Spillover effect of children’s education on parental physical functioning over the life course

Abstract

Parental well-being is linked to the life chances of adult children in later life. Despite accumulated knowledge on the role of children’s education on parental longevity in developed contexts, it remains unknown how children’s education may influence the trajectories of parental physical well-being over the aging process, particularly in developing contexts. Using a growth curve model and four-wave data from the China Health and Retirement Longitudinal Study, this study examines the association between children’s education and parental physical functioning trajectories as parents age. This study yields several findings. First, adult children’s schooling is negatively associated with the limited physical functioning of older parents. Second, consistent with the cumulative disadvantage perspective, this study confirms the diverging parental health trajectories across different children’s education groups as parents age. The linear rate of decline in parental physical functioning is slower among older adults with better-educated children. Third, the education returns of sons and daughters with regard to their parents’ physical functioning are similar to each other, implying the rationale for gender-blind attitudes in parenthood. Fourth, children’s education has a compensatory effect among parents lacking institutional old-age support. The association between children’s education and parental physical functioning is significant among rural older adults only.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-85570-6.

Introduction

The trajectory of physical functioning during the aging process is an influential predictor of long-term care needs^1,2. In countries that have experienced an epidemiological transition, chronic diseases and long-term healthcare needs, partially indicated by functional limitations, become the main concerns for population well-being^2–5. In this context, evidence on the social determinants of functional limitation trajectories is of critical importance in the era of rapid population aging. SES-health gradient is a classic topic in the field of population health^6,7. Given inter-connected life chances among family members, education is not only an important social determinant of personal health, but also has spillover effects on parental well-being^8–10. Despite weakened intergenerational solidarity between adult children and parents owing to the modernization¹¹, there is latent reciprocity within the intergenerational relationship. The supportive latent kinship matrix of family networks over the life span could be triggered by needs of family members^12,13. Previous studies documented the protective effect of children’s education on parental longevity^8,14–17. The positive association between adult children’s education and parents’ well-being could be explained by the social support theory in a life course perspective. The life chances of family members are mutually influenced over their life courses, and resources within the family networks are important sources of health^18–20. Lower-educated children may encounter more financial insecurities and provide less financial support to parents²¹. In contrast, parents may share resources and have easier access to the healthcare and medical services of higher-educated children, which could reduce the risks of limitations on their physical functioning²². In addition, since children can monitor parents’ health behaviors and influence parents’ attitudes toward the utilization of healthcare service, parents with better-educated adult children may be more capable of maintaining optimal physical functioning²³.

Similar to the education-health linkage in the long run^24,25, the importance of offspring education in parental well-being may vary over the life span²⁶. According to the SES-health gradient theory in a life course perspective, there are two competitive hypotheses on the effect of adult children’s education on parents’ health over the life course. According to the cumulative disadvantage perspective, the health disparities between advantaged and disadvantaged groups widen with age^27–29. Lower-educated people experience not only higher exposure to health risks, but also greater employment instability and more financial difficulties^24,27. These factors, taken together, seem to contribute to a widening gap in health between higher and lower educated groups. Meanwhile, according to the social support theory from a life course perspective, there are spillover effects of resources and stressors on other family members^18–20. Parents having better-educated adult children are more likely to have healthier behaviors and are more capable of managing diseases through healthcare utilization^15,30. Meanwhile, the intergenerational exchange between young children and parents responds to resources and needs over the lifespan^31,32. When parents age, they face higher risks of health deterioration and being diagnosed with chronic diseases, which may increase healthcare costs^2,3. Apart from this, parents place higher values on family connections with their spouse and children as they get older³³. Therefore, parents’ dependence on adult children may rise and thus the importance of children’s support in maintaining parental physical well-being may get stronger over the aging process.

In contrast to the cumulative disadvantage hypothesis, age-as-leveler hypothesis supports converging health trajectories across different education groups with age²⁵. The age-as-leveler hypothesis suggests that the age-dependence of health is stronger in late adulthood and the education returns to health weaken the older one gets, due to mortality selectivity at the aggregate level^24,25. Since age could be an overarching factor of health deterioration in parents’ later life and older adults with lower-educated children have a higher death hazard^17,26, the strength of the association between children’s education and parental physical health may weaken as people age. Lee²⁶ showed that the effect of children’s education on parental depressive symptoms weakens as parents age in Taiwan. However, it remains unknown how children’s education may contribute to the trajectories of older adults’ physical well-being as parents age.

Moreover, the salience of the intergenerational bond in later life may be contingent on the socio-cultural context³⁴. In East Asia, predominated by meritocracy culture, children’s education is taken as the key to family’s upward mobility^35,36. Moreover, older adults in East Asia emphasize filial piety and anticipate support from children³⁷. Parents may share the returns to children’s education given the persistent intergenerational solidarity in later life^{8,26,32,38,39}. However, parents who adhere to patrilocal norms may perceive the primary responsibilities of elder care as falling primarily on the son and daughter-in-law^40,41. Owing to the legacy of gender-specific role expectations on children^40,42, parents invest more in their son’s education so as to receive education returns from the son⁴³. Men’s advantages in education over women expand as the number of siblings increases⁴³. Nevertheless, a reversed gender gap in education has emerged among more recent cohorts under the college expansion policy⁴⁴. Paying more attention to these cohorts in China may help shed light on the rationale behind human capital investment strategies within the family in transitioning East Asian societies.

Furthermore, the importance of the educational returns for parents may be contingent upon the extent of need for old-age support. In advanced societies with complete old-age safety networks, intergenerational transfer from children to parents is less frequent. A study in Sweden using instrumental variables showed that children’s education is not associated with parental longevity⁴⁵. Meanwhile, some studies showed significant positive spillover effects from children’s education to parental survival in the USA¹⁵ and in Sweden³⁰. In contrast, the economic support related to children’s education is highly appreciated in developing contexts with no complete formal old-age support^46–48. However, the informal support from children to aging parents in developing contexts may be challenged by the uncertainties surrounding children’s life course transitions during periods of modernization⁴⁹. Specifically, geographic distance and the delayed transition to marriage may constrain daily contact, the flow of instrumental support, and emotional closeness between children and older parents^42,50. Thus, it becomes less clear how children’s education may protect the well-being of older adults in developing contexts that are moving rapidly towards modernization. As shown in previous studies, parents of higher-educated adult children have a lower death hazard compared with parents of lower-educated children^16,51. In South Africa, a one-year increase in adult children’s schooling is associated with 5% lower mortality risk for mothers and 6% lower mortality risk for fathers¹⁴. In China, parents with better-educated children show a lower rate in death hazard^38,51. Among few exceptions considering functional limitations, the findings of the association between children’s education and parental physical functioning are inconsistent. The study in Mexico showed that adult children’s education is not associated with parents’ physical functioning changes two years later¹⁶, whereas a study in Taiwan using cross-sectional data showed that parents with highly-educated children are 39% less likely to report physical functioning limitations than their counterparts with lowly-educated children⁵². This deserves more attention to the role of children’s education in the long run, particularly regarding its impact on physical functioning in developing contexts witnessing rapid population aging.

China context

The linked lives between children and parents last into parental later life in China⁴². Considering the contingency of the intergenerational tie on parental well-being across socio-cultural contexts³⁴, gender is a critical consideration in the association between adult children’s education and parental well-being in East Asian societies predominantly shaped by Confucianism and patriarchal systems. The legacy of Confucianism and patriarchal norms has fostered gender-specific expectations regarding filial piety, anticipating a higher dependence on sons⁴⁰. Adult sons are supposed to live with parents and take on the primary responsibility of elder care⁵³. The patrilocal tradition allows the sons to monitor parents’ health behaviors in a direct way, and provide them immediate social support^19,20. To take account of these normative gender-specific role expectations on children, adult sons’ schooling seems to be more important for parents than that of adult daughters. However, the practices of filial obligations deviate from the traditional gender-specific expectations in recent decades of contemporary China. In particular, daughters play a more important role in supporting aging parents than sons, consistent with the kin-keeper perspective^54–56. The practical reversal of gendered filial responsibilities may suggest a new trend in the educational returns between sons and daughters. The rise in women’s education owing to the expansion of college education system may have increased their bargaining power in family decisions^57–59. The perceived filial obligations among daughters-in-law tend to decline when the demands on eldercare conflict with their work^60,61. Then, the role of daughters-in-law in providing elder care may be weakened, whereas the role of daughters may grow with the rise in women’s education⁶¹. Taking account of these changes in feedback to parents in recent decades, daughters’ education returns to parents may be similar to or higher than that of sons.

China has been faced with rapid population aging since the beginning of the 21st century⁶². The graying population is expected to be a challenge for healthcare and the pension system⁶³. Due to limited expenditure on social welfare, the universal safety network to provide formal old-age support has not been established despite national efforts in recent years^47,64. In developing contexts where there is a lack of institutional old-age support, intergenerational support is significant for the well-being of older parents⁶⁵. The support from extended family remains an important source of old-age security although the economic independence of Chinese older adults has improved^39,47,55. The parental practical concerns over quality of later life may strengthen their desire for support from children^37,64. This is particularly true for older adults in rural areas⁶⁵. The rural-urban divide, described as “one country, two societies”, manifests China’s social inequality^66,67. Older adults in rural areas who receive a minimum monthly income around 150 RMB under the New Rural Pension Scheme^47,64 are disadvantaged in terms of old-age securities and economic independence compared with their urban counterparts. Additionally, urban parents who have a higher socioeconomic status are more capable of coping with life stressors, whereas their less advantaged rural counterparts tend to be more vulnerable and suffer from more severe stressors⁶⁸. Then, urban parents are less likely to be affected by the negative sides of lower-educated children and family strains⁶⁹. As a result, children’s support related to their education may compensate for the lack of infrastructural old-age support for their parents whereas children’s life strains may highlight parental vulnerabilities in rural areas. Then, the marginal effect of children’s education on older adults in rural areas could be higher than that for urban parents.

Despite accumulated knowledge on the role of children’s education on parental longevity in developed contexts, it remains unknown how children’s education may influence the trajectories of parental physical well-being over the aging process, particularly in developing contexts. To address the literature gap, this study innovatively examines the long-term association between children’s education and parental functional limitations within developing contexts using the China case. To the best of our knowledge, this is the first study to use physical functioning to examine the long-term effects of children’s education on parental physical health disparity trajectories, thereby advancing our understanding of the social determinants of physical functioning trajectories among older adults in developing contexts with similar historical changes. Compared to the use of cross-sectional data, employing panel data and the growth curve model in this study not only reveal the dynamics of physical functioning over the aging process but also offer methodological advantages by accounting for unobserved heterogeneity across individuals. Additionally, this study includes sub-sample analyses by child gender to test the normative ‘wisdom’ of son preference in family resource distribution, thereby highlighting the changes in the cultural contingency of intergenerational solidarity in East Asian societies as they move away from patriarchal traditions. Lastly, this study contributes to the resource-contingency of children’s education returns to parental well-being within a specific context by demonstrating the rural-urban divide in the linked lives.

Methods

Data and samples

This study uses four-wave data from the China Health and Retirement Longitudinal Study (CHARLS), implemented by Peking University in 2011, 2013, 2015 and 2018⁷⁰. The dataset in 2011 is the baseline survey. CHARLS is a national representative survey that took its design from the Health and Retirement Study (HRS) in the US for middle age and older adults, respondents who are 45 years or older. The CHARLS respondents (main respondents and their spouses) are followed up on every two years. In the analytic sample, we include older adults who have at least one child. Based on the assumption that most children have completed their education by age 25, we include samples whose children were over 25 at the start of the study⁸. Moreover, we restrict analytic samples to respondents aged over 60. At the baseline, there are 7,407 samples. After dropping respondents who had never married or had no living children, the sample size was 6,988. The sample size of respondents followed through four waves is 4504 and the number included in this study turned out to be 4,272. In this study, we focus on one child per parent and select the highest level of education among multiple children, to be consistent with De Neve and Harding¹⁴ and Jiang³⁸. Among parents whose son and daughter have the same education level, only one child’s information would be kept. In the analytic process, observations with complete information in our analytic variables are included. All methods were carried out in accordance with relevant guidelines and regulations. All experimental protocols were approved by Peking University. Informed consent was obtained from all subjects or their legal guardians.

Variables and measurements

In this study, we have included two dependent variables to measure physical functioning: instrumental activities of daily living (IADL), and activities of daily living (ADL). To measure IADL, respondents report the difficulty of performing five tasks, including walking 1 km, climbing several flights of stairs, kneeling/crouching, extending arms, and picking up coins, similar to Yahirun and colleagues’ study in Mexico¹⁶. In this study, the possible responses to the IADL questions are mapped on a four-point scale, with points ranging from 0 to 3 (no, I don’t have any difficulty = 0; I have difficulty but can still do it = 1; Yes, I have difficulty and need help = 2; I can’t do it = 3), which represent the level of difficulty in performing functional tasks in IADL. The summary score of the five items represents the level of IADL. To measure ADL, respondents report whether they have difficulties in dressing, bathing, eating, getting out of bed, using the toilet, or controlling urination and defecation⁷¹. The possible responses to the ADL questions include a four-point scale, with points ranging from 0 to 3 (no, I don’t have any difficulty = 0; I have difficulty but can still do it = 1; Yes, I have difficulty and need help = 2; I can not do it = 3), which represent levels of difficulty in performing functional tasks in ADL. The summary score of the six items represents the level of ADL.

The key independent variable is children’s schooling. Children’s education is referred to with a single question: “What’s the highest level of education your child completed?” The answers include various levels from no formal education (illiterate), did not finish primary school but is capable of reading or writing, Sishu/home school, elementary school, middle school, high school, vocational school, two-/three-year college/associate degree, four-year college/bachelor’s degree, post-graduate/master’s degree, or post-graduate/doctoral degree. We recode the responses to generate the year of the child’s schooling. Covariates include time-varying variables and time-constant variables. Time-varying covariates include age, marital status (married = 1, widowed/separated = 0), number of living children, and logged household expenditure per capital. Family resources is indicated by the household expenditure per capital⁷². For older adults, we assume their hukou status to be relatively stable. Since older adults’ hukou is highly associated with whether their midlife occupation is farming (r = 0.44), these analyses do not consider the occupation of the older adults. Therefore, we have three time-constant covariates, including gender (male = 1, female = 0), education and hukou status (urban = 1, rural = 0). Descriptive statistics of all independent variables used across all waves combined are presented in Table S1.

Analytic approach

The analytic approach in this study is Growth Curve Model (GCM, hereafter). GCM is a special multilevel modeling method, which examines both intra-individual and inter-individual differences in health trajectories⁷³. It includes both time-varying and time-constant covariates, to control both intra- and inter-individual confounds. Moreover, GCM allows random-effects variations across individuals, to take account of the unobserved heterogeneity across individuals over the aging process⁷⁴.

1

2

3

4

Level-1 primarily uses time and time-varying covariates, to show the effects of time-varying covariates on the changes of parents’ physical functioning limitations, similar to fixed effects. In Eq. (1), AgeC is the analytic time metric which is centered by the average age of all respondents (i.e., 70) across waves. In this way, the intercept represents the level of physical functioning limitation at age 70, and the slope represents the rate of increase in physical functioning limitation during the study period. The level-1 equation includes age centered on the mean (AgeC), marital status (Mar), number of living children (Child), logged household expenditure per capital (Lnexp), to predict the dependent variable. Specifically, , , are the coefficients of age, marital status, number of living children and logged household expenditure per capital on dependent variable. As the random intercept, captures the average level of difficulties in physical functioning. As the random slope, captures the rate of linear changes over time.

Level-2 includes two parts: using the intercept model to estimate in Eq. (1) whereas using the slope model to estimate in Eq. (1) for the random effects of time-constant variables. Specifically, as shown in Eq. (2) and Eq. (3), Level 2 includes the highest education of adult children (Cedu), gender (Male), education (Edu), and hukou (Urban) to estimate the intercept and slope in Eq. (1). In Eq. (2), , , and could be interpreted as the predicted changes in the mean of dependent variables for one unit increase in the time-constant variables. In Eq. (3), , , and could be interpreted as the predicted linear changes over time of dependent variables for one unit increase in the time-constant variables. The analytic tool is Mplus 8, and specific details could also refer to Mplus User’s Guide.

Results

Descriptive findings

We demonstrate the physical functioning trajectories from 2011 to 2018 among older adults in Figs. 1 and 2, respectively. Figures 1 and 2 display the increasing trends in the difficulty levels of instrumental activities of daily living (IADL, hereafter) and activities of daily living (ADL, hereafter) among older adults. That is, there are diverging physical functioning disparity trajectories across children’s education groups.

Fig. 1.

IADL scores by children’s education among older adults over time. Higher IADL scores indicate higher levels of functional limitations. Figure 1 shows the increasing trends in the difficulty levels of instrumental activities of daily living (IADL) from 2011 to 2018 across different children’s education groups among older adults. Older adults with lowly-educated children report a faster pace of functional decline.

Fig. 2.

ADL scores by children’s education among older adults over time. Higher ADL scores indicate higher levels of functional limitations. Figure 2 shows the increasing trends in the difficulty levels of activities of daily living (ADL) from 2011 to 2018 across different children’s education groups among older adults.

Analytic results

Table 1 shows the GCM results on IADL among older adults. Model 1 includes all time-varying and time-constant variables, and Model 2 further considers the residual covariance of the intercept and slope of the GCM. The results shown in Model 1 and Model 2 are largely consistent. As shown in both Model 1 and Model 2, the level of difficulty with IADL increases with age for older adults. Adult children’s schooling is negatively associated with the average level of difficulty with IADL, and the interactive effect of offspring education and age is negatively significant. Specifically, older adults with better-educated children have lower levels of difficulty with IADL, and higher levels of children’s education could moderate the rate of linear changes in the IADL trajectory. The health disparity according to children’s education levels widens as parents age. In addition, older adults’ own education, gender, and urban hukou are negatively associated with the average level of difficulty with IADL among older adults. A lower average level of difficulty with IADL is found among male parents, parents with more schooling, and those with an urban hukou. Higher levels of difficulty with IADL are found among those older adults with more children and those with higher levels of family expenditure.

Table 1.

Growth curve models predicting the effect of children’s education on IADL among older adults, China Health and Retirement Longitudinal Study, 2011 to 2018.

Variable	Model 1		Model 2
	Coef.	S.E.	Coef.	S.E.
Child schooling years	-0.066***	0.011	-0.071***	0.012
Age (centered on the mean)	0.199***	0.015	0.200***	0.015
Child schooling*age	-0.004**	0.001	-0.004**	0.001
Education	-0.052***	0.012	-0.053***	0.011
Male (ref., Female)	-1.078***	0.080	-0.971***	0.077
Urban (ref., Rural)	-0.392***	0.107	-0.305**	0.103
Married (ref., widowed/separated)	0.025	0.076	0.003	0.076
Number of living children	0.076***	0.024	0.077***	0.023
Logged household expenditure per capital	0.088***	0.019	0.095***	0.019
Random effects-variance component
Level 1: within person	4.553***	0.063	4.579***	0.063
Level 2: in intercept	4.121***	0.129	4.565***	0.139
Level 2: in linear growth rate	0.022***	0.002	0.018***	0.003
Constant	3.011***	0.222	3.004***	0.221
Residual correlation between intercept and linear growth rate			0.207***	0.013
AIC	81910.563		81572.874
BIC	82011.263		81681.320
ICC	0.470		0.470
n of persons	4272		4272
n of person-year observations	17,088		17,088

AIC = Akaike information criterion, BIC = Bayesian information criterion, the smaller the better; ref = reference; Coef.=coefficient; S.E.=standard error. p < 0.1 +, p < 0.05 *, p < 0.01 **, p < 0.001 *** (two tailed tests).

Table 2 shows the results of GCM estimating the effect of children’s education level on ADL among older adults. Model 2 in Table 2 also considers the residual covariance between the intercept and slope in addition to the covariates in Model 1. The model results in Model 1 and Model 2 are consistent. The model results show that adult children’s schooling is negatively associated with the average level of difficulty with ADL. That is, older adults with better-educated children report lower levels of difficulty with ADL. Meanwhile, older adults report more difficulty with ADL as they age. The interactive effect of adult children’s education and age is negatively significant. That is, adult children’s schooling could also attenuate the linear rate of increase in the difficulty with ADL over time. Moreover, older adults’ own education is also negatively associated with the average level of difficulty with ADL; older adults who are male and have urban hukou report lower levels of difficulty with ADL. Older adults with more children report more difficulty with ADL. The association between household expenditure and the likelihood of having difficulty with ADL is positive. A potential explanation is that providing instrumental support and medical support to older adults who have difficulty with ADL may increase household expenditures.

Table 2.

Growth curve models predicting the effect of children’s education on ADL among older adults, China Health and Retirement Longitudinal Study, 2011 to 2018.

Variable	Model 1		Model 2
	Coef.	SE	Coef.	SE
Child schooling years	-0.035***	0.007	-0.037***	0.007
Age (centered on the mean)	0.088***	0.011	0.084***	0.010
Child schooling*age	-0.003**	0.001	-0.003**	0.001
Education	-0.014***	0.007	-0.014***	0.007
Male (ref., Female)	-0.230***	0.049	-0.198***	0.045
Urban (ref., Rural)	-0.130**	0.066	-0.117**	0.061
Married (ref., widowed/separated)	0.015	0.049	0.008	0.048
Number of living children	0.034*	0.015	0.035*	0.015
Logged household expenditure per capital	0.072***	0.013	0.079***	0.013
Random effects-variance component
Level 1: within person	1.997***	0.028	1.994***	0.027
Level 2: in intercept	1.308***	0.047	1.640***	0.056
Level 2: in linear growth rate	0.020***	0.001	0.016***	0.001
Constant	0.546***	0.143	0.522***	0.142
Residual correlation between intercept and linear growth rate			0.121***	0.006
AIC	67741.529		67155.632
BIC	67842.229		67264.078
ICC	0.424		0.424
n of persons	4272		4272
n of person-year observations	17,088		17,088

AIC = Akaike information criterion, BIC = Bayesian information criterion, the smaller the better; ref = reference; Coef.=coefficient; S.E.=standard error. p < 0.1 +, p < 0.05 *, p < 0.01 **, p < 0.001 *** (two tailed tests).

Table 3 shows the GCM results on IADL and ADL among older adults, presenting adult sons’ and daughters’ samples separately. As shown in Model 1, the sons’ education has significantly negative effects on both the average level of difficulty with IADL and the rate of linear increase in difficulty with IADL among parents. Meanwhile, the sons’ education has a significantly negative effect on the average level of difficulty with ADL. However, the interaction term of the sons’ education and parental age is marginal significant (i.e. p < 0.1) in Model 3. Similarly, the daughters’ education has a significantly negative effect on the average level of difficulty in parental physical functioning. Additionally, the daughters’ education is not significantly associated with the linear increase in parental physical functioning, as shown in Models 2 and 4. That is, the strength of the association between the daughters’ education and parental physical functioning does not change as parents age. Overall, the effect of the daughters’ education on parental physical functioning is largely similar to that of the sons.

Table 3.

Growth curve models predicting the effect of children’s education on physical functioning among older adults by gender of children, China Health and Retirement Longitudinal Study, 2011 to 2018.

Variable	On IADL				On ADL
	Model 1 (Son)		Model 2 (Daughter)		Model 3 (Son)		Model 4 (Daughter)
	Coef.	S.E.	Coef.	S.E.	Coef.	S.E.	Coef.	S.E.
Child schooling years	-0.075***	0.015	-0.068***	0.019	-0.039***	0.010	-0.033**	0.012
Age (centered on the mean)	0.203***	0.021	0.188***	0.024	0.078***	0.014	0.083***	0.017
Child schooling*age	-0.004*	0.002	-0.003	0.215	-0.002+	0.001	-0.002	0.002
Education	-0.047**	0.016	-0.055**	0.019	-0.010	0.010	-0.019+	0.011
Male (ref., Female)	-1.072***	0.104	-0.759***	0.130	-0.261***	0.062	-0.104	0.076
Urban (ref., Rural)	-0.400**	0.147	-0.253	0.167	-0.167+	0.089	-0.083	0.098
Married (ref., widowed/separated)	-0.086	0.102	0.111	0.128	-0.009	0.065	0.020	0.083
Number of living children	0.050	0.032	0.116**	0.041	0.013	0.020	0.063*	0.025
Logged household expenditure per capital	0.146***	0.026	0.034	0.033	0.076***	0.017	0.098***	0.022
Random effects-variance component
Level 1: within person	4.616***	0.085	4.403***	0.103	1.972***	0.036	2.126***	0.049
Level 2: in intercept	4.542***	0.186	4.545***	0.233	1.531***	0.071	1.657***	0.100
Level 2: in linear growth rate	0.018***	0.004	0.016***	0.004	0.017***	0.002	0.016***	0.002
Constant	2.777***	0.298	3.181**	0.376	0.675***	0.190	0.180	0.247
Residual correlation between intercept and linear growth rate	0.202***	0.017	0.191***	0.020	0.102***	0.007	0.131***	0.010
AIC	45,769		28,317		37,581		23,711
BIC	45,869		28,411		37,682		23,804
ICC	0.469		0.471		0.433		0.402
n of persons	2388		1489		2388		1489
n of person-year observations	9552		5956		9552		5956

AIC = Akaike information criterion, BIC = Bayesian information criterion, the smaller the better; ref = reference; Coef.=coefficient; S.E.=standard error. p < 0.1 +, p < 0.05 *, p < 0.01 **, p < 0.001 *** (two tailed tests).

Table 4 displays the model results of the effect of children’s education on IADL and ADL among older adults, stratified by hukou status. As shown in Models 1 and 3, offspring schooling is not negatively associated with either the intercept or slope of parental IADL and ADL among urban older adults. In contrast, offspring schooling is negatively associated with both the average level and the rate of linear increase in parents’ functional limitations among rural older adults. Therefore, the findings confirm that adult children’s education is protective for the physical functioning of rural older adults only, who are more economically depend on their children^39,64. Additionally, we use sub-samples stratified by parental education, and further test the resource-contingency of the association between children’s education and parental physical functioning. The robustness check confirms that offspring education is more pronounced for lowly-educated parents, particularly in terms of IADL. The results have been shown in supplemental Tables 2 and 3.

Table 4.

Growth curve models predicting the effect of children’s education on IADL and ADL among older adults by hukou, China Health and Retirement Longitudinal Study, 2011 to 2018.

Variable	On IADL				On ADL
	Model 1 (Urban)		Model 2 (Rural)		Model 3 (Urban)		Model 4 (Rural)
	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
Child schooling years	-0.041+	0.025	-0.073***	0.013	-0.030+	0.016	-0.038***	0.008
Age (centered on the mean)	0.097*	0.040	0.203***	0.017	0.066*	0.029	0.087***	0.012
Child schooling*age	0.002	0.003	-0.004*	0.002	-0.001	0.002	-0.003*	0.001
Education	-0.077***	0.019	-0.041***	0.014	-0.016	0.011	-0.012	0.008
Male (ref., Female)	-0.661***	0.150	-1.056***	0.089	-0.132	0.088	-0.225***	0.052
Urban (ref., Rural)	/	/	/	/	/	/	/	/
Married (ref., widowed/separated)	0.072	0.155	-0.019	0.087	0.059	0.099	-0.007	0.055
Number of living children	0.133*	0.053	0.065*	0.026	0.038	0.033	0.033*	0.016
Logged household expenditure per capital	0.128***	0.040	0.090***	0.022	0.055*	0.026	0.084***	0.014
Random effects-variance component
Level 1: within person	3.042***	0.097	4.897***	0.075	1.357***	0.044	2.153***	0.033
Level 2: in intercept	3.441***	0.237	4.820***	0.163	1.496***	0.111	1.679***	0.065
Level 2: in linear growth rate	0.019***	0.005	0.018***	0.003	0.016***	0.002	0.015***	0.001
Constant	1.740***	0.512	3.145***	0.248	0.456***	0.335	0.516***	0.159
Residual correlation between intercept and linear growth rate	0.165***	0.021	0.218***	0.015	0.133***	0.012	0.118***	0.007
AIC	14,578		66,774		11,903		55,059
BIC	14,657		66,872		11,983		55,157
ICC	0.502		0.456		0.504		0.405
n of persons	823		3449		823		3449
n of person-year observations	3292		13,796		3292		13,796

AIC = Akaike information criterion, BIC = Bayesian information criterion, the smaller the better; ref = reference; Coef.=coefficient; S.E.=standard error. p < 0.1 ⁺, p < 0.05 *, p < 0.01 **, p < 0.001 *** (two tailed tests).

Discussion

The life chances of children and parents are interconnected over the life course, especially in the East Asian context, characterized by persistent intergenerational solidarity in parents’ later life^18,42,75. Meanwhile, East Asia shares the meritocracy traditions and the fact that children’s highest education level indicates the perceived family social status. Chinese parents who shoulder the responsibility of helping their children be high achievers in education may view their children’s highest education level as part of the fulfillment of parenthood. In this sense, children’s higher education level improves parents’ own appraisal of their success in parenthood. Moreover, parents whose children are better educated have a lower mortality risk^38,51. Despite that previous studies did not show any significant effect of offspring education on parental physical functioning changes over two years¹⁶, this study adds evidence of the long-term effect of offspring education on parental health indicated by physical functioning.

Second, our results show that children’s education has a significantly negative effect on the linear changes in parental physical functioning over time. That is, parents with better-educated children show a slower rate of deterioration in their physical functioning as they age. This finding provides support for the cumulative disadvantage hypothesis, which suggests a diverging health disparity trajectory across children’s education groups. In contrast to our finding, Lee and colleagues’ Taiwan study showed that the strength of the association between children’s education and parental depressive symptoms weakens as people age⁸. This may be contributed by an ‘inverted U-shaped’ distribution between age and depression among adults⁷⁶. The differentiated findings of this study also suggest that, compared with subjective well-being, physical health in later life may depend more on resources associated with education. In addition, our finding is consistent with the findings on the protective role of children’s education on parental longevity^8,51. Future studies may examine how the diverging physical functioning disparity trajectory may mediate the association between children’s education and parental mortality in the long run.

Third, this study demonstrates the cultural-contingency of linked lives within a specific context by testing the heterogeneity of the spillover effect of offspring education on parental physical functioning by child gender. This study shows that both sons’ and daughters’ education levels are negatively associated with the average level of limitation in parental physical functioning. The majority of adult children in the analytic samples (ages 25 to 49, born in 1961–1985) come from family with multiple children and are influenced by the son preference culture. Although patriarchal culture traditions emphasize the support from adult sons for older adults’ life securities, particularly according to the task-specific perspective⁴¹, this study does not provide evidence to support the hypothesis that an adult son’s education level is more important for parental physical functioning than an adult daughter’s education level. The findings are consistent with Torssander’s results on parents’ longevity in Sweden³⁰. Moreover, the largely similar effect of adult sons’ and daughters’ education levels on limitations in parental functioning may cast doubts on the traditional rationale of parenting strategy structured by the ‘son-preference’ culture⁴³. The norms of gendered education returns to parents in East Asian societies could be challenged by power dynamics in the family and children’s perceived filial obligations in the context of the expansion of the education system^44,60,61. As the adherence to filial piety declines, the closeness with daughters could become more important for parental well-being⁷⁷. The findings suggest that modernization may have changed the gendered practice of intergenerational relationships, which may further contribute to shifting childbearing and parenting preferences.

Fourth, this study also demonstrates the resource-contingency of linked lives within a specific context by testing the rural-urban divide in the spillover effect of offspring education on parental physical functioning. It confirms that the effect of children’s education on parental physical functioning is mainly significant for older adults with rural hukou. However, children’s education level is not significantly associated with urban parents’ physical functioning. Considering that urban older adults can use their pensions to cover the costs of daily living and care, while rural older adults are disadvantaged in receiving pensions and healthcare services, this divided access to formal support for older adults may structure their dependence on their children. Children’s education and associated resources may compensate for those parents who are in a disadvantaged status. In line with the findings of De Neve and Fink’s study⁴⁵, the findings on the rural-urban divide may suggest that the functional role of offspring education levels in protecting older adults may depend on whether institutional support is lacking. Furthermore, it suggests policy implications in terms of more emphasis on educational resources distribution to the disadvantaged group to address issues of stratification and health disparities. Moreover, older adults in rural areas who are often farmers, benefit more from earlier exit from their physically demanding work, leading to lower physical functioning limitations when they have highly-educated children. This may also help explain why the effect of children’s education is more pronounced among rural older adults. However, given the differences in cultural settings, it is uncertain that the conclusions based on China Context could be generalized to other emerging economies. Thus, it deserves more cross-cultural studies to test the link between adult children’s educational attainment and parental physical health in other emerging economies.

This study extends the health disparity trajectory over the life course and highlights the long-term effect of children’s education on parental physical functioning trajectories. Furthermore, our findings imply that the practices and perceptions of filial responsibility may be changing. In addition, our findings suggest that the strength of the intergenerational tie between adult children and parents may be contingent on the institutional support indicated by the rural-urban divide. However, this study is limited to show how the mechanisms underlying the linkage between adult children and parents may change as norms of filial responsibility for aging parents change across generations. Although we assume that biological factors may play a more important role in the deterioration of physical functioning in late adulthood, our findings mainly support the cumulative disadvantage perspective. However, this study may be limited in directly testing the age-as-leveler hypothesis and in capturing long-term trajectories due to using data covering only a 7-year study period. Given that pronounced changes in physical functioning may require long observation periods, the relatively short observation period of CHARLS, combined with a sample dominated by young older adults, present challenges in detecting significant marginal effects. A longer period of observation over the parental life course may provide us new information on the spillover effect of children’s education on parental quality of survival in later life. Even though the findings are derived from panel data, this study does not fully address potential unmeasured confounding factors, such as children’s caregiving behavior due to parental poor health, which may mediate the observed effects of offspring education on parental physical functioning. A more carefully designed experiment accounting for such confounders in future research may better imply the causal direction from children’s education to parental health.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Dan Chen solely conceived the study, developed the theoretical framework, conducted the empirical analysis, and wrote the manuscript.

Data availability

The data underlying this article are available and could be derived from sources in the public domain: https://charls.pku.edu.cn/.

Declarations

Competing interests

The authors declare no competing interests.