their health, potentially leading to negative health outcomes [25, 52]. Consequently, the positive effects of housework and caregiving on health are likely to diminish for rural residents. Rural areas also have more deeply ingrained traditional gender role ideologies that discourage women from formal employment [53]. Furthermore, rural hukou holders often engage in lower-pay, labor-intensive jobs or agricultural work, which are associated with worse health [28]. Therefore, we posit the following relationship: Hypothesis 3c Older women with urban hukou are less likely to experience health decline from intense or multiple forms of productive engagement than their rural hukou counterparts. Mediating pathways The “time-for-money” hypothesis suggests that older women taking up the caregiver and homemaker roles can exchange their labor for financial and other support from their adult children [25]. Such instrumental resources often correlate with better access to health resources, leading to enhanced overall health status [54]. Additionally, a stable family financial situation can strengthen older women’s sense of financial security, helping to mitigate the adverse health effects of work-related stress. We thus hypothesize an instrumental mediating pathway between productive aging and health. Hypothesis 4a Instrumental resources mediate the impact of productive engagement on health by amplifying the positive effects of housework and caregiving, while mitigating the negative effects of work. On the other hand, the emotional and social bonds fostered through shouldering family duties could have psychological benefits, which can be transferred to better overall health [55, 56]. Few studies have explored the relationship between work and the mental health of Chinese older women. However, some research suggests that work may negatively impact mental health by limiting opportunities for interpersonal support and increasing emotional distress [57–59], potentially leading to a decline in overall health status. Therefore, we hypothesize the mediating role of psychological resources as follows: Hypothesis 4b Housework and caregiving enhance psychological well-being, which in turn leads to better health, while work diminishes psychological well-being, contributing to worse health outcomes. Figure 1 summarizes the research questions and hypotheses in a conceptual framework. It illustrates how this study perceives whether and how the types and intensities of productive activities are associated with general and functional health among older Chinese women. Research framework Methods Data and sample This study uses data from the nationally representative China Family Panel Studies (CFPS), which collects biannual data from 2010 to 2022. It gathers extensive individual- and household-level information on family life, household dynamics, intergenerational relationships, and health. The survey employs multistage, stratified, probability-proportional-to-size (PPS) sampling and follows a panel design. CFPS’s panel structure and rich family context data make it particularly suitable for this study compared to other national datasets in China. We use three waves of data (2018, 2020, and 2022) as they provide consistent, longitudinal measures needed for this analysis, including work status, grandparenting, and housework participation. Other waves either lacked relevant data measured consistently or contained high proportions of missing values, exceeding 50%. Other productivity measures, such as volunteering or informal helping, are excluded as they are not covered in any wave. We first retain all female participants aged 55–80 in 2018. The lower cutoff of 55 was chosen to ensure all participants had reached the legal retirement age in China—50 for non-professionals and 55 for professional women—by the time of their participation. The upper cutoff of 80 was set considering China’s average life expectancy, which is approximately 77 years. We then include those who participated in any later waves (2020 or 2022), yielding a final sample of 4,821 participants in an imbalanced dataset: 1,709 participated in all three waves, 1,657 in 2018 only, 955 in 2018 and 2020, and 500 in 2018 and 2022. Older individuals, those with lower education, and those without spouses were more likely to drop out after the baseline wave (Appendix Table A.2). Following prior research, we retained participants from the first study wave (2018) to minimize selection bias and improve model estimation by preserving more data [60, 61]. Individual longitudinal weights in 2018 are used to adjust for sampling or non-response biases. Measurement Dependent variables are two measures of health. Functional capacities are measured by asking if participants can independently perform seven daily activities: outdoor activities, eating, doing kitchen activities, using public transportation, shopping, cleaning, and doing laundry. Responses in the original dataset were binary (yes = independent and no = not independent). We sum the scores of all the activities to create a 0–7 scale for functional capacities, with higher scores indicating higher capacities. Self-rated health status is coded from a five-level ordinal variable with 1 = not healthy, 2 = fair, 3 = relatively healthy, 4 = healthy, 5 = very healthy. We coded health status into a binary variable by combining ‘not healthy’ and ‘fair’ as ‘bad health’ (= 0), and classifying the remaining categories as ‘good health’ (= 1). This approach provides a more parsimonious interpretation while maintaining statistical robustness. Additionally, the even distribution between the two groups in our sample ensures the statistical power of the binomial logit regression model [62]. Independent variables are generated from three productive activities: work, housework, and caregiving [13]. Work intensity is the average work hours per week and has three levels: 0 = not working, 1 = fewer than 35 h/week, 2 = more than 35 h/week. Housework intensity is the time spent doing housework every day in three-levels with 0 = zero hours, 1 = fewer than three hours, 2 = more than three hours. Caregiving intensity is measured by the number of grandchild(ren) of whom participants take care as 0 = not taking care of any grandchild, 1 = one grandchild, 2 = 2 or more grandchildren. The intensity of productive engagement is an aggregated variable calculated as the sum of the three activity intensities, ranging from zero to six, with a higher score indicating a higher intensity. There are four moderating variables. Caregiving and financial support from adult children in the past six months are measured as two categorical variables: 1 = not receive, 2 = receive, and 3 = not applicable. Only participants aged 60 and above were required to answer these questions. To prevent listwise deletion of participants aged 55–59, they were coded as “not applicable.” Participants’ hukou status distinguishes rural from urban hukou. Spousal status is binary, coded as 0 = do not have a spouse (including never married, divorced, cohabitation, and widowed), 1 = have a spouse. Two sets of mediating variables are as follows. For the instrumental support pathway, we use whether the older women received children’s financial support and the logged per-capita household income (a continuous variable) to measure the financial security of the woman. For the psychological pathway, we use two variables to measure psychological well-being. Life satisfaction is a self-reported five-point ordinal variable, with higher scores indicating higher satisfaction. A second variable - depressive symptom - is measured based on questions from the CES-D index (Center for Epidemiological Studies Depression) [63] examining whether the participants feel emotionally down, do not sleep well, feel lonely, find it hard to do anything, feel happy, live a happy life, feel sad, and feel life cannot continue. A composite index is generated by summing the scores of each question (inversed coded for feeling happy and living a happy life), a common practice to treat the CES-D scale [64]. We include other covariates, based on previous literature, that are likely to affect older women’s health as control including her age, educational background, and pension insurance status (养老保险). Educational attainment is grouped into three categories to indicate whether the participant is illiterate, finished primary school, or finished junior high school and above. Pension insurance indicates whether the participant is receiving a pension through insurance or not. Analytic strategy We conduct hybrid models to examine the overall influences of productive engagement on older women’s health as this method fits the person-time longitudinal data structure and makes use of the temporal orders of variables [65]. The hybrid model combines the advantages of the fixed-effects model and the random-effects model. It allows fixed-effects estimates for level-one (within-person) variables while allowing for the inclusion of level-two (between-person) covariates, retaining the virtues of the random effects approach [66]. As such, level-one covariates are unbiased by level-two unobserved heterogeneity [67] while variation between individuals is allowed [65, 66]. More importantly, by differentiating within-person and between-person effects, the hybrid model distinguishes the influences caused by the change of an individual’s behavior or status over time and the initial differences between different persons [68]. The within-person effects therefore indicate the causal effects of the independent variable on the corresponding dependent variable. The basic hybrid model [67, 69] is given by: Here denotes the within-person variation and can be understood as ; denotes between-person variation, thus can be understood as . is a level one variable that varies both within and between clusters, is a level two variable that varies only between clusters. is level two error while is level one error. The hybrid model uses all available data points, which is particularly useful for longitudinal studies with imbalanced datasets [70]. We did not adopt multiple imputation to adjust for missing values because it provides limited added value for an imbalanced dataset [71] and the proportions of missing data in our variables are largely lower than the conventionally accepted 5% (see Appendix Table A.1). By retaining all participants from the 2018 wave and anyone attended follow-up waves in 2020 and 2022, such samples help provide more accurate estimation for between-person effects and error terms than deleting drop-out cases. Additionally, the fixed-effects part of the hybrid model enhances the estimation of within-person effects of time-varying predictors, reducing the endogeneities caused by unobserved influences [72]. Next, we examine moderating effects by incorporating interaction terms between productive aging and key variables, including whether received children’s caregiving or financial support, had a spouse, and the older women’s hukou status, into the hybrid model. Each moderator is analyzed separately. Finally, we conduct the Karlson-Holm-Breen (KHB) method to examine the mediators’ effects. The KHB method is designed for decomposing the total effect into direct and indirect effects for nested non-linear probability models such as logistic regression but can also be applied to linear regression [73]. Considering that one of our dependent variables is binary, the KHB method well suits our research purpose. The KHB model has other advantages over the traditional mediating model such as the classical three-step analysis. By comparing two regression models that one includes the mediators (say, M) and another contains a residualized version of M, the KHB method reveals the true impact of mediation/confounding on the coefficients of variables of interest [74] as the estimation is unaffected by the rescaling bias that arises in comparisons across multiple nested non-linear models [75]. It also allows multiple mediators to be examined simultaneously [74]. The KHB method stands as a valid and reliable approach for examining mediating mechanisms in statistical modeling as demonstrated by existing empirical studies [76]. Results Descriptive results Table 1 displays the descriptive statistics of the sample over the three waves. The average productive engagement intensity decreases slightly. Chinese women engage in relatively intense housework and maintain stable intensities of caregiving and work. Functional capacities decrease while self-rated health status improves slightly as they age. Table 1Descriptive statistics of the sample (weighted)mean/% (standard deviation)range201820202022work intensity (%) not working52.26%53.01%57.67% low intensity/part-time work23.91%23.18%24.37% high intensity/full-time work23.83%23.81%17.96%housework intensity (%) not doing housework6.69%5.03%6.93% low intensity/less than three hours per day57.66%58.15%60.09% high intensity/more than three hours per day35.65%36.81%32.99%grandparenting intensity (%) not taking care of any grandchildren59.68%62.50%57.30% low intensity/taking care of one grandchild32.56%30.00%35.49% high intensity/taking care of two or more grandchildren7.76%7.49%7.21%overall engagement intensity (mean)2.49 (1.26)2.47 (1.24)2.38 (1.23)0–6self-rated health status (binary, mean)0.50 (0.50)0.54 (0.50)0.56 (0.50)functional capacities (mean)6.57 (1.11)6.48 (1.18)6.46 (1.13)0–7age (mean)64.45 (6.68)65.62 (6.34)67.24 (6.13)55–84urban hukou (%)30%32%35%education level (%) illiterate51.77%45.60%42.43% primary school17.98%19.44%20.63% junior high school and above30.25%34.96%36.94%spousal status (have spouse, %)77%77%74%receiving pension insurance (%)56%59%67%Receive financial support from children (ref: no) Yes44.28%47.07%54.92% Not applicable28.33%25.45%13.71%Receive caregiving from children (ref: no) Yes32.48%34.06%39.9% Not applicable28.33%25.42%13.77%depressive symptoms (mean)14.45 (4.50)14.08 (4.45)14.33 (4.45)4–32life satisfaction (mean)4.21 (0.95)4.24 (0.89)4.19 (0.90)1–5Per capita household income (logged, mean)10.54 (1.06)10.62 (1.05)10.73 (1.06)0.31–16.15Receive financial support/caregiving from children in the past six months are applicable for only participants aged 60 and above Descriptive statistics of the sample (weighted) Receive financial support/caregiving from children in the past six months are applicable for only participants aged 60 and above The hukou status and educational background of participants remain stable over time. Slightly above 30% of the sample have urban hukou. Around 50% of the participants are illiterate or almost illiterate. This cohort of older adults was born before 1963, and they had limited access to education. As expected, the percentage of older women having a spouse decreases while the share of older women receiving pension insurance increases. Participants’ wealth grows over the years. Older women’s mental health and life satisfaction remain stable. As Fig. 2 shows, in general, most participants engage in housework in combination with other activities. Very few engage in only paid work or caregiving or the combination of these two activities. The share of women engaging solely in housework or combining housework and caregiving has increased over time, while the proportion combining work and housework has decreased, dropping from the largest category (30.0%) to the second largest (24.1%). This trend reflects traditional gender norms and the retirement system that encourage women to leave the labor market and take on more domestic responsibilities. Notably, around 20% have consistently participated in all three types of productive activities. Also, very few older women (4%) do not engage in any productive activities, indicating that older women in the sample are very productive. Despite the stipulated retirement age of 50 for workers and 55 for professionals, our sample shows high work participation rates due to the inclusion of rural older women who continue agricultural work and do not “retire” from it. Combination of types of productive activities (percentage) Multivariate analysis results Direct effects Table 2 presents the within-person estimates for productive engagement types and intensity on functional capacities and self-rated health status. Between-person estimates are shown in Appendix Figure A.1 as we primarily rely on within-person estimates for inference. Table 2Hybrid model of productive engagement types and intensity on health outcomesFunctional capacitiesSelf-rated health status (OR)Within_low work intensity0.050.96Within_high work intensity0.031.20Within_low housework intensity0.80***1.33Within_high housework intensity0.89***1.27Within_low grandparenting intensity0.051.22Within_high grandparenting intensity0.17**0.91Within_overall productive intensity0.10***1.06Within_age−0.07***−0.05***−0.07***−0.06***1.051.051.041.05*Within_depressive symptoms−0.03***−0.03***−0.03***−0.03***0.91***0.91***0.91***0.91***Within_spousal status0.02−0.030.01−0.010.53*0.52*0.51*0.52*Urban hukou0.16***0.070.030.17***1.030.830.800.96Education level (base: illiteracy)primary school0.20***0.19***0.22***0.23***1.50**1.41**1.49**1.55***junior high school and above0.30***0.24***0.28***0.31***1.90***1.65***1.76***1.88***Per capita household income0.05**0.04*0.04*0.04*1.091.061.071.07Pension status0.060.030.040.021.34**1.30**1.29**1.29**constants6.35***4.25***6.63***6.02***N71027314735270668091832383708046OR: Exponentiated coefficients. Random effects estimation is conducted on covariates (hukou, education, per capita household income, pension) due to little changes across different waves* p < 0.05, ** p < 0.01, *** p < 0.001 Hybrid model of productive engagement types and intensity on health outcomes OR: Exponentiated coefficients. Random effects estimation is conducted on covariates (hukou, education, per capita household income, pension) due to little changes across different waves * p < 0.05, ** p < 0.01, *** p < 0.001 The within-individual effects show that older women who switch from non-engagement to lower-intensity or higher-intensity housework and high-intensity grandparenting will gain significantly better functional capacities. Engaging in low-intensity housework brings improvements in functional capacities by 0.8 units and high-intensity housework 0.89 units. High-intensity grandparenting (coefficient: 0.17) and greater combined intensities of productive activities (coefficient: 0.1) improve functional capacities too. However, no single type of activity nor combined intensities of productive activities significantly benefit their self-rated health. As a woman become older, her functional status declines but self-rated health improves. Having poorer psychological wellbeing leads to worse functional capacities and self-rated health. Losing a spouse does not seem to affect an older woman’s functioning capacities but improves their self-rated health status significantly. Urban hukou holders report better functional capacities. A higher educational level is related to better health. Those who have higher per capita household income enjoy significantly better functional capacities. Receiving a pension through insurance is related to better self-rated health. Moderating effects Table 3 provides the result of the interactional effects of productive engagement intensity and the moderators. Table 3Hybrid model with interactional effects of child support/spousal status/Hukou and overall productive engagement intensity on healthFunctional capacitiesSelf-rated health status (OR)Within_overall productive intensity0.08**0.070.18**0.07**1.021.160.991.09Children’s care (ref: not receive)Receive children’s care, age 60+−0.86***0.39***Age 55–59, not applicable−0.050.57Within_productive intensity*children’s care (ref: not receive)Intensity * Receive children’s care, age 60+0.061.34*Intensity * Age 55–59, not applicable−0.020.82Children’s financial supportReceive children’s financial support, age 60+−0.24*0.82Age 55–59, not applicable0.180.77Within_productive intensity*children’s financial supportIntensity* Receive children’s financial support, age 60+0.061.03Intensity* Age 55–59, not applicable−0.010.73Within_productive intensity*have a spouse−0.101.09Within_productive intensity*urban hukou0.15**0.90ControlsYesYesYesYesYesYesYesYesConstants6.39***6.11***5.83***6.02***N70667064706670668046804480468046OR: Exponentiated coefficients* p < 0.05, ** p < 0.01, *** p < 0.001 Hybrid model with interactional effects of child support/spousal status/Hukou and overall productive engagement intensity on health Within_productive intensity* children’s care (ref: not receive) Within_productive intensity* children’s financial support Within_productive intensity* have a spouse Within_productive intensity* urban hukou OR: Exponentiated coefficients * p < 0.05, ** p < 0.01, *** p < 0.001 Table 3 shows that Chinese older women receiving care from children are more likely (OR: 1.34) to improve their self-rated health through productive activities compared to non-receivers. Similarly, urban older women exhibit greater improvements in functional capacities (coefficient: 0.15) from productive engagement than their rural counterparts. However, children’s financial support and spousal status do not moderate the relationship between productive aging and health. Mediating effects The KHB mediation analysis shows the estimated total effect, direct effect, and indirect effects of the type of productive activity. Statistically significant results are presented in Table 4. Table 4 reports the decomposition of all the mediators’ contributions. Sizes of total effects, direct effects, and indirect (mediating) effects of independent variables on dependent variables are shown in the same row of each activity type. Estimation of each mediator’s mediating effect is listed in the row of Indirect effects; the percentages of each mediator’ contribution to the indirect effects are shown in the row of P_indirect effects(%). A higher contribution percentage reflects a stronger mediating effect. Table 4Contribution of each mediator to the indirect effects on functional capacities and self-rated health status by the KHB methodFunctional capacitySelf-rated health (binary; OR)Total effectsDirect effectsIndirect effectsP_indirect effects (%)Total effectsDirect effectsIndirect effectsP_ indirect effects (%)High-intensity work1.79***1.94***0.92*Depressive symptoms0.9386.2Life satisfaction0.995.35Per capita household income (logged)0.996.38Children’s financial support0.992.08Low-intensity housework1.86***1.80***0.06***3.61***2.90***1.24***Depressive symptoms0.0583.171.2292.94Life satisfaction−0.001−2.381.016.14Per capita household income (logged)0.0048.411.015.01children’s financial support0.00610.800.99−4.09High-intensity housework1.92***1.87***0.05**3.39***2.95***1.15*Depressive symptoms0.00873.551.15100.40Life satisfaction0.0010.271.000.07Per capita household income (logged)0.0039.031.017.33Children’s financial support0.00317.150.99−7.8Low-intensity grandparenting0.14***0.11***0.03*Depressive symptoms0.00119.19Life satisfaction−0.00−0.52Per capita household income (logged)0.00513.47Children’s financial support0.0267.86High-intensity grandparenting0.24***0.20***0.04**Depressive symptoms0.0249.24Life satisfaction−0.00−0.83Per capita household income (logged)−0.001−4.28Children’s financial support0.0255.86Overall engagement1.24***1.27***0.98*Depressive symptoms0.9914.02Life satisfaction0.998.29Per capita household income (logged)1.00−1.26Children’s financial support0.9878.95Only significant indirect (mediating) effects are shown; OR: exponentiated coefficients; P_indirect effects: contribution of each mediator to the indirect effects in percentage* p < 0.05, ** p < 0.01, *** p < 0.001 Contribution of each mediator to the indirect effects on functional capacities and self-rated health status by the KHB method Only significant indirect (mediating) effects are shown; OR: exponentiated coefficients; P_indirect effects: contribution of each mediator to the indirect effects in percentage * p < 0.05, ** p < 0.01, *** p < 0.001 The results indicate that high-intensity work has a positive direct effect on self-rated health (OR: 1.8), but mediators collectively exert negative indirect effects (OR: 0.9). Depressive symptoms are the primary mediator (OR: 0.9), accounting for 86% of the total indirect effects, suggesting that high-intensity work increases depressive symptoms, diminishing its health benefits for Chinese older women. The health outcomes of both low- and high-intensity housework are significantly mediated, with depressive symptoms contributing between 70% and 100% of these effects. Housework alleviates depressive symptoms, improving functional capacities (coefficients: 0.05 and 0.03) and self-rated health (OR: 1.2 for both intensities). Additionally, housework increases the likelihood of receiving financial support from children, which enhances functional capacities (coefficients: 0.006 and 0.008) but slightly decreases self-rated health (OR: 0.99). Children’s financial support serves as the key mediator linking low- and high-intensity grandparenting to functional capacities, accounting for 68% and 56% of the mediating effects, respectively. Depressive symptoms play a crucial role as a mediator for high-intensity grandparenting, contributing 49% of the effects. Grandparenting helps alleviate depressive symptoms (coefficients: 0.006 and 0.02) and increases children’s financial support (coefficients: 0.02 for both intensities), ultimately improving functional capacities. However, combined intensities of productive engagement benefit self-rated health less than expected due to the suppressive effects from children’s financial support (79%) and depressive symptoms (14%). Higher levels of productive engagement reduce financial support from children (OR: 0.98) and increase depressive symptoms (OR: 0.99), offsetting its positive health impacts. Discussion The cohort of older Chinese women in this study shares a unique historical background as they were born in Mao’s regime and immersed in the socialist gender movement, leading them to shoulder the double burden of both paid work and family duties and having to retire early [77]. The data used for this study supports such claims as half of the sample engage only in family-related productive activities while over 40% bear the double burden. This study examines with longitudinal data whether this double burden for older Chinese women will result in health risks and unpack the various ways in which moderating and mediating factors are involved in this relationship. This research adopts a longitudinal design by tracking the same group of older women throughout four waves of a study and conducting the hybrid models analyses that separate within-person and between-person effects of productive engagement on health. Model results show that Chinese older women’s health benefits from family responsibilities (housework and grandparenting) but not from work, including part-time, full-time, and agricultural work. Their health also benefits from multiple productive roles, and more intense productive engagement, supporting the role enhancement/accumulation perspective. These findings support Hypothesis 1a, which posits the positive influence of family duties on health, but contradict Hypothesis 2, which suggests that multiple or more intense productive engagements negatively impact health. We adopt roles’ normativity and multiplicity concepts to explain such findings that counteract the double-burden concern. On the one hand, this cohort of Chinese older women are expected to shoulder primarily homemaking responsibilities because they are immersed in the traditional gender norms [78]. Such an expectation lays a foundation for family-related productive roles to exert normative effects on health positively. Moreover, older Chinese women in our study appear capable of managing multiple roles concurrently. We attribute this ability to the relatively younger age of our sample group, with a mean age of 64 during the first wave and the early mandatory retirement age for women. At this age, they remain functionally able to handle the demands of intensive family provider or caregiver roles. For these women, various productive roles might serve as opportunities for physical activities, help in preventing functional decline, and foster engagement in society. Such roles can potentially impart a sense of fulfilment and self-efficacy, rather than pose as burdensome or demanding during the aging process [37, 79]. Several potential moderators and mediators are examined in this study to understand Chinese older women’s productive aging contextually. Results show that those who receive intergenerational transfers of caregiving from their children benefit more from productive aging, partially supporting hypothesis 3a. Such results highlight the significance of family networks and family support for older women’s productive aging, consistent with existing studies that emphasizes family’s essential role in Chinese older women’s later lives [24, 25]. In other words, Chinese older women who experience a reciprocal relationship with their children are more likely to gain good health, aligning well with existing literature that highlights the significance of intergenerational transfer in sustaining Chinese older adults’ later life health [80]. Additionally, having an urban hukou work as a positive moderator, partially supporting hypothesis 3c. Rural older women are often left behind to take on grandparenting or agricultural work without direct assistance from their adult children, who migrate to cities [52]. This lack of assistance may diminish the health benefits they derive from productive aging. These findings suggest that supportive environments at both family and community levels contribute to the beneficial effect of productive aging. Moreover, we examine two mediating pathways: the psychological and the instrumental. The KHB mediation analysis shows that the psychological pathway consistently exerts significant mediating effects. Such a finding is consistent with previous studies that attest to a psychological pathway between productive aging and health [55]. More specifically, engaging in housework and caregiving reduces depressive symptoms, leading to enhanced functional capacities and improved self-rated health. In contrast, low-intensity work exacerbates depressive symptoms, diminishing older women’s potential to reap the benefit of enhanced self-rated health from working. While existing literature indicates that post-retirement employment negatively affects the psychological well-being of Chinese older women [59], our findings reveal how these impacts translate into poorer self-rated health. Additionally, our results regarding housework align with studies that highlight the mediating role of psychological well-being between housework and other health metrics, such as the number of survival days [81]. Our investigation of housework is also among the few studies that examine the relationship between household chores—which constitute a significant portion of the productivity of Chinese older women—and health. The instrumental pathway, particularly financial support from children, serves as the primary mediator between grandparenting and functional capacities. Studies on intergenerational support suggest that older Chinese adults provide childcare for their grandchildren in exchange for financial support from their children, a concept known as the “time-for-money” hypothesis [52]. Our findings confirm that this exchange not only facilitates intergenerational reciprocity but also contributes to improved health outcomes for older adults. That said, the findings provide partial support for both Hypotheses 4a and 4b. Based on the findings, this study suggests that much can be done to create a more conducive environment for Chinese older women’s productive engagement. Government policies should create more opportunities for older women to engage in diverse productive activities apart from family responsibilities, such as paid work, as such engagement is beneficial to their health. Currently, Chinese older women’s involvement in the formal labor market is limited as the legal retirement age for females is in the 50 s, hampering their chances of gaining financial security and health benefits from formal work. Sieber posited that individuals can manage multiple roles, letting each role buffer the stress brought by other roles, and transfer resources between different roles. Our findings affirm that Chinese older women possess this capability [37]. Policy considerations might advocate for a later retirement age since Chinese older women are adept at managing multiple roles both in the workplace and the family. This is particularly relevant as their caregiving burden decreases due to declining fertility rates. Additionally, older women’s productive roles can help ease financial pressures on families and reduce strain on the pension system. In the meantime, more public care facilities should be provided to release older women from onerous domestic chores and care responsibilities and let them manage different roles and life goals more flexibly. Such attempts to promote older women’s economic productivity, while may not specifically tailored to the current cohort of older women, align with broader shifts in Chinese women’s increasing economic participation amid modernization—and hold significant implications for future generations [82]. Also, incentives should be designed to foster stronger interactions between older women and their family members, particularly adult children. Our findings underscore the significant role that family caregiving plays in promoting productive aging, as evidenced by its positive impact on older women’s self-rated health. Strengthening these familial bonds through targeted support can significantly enhance their overall well-being and health outcomes. Policies providing funding for community programs that encourage family engagement can strengthen these crucial support networks. Our findings diverge in part from foundational literature rooted primarily in Western contexts, despite the shared observation that work tends to positively influence well-being [19]. In the case of Chinese older women, all types of productive activities examined show positive associations with well-being—a pattern that may reflect their unique resilience at the intersection of traditional gender norms and a halted gender equality movement [27, 83]. These women have historically been modeled into “superwomen,” taking on a wide range of roles in both the household and workforce. This underscores the importance of situating older women’s productive aging within specific cultural, societal, and historical contexts. Nonetheless, the limited number of studies focused specifically on older women’s productivity, combined with inconsistent measures of well-being, continues to hinder meaningful cross-contextual comparisons. Future research should examine how distinct cultural and historical trajectories shape women’s productive roles and their implications for later-life well-being specifically, which might differ from men’s trajectories significantly. This study also represents one of the first attempts to consolidate two key mechanisms through which roles influence well-being in older age: the normative effects and the cumulative effects of multiple roles. These mechanisms are further examined with the dual lenses of the role strain perspective and the role enhancement/accumulation perspective. By integrating these theoretical strands into a coherent role theory framework, this study offers a novel explanation for the varying effects of role types and intensities on well-being. Such critical reflection and application of role theory are rare in the current literature, and thus constitute a key theoretical contribution of this study to the fields of productive aging and role theory research. Such a framework would benefit from further research efforts incorporating more nuanced, empirically validated measures of role intensity, perceived role pressure, and the subjective meaning of roles. This study has several limitations. Due to data availability, we do not use measures of productivity activities such as volunteering or education. Yet, given the relatively low rates of volunteer engagement among Chinese older adults [32], this omission does not substantially affect the overall contribution of the manuscript. Data availability also prevents us from exploring the effects of different occupations on productive aging though women’s education is closely related to occupation. As around half of the sample are illiterate and 20% have primary school level education, it is plausible to assume that many of them work in lower pay sectors. With better occupation indicators, future studies may be able to investigate various occupations of older women, which could be illuminating considering female labor force is becoming increasingly important in China. Additionally, the longitudinal design and within-person estimation cannot entirely rule out the possibility of reverse causality, a common methodological challenge faced by many researchers. Despite our effort to approach causal inferences through careful statistical analysis, we acknowledge the limitation that healthier older women may be more likely to remain productive in later life. Last but not least, we opted for a simpler and more transparent operationalization—summing the intensity levels of different productive activities—to facilitate straightforward interpretation and comparison across activity types. Future exploration may consider adopting more nuanced approaches such as the latent class/profile analysis to realistically reflect heterogeneous role configurations. Conclusions Based on longitudinal data analysis, this study confirms that productive aging exerts enhancing effects on Chinese older women’s health. This study contributes theoretically to productive aging research by differentiating the effects of multiple roles (number of roles) and roles’ normative effects (role types) through examining the intensity and types of productive activities. Empirically, we provide new evidence to the field by examining potential moderators and the potential mediating pathways in a developing country. This study shows that cultural and policy contexts matter as gender differences in mandatory retirement age and hukou differences in labor market engagement in China are still significant. As such disparities began to narrow in recent years [84], we expect to see smaller differences among future cohorts of older adults. Policy efforts could support later retirement and promote integrative measures that strengthen intergenerational bonds for older women. Acknowledgements This paper uses restricted data from the China Family Panel Studies that cannot be made publicly available. The author can guide interested researchers on how to obtain data used in this paper. We thank the editors and reviewers for their valuable comments and feedback on earlier versions of the manuscript. Abbreviations China Family Panel Studies Karlson–Holm–Breen method Appendix Table 5Missing counts and percentages by interview year Variable 2018 2020 2022 count % count % count % Work intensity 1885.52341.77120.75 Housework intensity 000000 Grandparenting intensity 50.14753.96724.47 Overall engagement intensity 1935.661005.29825.06 Self-rated health status 441.29794.14845.18 Functional capacities 1705170.9311.91 Age 000000 Hukou status 371.08281.49613.77 Education level 000000 Spousal status 0020.1200 Receiving pension insurance 371.1552.87684.21 Depressive symptoms 1885.52341.77120.75 Life satisfaction 000000 Receiving caregiving from children 50.14753.96724.47 Receiving financial support from children 1935.661005.29825.06 Per capita household income 441.29794.14845.18All shown variables are missing at random according to missing data analysis based on logit regression method Missing counts and percentages by interview year All shown variables are missing at random according to missing data analysis based on logit regression method Table 6Sample attrition status (weighted)Attended follow-up wavesOnly baseline waveAge (mean)63.765.8Education Level Illiteracy50.0%62.6% Primary School19.5%15.8% Junior high school and above30.5%21.6%Have a spouse83.6%75.1%Demographic statistics are from the 2018 wave Sample attrition status (weighted) Demographic statistics are from the 2018 wave Between-person estimates of productive engagement types and intensities. Between-person effects indicate different productive engagement types and intensities’ impact on health in comparison to non-engagement. For instance, Chinese older women engaging in low-intensity work generally have better functional capacities by 4.2 units than those who do not have such engagement. A between-person one unit increase in overall engagement intensity is associated with better functional capacities by 2.7 units Authors’ contributions Xinyi Chen and Wei-jun Jean Yeung conceived and designed the study, Xinyi Chen conducted data analysis and drafted the manuscript, Wei-jun Jean Yeung substantively revised it. Both authors have read and approved the manuscript. Funding No funding was received for conducting this study. Data availability The data that support the findings of this study are available from China Family Panel Studies (CFPS) but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data is however available from the authors upon reasonable request and with permission from CFPS. CFPS data are available on the SHARE website upon request. The datasets were derived from sources in the public domain: China Family Panel Studies, https://www.isss.pku.edu.cn/cfps/en/. Declarations Ethics approval and consent to participate The data are from China Family Panel Studies (CFPS), funded by 985 Program of Peking University and carried out by the Institute of Social Science Survey of Peking University. The original study protocol was approved by the Peking University Institutional Review Board (Approval No: IRB00001052-14010). Prior to the original data collection, written informed consent was obtained from all participants. Data was collected in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki. All methods were carried out in accordance with relevant guidelines and regulations or Declaration of Helsinki. For this secondary analysis, we used fully anonymized data, and therefore, our study was considered exempt from further ethical review by our institutional ethics committee. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Xinyi Chen, Email: xinyic@u.nus.edu. Wei-Jun Jean Yeung, Email: Jean_Yeung@a-star.edu.sg. References Associated Data Data Availability Statement The data that support the findings of this study are available from China Family Panel Studies (CFPS) but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data is however available from the authors upon reasonable request and with permission from CFPS. CFPS data are available on the SHARE website upon request. The datasets were derived from sources in the public domain: China Family Panel Studies, https://www.isss.pku.edu.cn/cfps/en/.