Analytical framework in the Chinese context

Based on the analytical frameworks developed by Kemper (Reference Kemper1992) and Vlachantoni et al. (Reference Vlachantoni, Shaw, Evandrou and Falkingham2015), we divide the potential factors affecting the receipt of informal care into three groups: need factors, demographic factors and socio-economic factors. Studies have been conducted to examine the impact of these factors in several countries. First, need has been considered to be the direct reason for using informal care (Agree Reference Agree1999). Kemper's research suggested that ‘[t]he greater the need for care, the more … informal care the disabled elderly are expected to use’ (Reference Kemper1992: 425). Need for care has been measured by severity of physical disability. Most empirical studies have found that disability is an important factor affecting the receipt of informal care (Kemper Reference Kemper1992; Larsson and Silverstein Reference Larsson and Silverstein2004; Pickard et al. Reference Pickard, Wittenberg, Comas-Herrera, Davies and Darton2000; Suanet, Van Groenou and Van Tilburg Reference Suanet, Van Groenou and Van Tilburg2012; Vlachantoni et al. Reference Vlachantoni, Shaw, Evandrou and Falkingham2015).

Second, demographic factors such as age, gender, living arrangements and marital status have also been conceptualised as important determinants of care receipt. A strong association had been found between the receipt of care and older people's living arrangements (Pickard et al. Reference Pickard, Wittenberg, Comas-Herrera, Davies and Darton2000) or marital status (Glaser et al. Reference Glaser, Stuchbury, Tomassini and Askham2008; Vlachantoni et al. Reference Vlachantoni, Shaw, Evandrou and Falkingham2015). Older people living with someone else in the same household or those having a surviving partner have been found to be more likely to receive informal care. This suggests that the receipt of informal care is not only the result of the need for care, but also hinges heavily upon the availability of care. However, due to the heterogeneity of the population under investigation, the statistical significance of these demographic factors tends to vary considerably from one group of people to another.

The most commonly explored socio-economic factors in the literature are income and educational qualifications. The extent to which these factors play an important role is still an issue under debate. For those studies which have found a statistically significant impact, higher income levels or educational qualifications have been negatively associated with the receipt of informal care (Kemper Reference Kemper1992; Peek, Coward and Peek Reference Peek, Coward and Peek2000; Larsson and Silverstein Reference Larsson and Silverstein2004). When information on income has been unavailable, housing tenure or ownership of a car has often been used as a marker of income (Macintyre et al. Reference Macintyre, Ellaway, Der, Ford and Hunt1998).

It should be pointed out that all of the studies discussed above were conducted in developed countries. None of them examined the receipt of informal care in China, a developing country where the social, economic and political contexts are totally different from those of developed countries. To investigate the issue in the Chinese context, features unique to Chinese society should be given due attention. Like many developing countries, China has a dual economy structure where highly developed industrial sectors in cities and backward agricultural sectors in rural areas co-exist (Cai Reference Cai2012). This creates a divide in social and family structures between rural and urban China. One the one hand, nuclear families are becoming increasingly common in the cities. Filial piety, as the corner stone of Chinese traditional values, has been weakened (Cheung and Kwan Reference Cheung and Kwan2009). On the other hand, extended families are still the preferred family model in rural China, and traditional values have not undergone as much change in rural areas as in urban China (Silverstein, Cong and Li Reference Silverstein, Cong and Li2006). These rural–urban variations directly affect older people's access to, and their expectations of, informal care. Therefore, it is hypothesised in this paper that the receipt of informal care is affected by whether older people live in rural or urban areas in China.

Another factor is the availability of financial assistance from children. As mentioned above, the government policy stipulates that adult children who are unable to care for their parents should provide financial support instead. The existing literature suggests that the provision of financial assistance is a common practice among adult children in China (Zimmer and Kwong Reference Zimmer and Kwong2003). First, financial assistance changes the socio-economic status of older people. According to Kemper's (Reference Kemper1992) theories, this might affect their chances of receiving informal care. Second, it might affect older people's preference for different types of informal care. For example, older people receiving financial support from their children might turn to their spouses or other relatives for help so that their children will not be overburdened. In either case, financial assistance should be incorporated into the analytical framework in the Chinese context.

Research methods

Data analysis

Regression analyses were conducted to examine the effects of need, and demographic and socio-economic factors on the receipt of informal care. For the dependent variables with binary categories, we built binary logit regression models that could be expressed as follows:

(1) $$\displaystyle{{\Pr \left( {{\rm Car}{\rm e}_i = 1} \right)} \over {1 - \Pr \left( {{\rm Car}{\rm e}_i = 1} \right)}} = {\rm e}^{[{\rm \beta} _0 + \mathop \sum \nolimits_{k = 1}^m ({\rm \beta} _kx_{ik})]}$$

where $\Pr \left( {{\rm Car}{\rm e}_i = 1} \right)$ denotes the probability of an old person i receiving informal care, x _k (k = 1, 2 … m) are the independent variables and denote the need, and demographic and socio-economic characteristics of this person, and β _k (k = 0, 1 … m) are the coefficients of x _k . We calculated e^{β _k} for each independent variable, which could be interpreted as the odds ratios of care receipt when an older person's characteristics change by one unit or from one category to another.

The second group of dependent variables with multiple categories was examined in multinomial logit regression models that took the following form:

(2) $$\displaystyle{{\Pr \left( {{\rm Car}{\rm e}_i = j} \right)} \over {\Pr \left( {{\rm Car}{\rm e}_i = 1} \right)}} = {\rm e}^{\mathop \sum \nolimits ({\rm \beta} _0 + {\rm \beta} _k^{\left( j \right)} x_{ik})},\; j = {\rm} 2\;{\rm or}\;3.$$

$\Pr \left( {{\rm Car}{\rm e}_i = j} \right)$ denotes the probability of receiving informal care from a particular source j (j = 1, 2 or 3), x _ik are the characteristics of each person and ${\rm \beta} _k^{\left( j \right)} $ are the coefficients for x _ik under the source j. The model estimates the relative probabilities (or risks) of receiving informal care from source j against an arbitrarily selected base outcome. In the equation, outcome 1 is chosen as the base outcome. We calculated ${\rm e}^{{\rm \beta} _k^{\left( j \right)}} $ for each independent variable k and each care source j, which could be interpreted as the relative-risk ratios of receiving care from source j against the base outcome when an older person's characteristics change by one unit or from one category to another.

To investigate the determinants of informal care intensity, we built a two-part model. The first part is a logit model taking the form of Equation (1). The second part is a linear regression model which can be expressed as follows:

(3) $${\rm Hou}{\rm r}_i = \; {\rm \beta} _0 + \mathop \sum \nolimits_{k = 1}^m ({\rm \beta} _kx_{ik}) + \varepsilon _i$$

Hour _i denotes the weekly hours of informal care an old person i receives from his or her children, x _k (k = 1, 2 … m) are the independent variables, and β _k (k = 0, 1 … m) are the coefficients of x _k . It must be pointed out that the second part is a conditional model. Conditional on receipt of informal care, an older person's weekly hours of care from children change by β _k if this person's characteristics change by one unit or from one category to another. Combining Equations (1) and (3), we can calculate the overall marginal effects of the independent variables on care intensity for the entire sample.

The CHARLS used the cluster sampling under the PPS rule (see above), which means that the clusters were not sampled with equal probabilities. Meanwhile, there was non-response in the course of sampling. To compensate for the potential bias of the descriptive estimates, a range of sample weights were generated by the survey (Zhao et al. Reference Zhao, Strauss, Yang, Giles, Hu, Hu, Lei, Liu, Park, Smith and Wang2013). Since the focus of our analyses was the receipt of informal care at the individual level, we calculated the sample characteristics weighted by the individual weights with household and individual non-response adjustment. For the regression models, the sampling probabilities varied exogenously by design. In this case, both the weighted and unweighted coefficients were consistent, but the weighted results tended to be less precise (i.e. larger standard errors) (Solon, Haider and Wooldridge Reference Solon, Haider and Wooldridge2013). Therefore, we only report the unweighted regression results in this paper.

There was a non-trivial proportion of older people (17%) who did not report their income in the survey. We first tested the income variable with the missing values excluded from the sample in the regression analysis, and found that it was not a significant variable in any of the regression models. We then conducted multiple imputation on the income variable to impute the missing values (Rubin Reference Rubin1987), and included this variable in the regression models. Again we did not find a significant impact in any of the models. Finally, we excluded the income variable altogether and fitted the models again. Comparing the two sets of results, we found that excluding this variable made little difference to the modelling results. In light of these analyses, it became clear that the income variable was not an important determinant, and thus we decided not to report the regression results of the income variable. But we will discuss the implications of these results at the end of the paper.

For the independent variables, we report the odds ratios in the case of the binary logit models or the relative-risk ratios in the case of the multinomial logit regression models (see above). We report the robust standard errors and two-tailed significance levels of the z-statistic. We conducted a series of model specification tests. Since different tests do not always provide a consistent story, a simultaneous examination of these tests enables us to evaluate comprehensively whether the models are well-fitted. For the binary logit models, we report the results of the joint significance test, the McFadden's Pseudo-R ² statistic, the results of the Pearson goodness-of-fit test and the results of the Hosmer-Lemeshow (H-L) goodness-of-fit test. For the multinomial logit models, apart from the joint significance test, the McFadden's Pseudo-R ² statistic and the H-L goodness-of-fit test, we also report the results of the generalised Hausman test.Footnote ⁴ In the multinomial logit regression models, we explicitly exclude those older people who do not receive informal care from the analysis by design. Such a model design is based on the assumption of the independence of irrelevant alternatives (IIA). The generalised Hausman test indicates whether the IIA assumption is met and, as a consequence, whether the models with a partial sample generate unbiased estimates of relative-risk ratios.Footnote ⁵ The analyses were conducted using Stata version 13.

Conclusion and discussion

The proportion of older people in China is increasing rapidly, and so is the demand for social care. Since formal care services are not widely available yet, older people have to rely heavily on care provided by a spouse, children or other relatives in the family to meet their needs in their daily lives. So far very little is known in the existing literature about older people's informal care and support in the Chinese context. This paper uses a nationally representative sample – the second wave data of the CHARLS collected in 2013 – to examine the factors affecting receipt of informal care among people aged 60 and over in China. This concluding section will discuss three issues: the contribution of this paper to the international literature, the future landscape of care demand and possible pathways of building the social care system in China.

It was found that individuals' needs were the most important factor driving the receipt of informal care. For all of the regression models without exception, severity of disability had a significant impact. People with more severe disability were more likely to receive care from various sources. These findings are highly consistent with those in the developed countries (Glaser et al. Reference Glaser, Stuchbury, Tomassini and Askham2008; Kemper Reference Kemper1992; Larsson and Silverstein Reference Larsson and Silverstein2004; Pickard et al. Reference Pickard, Wittenberg, Comas-Herrera, Davies and Darton2000). Comparing different sources of care, we found that disabled people were more likely than independent people to receive care from a spouse or children, rather than receive care from other relatives in the family. The existing literature has shown that people tend to rely on those people who are closer to them in kinship for informal care (Dujardin et al. Reference Dujardin, Farfan-Portet, Mitchell, Popham, Thomas and Lorant2011; Hirst Reference Hirst2001). Our results suggest that this pattern is even stronger among older people with more severe disability.

With regard to demographic factors, living arrangements were found to be the most important determinant. People living alone were less likely than those living with someone to receive informal care. Among the recipients of informal care, people living alone were more likely to receive this care from other relatives in the family. Consistent with the existing literature (Glaser et al. Reference Glaser, Stuchbury, Tomassini and Askham2008; Vlachantoni et al. Reference Vlachantoni, Shaw, Evandrou and Falkingham2015), we found that in China older people's likelihood of receiving informal care increased with age. The only exception was that people in the higher age groups were not necessarily more likely to receive informal care from other relatives.

Education seemed to be the most important socio-economic factor. People with lower levels of education were more likely to receive informal care. We also investigated, separately, the impact of education for those older people with different levels of disability. The negative relationship between the educational qualifications and the likelihood of receiving informal care remains statistically significant.Footnote ⁶ These findings are consistent with studies conducted in developed countries (Kemper Reference Kemper1992; Larsson and Silverstein Reference Larsson and Silverstein2004; Peek, Coward and Peek Reference Peek, Coward and Peek2000). However, the results seem inconsistent with those reported by Lu, Liu and Piggott (Reference Lu, Liu and Piggott2015) on the basis of CHARLS Wave 1 data. This discrepancy might be attributed to the fact that the questionnaire design in relation to using informal care in Wave 2 is completely different from that in Wave 1.

We examined two factors that are unique to Chinese society, namely hukou status and regular financial assistance from children. It was found that people's hukou status played a significant role. People living in rural areas were more likely to receive informal care from other relatives in the family or to receive care from more than one source. Silverstein, Cong and Li (Reference Silverstein, Cong and Li2006) argued that extended families are still the norm in rural areas. Rural residents expect other relatives in the family to care for them, and other relatives are also more accessible as a source of care. Our results seem to support their theory.

We found no evidence that financial assistance from children affects the receipt of informal care. None of the regression models showed that this was a significant factor. We also investigated the impact of household income on receipt of informal care (see the Data analysis section). Similar to financial assistance from children, we did not find any evidence that household income was a significant factor in any of the models. These results seem to suggest that the receipt of informal care in China is not based on financial considerations.

The analysis in this paper was based on the framework developed by Kemper (Reference Kemper1992) and Vlachantoni et al. (Reference Vlachantoni, Shaw, Evandrou and Falkingham2015). This framework also has some overlap with Anderson's (Reference Anderson1995) behavioural model which has been widedly used to examine health-care utilisation. The advantages of using this kind of framework are twofold. First, it provides a useful tool to identify the potential determinants in a structured and systematic manner. Second, the potential determinants proposed in the framework are largely comparable in different contexts. To some extent, this facilitates international comparisons.

However, the limitations of this framework should also be given due attention. The framework assumes that all of the determinants have a one-way causal impact on care receipt. This assumption is worthy of scrutiny especially when some people may receive both informal care and health care. Previous empirical studies have treated the receipt of health care as an indicator of need (Murphy, Whelan and Norman Reference Murphy, Whelan and Normand2015). This is also the approach adopted in this paper. However, it should be borne in mind that such an approach assumes away the possbile joint determination between the use of informal care and health care. Put differently, it assumes that the use of health care is not an endogenous factor.Footnote ⁷ Given the focus of this paper, we did not test the endogeneity of using health care and thus did not specify the causal relationships between these two variables in the Chinese context. But this is certainly an important issue that merits a separate study in the future.

At present, approximately one in four older people aged 60 and over in China are receiving informal care from a family member to help with their daily activities (Table 1). As the Chinese older population reached 212 million in 2014, this translates into 53 million informal care recipients. To set this in context, the figure is comparable to the entire population of England (54 million in 2014) (Office for National Statistics UK 2015a ).

The research findings in this paper enable us to paint a crude picture of the demand for social care in the following decades. Given the rapid change in the age structure, there seems little doubt that China will face huge pressure to meet old people's care needs. Due to low fertility rates, concern has been raised that there may not be enough younger adults to look after their parents in the future (Li and Zhang Reference Li and Zhang2013). Two consequences are likely to follow. First, more people in old age will have unmet needs because of the unavailability of care from children. Second, younger adults will be overstretched by their care responsibilities. They might have to give up their jobs to care for their parents, or they might fall sick themselves and need care from others.

In response to the ageing population in China, the Chinese government has decided to relax the one-child policy and encourage people to have more children (Central Committee of the Communist Party of PRC 2015: Article 7.8). However, this will not necessarily lead to an immediate rise in the fertility rate. As a matter of fact, due to the high costs of child rearing in China, some young couples cannot afford to raise a second child, even though they may very much want to do so (Ruan et al. Reference Ruan, Si, Song and Tang2015). This implies that one or two isolated policies will not suffice to address the issue of population ageing in China. What is needed is a co-ordinated welfare system that addresses the issue from multiple dimensions. For example, the government could consider reforming the existing child benefit policy alongside the reform of the family planning policy. Financial subsidies could be provided to those couples who wish to have a second child to relieve them of the financial burden.

Demographic patterns do not always bear bad news. Our analyses show that, apart from children, spouses are another major source of care: 80 per cent of married care recipients received care from a spouse. The majority of people (76%) are still married in old age in China. This stands in stark contrast to developed countries, which have low marriage rates among the older population.Footnote ⁸ Therefore, it can be argued that in the case of China, spouse care to some extent could buffer the future decline in care from children. Moreover, China has a large gap in life expectancy between men (72.4 years) and women (77.4 years) at the moment (National Bureau of Statistics of PRC 2015). Many developed countries have seen this gender gap decreasing over the past decades as a result of a decline in avoidable mortality of older males (e.g. reduction in lung cancer caused by smoking) (Pattison et al. Reference Pattison, Mcpherson, Blakemore and Haberman2012). If China could follow the same demographic trajectory in the future, there would be more male spouse carers available to provide informal care.

Second, the Chinese government has invested heavily in the education sector in the past three decades, which has led to a rapid increase in the numbers and proportions of people with higher levels of educational qualifications (Chan, Ngok and Phillips Reference Chan, Ngok and Phillips2008). This means that older people in future cohorts will be better educated than those examined in this paper. Given the relationship between educational qualifications and the receipt of unpaid care, it seems reasonable to argue that reliance on informal care will be lessened among older people in future cohorts.

It must also be noted that informal care in China is not equally distributed among different groups of people. Some might have substantially higher unmet needs than others. All other things being equal, disabled older people living alone (predicted probability = 0.31) were only 50 per cent as likely as those living with their children (predicted probability = 0.60) to receive informal care. Formal care will be absolutely crucial to ease the pressure imposed on informal care. As the government puts in more resources to expand the capacity of formal care services across the country, the issue of equality should be seriously considered. The government might want to prioritise formal care services for those older people who have care needs but have no access to informal care, so that the government support can be directed to those who need it most. Formal care facilities, such as elderly care centres in the community or care homes, could concentrate on those people who live alone or have no children. This means that formal care services could act as a complement to informal care. The foremost objective at this stage is to identify unmet needs and fill in the gap in care.

Once the formal care sector has sufficient capacity and professional carers become more experienced in service provision, the government could consider moving the system in a more ‘carer-blind’ direction. In this case, the provision of formal care is based on people's needs for care, rather than the availability of informal care. Such a system not only helps people with similar care needs achieve equal health outcomes, but also relieves informal carers of caring burdens that might adversely affect their health, employment or quality of life in general.