Economic strain and well-being in late life: Findings from an 18-year population-based longitudinal study of older Taiwanese adults

(1)

Economic strain and well-being in late life: findings

from an 18-year population-based longitudinal study

of older Taiwanese adults

Chi Chiao

1

_{, Li-Jen Weng}

2

_{, Amanda L. Botticello}

3 1

Institute of Health and Welfare Policy, College of Medicine, National Yang-Ming University, No. 155, Sec. 2, Li-Nong Street, 112, Taipei, Taiwan, ROC 2

Department of Psychology, College of Science, National Taiwan University, Taipei, Taiwan, ROC 3

Kessler Foundation Research Center and Department of Physical Medicine & Rehabilitation, UMDNJ-New Jersey Medical School, NJ, USA Address correspondence to Chi Chiao, E-mail: cchiao@ym.edu.tw

A B S T R AC T

Background This study estimates the concurrent and longitudinal effects of perceived economic strain and socioeconomic status (SES) on well-being of older adults in Taiwan.

Methods This study uses data from the Taiwan Longitudinal Study on Aging, a nationally representative sample (n ¼ 3602) of older adults aged 60 and above. Participants were interviewed and followed for 18 years. Individual well-being is measured by self-reported life

satisfaction, psychological distress and perceived health status. Generalized linear modeling with the generalized estimating equation estimates is used to predict the relationships between perceived economic strain, SES and well-being cross-sectionally and longitudinally, controlling for individual background characteristics, physical health and survival status.

Results Older adults who experienced economic strain had significantly poorer well-being in comparison to older adults without strain, both cross-sectionally and longitudinally, controlling for SES and other covariates. In contrast, SES indicators did not consistently predict well-being in the cross-sectional and longitudinal analyses.

Conclusions These findings suggest a strong, cumulative, negative effect of perceived economic strain on well-being among older adults. Health-care initiatives aiming at promoting well-being among older adults should consider the impact of economic strain, which may increase at the end of the life course and threaten health and functioning.

Keywords longitudinal analysis, older adults, perceived economic strain, socioeconomic status, Taiwan, well-being and health

Introduction

In Taiwan, the proportion of persons over the age of 65 has increased steadily in recent decades. However, well-being does not always accompany increased longevity, and there-fore, public health concern for the maintenance of good health and quality of life among this population has also increased. Of particular concern is how socioeconomic status (SES) and economic strain in later life may contribute to differences in health and well-being among older adults.1,2

Well-being is the self-evaluation of individual lives3,4 assessed by life satisfaction, psychological distress and per-ceived health.4–6 Previous studies have demonstrated that

poor well-being negatively inﬂuences physical health, functioning and quality of life for individuals and their families7–10 and even contributes to the increased risk for mortality.11,12 In studies of Western populations, poor health and well-being is consistently associated with low income13–15and material factors such as individual material and resource deprivation to poverty in the local commu-nity.15–17 Insights provided by the stress process model18,19

Chi Chiao, Associate Professor Li-Jen Weng, Professor of Psychology

Amanda L. Botticello, Research Scientist and Assistant Professor

Journal of Public Health | pp. 1 – 11 | doi:10.1093/pubmed/fdr069

at University of California, Los Angeles on September 13, 2011

jpubhealth.oxfordjournals.org

(2)

indicated that low-status groups are more vulnerable to poor health and well-being than others based on disadvantages inherent to their social location. These disadvantages, known as ‘stressors’, are socially patterned disproportionately experi-enced by members of low-status groups. For example, low SES persons may be at increased risk for poor well-being because they more frequently experience economic strain. Research examining the effect of financial difficulties on health and well-being supports the idea that the subjective experience of ‘economic strain’ is related to, but independent of, SES.20–25 Kahn and Pearlin26 provides evidence that financial strain has a significant influence on a range of health outcomes. Using retrospective data from 1167 older adults, their findings suggest sustained financial hardship that has a cumulative negative effect on health outcomes later in life, independent of income and other SES indicators.

Economic strain, as stressor,19,27could play an important, albeit different, role in the experience of health and well-being among older adults in Taiwan.28–31The social values of Confucius such as in Taiwan emphasize the adequacy of how basic needs are met rather than overall levels of income and wealth. However, the concept of strain inherent suggests difﬁculty in the ability to meet basic needs. The goal of this study is to investigate both the concurrent and cumulative effects of perceived economic strain on well-being using prospective, longitudinal data collected from a nationally representative sample of older adults, in Taiwan, controlling for economic resources such as income and home owner-ship, SES, sociodemographic characteristics, physical health status, social functioning and survival status in a multivariate model.

Methods Data and sample

The data for this analysis are from the Taiwan Longitudinal Study on Aging (TLSA), a nationally representative survey designed to study the impact of socioeconomic development on the physical and emotional health of older adults in Taiwan. Data were collected by the Bureau of Health Promotion of the Taiwan Public Health Department from 1989 to 2007. The baseline sample was derived using a three-stage sampling framework. A total of 4049 older adults were ﬁrst interviewed in-person in 1989, with four follow-up interviews conducted between 1996 and 2007 for surviving participants. Information on TLSA can be found at www.bhp.doh.gov.tw and further details on the TLSA sampling and design are reported elsewhere.32,33 For this study, the analytic sample was restricted to the baseline, with

complete data on SES, economic strain and indicators of well-being. The sample ﬂow is shown in Fig.1.

Measures

Well-being

Well-being was assessed by three self-reported measures on life satisfaction, psychological distress and health status. The Life Satisfaction Index (LSI) is a 10-item scale adaptation of the original 20-item LSI.34 LSI items include ‘Has your life been better than most people’s lives?’ and ‘Are you satisﬁed with your life?’ and were rated yes or no. Items were reverse scored when necessary and summed so that higher LSI scores corresponded with better life satisfaction. Psychological distress was measured by a 10-item version of the Center of Epidemiological Studies-Depression (CES-D) scale.35Each item was rated on a four-point scale, indicating the frequency of experiencing each symptom in the past week. Responses were reversely scored when necessary so that higher scores represented higher levels of depressive symptoms. Previous analysis of the TSLA data demon-strated two distinct domains in the 10-item CES-D: negative affect and lack of positive affect.32,36 Items were summed within each of the two domains.24 Health status was measured by the SF-36 item which asked individuals to rate their health as ‘poor,’ ‘fair,’ ‘good,’ ‘very good’ or ‘excellent’ on a scale of 1 – 5.37

Economic strain

Economic strain was a time-varying covariate that was assessed whether individuals had enough living expenses. The responses ranged from more than enough to very insuf-ficient but the points of the rating scales used varied across TLSA waves. A dichotomous measure of economic strain was created for the analysis 1 ¼ ‘insufficient’ (i.e. strain) and 0 ¼ ‘sufficient’ (i.e. no strain).

The SES measures

The SES measures included time-varying indicators of econ-omic resources (i.e. household income and home owner-ship), work status and a baseline measure of education. Household monthly income categories varied across waves; for the analysis, quartiles were used to for consistency in the ranking of household income over time. Home ownership at each wave was coded 1 if the respondent’s owned their current residence and 0 if not. Work status was categorized as unemployed, assisting family, full- or part-time work and retired for each wave. Education consists of four categories (illiterate, incomplete primary education, completed primary education and high school or above).

2 J O U R N A L O F P U B L I C H E A LT H

(3)

Controls

The analyses were adjusted for several time-varying health indicators. Physical disability was measured by the Instrumental Activities of Daily Living (IADL) and Nagi scales.38 The IADL scale assessed if respondents had diffi-culty with shopping, managing money, using transportation, doing heavy housework or using a telephone. The Nagi scale assessed difficulty with crouching, raising hands over-head, grasping small objects with their fingers, lifting heavy objects, walking 200 – 300 m or climbing stairs. These two physical disability measures were dichotomized for the

analyses (1 ¼ no limitations, 0 ¼ at least one limitation) based on approaches used in prior studies.39–43 Respondents indicated a history of cardiovascular disease (CVD) if: (i) a doctor ever told respondents they had a heart attack, coronary heart disease or other heart problems; (ii) they had a stroke; and (iii) they have diabetes. Endorsing any of these conditions was coded as 1 and otherwise 0.44

Several family and demographic characteristics were examined including a time-varying measure of marital status (married, widowed and other) and the number of children reported at baseline. Age and gender were measured by the

Participants n =3954 (89.62%) Dropouts n=363 Aborigine n=95 (10.38%) Analytical sample: n=3602 Exclusion: Proxy n=211; Missing n=141 Participants n =2624 (66.36%) LFU n =309 (7.81%) Deaths n =1021 (25.83%) Analytical sample: n=2143 Exclusion: Proxy n=254; Missing n=277 Total sample in 1989 N=4412 Participants n =2267 (77.29%) LFU n =245 (8.36%) Deaths n =421 (14.35%) Analytical sample: n=1874 Exclusion: Proxy n=235; Missing n=158 Participants n =1711 (68.11%) LFU n =167 (6.65%) Deaths n =634 (25.24%) Participants n =1248 (66.45%) LFU n =131 (6.98%) Deaths n =499 (26.57%) Analytical sample n=829 Exclusion: Proxy n=270 Missing n=149 1989 1996 1999 2003 2007 Analytical sample: n=1320 Exclusion: Proxy n=264; Missing n=127

Fig. 1 Participants in serial surveys in the TLSA from 1989 to 2007. LFU is lost to follow-up mainly due to moving and rejection to be interviewed. Missing is incomplete data on major constructs.

(4)

standard survey questions. Ethnic groups reﬂect post-industrial patterns of Taiwanese migration (Fukianese, Hakka and Mainlander).

Statistical analysis

Parallel linear regression models were used to estimate cross-sectional and longitudinal associations of economic strain and SES with each measure of well-being (i.e. life satisfaction, psychological distress and self rated health). Generalized esti-mating equations (GEE) with robust standard error estimates were used to take into account within-subject correlations during the 18-year follow-up period.45–47Estimated associ-ations were described in the form of adjusted coefﬁcients and STATA version 9.048 was used to manage and analyze data. Data from the 1989 baseline, 1996, 1999, 2003 and 2007 follow-up interviews were assessed simultaneously in all analyses.

The initial analyses examined the cross-sectional associ-ation between economic strain and each outcome well-being. Each model was then adjusted for the health and demo-graphic covariates and whether the respondent was deceased in 2007, and finally for SES. The baseline measures of well-being were included to reduce unobserved heterogeneity. Given the considerable changes to family structure, SES and physical health at the end of the life course, we assessed the robustness of these associations over time. The longitudinal (transition) models included measures of well-being from previous waves to examine associations between economic strain and subsequent well-being.49Product terms represent-ing interactions between economic strain and SES and gender were tested and excluded from the final models because of statistical non-significance.

Results

Sample characteristics

The characteristics of the analytic sample at baseline (n ¼ 3602) are shown in Table 1. The sample was approxi-mately equally distributed by gender, the majority (46.7%) was between the ages of 60 and older, and persons of Fukianese background comprised the largest ethnic group (61.9%). Overall, this sample indicated a notable amount of SES disadvantages as the majority of the sample was illiterate (40.0% and did not work (39.5%). However, over two-thirds of the respondents owned their home. Approximately 40% had experienced some physical limitations and 28% some form of CVD. Family support was evident at baseline, with the majority of respondents reporting being married (65.5%) and having large families (70.6%).

Economic strain was common, with one out of every four to five persons indicating that they had insufficient resources in the past year. However, reports of economic strain varied significantly by individual characteristics, as shown in Table 1. Economic strain was common among persons with low education (x2_ð3Þ¼ 108.30; P , 0.001) and low income (x2

ð3Þ¼ 171.76; P , 0.001). Strain was also more

likely among the Fukianese (x2_ð2Þ¼ 7.96; P ¼ 0.019) and those persons with fewer children (x2

ð2Þ¼ 19.51; P , 0.001).

Economic strain was also more likely among persons report-ing CVD and physical disability. And ﬁnally, older adults who reported economic strain at baseline were also more likely to be deceased before 2007 (x2

ð1Þ¼ 15.69; P , 0.001).

Cross-sectional associations

Table 2 presents multivariate regression models testing the cross-sectional associations between economic strain and each of the three indicators of well-being. Model 1 for each outcome adjusts for the effects of the sociodemographic characteristics, family background and concurrent physical health. We found that older adults who reported economic strain perceived lower levels of life satisfaction (b ¼ 21.21; P ,0.001) and health status (b ¼ 20.24; P , 0.001), and higher levels of psychological distress on both domains (b ¼ 1.52, P , 0.001 for negative affect; b ¼ 0.67, P , 0.001 for lack of positive affect) in comparison to older adults with no strain. Model 2 for each outcome adjusts the relationships between economic strain and well-being for the effects of SES, as measured by household income, home ownership, education and work status. The relationship between economic strain and life satisfaction remained largely independent of SES, although being in higher edu-cation and income strata modestly improved life satisfaction. The strain effect on negative affect did not have a substan-tial change after inclusion of the SES variables. Relative to the effect observed for economic strain, the associations between current SES and well-being were not as large, with the exception of education which had a signiﬁcant positive effect in predicting life satisfaction and perceived health. Notable among the control variables, all outcomes decreased signiﬁcantly with the presence of CVD and limitation in physical functioning.

Longitudinal associations

Table3presents multivariate regression results for longitudi-nal (transition) models, which are adjusted for both static (i.e. gender, ethnicity and education at baseline) and the time-speciﬁc covariates (i.e. SES, physical health and prior levels of strain and well-being). As in the previous analyses, 4 J O U R N A L O F P U B L I C H E A LT H

(5)

we observe that persons perceiving economic strain have lower life satisfaction over time in Model 1 (b ¼ 20.72; P ,0.001) and that this significant association remains even when the changing SES circumstances in the sample are taken into account (b ¼ 20.65; P , 0.001). A similar pattern emerges for both domains of depressive symptoms; the transition effect of economic strain was also found to be significantly associated with increasing subsequent psycho-logical distress on both domains (b ¼ 0.85, P , 0.001 for negative affect; b ¼ 0.37, P , 0.001 for lack of positive affect) and decreasing subsequent perceived health (b ¼ 20.14; P , 0.001) after adjusting for prior well-being and the other covariates. Although the magnitude of these effects was smaller relative to the cross-sectional analysis due to the inclusion of prior measures of strain and each outcome, overall, these findings suggest that the lasting effects of economic strain accumulate over time.

A sensitivity analysis was performed to assess the effect of loss to follow-up over time. Attrition rates in TSLA are similar longitudinal surveys of older adults such as the HRS/AHEAD study in the USA where attrition is largely due to death.50 We assessed differences in SES, individual background and physical health between continuing partici-pants and dropouts across waves (the results not tabled). Due to a large number of male migrants from China in 1949, males comprised 57.94% of the baseline analytical sample. The proportion of males in the sample relative to females continued to decline over the course of the data col-lection: 56.23% in 1996; 55.66% in 1999; 54.85% in 2003; and 54.16% in 2007. The analyses indicated that the pro-portions of older adults with economic strain in the analytic samples slightly increased from 22.02% in 1989 and 25.66% Table 1 Baseline (1989) sample characteristics of the TLSA and

bivariate differences in self-reported economic strain (N ¼ 3602) Total Economic strain

Yes No Gender (%) Female 42.06 24.49 75.51 Male 57.94 20.22 79.78 Age (%) 60 – 64 years 37.62 21.85 78.15 65 – 74 years 46.70 21.94 78.06 75 – 84 years 14.30 22.91 77.09 85 years 1.39 20.00 80.00 Ethnicity (%) Fukianese 61.94 23.53 76.47 Hakka 14.85 20.00 80.00 Mainlander 23.21 19.26 80.74* Education (%) Illiterate 40.01 28.77 71.23

Incomplete primary education 15.87 26.09 73.91 Completed primary education 24.30 17.39 82.61 High-school graduate and above 19.82 10.66 89.34*** Work status (%)

No work 39.47 22.10 77.90

Assisting family 24.94 23.16 76.84

Full- or part-time work 21.72 21.87 78.13

Retired 13.86 20.04 79.96

Monthly household income (%)

The lowest 32.28 28.92 71.08

The second quartile 27.87 28.46 71.54 The third quartile 25.67 16.21 83.79 The highest quartile 14.17 3.77 96.23*** Home ownership (%) Owned 63.86 21.00 79.00 None 36.14 23.63 76.37 CVD related disease (%) Yes 28.73 28.03 71.97 No 71.27 19.59 80.41*** IADL (%)

Any of five selected limited functions 40.14 30.50 69.50

None 59.86 16.33 83.67***

Nagi (%)

Any of six selected limited functions 43.36 28.96 71.04

None 56.64 16.72 83.28*** Marital status (%) Married 65.53 21.24 78.76 Widowed 27.42 23.20 76.80 Separated/divorced/never married 07.06 24.80 75.20 Number of children (%) 0 – 3 29.37 26.64 73.36 Continued Table 1 Continued

Total Economic strain

Yes No 4 – 5 32.94 20.52 79.48 6 37.69 19.39 80.61*** Survival status in 2007 (%) Deceased in 2007 63.24 24.10 75.90 Alive until 2007 35.90 18.33 81.67 Lost to follow-up 0.86 22.58 77.42***

x2test was used to test for group differences in economic strain. *P , 0.05.

**P , 0.01. ***P , 0.001.

(6)

Table 2 Adjusted cross-sectional associations of subjective economic strain and economic status with life satisfaction, psychological distress and self-rated health, TLSA 1989 –2007

Life satisfaction Depressive symptomatology Self-rated health

Negative affect Lack of positive affect

Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2

Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Economic strain

Concurrent economic strain (ref ¼ no)

21.21*** (21.31, 21.10) 21.16*** (21.27, 21.05) 1.52*** (1.33, 1.71) 1.55*** (1.35, 1.74) 0.67*** (0.57, 0.78) 0.67*** (0.56, 0.77) 20.24*** (20.28, 20.20) 20.23*** (20.27, 20.19) Socioeconomic status

Household income (ref ¼ lowest quartile)

The second quartile 0.06

(20.06, 0.17) 20.20 (20.41, 0.01) 0.02 (20.09, 0.14) 0.03 (20.02, 0.08)

The third quartile 0.05

(20.07, 0.18) 0.08 (20.15, 0.31) 20.01 (20.13, 0.11) 0.01 (20.03, 0.06)

The highest quartile 0.16*

(0.02, 0.30) 0.18 (20.08, 0.44) 0.05 (20.09, 0.19) 20.001 (20.06, 0.05) Home ownership (ref ¼ no) 0.05 (20.05, 0.14) 0.14 (20.03, 0.31) 0.02 (20.07, 0.11) 20.02 (20.06, 0.01) Education (ref ¼ illiterate)

Incomplete primary education 0.11 (20.04, 0.27) 20.05 (20.33, 0.22) 20.04 (20.17, 0.09) 20.03 (20.03, 0.08) Completed primary education 0.27*** (0.14, 0.41) 20.20 (20.44, 0.05) 20.11 (20.22, 0.01) 20.05 (20.004, 0.10) High-school graduate and above 0.30*** (0.14, 0.47) 20.15 (20.44, 0.15) 20.10 (20.24, 0.04) 0.12*** (0.05, 0.18) Employment status (ref ¼ unemployed)

Assisting family 0.04 (20.07, 0.16) 20.38*** (20.59, 20.17) 20.04 (20.15, 0.08) 0.04 (20.001, 0.09)

Full- or part-time work 0.07

(20.08, 0.22) 20.21 (20.48, 0.06) 20.04 (20.18, 0.11) 0.10** (0.04, 0.16) Retired 0.01 (20.13, 0.16) 20.14 (20.39, 0.12) 20.01 (20.14, 0.12) 0.02 (20.04, 0.07) Physical health control

CVD (ref ¼ no) 20.19*** (20.28, 20.09) 20.20*** (20.30, 20.10) 0.63*** (0.45, 0.80) 0.63*** (0.45, 0.80) 0.09* (0.0003, 0.18) 0.09* (0.003, 0.19) 20.29*** (20.33, 20.25) 20.29*** (20.33, 20.25) 6 JOURN A L OF P UBLIC H EAL T H

(7)

in 1996 to 27.59% in 1999 and then turned to decline from 25.08% in 2003 to 22.80% in 2007.

Discussion

Main finding of this study

This study supports previous studies demonstrating a robust cross-sectional association between economic strain and well-being.8,21,24,31 Our analyses further demonstrate that economic strain predicts well-being over and above SES both cross-sectionally and longitudinally. The multivariate longitudinal analysis indicates that the harmful effect of economic strain accrues over time, suggesting that economic strain is cumulatively associated with declining well-being in later life. This association is consistent across three aspects of well-being—life satisfaction, depressive symptoms and perceived health status—suggesting that economic strain broadly affects health and quality of life at older ages.

What is already known on this topic

Subjective well-being has been shown to differ by indicators of social stratiﬁcation.13–17 Disparities by SES factors are frequently a focus of research related to aging and well-being in industrialized countries.1 Comparatively less attention has been concentrated on social group differences in well-being among the older members of the population in Asian countries. The stress process literature suggested perceived socioeconomic disadvantage to be associated with well-being.22–28This analysis suggests that the negative effects of economic strain may accumulate and diminish well-being among older adults in Asian countries.

What this study adds

To our knowledge, this is the first study that examines the longitudinal relationship between economic strain and well-being for older adults in Taiwan. Our results suggest that economic strain is more predictive of well-being than income. That is, perceived difficulty in meeting one’s basic needs compromises well-being in later life, suggesting that quality-of-life gains among older adults have not kept pace with gains in longevity. In contrast, a robust, positive cross-sectional association was observed between higher levels of education and life satisfaction, particularly among high-school graduates versus those with incomplete primary education. This suggests that education is the most potent SES resource against declining well-being in older adult-hood. However, education level does little to mitigate the strains imposed on well-being by financial hardship. These findings may be interpreted in the hedonic adaptation3–5

IADL (r ef ¼ no limita tion) 2 0.44*** (2 0.55, 2 0.34) 2 0.41*** (2 0.52, 2 0.30) 1.15*** (0.96, 1.35) 1.12*** (0.92, 1.31) 0.30*** (0.19, 0.40) 0.29*** (0.18, 0.39) 2 0.28*** (2 0.33, 2 0.24) 2 0.27*** (2 0.31, 2 0.23) Nagi (r ef ¼ no limita tion) 2 0.28*** (2 0.39, 2 0.17) 2 0.28*** (2 0.39, 2 0.17) 0.75*** (0.55, 0.95) 0.74*** (0.54, 0.93) 0.20*** (0.09, 0.31) 0.20*** (0.09, 0.31) 2 0.33*** (2 0.38, 2 0.29) 2 0.33*** (2 0.37, 2 0.28) C omparison of Model 2 to Model 1 x 2 72.26*** 52.61*** 8.22 68.90*** Degr ee of fr eedom 10 10 10 10 All models adjus ted for age, gender , ethnicity , marital sta tus, number of childr en, survival sta tus, w av es of intervie w , baseline values of life sa tisfa ction, ps ychological dis tr ess and self -r ated health. * P , 0.05. ** P , 0.01. *** P , 0.001.

(8)

Table 3 Adjusted longitudinal (transition) associations of previous subjective economic strain and economic status with subsequent life satisfaction, psychological distress and self-rated health, TLSA 1989 – 2007

Life satisfaction Depressive symptomatology Self-rated Health

Negative affect Lack of positive affect

Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2

Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Coeff. (95% CI) Economic strain

Previous economic strain (ref ¼ no) 20.72*** (20.89, 20.55) 20.65*** (20.82, 20.48) 0.87** (0.56, 1.17) 0.85*** (0.54, 1.16) 0.41*** (0.26, 0.56) 0.37*** (0.21, 0.52) 20.16*** (20.22, 20.10) 20.14*** (20.21, 20.08) Socioeconomic status

Previous household income (ref ¼ lowest quartile)

The second quartile 0.12

(20.06, 0.29) 20.01 (20.34, 0.32) 20.05 (20.21, 0.12) 0.0002 (20.07, 0.07)

The third quartile 0.05

(20.14, 0.24) 0.09 (20.26, 0.44) 20.05 (20.22, 0.13) 20.03 (20.10, 0.04)

The highest quartile 20.07

(20.27, 0.14) 0.17 (20.21, 0.55) 20.11 (20.30, 0.08) 0.003 (20.08, 0.08)

Previous home ownership (ref ¼ no) 0.05

(20.10, 0.18) 0.18 (20.07, 0.43) 0.03 (20.10, 0.16) 20.04 (20.08, 0.01) Education (ref ¼ illiterate)

Incomplete primary education 0.18

(20.03, 0.39) 20.08 (20.46, 0.29) 20.17 (20.37, 0.03) 0.05 (20.03, 0.13)

Completed primary education 0.61***

(0.43, 0.79) 20.31 (20.63, 0.01) 20.24** (20.41, 20.07) 0.07* (0.003, 0.14)

High-school graduate and above 0.77***

(0.55, 0.99) 20.41* (20.80, 20.02) 20.37*** (20.57, 20.16) 0.24*** (0.16, 0.32) Employment status (ref ¼ unemployed)

Assisting family 0.02 (20.15, 0.19) 20.18 (20.49, 0.13) 0.02 (20.14, 0.18) 0.001 (20.06, 0.07)

Full- or part-time work 0.10

(20.10, 0.31) 20.11 (20.49, 0.27) 20.07 (20.26, 0.12) 0.001 (20.08, 0.08) Retired 0.06 (20.15, 0.27) 20.17 (20.55, 0.20) 20.003 (20.20, 0.19) 0.04 (20.04, 0.12) Physical health controls

CVD (ref ¼ no) 20.10 (20.24, 0.03) 20.13 (20.27, 0.01) 0.54*** (0.29, 0.79) 0.55*** (0.30, 0.80) 0.06 (20.07, 0.19) 0.08 (20.05, 0.21) 20.30*** (20.36, 20.25) 20.32*** (20.36, 20.26) 8 JOURN A L OF P UBLIC H EAL T H

(9)

and homeostasis51in which individuals may quickly adapt to static factors like their education level, but may less readily adapt to not having enough money on a day-to-day basis.

Taken together, our results suggest that economic strain independently contributes to poorer well-being for older adults in Taiwan. This analysis represents an important ﬁrst step in examining multiple aspects of SES and poverty among non-Western industrialized countries. The industrial-ization of Taiwan in the last decades has dramatically changed labor market structure with large numbers of people seeking highly paid positions such as salesperson. As a result, income is often the focus of understanding health disparities and well-being among older adults in planning social and health initiatives. Our analyses suggested that the experience of economic strain rather than income threatens well-being and that the negative effects of strain accumulate. Therefore, future initiatives aimed at eliminating health dis-parities among older adults will beneﬁt from considering economic strain as well as economic resources as risk factors for diminished quality of life.52

A particular strength of this analysis was the use of longi-tudinal data, which allowed us to examine change in well-being as a function of economic strain at a prior time and other dynamic indicators. While this analytical strategy intends to establish a causal link of economic strain to well-being, reciprocal causation could arise from a process of multiple system strain. That is, decline in well-being leads to economic strain, which in turn accelerates decline of well-being. In a separate analysis, we further restricted the sample to those who reported their health as fair and better or those whose baseline well-being was above 50 percentile. Economic hardship was still a signiﬁcant predictor of subsequent well-being decline among the selected sample.

Limitations of this study

Although our research provides new information via a longi-tudinal analysis regarding economic strain, SES and well-being for older adults, this work is not without constraints. First, well-being as measured by the LSI, CES-D and self-rated health is subject to recall bias. Second, the assessment of other important covariates to predict relationships between economic strain and well-being is limited by the use of existing data in the TLSA, which lacks measures of related constructs such as ﬁnancial support that ameliorated strain. Third, health measures are self-reported and the use of objective information, such as medical records, is pre-cluded by the use of an existing data set. Fourth, this analy-sis covered an 18-year period, but it was obtained on a series of combined periods. The problem with measuring

IADL (r ef ¼ no limita tion) 2 0.56*** (2 0.72, 2 0.40) 2 0.52*** (2 0.68, 2 0.36) 1.52*** (1.22, 1.82) 1.50*** (1.20, 1.80) 0.44*** (0.29, 0.59) 0.42*** (0.27, 0.57) 2 0.34*** (2 0.40, 2 0.28) 2 0.34*** (2 0.40, 2 0.28) Nagi (r ef ¼ no limita tion) 2 0.37*** (2 0.53, 2 0.20) 2 0.37*** (2 0.53, 2 0.20) 1.17*** (0.86, 1.47) 1.17** (0.86, 1.47) 0.27** (0.12, 0.43) 0.27** (0.11, 0.42) 2 0.44*** (2 0.51, 2 0.38) 2 0.44*** (2 0.50, 2 0.38) C omparison of Model 2 to Model 1 x 2 44.67*** 6.72 13.70 52.98*** Degr ee of fr eedom 10 10 10 10 All models adjus ted for age, gender , ethnicity , marital sta tus, number of childr en, survival sta tus, w av es of intervie w , pr evious values of life sa tisfa ction, ps ychological dis tr ess and self-r a ted health. * P , 0.05. ** P , 0.01. *** P , 0.001.

(10)

well-being decline over longer periods of time is that there is greater risk of attrition due to mortality, particularly for older adults. We included the variable of survival status in the analysis and found that survivors had signiﬁcant higher levels for all three well-being measures.

Funding

This research work was supported by the National Science Council in Taiwan under grant NSC97-2314-B-010-047-MY3.

References

1 Adler NE, Boyce T, Chesney MA et al. Socioeconomic inequalities in health: no easy solution. JAMA 1993;269(24):3140 – 5.

2 Dolan P, Metcalfe R, Munro V et al. Valuing lives and life years: anomalies, implications, and an alternative. Health Econ Policy Law 2008;3(Pt 3):277 – 300.

3 Diener E. Subjective well-being: the science of happiness and a pro-posal for a national index. Am Psychol 2000;55:34 – 43.

4 Diener E, Suh EM, Lucas RE et al. Subjective well-being: three decades of progress. Psychol Bull 1999;125(2):276 – 302.

5 Campbell A. Subjective measures of well-being. Am Psychol 1976;31(2):117 – 124.

6 Morrow-Howell N, Hinterlong J, Rozario PA et al. Effects of

volun-teering on the well-being of older adults. J Gerontol

2003;58(3):S137 – 45.

7 Collins AL, Goldman N, Rodriguez G. Is positive well-being pro-tective of mobility limitations among older adults? J Gerontol 2008;63(6):P321 – 7.

8 George LK. Economic status and subjective well-being: a review of the literature and an agenda for future research. In: Culter NE, Gregg DW, Lawton MP (eds). Aging, Money, and Life Satisfaction: Aspects of Financial Gerontology. New York: Springer Pub. Co., 1992,69 – 99.

9 Rowan PJ, Al-Jurdi R, Tavakoli-Tabasi S et al. Physical and psycho-social contributors to quality of life in veterans with hepatitis C not on antiviral therapy. J Clin Gastroenterol 2005;39(8):731 – 6.

10 Steffens DC, Otey E, Alexopoulos GS et al. Perspectives on depression, mild cognitive impairment, and cognitive decline. Arch Gen Psychiatry2006;63(2):130 – 8.

11 Idler EL, Kasl S. Health perceptions and survival: do global evalu-ations of health status really predict mortality? J Gerontol 1991;46(2):S55 – 65.

12 Mossey JM, Shapiro E. Self-rated health: a predictor of mortality among the elderly. Am J Public Health 1982;72(8):800 – 8.

13 Diener E, Biswas-Diener R. Will money increase subjective well-being? Soc Indicators Res 2002;57(2):119 – 69.

14 Pappas G, Queen S, Hadden W et al. The increasing disparity in mortality between socioeconomic groups in the United States, 1960 and 1986. N Engl J Med 1993;329(2):103 – 9.

15 Dolan P, Peasgooda T, Whiteb M. Do we really know what makes us happy: a review of the economic literature on the factors associ-ated with subjective well-being. J Econ Psychol 2008;29(1):94 – 122. 16 Elstad JI. The psycho-social perspective on social inequalities in

health. Sociol Health Illness 1998;20(5):598 – 618.

17 Graham H (eds). Understanding Health Inequalities. Buckingham: Open University Press, 2001.

18 Pearlin LI. The sociological study of stress. J Health Soc Behav 1989;30:241 – 56.

19 Pearlin LI, Schieman S, Fazio EM et al. Stress, health, and the life course: some conceptual perspectives. J Health Soc Behav 2005;46(2):205 – 19.

20 Chi I, Chou K-L. Financial strain and depressive symptoms among Hong Kong Chinese elderly: a longitudinal study. J Gerontol Soc Work2000;32(4):41 – 60.

21 Cheng YH, Chi I, Boey KW et al. Self-rated economic condition and the health of elderly persons in Hong Kong. Soc Sci Med 2002;55(8):1415 – 24.

22 Krause N, Jay G, Liang J. Financial strain and psychological well-being among the American and Japanese elderly. Psychol Aging 1991;6(2):170 – 81.

23 Li Y, Aranda MP, Chi I. Health and life satisfaction of ethnic min-ority older adults in mainland China: effects of ﬁnancial strain. Int J Aging Hum Dev2007;64(4):361 – 79.

24 Ferraro KF, Su Y. Financial strain, social relations, and psychological distress among older people: a cross-cultural analysis. J Gerontol 1999;54(1):S3 – 15.

25 Krause N. Chronic ﬁnancial strain, social support, and depressive symptoms among older adults. Psychol Aging 1987;2(2):185 – 92. 26 Kahn JR, Pearlin LI. Financial strain over the life course and health

among older adults. J Health Soc Behav 2006;47(1):17 – 31.

27 Lazarus RS, Folkman S. Stress, Appraisal, and Coping. New York: Springer, 1984.

28 Ferrie JE, Martikainen P, Shipley MJ et al. Self-reported economic difﬁculties and coronary events in men: evidence from the Whitehall II study. Intern J Epidemiol 2005;34:640 – 8.

29 Rosengren A, Hawken S, Ounpuu S et al. Association of psychoso-cial risk factors with risk of acute myocardial infarction in 11119 cases and 13648 controls from 52 countries (the INTERHEART study): case – control study. Lancet 2004;364:953 – 6.

30 Lantz PM, House JS, Mero RP et al. Stress, life events, and socioe-conomic disparities in health: results from the Americans’ changing lives study. J Health Soc Behav 2005;46:274 – 88.

31 Castro Ad, GC G, DT T. Examining alternative measures of social disadvantage among Asian Americans: the relevance of economic opportunity, subjective social status, and ﬁnancial strain for health. J Immigr Minor Health2009;0:1557 – 912.

32 Chiao C, Weng LJ, Botticello A. Do older adults become more depressed with age in Taiwan? The role of social position and Birth cohort. J Epidemiol Community Health 2009;63:625 – 32.

33 Glei DA, Landau DA, Goldman N et al. Participating in social activities helps preserve cognitive function: an analysis of a longi-tudinal, population-based study of the elderly. Intern J Epidemiol 2005;34(4):864 – 71.

10 J O U R N A L O F P U B L I C H E A LT H

(11)

34 Neugarten BL, Havighurst RJ, Tobin SS. The measurement of life satisfaction. J Gerontol 1961;16:134 – 43.

35 Radloff LS. The CES-D Scale: a self-report depression scale for

research in the general population. Appl Psychol Meas

1977;1(3):385 – 401.

36 Lee K-L, Ou Y-L, Chen S-H et al. The psychometric properties of a short form of the CES-D used in the Taiwan longitudinal study on aging. J Formos Ment Health 2009;22(4):383 – 410.

37 McHorney CA, Ware JE Jr, Raczek AE. The MOS 36-Item short-form health survey (SF-36): II. Psychometric and clinical tests of validity in measuring physical and mental health constructs. Med Care1993;31:247 – 63.

38 Nagi SZ. Some conceptual issues in disability and rehabilitation. In: Sussman MB (ed.). Sociology and Rehabilitation. Washington, DC: American Sociological Association, 1965,100 – 13.

39 Freedman VA, Crimmins E, Schoeni RF et al. Resolving inconsisten-cies in trends in old-age disability: report from a technical working group. Demography 2004;41(3):417 – 41.

40 Manton KG, Corder LS, Stallard E. Estimates of change in chronic disability and institutional incidence and prevalence rates in the U.S. elderly population from the 1982, 1984, and 1989 National Long Term Care Survey. J Gerontol: Soc Sci 1993;48(4):S153 – 66.

41 Ofstedal M, Zimmer Z, Hermalin A et al. Short-term trends in functional limitation and disability among older Asians: a compari-son of ﬁve Asian settings. J Cross-Cult Gerontol 2007;22(3):243 – 61. 42 Pavalko EK, Mossakowski KN, Hamilton VJ. Does perceived

dis-crimination affect health? Longitudinal relationships between work discrimination and women’s physical and emotional health. J Health Soc Behav2003;44(1):18 – 33.

43 Schoeni RF, Liang J, Bennett J et al. Trends in old-age functioning and disability in Japan, 1993 – 2002. Popul Stud 2006;60(1):39 – 53.

44 Expert Panel on Detection, Evaluation, and Treatment of High Blood Pressure in Adults. Executive summary of the third report of the National Cholesterol Education Program (NCEP) expert panel on detection, evaluation, and treatment of high blood

cholesterol in adults (Adult Treatment Panel III). JAMA

2001;285(19):2486 – 97.

45 Liang KY, Zeger SL. Regression analysis for correlated data. Annu Rev Public Health1993;14:43 – 68.

46 Rabe-Hesketh S, Everitt BS. A Handbook of Statistical Analyses: Using Stata, 4th edn. USA: Chapman & Hall/CRC, 2006.

47 Zeger SL, Liang KY, Albert PS. Models for longitudinal data: a generalized estimating equation approach. Biometrics 1988;44(4): 1049 – 60.

48 StataCorp. Stata Statistical Software: Release 9.0. College Station, TX: Stata Corporation, 2005.

49 Diggle PJ, Liang K-Y, Zeger SL. Analysis of Longitudinal Data. New York: Oxford University Press Inc., 1994.

50 Chodosh J, Miller-Martinez D, Aneshensel CS et al. Depressive symptoms, chronic diseases, and physical disabilities as predictors of cognitive functioning trajectories in older Americans. J Am Geriatr Soc2010;58(12):2350 – 57.

51 Cummins RA. Objective and subjective quality of life: an interactive model. Soc Indicators Res 2000;52:55 – 72.

52 Krause N, Newsom JT, Rook KS. Financial strain, negative social interaction, and self-rated health: evidence from two United States nationwide surveys. Ageing Soc 2008;28:1001 – 23.