Wealth in Middle and Later Life: Examining the Life Course Timing of Women’s Health Limitations

Wealth in Middle and Later Life: Examining the Life Course Timing of Women’s Health Limitations Abstract Background and Objectives Guided by cumulative inequality theory, this study poses two main questions: (a) Does women’s poor health compromise household financial assets? (b) If yes, is wealth sensitive to the timing of women’s health limitations? In addressing these questions, we consider the effect of health limitations on wealth at older ages, as well as examine how health limitations influence wealth over particular segments of the life course, giving attention to both the onset and duration of health limitations. Research Design and Methods Using 36 years of data from the National Longitudinal Survey of Mature Women, piecewise growth curve and linear regression models were used to estimate the effects of life course timing and duration of health limitations on household wealth. Results The findings reveal that women who experienced health limitations accumulated substantially less wealth over time, especially if the health limitations were manifest during childhood or early adulthood. Discussion and Implications This study identifies how early-life health problems lead to less wealth in later life. Women’s health, Financial assets, Health selection, Cumulative inequality The relationship between health and wealth has garnered attention in the last decade as researchers have sought to explicate the ways in which these life domains influence one another over time. Although most research has focused on how income and, to a lesser extent, wealth influence health, there is growing evidence that health also influences income and wealth. For instance, individuals in poor health face both social and economic challenges, and these may have important implications for educational attainment, labor force participation, and even retirement decisions (Cai, Mavromaras, & Oguzoglu, 2014; Lundborg, Nilsson, & Rooth, 2014; Lynch & von Hippel, 2016; Schinkel-Ivy, Mosca, & Mansfield, 2017). By contrast, those in good health generally have less complicated paths to status attainment and greater opportunity to accumulate assets. A life course approach to identifying the relationships between health and wealth gives priority to timing in a person’s life (Elder, 1998). There is substantial variability in the timing of health problems (both onset and duration) and this variability may have financial repercussions. Although many people experience health problems of relatively short duration, others may struggle with illness over long periods of time. Additionally, some individuals are disadvantaged from birth or encounter health problems early in life while others are not faced with these problems until advanced age. This variability in the timing of health problems may have a major influence on wealth accumulation; however, we are unaware of any studies that systematically examine how the timing of health problems influences wealth accumulation over a large span of the life course. Based on the life course perspective and cumulative inequality theory, this study examines the long-term consequences of women’s health limitations on household wealth. Our research questions are twofold: Does women’s poor health compromise household financial assets? If yes, is wealth sensitive to the timing of women’s health limitations? We address these and related questions with data from the National Longitudinal Survey of Mature Women (NLSMW), a 36-year panel study. Women’s Health and Status Attainment Empirical research has long established a strong relationship between health and socioeconomic status (SES), but debate over the precise nature of this association persists. Evidence in support of social causation—the effect of social status on health—dates back nearly three decades to the Whitehall study of British civil servants in which Marmot, Shipley, and Rose (1984) revealed an inverse relationship between employment grade and mortality over 10 years. Since then, the study has expanded to women and examined morbidity and a range of health outcomes showing compelling evidence in support of social causation (Marmot et al., 1991). Moreover, it stimulated scores of studies examining how SES affects various dimensions of health (e.g., Hammarström, Stenlund, & Janlert, 2011). Comparatively less attention, however, has been paid to health selection—the effect of health on status attainment. Health selection refers to the process through which good health leads to opportunity or, conversely, poor health leads to other risks. Despite the predominant focus on the social gradient, studies have revealed several ways in which life course transitions can be complicated by poor health and lead to lower status attainment, including reducing the likelihood that one will enroll in college (Jackson, 2009) and impinging on one’s ability to work—even forcing early retirement (Wu, 2003)—which may result in losses to both income and wealth (Smith, 1999, 2004). Although the costs of health selection are considerable, the effect of health on wealth has been among the largest across life stages: Haas (2006) found that excellent childhood health was associated with 104% greater wealth in adulthood than poor childhood health; Smith (2004) showed that among adults aged 51 years and older a new health problem reduces wealth by as much as $50,000 over an 8-year period. Few studies, however, examine this phenomenon separately for women. This is important given that women not only experience a higher prevalence of disease and disability compared to men, but their health may influence wealth in distinct ways. For example, marriage provides a wealth advantage (Wilmoth & Koso, 2002; Yamokoski & Keister, 2006), but less healthy individuals are less likely to marry, and this selection effect is often stronger for women (Conley & Glauber, 2007). Childbearing also uniquely affects women, which is tightly intertwined with women’s long-term health and status attainment. Early motherhood, in particular, contributes to lower status attainment (Kane, Morgan, Harris, & Guilkey, 2013), and when experienced outside marriage, the consequences for wealth are even greater (Painter, Frech, & Williams, 2015). Later in life, new health problems among older women, but not older men, increase the risk of divorce, and divorce is especially detrimental to women’s wealth (Wilmoth & Koso, 2002; Yamokoski & Keister, 2006). The majority of women today work while also performing a disproportionate share of household labor (Lachance-Grzela & Bouchard, 2010). Compared to men, the onset of a new health problem may contribute to a greater overall reduction in labor, both at home and at work. Indeed, Wu (2003) showed that women’s health problems are actually quite consequential to household wealth—even more so than men’s—possibly due to women’s greater share of nonmarket labor in the household. To our knowledge, this is one of the first studies to investigate the effects of women’s health limitations at each stage of the life course, as can be done with the NLSMW. In this way, the study contributes not only to the study of women’s role in wealth accumulation but also to the study of the life course and women’s financial vulnerability in later life (Baker, West, & Wood, 2017). An assessment of the health selection hypothesis, however, needs to account for parental SES, especially if one is examining the early onset of health limitations. Among those who experience serious health events, especially in early life, financial support from parents may curtail some of the costs to status attainment. Findings indicate that intergenerational financial transfers may account for as much as one-fourth of wealth accumulation (Modigliani, 1988). Additionally, because health problems are more common among those of lower socioeconomic status, we risk overstating the relationship between health limitations and wealth if we fail to model parental SES. Theoretical Framework In addition to empirical research, this study draws on cumulative inequality theory to guide our understanding of the relationship between women’s health and household wealth. First, the theory specifies that disadvantage increases exposure to risk, which may result in further disadvantage (Ferraro & Shippee, 2009). When applied to the study of health and wealth, those in poor health may be more susceptible to risks, including higher medical costs and the inability to work, which can lead to less wealth accumulation over time. In this sense, health status shapes the mechanisms that are related to financial accumulation. Whereas women have historically been more socioeconomically disadvantaged compared to men, one anticipates that it is important to consider how health problems may contribute to women’s additional disadvantages encountered at various points in the life course. Cumulative inequality theory also considers the importance of timing and duration in shaping trajectories of inequality (Ferraro & Shippee, 2009). The theory specifies that long-term exposure to disadvantage is more consequential than short-term exposure, and that trajectories are influenced not only by the amount of exposure, but also by the timing of events. For example, earlier onset of health problems may lead to a host of outcomes not encountered otherwise, including a lower probability of marrying, fewer years of schooling, and interrupted labor force participation. In the present study, we examine whether the life course timing and duration of health limitations are related to household wealth. Health limitations are a type of disadvantage, exposing people to additional risks, and the longer one faces such exposure, the greater the anticipated effect on wealth. Finally, it is well known that there is a strong intergenerational component to the accumulation of wealth and, according to the theory, “family lineage is a key source of life course inequality, especially for the early stages of the life course” (Ferraro, Shippee, & Schafer, 2009, p. 418). Although the NLSMW does not have wealth data on the respondent’s parents, SES is available. We expect that high parental SES is associated with greater household wealth—evidence of an intergenerational transmission of inequality—and that health limitations influence this relationship. Life Course Analysis of Health and Wealth The present study is distinct from prior investigations of health selection in several important ways. First, it draws on 36 years of data from the NLSMW. These data follow women from early middle adulthood (30–44 years of age) to older age (66–80) and include measures of health and SES at multiple points in time. We combine both retrospective and prospective data on health limitations to construct a comprehensive history of health limitations from childhood to older ages. Moreover, these data permit an examination of the health–wealth relationship among women, who have been studied less frequently than men. Second, this is an extension of studies which examine health selection at a single stage of the life course. Although the importance of both early- and later-life health on wealth has been shown repeatedly, few studies investigate the effect of health limitations across life stages. Adding this dimension is important given that (a) health problems may have their origins in earlier life course stages, (b) beyond their initial impact, health problems may have a prolonged influence on status attainment, and (c) the onset of health problems may have differential effects on wealth over time (Smith, 1999, 2004). Using women’s health histories, we can evaluate the long-term financial repercussions. Third, wealth is a household indicator, and the NLSMW enables us to consider important household and family influences on wealth. Unlike many previous studies, we include information on parental SES in estimating the effect of health limitations on wealth. Design and Methods Sample The NLSMW collected 21 waves of data over 36 years (1967–2003). Multistage probability sampling was used to draw a sample of 5,083 civilian, noninstitutionalized women aged 30–44 years in 1967, with an oversample of black women (n = 1,390). Respondents were resurveyed approximately every 2 years, and most of the interviewing was conducted in-person. By the conclusion of the study in 2003, the women were 66–80 years of age. The final survey yielded 2,237 respondents or 44% of the original sample. Attrition in the NLSMW was largely due to mortality (29%) and refusal (20%). Compared to the stayers, the attritors were, on average, older and less educated. They also reported less household wealth and lower self-ratings of health. We explicitly address attrition with our analytic methods, described below, that use either a Heckman selection model (Heckman, 1979) or a growth curve analysis so that respondents can contribute to the estimation of growth curves regardless of whether they missed one or more measurement occasions. Measures Household Wealth Respondents were asked detailed questions related to household assets and debts, including but not limited to housing, mortgages, savings, and loans. Debts for each household were summed and then subtracted from total household assets by the data investigators. Wealth was bottom-coded at $-999,999 and top-coded at $999,999 in most, but not all, survey years to ensure respondent confidentiality. Measures were top-coded at this value in the growth curves only to maintain consistency across survey waves. Wealth was also inflation-adjusted to 2003 dollars using the Consumer Price Index. Health Limitations The NLSMW provides both retrospective and prospective information on health limitations that enable us to trace women’s health limitations back to childhood—and construct a measure of lifetime health limitations. Beginning in 1967 at the initial wave, women were asked a series of questions about whether their health or physical condition limited what they could do. We created a binary variable at all but two survey waves (health limitations were not available in 1968 or 1969) to measure women’s health limitations from any of four activities: working at all, amount or kind of work one can do, housework, and other activities. In addition, if respondents reported any limitations due to their health at the initial wave, they were asked, “How long have you been limited in this way?” Prior health limitations were reported in years and in months. Although the retrospective data provide information on health limitations prior to the initial survey, this is true only for health problems manifest in 1967. Early-life health limitations that subsided by 1967 when women were 30–44 years of age were not captured by the study, resulting in conservative estimates for childhood- and early-adult health problems. Similar measures are used by other studies (e.g., Garbaski, 2014; Gee & Walsemann, 2009). We combine the retrospective and prospective indicators to construct two sets of measures: timing and duration of health limitations. Timing or onset is measured using a system of binary variables that represent the life stage during which a health limitation was first reported: childhood (<18), early adulthood (18–34), early middle age (35–49), late middle age (50–64), and older age (65+). Other studies use similar ages to differentiate life stages (e.g., Kahn & Pearlin, 2006). By contrast, duration is the summation of years in which women reported health limitations (top-coded at the top 2% due to the small number of responses in higher categories). Specifically, we added together the number of years women retrospectively reported living with a health limitation at baseline and the number of survey waves with an affirmative response—treating health limitations in consecutive survey waves as uninterrupted episodes. We examined alternative strategies to coding health limitations and obtained similar results. Personal and Household Characteristics A host of personal and household characteristics are available to adjust our linear regression estimates. Chronological age is measured in years. Black is a binary variable coded 1 for black and 0 for non-black respondents. Continuous measures capture the proportion of time during the study women spent married, working full-time and part-time, and living with a spouse who had health problems. On spousal variables, nonmarried respondents were coded 0. The number of children belonging to each respondent is a count variable measured at baseline when women were aged 30–44 (top-coded at the top 2%). Education was measured by years of schooling completed. Both the respondent and her parent’s occupational prestige were measured using the Duncan Socioeconomic Index (SEI; Duncan, 1961). Duncan SEI consists of prestige ratings for occupations, ranging from 0 to 97, based on their income and education distributions (U.S. Department of Labor, 2001). The coding and descriptive statistics for the variables are displayed in Table 1. Table 1. Means and Standard Deviations of Variables in the NLSMW, 1967–2003   Description  Range  Mean  SD  Dependent Variable  Household Wealth, 1967a  In thousands  −99.999–482.500  10.037  22.102  Household Wealth, 2003  In thousands  −99.999–4109.665  235.937  377.782  Independent Variables  Duration of Health Limitations  Duration  Years  0–44  8.276  10.478  Timing of Health Limitationsb  Childhood  0–17 years  0/1  0.018  c  Early Adulthood  18–34 years  0/1  0.102    Early Middle Age  35–49 years  0/1  0.342    Late Middle Age  50–64 years  0/1  0.185    Older Age  65+ years  0/1  0.069    Personal and Household Characteristics  Age, 1967  Years  30–44  37.230  4.358  Black, 1967  1 = black, 0 = non-black  0/1  0.280    Married, 1967–2003d  Proportion of time  0–1  0.689  0.379  Husband’s Health Limitations, 1967–2003d  Proportion of time  0–1  0.147  0.231  Children, 1967  Number of children  0–9  3.160  2.205  Education, 1967  Years of education  0–18  10.918  2.842  Parent’s Prestige, 1967  Duncan SEI  4–96  26.415  21.542  Respondent’s Prestige, 1967  Duncan SEI  0–93  33.763  21.116  Full-Time Employment, 1967–2003d  Proportion of time  0–1  0.288  0.291  Part-Time Employment, 1967–2003d  Proportion of time  0–1  0.156  0.197    Description  Range  Mean  SD  Dependent Variable  Household Wealth, 1967a  In thousands  −99.999–482.500  10.037  22.102  Household Wealth, 2003  In thousands  −99.999–4109.665  235.937  377.782  Independent Variables  Duration of Health Limitations  Duration  Years  0–44  8.276  10.478  Timing of Health Limitationsb  Childhood  0–17 years  0/1  0.018  c  Early Adulthood  18–34 years  0/1  0.102    Early Middle Age  35–49 years  0/1  0.342    Late Middle Age  50–64 years  0/1  0.185    Older Age  65+ years  0/1  0.069    Personal and Household Characteristics  Age, 1967  Years  30–44  37.230  4.358  Black, 1967  1 = black, 0 = non-black  0/1  0.280    Married, 1967–2003d  Proportion of time  0–1  0.689  0.379  Husband’s Health Limitations, 1967–2003d  Proportion of time  0–1  0.147  0.231  Children, 1967  Number of children  0–9  3.160  2.205  Education, 1967  Years of education  0–18  10.918  2.842  Parent’s Prestige, 1967  Duncan SEI  4–96  26.415  21.542  Respondent’s Prestige, 1967  Duncan SEI  0–93  33.763  21.116  Full-Time Employment, 1967–2003d  Proportion of time  0–1  0.288  0.291  Part-Time Employment, 1967–2003d  Proportion of time  0–1  0.156  0.197  Note: NLSMW = National Longitudinal Survey of Mature Women. aWealth was measured on 12 occasions during the study; the first and last measurements are displayed. bTiming refers to the onset of a health limitation; reference group = no health limitations (28.4% of respondents). cThe standard deviation of a dichotomous variable is omitted because it is a function of the mean. dTime-varying covariate in the growth curve analysis. View Large Table 1. Means and Standard Deviations of Variables in the NLSMW, 1967–2003   Description  Range  Mean  SD  Dependent Variable  Household Wealth, 1967a  In thousands  −99.999–482.500  10.037  22.102  Household Wealth, 2003  In thousands  −99.999–4109.665  235.937  377.782  Independent Variables  Duration of Health Limitations  Duration  Years  0–44  8.276  10.478  Timing of Health Limitationsb  Childhood  0–17 years  0/1  0.018  c  Early Adulthood  18–34 years  0/1  0.102    Early Middle Age  35–49 years  0/1  0.342    Late Middle Age  50–64 years  0/1  0.185    Older Age  65+ years  0/1  0.069    Personal and Household Characteristics  Age, 1967  Years  30–44  37.230  4.358  Black, 1967  1 = black, 0 = non-black  0/1  0.280    Married, 1967–2003d  Proportion of time  0–1  0.689  0.379  Husband’s Health Limitations, 1967–2003d  Proportion of time  0–1  0.147  0.231  Children, 1967  Number of children  0–9  3.160  2.205  Education, 1967  Years of education  0–18  10.918  2.842  Parent’s Prestige, 1967  Duncan SEI  4–96  26.415  21.542  Respondent’s Prestige, 1967  Duncan SEI  0–93  33.763  21.116  Full-Time Employment, 1967–2003d  Proportion of time  0–1  0.288  0.291  Part-Time Employment, 1967–2003d  Proportion of time  0–1  0.156  0.197    Description  Range  Mean  SD  Dependent Variable  Household Wealth, 1967a  In thousands  −99.999–482.500  10.037  22.102  Household Wealth, 2003  In thousands  −99.999–4109.665  235.937  377.782  Independent Variables  Duration of Health Limitations  Duration  Years  0–44  8.276  10.478  Timing of Health Limitationsb  Childhood  0–17 years  0/1  0.018  c  Early Adulthood  18–34 years  0/1  0.102    Early Middle Age  35–49 years  0/1  0.342    Late Middle Age  50–64 years  0/1  0.185    Older Age  65+ years  0/1  0.069    Personal and Household Characteristics  Age, 1967  Years  30–44  37.230  4.358  Black, 1967  1 = black, 0 = non-black  0/1  0.280    Married, 1967–2003d  Proportion of time  0–1  0.689  0.379  Husband’s Health Limitations, 1967–2003d  Proportion of time  0–1  0.147  0.231  Children, 1967  Number of children  0–9  3.160  2.205  Education, 1967  Years of education  0–18  10.918  2.842  Parent’s Prestige, 1967  Duncan SEI  4–96  26.415  21.542  Respondent’s Prestige, 1967  Duncan SEI  0–93  33.763  21.116  Full-Time Employment, 1967–2003d  Proportion of time  0–1  0.288  0.291  Part-Time Employment, 1967–2003d  Proportion of time  0–1  0.156  0.197  Note: NLSMW = National Longitudinal Survey of Mature Women. aWealth was measured on 12 occasions during the study; the first and last measurements are displayed. bTiming refers to the onset of a health limitation; reference group = no health limitations (28.4% of respondents). cThe standard deviation of a dichotomous variable is omitted because it is a function of the mean. dTime-varying covariate in the growth curve analysis. View Large The growth curve analysis also includes these control variables, but three of them are treated as time-varying covariates: marital status, husband’s health, and employment. In the piecewise growth curve models, marital status distinguishes between married, widowed, divorced/separated, and never married (reference group) respondents. Husband’s health is coded 1 for any health problem or condition that limits his ability to work and 0, otherwise. Employment status is measured using binary variables for full-time and part-time employment. Analytic Plan Two research questions are addressed with multiple analytic strategies, and we organize the analysis into two stages. We began with an ordinary least squares regression (OLS) analysis of the women who survived to 2003 to provide a straightforward test of our first research question. We applied a hyperbolic sine transformation to the dependent variable to handle the skewed continuous distribution of wealth, which also includes negative values and values of zero (Burbidge, Magee, & Robb, 1988). In supplementary analyses, we tested alternative transformations and the substantive conclusions were unchanged. We also adjusted measures for the number of household members and reached similar conclusions. To account for attrition due to mortality, we estimated a Heckman selection model, using unique information in the selection equation (e.g., self-rated health, requires help with personal care, and parental life status), and included the variable expressing the nonselection hazard (inverse of Mills’ ratio) in our linear regression analysis. We also examined other sources of attrition, including nonresponse, but the conclusions were the same. After accounting for selective mortality, these analyses permit one to gauge the effect of women’s health limitations on wealth accumulation in 2003. We also incorporate tests for differences in the onset of women’s health limitations—from childhood to later life—to discern their impact on wealth when most of the women were retired. The second stage of the analysis consisted of piecewise growth curve models. The piecewise model enables us to capitalize on available data between 1967 and 2003 by allowing each woman to contribute as many observations as possible, even those who do not survive to the final wave. We model trajectories of wealth from 12 occasions and examine the effect of health limitations at various stages of the life course on the intercept and rate of growth in wealth over time, while modeling separate slopes by life course stage. One advantage of the piecewise model is that it allows us to include onset predictors only for slopes in ages that are causally prior. From a policy vantage point, these analyses provide a window into key life course periods when the effects of health limitations are having their greatest effect. We used age as the time metric in the piecewise growth curve models. The variable for age was broken into three pieces: age 34–49, 49–65, and 65–78, and centered at age 34, which allows us to interpret the intercept as women’s household wealth at age 34, as well as examine the rate of change over particular segments of the life course as these women grow older. Observations with an age below 34 and an age above 78 were dropped from the growth curve analysis because there were far fewer of these compared to other ages. This also makes the splines more balanced in terms of the number of years each piece covers. In addition, we mean-centered education, occupational prestige, and number of children to provide a more meaningful constant to interpret. Results As shown in Table 1, the mean household wealth expressed in thousands during 1967 (when the women were 30–44) was 10.037 or $10,037 inflation-adjusted dollars. By comparison, the mean household wealth in 2003 (when the women were 66–80) was 235.937 or $235,937 dollars. Women in the sample were, on average, 37 years of age in 1967 and more than one-quarter of respondents had not reported any health limitations by 2003 (28%). The majority of women, however, first reported health limitations in early middle age (34%), with the average duration of health limitations equal to eight years. For the first stage of the analysis, Table 2 presents the results from four OLS regression models, with variables introduced in blocks, to address whether women’s health limitations are associated with less household wealth in later life. Model 1 examines personal and household characteristics in 1967 on wealth accumulation (while adjusting for selective mortality). Both the respondent and her parent’s occupational prestige had positive effects on household wealth, although the effect of the respondent’s own occupational prestige was stronger. We tested interactions between parental occupational prestige and women’s health limitations, but none were significant (results not shown). Table 2. OLS Regression of Accumulated Household Wealth in 2003   Model 1  Model 2  Model 3  Model 4  Personal and Household Characteristics  Age, 1967  0.023a  0.023  0.005  0.028  (0.023)  (0.023)  (0.023)  (0.022)  Black, 1967  −1.081***  −1.165***  −1.124***  −0.488*  (0.209)  (0.204)  (0.206)  (0.203)  Children, 1967  −0.039  −0.019  −0.028  −0.071*  (0.037)  (0.036)  (0.037)  (0.035)  Education, 1967  0.098*  0.066  0.082*  0.078*  (0.038)  (0.037)  (0.037)  (0.035)  Parent’s Prestige, 1967  0.010**  0.009**  0.008*  0.008*  (0.004)  (0.003)  (0.003)  (0.003)  Respondent’s Prestige, 1967  0.018***  0.014**  0.016***  0.014**  (0.004)  (0.004)  (0.004)  (0.004)  Mortality Selection  −1.568**  −1.175**  −1.193**  −1.171**  (0.455)  (0.447)  (0.451)  (0.427)  Duration of Health Limitations  Duration    −0.073***        (0.018)      Duration2    0.001        (0.000)      Timing of Health Limitationsb   Childhood      −2.648***  −2.127***      (0.558)  (0.531)   Early Adulthood      −1.355***  −0.986***      (0.285)  (0.274)   Early Middle Age      −0.876***  −0.607**      (0.198)  (0.192)   Late Middle Age      −0.621**  −0.421*      (0.210)  (0.199)   Older Age      −0.598*  −0.563*      (0.262)  (0.246)  Life Course Events  Married, 1967–2003c        2.319***        (0.216)  Husband’s Health Limitations, 1967–2003c        −0.788*        (0.334)  Full-Time Employment, 1967–2003c        0.342        (0.295)  Part-Time Employment, 1967–2003c        0.218        (0.380)  Constant  3.602***  4.311***  4.910***  2.466**  (0.727)  (0.714)  (0.749)  (0.767)  Adjusted R2  0.282  0.322  0.310  0.394  N  902  902  902  902    Model 1  Model 2  Model 3  Model 4  Personal and Household Characteristics  Age, 1967  0.023a  0.023  0.005  0.028  (0.023)  (0.023)  (0.023)  (0.022)  Black, 1967  −1.081***  −1.165***  −1.124***  −0.488*  (0.209)  (0.204)  (0.206)  (0.203)  Children, 1967  −0.039  −0.019  −0.028  −0.071*  (0.037)  (0.036)  (0.037)  (0.035)  Education, 1967  0.098*  0.066  0.082*  0.078*  (0.038)  (0.037)  (0.037)  (0.035)  Parent’s Prestige, 1967  0.010**  0.009**  0.008*  0.008*  (0.004)  (0.003)  (0.003)  (0.003)  Respondent’s Prestige, 1967  0.018***  0.014**  0.016***  0.014**  (0.004)  (0.004)  (0.004)  (0.004)  Mortality Selection  −1.568**  −1.175**  −1.193**  −1.171**  (0.455)  (0.447)  (0.451)  (0.427)  Duration of Health Limitations  Duration    −0.073***        (0.018)      Duration2    0.001        (0.000)      Timing of Health Limitationsb   Childhood      −2.648***  −2.127***      (0.558)  (0.531)   Early Adulthood      −1.355***  −0.986***      (0.285)  (0.274)   Early Middle Age      −0.876***  −0.607**      (0.198)  (0.192)   Late Middle Age      −0.621**  −0.421*      (0.210)  (0.199)   Older Age      −0.598*  −0.563*      (0.262)  (0.246)  Life Course Events  Married, 1967–2003c        2.319***        (0.216)  Husband’s Health Limitations, 1967–2003c        −0.788*        (0.334)  Full-Time Employment, 1967–2003c        0.342        (0.295)  Part-Time Employment, 1967–2003c        0.218        (0.380)  Constant  3.602***  4.311***  4.910***  2.466**  (0.727)  (0.714)  (0.749)  (0.767)  Adjusted R2  0.282  0.322  0.310  0.394  N  902  902  902  902  Note: OLS = Ordinary least squares regression analysis. aUnstandardized regression coefficient (standard error in parentheses). bTiming refers to the onset of a health limitation; reference group = no health limitations. cProportion of time in the study period. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Table 2. OLS Regression of Accumulated Household Wealth in 2003   Model 1  Model 2  Model 3  Model 4  Personal and Household Characteristics  Age, 1967  0.023a  0.023  0.005  0.028  (0.023)  (0.023)  (0.023)  (0.022)  Black, 1967  −1.081***  −1.165***  −1.124***  −0.488*  (0.209)  (0.204)  (0.206)  (0.203)  Children, 1967  −0.039  −0.019  −0.028  −0.071*  (0.037)  (0.036)  (0.037)  (0.035)  Education, 1967  0.098*  0.066  0.082*  0.078*  (0.038)  (0.037)  (0.037)  (0.035)  Parent’s Prestige, 1967  0.010**  0.009**  0.008*  0.008*  (0.004)  (0.003)  (0.003)  (0.003)  Respondent’s Prestige, 1967  0.018***  0.014**  0.016***  0.014**  (0.004)  (0.004)  (0.004)  (0.004)  Mortality Selection  −1.568**  −1.175**  −1.193**  −1.171**  (0.455)  (0.447)  (0.451)  (0.427)  Duration of Health Limitations  Duration    −0.073***        (0.018)      Duration2    0.001        (0.000)      Timing of Health Limitationsb   Childhood      −2.648***  −2.127***      (0.558)  (0.531)   Early Adulthood      −1.355***  −0.986***      (0.285)  (0.274)   Early Middle Age      −0.876***  −0.607**      (0.198)  (0.192)   Late Middle Age      −0.621**  −0.421*      (0.210)  (0.199)   Older Age      −0.598*  −0.563*      (0.262)  (0.246)  Life Course Events  Married, 1967–2003c        2.319***        (0.216)  Husband’s Health Limitations, 1967–2003c        −0.788*        (0.334)  Full-Time Employment, 1967–2003c        0.342        (0.295)  Part-Time Employment, 1967–2003c        0.218        (0.380)  Constant  3.602***  4.311***  4.910***  2.466**  (0.727)  (0.714)  (0.749)  (0.767)  Adjusted R2  0.282  0.322  0.310  0.394  N  902  902  902  902    Model 1  Model 2  Model 3  Model 4  Personal and Household Characteristics  Age, 1967  0.023a  0.023  0.005  0.028  (0.023)  (0.023)  (0.023)  (0.022)  Black, 1967  −1.081***  −1.165***  −1.124***  −0.488*  (0.209)  (0.204)  (0.206)  (0.203)  Children, 1967  −0.039  −0.019  −0.028  −0.071*  (0.037)  (0.036)  (0.037)  (0.035)  Education, 1967  0.098*  0.066  0.082*  0.078*  (0.038)  (0.037)  (0.037)  (0.035)  Parent’s Prestige, 1967  0.010**  0.009**  0.008*  0.008*  (0.004)  (0.003)  (0.003)  (0.003)  Respondent’s Prestige, 1967  0.018***  0.014**  0.016***  0.014**  (0.004)  (0.004)  (0.004)  (0.004)  Mortality Selection  −1.568**  −1.175**  −1.193**  −1.171**  (0.455)  (0.447)  (0.451)  (0.427)  Duration of Health Limitations  Duration    −0.073***        (0.018)      Duration2    0.001        (0.000)      Timing of Health Limitationsb   Childhood      −2.648***  −2.127***      (0.558)  (0.531)   Early Adulthood      −1.355***  −0.986***      (0.285)  (0.274)   Early Middle Age      −0.876***  −0.607**      (0.198)  (0.192)   Late Middle Age      −0.621**  −0.421*      (0.210)  (0.199)   Older Age      −0.598*  −0.563*      (0.262)  (0.246)  Life Course Events  Married, 1967–2003c        2.319***        (0.216)  Husband’s Health Limitations, 1967–2003c        −0.788*        (0.334)  Full-Time Employment, 1967–2003c        0.342        (0.295)  Part-Time Employment, 1967–2003c        0.218        (0.380)  Constant  3.602***  4.311***  4.910***  2.466**  (0.727)  (0.714)  (0.749)  (0.767)  Adjusted R2  0.282  0.322  0.310  0.394  N  902  902  902  902  Note: OLS = Ordinary least squares regression analysis. aUnstandardized regression coefficient (standard error in parentheses). bTiming refers to the onset of a health limitation; reference group = no health limitations. cProportion of time in the study period. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Despite Models 2 and 3 being similar in design, they differ in their measurement of health limitations: Model 2 uses a measure of duration (along with a quadratic term), whereas Model 3 draws on a system of binary indicators that capture the life course timing of health limitations. Prior research documented the correlation between duration and timing and the difficulty in disentangling such measures (Hallqvist, Lynch, Bartley, Lang, & Blane, 2004). For that reason, we model the duration and timing of health limitations separately. First, results from Model 2 reveal a negative relationship between duration of exposure and accumulated wealth at older ages. By contrast, Model 3 presents the timing of health limitations by life course stage. Women who reported health limitations during any period in the life course had less household wealth in older adulthood than those with no health limitations. Most notably, among those who reported health limitations, there was evidence of a health gradient, with earlier onset being more consequential to wealth. Model 4 adjusts for subsequent life changes over the study period (in particular: marriage, husband’s health, and employment). Health limitations at each life stage remained a significant predictor of household wealth even after accounting for changes in family and employment. Even though the effects are somewhat attenuated, the timing of health limitations has an important influence on household wealth. Although the OLS regressions illustrate the magnitude of the problem, we now turn to the piecewise growth curve analyses to more precisely quantify the impact of health limitations and the processes involved. The results from this second stage of the analysis are presented in Table 3. In specifying the models, Model 1 is an unconditional growth model with random intercept and slope components, revealing that wealth increases in middle age as women grow older, but decreases after age 65. Table 3. Piecewise Growth Curve of Women’s Household Wealth, 1967–2003   Model 1  Model 2  Model 3  Model 4  Initial Status   Constant  2.802***,a  3.305***  3.351***  1.959***  (0.048)  (0.050)  (0.053)  (0.126)  Personal and Household Characteristics   Black, 1967    −1.577***  −1.577***  −1.288***    (0.073)  (0.072)  (0.068)   Children, 1967    −0.028*  −0.027*  −0.058***    (0.014)  (0.014)  (0.013)   Education, 1967    0.171***  0.163***  0.158***    (0.014)  (0.014)  (0.013)   Parent’s Prestige, 1967    0.003  0.002  0.003*    (0.001)  (0.001)  (0.001)   Respondent’s Prestige, 1967    0.017***  0.017***  0.016***    (0.002)  (0.002)  (0.002)  Timing of Health Limitationsb   Childhood      −0.735*  −0.607      (0.351)  (0.337)   Early Adulthood      −0.330*  −0.246      (0.149)  (0.143)  Life Course Events   Married, 1967–2003c        1.490***        (0.118)   Widowed, 1967–2003c        0.890***        (0.124)   Divorced, 1967–2003c        0.216        (0.124)   Husband’s Health Limitations, 1967–2003c        −0.146***        (0.040)   Full-Time Employment, 1967–2003c        0.114**        (0.034)   Part-Time Employment, 1967–2003c        0.080*        (0.036)  Rate of Change, Ages 34–49   Intercept  0.067***  0.067***  0.066***  0.072***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      0.009  0.021      (0.027)  (0.026)   Early Adulthood      0.004  0.004      (0.011)  (0.011)  Rate of Change, Ages 49–65   Intercept  0.050***  0.050***  0.058***  0.069***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      −0.035  −0.048*      (0.024)  (0.023)   Early Adulthood      −0.032**  −0.033**      (0.011)  (0.010)   Early Middle Age      −0.015**  −0.016**      (0.005)  (0.005)  Rate of Change, Ages 65–78           Intercept  −0.017**  −0.016*  −0.001  0.008  (0.006)  (0.006)  (0.015)  (0.015)  Timing of Health Limitationsb   Childhood      −0.054  −0.061      (0.055)  (0.054)   Early Adulthood      −0.009  −0.002      (0.029)  (0.029)   Early Middle Age      −0.012  −0.014      (0.019)  (0.018)   Late Middle Age      −0.027  −0.023      (0.018)  (0.018)   Older Age      −0.016  −0.014      (0.021)  (0.021)  N of Persons  4,664  3,989  3,989  3,989    Model 1  Model 2  Model 3  Model 4  Initial Status   Constant  2.802***,a  3.305***  3.351***  1.959***  (0.048)  (0.050)  (0.053)  (0.126)  Personal and Household Characteristics   Black, 1967    −1.577***  −1.577***  −1.288***    (0.073)  (0.072)  (0.068)   Children, 1967    −0.028*  −0.027*  −0.058***    (0.014)  (0.014)  (0.013)   Education, 1967    0.171***  0.163***  0.158***    (0.014)  (0.014)  (0.013)   Parent’s Prestige, 1967    0.003  0.002  0.003*    (0.001)  (0.001)  (0.001)   Respondent’s Prestige, 1967    0.017***  0.017***  0.016***    (0.002)  (0.002)  (0.002)  Timing of Health Limitationsb   Childhood      −0.735*  −0.607      (0.351)  (0.337)   Early Adulthood      −0.330*  −0.246      (0.149)  (0.143)  Life Course Events   Married, 1967–2003c        1.490***        (0.118)   Widowed, 1967–2003c        0.890***        (0.124)   Divorced, 1967–2003c        0.216        (0.124)   Husband’s Health Limitations, 1967–2003c        −0.146***        (0.040)   Full-Time Employment, 1967–2003c        0.114**        (0.034)   Part-Time Employment, 1967–2003c        0.080*        (0.036)  Rate of Change, Ages 34–49   Intercept  0.067***  0.067***  0.066***  0.072***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      0.009  0.021      (0.027)  (0.026)   Early Adulthood      0.004  0.004      (0.011)  (0.011)  Rate of Change, Ages 49–65   Intercept  0.050***  0.050***  0.058***  0.069***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      −0.035  −0.048*      (0.024)  (0.023)   Early Adulthood      −0.032**  −0.033**      (0.011)  (0.010)   Early Middle Age      −0.015**  −0.016**      (0.005)  (0.005)  Rate of Change, Ages 65–78           Intercept  −0.017**  −0.016*  −0.001  0.008  (0.006)  (0.006)  (0.015)  (0.015)  Timing of Health Limitationsb   Childhood      −0.054  −0.061      (0.055)  (0.054)   Early Adulthood      −0.009  −0.002      (0.029)  (0.029)   Early Middle Age      −0.012  −0.014      (0.019)  (0.018)   Late Middle Age      −0.027  −0.023      (0.018)  (0.018)   Older Age      −0.016  −0.014      (0.021)  (0.021)  N of Persons  4,664  3,989  3,989  3,989  Note: aUnstandardized regression coefficient (standard error in parentheses). bTiming refers to the onset of a health limitation; reference group = no health limitations. cTime-varying covariate. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Table 3. Piecewise Growth Curve of Women’s Household Wealth, 1967–2003   Model 1  Model 2  Model 3  Model 4  Initial Status   Constant  2.802***,a  3.305***  3.351***  1.959***  (0.048)  (0.050)  (0.053)  (0.126)  Personal and Household Characteristics   Black, 1967    −1.577***  −1.577***  −1.288***    (0.073)  (0.072)  (0.068)   Children, 1967    −0.028*  −0.027*  −0.058***    (0.014)  (0.014)  (0.013)   Education, 1967    0.171***  0.163***  0.158***    (0.014)  (0.014)  (0.013)   Parent’s Prestige, 1967    0.003  0.002  0.003*    (0.001)  (0.001)  (0.001)   Respondent’s Prestige, 1967    0.017***  0.017***  0.016***    (0.002)  (0.002)  (0.002)  Timing of Health Limitationsb   Childhood      −0.735*  −0.607      (0.351)  (0.337)   Early Adulthood      −0.330*  −0.246      (0.149)  (0.143)  Life Course Events   Married, 1967–2003c        1.490***        (0.118)   Widowed, 1967–2003c        0.890***        (0.124)   Divorced, 1967–2003c        0.216        (0.124)   Husband’s Health Limitations, 1967–2003c        −0.146***        (0.040)   Full-Time Employment, 1967–2003c        0.114**        (0.034)   Part-Time Employment, 1967–2003c        0.080*        (0.036)  Rate of Change, Ages 34–49   Intercept  0.067***  0.067***  0.066***  0.072***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      0.009  0.021      (0.027)  (0.026)   Early Adulthood      0.004  0.004      (0.011)  (0.011)  Rate of Change, Ages 49–65   Intercept  0.050***  0.050***  0.058***  0.069***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      −0.035  −0.048*      (0.024)  (0.023)   Early Adulthood      −0.032**  −0.033**      (0.011)  (0.010)   Early Middle Age      −0.015**  −0.016**      (0.005)  (0.005)  Rate of Change, Ages 65–78           Intercept  −0.017**  −0.016*  −0.001  0.008  (0.006)  (0.006)  (0.015)  (0.015)  Timing of Health Limitationsb   Childhood      −0.054  −0.061      (0.055)  (0.054)   Early Adulthood      −0.009  −0.002      (0.029)  (0.029)   Early Middle Age      −0.012  −0.014      (0.019)  (0.018)   Late Middle Age      −0.027  −0.023      (0.018)  (0.018)   Older Age      −0.016  −0.014      (0.021)  (0.021)  N of Persons  4,664  3,989  3,989  3,989    Model 1  Model 2  Model 3  Model 4  Initial Status   Constant  2.802***,a  3.305***  3.351***  1.959***  (0.048)  (0.050)  (0.053)  (0.126)  Personal and Household Characteristics   Black, 1967    −1.577***  −1.577***  −1.288***    (0.073)  (0.072)  (0.068)   Children, 1967    −0.028*  −0.027*  −0.058***    (0.014)  (0.014)  (0.013)   Education, 1967    0.171***  0.163***  0.158***    (0.014)  (0.014)  (0.013)   Parent’s Prestige, 1967    0.003  0.002  0.003*    (0.001)  (0.001)  (0.001)   Respondent’s Prestige, 1967    0.017***  0.017***  0.016***    (0.002)  (0.002)  (0.002)  Timing of Health Limitationsb   Childhood      −0.735*  −0.607      (0.351)  (0.337)   Early Adulthood      −0.330*  −0.246      (0.149)  (0.143)  Life Course Events   Married, 1967–2003c        1.490***        (0.118)   Widowed, 1967–2003c        0.890***        (0.124)   Divorced, 1967–2003c        0.216        (0.124)   Husband’s Health Limitations, 1967–2003c        −0.146***        (0.040)   Full-Time Employment, 1967–2003c        0.114**        (0.034)   Part-Time Employment, 1967–2003c        0.080*        (0.036)  Rate of Change, Ages 34–49   Intercept  0.067***  0.067***  0.066***  0.072***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      0.009  0.021      (0.027)  (0.026)   Early Adulthood      0.004  0.004      (0.011)  (0.011)  Rate of Change, Ages 49–65   Intercept  0.050***  0.050***  0.058***  0.069***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      −0.035  −0.048*      (0.024)  (0.023)   Early Adulthood      −0.032**  −0.033**      (0.011)  (0.010)   Early Middle Age      −0.015**  −0.016**      (0.005)  (0.005)  Rate of Change, Ages 65–78           Intercept  −0.017**  −0.016*  −0.001  0.008  (0.006)  (0.006)  (0.015)  (0.015)  Timing of Health Limitationsb   Childhood      −0.054  −0.061      (0.055)  (0.054)   Early Adulthood      −0.009  −0.002      (0.029)  (0.029)   Early Middle Age      −0.012  −0.014      (0.019)  (0.018)   Late Middle Age      −0.027  −0.023      (0.018)  (0.018)   Older Age      −0.016  −0.014      (0.021)  (0.021)  N of Persons  4,664  3,989  3,989  3,989  Note: aUnstandardized regression coefficient (standard error in parentheses). bTiming refers to the onset of a health limitation; reference group = no health limitations. cTime-varying covariate. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Model 2 adds personal and household characteristics in 1967 to examine the effect of each covariate on wealth, including parental occupational prestige. The relationship between parental occupational prestige and initial household wealth was nonsignificant in Model 2. We also tested whether parental occupational prestige influenced the rate of change in wealth over time, but the parameter was nonsignificant (not shown). Model 3 incorporates binary indicators that capture the life course timing of health limitations. The model reveals that health limitations in childhood and early adulthood lowered the intercept on average, although neither produced an effect on the slope between 34 and 49 years of age. Between the ages of 49 and 65, there was significant growth in wealth; however, the onset of health limitations in early adulthood slowed the rate of growth in household wealth during this period of time, as did health limitations in early middle age. Earlier onset had larger effects on both the intercept and slope. Model 4 examines the effect of health limitations net of subsequent life changes treated as time-varying covariates. The results are not markedly different from Model 3; however, the effects of childhood- and early-adult health problems on the intercept became nonsignificant after controlling for changes in family and employment during the life course, though both slowed the rate of growth between the ages of 49 and 65. Estimates for personal and household characteristics in 1967 were slightly attenuated from Model 3, with the exception of parental occupational prestige which was significant: higher parental occupational prestige was associated with a higher average intercept. Supplementary Analyses Most women in our sample were married in 1967 (80%), with a modest percentage of these women divorcing during the span of the study (17%). This sample, therefore, represents a rather homogeneous cohort of women, which helps in some aspects to isolate how health limitations influence wealth. Studies on marriage and wealth nonetheless show the importance of marital histories for wealth accumulation, finding that those who are continuously married fare better than those who are not (Wilmoth & Koso, 2002). We thus examined our findings in the context of marital histories more closely. Extending our analyses in Table 2, we compared the effect of health limitations across marital histories (1 = continuously married, 0 = noncontinuously married) over the study period for those who survived to 2003. We examined interaction terms between the health and marital status variables, but none were significant. Continuously married women, however, were less likely to experience health limitations over the life course and reported shorter duration, on average. Given that wealth is measured at the household level, we tested additional information from the husband, including education and occupational prestige; however, our substantive conclusions were unchanged. We also tested for mediation with nonparametric bootstrap in R using the mediation package (Imai, Keele, & Tingley, 2010) but found none for any of the onset measures. Discussion and Policy Implications It has long been known that health and wealth are related, but most research on the topic has focused on how economic resources are salubrious to health—and how economic deprivation is bad for health. The purpose of this research was to examine the effect of health limitations on wealth by capitalizing on exceptional longitudinal data on health limitations and asset accumulation. Previous studies of health selection illustrate the extent to which poor health influences wealth accumulation, but we are unaware of any studies that isolated how the onset of health limitations at various points in the life course influence wealth accumulation. We found that women’s long-term health limitations take a substantial toll on household wealth that is independent of the woman’s marital status and employment, as well as her husband’s health. The results presented herein are consistent with some studies examining health and wealth in later life (Michaud & van Soest, 2008; Wu, 2003), but distinctive by identifying how early-life health problems lead to less wealth in later life. These findings have implications for research on health and social stratification and support elements of cumulative inequality theory. With a focus on how social stratification unfolds over the life course, the theory specifies that trajectories are shaped by the “onset, duration, and magnitude of exposures” (Ferraro et al., 2009, p. 420). Findings from the NLSMW revealed that life course timing and duration are critical for how health limitations lead to less wealth. Also consistent with the theory and prior studies is the influence of family lineage: parental occupational prestige is related to the wealth of the child’s household, independent of the respondent’s occupational prestige (Andrew & Ruel, 2010). The implications for this study are threefold. First, the findings contribute to our understanding of how health problems reduce household wealth by considering “upstream” factors that shape financial security in later life. There is substantial variability in the life course timing of health limitations, and without this knowledge, studies may overlook the effect of health limitations on wealth at particular life stages. To explicate the health–wealth relationship in later life, it is important to integrate elements of earlier life. Second, a life course study of health selection is relevant not only for future studies but also contributes to effective interventions. An increasing number of Americans are becoming bankrupt due to out-of-pocket medical expenses (Himmelstein, Thorne, Warren, & Woolhandler, 2009), and identifying the life course pathways could potentially broaden our understanding and deliver more targeted help. We found that the onset of health limitations reduced wealth, regardless of life stage, but onset in early life was most detrimental, slowing the rate of growth during middle age. By contrast, there was little change in wealth beyond age 65 due to onset of health limitations. Whereas the financial loss was manifest early, there is a need for interventions for those with early-life health problems so that fewer women become poor in later life. Future studies, however, should also consider the role of entitlement programs (e.g., Social Security, Medicare, and Medicaid) in the health–wealth relationship in older adulthood, which may at least partially offset the cost of health limitations. Finally, given recent social and demographic trends related to marriage, divorce, fertility, and work in the United States, there is an urgent need to develop social policies to enhance financial security in later life. A rising divorce rate among persons aged 50 years and older (Brown & Lin, 2012), coupled with a projected retirement crisis, will likely translate into a growing number of older adults facing financial insecurity. As shown herein, women’s health problems during early and middle adulthood are associated with lower household wealth: the financial loss affects both women and men. Indeed, despite the notable rise in women’s labor force participation, the Social Security replacement rate for married couples has actually declined in recent decades (Ellis, Munnell, & Eschtruth, 2014). In addition, the growing population of older divorced women is likely to face financial insecurity because of “relatively low Social Security benefits and high poverty levels” (Lin, Brown, & Hammersmith, 2017, p. 105). Rather than an exclusive reliance on Supplemental Security Income, beginning at age 65, policies to accumulate assets earlier may actually result in lower overall costs. There are several strengths of this study, but we also note its limitations. First, health limitations are based on fairly simple self-reported measures making it is difficult to determine the nature and severity of women’s health problems. The strength of the data is the many measurement occasions encompassing both retrospective and prospective data, but the measures are basic indicators of health limitations and susceptible to reporting bias and measurement error. In addition, as we noted earlier, the data are unable to capture health problems that were resolved before the beginning of the study, which we may lead to an underestimate of the health gradient in wealth. Second, the 36-year duration of the NLSMW is extraordinary, but the study’s discontinuation in 2003 constrains generalizability. Given the demographic trends noted earlier, we speculate that NLSMW estimates of the effect of women’s health limitations on wealth are an underestimate of the current and future reality. Thus, we invite future studies to confirm or refute that interpretation. Third, growth curve models allow respondents to contribute observations even if they missed one or more measurement occasions; however, estimates are still susceptible to selection bias if attrition is selective on both health and wealth. Despite these limitations, we believe that findings from this study contribute to a better understanding of the health–wealth relationship. The results from our analyses reveal that the effect of health limitations on household wealth is substantial, especially when the onset of health limitations occurs early in a woman’s life. In order to prevent poverty among the older adult population, we recommend greater attention to identifying, and compensating for, the influence of women’s health limitations on wealth. Funding Support for this research was provided by the National Institute on Aging (T32AG025671-02 and R01AG043544) and the Purdue Center on Aging and the Life Course. Conflict of Interest None reported. Acknowledgments The authors thank Ann Howell, Patricia Morton, Markus Schafer, and Dario Spini for helpful comments on an earlier draft of this manuscript. References Andrew, M. & Ruel, E. ( 2010). Intergenerational health selection and wealth: A first look at parents’ health events and inter vivos financial transfers. Social Science Research , 39, 1126– 1136. doi: 10.1016/j.ssresearch.2010.06.004 Google Scholar CrossRef Search ADS   Baker, A. C., West, S., & Wood, A. ( 2017). Asset depletion, chronic financial stress, and mortgage trouble among older female homeowners. The Gerontologist , gnx137, doi: 10.1093/geront/gnx137 Brown, S. L., & Lin, I. F. ( 2012). The gray divorce revolution: Rising divorce among middle-aged and older adults, 1990-2010. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 67, 731– 741. doi: 10.1093/geronb/gbs089 Google Scholar CrossRef Search ADS   Burbidge, J. B., Magee, L., & Robb, A. L. ( 1988). Alternative transformations to handle extreme values of the dependent variable. Journal of the American Statistical Association , 83, 123– 127. doi: 10.1080/01621459.1988.10478575 Google Scholar CrossRef Search ADS   Cai, L., Mavromaras, K., & Oguzoglu, U. ( 2014). The effects of health status and health shocks on hours worked. Health Economics , 23, 516– 528. doi: 10.1002/hec.2931 Google Scholar CrossRef Search ADS PubMed  Conley, D., & Glauber, R. ( 2007). Gender, body mass, and socioeconomic status: New evidence from the PSID. Advances in Health Economics and Health Services Research , 17, 253– 275. doi:10.1016/S0731-2199(06)17010-7 Google Scholar CrossRef Search ADS PubMed  Duncan, O. D. ( 1961). A socioeconomic index for all occupations. In A. J. Reiss, Jr. (Ed.), Occupations and social status  (pp. 109– 138). New York: Free Press. Elder, G. H. Jr. ( 1998). The life course as developmental theory. Child Development , 69, 1– 12. doi:10.1111/j.1467-8624.1998.tb06128.x Google Scholar CrossRef Search ADS PubMed  Ellis, C. D., Munnell, A. H., & Eschtruth, A. D. ( 2014). Falling short: The coming retirement crisis and what to do about it . New York: Oxford University Press. Google Scholar CrossRef Search ADS   Ferraro, K. F., & Shippee, T. P. ( 2009). Aging and cumulative inequality: How does inequality get under the skin? The Gerontologist , 49, 333– 343. doi: 10.1093/geront/gnp034 Google Scholar CrossRef Search ADS PubMed  Ferraro, K. F., Shippee, T. P., & Schafer, M. H. ( 2009). Cumulative inequality theory for research on aging and the life course. In V. L. Bengtson, D. Gans, N. M. Putney, & M. Silverstein (Eds.), Handbook of theories of aging  ( 2nd ed., pp. 413– 433). New York: Springer. Garbaski, D. ( 2014). The interplay between child and maternal health: Reciprocal relationships and cumulative disadvantage during childhood and adolescence. Journal of Health and Social Behavior , 55, 91– 106. doi: 10.1177/0022146513513225 Google Scholar CrossRef Search ADS PubMed  Gee, G., & Walsemann, K. ( 2009). Does health predict the reporting of racial discrimination or do reports of discrimination predict health? Findings from the National Longitudinal Study of Youth. Social Science & Medicine , 68, 1676– 1684. doi: 10.1016/ j.socscimed.2009.02.002 Google Scholar CrossRef Search ADS   Haas, S. A. ( 2006). Health selection and the process of social stratification: The effect of childhood health on socioeconomic attainment. Journal of Health and Social Behavior , 47, 339– 354. doi: 10.1177/002214650604700403 Google Scholar CrossRef Search ADS PubMed  Hallqvist, J., Lynch, J., Bartley, M., Lang, T., & Blane, D. ( 2004). Can we disentangle life course processes of accumulation, critical period and social mobility? An analysis of disadvantaged socio-economic positions and myocardial infarction in the Stockholm Heart Epidemiology Program. Social Science & Medicine , 58, 1555– 1562. doi: 10.1016/S0277-9536(03)00344-7 Google Scholar CrossRef Search ADS   Hammarström, A., Stenlund, H., & Janlert, U. ( 2011). Mechanisms for the social gradient in health: Results from a 14-year follow-up of the Northern Swedish Cohort. Public Health , 125, 567– 576. doi: 10.1016/j.puhe.2011.06.010 Google Scholar CrossRef Search ADS PubMed  Heckman, J. J. ( 1979). Sample selection bias as a specification error. Econometrica , 47, 153– 161. doi: 10.2307/1912352 Google Scholar CrossRef Search ADS   Himmelstein, D. U., Thorne, D., Warren, E., & Woolhandler, S. ( 2009). Medical bankruptcy in the United States, 2007: Results of a national study. The American Journal of Medicine , 122, 741– 746. doi: 10.1016/j.amjmed.2009.04.012 Google Scholar CrossRef Search ADS PubMed  Imai, K., Keele, L., & Tingley, D. ( 2010). A general approach to causal mediation analysis. Psychological Methods , 15, 309– 334. doi: 10.1037/a0020761 Google Scholar CrossRef Search ADS PubMed  Jackson, M. I. ( 2009). Understanding links between adolescent health and educational attainment. Demography , 46, 671– 694. doi:10.1353/dem.0.0078 Google Scholar CrossRef Search ADS PubMed  Kahn, J. R., & Pearlin, L. I. ( 2006). Financial strain over the life course and health among older adults. Journal of Health and Social Behavior , 47, 17– 31. doi: 10.1177/002214650604700102 Google Scholar CrossRef Search ADS PubMed  Kane, J. B., Philip Morgan, S., Harris, K. M., & Guilkey, D. K. ( 2013). The educational consequences of teen childbearing. Demography , 50, 2129– 2150. doi: 10.1007/s13524-013-0238-9 Google Scholar CrossRef Search ADS PubMed  Lachance-Grzela, M. & Bouchard, G. ( 2010). Why do women do the lion’s share of housework? A decade of research. Sex Roles , 63, 767– 780. doi: 10.1007/s11199-010-9797-z Google Scholar CrossRef Search ADS   Lin, I. F., Brown, S. L., & Hammersmith, A. M. ( 2017). Marital biography, social security receipt, and poverty. Research on Aging , 39, 86– 110. doi: 10.1177/0164027516656139 Google Scholar CrossRef Search ADS PubMed  Lundborg, P., Nilsson, A., & Rooth, D. O. ( 2014). Adolescent health and adult labor market outcomes. Journal of Health Economics , 37, 25– 40. doi: 10.1016/j.jhealeco.2014.05.003 Google Scholar CrossRef Search ADS PubMed  Lynch, J. L., & von Hippel, P. T. ( 2016). An education gradient in health, a health gradient in education, or a confounded gradient in both? Social Science & Medicine , 154, 18– 27. doi: 10.1016/ j.socscimed.2016.02.029 Google Scholar CrossRef Search ADS   Marmot, M. G., Shipley, M. J., & Rose, G. ( 1984). Inequalities in death—specific explanations of a general pattern? Lancet , 1, 1003– 1006. doi: 10.1016/S0140-6736(84)92337-7 Google Scholar CrossRef Search ADS PubMed  Marmot, M. G., Smith, G. D., Stansfeld, S., Patel, C., North, F., Head, J., … Feeney, A. ( 1991). Health inequalities among British civil servants: The Whitehall II study. Lancet , 337, 1387– 1393. doi:10.1016/0140-6736(91)93068-K Google Scholar CrossRef Search ADS PubMed  Michaud, P. C., & van Soest, A. ( 2008). Health and wealth of elderly couples: Causality tests using dynamic panel data models. Journal of Health Economics , 27, 1312– 1325. doi: 10.1016/ j.jhealeco.2008.04.002 Google Scholar CrossRef Search ADS PubMed  Modigliani, F. ( 1988). The role of intergenerational transfers and life cycle saving in the accumulation of wealth. Journal of Economic Perspectives , 2, 15– 40. doi: 10.1257/jep.2.2.15 Google Scholar CrossRef Search ADS   Painter, M., Frech, A., & Williams, K. ( 2015). Nonmarital fertility, union history, and women’s wealth. Demography , 52, 153– 182. doi: 10.1007/s13524-014-0367-9 Google Scholar CrossRef Search ADS PubMed  Schinkel-Ivy, A., Mosca, I., & Mansfield, A. ( 2017). Factors contributing to unexpected retirement and unemployment in adults over 50 years old in Ireland. Gerontology & Geriatric Medicine , 3, 2333721417722709. doi: 10.1177/2333721417 722709 Google Scholar PubMed  Smith, J. P. ( 1999). Healthy bodies and thick wallets: The dual relation between health and economic status. Journal of Economic Perspectives , 13, 145– 166. doi: 10.1257/jep.13.2.145 Google Scholar CrossRef Search ADS   Smith, J. P. ( 2004). Unraveling the SES-health connection. Population and Development Review , 30, 108– 132. Google Scholar CrossRef Search ADS   U.S. Department of Labor. ( 2001). NLS of mature women user’s guide . Retrieved from https://www.bls.gov/nls/mwguide/2001/nlsmwg0.pdf Wilmoth, J. & Koso, G. ( 2002). Does marital history matter? Marital status and wealth outcomes among preretirement adults. Journal of Marriage and Family , 64, 254– 268. doi: 10.1111/j.1741-3737.2002.00254.x Google Scholar CrossRef Search ADS   Wu, S. ( 2003). The effect of health events on the economic status of married couples. Journal of Human Resources , 38, 219– 230. doi: 10.2307/1558762 Google Scholar CrossRef Search ADS   Yamokoski, A. & Keister, L. A. ( 2006). The wealth of single women: Marital status and parenthood in the asset accumulation of young baby boomers in the United States. Feminist Economics , 12, 167– 194. doi: 10.1080/13545700500508478 Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Gerontologist Oxford University Press

Wealth in Middle and Later Life: Examining the Life Course Timing of Women’s Health Limitations

Loading next page...
 
/lp/ou_press/wealth-in-middle-and-later-life-examining-the-life-course-timing-of-zPAn0zHNcr
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0016-9013
eISSN
1758-5341
D.O.I.
10.1093/geront/gny048
Publisher site
See Article on Publisher Site

Abstract

Abstract Background and Objectives Guided by cumulative inequality theory, this study poses two main questions: (a) Does women’s poor health compromise household financial assets? (b) If yes, is wealth sensitive to the timing of women’s health limitations? In addressing these questions, we consider the effect of health limitations on wealth at older ages, as well as examine how health limitations influence wealth over particular segments of the life course, giving attention to both the onset and duration of health limitations. Research Design and Methods Using 36 years of data from the National Longitudinal Survey of Mature Women, piecewise growth curve and linear regression models were used to estimate the effects of life course timing and duration of health limitations on household wealth. Results The findings reveal that women who experienced health limitations accumulated substantially less wealth over time, especially if the health limitations were manifest during childhood or early adulthood. Discussion and Implications This study identifies how early-life health problems lead to less wealth in later life. Women’s health, Financial assets, Health selection, Cumulative inequality The relationship between health and wealth has garnered attention in the last decade as researchers have sought to explicate the ways in which these life domains influence one another over time. Although most research has focused on how income and, to a lesser extent, wealth influence health, there is growing evidence that health also influences income and wealth. For instance, individuals in poor health face both social and economic challenges, and these may have important implications for educational attainment, labor force participation, and even retirement decisions (Cai, Mavromaras, & Oguzoglu, 2014; Lundborg, Nilsson, & Rooth, 2014; Lynch & von Hippel, 2016; Schinkel-Ivy, Mosca, & Mansfield, 2017). By contrast, those in good health generally have less complicated paths to status attainment and greater opportunity to accumulate assets. A life course approach to identifying the relationships between health and wealth gives priority to timing in a person’s life (Elder, 1998). There is substantial variability in the timing of health problems (both onset and duration) and this variability may have financial repercussions. Although many people experience health problems of relatively short duration, others may struggle with illness over long periods of time. Additionally, some individuals are disadvantaged from birth or encounter health problems early in life while others are not faced with these problems until advanced age. This variability in the timing of health problems may have a major influence on wealth accumulation; however, we are unaware of any studies that systematically examine how the timing of health problems influences wealth accumulation over a large span of the life course. Based on the life course perspective and cumulative inequality theory, this study examines the long-term consequences of women’s health limitations on household wealth. Our research questions are twofold: Does women’s poor health compromise household financial assets? If yes, is wealth sensitive to the timing of women’s health limitations? We address these and related questions with data from the National Longitudinal Survey of Mature Women (NLSMW), a 36-year panel study. Women’s Health and Status Attainment Empirical research has long established a strong relationship between health and socioeconomic status (SES), but debate over the precise nature of this association persists. Evidence in support of social causation—the effect of social status on health—dates back nearly three decades to the Whitehall study of British civil servants in which Marmot, Shipley, and Rose (1984) revealed an inverse relationship between employment grade and mortality over 10 years. Since then, the study has expanded to women and examined morbidity and a range of health outcomes showing compelling evidence in support of social causation (Marmot et al., 1991). Moreover, it stimulated scores of studies examining how SES affects various dimensions of health (e.g., Hammarström, Stenlund, & Janlert, 2011). Comparatively less attention, however, has been paid to health selection—the effect of health on status attainment. Health selection refers to the process through which good health leads to opportunity or, conversely, poor health leads to other risks. Despite the predominant focus on the social gradient, studies have revealed several ways in which life course transitions can be complicated by poor health and lead to lower status attainment, including reducing the likelihood that one will enroll in college (Jackson, 2009) and impinging on one’s ability to work—even forcing early retirement (Wu, 2003)—which may result in losses to both income and wealth (Smith, 1999, 2004). Although the costs of health selection are considerable, the effect of health on wealth has been among the largest across life stages: Haas (2006) found that excellent childhood health was associated with 104% greater wealth in adulthood than poor childhood health; Smith (2004) showed that among adults aged 51 years and older a new health problem reduces wealth by as much as $50,000 over an 8-year period. Few studies, however, examine this phenomenon separately for women. This is important given that women not only experience a higher prevalence of disease and disability compared to men, but their health may influence wealth in distinct ways. For example, marriage provides a wealth advantage (Wilmoth & Koso, 2002; Yamokoski & Keister, 2006), but less healthy individuals are less likely to marry, and this selection effect is often stronger for women (Conley & Glauber, 2007). Childbearing also uniquely affects women, which is tightly intertwined with women’s long-term health and status attainment. Early motherhood, in particular, contributes to lower status attainment (Kane, Morgan, Harris, & Guilkey, 2013), and when experienced outside marriage, the consequences for wealth are even greater (Painter, Frech, & Williams, 2015). Later in life, new health problems among older women, but not older men, increase the risk of divorce, and divorce is especially detrimental to women’s wealth (Wilmoth & Koso, 2002; Yamokoski & Keister, 2006). The majority of women today work while also performing a disproportionate share of household labor (Lachance-Grzela & Bouchard, 2010). Compared to men, the onset of a new health problem may contribute to a greater overall reduction in labor, both at home and at work. Indeed, Wu (2003) showed that women’s health problems are actually quite consequential to household wealth—even more so than men’s—possibly due to women’s greater share of nonmarket labor in the household. To our knowledge, this is one of the first studies to investigate the effects of women’s health limitations at each stage of the life course, as can be done with the NLSMW. In this way, the study contributes not only to the study of women’s role in wealth accumulation but also to the study of the life course and women’s financial vulnerability in later life (Baker, West, & Wood, 2017). An assessment of the health selection hypothesis, however, needs to account for parental SES, especially if one is examining the early onset of health limitations. Among those who experience serious health events, especially in early life, financial support from parents may curtail some of the costs to status attainment. Findings indicate that intergenerational financial transfers may account for as much as one-fourth of wealth accumulation (Modigliani, 1988). Additionally, because health problems are more common among those of lower socioeconomic status, we risk overstating the relationship between health limitations and wealth if we fail to model parental SES. Theoretical Framework In addition to empirical research, this study draws on cumulative inequality theory to guide our understanding of the relationship between women’s health and household wealth. First, the theory specifies that disadvantage increases exposure to risk, which may result in further disadvantage (Ferraro & Shippee, 2009). When applied to the study of health and wealth, those in poor health may be more susceptible to risks, including higher medical costs and the inability to work, which can lead to less wealth accumulation over time. In this sense, health status shapes the mechanisms that are related to financial accumulation. Whereas women have historically been more socioeconomically disadvantaged compared to men, one anticipates that it is important to consider how health problems may contribute to women’s additional disadvantages encountered at various points in the life course. Cumulative inequality theory also considers the importance of timing and duration in shaping trajectories of inequality (Ferraro & Shippee, 2009). The theory specifies that long-term exposure to disadvantage is more consequential than short-term exposure, and that trajectories are influenced not only by the amount of exposure, but also by the timing of events. For example, earlier onset of health problems may lead to a host of outcomes not encountered otherwise, including a lower probability of marrying, fewer years of schooling, and interrupted labor force participation. In the present study, we examine whether the life course timing and duration of health limitations are related to household wealth. Health limitations are a type of disadvantage, exposing people to additional risks, and the longer one faces such exposure, the greater the anticipated effect on wealth. Finally, it is well known that there is a strong intergenerational component to the accumulation of wealth and, according to the theory, “family lineage is a key source of life course inequality, especially for the early stages of the life course” (Ferraro, Shippee, & Schafer, 2009, p. 418). Although the NLSMW does not have wealth data on the respondent’s parents, SES is available. We expect that high parental SES is associated with greater household wealth—evidence of an intergenerational transmission of inequality—and that health limitations influence this relationship. Life Course Analysis of Health and Wealth The present study is distinct from prior investigations of health selection in several important ways. First, it draws on 36 years of data from the NLSMW. These data follow women from early middle adulthood (30–44 years of age) to older age (66–80) and include measures of health and SES at multiple points in time. We combine both retrospective and prospective data on health limitations to construct a comprehensive history of health limitations from childhood to older ages. Moreover, these data permit an examination of the health–wealth relationship among women, who have been studied less frequently than men. Second, this is an extension of studies which examine health selection at a single stage of the life course. Although the importance of both early- and later-life health on wealth has been shown repeatedly, few studies investigate the effect of health limitations across life stages. Adding this dimension is important given that (a) health problems may have their origins in earlier life course stages, (b) beyond their initial impact, health problems may have a prolonged influence on status attainment, and (c) the onset of health problems may have differential effects on wealth over time (Smith, 1999, 2004). Using women’s health histories, we can evaluate the long-term financial repercussions. Third, wealth is a household indicator, and the NLSMW enables us to consider important household and family influences on wealth. Unlike many previous studies, we include information on parental SES in estimating the effect of health limitations on wealth. Design and Methods Sample The NLSMW collected 21 waves of data over 36 years (1967–2003). Multistage probability sampling was used to draw a sample of 5,083 civilian, noninstitutionalized women aged 30–44 years in 1967, with an oversample of black women (n = 1,390). Respondents were resurveyed approximately every 2 years, and most of the interviewing was conducted in-person. By the conclusion of the study in 2003, the women were 66–80 years of age. The final survey yielded 2,237 respondents or 44% of the original sample. Attrition in the NLSMW was largely due to mortality (29%) and refusal (20%). Compared to the stayers, the attritors were, on average, older and less educated. They also reported less household wealth and lower self-ratings of health. We explicitly address attrition with our analytic methods, described below, that use either a Heckman selection model (Heckman, 1979) or a growth curve analysis so that respondents can contribute to the estimation of growth curves regardless of whether they missed one or more measurement occasions. Measures Household Wealth Respondents were asked detailed questions related to household assets and debts, including but not limited to housing, mortgages, savings, and loans. Debts for each household were summed and then subtracted from total household assets by the data investigators. Wealth was bottom-coded at $-999,999 and top-coded at $999,999 in most, but not all, survey years to ensure respondent confidentiality. Measures were top-coded at this value in the growth curves only to maintain consistency across survey waves. Wealth was also inflation-adjusted to 2003 dollars using the Consumer Price Index. Health Limitations The NLSMW provides both retrospective and prospective information on health limitations that enable us to trace women’s health limitations back to childhood—and construct a measure of lifetime health limitations. Beginning in 1967 at the initial wave, women were asked a series of questions about whether their health or physical condition limited what they could do. We created a binary variable at all but two survey waves (health limitations were not available in 1968 or 1969) to measure women’s health limitations from any of four activities: working at all, amount or kind of work one can do, housework, and other activities. In addition, if respondents reported any limitations due to their health at the initial wave, they were asked, “How long have you been limited in this way?” Prior health limitations were reported in years and in months. Although the retrospective data provide information on health limitations prior to the initial survey, this is true only for health problems manifest in 1967. Early-life health limitations that subsided by 1967 when women were 30–44 years of age were not captured by the study, resulting in conservative estimates for childhood- and early-adult health problems. Similar measures are used by other studies (e.g., Garbaski, 2014; Gee & Walsemann, 2009). We combine the retrospective and prospective indicators to construct two sets of measures: timing and duration of health limitations. Timing or onset is measured using a system of binary variables that represent the life stage during which a health limitation was first reported: childhood (<18), early adulthood (18–34), early middle age (35–49), late middle age (50–64), and older age (65+). Other studies use similar ages to differentiate life stages (e.g., Kahn & Pearlin, 2006). By contrast, duration is the summation of years in which women reported health limitations (top-coded at the top 2% due to the small number of responses in higher categories). Specifically, we added together the number of years women retrospectively reported living with a health limitation at baseline and the number of survey waves with an affirmative response—treating health limitations in consecutive survey waves as uninterrupted episodes. We examined alternative strategies to coding health limitations and obtained similar results. Personal and Household Characteristics A host of personal and household characteristics are available to adjust our linear regression estimates. Chronological age is measured in years. Black is a binary variable coded 1 for black and 0 for non-black respondents. Continuous measures capture the proportion of time during the study women spent married, working full-time and part-time, and living with a spouse who had health problems. On spousal variables, nonmarried respondents were coded 0. The number of children belonging to each respondent is a count variable measured at baseline when women were aged 30–44 (top-coded at the top 2%). Education was measured by years of schooling completed. Both the respondent and her parent’s occupational prestige were measured using the Duncan Socioeconomic Index (SEI; Duncan, 1961). Duncan SEI consists of prestige ratings for occupations, ranging from 0 to 97, based on their income and education distributions (U.S. Department of Labor, 2001). The coding and descriptive statistics for the variables are displayed in Table 1. Table 1. Means and Standard Deviations of Variables in the NLSMW, 1967–2003   Description  Range  Mean  SD  Dependent Variable  Household Wealth, 1967a  In thousands  −99.999–482.500  10.037  22.102  Household Wealth, 2003  In thousands  −99.999–4109.665  235.937  377.782  Independent Variables  Duration of Health Limitations  Duration  Years  0–44  8.276  10.478  Timing of Health Limitationsb  Childhood  0–17 years  0/1  0.018  c  Early Adulthood  18–34 years  0/1  0.102    Early Middle Age  35–49 years  0/1  0.342    Late Middle Age  50–64 years  0/1  0.185    Older Age  65+ years  0/1  0.069    Personal and Household Characteristics  Age, 1967  Years  30–44  37.230  4.358  Black, 1967  1 = black, 0 = non-black  0/1  0.280    Married, 1967–2003d  Proportion of time  0–1  0.689  0.379  Husband’s Health Limitations, 1967–2003d  Proportion of time  0–1  0.147  0.231  Children, 1967  Number of children  0–9  3.160  2.205  Education, 1967  Years of education  0–18  10.918  2.842  Parent’s Prestige, 1967  Duncan SEI  4–96  26.415  21.542  Respondent’s Prestige, 1967  Duncan SEI  0–93  33.763  21.116  Full-Time Employment, 1967–2003d  Proportion of time  0–1  0.288  0.291  Part-Time Employment, 1967–2003d  Proportion of time  0–1  0.156  0.197    Description  Range  Mean  SD  Dependent Variable  Household Wealth, 1967a  In thousands  −99.999–482.500  10.037  22.102  Household Wealth, 2003  In thousands  −99.999–4109.665  235.937  377.782  Independent Variables  Duration of Health Limitations  Duration  Years  0–44  8.276  10.478  Timing of Health Limitationsb  Childhood  0–17 years  0/1  0.018  c  Early Adulthood  18–34 years  0/1  0.102    Early Middle Age  35–49 years  0/1  0.342    Late Middle Age  50–64 years  0/1  0.185    Older Age  65+ years  0/1  0.069    Personal and Household Characteristics  Age, 1967  Years  30–44  37.230  4.358  Black, 1967  1 = black, 0 = non-black  0/1  0.280    Married, 1967–2003d  Proportion of time  0–1  0.689  0.379  Husband’s Health Limitations, 1967–2003d  Proportion of time  0–1  0.147  0.231  Children, 1967  Number of children  0–9  3.160  2.205  Education, 1967  Years of education  0–18  10.918  2.842  Parent’s Prestige, 1967  Duncan SEI  4–96  26.415  21.542  Respondent’s Prestige, 1967  Duncan SEI  0–93  33.763  21.116  Full-Time Employment, 1967–2003d  Proportion of time  0–1  0.288  0.291  Part-Time Employment, 1967–2003d  Proportion of time  0–1  0.156  0.197  Note: NLSMW = National Longitudinal Survey of Mature Women. aWealth was measured on 12 occasions during the study; the first and last measurements are displayed. bTiming refers to the onset of a health limitation; reference group = no health limitations (28.4% of respondents). cThe standard deviation of a dichotomous variable is omitted because it is a function of the mean. dTime-varying covariate in the growth curve analysis. View Large Table 1. Means and Standard Deviations of Variables in the NLSMW, 1967–2003   Description  Range  Mean  SD  Dependent Variable  Household Wealth, 1967a  In thousands  −99.999–482.500  10.037  22.102  Household Wealth, 2003  In thousands  −99.999–4109.665  235.937  377.782  Independent Variables  Duration of Health Limitations  Duration  Years  0–44  8.276  10.478  Timing of Health Limitationsb  Childhood  0–17 years  0/1  0.018  c  Early Adulthood  18–34 years  0/1  0.102    Early Middle Age  35–49 years  0/1  0.342    Late Middle Age  50–64 years  0/1  0.185    Older Age  65+ years  0/1  0.069    Personal and Household Characteristics  Age, 1967  Years  30–44  37.230  4.358  Black, 1967  1 = black, 0 = non-black  0/1  0.280    Married, 1967–2003d  Proportion of time  0–1  0.689  0.379  Husband’s Health Limitations, 1967–2003d  Proportion of time  0–1  0.147  0.231  Children, 1967  Number of children  0–9  3.160  2.205  Education, 1967  Years of education  0–18  10.918  2.842  Parent’s Prestige, 1967  Duncan SEI  4–96  26.415  21.542  Respondent’s Prestige, 1967  Duncan SEI  0–93  33.763  21.116  Full-Time Employment, 1967–2003d  Proportion of time  0–1  0.288  0.291  Part-Time Employment, 1967–2003d  Proportion of time  0–1  0.156  0.197    Description  Range  Mean  SD  Dependent Variable  Household Wealth, 1967a  In thousands  −99.999–482.500  10.037  22.102  Household Wealth, 2003  In thousands  −99.999–4109.665  235.937  377.782  Independent Variables  Duration of Health Limitations  Duration  Years  0–44  8.276  10.478  Timing of Health Limitationsb  Childhood  0–17 years  0/1  0.018  c  Early Adulthood  18–34 years  0/1  0.102    Early Middle Age  35–49 years  0/1  0.342    Late Middle Age  50–64 years  0/1  0.185    Older Age  65+ years  0/1  0.069    Personal and Household Characteristics  Age, 1967  Years  30–44  37.230  4.358  Black, 1967  1 = black, 0 = non-black  0/1  0.280    Married, 1967–2003d  Proportion of time  0–1  0.689  0.379  Husband’s Health Limitations, 1967–2003d  Proportion of time  0–1  0.147  0.231  Children, 1967  Number of children  0–9  3.160  2.205  Education, 1967  Years of education  0–18  10.918  2.842  Parent’s Prestige, 1967  Duncan SEI  4–96  26.415  21.542  Respondent’s Prestige, 1967  Duncan SEI  0–93  33.763  21.116  Full-Time Employment, 1967–2003d  Proportion of time  0–1  0.288  0.291  Part-Time Employment, 1967–2003d  Proportion of time  0–1  0.156  0.197  Note: NLSMW = National Longitudinal Survey of Mature Women. aWealth was measured on 12 occasions during the study; the first and last measurements are displayed. bTiming refers to the onset of a health limitation; reference group = no health limitations (28.4% of respondents). cThe standard deviation of a dichotomous variable is omitted because it is a function of the mean. dTime-varying covariate in the growth curve analysis. View Large The growth curve analysis also includes these control variables, but three of them are treated as time-varying covariates: marital status, husband’s health, and employment. In the piecewise growth curve models, marital status distinguishes between married, widowed, divorced/separated, and never married (reference group) respondents. Husband’s health is coded 1 for any health problem or condition that limits his ability to work and 0, otherwise. Employment status is measured using binary variables for full-time and part-time employment. Analytic Plan Two research questions are addressed with multiple analytic strategies, and we organize the analysis into two stages. We began with an ordinary least squares regression (OLS) analysis of the women who survived to 2003 to provide a straightforward test of our first research question. We applied a hyperbolic sine transformation to the dependent variable to handle the skewed continuous distribution of wealth, which also includes negative values and values of zero (Burbidge, Magee, & Robb, 1988). In supplementary analyses, we tested alternative transformations and the substantive conclusions were unchanged. We also adjusted measures for the number of household members and reached similar conclusions. To account for attrition due to mortality, we estimated a Heckman selection model, using unique information in the selection equation (e.g., self-rated health, requires help with personal care, and parental life status), and included the variable expressing the nonselection hazard (inverse of Mills’ ratio) in our linear regression analysis. We also examined other sources of attrition, including nonresponse, but the conclusions were the same. After accounting for selective mortality, these analyses permit one to gauge the effect of women’s health limitations on wealth accumulation in 2003. We also incorporate tests for differences in the onset of women’s health limitations—from childhood to later life—to discern their impact on wealth when most of the women were retired. The second stage of the analysis consisted of piecewise growth curve models. The piecewise model enables us to capitalize on available data between 1967 and 2003 by allowing each woman to contribute as many observations as possible, even those who do not survive to the final wave. We model trajectories of wealth from 12 occasions and examine the effect of health limitations at various stages of the life course on the intercept and rate of growth in wealth over time, while modeling separate slopes by life course stage. One advantage of the piecewise model is that it allows us to include onset predictors only for slopes in ages that are causally prior. From a policy vantage point, these analyses provide a window into key life course periods when the effects of health limitations are having their greatest effect. We used age as the time metric in the piecewise growth curve models. The variable for age was broken into three pieces: age 34–49, 49–65, and 65–78, and centered at age 34, which allows us to interpret the intercept as women’s household wealth at age 34, as well as examine the rate of change over particular segments of the life course as these women grow older. Observations with an age below 34 and an age above 78 were dropped from the growth curve analysis because there were far fewer of these compared to other ages. This also makes the splines more balanced in terms of the number of years each piece covers. In addition, we mean-centered education, occupational prestige, and number of children to provide a more meaningful constant to interpret. Results As shown in Table 1, the mean household wealth expressed in thousands during 1967 (when the women were 30–44) was 10.037 or $10,037 inflation-adjusted dollars. By comparison, the mean household wealth in 2003 (when the women were 66–80) was 235.937 or $235,937 dollars. Women in the sample were, on average, 37 years of age in 1967 and more than one-quarter of respondents had not reported any health limitations by 2003 (28%). The majority of women, however, first reported health limitations in early middle age (34%), with the average duration of health limitations equal to eight years. For the first stage of the analysis, Table 2 presents the results from four OLS regression models, with variables introduced in blocks, to address whether women’s health limitations are associated with less household wealth in later life. Model 1 examines personal and household characteristics in 1967 on wealth accumulation (while adjusting for selective mortality). Both the respondent and her parent’s occupational prestige had positive effects on household wealth, although the effect of the respondent’s own occupational prestige was stronger. We tested interactions between parental occupational prestige and women’s health limitations, but none were significant (results not shown). Table 2. OLS Regression of Accumulated Household Wealth in 2003   Model 1  Model 2  Model 3  Model 4  Personal and Household Characteristics  Age, 1967  0.023a  0.023  0.005  0.028  (0.023)  (0.023)  (0.023)  (0.022)  Black, 1967  −1.081***  −1.165***  −1.124***  −0.488*  (0.209)  (0.204)  (0.206)  (0.203)  Children, 1967  −0.039  −0.019  −0.028  −0.071*  (0.037)  (0.036)  (0.037)  (0.035)  Education, 1967  0.098*  0.066  0.082*  0.078*  (0.038)  (0.037)  (0.037)  (0.035)  Parent’s Prestige, 1967  0.010**  0.009**  0.008*  0.008*  (0.004)  (0.003)  (0.003)  (0.003)  Respondent’s Prestige, 1967  0.018***  0.014**  0.016***  0.014**  (0.004)  (0.004)  (0.004)  (0.004)  Mortality Selection  −1.568**  −1.175**  −1.193**  −1.171**  (0.455)  (0.447)  (0.451)  (0.427)  Duration of Health Limitations  Duration    −0.073***        (0.018)      Duration2    0.001        (0.000)      Timing of Health Limitationsb   Childhood      −2.648***  −2.127***      (0.558)  (0.531)   Early Adulthood      −1.355***  −0.986***      (0.285)  (0.274)   Early Middle Age      −0.876***  −0.607**      (0.198)  (0.192)   Late Middle Age      −0.621**  −0.421*      (0.210)  (0.199)   Older Age      −0.598*  −0.563*      (0.262)  (0.246)  Life Course Events  Married, 1967–2003c        2.319***        (0.216)  Husband’s Health Limitations, 1967–2003c        −0.788*        (0.334)  Full-Time Employment, 1967–2003c        0.342        (0.295)  Part-Time Employment, 1967–2003c        0.218        (0.380)  Constant  3.602***  4.311***  4.910***  2.466**  (0.727)  (0.714)  (0.749)  (0.767)  Adjusted R2  0.282  0.322  0.310  0.394  N  902  902  902  902    Model 1  Model 2  Model 3  Model 4  Personal and Household Characteristics  Age, 1967  0.023a  0.023  0.005  0.028  (0.023)  (0.023)  (0.023)  (0.022)  Black, 1967  −1.081***  −1.165***  −1.124***  −0.488*  (0.209)  (0.204)  (0.206)  (0.203)  Children, 1967  −0.039  −0.019  −0.028  −0.071*  (0.037)  (0.036)  (0.037)  (0.035)  Education, 1967  0.098*  0.066  0.082*  0.078*  (0.038)  (0.037)  (0.037)  (0.035)  Parent’s Prestige, 1967  0.010**  0.009**  0.008*  0.008*  (0.004)  (0.003)  (0.003)  (0.003)  Respondent’s Prestige, 1967  0.018***  0.014**  0.016***  0.014**  (0.004)  (0.004)  (0.004)  (0.004)  Mortality Selection  −1.568**  −1.175**  −1.193**  −1.171**  (0.455)  (0.447)  (0.451)  (0.427)  Duration of Health Limitations  Duration    −0.073***        (0.018)      Duration2    0.001        (0.000)      Timing of Health Limitationsb   Childhood      −2.648***  −2.127***      (0.558)  (0.531)   Early Adulthood      −1.355***  −0.986***      (0.285)  (0.274)   Early Middle Age      −0.876***  −0.607**      (0.198)  (0.192)   Late Middle Age      −0.621**  −0.421*      (0.210)  (0.199)   Older Age      −0.598*  −0.563*      (0.262)  (0.246)  Life Course Events  Married, 1967–2003c        2.319***        (0.216)  Husband’s Health Limitations, 1967–2003c        −0.788*        (0.334)  Full-Time Employment, 1967–2003c        0.342        (0.295)  Part-Time Employment, 1967–2003c        0.218        (0.380)  Constant  3.602***  4.311***  4.910***  2.466**  (0.727)  (0.714)  (0.749)  (0.767)  Adjusted R2  0.282  0.322  0.310  0.394  N  902  902  902  902  Note: OLS = Ordinary least squares regression analysis. aUnstandardized regression coefficient (standard error in parentheses). bTiming refers to the onset of a health limitation; reference group = no health limitations. cProportion of time in the study period. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Table 2. OLS Regression of Accumulated Household Wealth in 2003   Model 1  Model 2  Model 3  Model 4  Personal and Household Characteristics  Age, 1967  0.023a  0.023  0.005  0.028  (0.023)  (0.023)  (0.023)  (0.022)  Black, 1967  −1.081***  −1.165***  −1.124***  −0.488*  (0.209)  (0.204)  (0.206)  (0.203)  Children, 1967  −0.039  −0.019  −0.028  −0.071*  (0.037)  (0.036)  (0.037)  (0.035)  Education, 1967  0.098*  0.066  0.082*  0.078*  (0.038)  (0.037)  (0.037)  (0.035)  Parent’s Prestige, 1967  0.010**  0.009**  0.008*  0.008*  (0.004)  (0.003)  (0.003)  (0.003)  Respondent’s Prestige, 1967  0.018***  0.014**  0.016***  0.014**  (0.004)  (0.004)  (0.004)  (0.004)  Mortality Selection  −1.568**  −1.175**  −1.193**  −1.171**  (0.455)  (0.447)  (0.451)  (0.427)  Duration of Health Limitations  Duration    −0.073***        (0.018)      Duration2    0.001        (0.000)      Timing of Health Limitationsb   Childhood      −2.648***  −2.127***      (0.558)  (0.531)   Early Adulthood      −1.355***  −0.986***      (0.285)  (0.274)   Early Middle Age      −0.876***  −0.607**      (0.198)  (0.192)   Late Middle Age      −0.621**  −0.421*      (0.210)  (0.199)   Older Age      −0.598*  −0.563*      (0.262)  (0.246)  Life Course Events  Married, 1967–2003c        2.319***        (0.216)  Husband’s Health Limitations, 1967–2003c        −0.788*        (0.334)  Full-Time Employment, 1967–2003c        0.342        (0.295)  Part-Time Employment, 1967–2003c        0.218        (0.380)  Constant  3.602***  4.311***  4.910***  2.466**  (0.727)  (0.714)  (0.749)  (0.767)  Adjusted R2  0.282  0.322  0.310  0.394  N  902  902  902  902    Model 1  Model 2  Model 3  Model 4  Personal and Household Characteristics  Age, 1967  0.023a  0.023  0.005  0.028  (0.023)  (0.023)  (0.023)  (0.022)  Black, 1967  −1.081***  −1.165***  −1.124***  −0.488*  (0.209)  (0.204)  (0.206)  (0.203)  Children, 1967  −0.039  −0.019  −0.028  −0.071*  (0.037)  (0.036)  (0.037)  (0.035)  Education, 1967  0.098*  0.066  0.082*  0.078*  (0.038)  (0.037)  (0.037)  (0.035)  Parent’s Prestige, 1967  0.010**  0.009**  0.008*  0.008*  (0.004)  (0.003)  (0.003)  (0.003)  Respondent’s Prestige, 1967  0.018***  0.014**  0.016***  0.014**  (0.004)  (0.004)  (0.004)  (0.004)  Mortality Selection  −1.568**  −1.175**  −1.193**  −1.171**  (0.455)  (0.447)  (0.451)  (0.427)  Duration of Health Limitations  Duration    −0.073***        (0.018)      Duration2    0.001        (0.000)      Timing of Health Limitationsb   Childhood      −2.648***  −2.127***      (0.558)  (0.531)   Early Adulthood      −1.355***  −0.986***      (0.285)  (0.274)   Early Middle Age      −0.876***  −0.607**      (0.198)  (0.192)   Late Middle Age      −0.621**  −0.421*      (0.210)  (0.199)   Older Age      −0.598*  −0.563*      (0.262)  (0.246)  Life Course Events  Married, 1967–2003c        2.319***        (0.216)  Husband’s Health Limitations, 1967–2003c        −0.788*        (0.334)  Full-Time Employment, 1967–2003c        0.342        (0.295)  Part-Time Employment, 1967–2003c        0.218        (0.380)  Constant  3.602***  4.311***  4.910***  2.466**  (0.727)  (0.714)  (0.749)  (0.767)  Adjusted R2  0.282  0.322  0.310  0.394  N  902  902  902  902  Note: OLS = Ordinary least squares regression analysis. aUnstandardized regression coefficient (standard error in parentheses). bTiming refers to the onset of a health limitation; reference group = no health limitations. cProportion of time in the study period. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Despite Models 2 and 3 being similar in design, they differ in their measurement of health limitations: Model 2 uses a measure of duration (along with a quadratic term), whereas Model 3 draws on a system of binary indicators that capture the life course timing of health limitations. Prior research documented the correlation between duration and timing and the difficulty in disentangling such measures (Hallqvist, Lynch, Bartley, Lang, & Blane, 2004). For that reason, we model the duration and timing of health limitations separately. First, results from Model 2 reveal a negative relationship between duration of exposure and accumulated wealth at older ages. By contrast, Model 3 presents the timing of health limitations by life course stage. Women who reported health limitations during any period in the life course had less household wealth in older adulthood than those with no health limitations. Most notably, among those who reported health limitations, there was evidence of a health gradient, with earlier onset being more consequential to wealth. Model 4 adjusts for subsequent life changes over the study period (in particular: marriage, husband’s health, and employment). Health limitations at each life stage remained a significant predictor of household wealth even after accounting for changes in family and employment. Even though the effects are somewhat attenuated, the timing of health limitations has an important influence on household wealth. Although the OLS regressions illustrate the magnitude of the problem, we now turn to the piecewise growth curve analyses to more precisely quantify the impact of health limitations and the processes involved. The results from this second stage of the analysis are presented in Table 3. In specifying the models, Model 1 is an unconditional growth model with random intercept and slope components, revealing that wealth increases in middle age as women grow older, but decreases after age 65. Table 3. Piecewise Growth Curve of Women’s Household Wealth, 1967–2003   Model 1  Model 2  Model 3  Model 4  Initial Status   Constant  2.802***,a  3.305***  3.351***  1.959***  (0.048)  (0.050)  (0.053)  (0.126)  Personal and Household Characteristics   Black, 1967    −1.577***  −1.577***  −1.288***    (0.073)  (0.072)  (0.068)   Children, 1967    −0.028*  −0.027*  −0.058***    (0.014)  (0.014)  (0.013)   Education, 1967    0.171***  0.163***  0.158***    (0.014)  (0.014)  (0.013)   Parent’s Prestige, 1967    0.003  0.002  0.003*    (0.001)  (0.001)  (0.001)   Respondent’s Prestige, 1967    0.017***  0.017***  0.016***    (0.002)  (0.002)  (0.002)  Timing of Health Limitationsb   Childhood      −0.735*  −0.607      (0.351)  (0.337)   Early Adulthood      −0.330*  −0.246      (0.149)  (0.143)  Life Course Events   Married, 1967–2003c        1.490***        (0.118)   Widowed, 1967–2003c        0.890***        (0.124)   Divorced, 1967–2003c        0.216        (0.124)   Husband’s Health Limitations, 1967–2003c        −0.146***        (0.040)   Full-Time Employment, 1967–2003c        0.114**        (0.034)   Part-Time Employment, 1967–2003c        0.080*        (0.036)  Rate of Change, Ages 34–49   Intercept  0.067***  0.067***  0.066***  0.072***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      0.009  0.021      (0.027)  (0.026)   Early Adulthood      0.004  0.004      (0.011)  (0.011)  Rate of Change, Ages 49–65   Intercept  0.050***  0.050***  0.058***  0.069***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      −0.035  −0.048*      (0.024)  (0.023)   Early Adulthood      −0.032**  −0.033**      (0.011)  (0.010)   Early Middle Age      −0.015**  −0.016**      (0.005)  (0.005)  Rate of Change, Ages 65–78           Intercept  −0.017**  −0.016*  −0.001  0.008  (0.006)  (0.006)  (0.015)  (0.015)  Timing of Health Limitationsb   Childhood      −0.054  −0.061      (0.055)  (0.054)   Early Adulthood      −0.009  −0.002      (0.029)  (0.029)   Early Middle Age      −0.012  −0.014      (0.019)  (0.018)   Late Middle Age      −0.027  −0.023      (0.018)  (0.018)   Older Age      −0.016  −0.014      (0.021)  (0.021)  N of Persons  4,664  3,989  3,989  3,989    Model 1  Model 2  Model 3  Model 4  Initial Status   Constant  2.802***,a  3.305***  3.351***  1.959***  (0.048)  (0.050)  (0.053)  (0.126)  Personal and Household Characteristics   Black, 1967    −1.577***  −1.577***  −1.288***    (0.073)  (0.072)  (0.068)   Children, 1967    −0.028*  −0.027*  −0.058***    (0.014)  (0.014)  (0.013)   Education, 1967    0.171***  0.163***  0.158***    (0.014)  (0.014)  (0.013)   Parent’s Prestige, 1967    0.003  0.002  0.003*    (0.001)  (0.001)  (0.001)   Respondent’s Prestige, 1967    0.017***  0.017***  0.016***    (0.002)  (0.002)  (0.002)  Timing of Health Limitationsb   Childhood      −0.735*  −0.607      (0.351)  (0.337)   Early Adulthood      −0.330*  −0.246      (0.149)  (0.143)  Life Course Events   Married, 1967–2003c        1.490***        (0.118)   Widowed, 1967–2003c        0.890***        (0.124)   Divorced, 1967–2003c        0.216        (0.124)   Husband’s Health Limitations, 1967–2003c        −0.146***        (0.040)   Full-Time Employment, 1967–2003c        0.114**        (0.034)   Part-Time Employment, 1967–2003c        0.080*        (0.036)  Rate of Change, Ages 34–49   Intercept  0.067***  0.067***  0.066***  0.072***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      0.009  0.021      (0.027)  (0.026)   Early Adulthood      0.004  0.004      (0.011)  (0.011)  Rate of Change, Ages 49–65   Intercept  0.050***  0.050***  0.058***  0.069***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      −0.035  −0.048*      (0.024)  (0.023)   Early Adulthood      −0.032**  −0.033**      (0.011)  (0.010)   Early Middle Age      −0.015**  −0.016**      (0.005)  (0.005)  Rate of Change, Ages 65–78           Intercept  −0.017**  −0.016*  −0.001  0.008  (0.006)  (0.006)  (0.015)  (0.015)  Timing of Health Limitationsb   Childhood      −0.054  −0.061      (0.055)  (0.054)   Early Adulthood      −0.009  −0.002      (0.029)  (0.029)   Early Middle Age      −0.012  −0.014      (0.019)  (0.018)   Late Middle Age      −0.027  −0.023      (0.018)  (0.018)   Older Age      −0.016  −0.014      (0.021)  (0.021)  N of Persons  4,664  3,989  3,989  3,989  Note: aUnstandardized regression coefficient (standard error in parentheses). bTiming refers to the onset of a health limitation; reference group = no health limitations. cTime-varying covariate. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Table 3. Piecewise Growth Curve of Women’s Household Wealth, 1967–2003   Model 1  Model 2  Model 3  Model 4  Initial Status   Constant  2.802***,a  3.305***  3.351***  1.959***  (0.048)  (0.050)  (0.053)  (0.126)  Personal and Household Characteristics   Black, 1967    −1.577***  −1.577***  −1.288***    (0.073)  (0.072)  (0.068)   Children, 1967    −0.028*  −0.027*  −0.058***    (0.014)  (0.014)  (0.013)   Education, 1967    0.171***  0.163***  0.158***    (0.014)  (0.014)  (0.013)   Parent’s Prestige, 1967    0.003  0.002  0.003*    (0.001)  (0.001)  (0.001)   Respondent’s Prestige, 1967    0.017***  0.017***  0.016***    (0.002)  (0.002)  (0.002)  Timing of Health Limitationsb   Childhood      −0.735*  −0.607      (0.351)  (0.337)   Early Adulthood      −0.330*  −0.246      (0.149)  (0.143)  Life Course Events   Married, 1967–2003c        1.490***        (0.118)   Widowed, 1967–2003c        0.890***        (0.124)   Divorced, 1967–2003c        0.216        (0.124)   Husband’s Health Limitations, 1967–2003c        −0.146***        (0.040)   Full-Time Employment, 1967–2003c        0.114**        (0.034)   Part-Time Employment, 1967–2003c        0.080*        (0.036)  Rate of Change, Ages 34–49   Intercept  0.067***  0.067***  0.066***  0.072***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      0.009  0.021      (0.027)  (0.026)   Early Adulthood      0.004  0.004      (0.011)  (0.011)  Rate of Change, Ages 49–65   Intercept  0.050***  0.050***  0.058***  0.069***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      −0.035  −0.048*      (0.024)  (0.023)   Early Adulthood      −0.032**  −0.033**      (0.011)  (0.010)   Early Middle Age      −0.015**  −0.016**      (0.005)  (0.005)  Rate of Change, Ages 65–78           Intercept  −0.017**  −0.016*  −0.001  0.008  (0.006)  (0.006)  (0.015)  (0.015)  Timing of Health Limitationsb   Childhood      −0.054  −0.061      (0.055)  (0.054)   Early Adulthood      −0.009  −0.002      (0.029)  (0.029)   Early Middle Age      −0.012  −0.014      (0.019)  (0.018)   Late Middle Age      −0.027  −0.023      (0.018)  (0.018)   Older Age      −0.016  −0.014      (0.021)  (0.021)  N of Persons  4,664  3,989  3,989  3,989    Model 1  Model 2  Model 3  Model 4  Initial Status   Constant  2.802***,a  3.305***  3.351***  1.959***  (0.048)  (0.050)  (0.053)  (0.126)  Personal and Household Characteristics   Black, 1967    −1.577***  −1.577***  −1.288***    (0.073)  (0.072)  (0.068)   Children, 1967    −0.028*  −0.027*  −0.058***    (0.014)  (0.014)  (0.013)   Education, 1967    0.171***  0.163***  0.158***    (0.014)  (0.014)  (0.013)   Parent’s Prestige, 1967    0.003  0.002  0.003*    (0.001)  (0.001)  (0.001)   Respondent’s Prestige, 1967    0.017***  0.017***  0.016***    (0.002)  (0.002)  (0.002)  Timing of Health Limitationsb   Childhood      −0.735*  −0.607      (0.351)  (0.337)   Early Adulthood      −0.330*  −0.246      (0.149)  (0.143)  Life Course Events   Married, 1967–2003c        1.490***        (0.118)   Widowed, 1967–2003c        0.890***        (0.124)   Divorced, 1967–2003c        0.216        (0.124)   Husband’s Health Limitations, 1967–2003c        −0.146***        (0.040)   Full-Time Employment, 1967–2003c        0.114**        (0.034)   Part-Time Employment, 1967–2003c        0.080*        (0.036)  Rate of Change, Ages 34–49   Intercept  0.067***  0.067***  0.066***  0.072***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      0.009  0.021      (0.027)  (0.026)   Early Adulthood      0.004  0.004      (0.011)  (0.011)  Rate of Change, Ages 49–65   Intercept  0.050***  0.050***  0.058***  0.069***  (0.003)  (0.003)  (0.004)  (0.004)  Timing of Health Limitationsb   Childhood      −0.035  −0.048*      (0.024)  (0.023)   Early Adulthood      −0.032**  −0.033**      (0.011)  (0.010)   Early Middle Age      −0.015**  −0.016**      (0.005)  (0.005)  Rate of Change, Ages 65–78           Intercept  −0.017**  −0.016*  −0.001  0.008  (0.006)  (0.006)  (0.015)  (0.015)  Timing of Health Limitationsb   Childhood      −0.054  −0.061      (0.055)  (0.054)   Early Adulthood      −0.009  −0.002      (0.029)  (0.029)   Early Middle Age      −0.012  −0.014      (0.019)  (0.018)   Late Middle Age      −0.027  −0.023      (0.018)  (0.018)   Older Age      −0.016  −0.014      (0.021)  (0.021)  N of Persons  4,664  3,989  3,989  3,989  Note: aUnstandardized regression coefficient (standard error in parentheses). bTiming refers to the onset of a health limitation; reference group = no health limitations. cTime-varying covariate. *p < .05; **p < .01; ***p < .001 (two-tailed tests). View Large Model 2 adds personal and household characteristics in 1967 to examine the effect of each covariate on wealth, including parental occupational prestige. The relationship between parental occupational prestige and initial household wealth was nonsignificant in Model 2. We also tested whether parental occupational prestige influenced the rate of change in wealth over time, but the parameter was nonsignificant (not shown). Model 3 incorporates binary indicators that capture the life course timing of health limitations. The model reveals that health limitations in childhood and early adulthood lowered the intercept on average, although neither produced an effect on the slope between 34 and 49 years of age. Between the ages of 49 and 65, there was significant growth in wealth; however, the onset of health limitations in early adulthood slowed the rate of growth in household wealth during this period of time, as did health limitations in early middle age. Earlier onset had larger effects on both the intercept and slope. Model 4 examines the effect of health limitations net of subsequent life changes treated as time-varying covariates. The results are not markedly different from Model 3; however, the effects of childhood- and early-adult health problems on the intercept became nonsignificant after controlling for changes in family and employment during the life course, though both slowed the rate of growth between the ages of 49 and 65. Estimates for personal and household characteristics in 1967 were slightly attenuated from Model 3, with the exception of parental occupational prestige which was significant: higher parental occupational prestige was associated with a higher average intercept. Supplementary Analyses Most women in our sample were married in 1967 (80%), with a modest percentage of these women divorcing during the span of the study (17%). This sample, therefore, represents a rather homogeneous cohort of women, which helps in some aspects to isolate how health limitations influence wealth. Studies on marriage and wealth nonetheless show the importance of marital histories for wealth accumulation, finding that those who are continuously married fare better than those who are not (Wilmoth & Koso, 2002). We thus examined our findings in the context of marital histories more closely. Extending our analyses in Table 2, we compared the effect of health limitations across marital histories (1 = continuously married, 0 = noncontinuously married) over the study period for those who survived to 2003. We examined interaction terms between the health and marital status variables, but none were significant. Continuously married women, however, were less likely to experience health limitations over the life course and reported shorter duration, on average. Given that wealth is measured at the household level, we tested additional information from the husband, including education and occupational prestige; however, our substantive conclusions were unchanged. We also tested for mediation with nonparametric bootstrap in R using the mediation package (Imai, Keele, & Tingley, 2010) but found none for any of the onset measures. Discussion and Policy Implications It has long been known that health and wealth are related, but most research on the topic has focused on how economic resources are salubrious to health—and how economic deprivation is bad for health. The purpose of this research was to examine the effect of health limitations on wealth by capitalizing on exceptional longitudinal data on health limitations and asset accumulation. Previous studies of health selection illustrate the extent to which poor health influences wealth accumulation, but we are unaware of any studies that isolated how the onset of health limitations at various points in the life course influence wealth accumulation. We found that women’s long-term health limitations take a substantial toll on household wealth that is independent of the woman’s marital status and employment, as well as her husband’s health. The results presented herein are consistent with some studies examining health and wealth in later life (Michaud & van Soest, 2008; Wu, 2003), but distinctive by identifying how early-life health problems lead to less wealth in later life. These findings have implications for research on health and social stratification and support elements of cumulative inequality theory. With a focus on how social stratification unfolds over the life course, the theory specifies that trajectories are shaped by the “onset, duration, and magnitude of exposures” (Ferraro et al., 2009, p. 420). Findings from the NLSMW revealed that life course timing and duration are critical for how health limitations lead to less wealth. Also consistent with the theory and prior studies is the influence of family lineage: parental occupational prestige is related to the wealth of the child’s household, independent of the respondent’s occupational prestige (Andrew & Ruel, 2010). The implications for this study are threefold. First, the findings contribute to our understanding of how health problems reduce household wealth by considering “upstream” factors that shape financial security in later life. There is substantial variability in the life course timing of health limitations, and without this knowledge, studies may overlook the effect of health limitations on wealth at particular life stages. To explicate the health–wealth relationship in later life, it is important to integrate elements of earlier life. Second, a life course study of health selection is relevant not only for future studies but also contributes to effective interventions. An increasing number of Americans are becoming bankrupt due to out-of-pocket medical expenses (Himmelstein, Thorne, Warren, & Woolhandler, 2009), and identifying the life course pathways could potentially broaden our understanding and deliver more targeted help. We found that the onset of health limitations reduced wealth, regardless of life stage, but onset in early life was most detrimental, slowing the rate of growth during middle age. By contrast, there was little change in wealth beyond age 65 due to onset of health limitations. Whereas the financial loss was manifest early, there is a need for interventions for those with early-life health problems so that fewer women become poor in later life. Future studies, however, should also consider the role of entitlement programs (e.g., Social Security, Medicare, and Medicaid) in the health–wealth relationship in older adulthood, which may at least partially offset the cost of health limitations. Finally, given recent social and demographic trends related to marriage, divorce, fertility, and work in the United States, there is an urgent need to develop social policies to enhance financial security in later life. A rising divorce rate among persons aged 50 years and older (Brown & Lin, 2012), coupled with a projected retirement crisis, will likely translate into a growing number of older adults facing financial insecurity. As shown herein, women’s health problems during early and middle adulthood are associated with lower household wealth: the financial loss affects both women and men. Indeed, despite the notable rise in women’s labor force participation, the Social Security replacement rate for married couples has actually declined in recent decades (Ellis, Munnell, & Eschtruth, 2014). In addition, the growing population of older divorced women is likely to face financial insecurity because of “relatively low Social Security benefits and high poverty levels” (Lin, Brown, & Hammersmith, 2017, p. 105). Rather than an exclusive reliance on Supplemental Security Income, beginning at age 65, policies to accumulate assets earlier may actually result in lower overall costs. There are several strengths of this study, but we also note its limitations. First, health limitations are based on fairly simple self-reported measures making it is difficult to determine the nature and severity of women’s health problems. The strength of the data is the many measurement occasions encompassing both retrospective and prospective data, but the measures are basic indicators of health limitations and susceptible to reporting bias and measurement error. In addition, as we noted earlier, the data are unable to capture health problems that were resolved before the beginning of the study, which we may lead to an underestimate of the health gradient in wealth. Second, the 36-year duration of the NLSMW is extraordinary, but the study’s discontinuation in 2003 constrains generalizability. Given the demographic trends noted earlier, we speculate that NLSMW estimates of the effect of women’s health limitations on wealth are an underestimate of the current and future reality. Thus, we invite future studies to confirm or refute that interpretation. Third, growth curve models allow respondents to contribute observations even if they missed one or more measurement occasions; however, estimates are still susceptible to selection bias if attrition is selective on both health and wealth. Despite these limitations, we believe that findings from this study contribute to a better understanding of the health–wealth relationship. The results from our analyses reveal that the effect of health limitations on household wealth is substantial, especially when the onset of health limitations occurs early in a woman’s life. In order to prevent poverty among the older adult population, we recommend greater attention to identifying, and compensating for, the influence of women’s health limitations on wealth. Funding Support for this research was provided by the National Institute on Aging (T32AG025671-02 and R01AG043544) and the Purdue Center on Aging and the Life Course. Conflict of Interest None reported. Acknowledgments The authors thank Ann Howell, Patricia Morton, Markus Schafer, and Dario Spini for helpful comments on an earlier draft of this manuscript. References Andrew, M. & Ruel, E. ( 2010). Intergenerational health selection and wealth: A first look at parents’ health events and inter vivos financial transfers. Social Science Research , 39, 1126– 1136. doi: 10.1016/j.ssresearch.2010.06.004 Google Scholar CrossRef Search ADS   Baker, A. C., West, S., & Wood, A. ( 2017). Asset depletion, chronic financial stress, and mortgage trouble among older female homeowners. The Gerontologist , gnx137, doi: 10.1093/geront/gnx137 Brown, S. L., & Lin, I. F. ( 2012). The gray divorce revolution: Rising divorce among middle-aged and older adults, 1990-2010. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 67, 731– 741. doi: 10.1093/geronb/gbs089 Google Scholar CrossRef Search ADS   Burbidge, J. B., Magee, L., & Robb, A. L. ( 1988). Alternative transformations to handle extreme values of the dependent variable. Journal of the American Statistical Association , 83, 123– 127. doi: 10.1080/01621459.1988.10478575 Google Scholar CrossRef Search ADS   Cai, L., Mavromaras, K., & Oguzoglu, U. ( 2014). The effects of health status and health shocks on hours worked. Health Economics , 23, 516– 528. doi: 10.1002/hec.2931 Google Scholar CrossRef Search ADS PubMed  Conley, D., & Glauber, R. ( 2007). Gender, body mass, and socioeconomic status: New evidence from the PSID. Advances in Health Economics and Health Services Research , 17, 253– 275. doi:10.1016/S0731-2199(06)17010-7 Google Scholar CrossRef Search ADS PubMed  Duncan, O. D. ( 1961). A socioeconomic index for all occupations. In A. J. Reiss, Jr. (Ed.), Occupations and social status  (pp. 109– 138). New York: Free Press. Elder, G. H. Jr. ( 1998). The life course as developmental theory. Child Development , 69, 1– 12. doi:10.1111/j.1467-8624.1998.tb06128.x Google Scholar CrossRef Search ADS PubMed  Ellis, C. D., Munnell, A. H., & Eschtruth, A. D. ( 2014). Falling short: The coming retirement crisis and what to do about it . New York: Oxford University Press. Google Scholar CrossRef Search ADS   Ferraro, K. F., & Shippee, T. P. ( 2009). Aging and cumulative inequality: How does inequality get under the skin? The Gerontologist , 49, 333– 343. doi: 10.1093/geront/gnp034 Google Scholar CrossRef Search ADS PubMed  Ferraro, K. F., Shippee, T. P., & Schafer, M. H. ( 2009). Cumulative inequality theory for research on aging and the life course. In V. L. Bengtson, D. Gans, N. M. Putney, & M. Silverstein (Eds.), Handbook of theories of aging  ( 2nd ed., pp. 413– 433). New York: Springer. Garbaski, D. ( 2014). The interplay between child and maternal health: Reciprocal relationships and cumulative disadvantage during childhood and adolescence. Journal of Health and Social Behavior , 55, 91– 106. doi: 10.1177/0022146513513225 Google Scholar CrossRef Search ADS PubMed  Gee, G., & Walsemann, K. ( 2009). Does health predict the reporting of racial discrimination or do reports of discrimination predict health? Findings from the National Longitudinal Study of Youth. Social Science & Medicine , 68, 1676– 1684. doi: 10.1016/ j.socscimed.2009.02.002 Google Scholar CrossRef Search ADS   Haas, S. A. ( 2006). Health selection and the process of social stratification: The effect of childhood health on socioeconomic attainment. Journal of Health and Social Behavior , 47, 339– 354. doi: 10.1177/002214650604700403 Google Scholar CrossRef Search ADS PubMed  Hallqvist, J., Lynch, J., Bartley, M., Lang, T., & Blane, D. ( 2004). Can we disentangle life course processes of accumulation, critical period and social mobility? An analysis of disadvantaged socio-economic positions and myocardial infarction in the Stockholm Heart Epidemiology Program. Social Science & Medicine , 58, 1555– 1562. doi: 10.1016/S0277-9536(03)00344-7 Google Scholar CrossRef Search ADS   Hammarström, A., Stenlund, H., & Janlert, U. ( 2011). Mechanisms for the social gradient in health: Results from a 14-year follow-up of the Northern Swedish Cohort. Public Health , 125, 567– 576. doi: 10.1016/j.puhe.2011.06.010 Google Scholar CrossRef Search ADS PubMed  Heckman, J. J. ( 1979). Sample selection bias as a specification error. Econometrica , 47, 153– 161. doi: 10.2307/1912352 Google Scholar CrossRef Search ADS   Himmelstein, D. U., Thorne, D., Warren, E., & Woolhandler, S. ( 2009). Medical bankruptcy in the United States, 2007: Results of a national study. The American Journal of Medicine , 122, 741– 746. doi: 10.1016/j.amjmed.2009.04.012 Google Scholar CrossRef Search ADS PubMed  Imai, K., Keele, L., & Tingley, D. ( 2010). A general approach to causal mediation analysis. Psychological Methods , 15, 309– 334. doi: 10.1037/a0020761 Google Scholar CrossRef Search ADS PubMed  Jackson, M. I. ( 2009). Understanding links between adolescent health and educational attainment. Demography , 46, 671– 694. doi:10.1353/dem.0.0078 Google Scholar CrossRef Search ADS PubMed  Kahn, J. R., & Pearlin, L. I. ( 2006). Financial strain over the life course and health among older adults. Journal of Health and Social Behavior , 47, 17– 31. doi: 10.1177/002214650604700102 Google Scholar CrossRef Search ADS PubMed  Kane, J. B., Philip Morgan, S., Harris, K. M., & Guilkey, D. K. ( 2013). The educational consequences of teen childbearing. Demography , 50, 2129– 2150. doi: 10.1007/s13524-013-0238-9 Google Scholar CrossRef Search ADS PubMed  Lachance-Grzela, M. & Bouchard, G. ( 2010). Why do women do the lion’s share of housework? A decade of research. Sex Roles , 63, 767– 780. doi: 10.1007/s11199-010-9797-z Google Scholar CrossRef Search ADS   Lin, I. F., Brown, S. L., & Hammersmith, A. M. ( 2017). Marital biography, social security receipt, and poverty. Research on Aging , 39, 86– 110. doi: 10.1177/0164027516656139 Google Scholar CrossRef Search ADS PubMed  Lundborg, P., Nilsson, A., & Rooth, D. O. ( 2014). Adolescent health and adult labor market outcomes. Journal of Health Economics , 37, 25– 40. doi: 10.1016/j.jhealeco.2014.05.003 Google Scholar CrossRef Search ADS PubMed  Lynch, J. L., & von Hippel, P. T. ( 2016). An education gradient in health, a health gradient in education, or a confounded gradient in both? Social Science & Medicine , 154, 18– 27. doi: 10.1016/ j.socscimed.2016.02.029 Google Scholar CrossRef Search ADS   Marmot, M. G., Shipley, M. J., & Rose, G. ( 1984). Inequalities in death—specific explanations of a general pattern? Lancet , 1, 1003– 1006. doi: 10.1016/S0140-6736(84)92337-7 Google Scholar CrossRef Search ADS PubMed  Marmot, M. G., Smith, G. D., Stansfeld, S., Patel, C., North, F., Head, J., … Feeney, A. ( 1991). Health inequalities among British civil servants: The Whitehall II study. Lancet , 337, 1387– 1393. doi:10.1016/0140-6736(91)93068-K Google Scholar CrossRef Search ADS PubMed  Michaud, P. C., & van Soest, A. ( 2008). Health and wealth of elderly couples: Causality tests using dynamic panel data models. Journal of Health Economics , 27, 1312– 1325. doi: 10.1016/ j.jhealeco.2008.04.002 Google Scholar CrossRef Search ADS PubMed  Modigliani, F. ( 1988). The role of intergenerational transfers and life cycle saving in the accumulation of wealth. Journal of Economic Perspectives , 2, 15– 40. doi: 10.1257/jep.2.2.15 Google Scholar CrossRef Search ADS   Painter, M., Frech, A., & Williams, K. ( 2015). Nonmarital fertility, union history, and women’s wealth. Demography , 52, 153– 182. doi: 10.1007/s13524-014-0367-9 Google Scholar CrossRef Search ADS PubMed  Schinkel-Ivy, A., Mosca, I., & Mansfield, A. ( 2017). Factors contributing to unexpected retirement and unemployment in adults over 50 years old in Ireland. Gerontology & Geriatric Medicine , 3, 2333721417722709. doi: 10.1177/2333721417 722709 Google Scholar PubMed  Smith, J. P. ( 1999). Healthy bodies and thick wallets: The dual relation between health and economic status. Journal of Economic Perspectives , 13, 145– 166. doi: 10.1257/jep.13.2.145 Google Scholar CrossRef Search ADS   Smith, J. P. ( 2004). Unraveling the SES-health connection. Population and Development Review , 30, 108– 132. Google Scholar CrossRef Search ADS   U.S. Department of Labor. ( 2001). NLS of mature women user’s guide . Retrieved from https://www.bls.gov/nls/mwguide/2001/nlsmwg0.pdf Wilmoth, J. & Koso, G. ( 2002). Does marital history matter? Marital status and wealth outcomes among preretirement adults. Journal of Marriage and Family , 64, 254– 268. doi: 10.1111/j.1741-3737.2002.00254.x Google Scholar CrossRef Search ADS   Wu, S. ( 2003). The effect of health events on the economic status of married couples. Journal of Human Resources , 38, 219– 230. doi: 10.2307/1558762 Google Scholar CrossRef Search ADS   Yamokoski, A. & Keister, L. A. ( 2006). The wealth of single women: Marital status and parenthood in the asset accumulation of young baby boomers in the United States. Feminist Economics , 12, 167– 194. doi: 10.1080/13545700500508478 Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

The GerontologistOxford University Press

Published: Jun 4, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off