| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Polisher Research Institute, Madlyn and Leonard Abramson Center for Jewish Life, North Wales, Pennsylvania.
2 Department of Sociology
3 Office of Population Research, Princeton University, New Jersey.
4 Center for Population and Health, Georgetown University, Washington, DC.
5 Bureau of Health Promotion, Department of Health, Taiwan, Republic of China.
Address correspondence to Jennifer C. Cornman, Polisher Research Institute, Madlyn and Leonard Abramson Center for Jewish Life, 1425 Horsham Rd., North Wales, PA. 19454. E-mail: jcornman{at}abramsoncenter.org
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. This study uses data from four waves of the Survey of Living Status of the Elderly in Taiwan that spans a 10-year period and employs latent growth curve models to examine trajectories of perceived support and the sociocultural factors that may explain variability in baseline levels of support and variability in changes in support as respondents age.
Results. Perceptions about social support appear to follow a linear trajectory across age, with significant variation in baseline perceptions and in patterns of change in perceived support across the sample. Sociocultural factors primarily explain differences in initial levels of support but also have some effect on changes in support.
Discussion. Despite the increasing pressures and demands on adult children's time that are associated with social and economic development, the elderly in Taiwan on average feel supported by their social networks, with the perceived availability of support increasing with age.
Most prior research on perceived social support has focused on Western populations. Less is known about elderly persons' perceptions about their social support in non-Western settings, such as the newly industrialized societies in East Asia, where most social support to the elderly is expected to be, and is, provided by the family (Hermalin, 2000
; Sung, 1990). These societies have relatively few social welfare programs for the elderly population, and many are experiencing rapid social and economic developments that are providing adult children with social and economic opportunities outside the family as well as opportunities for greater geographic mobility. Such development, coupled with rapid population aging, may be making traditional forms of family support more difficult to maintain (Lin, 1994; Ogawa & Rutherford, 1997).
Over the last 50 years, Taiwan has experienced rapid declines in fertility and increases in life expectancy concurrently with dramatic changes in labor force composition and increases in educational attainment for both men and women. Between 1952 and 1995, total fertility declined from 6.6 to 1.8, and life expectancy increased from 56 to 72 years for men and from 56 to 78 years for women, yielding a near tripling in the percentage of the population aged 65 and older (from 3% to 8%; Cornman, 1999). The percentage of the labor force employed in the agricultural sector fell from 56% in 1952 to only 8% in 2000, and female labor force participation rose from 19% in 1956 to 45% in 1995 (Cornman, 1999). School enrollment rates also increased dramatically in this period. For senior high school (15–17 year olds), enrollment rates increased from 8% to 87% between 1952 and 2000, and college enrollment rates (18–21 year olds) increased from 2% to 39% (Hermalin, Liu, & Freedman, 1994; Ministry of Education, 2001).
These social, demographic, and economic changes are likely to affect the availability of traditional family-based support. Lower fertility means there will be fewer children available to provide care. Increases in educational and employment opportunities can lead to adult children spending greater amounts of time outside the family unit as well as to greater demands on adult children's time. Previous work in Taiwan, for example, has shown that higher levels of education are associated with lower levels of extended family living (Freedman, Thornton, & Yang, 1994). Although the proportion of elders living with children (traditionally, a culturally desirable living arrangement) is still high compared with many Western societies, Hermalin (2000) estimates that the proportion of Taiwanese aged 65 and over living with children declined from 67% to 51% between 1976 and 1996.
In this research, we use a nationally representative sample of older Taiwanese adults to examine trajectories of perceived support across four periods of observation over a 10-year period. We attempt to determine (a) whether perceived support increases or decreases across time for Taiwanese elders, (b) whether marked variability in support trajectories exists, and (c) what factors predict patterns of support. Previous research has demonstrated that positive perceptions about one's support networks promote health and health care utilization (Cassel, 1976; Cobb, 1976; Cornman, Goldman, Glei, Weinstein, & Chang, 2003; Doeglas et al., 1996; Schmitz, Russell, & Cutrona, 1997) and are as influential as received support in predicting various health outcomes (Chi & Chou, 2001; Doeglas et al., 1996; Wallsten, Tweed, Blazer, & George, 1999). Therefore, understanding whether older adults perceive a loss in support as they age is an important research and policy question. This is particularly salient in settings like Taiwan, where social support is primarily provided by informal sources and where national governments actively promote the continuation of family and other forms of informal support as the primary source for caring for elders (e.g., Ogawa & Rutherford, 1997; Sung, 2001).
Background
Social support is a multidimensional construct that has been conceptualized and measured in a variety of ways. Most measures of support fall into one of three categories: measures of the degree of social integration (e.g., social network characteristics), measures of received support, and measures of perceived support that capture an individual's beliefs about the availability of support (Sarason, Sarason, & Pierce, 1990). Studies that have examined the stability of social support among the elderly have typically focused on network composition factors and received support. These studies, in general, find that network size and contact with close family and friends are stable during late life, although contact with distant relatives and peripheral friendships show some instability (Bergman, Neiderhiser, Pedersen, & Plomin, 2001; Field, 1999; Field & Minkler, 1988; Krause, 1999; Martire, Schultz, Mittlemark, & Newsom, 1999; Stoller & Pugliesi, 1991; van Tilburg, 1998). Previous research also shows that the support (tangible, emotional, informational, or instrumental) received by elderly persons is either stable or tends to increase over time, possibly because of an increased need for support (Kelman, Thomas, & Tanaka, 1994; Krause, 1999; Matt & Dean, 1993; van Tilburg, 1998).
A few studies have focused specifically on stability and change in perceived support in old age. Martire and colleagues (1999), for example, report small increases in the perceived availability of emotional and tangible support and no change in the perceived availability of informational support over a 5-year period. Field and Minkler (1988) find that positive feelings of closeness with children increased and satisfaction with relations with grandchildren did not change during the 14 years covered by their study. Krause (1999), examining measures of network contact, received support, and perceived support, shows that perceived support is the most stable dimension, changing little during the 3 years between baseline and follow-up.
Along with the identification of these general patterns, some studies also note significant individual-level variation in the patterns of change for some types of support in late life (e.g., Krause, 1999; Martire et al., 1999). Although relatively few studies examine personal factors associated with change, this variability in patterns of change might be explained by social, economic, and network characteristics. Cross-sectional studies show differences in perceptions of support by sex (e.g., Connidis, 1994), socioeconomic status (e.g., Krause & Borawski-Clark, 1995), marital status (e.g., Curran, McLanahan, & Knab, 2003; Lynch et al., 1999), network characteristics (Cutrona, 1986; Seeman & Berkman, 1988), and social involvement (Lynch et al., 1999; Seeman & Berkman, 1988). Most of the studies that have examined the relationships between social and demographic characteristics and change in social support examine change using two waves of data. These studies find that improved social support at follow-up is associated with more social contact (Seeman & Berkman, 1988; Walter-Ginzburg, Blumstein, Chetrit, Gindin, & Modan, 1999), larger social networks (Seeman & Berkman, 1988), higher socioeconomic status (Field, Minkler, Falk, & Leino, 1993), and having children living nearby (Walter-Ginzburg et al., 1999). Field and Minkler (1988) also find that contact with friends declined for men, but not for women.
At the same time, several other studies find that social characteristics have little influence on change in social support. For example, Martire and colleagues (1999), who examined change across five waves of data, show that the social and demographic characteristics they examine tend to more consistently predict variability in baseline support rather than variability in change in support. Similarly, Kelman and colleagues (1994) report that health and functional status influence change in support more than social and economic characteristics.
There is little research on Taiwanese elders' perceptions about the availability of social support to guide expectations about perceptions of social support or how they may change with age. It is quite possible that older Taiwanese adults may hold negative perceptions about the availability of social support as the burden for parental care and pressures from life outside the family make it increasingly difficult for families to provide support. At the same time, many members of this current generation of elders are embedded in large familial networks and have at least one child living nearby (Hermalin, 2000). Many also have adult children that hold traditional beliefs regarding children's obligations to support their parents in old age (Cornman, 1999). Consequently, Taiwanese elders may have positive perceptions about the availability of support. Our expectations about changes in perceived support are guided by findings from studies of Western populations. Therefore, even if Taiwanese elders hold negative perceptions, we can only hypothesize that perceptions about the availability of social support will remain stable or improve with age.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Multinomial logit models of the missing cases indicated that those who were lost to follow-up were more educated, less likely to be co-residing with children, and less likely to be married and had worse perceptions about their social support. Those who had died were older, more likely to be male and unmarried, had fewer children, were involved in fewer social activities, and had worse perceptions about the availability of social support. Because we drop these respondents from our analyses, we lose the part of the sample with the worst perceptions. Our analyses therefore most likely underestimate the degree of, and variability in, change in perceptions. To understand the potential effects on our results of excluding these respondents, we also examine alternative models that include the deceased and cases lost to follow-up. These models are discussed in detail at the end of Results.
Measures
Perceived support
In each survey year, respondents were asked to rate the availability of emotional and instrumental support: willingness of others to listen when the respondent needs to talk (emotional support), level of caring provided to the respondent (emotional support), and reliability of others to provide sick care (instrumental support). The structure and wording of the perception questions and the response categories, however, were not identical over the four waves of data collection. In 1989, the support questions were asked separately about spouses, children, and friends. To make these support measures more comparable with the measures in other years, a single support measure for each type of support in 1989 was created by averaging the responses regarding spouses, children, and friends and regrouping the average into the five categories of the original measures (Table 1).
|
To make these measures more comparable across surveys, we dichotomized each measure to reflect positive versus nonpositive perceptions (see Table 1) and summed the three items at each wave to produce a perceived support scale, which ranges from 0 to 3 (see Table 1). To determine whether our reported findings were robust to alternate approaches to measurement, we implemented several measurement and modeling strategies, which are described in Results.
Sociodemographic characteristics
We examine differences in perceptions by age (mean = 65.7 years), sex (56.2% male), male/husband's education (mean = 5.3 years), ethnicity (59.9% Fukienese, 17.1% Hakka, 24.0% Mainlander), and living arrangements (73.6% live with children), all measured at baseline (Table 2). Although elderly women outnumber elderly men in most populations, the gender ratio of elderly persons in Taiwan currently favors men because of the immigration of approximately 1 million Nationalist military and civilian supporters from Mainland China in 1949, many of whom were unmarried soldiers. There are three main ethnic groups in Taiwan: the Fukienese and the Hakka, who began their migration from Mainland China in the 1600s, and Mainlanders, who migrated to Taiwan at the end of World War II.
|
We include living arrangements because considerable social support derives from co-resident children and because the co-residence of elderly parents and adult children is an important cultural norm. We distinguish between respondents living with children and respondents not living with children.
Network characteristics
We examine three network characteristics measured at baseline: marital status, number of living children, and number of friends (the latter two are treated as continuous). We recognize that changes in the make-up of one's network can affect the perceived availability of support; however, with the exception of marital status, these network characteristics show little change over the 10-year period. Furthermore, previous research suggests that when a significant change in network membership occurs, other network members tend to fill in where needed (e.g., Stoller & Pugliesi, 1991). Finally, we ran additional models where these network characteristics were treated as time varying and allowed to directly affect the respective support measure from each wave. Results (not shown, but available upon request) demonstrated that changes in network composition had little effect on changes in perceptions. At baseline, 74% of respondents were married, most had more than two children, and, on average, six friends were seen at least once per week (see Table 2).
Social involvement
Respondents were asked to report whether and how often they participated in six social activities: socializing with friends/neighbors and/or memberships in a religious group, business association, political group, clan association, and/or an elderly organization. The mean number of social activities in which respondents had weekly participation is 1.5 (see Table 2), with roughly 20% not participating in any activity.
Health
We examine both the existence of disability (a dichotomous measure indicating difficulty with one or more of the following activities: crouching, climbing stairs, lifting 11–12 kg, walking 200–300 m, performing physical work around the house, and bathing) and counts of depressive symptoms (using an abbreviated, 10-item Center for Epidemiological Study–Depression Scale), because both are expected to have a direct bearing on perceptions of support. For example, the presence of physical impairment may concretize perceptions of sick-care support, whereas depressive symptoms may reduce perceptions of support. Both disability levels and depressive symptoms show an increase between 1989 and 1999 (see Table 2).
Analysis
For these analyses, we are interested in modeling stability and change—trajectories—of perceived social support over time among older Taiwanese adults. An ideal method for examining patterns over time at the individual level, while still capturing between-individual differences, is growth curve modeling. In a typical growth curve model, time-specific measures are assumed to be produced by two factors: the latent process of interest and random error. If the latent pattern is assumed to be linear across time, then two latent parameters define the process: an intercept and a slope. Thus, individuals' measures can be modeled as:
|
|
2, which may or may not vary across periods of measurement. In this equation, yit is the time t measure of y for individual i, 
i is an individual-specific intercept, and ßi is an individual-specific slope. The intercepts and slopes allow each individual to have a different linear pattern for support. The second level of the model specifies a distribution for these random effects:
|
|
0 and ß0, respectively, variances of 
and ß, and a covariance between the intercepts and slopes of 
ß. Beyond modeling the distribution of random effects, we often are interested in explaining between-individual variation in these random effects. This result can be obtained by representing 0 and ß0 as a linear function of an adjusted intercept and covariates of interest that differentiate individuals in a sample:
|
|
Growth models can generally be estimated within a structural equation modeling framework (see Meredith & Tisak, 1990; Willett & Sayer, 1994). In these analyses, we model trajectories of perceived support and the effect of covariates on these trajectories. Most covariates are included as predictors of the latent intercept and slope of the growth model, but the health measures are allowed to directly influence support at each time point. We first model individual trajectories and examine the between-individual variance present in the sample (Model 1). In our second model, we include variables thought to influence the latent baseline perceived support and growth in support over time. Figure 1 presents a structural equation based path diagram for the model we estimate.
|
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
) and item reliabilities (R2) for the indicators of support at each wave. The factor loadings were set for a linear growth specification, but we allowed the effect of the latent intercept on the wave 1 measure of support to be freely estimated. The wave 1 measure, an average of perceptions across network members for each dimension of support, is likely to produce more positive perceptions of support than items from the other three waves that ask whether network members in general were available to provide support. Indeed, the factor loading for the effect of the latent intercept on the wave 1 measure is > 40% larger ( = 1.42) than the fixed value of 1 for the other measures.
|
Based on the results of the second-level equation, the aggregate mean for baseline support is 2.04, and the estimated aggregate mean for the growth rate is 0.15. This result implies that the expected level of support at baseline is 2.04 units (x 1.42 = 2.90, given the factor loading estimate for the wave 1 measure), and this level is expected to increase by 0.15 unit per wave across the decade. Thus, the expected level of support across waves is 2.90 at wave 1, 2.19 at wave 2, 2.34 at wave 3, and 2.49 at wave 4. Compared with the descriptive statistics for perceived support in Table 2, the model very closely reproduces the observed means, despite the model fit indexes, which suggest that this base model does not fit the data particularly well (the incremental fit index [IFI] is <.9, the root mean squared error of approximation [RMSEA] is >.05, and the chi-square value is significant; see Bollen, 1989).
The second level of the model also estimates the variation in the sample around this aggregate pattern. The variance for baseline support is 0.06, indicating that individual-level baseline support should range from 2.4 to 3.4 units (2.9 ± 2[.06]
) for most of the sample (although 3.0 is the maximum for the scale). The estimated variance in growth in support across waves is 0.04, implying that growth in support ranges from –0.25 unit (a decline in support) to 0.55 unit per wave. The variances in both baseline and growth in support are highly statistically significant, indicating considerable between-individual heterogeneity in trajectories of support across waves.
Table 4 presents the results of the second model in which covariates were included to explain variation in growth patterns. The model appears to fit the data quite well, with an RMSEA below.05 and an IFI above.95. Similarly, although the model chi-square value is statistically significant, the ratio of the chi square to its degrees of freedom is well below 4, indicating a good fit (see Bollen, 1989).
|
In this model, a number of the covariates introduced to explain variation in the individual-level growth patterns significantly predict baseline differences in support. Perceived availability of support at baseline increases with age and education, married respondents have more positive perceptions than the unmarried, both Fukienese and Hakka individuals have more positive perceptions at baseline than Mainlanders, and living with one's children (but not the number of children) translates into more positive perceptions. Overall, 11% of the variance in baseline support is explained by the covariates. Slightly fewer of the covariates affect the latent growth rate. Education, being married, being Fukienese or Hakka, and the number of living children are all significantly associated with increases in the rate of growth in positive perceptions about the availability of support. The explained variance in the growth rate is comparable with that for latent baseline support: 11%.
In an effort to improve model fit and test the robustness of our findings, we examined a considerable number of alternate models and measurement strategies. In some models, we treated support at each wave as latent variables, with each of the three measures of support as indicators of latent support. We also tried alternate categorizations of the support measures. In some models, we excluded the first wave of measurement, given the large differences in measurement between wave 1 and the subsequent waves. In each of these models, we also altered our approach to handling the missing data. In the models we report, we used list-wise deletion of all missing data. In other models, we used hotdecking, mean/mode imputation, or full information maximum likelihood estimation on the entire sample, including decedents (see Allison, 2002). The results of all additional analyses (not reported, but available upon request) were virtually identical to those we report, with only slight exceptions. Specifically, in some models, fewer covariates predicted growth in perceived support. However, in a number of these models, convergence was questionable, and negative error variances were encountered.
Finally, we also attempted some nonlinear growth models, including allowing the coefficient for the effect of the latent growth factor on the fourth-wave support factor to be estimated and allowing the coefficients for the effect of the latent growth factor on both the third- and the fourth-wave support factors to vary. Parameter estimates for the effect of the latent growth factor on the wave support factors in these two models were not significantly different from those specifying linear growth in support. The substantive results were also identical to those we report here: Both indicate growth in perceived support across time, with identical covariates predicting between-individual variation in trajectories at statistically significant levels.
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Our results are consistent with many studies of perceived support in late life in Western populations. Several studies found that perceptions about social support improved over the study period (e.g., Field & Minkler, 1988; Martire et al., 1999). Studies in Western populations have also found that baseline support and change in support differ by socioeconomic status, marital status, ethnicity, and social involvement. Martire and colleagues (1999), however, found that variability in support is generally not explained by personal characteristics, which may be due in part to their relatively short study period (5 years).
We speculated that with the increasing pressures and demands on adult children's time that are associated with social and economic development, older Taiwanese adults may hold negative perceptions about the availability of social support. However, despite the changes occurring in Taiwanese society, older Taiwanese adults generally feel supported by their social networks, with the perceived availability of support increasing with age. There are several plausible explanations for these findings. First, support perceptions may be influenced more by personality traits that have been present throughout life than by current life circumstances (Sarason et al., 1990). For instance, people who have positive outlooks on life may view interactions with network members more positively than those who have negative life views.
Second, adjustments to, and expectations about, support arrangements may be continually adjusted such that the availability of support appears to be stable or improving over time (Martire et al., 1999). Our findings regarding disability are consistent with this explanation. The presence of disability is associated with less positive perceptions at earlier waves but over time becomes associated with more positive perceptions, suggesting that adjustments in care arrangements may have been made. The possibility that adjustments to expectations about caregiving lead to more positive perceptions over time is supported in part by a recent study by Cornman (1999). Results from the study, which used the 1993 wave of these data that included interviews with the respondents' adult children, show that both elderly parents and their adult children have a preference for independent living while elderly parents can support and care for themselves. However, when a parent needs support, as the result of health problems or the loss of a spouse, both parents and children agree that parents should live with an adult child.
Although our data did not allow us to examine the effects of received support, perceptions of support could also be driven in part by the amounts and types of support received, which in turn may be a function of the need for support. Generally, late-life declines in health are often accompanied by an increase in the need for, and provision of, various types of support. As received support becomes more apparent, respondents may perceive an increase in the availability of support. Additional research examining how social support is maintained, how expectations for (as opposed to perceptions about) support change, and how personality and changes in received support might affect support perceptions is needed to better understand why support perceptions tend to become more positive as older adults age.
This study is not without limitations. First, the wording of questions for the support measures is not consistent across all waves of data, making it difficult to distinguish between true change and change attributable to measurement. Even so, we provide evidence that our results are robust to a variety of different measurement strategies, indicating that our results are valid despite the shortcomings in measurement across waves. Second, owing to missing data issues, we drop the portion of our sample with the worst perceptions. Our analyses therefore most likely underestimate the degree of, and variability in, change in perceptions. The additional analyses following alternate strategies for handling missing data (discussed above) were consistent with this assumption. A related limitation is that our results are generalizable to only those who survived and provided complete data at all four waves. Finally, we examine perceptions about the availability of only two types of social support: emotional and instrumental support. Perceptions about other types of support (e.g., informational or financial) may show different patterns of change.
Overall, older Taiwanese adults hold positive perceptions about the availability of social support from friends and family, and perceptions become more positive as they age. These results suggest that policy efforts to promote and sustain informal sources of social support may be beneficial to older adults in Taiwan. However, it should be kept in mind that the generation of elders upon which this study is based tends to have large families from which to draw support. Future generations of elders will have fewer children. If the government continues to endorse the use of informal care, it may become increasingly important to promote policies that mobilize resources outside the family, such as neighborhood community resources (Chen & Lu, 2001).
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received for publication February 5, 2004. Accepted for publication June 15, 2004.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  B. A. Shaw, N. Krause, J. Liang, and J. Bennett Tracking Changes in Social Relations Throughout Late Life J. Gerontol. B. Psychol. Sci. Soc. Sci., March 1, 2007; 62(2): S90 - S99. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
