## Abstract

Analyses of the Health and Retirement Study (HRS) between 1992 and 2014 compare the relationship between different levels and forms of debt and heart attack risk trajectories across four cohorts. Although all cohorts experienced growing household debt, including the increase of both secured and unsecured debt, they nevertheless encountered different economic opportunity structures and crises at sensitive times in their life courses, with implications for heart attack risk trajectories. Results from frailty hazards models reveal that unsecured debt is associated with increased risk of heart attack across all cohorts. Higher levels of housing debt, however, predict higher rates of heart attack among only the earlier cohorts. Heart attack risk trajectories for Baby Boomers with high levels of housing debt are lower than those of same-aged peers with no housing debt. Thus, the relationship between debt and heart attack varies by level and form of debt across cohorts but distinguishes Baby Boomer cohorts based on their diverse exposures to volatile housing market conditions over the sensitive household formation period of the life course.

## Introduction

The incidence and prevalence of some mortality risks, such as heart attack in aging cohorts, have declined or leveled off over recent decades. Socioeconomic status (SES) disparities in these risks across cohorts, however, appear to be widening. Higher levels of education, income, and wealth are typically buffers against stress, operating through healthy behaviors, access to health care resources, and reduced risks of disease and mortality with age. An ignored SES-related factor is debt, even though its level has risen across cohorts that have experienced different educational, labor, and housing markets, business cycles, and earnings trajectories across the life course and its form has differentiated into several types of secured and unsecured debt with differential impact on subgroups of aging populations (Dwyer 2018). Few studies have examined trends in the effects of different levels and forms of debt on morbidity or mortality risk. Furthermore, research on the relationship between debt and health has focused more on general health indicators despite evidence showing that the socioeconomic gradient in health varies by disease, and is particularly strong for cardiovascular diseases (Sommer et al. 2015).

This study analyzes data from the Health and Retirement Study (HRS) between 1992 and 2014 to compare the relationship between secured and unsecured debt to the risk of heart attack across four HRS cohorts. The basic rationale of the study is that although succeeding cohorts all exhibit the growth of household debt, including the increase of both secured and unsecured debt, they have nevertheless encountered different economic opportunity structures and crises at vulnerable times in the life course that anchor lifetime capacities to carry types of debt and to leverage them to advantage. We argue that these encounters demarcate the fortunes of successive cohorts with consequences for cardiovascular health.

## Sailing Into a Perfect Storm: Cohort Exposures to Recessions and Mortgage Crises

Successive twentieth-century cohorts have encountered different economic contexts in young adulthood when household formation and first mortgage acquisitions normatively occur. Usually associated with cohorts in their 30s, household formation is best conceptualized within a life course framework as a sensitive period in financial careers when future trajectories of wealth and debt are grounded in a baseline of resources (Ben-Shlomo and Kuh 2002; Streeter et al. 2018). Household formation has been represented as an absorbing transition to adult independence signaled by movement out of the parental home and the establishment of a new household, often followed by homeownership. Delays in household formation slow the process of wealth accumulation based on home equity, the major source of assets in addition to pension wealth at older ages. In this vein, Myers and Lee (2016) argued that homeownership is a cumulative status, anchored by financial opportunity structures and differential individual resources that form each cohort’s legacy and set the momentum of homeownership careers. Their analyses of Current Population Surveys and Annual Social and Economic Characteristics between 1964 and 2013 documented that earlier cohorts garnered the greatest advantage during this sensitive period with earlier starts in homeownership and less exposure to risky mortgage products, rising housing prices, and stagnating wages.

Masnick et al. (2006), using data from Surveys of Consumer Finance, further proposed that homeownership across cohorts displays differential trajectories of value, in which successive cohorts entered housing markets with changing financial regimes and successively higher housing values for first-time buyers in the more recent cohorts and for trade-up buyers in earlier cohorts. They found that between 1989 and 2001, later cohorts paid higher prices for homes. Housing debt also cut into their equity. These trends appear to have had selective effects: higher housing values have either leveraged and protected wealth for more advantaged subgroups, or increased the risks of delinquency and foreclosure for more disadvantaged subgroups. Later cohorts have faced higher levels of inequality, stagnating wages, easily available consumer credit, and rising costs of college education—all of which have contributed to greater debt burdens and inequality for recent cohorts compared with their predecessors (Houle 2014).

Figure 1 shows four HRS cohorts’ placements in different debt-related financial periods when they were ages 30–39. The figure identifies recessions since 1970 defined by the National Bureau of Economic Research (National Bureau of Economic Research 2020) and two periods of mortgage crises that followed in succession between 1986 and 2006. Members of the Prewar and War Baby cohorts encountered three relatively mild recessions (compared to the Great Recession) in their 30s. With the average age of homeownership at 28–29 for these cohorts (Streeter et al. 2018), most new homeowners did not encounter the higher climbing interest rates of the late 1970s and 1980s. Their average mortgage debt diminished over the life course, often ending in mortgage burning celebrations (Rosenbaum 2014). The Early and Mid-Boomers, in contrast, took on more mortgage debt with age as they experienced successive recessions and two mortgage bubbles (Masnick et al. 2006), the Savings and Loan Crisis of the late 1980s and early 1990s, and the longer lasting real estate bubble ending in the subprime crisis and the Great Recession. Over the increasingly volatile economic period between 1986 and 2006, Boomers’ debt portfolios shifted away from potentially wealth-building home mortgage debt to unsecured debt in the forms of medical bills, credit cards, and other forms of noncollateralized debt (Houle 2014).

As such, successive twentieth-century cohorts encountered recessions and mortgage bubbles at different phases of the life course, with differential impacts on their homeownership and debt trajectories. These changing economic contexts and complements of debts have likely shaped the relationship between financial strain and health, particularly because this strain has occurred over a sensitive period in the financial life course of Baby Boomer cohorts. More advantaged members within these cohorts who purchased and maintained homes during the sensitive period of household formation were likely to be less vulnerable to the housing crisis in the mid-2000s.

## Forms and Levels of Household Debt Since World War II

Total household debt in the United States reached a new peak of $13.8 trillion in the second quarter of 2019, a sum higher than that at the height of the credit bubble in 2008 (Federal Reserve Bank of New York 2019). The level of household debt grew from 64% of disposable income in 1974 to nearly 100% of disposable income in 2018 (Ahn et al. 2018). Mortgage balances represented the largest component ($9.1 trillion), with auto loans at $1.3 trillion, credit card debt at$.87 trillion, and student loans at $1.5 trillion. Mortgages and auto loans are secured debts, offering some kind of collateral (e.g., a house or a car) to the lender. Credit card debt is unsecured, offering no collateral to the vendor and often subject to third-party collections and lawsuits. Personal and payday loans are other examples of unsecured debts. Student loans are not always treated as unsecured debt (for an exception, see Richardson et al. 2013) but, like unsecured debt, serve to burden the financial capacity to initiate homeownership. In this study, student debt is not investigated because it is not measured separately in the HRS. Secured debts have been broadly familiar at least since World War II when the Federal Housing Administration (FHA) and GI Bill institutionalized low–down payment, low-interest mortgages amortized over 30 years for veterans (Rosenbaum 2014), although these benefits largely excluded non-Whites (Ehlers and Hinkson 2017). This program brought homeownership within the reach of more than one-half of households: 62% of households had mortgages by 1960, compared with 44% in 1940 (Rosenbaum 2014). Owning a home became associated with household formation among young adults, starting with the War Babies and Baby Boomers, who grew up during the largest expansion of homebuilding by that time in history. However, the growth of homeownership slowed by 1980 to 64% of households (Rosenbaum 2014), largely due to (1) delayed household formation as a result of increases in post-secondary educational participation; (2) dislocations related to economic restructuring, which especially impacted younger Baby Boomers (Bernhardt et al. 2001); and (3) increasing interest rates for 30-year mortgages that rose to 16% in the early 1980s, dropping afterward to recent levels between 3% and 4% (Freddie Mac 2017). After 1980, a succession of new financial policies continuously altered the mortgage landscape. The deregulation of the financial industry led to a loosening of lending practices and to the first mortgage crisis, often associated with failed savings and loan enterprises, in the late 1980s and early 1990s (Bratt and Immergluck 2015). New mortgage products with varying terms and amortization schedules proliferated over the next decade or more and were made available to high risk borrowers during the Clinton Administration. These changes paved the way for a mortgage lending boom between 1995 and 2006. The boom was also fed by an escalation in housing prices over the entire period, generating investment optimism among prospective homebuyers (Masnick et al. 2006; Rosenbaum 2014) and by a cultural shift toward the use of multiple forms of debt as means to enhance life chances (Fligstein and Goldstein 2015). In addition, this period initiated a new pattern of substitution of mortgage debt for consumer debt when record amounts of equity from homes were used to cover other consumer debts (Olegario 2019). Finally, the Great Recession of 2007–2009, which adversely affected 4 of 10 households with unemployment, negative home equity, or mortgage delinquency (Hurd and Rohwedder 2010), brought the highest foreclosure and delinquency rates (5%) since the 1970–1987 period, when they stayed at approximately 1%. Rates returned to lower levels of approximately 2% by 2014 (Ahmad 2015). Unsecured debt has grown significantly over recent decades. Levels of credit card debt rose steadily between 1980 and 2008, declined until 2011, but then increased very steeply after 2014 (Federal Reserve Bank of New York 2019). People increasingly turn to credit cards for daily consumption needs, leading to more than 50% more credit card debt today than a decade ago (Federal Reserve Bank of New York 2019). Student loan debt has surpassed credit card debt, slowing recent cohorts’ entry into household formation and homeownership (Myers and Lee 2016), leaving some with little equity and damaged credit (Draut and Silva 2004; Houle and Warner 2017). The negative impact of debt has also reached older persons, whose median debt in 2016 was 2.5 times more than it was in 2001 (Goodman et al. 2017). Not all debt is problematic. Debt is not always an indicator of financial problems but also signals access to financial resources for status maintenance or upward socioeconomic mobility. Debt can be leveraged to increase wealth among more advantaged households. But it can burden moderate-income households, especially in the face of economic shocks. Finally, because credit is less accessible to the most disadvantaged (Kumhof et al. 2015), the absence of debt can mask serious economic vulnerability. Hence, debt is embedded in complex life course patterns of social inequality encompassing selective factors that cumulatively stratify aging cohorts. ## Debt, Health, and Cardiovascular Disease Although wealth or net worth is more commonly included in studies of SES and health among adults as putatively protective of health, debt itself as a unique stressor has recently attracted researchers’ attention. A small but growing literature has linked severe or chronic debt, noncollateralized debt, and chronic financial strain to cumulative physiological stress. Financial strain contributes to differential morbidity and mortality rates, in part, because the induced stress of debt increases risk of obesity, substance abuse, depression, suicide, and poor self-related health (Richardson et al. 2013; Turunen and Hiilamo 2014). The majority of the studies we have identified (mostly cross-sectional) linked debt to perceived stress, depression, and poor self-rated health (e.g., Drentea and Reynolds 2012, 2015; Gathergood 2012; Knapp and Dean 2018), with a growing number of studies focused on later cohorts who face proportionately higher debt burdens from unsecured debt and college loans (Dwyer et al. 2016; Kim and Chatterjee 2019; Sweet et al. 2013; Walsemann et al. 2015). Not all forms of debt, however, are linked to worse health outcomes. In their study of 17 European countries between 1995 and 2012, Clayton et al. (2015) found that whereas long-term aggregate household debt was associated with poor health outcomes, short-term aggregate household debt was predictive of positive health outcomes. Their findings showed that the health impact of aggregated household debt depends on its form, which might be the case at the individual level as well. Mortgage debt still carries a social valuation as “good debt” compared with unsecured “bad debt” based on normative and wealth-building criteria (Dwyer 2018). Unsecured debt may exert a unique impact on everyday emotional experiences with a more immediate and stigmatizing concreteness, especially increasing financial distress when income is low or insecure (Berger et al. 2016; Dunn and Mirzaie 2016; Kahnemann and Deaton 2010). However, the recent climate for mortgages raises the question of the unique impacts of secured debt and unsecured debt on health across cohorts over this period. The four HRS cohorts tracked in Fig. 1 have lived over a sufficiently volatile period to examine how forms and levels of debt influence health, particularly in adulthood when risks for physical illness increase in response to aging and cumulative stress. Moreover, the large-sized Baby Boom cohorts have been repeatedly identified as being particularly vulnerable to stress-related factors associated with the competitive pressures emanating from sheer cohort size (Easterlin 1980) and from changing economic conditions over the second half of the twentieth century (e.g., Bernhardt et al. 2001; Hughes and O’Rand 2005; Yang 2008). Notably, the relationship between debt and health is bidirectional. Health problems themselves can lead to substantial debt (Dwyer 2018). In fact, medical debt is the single largest cause of consumer bankruptcy (Austin 2014). In 2015, 26% of Americans aged 18–64 struggled to pay medical bills, with 70% of those cutting back on basic expenses to pay them (Hamel et al. 2016). People with high levels of medical debt are more likely to avoid going to the doctor or dentist (Kalousova and Burgard 2013), creating a cycle of health problems leading to debt, which in turn increases the risk of more health problems and increased debt. Dobkin et al. (2018) observed this complex link by examining the economic consequences of hospital admissions for non-elderly HRS respondents. Respondents admitted to the hospital had a higher risk of future bankruptcy and reduced access to consumer credit. Disentangling this reciprocal relationship presents methodological and theoretical challenges that this study only partially addresses with a dynamic longitudinal multivariate cohort design. Few studies have examined the relationship between forms of debt and specific diseases in aging cohorts, even though the strength of the relationship between socioeconomic factors and health varies by disease (Sommer et al. 2015). We focus on risk of heart attack for multiple reasons. First, socioeconomic disparities in cardiovascular disease (CVD) have recently become more pronounced despite relative declines in CVD across cohorts (Crimmins 2018). The persistence of health disparities in CVD raises questions about the role of debt. Second, depression and poor health behaviors are predictive of heart attack risk and are more common among those in debt (Drentea and Lavrakas 2000; Gathergood 2012; Turunen and Hiilamo 2014). Finally, the few studies that have examined health outcomes other than mental or general health have found links between indicators of cardiovascular diseases and debt (Dean et al. 2019; Sweet 2018). We examine cohort differences in the relationship between heart attack risk and levels and forms of debt across four HRS cohorts guided by three hypotheses: • Debt level hypothesis: Higher levels of debt are predictive of higher risks of heart attack for all cohorts. • Debt form hypothesis: The stress-inducing, health-depleting effects of debt are greater for unsecured debt than secured debt, resulting in a stronger relationship between elevated heart attack risk and unsecured debt compared with secured debt. • Cohort sensitive-period hypothesis: The relationships among elevated heart attack risk, secured debt, and unsecured debt are distinctive for later cohorts who have benefitted from declining heart attack rates but who distinctively encountered sustained volatile housing markets over the sensitive period of household formation that have differentiated their heart attack risks. ## Data and Methods ### Data To test these hypotheses, we analyze 22 years of data collected by the Health and Retirement Study (HRS) between 1992 and 2014. Launched in 1992 by the National Institute on Aging at the University of Michigan’s Institute for Social Research, the HRS is a leading source of information on the financial, health, work, and retirement well-being of adults over age 50 in the United States (Sonnega et al. 2014). Every two years, the longitudinal panel study surveys a nationally representative sample of approximately 20,000 people in the United States in their 50s and older. In 1992, the HRS began by sampling a nationally representative group of Americans aged 51–61 who were born between 1931 and 1941. This is the Prewar cohort. To facilitate same-age comparisons across cohorts, our sample includes those born in 1936 to 1941. The HRS implements a steady state design, adding a new cohort of respondents aged 51–56 every six years to replenish earlier cohorts as they age. In 1998, the HRS began surveying the War Baby cohort, born between 1942 and 1947. The Early Boomer cohort, born between 1948 and 1953, was added to the HRS sample in 2004. In 2010, the sample was refreshed again with a cohort born in 1954 to 1959, known as the Mid-Boomers. The sample design of HRS thus enables cohort comparisons but also allows us to address selective mortality. For panel members who died between survey years, proxy informants knowledgeable about the decedents’ health and financial situation participate in exit interviews. The deceased respondents’ spouse, child, close friend, or other family member report information about the respondent from the time of last interview until the death. If exit interviews are incomplete during one wave, usually because estates have not been settled, post-exit interviews fill in missing information. By tracking through the Social Security Death Index and the National Death Index, HRS staff confirm the status of deceased respondents and locate exit interview proxies. As of the end of 2010, exit interviews were completed for 93% of decedents (Sonnega et al. 2014). For purposes of this investigation, exit interviews allow us to include detailed information about respondents who had heart attacks but were not captured in the core data. ### Measures The key outcome of interest is whether and when respondents had heart attacks. In each survey year, respondents or their proxies reported whether a doctor had ever (or since the last interview) told them they had had a heart attack. Those who answered yes report the date of heart attack. For survival analyses, we code the event variable as 1 if respondents experienced a heart attack. For those reporting a heart attack, we subtract their birth year from year of first heart attack to calculate duration. For respondents who did not report having a heart attack, duration equals age at last survey. To test our debt level hypothesis, which predicts elevated heart attack risk among those with higher levels of debt across cohorts, our key independent variables measure the amount and type of respondents’ debt. In every survey year between 1992 and 2014, respondents reported the dollar amount of their housing debt (mortgages and home equity loans), other forms of secured debt (business, land, rental property, vehicle), and nonsecured debt (credit card debt, medical bills, personal loans, or other debt that does not have collateral). We convert reported dollar amounts to the value of 2014 U.S. dollars and apply a logarithm transformation to account for the skewed distribution. In descriptive analyses, we compare respondents who have no unsecured and secured debt with those with high amounts of debt, or those in the top quartile of the unsecured/secured debt distribution for their cohort at each survey year. These measures also allow us to test our debt form hypothesis, which predicts that increasing amounts of unsecured debt raise the risk of heart attack to a greater extent than increasing amounts of secured debt. For comparative purposes, we compute net worth at each survey year as respondents’ total assets minus their total debts, in constant 2014 dollars. Because net worth is highly skewed and can be negative, respondents at each survey year are classified by their net worth percentile within their cohort, with values ranging from 1 to 10. As another indicator of financial stability, we divide respondents’ total liabilities by their total assets at each survey year and create a dichotomized measure of whether or not respondents have a debt-to-asset ratio of .5 or greater. A high debt-to-asset ratio means that creditors own at least one-half of respondents’ assets. To test our cohort sensitive-period hypothesis—predicting a distinctive, significant relationship between housing debt and heart attack risk among later cohorts than earlier cohorts—we include dichotomous measures of membership in the Prewar, War Baby, Early Boomer, or Mid-Boomer cohorts. Our models also include demographic variables of sex, race, and years of education measured in the cohort entry interview wave when respondents were aged 51–56. We compare respondents who identify themselves as White non-Hispanic with those who are Black non-Hispanic, Hispanic, or a combined category of other races and ethnicities. For every wave of data, we use an indicator of whether respondents are married or partnered and compare them with respondents who are separated, divorced, widowed, or never married. HRS respondents also provide detailed employment information during each interview wave. We include time-varying measures of whether respondents are currently employed and whether their longest occupation is blue collar (crafts, labor, farm, and operative occupations), lower white-collar (sales and clerical occupations), or upper white-collar (executive, managerial, and professional occupations). For every survey year, we compare respondents who report having health insurance with those who do not. Other time-varying covariates include household income in 2014 dollars and homeownership. To examine a possible link between debt and risk of heart attack, we explore measures of financial strain. We compare those who reported it is “somewhat, very, or completely difficult to meet monthly payments of their family’s bills” with those who reported it is “not very or not at all difficult.” Data limitations restrict our use of other indicators of debt-related stress. Beginning in 2006, a randomly selected half of HRS respondents answered a leave-behind psychosocial questionnaire that contains key measures of financial and housing problems. In 2008, the other half of the sample answered the questionnaire, and this alternating-year survey design continues. We report responses to these leave-behind questionnaires in descriptive analyses of 2010 data, but we do not use the measures in the full-sample multivariate analyses. We report the percentage of 2010 respondents who fell two or more months behind on mortgage payments or went through foreclosure of their home in the previous two years. We compare 2010 respondents who reported it somewhat or very likely that they will fall behind on mortgage payments over the next six months with those who said it is unlikely. Dichotomous measures indicate whether they have been unfairly denied a bank loan in the previous two years, are currently unsatisfied with their financial situation, and are very upset by financial strains or housing problems that have lasted for at least the past year. Other key variables measure depression and unhealthy behaviors. The HRS measures a subset of symptoms from the Center for Epidemiologic Studies Depression scale (CES-D). Higher scores indicate greater numbers and/or frequencies of depressive symptoms, such as loneliness, restless sleep, not enjoying life, and feeling sad. Because debt may increase risk of poor health by reducing the ability to pursue health care, we include time-varying measures of the frequency respondents go to doctors. In every survey year, respondents report the frequency they talked to medical doctors about their health over the previous two years. Respondents who reported a recent heart attack are likely to see doctors because of their heart attacks, so we use a lagged measure of the number of doctor visits in the two years preceding heart attack for these respondents. HRS respondents report every year whether and how often they consume alcohol. Using guidelines recommended by the Centers for Disease Control and Prevention (2016), we identify heavy drinkers as men who consume 15 or more drinks per week and women who consume 8 drinks or more per week. Moderate drinkers are those who have up to one drink per day for women and two drinks per day for men. These groups are compared with nondrinkers. The time-varying covariate for smoking compares respondents who currently smoke with those who do not. Respondents report every year how often they participate in sports or other vigorous activities, such as jogging, swimming, aerobics, or tennis. We compare respondents who engage in vigorous activities at least three times per week with all others. Finally, we include a time-varying measure for respondents’ body mass index, calculated from their weight and height at each survey year. ### Multivariable Analytic Strategy To test our hypotheses about cohort differences in the relationship between different forms of debt and heart attack risk trajectories, we use survival analyses. The Cox proportional hazards model is a frequent choice for handling survival data, which measures time to an event or censoring (Guo 2010). The Cox method does not require specification of the probability distribution of survival times and can model time-dependent covariates. However, the HRS design gives repeated measures within individuals, which results in correlated measures within observations. This violates the Cox hazards model’s assumption of independent observations. Furthermore, it is unlikely that all relevant covariates can be included in a Cox model. This leads to unobserved heterogeneity between individuals. Multivariate frailty models take into account the correlation between observations and unobserved heterogeneity (Wienke 2010). Vaupel et al. (1979) introduced the term frailty to models of survival data, which they considered to be a random variable measuring variation in risk of mortality between individuals. They separated heterogeneity due to observed and unobserved factors by estimating the individual hazard rate as a function of the frailty variable multiplied by a basic hazard rate shared by all respondents. In other words, the unobserved frailty factor can be represented by an unobservable random effect that impacts the baseline hazard function multiplicatively. The hazard function for individual i at time t given the underlying hazard function λ0 is $λit=Ziλ0texpβXi,$ where Zi is the frailty score for individual i, and exp[βXi] is the exponentiated linear function of a set of fixed covariates. Frailty scores measure the amount of variance left unexplained by the covariates. If the frailty score is 0, measured covariates fully explain the relationship between the dependent and independent variables. Essentially, this is a random-effects proportional hazards model (Liu 2014). Because of its flexible mathematical properties, we specify a gamma distribution for the random effect Z. The frailty approach has biological underpinnings (Aalen et al. 2015). The incidence of heart attack, for example, increases with age, and then the risk begins to decline (Akushevich et al. 2012). One interpretation of the peaking incidence rate reflects the frailty view that the risk of heart attack increases throughout life, but some people are more highly susceptible and more likely to develop the disease early and die (Afilalo et al. 2014), leaving the population with a smaller proportion of susceptible individuals. Thus, the incidence rate drops. In diseases with some heritable component, such as CVD (Kathiresan and Srivastava 2012), frailty models allow us to acknowledge the influence of identifiable, measurable risk factors and varying predisposition to disease. Using standard survival analysis can result in biased parameter estimates and incorrect model-based predictions (Liu 2014). Frailty models, however, introduce random effects to account for unobserved heterogeneity and associations between event times and model covariates. For all survival analyses, we first run Cox proportional hazards models and then run gamma frailty hazard models. Because they provide a better fit to the data, with no significant changes in parameter estimates, we present results from frailty models. ## Results ### Descriptive Statistics Table 1 reports cohort differences in dependent and independent variables when respondents first entered the study. Analysis of variance (ANOVA) tests show that mean differences across cohorts for designated variables are statistically significant at the p < .01 level. Slightly less than one-half the sample is male, with ethnic/racial diversity increasing among more recently added cohorts. War Babies and members of the Prewar cohort completed fewer average years of schooling than Boomers. However, Mid-Boomers who were in their 50s in 2010 are less likely to be employed than same-aged earlier cohorts and more likely to be working in lower white-collar occupations if they are employed. As shown in Table 1, 14% of the Prewar cohort reported having a heart attack by 2014 when they were aged 73–78. Approximately 8% of the War Babies who were 67–72 years old in 2014 reported a heart attack. Only 4% and 2% of the Early Boomers and Mid-Boomers, respectively, reported having a heart attack when Early Boomers were 61–66 years old and Mid-Boomers were aged 55–60. In addition to higher rates of ever having a heart attack, the two earlier cohorts are significantly more likely than Boomers to have had a heart attack upon survey entry when all respondents were in their early to mid-50s. The Boomers are less likely to smoke, more likely to exercise, and report higher levels of depression and being overweight than the earlier cohorts. Mid-Boomers, on average, go the doctor and drink alcohol more frequently than members of the Prewar cohort. The bottom panel in Table 1 shows cohort differences in financial status and resources at ages 51–56. War Babies, who were in their early to mid-50s in 1998, benefited from the strong economy of that year. War Babies have, on average, the highest household incomes and homeownership rates and are the least likely cohort to have high debt-to-asset ratios. In contrast, same-aged Mid-Boomers in 2010 have the lowest average household incomes and rates of homeownership. They are the most likely cohort to have high debt-to-asset ratios and reported the highest mean amount of secured and unsecured debt at the ages of 51 to 51–56. For example, whereas Prewar homeowners in their 50s owed an average of$50,600 in housing debt in 1992, same-aged Mid-Boomers owed $96,500 in housing debt in 2010. All debt amounts are measured in the value of 2014 dollars. Early Boomers, as well, acquired significantly more debt in their 50s than same-aged members of the Prewar and War Baby cohorts. Compared with those who were in their 50s during the 1990s, Early and Mid-Boomers had$5,000 to $10,000 more in secured, non-housing debt in 2004 and 2010. More than one-half of all cohort members did not have unsecured debt when they entered the survey at ages 51–56. However, Boomers in the 2000s had substantially higher average amounts of unsecured debt. For example, the mean unsecured debt amount for Prewar respondents was$6,545 in 1992 compared with an average of more than \$20,000 for same-aged Boomers in 2004 and 2010. Previous research suggests that Boomers’ weaker financial resources and higher debt burdens would contribute to higher rates of heart attacks as they age. However, descriptive results show that War Babies and Prewar respondents were more likely to experience heart attacks over time than same-aged Boomers.

Table 1 compares key measures across cohorts when all respondents are aged 51–56, and Table 2 examines cross-cohort differences in experiencing financial challenges during and after the Great Recession. Mid-Boomers were in their prime earning years, and members of the Prewar cohort were primarily retired. Across measures, Boomers reported higher rates of housing and financial problems during and after the recession. For example, Mid-Boomers were about four times more likely than the earliest cohort members to be denied a bank loan, fall behind on their mortgage, and experience ongoing distressing housing problems. Although only 6% of War Babies and 4% of the Prewar cohort reported experiencing very upsetting financial problems for a year or more, 10% of Early Boomers and 14% of Mid-Boomers did. Boomer homeowners had higher rates of past and predicted future home foreclosures and were more likely to have higher debt-to-asset ratios, associated with their dissatisfaction with their current financial situation. In fact, almost one-half of Mid-Boomers reported that they had difficulty paying their monthly bills in 2010. These descriptive statistics indicate worse financial and housing outcomes for Boomers after the Great Recession that could have implications for heart attack risk trajectories.

### Debt and Heart Attack Risk

Table 3 presents results from frailty models that test the debt level hypothesis, which predicts increasing levels of debt are associated with elevated heart attack risk for all cohorts, and the debt form hypothesis, which predicts stronger positive relationships between unsecured debt and heart attack risk than secured debt. We run models separately for each cohort. The debt measures in the first models include secured debt backed by housing property, non–housing secured debt, and unsecured debt. The models also control for assets, age, sex, race, ethnicity, marital status, education, income, employment status, occupation, health insurance coverage, difficulty paying bills, doctor visit frequency, depression, alcohol consumption, smoking, exercise frequency, and body mass index.

Results from Model 1 only partially support our hypothesis that debt is associated with increased risk of heart attack. Whereas increasing amounts of housing debt are associated with higher risk of heart attack for War Babies and members of the Prewar Cohort, the opposite is the case for Boomers. Ad hoc t tests comparing betas for mortgage debt across cohorts show that parameter estimates for secured housing debt for Boomers are statistically significantly different than for War Babies or Prewar cohort members. When all other covariates are controlled for, Boomers with higher amounts of mortgage debt have a lower risk of heart attack than their peers with less mortgage debt. In sharp contrast, higher amounts of unsecured debt are predictive of higher hazard of heart attack for all cohorts. Non–housing secured debt is not statistically significantly associated with heart attack risk for any cohort. Thus, we reject the debt level hypothesis because higher levels of debt are not consistently predictive of elevated heart attack hazard across cohorts. We find support for the debt form hypothesis: higher amounts of unsecured debt across cohorts are predictive of increased heart attack risk, and secured debt is not. The relationship between debt and heart attack varies by type of debt.

To test the robustness of our findings, Table 3 presents coefficients from three other models that include alternative measures of debt. As in Model 1, we run Models 2–4 separately for each cohort and present the net effects of debt and assets on heart attack risk after controlling for other covariates. Model 2 shows that respondents from all cohorts with more assets are less likely to experience having a heart attack. Debt burden, however, is associated with higher heart attack hazard for members of only the Prewar and War Baby cohorts. When forms of debt are not differentiated, it appears that debt is not statistically significantly associated with heart attack risk among Boomers. Model 3 shows that regardless of cohort, moving up the ladder of net worth is associated with reduced heart attack risk, with other covariates controlled for. Finally, Model 4 supports findings in Models 1 and 2 that the relationship between heart attack risk and debt varies across cohorts when key predictors of heart attack risk are controlled for. For the earlier cohorts, higher debt-to-asset ratios are associated with elevated risk of heart attack. The opposite is true for Boomers.

Because Table 3 presents the relationship between logged amounts of debt and heart attack risk, the coefficient interpretations are not straightforward. For example, for Prewar residents, a one-unit increase in the log of housing debt is associated with a 1% increase in the hazard of heart attack. Figures 2 and 3 better illustrate how the relationship between different forms of debt and heart attack risk vary over time and across cohorts. They depict the cumulative hazard of heart attack by age and cohort for those with no debt and high debt, identified as respondents in the upper quartile of debt for their cohort and survey year.

Figure 2 shows that a high level of unsecured debt is associated with elevated risk of heart attack regardless of birth cohort. The gap between those with no unsecured debt and high unsecured debt grows bigger as Prewar and War Babies age.

Figure 3 presents a different picture of the relationship between housing debt and heart attack risk trajectories. For the Prewar cohort, respondents with high levels of housing debt have elevated risk of heart attack beginning in their mid-30s and extending into their 70s. Among War Babies, those with large mortgages or home equity loans experience about the same risk of heart attack as War Babies with no housing debt. In contrast, Boomers with high levels of housing debt have lower rates of heart attack than their age mates with no housing debt. In fact, Mid-Boomers with no housing debt have higher heart attack risk trajectories that run closely parallel to that of the oldest cohort with high secured debt. Thus, we reject the debt level hypothesis: higher levels of debt are inconsistently associated with risk of heart attack across cohorts. Only increasing amounts of unsecured debt are associated with increased heart attack hazard across cohorts, thus supporting the debt form hypothesis.

### Cohort Sensitive Period and Risk of Heart Attack

Finally, Table 4 builds on these findings and presents results testing the cohort sensitive-period hypothesis that the relationship between housing debt and heart attack risk is significantly different for Boomers than earlier cohorts due, in part, to the sustained volatile housing and financial markets Boomers experienced between their 30s and early 50s. Table 3 reports and tests the statistical difference between parameter estimates across cohorts, and Table 4 shows the results of the frailty analysis on the full HRS sample. We capture cohort differences in risk of heart attack by comparing Mid-Boomers, Early Boomers, and War Babies with members of the Prewar cohort and including cohort interaction terms for our key measures of debt. Interactions between debt and War Babies are not statistically significant and do not improve model fit. For ease of presentation, we do not include them in the model.

Table 4 shows that the risk of heart attack is lower among later cohorts than respondents born between 1936 and 1941. Even when controlling for age and other key indicators, Mid-Boomers, for example, have a 75% (0.25 – 1) lower hazard of heart attack than members of the Prewar cohort. In support of the debt level hypothesis, the hazard of heart attack increases with rising amounts of unsecured debt for all cohorts. Nonsignificant interaction terms show that the relationship between unsecured debt and hazard of heart attack does not vary by cohort. Overall, higher housing debt predicted elevated heart attack risk, but the negative betas for interactions between housing debt and Boomer cohorts show (as in Table 3) that the relationship between housing debt and hazard of heart attack is not consistent across cohorts. With other variables held constant, higher amounts of housing debt predict increased risk of heart attack for the Prewar cohort and War Babies (β = .02). In contrast, holding all other factors constant, high levels of housing debt are associated with reduced hazard of heart attack for Boomers (.02 – .03 = –.01). The relationship between housing debt and heart attack does not statistically vary between War Babies and Prewar cohort members. For all cohorts, non–housing secured debt is not statistically linked to heart attack risk.

Table 4 presents findings for variables known to predict heart attack risk. Age, sex, race/ethnicity, marital status, education, and income are related to the hazard of heart attack in predictable ways. Risk of heart attack increases with advancing age. Men’s risk of heart attack is 2.4 times as great as women’s risk, and White non-Hispanic respondents are more likely to experience a heart attack than Black non-Hispanics, Hispanics, or those of other races/ethnicities. Each additional year of education reduces the hazard of heart attack by 3%. Married respondents are at reduced risk of heart attack than unpartnered respondents. Employed respondents have an 8% lower risk of heart attack than those out of the labor force, but employees of blue-collar occupations are 25% more likely than upper white-collar workers to have a heart attack. Higher levels of household income, homeownership, and having health insurance are associated with decreases in the hazard of heart attack. Financial stress is strongly linked to risk of heart attack. When income and debt are controlled for, respondents who struggled to pay their bills were 50% more likely to experience a heart attack than those who could easily meet their financial obligations.

Finally, our model includes important measures of health behaviors known to increase the risk of heart attack. These variables operate in predictable ways but do not account for the statistically significant relationship between debt and heart attack risk. Hazard of heart attack is lower among those who exercise regularly and drink alcohol moderately. Depression, smoking, and heavy weight are predictive of increased risk of heart attack. Even after we control for these key health predictors, higher amounts of unsecured debt are associated with increased hazard of heart attack for all cohorts. Members of the Prewar and War Baby cohorts have higher risks of heart attack as their housing debt increases. The opposite is true for Early and Mid-Boomers. The statistically significant random effect indicates unobserved heterogeneity remains even after controlling for measured covariates.

## Discussion

The relationship between risk of heart attack and debt is not uniform across cohorts or forms of debts. Unsecured debt is associated with increased risks of heart attack across all cohorts. However, secured debt related to homeownership is not. Compared with earlier cohorts, some Early Boomers and Mid-Boomers who survived a series of mortgage bubbles during adulthood and the Great Recession experienced a seemingly protective effect of high levels of mortgage debt against heart attack. By the end of the Great Recession, the Boomer cohorts were worse off financially in many respects compared with their older counterparts, with Mid-Boomers especially disadvantaged on average. However, the seemingly protective effect of high mortgage debt against heart attack probably reflects selective processes that benefitted the more advantaged homeowners in these cohorts. Those with greater resources who survived volatile housing markets with high levels of secured housing debt were less likely to experience heart attack. The least advantaged subgroups in these cohorts, who were without mortgage debt in their mid-50s, displayed relatively increased rates of heart attack that reached the levels observed in the oldest cohort with high debt levels.

Focusing on debt as opposed to wealth or net worth permits a closer view of the impact of economic insecurity on the stress process and health. Greater wealth is associated with better health and also with higher debt levels. Leveraging debt can also build wealth, as in the case of homeownership or business-related debt. And wealth buffers the effects of debt, with improved access to resources including the use of housing equity to absorb consumer debt. Nevertheless, debt and wealth are distinctive phenomena. First, debt generally has a lower social valuation as “bad” or stigmatizing, especially if unpaid debt increases or jeopardizes other resources that may range from credit scores to job placement. Second, debt has an everyday salience related to the recurrent flow of liabilities with stress-provoking effects on lives as bills come due or overdue. This is in contrast to wealth, which is arguably a more abstract and temporally diffuse phenomenon best represented as the “estimated stock” of resources based on future expectations. Accordingly, the effects of debt on stress are more immediate and concrete, and can only be buffered by wealth.

Many studies of the impact of the Great Recession have revealed the short-term impact of this event on population subgroups with differential economic resources. This longitudinal study has situated the Great Recession in a longer history of differential cohort exposures to crises and bubbles. The approach here has revealed that in the case of two Boomer cohorts, their experiences have differentiated them from earlier cohorts. By following a dynamic life course approach anchored in homeownership careers and household debt trajectories emanating from the sensitive period of normative household formation, these analyses add to our understanding of how the Boomer cohorts weathered the volatile housing and credit markets of the last 30 years.

More granular analyses are now needed to examine the patterns of debt inequality within these cohorts that contributed to differential heart attack risk. Recessions and mortgage crises tend to accentuate and lock in inequalities within cohorts (Picketty and Saez 2015). Although low-income and minority groups benefitted from the mid-1990s initiatives, they faced additional institutional barriers to becoming and remaining homeowners (Kochkar et al. 2009). Racial gaps in housing values persisted throughout the twentieth century (Ehlers and Hinkson 2017). Furthermore, Blacks in the1990s were still less likely to apply for a mortgage and twice as likely to be rejected (Charles and Hurst 2002). Predatory lending practices resulted in steep reversals of homeownership for Blacks and Hispanics after 2005 who were disproportionately exposed to subprime mortgage rates, delinquency, and foreclosure (Ehlers and Hinkson 2017). These exposures across the life course of minority groups have resulted in the absence of credit across the life course with consequences for health at older ages.

Finally, this analysis does not escape the challenge of disentangling causality in at least three important relationships. The first is between forms of debt and health events such as heart attacks. SES and health are interdependent over time, and specific forms of debt and types of disease may be involved in quite different relationships over time that cannot be readily specified with these data. The Boomer cohorts encountered predatory housing markets for nearly two decades, making the precise timing of the impact of housing markets on heart attacks less transparent. Second, the unitary focus on heart attacks excludes observations of health sequelae or comorbidities that may identify other causes of heart attacks that may themselves be influenced by financial stress or may have been causes of debt prior to heart attack. A third relationship cannot be specified: the relationship between household formation and other confounders besides housing markets that can influence patterns of household formation and debt. The well-documented diversity in Boomer patterns of delayed marriage, disrupted marriage, cohabitation, and nonmarriage that could influence the association of homeownership with heart attack cannot be accounted for in these data. All three issues deserve more attention as data permit. Nevertheless, this cohort story has broken new ground in the life course analysis of SES and health with the focus on levels and forms of debt and heart attacks with age.

## Acknowledgments

This work was supported by NIA/NIH Grant P30-AG034424 to the Duke Center for Population Health and Aging and by the Duke University Trinity College of Arts & Sciences. We thank Scott M. Lynch for feedback on statistical modeling and Bryce Bartlett for early assistance with graphics.

## Authors’ Contributions

Angela M. O’Rand contributed to the study concept and design. Data preparation and analyses were performed by Jenifer Hamil-Luker with feedback from O’Rand. The first draft of the manuscript was drafted jointly by Angela M. O’Rand and Jenifer Hamll-Luker. Both authors edited subsequent versions and approved the final manuscript.

## Data Availability

All data sets used are publicly available from the Health and Retirement Study at the University of Michigan.

## Compliance With Ethical Standards

### Ethics and Consent

The authors report no ethical issues.

### Conflict of interest

The authors report no conflicts of interest.

## References

Aalen, O. O., Valberg, M., Grotmol, T., & Tretli, S. (
2015
).
Understanding variation in disease risk: The elusive concept of frailty
.
International Journal of Epidemiology
,
44
,
1408
1421
.
Afilalo, J., Alexander, K. P., Mack, M. J., Maurer, M. S., Green, P., Allen, L. A., & Forman, D. E. (
2014
).
Frailty assessment in the cardiovascular care of older adults
.
Journal of the American College of Cardiology
,
63
,
747
762
.
2015
).
Mortgage foreclosures and delinquencies continue to drop
.
Mortgage Bankers Association
.
Ahn, M., Batty, M., & Meisenzahl, R. R. (
2018
).
Household debt-to-income ratios in enhanced financial accounts
(FEDS Notes 2018-01-11).
Washington, DC
:
Board of Governors of the Federal Reserve System
.
Akushevich, I., Kravchenko, J., Ukraintseva, S., Arbeev, K., & Yashin, A. I. (
2012
).
Age patterns of incidence of geriatric disease in the U.S. elderly population: Medicare-based analysis
.
Journal of the American Geriatrics Society
,
60
,
323
327
.
Austin, D. (
2014
).
Medical debt as a cause of consumer bankruptcy
.
Maine Law Review
,
67
,
1
34
.
Ben-Shlomo, Y., & Kuh, D. (
2002
).
A life course approach to chronic disease epidemiology: Conceptual models, empirical challenges, and interdisciplinary perspectives
.
International Journal of Epidemiology
,
31
,
285
293
.
Berger, L. M., Collins, J. M., & Cuesta, L. (
2016
).
Household debt and adult depressive symptoms in the United States
.
Journal of Family and Economic Issues
,
37
,
42
57
.
Bernhardt, A., Morris, M., Handcock, M. S., & Scott, M. A. (
2001
).
Divergent paths: Economic mobility in the new American labor market
.
New York, NY
:
Russell Sage
.
Bratt, R. G., & Immergluck, D. (
2015
).
The mortgage crisis: Historical context and recent responses
.
Journal of Urban Affairs
,
37
,
32
37
.
Centers for Disease Control and Prevention
. (
2016
).
[Fact sheet].
Atlanta, GA
:
Centers for Disease Control and Prevention
Charles, K. K., & Hurst, E. (
2002
).
The transition to home ownership and the Black-White wealth gap
.
Review of Economics and Statistics
,
84
,
281
297
.
Clayton, M., Liñares-Zegarra, J., & Wilson, J. O. (
2015
).
Does debt affect health? Cross country evidence on the debt-health nexus
.
Social Science & Medicine
,
130
,
51
58
.
Crimmins, E. (
2018
).
Trends in mortality, disease, and physiological status in the older population
. In M. Hayward, & M. Majumdar (Eds.),
Future directions for the demography of aging
(pp.
3
30
).
Washington, DC
:
.
Curry, T., & Shibut, L. (
2000
).
The cost of the savings and loan crisis: Truth and consequences
.
FDIC Banking Review
,
13
(
2
),
26
35
.
Dean, L. T., Knapp, E. A., Snguon, S., Ransome, Y., Qato, D. M., & Visvanathan, K. (
2019
).
Consumer credit, chronic disease and risk behaviours
.
Journal of Epidemiology & Community Health
,
73
,
73
78
.
Dobkin, C., Finkelstein, A., Kluender, R., & Notowidigdo, M. J. (
2018
).
The economic consequences of hospital admissions
.
American Economic Review
,
108
,
308
352
.
Draut, T., & Silva, J. (
2004
).
Generation broke: The growth of debt among young Americans
(Policy Brief).
New York, NY
:
Dēmos
Drentea, P., & Lavrakas, P. J. (
2000
).
Over the limit: The association among health status, race and debt
.
Social Science & Medicine
,
50
,
517
529
.
Drentea, P., & Reynolds, J. R. (
2012
).
Neither a borrower nor a lender be: The relative importance of debt and SES for mental health among older adults
.
Journal of Aging and Health
,
24
,
673
695
.
Drentea, P., & Reynolds, J. R. (
2015
).
Where does debt fit in the stress process model?
Society and Mental Health
,
5
,
16
32
.
Dunn, L. F., & Mirzaie, I. A. (
2016
).
Consumer debt stress, changes in household debt, and the Great Recession
.
Economic Inquiry
,
54
,
201
214
.
Dwyer, R. E. (
2018
).
Credit, debt, and inequality
.
Annual Review of Sociology
,
44
,
237
261
.
Dwyer, R. E., Neilson, L. A., Nau, M., & Hodson, R. (
2016
).
Mortgage worries: Young adults and the US housing crisis
.
Socio-Economic Review
,
14
,
483
505
.
Easterlin, R. A. (
1980
).
Birth and fortune: The impact of numbers on personal welfare
.
Chicago, IL
:
University of Chicago
.
Ehlers, N., & Hinkson, L. R. (
2017
).
Subprime health: Debt and race in U.S. medicine
.
Minneapolis
:
University of Minnesota Press
.
Federal Reserve Bank of New York, Center for Microeconomic Data
. (
2019
).
Household debt and credit report
, Q2 2019.
New York
:
Federal Reserve Bank of New York
. https://www.newyorkfed.org/medialibrary/interactives/householdcredit/data/pdf/hhdc_2019q2.pdf
Fligstein, N., & Goldstein, A. (
2015
).
The emergence of a financial culture in American households, 1989–2007
.
Socio-Economic Review
,
13
,
575
601
.
Freddie Mac
. (
2017
).
30-year fixed rate mortgages since 1971
.
McLean, VA
:
Freddie Mac
Gathergood, J. (
2012
).
Debt and depression: Causal links and social norm effects
.
Economic Journal
,
122
,
1094
1114
.
Goodman, L., Kaul, K., & Zhu, J. (
2017
).
What the 2016 survey of consumer finances tells us about senior homeowners
.
Washington, DC
:
Urban Institute
.
Guo, G. (
2010
).
Survival analysis
.
New York, NY
:
Oxford University Press
.
Hamel, L., Norton, M., Pollitz, K., Levitt, L., Claxton, G., & Brodie, M. (
2016
).
The burden of medical debt: Results from the Kaiser Family Foundation/New York Times Medical Bills Survey
.
San Francisco, CA
:
Kaiser Family Foundation
.
Houle, J. N. (
2014
).
A generation indebted: Young and adult debt across three cohorts
.
Social Problems
,
61
,
448
465
.
Houle, J. N., & Warner, C. (
2017
).
In the red and back to the nest? Student debt, college completion and returning to the parental home among young adults
.
Sociology of Education
,
90
,
89
108
.
Hughes, M. E., & O’Rand, A. M. (
2005
).
The lives and times of the baby boomers
. In R. Farley, & J. Haaga (Eds.),
The American people: Census 2000
(pp.
224
255
).
New York, NY
:
Russell Sage
.
Hurd, M. D., & Rohwedder, S. (
2010
).
Effects of the financial crisis and great recession on American households
(NBER Working Paper No. 16407).
Cambridge, MA
:
National Bureau of Economic Research
.
Kahnemann, D., & Deaton, A. (
2010
).
High income improves evaluation of life but not emotional well-being
.
Proceedings of the National Academy of Sciences
,
107
,
16489
16493
.
Kalousova, L., & Burgard, S. A. (
2013
).
Debt and foregone medical care
.
Journal of Health and Social Behavior
,
54
,
203
219
.
Kathiresan, S., & Srivastava, D. (
2012
).
Genetics of human cardiovascular disease
.
Cell
,
148
,
1242
1257
.
Kim, J., & Chatterjee, S. (
2019
).
Student loans, health, and life satisfaction of US households: Evidence from a panel study
.
Journal of Family and Economic Issues
,
40
,
36
50
.
Knapp, E. A., & Dean, L. T. (
2018
).
Consumer credit scores as a novel tool for identifying health in urban U.S. neighborhoods
.
Annals of Epidemiology
,
28
,
724
729
.
Kochkar, R., Gonzalez-Barrera, A., & Ockterman, D. (
2009
).
Through boom and bust: Minorities, immigrants and homeownership
.
Washington, DC
:
Pew Hispanic Center
.
Kumhof, M., Rancière, R., & Winant, P. (
2015
).
Inequality, leverage and crises
.
American Economic Review
,
105
,
1217
1245
.
Liu, X. (
2014
).
Survival models on unobserved heterogeneity and their applications in analyzing large-scale survey data
.
Journal of Biometrics & Biostatistics
,
5
(
2
),
1
23
.
Masnick, G. S., Di, Z. X., & Belsky, E. S. (
2006
).
Emerging trends in housing debt and home equity
.
Housing Policy Debate
,
17
,
491
527
.
Myers, D., & Lee, H. (
2016
).
Cohort momentum and future homeownership: The outlook to 2050
.
Cityscape
,
18
(
1
),
131
144
.
National Bureau of Economic Research
. (
2020
).
Odekon, M. (
2015
).
Booms and busts: An encyclopedia of economic history from the first stock market crash of 1792 to the current global economic crisis
.
New York, NY
:
Routledge
.
Olegario, R. (
2019
).
The history of credit in America
. In J. Butler (Ed.),
Oxford research encyclopedia of American history
.
Oxford, UK
:
Oxford University Press
. Available from https://oxfordre.com/americanhistory.
Picketty, T., & Saez, E. (
2015
).
Top incomes and the Great Recession: Recent evolutions and policy implications
.
IMF Economic Review
,
61
,
456
478
.
Richardson, T., Elliott, P., & Roberts, R. (
2013
).
The relationship between personal unsecured debt and mental and physical health: A systematic review and meta-analysis
.
Clinical Psychology Review
,
33
,
1148
1162
.
Rosenbaum, E. (
2014
).
Cohort trends in housing and household formation since 1990
. In J. R. Logan (Ed.),
Diversity and disparities: America enters a new century
(pp.
181
207
).
New York, NY
:
Russell Sage
.
Sommer, I., Griebler, U., Mahlknecht, P., Thaler, K., Bouskill, K., Gartlehner, G., & Mendis, S. (
2015
).
Socioeconomic inequalities in non-communicable diseases and their risk factors: An overview of systematic reviews
.
BMC Public Health
,
15
,
914
. 10.1186/s12889-015-2227-y.
Sonnega, A., Faul, J. D., Ofstedal, M. B., Langa, K. M., Phillips, J. W., & Weir, D. R. (
2014
).
Cohort profile: The Health and Retirement Study (HRS)
.
International Journal of Epidemiology
,
43
,
576
585
.
Streeter, J., Sims, T., & Deevy, M. (
2018
).
Generational shifts in life course trajectories: Implications for homeownership by age 30
.
Stanford, CA
:
Stanford Center on Longevity
.
Sweet, E. (
2018
).
“Like you failed at life”: Debt, health and neoliberal subjectivity
.
Social Science & Medicine
,
212
,
86
93
.
Sweet, E., Nandi, A., Adam, E. K., & McDade, T. W. (
2013
).
The high price of debt: Household financial debt and its impact on mental and physical health
.
Social Science & Medicine
,
91
,
94
100
.
Turunen, E., & Hiilamo, H. (
2014
).
Health effects of indebtedness: A systematic review
.
BMC Public Health
,
14
,
489
. 10.1186/1471-2458-14-489.
Vaupel, J. W., Manton, K. G., & Stallard, E. (
1979
).
The impact of heterogeneity in individual frailty on the dynamics of mortality
.
Demography
,
16
,
439
454
.
Walsemann, K. M., Gee, G. C., & Gentile, D. (
2015
).
Sick of our loans: Student borrowing and the mental health of young adults in the United States
.
Social Science & Medicine
,
124
,
85
93
.
Wienke, A. (
2010
).
Frailty models in survival analysis
.
New York, NY
:
CRC Press
.
Yang, Y. (
2008
).
Social inequalities in happiness in the United States, 1972–2004: An age-period-cohort analysis
.
American Sociological Review
,
73
,
204
226
.

## Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.