Living Arrangements of the Elderly in China and Consequences for their Emotional Well-being

We study the living arrangements and consequences for emotional well-being of the elderly in China using data from a national probability sample survey conducted in 2010, part of the China Family Panel Studies: 14,960 households were included and information was collected for each family member. We study 7,015 people in the sample age 60+. We find that, compared to living independently with one’s spouse, elderly respondents living with grown children are less happy, have less life satisfaction, and are more depressed, especially when the spouse is not sharing the household. The negative effects largely disappear when there also are grandchildren in the household although widows and widowers remain more prone to depression. Elderly people living in “generation-skipping” families suffer the same fate as living with adult children but no grandchildren—they are less happy and more depressed and, when not sharing responsibilities with a spouse, less satisfied with life than independent elderly couples. Finally, living alone or living with other relatives results in a significant degradation of emotional health. But the very small fraction of elderly respondents living with non-relatives enjoy the greatest happiness and the least depression. Living Arrangements and Emotional Well-being among Elderly in China 3 INTRODUCTION Due to a combination of increases in longevity and the one child policy, which has dramatically slowed population replacement, China’s population is aging very rapidly (Banister 1992; Chen and Liu 2009) and will age even more rapidly in the future (United Nations 2002, cited by Zeng and Wang 2003:98). Thus, China will soon have the largest population of elderly people in the world. Traditionally, old age security in China was managed mainly via co-residence with an adult child, nominally the eldest son, who continued to live in the parental household, or nearby, even after marriage and the appearance of children (Davis-Friedmann 1983:34-35; Whyte 2003:5). Even when new nuclear households were formed, elderly parents often moved to the home of one of their children when it became difficult for them to care for themselves or when they were widowed (Korinek, Zimmer, and Gu 2011). This pattern was particularly pronounced in rural China—which, until recently, included the bulk of the population—since there was (Chow 1991; see also Lee and Xiao 1998) and is (Cai et al. 2012) essentially no state support for the elderly in rural areas . To be sure, some (e.g., Levy 1949) have suggested that the elderlyparent-plus-adult-child household pattern was more the norm than the rule, at least since the middle of the 20 century. However, Cartier (1995:320) reports that as late as 1982 nearly 75% of elderly Chinese lived with their adult children. This no longer the case. Due to a combination of increased urbanization; changes in housing stock—the replacement of hutongs (dwellings arranged around courtyards and housing multiple, often related, families) with high rise apartments containing small units suitable only for nuclear families; and increased migration, which results in many adults living a great distance away from their parents, the proportion of elderly Chinese living with their adult children has declined substantially between 1982 and 2010 and the proportion living with their spouse and no one else has increased concomitantly (Zeng and Wang 2003). Specifically, Zeng and Wang showed from analysis of 1982, 1990, and 2000 census data that the percentage of men age 65+ living with children declined from 68% to 60% and the percentage living only with a spouse increased from 17% to 29%; the corresponding percentages for women were 74% to 69% and 11% to 19% (2003:Table 2). This trend has continued through 2010. In the data analyzed here the percentage of men age 65+ living with children declined to 33% and for women to 42% while the proportion of men living only with a spouse increased to 43% and for women to 30%. In addition, as internal Living Arrangements and Emotional Well-being among Elderly in China 4 labor migration has increased, many children have been left behind or sent back to live with grandparents by parents too busy to care for their children or lacking access to education or suitable housing for their children (Zeng and Wang 2003:104; Silverstein, Cong, and Li 2006). Thus, over the past 20 years or so, a new form of household has emerged, the “generation-skipping” household consisting of grandparent(s) and one or more non-adult grandchildren but no members of the middle generation. The concern of this paper is how the living arrangements of elderly Chinese affect their emotional well-being. While the effect of living arrangements on the well-being of the elderly in China has received considerable attention by researchers, our understanding is as yet incomplete. One difficulty is that researchers use a variety of classifications of living arrangements, often simply one specific category against all others, which makes comparisons across studies difficult (e.g., Chen and Silverstein 2000; Silverstein et al. 2006; Liu and Guo 2008; Sun et al. 2011; Ye and Chen 2014). In what follows, we explicitly contrast all pairs of living arrangements. Because we have data only on the non-institutionalized population, we do not consider the institutionalized elderly, who in any event constitute only a tiny fraction (less than 0.4%) of the elderly population (Zeng and Wang 2003:106 [Table 2]). What has been firmly established is that living alone is an undesirable condition. Those who live alone are less satisfied with life (Wang, Chen, and Han 2013:5[Table 3]), more depressed (Sun et al. 2011), and have lower overall psychological well-being (Chen and Short 2008:17[Table 4]; Wang et al. 2013:5[Table 3]) than those who live with others. This outcome may be due in part to the fact that those who live alone are usually widowed (in our data, 76% of those who live alone are widowed—92% among women and 57% among men; interestingly 23% of men live alone have never married, compared to only 1% of women who live alone, which reflects the fact that for women in China marriage is nearly universal). Widows who live alone have a double disadvantage—they must cope with the loss of a spouse, which is known to increase the risk of depression (Li et al. 2005), as well as the lack of other companionship. What is as yet unclear is whether living with children or grandchildren promotes or undercuts emotional well-being. The best evidence comes from the ongoing study of the Oldest Old (Zeng et al. 2002), a nationally representative panel survey, with an initial survey of people age 80+ in 1998 and new waves approximately every two years. In 2002 the survey was expanded to include people age 65-79 (Zeng and Vaupel 2004). Still, the available evidence to date comes from analysis of the oldest old population, those age 80 or more. Since 88% of the population age 60 or Living Arrangements and Emotional Well-being among Elderly in China 5 older is less than age 80 (computations from our data), results from the original oldest-old sample can hardly be taken as typical of elderly Chinese. Still, these results are suggestive and we report them here. Wang et al. (2013) studied life satisfaction (using a single item scale) and emotional well-being (the propensity to look on the bright side of things, to feel anxious or fearful, to feel lonely or isolated, to feel that the older you get the more useless you are, and to be as happy now as when younger, which they combined into a single scale) and found that married couples living with children and married couples living independently did not differ significantly with respect to either measure. Chen and Short (2008:Table 5) studied the same population and used the same measures, except that they combined them into two 3category measures: positive well-being (the quality of life now, the propensity to look on the bright side of things, the propensity to be as happy as when younger); and negative well-being (the propensity to feel fearful, lonely, or useless). They found no significant differences between those living with their spouse but not with a child, those living with a child but not with a spouse, and those living with both a child and a spouse. Several studies (Chen and Silverstein 2000; Silverstein et al. 2006; Liu and Guo 2008) find that living with children is beneficial, but they fail to distinguish between elderly couples living independently and elderly people living alone. To date there has been to our knowledge only one study focusing on the consequences for the elderly of living in “generation-skipping” households, that is, with grandchildren but not with children. Silverstein et al. (2006) showed that grandparents in generation-skipping households were less depressed than other elderly people in their sample, especially when remittances were sent back by parents who had gone out for work. But their data were restricted to a sample of elderly people from rural townships within Chaohu City, a primarily agricultural city in central Anhui Province with high rates of temporary out-migration of adults for work. Given the specificity of the sample, it is not at all clear whether their results would hold for China in general. In our analysis (described in detail below) we distinguish 10 categories of living arrangements, which enables us to assess the effect of various combinations—elderly living with spouse, with spouse and grown children, with spouse, children, and grandchildren, and so on. There are two competing hypotheses regarding the effect of living with grown children. The family support hypothesis posits a benefit of living with children on the ground that such arrangements facilitate material support (financial support and aid in daily life) and minimize loneliness and isolation, which may be a problem not only for those who live alone but also for elderly couples living independently. This is consistent with the claim of Davis-Friedmann Living Arrangements and Emotional Well-being among Elderly in China 6 (1983:49) that, at least until the 1980s, “Life-long interdependence remains the preferred parentchild relationship among both young and old.” For a contrasting view, see Logan, Bian, and Bian (1998), who find, using data from Tianjin and Shanghai, that both older and younger respondents prefer that the old people live separately. The family conflict hypothesis posits that the irritations of family life may undercut any advantages of social interaction, even if material support is enhanced. Such irritations may be particularly pronounced when an elderly person or couple live with a son because then the primary caregiver is likely to be a daughter-in-law and mother-in-law/daughter-inlaw conflicts are legendary in many cultures, not least China (Wolf 1968:869-870). Thus, we expect that those who live with their daughters, who are a small but non-trivial fraction of the elderly (over 10% of those living with a grown child in rural areas and over 20% in urban areas [Zeng and Wang 2003:111]), will be emotionally better off than those who live with their sons and that this gap will be more pronounced for elderly women than for elderly men. It often has been claimed that elderly Chinese value close relations with their children and their children’s families. As Silverstein et al. note (2006:S257-S258) ...what is distinctive about Chinese grandparents is that their contributions take place within a cultural system of filial piety. In a society that emphasizes collective family goals over individual goals, the contributions of grandparents to the welfare of their children’s families fulfill a cultural mandate and are highly valued.... Such provisions allow grandparents to command greater respect from younger generations and to better secure claims to filial piety, which in turn enhances their sense of purpose and self-worth within the family.... (See also Strom et al. 1999; Yan 2003; Mjelde-Mossey, Chi, and Lou 2005.) Insofar as this is so, living in households with both grown children and grandchildren should enhance emotional wellbeing compared to other living arrangements. Whether living in generation-skipping households also enhances emotional well-being is more problematic, because whatever benefit is derived from interaction with grandchildren and from contributing to the collective well-being of the family may be offset by the burden of caring for them. Moreover, in China as elsewhere, adolescent children can be difficult (Jessor et al. 2003). Thus, we might well expect differences in the feelings of the grandparents depending on the age of the child. Finally, we expect to replicate the widely observed negative consequences of living alone for the emotional well-being of the elderly due to the psychological costs of social isolation. Living Arrangements and Emotional Well-being among Elderly in China 7 DATA AND ANALYTIC STRATEGY Data The data used in this analysis are from the 2010 wave of the China Family Panel Studies, a (nearly) national multi-stage probability sample of Chinese families. In the 2010 wave, 14,960 households were included in the sample and interviews were conducted with all family members age 10 or older, resulting in a total sample of 57,115. Our sample is restricted to 7,038 people age 60 and over who responded to the adult questionnaire (see Table 2). We further restrict the sample to 7,015 by excluding the 23 elderly respondents living with children under age 16 to ensure that those living with children are living with adult children. The choice of age 60 as the cutoff point for defining the elderly population reflects the nominal retirement age for male workers (for women it is age 50 for ordinary workers and 55 for civil servants [Wikipedia 2013]). Davis-Friedmann (1983:3) asserts that “Age 60 ... marks a ... universally accepted point in time for entry into the oldest generation, and among those who have celebrated this birthday there are few who still identify themselves as middle-aged.” To be sure, the Chinese National Bureau of Statistics defines old age as beginning at age 65, as indicated by the fact that since 1982 separate statistics have been presented for this age group. Still, we think the nominal retirement age is the optimal cutting point. This follows conventional practice when an upper age limit is imposed in population surveys. For example, many U.S. studies restrict samples to those no older than 64. Details of the sample design are given in Xie, Qiu, and Lü (2012). Here it suffices to note that six strata were initially specified: four provinces (Gansu, Guangdong, Henan, and Liaoning) and a provincial-level city (Shanghai) were each treated as separate strata and a sixth stratum consisted of the remaining 20 provinces sampled. Within each of the four singleprovince strata, 16 counties were chosen at random but with probability proportional to size (PPS); however, in Shanghai 32 townships (which in urban areas are known as “streets”—jiedao) were chosen PPS as the first stage. Within the 20-province stratum, 80 counties were chosen PPS. Within each county four villages or neighborhoods were chosen PPS; for Shanghai, two villages or neighborhoods were chosen PPS within each jiedao. Within each village/neighborhood, 25 households were chosen at random. Because of the multistage design, it is necessary to take account of the resulting clustering of the sample; we specify the village/neighborhood as the cluster variable. In addition, we weight the data using “poststratification adjustment weights” for the Living Arrangements and Emotional Well-being among Elderly in China 8 adult sample, which take account of the differential sampling rates implied by the sample design, a correction for differential non-response rates, and a final adjustment to replicate the age-by-sex distribution of the 2010 census (Lü and Xie 2012). Since several variables used in our analysis have missing data, including a family income variable (592 missing values) and a measure of closeness to children (1,362 missing values, including those without living children), we imputed all missing values using Stata 13’s multiple imputation (-mi-) procedures, carrying out 10 imputations, and conducted our analysis using -miprocedures (StataCorp 2013). Our analysis is based on 7,015 completed cases. The legitimacy of multiple imputation turns on the plausibility of the assumption that missing values for the variables to be imputed are “missing at random” (MAR) (Rubin 1987; Little and Rubin 2002)—that is, that net of predictors of these variables included in a model there is no correlation between the true value of the variable and the likelihood that the value is missing in the data set. This is a bit problematic with respect to family income since those in high income families might be concerned about information being shared with the tax authorities. But since the variable refers to family rather than individual income, this is unlikely to be an important concern on the part of respondents. Moreover, despite vivid newspaper accounts about the Chinese nouveau riche, they represent only a tiny fraction of China’s population and are known to be resistant to being interviewed and hence are unlikely to be found in survey samples. The large amount of missing data on the “closeness to children” variable also is problematic. Of the 1,362 missing values, 182 are due to the fact that the respondent has no living children. However, the remainder are troublesome. One possibility is that people who are not close to their children are less likely to answer the “closeness” question because of the discomfort involved in admitting to themselves (and to the interviewer) that, against strong norms, they are not close to their children. Although no direct assessment of this conjecture is possible, it is possible to carry out a partial indirect test—if those who failed to respond to the closeness questions are significantly less happy and satisfied and more depressed than otherwise comparable people who did respond, we have a basis for suspecting that the missing data on the “closeness” reflect a lack of closeness to children, given that—as we will see below—closeness to children is positively associated with happiness and satisfaction and negatively associated with depression. To assess this possibility, we added a dummy variable to the set of variables in Table 4, coded 1 if a response to the “closeness” variable was missing and coded 0 otherwise. Living Arrangements and Emotional Well-being among Elderly in China 9 We also imputed missing values for the “closeness” variable. Given this, the coefficients associated with the dummy variables for missingness on “closeness” can be interpreted as the expected difference with respect to happiness, life satisfaction, and depression between those missing information on closeness and others. It turns out that those missing information on closeness are more happy and satisfied than expected, not less. They also are less depressed, although the coefficient for depression is not significant. From these results we conclude that the hypothesis that those who are not close to their children are less likely to respond to the closeness question is unlikely. Since we have no other hypotheses regarding the possibility that the closeness variable is not MAR, we treat it as MAR and impute the missing cases. Analytic Strategy The analysis we conduct here can be divided into three parts. Our first task is to describe the distribution of non-institutional living arrangements experienced by the elderly population. Second, we study the determinants of living arrangements, about which relatively little is known. Third, we consider the consequences of living arrangements for the emotional well-being of the elderly. For each of these tasks, we first introduce the variables studied and the models estimated and then discuss the results. Descriptive statistics for all variables used in the analysis are shown in Tables 1 and 2. DISTRIBUTION OF LIVING ARRANGEMENTS OF THE ELDERLY In our introductory discussion we noted that the living arrangements of the elderly in China have been changing rapidly as China has urbanized, the urban housing stock has become increasingly dominated by small apartments designed for nuclear families, and internal migration has increased. However, until now there has been no definitive assessment of the contemporary living patterns of the elderly or of trends in these patterns over time. Table 2 provides such estimates for 2010. In constructing the table, we classified our elderly respondents on the basis of the focal relationships identified in each line of the table. The first category consists of married 9 couples living together but without children or grandchildren —although there could be other relatives or non-relatives in the household as well. This is the reference category for the models shown below. The guiding principle in constructing our typology was to be able to distinguish the presence of a spouse, a grown child (or children, including inlaws), and a grandchild or children. To make these distinctions, and also distinguish between Living Arrangements and Emotional Well-being among Elderly in China 10 people living with other relatives, people living with non-relatives, and people living alone, requires 10 categories. We tried to reduce the number of categories by testing the significance of differences between pairs of coefficients when studying the effect of living arrangements on emotional outcomes (e.g., contrasting Category 2 with Category 5 to assess whether living with a spouse mattered among those living with grown children but without grandchildren), but concluded that there were too many instances in which corresponding coefficients were significantly different to warrant combining categories. Since 17% of elderly respondents identified in the household questionnaire failed to respond to the adult questionnaire, it is important to assess whether the respondents constitute an unbiased subset of all elderly adults. Lacking adequate data on the personal characteristics of nonrespondents, we resorted to the simple expedient of comparing the distribution of living arrangements among the elderly counted as household members with the distribution of elderly respondents to the adult questionnaire. Inspection of the right and left panels of Table 2 makes it clear that there is relatively little difference between the distributions (the indexes of dissimilarity, Δ, are, respectively, 4.4, 9.5, and 7.6 for rural, urban and all respondents). We thus are comfortable treating the sample of respondents as representative of non-institutionalized elderly people in China. From the left-hand column of the table, it is evident that 65% of the elderly population continue to live with their spouses and that of these fewer than half (40%) live with grown children or grandchildren. The tendency to live with a spouse but not with children or grandchildren is somewhat more common among the urban elderly than among the rural elderly (42% vs. 35%) and the tendency to live with grandchildren, whether in 3-generation or generation-skipping households, is somewhat more common among the rural elderly (35% vs. 27%); there are no other rural-urban differences of note. It also is evident that even in rural China, the 3-generation household is no longer dominant—if it ever was: 28% of the rural elderly live in 3-generation households, as do 22% of the urban elderly. Interestingly, ruralurban differences are more pronounced in the lefthand panel than in the right-hand panel, where—with the exception of the propensity for couples to live independently (Category 1)—they essentially disappear. This probably reflects rural-urban differences in the living arrangements of the oldest old, who were disproportionately unlikely to respond to the adult questionnaire due to higher rates of physical infirmity and cognitive impairment. Living Arrangements and Emotional Well-being among Elderly in China 11 Much has been made (e.g. China Youth Research Center 2006; Guo 2008; Chan 2009:9) of an increasing propensity to send children to live with their grandparent(s). The claim is that when two parents both go out for work, it is difficult to care for their children due to inadequate housing, inadequate supervision, and difficulties in arranging suitable schooling (Liang and Chen 2007; Chan and Buckingham 2008; Liang, Guo, and Duan 2008). Similarly, young professionals may lack the time to care for their children. Thus, they send them to live with grandparents, who presumably are happy to take on the task of caring for grandchildren. In the third section of the paper, we study whether this last claim is correct. For now, we simply note that “generation-skipping” households are quite uncommon; only about 6% of the elderly live in such households and most (81%) of these are households in which both grandparents are present. Finally, very few people live with other relatives or non-relatives but about 17% live alone. DETERMINANTS OF LIVING ARRANGEMENTS OF THE ELDERLY