Allocation of Eldercare Responsibilities Between Children and the Government in China: Does the Sense of Injustice Matter?

Abstract

China’s large aging population poses grim challenges to eldercare provision. Against the background of withering traditional kinship-based eldercare and increasing significance of government-sponsored support programs, this study draws on data from the 2013 Chinese General Social Survey to investigate not only the correlation between the sense of social injustice and the preference of allocating eldercare responsibilities between public and private agents, but also how this correlation varies between urban and rural regions. We find that perceived social injustice is significantly correlated with the odds of designating the government, instead of family members, to shoulder eldercare responsibilities. Further mediation analysis suggests that this correlation is mediated through concerns about eldercare. On average, the link between perceived social injustice and preference of eldercare duty allocation is weaker in rural than in urban China. Theoretical and policy implications are discussed.

Appendices

Appendices

Appendix I

See Table 4.

Table 4 Descriptive statistics of variables

Appendix II

Methodological Details of the KHB Method

To examine the mediating effect, the following two models are fitted to predict a latent continuous variable of the way of allocating eldercare duties, as denoted to be allocation of eldercare duties*:

$$ {\text{Allocation}}\;{\text{of}} \;{\text{Eldercare}} \;{\text{Duties}}^{*} \; = \;\alpha_{\text{F}} + \beta_{\text{F}} {\text{Sense}}\; {\text{of}} \;{\text{Social}}\; {\text{Injustice}}\; + \;\gamma_{\text{F}} {\text{Concern}}\;{\text{about}} \;{\text{Eldercare + }}\delta_{\text{F}} {\text{Covariates}} + \varepsilon_{1} $$

(1)

$$ {\text{Allocation}}\; {\text{of}}\; {\text{Eldercare}} \;{\text{Duties}}^{*} = \alpha_{\text{R}} + \beta_{\text{R}} {\text{Sense}}\;{\text{of}}\;{\text{Social}} \;{\text{Injustice}} + \delta_{R} {\text{Covariates}} + \varepsilon_{2} . $$

(2)

Model (1) is the full model and model (2) is the reduced model, so the mediation effect of the concerns about eldercare is captured by the difference between β_R and β_F. However, the latent variable of allocation of eldercare duties* is unobservable. Instead, it is measured with a dichotomous variable that has a value of one if allocation of eldercare duties* ≥ 0 and zero otherwise. What follows is that the estimated coefficients—denoted, respectively, by b_R and b_F—introduce assumptions about the distribution of the error terms (denoted, respectively, as σ_R and σ_F), as in $b_{R} = \frac{{\beta_{R} }}{{\sigma_{R} }}$ and $b_{F} = \frac{{\beta_{F} }}{{\sigma_{F} }}$. Therefore, direct comparison between b_R and b_F is problematic, since σ_R is usually not equal to σ_F.

To resolve this problem, we adopt the KHB method. This method starts from regressing the concerns about eldercare situation on the sense of social injustice, and then uses the model residual instead of the original observed measure of perceived social injustice to fit model (1), as in model (3).

$$ Allocation\;of\;Eldercare\;Duties^{*} \, = \,\tilde{\alpha }_{R} + \tilde{\beta }_{R} Sense\;of\;Social\;Injustice + \tilde{\gamma }_{R} Concern\;about\;Eldercare_{{residual}} + \tilde{\delta }_{R} {\text{Covariates}} + \varepsilon _{3} . $$

(3)

Evidently, the variances of $\varepsilon_{3}$ and $\varepsilon_{1}$ are the same (i.e., denote the variance of $\varepsilon_{3}$ to be $\tilde{\sigma }_{R}$, so that $\tilde{\sigma }_{R}$ = σ_F). Also, $\tilde{\beta }_{R}$ = β_R. Hence, we have $\tilde{b}_{R} - b_{F} = \frac{{\tilde{\beta }_{R} }}{{\tilde{\sigma }_{R} }} - \frac{{\beta_{F} }}{{\sigma_{F} }} = \frac{{\beta_{R} - \beta_{F} }}{{\sigma_{F} }} = \frac{{\gamma_{F} }}{{\sigma_{F} }}b$. Furthermore, a test statistic for the null hypothesis of $\tilde{b}_{R} - b_{F} = 0$ can be constructed using the delta method, and we have ${{\sqrt N \left( {\tilde{b}_{R} - b_{F} } \right)} \mathord{\left/ {\vphantom {{\sqrt N \left( {\tilde{b}_{R} - b_{F} } \right)} {\sqrt {\left( {\frac{{\gamma_{F} }}{{\sigma_{F} }},b} \right)\sum {\left( {\frac{{\gamma_{F} }}{{\sigma_{F} }},b} \right)^{\prime } } } }}} \right. \kern-0pt} {\sqrt {\left( {\frac{{\gamma_{F} }}{{\sigma_{F} }},b} \right)\sum {\left( {\frac{{\gamma_{F} }}{{\sigma_{F} }},b} \right)^{\prime } } } }}$ ~ N(0, 1), where Σ is the covariance matrix for $\gamma_{F}$ and b. If the test statistic is significant, the mediation effect of concerns about eldercare exists. Note that for research question 2, the estimation of $\tilde{b}_{R}$ and $b_{F}$ is based on the logistic regression model due to the binary nature of the allocation of eldercare duties.

Appendix III: Cohort Variations

Cohort heterogeneity in the link between perceived social injustice and the preference of eldercare responsibility allocation is worth exploring for at least two reasons. The first reason is that cohort difference can be used to proxy the distinction between caregivers and care receivers since care receivers are always more senior than caregivers. This distinction matters because caregivers and care receivers could have contrasting eldercare preferences. For example, caregivers might prefer the government to take more responsibilities so as to reduce their own burden, while care receivers might prefer their children as the care provider due to influences of traditional ideas (Liang and Marier 2017).

The other reason to consider cohort heterogeneity is the potential influences of the one-child policy. Regulating that each couple can only have one child,⁵ this policy effectively shrinks the number of potential family caregivers, discouraging older parents to rely on the only child for eldercare. This preference for the government is also likely to the case for the only child, because, without siblings, their eldercare burden can be overwhelming. However, it is also possible that those affected by the one-child policy are reluctant to turn to the government. For example, one propaganda strategy of the government, when instigating the one-child policy in the 1980s, is to promise to take care of the older adults, but this promise apparently failed to be honored due to the huge financial pressure (Chen and Xia 2013). Against this background, recent propaganda and legal practices started to call for family-based eldercare. The feeling of being cheated on the part of the only child or their older parents can dampen their confidence in the government.

A binary variable is constructed to measure whether or not one belongs to the cohorts affected by the one-child policy. A convenient age cutoff is set at the age of 56. That is, respondents who are 56 years old or younger are treated as those affected by the one-child policy. This cutoff is determined by the following consideration. The one-child policy was promoted to the national policy and written into the Constitution in 1982. Assuming the couples in the 1980s gave birth to their child at around 25, those who were aged 56 or younger in 2013 should then be the individuals affected by the one-child policy. We also tested other age cutoffs, and similar results are obtained. No significant effect is detected for the interaction with either individuals affected by the one-child policy or cohort groups (Model 1 and Model 2 in Table 5), suggesting that the correlation between perceived social injustice and the preference of eldercare responsibility allocation is not altered by whether or not one is subject to the one-child policy. Also, since cohort is used to proxy the distinction between caregivers and care receivers, this distinction is also independent from the pattern of association between one’s sense of social injustice and allocation of eldercare duties.

Table 5 Result of the interaction effects: the heterogeneous choice model

See Table 5.