The role of consumer and mortgage debt for financial stress

Abstract

Objectives

Financial debt held by older adults in the U.S. has grown over the past two decades. This study examines the extent to which credit cards, other consumer debts, and mortgage debt increase financial stress. Outcome measures of financial stress include the material domain (“bill-paying difficulty”) and psychological domain (“ongoing financial strain”).

Method

We analyzed adults age 62 and older in the 2004 to 2016 waves of the Health and Retirement Study using random-effects logit regressions.

Results

Unsecured consumer debt is associated with more financial stress per dollar than mortgage debt. A detailed assessment of mortgage debt finds that greater levels of both first and secondary mortgages are associated with greater bill-paying difficulty and greater ongoing financial strain. An increase in new mortgage debt obtained after age 62 is associated with an increase in bill-paying difficulty, but is not significantly associated with ongoing financial strain. In contrast, a reduction in mortgage debt since age 62 is associated with lower bill-paying difficulty and lower levels of ongoing financial strain.

Conclusion

The relationship between consumer debt, mortgages, and financial stress is nuanced, and depends on both the type and timing of the debt.

KEYWORDS:

• Quality of life/ well-being
• emotion/ worry
• financial stress
• mortgage borrowing

Introduction

The amount of financial debt held by older adults in the United States has grown substantially over the past decade, both in dollar terms and as a proportion of older adults carrying debt into retirement (Collins, Hembre, & Urban, 2018; Fichtner, 2019; Lusardi, Mitchell, & Oggero, 2018). The median total consumer debt of older adults in 2016 was $31,300, an increase of 250% from 2001 levels (National Council on Aging, 2018). Mortgage debt is the predominant form of debt held by older adults and is responsible for the largest increase in debt (Butrica & Karamcheva, 2019). The number of older adults who carry mortgage debt into retirement has doubled over the past two decades from 20% in 1992 to more than 40% in 2016.¹ In addition older adults rely on new mortgage borrowing to meet health care expenditures. Estimations based on the PSID show that a negative health condition is associated with a 13% reduction in home equity (Dalton & LaFave, 2017). Households borrow an average amount of $12,101 against home equity if married, and $4815 if unmarried, to cover the costs of acute, chronic, and psychosocial conditions if these are severely limiting an older adults’ daily activities (Dalton & LaFave, 2017).

Evidence suggests that indebtedness is associated with perceived stress across age groups (Drentea & Reynolds, 2015; Dunn & Mirzaie, 2016), which negatively affects individuals’ well-being in many areas of life, ranging from marital stability, career advancement, healing from illness, to social embeddedness (Addo, 2014; de Souza, Yap, Wroblewski, Blinder, & Arau, 2017; Rothstein & Rouse, 2011). Consequently, understanding the determinants of financial stress is an important area of research.

In this article, we focus specifically on older adults and their experience of financial stress. We consider both material and psychological domains of financial stress. We assess consumer debts, including credit cards and other non-housing consumer debts, and mortgage loans. We pay particular attention to different types and timing of mortgage debt for this population given the prominence of housing wealth as the primary asset in the retirement portfolio for a majority of older households (Goodman, Kaul, & Zhu, 2017; Mayer, 2017; Moulton, Haurin, Dodini, & Schmeiser, 2016). Data for the analysis come from multiple panels of the Health and Retirement Study (HRS).

Literature review

The use of debt can be part of a strategy to spend down assets in retirement, including otherwise illiquid home equity (Poterba, Venti, & Wise, 2011). An increasing trend toward use of debt in retirement may simply reflect rational behavior. On the other hand, debt may be associated with financial stress, with negative implications for economic, physical, and social well-being (Addo, 2014; de Souza et al., 2017; Rothstein & Rouse, 2011). We rely on two theoretical frameworks to investigate these assumptions. Our first theoretical framework is the stress process model (Pearlin, 1999), applied to the role of debt as suggested by Drentea and Reynolds (2015). The stress process model posits that a person’s economic circumstances affect their exposure to stress (Pearlin, Menaghan, Lieberman, & Mullan, 1981; Turner, 2013), such as debt accumulated in credit cards, personal loans, and mortgages can lead to worry about difficulty paying bills and feeling upset about financial burdens. When applying the stress process model to debt, Drentea and Reynolds (2015) identify financial debt as having a direct relationship with stress, rather than one that is mediated by social and personal resources. We also rely on the theoretical model of strength and vulnerability integration in the aging process (Charles, 2010). The framework states that older adults can experience stress when facing longer-term exposure to stressors, despite the strengths of older age that is grounded in lifelong experience, a change in perspective, and adapting to loss over a lifetime. We test this framework in the context of financial debt, by conceptualizing debt as ‘sustained exposure to highly arousing, inescapable negative situations’ (p. 1076) (Charles, 2010).

The current study builds on these frameworks to improve our understanding of whether different types of debt can create distinct stress exposures in older age. Specifically, we conceptualize financial stress as having two broad domains, a material domain and a psychological domain (for a literature review, see Altice, Banegas, Tucker-Seeley, & Yabroff, 2017; Tay, Batz, Parrigon, & Kuykendall, 2017). The material domain of financial stress reflects difficulty paying monthly bills and difficulty accumulating sufficient savings for emergencies or retirement (de Souza et al., 2014; Yabroff et al., 2016). This domain is well-established in the material deprivation literature (Whelan & Maître, 2013). Because debt requires repayment, it can increase financial stress. About 45% of older homeowners with mortgages spend more than 30% of their incomes on monthly housing costs, which has been shown to directly affect their ability to afford food, healthcare, and transportation (JCHS, 2016).

The psychological domain of financial stress is related to individuals’ sense of mastery and coping abilities and, in turn, to increased feelings of hopelessness, frustration, and anxiety (Cohen, Salonen, & Kaplan, 1999; Drentea & Reynolds, 2015; Gallo & Matthews, 2003). Older adults are later in their life cycle with less time available to repay mortgage debt, which may make older adults feel more constrained by their debt and intensify feelings of financial stress (Lusardi, 2012). A small number of studies have documented the association of debt and mental health in older adults (Hiilamo & Grundy, 2020; Richardson, Elliott, & Roberts, 2013). Drentea and Reynolds (2012) find that the debt-to-depression relationship is moderated by financial stress and that debt is a consistent and more important socio-economic predictor of mental health in older adults than income or assets. The role of individual types of debt, such as credit card debt, other non-housing consumer debt, and mortgage debt, has received limited attention. An exception is a recent European study based on SHARE data for Germany, Belgium, and France that distinguishes between housing debt and non-housing debt, even if only in two broad categories, as predictors of depression of adults aged 50 and older (Hiilamo & Grundy, 2020). They find that housing and non-housing debts are similarly strong predictors of depression in women; among men, non-housing debt is a stronger predictor of depression than housing debt (Hiilamo & Grundy, 2020). Except for this study, there is a lack of understanding of how different types of debt carried into older age are associated with financial stress.

Research objectives

Our first research objective is to investigate the association of the level of financial stress in the material and psychological domains with housing and non-housing debt, following suggestions in the stress process model literature. Different types of debt have been associated with different levels of stress in prior research. In these studies, a composite stress measure is used which does not allow distinguishing between different domains of stress. Individuals report the highest levels of financial stress for payday loans, followed by credit card debt, student loans, informal loans from family and friends, and bank loans (Dunn & Mirzaie, 2016; Hojman, Miranda, & Ruiz-Tagle, 2016). Debt that is secured by property is perceived as less stressful, with stress being lower for loans for cars, appliances, and furniture and lowest for mortgage debt (Dunn & Mirzaie, 2016; Shen, Sam, & Jones, 2014). While prior studies have been conducted on the general population, we expect to observe similar responses for older adults, with consumer debt that is not backed by property being associated with significantly higher financial stress than mortgage debt in both the material and psychological domains.

Not all mortgage debt is the same, and thus, our second research objective is to examine whether financial stress in the material and psychological domains differ for the two broad types of mortgages, first and secondary mortgages. First mortgages are paid off before secondary mortgages if a homeowner sells the home or is unable to make mortgage payments and the home is liquidated. Typically, first mortgages are the loans that a household used to buy or build the home. Secondary mortgages typically include second mortgages, home equity lines of credit (HELOCs), and other mortgages, such as construction loans. Dunn and Mirzaie (2016) evaluate the financial stress associated with two types of mortgages, first mortgages and HELOCs. While first mortgages provide homeowners with a lump sum payment at origination of the loan, HELOCs work similar to a credit card and allow homeowners to withdraw money over a set period of time, usually up to 10 years. HELOCs were associated with slightly greater financial stress compared to debt from a first mortgage. Secondary mortgages typically have a shorter repayment term and a higher interest rate and, for this reason, may be associated with higher levels of financial stress than first mortgages among older adults as well. This approach to distinguish first from secondary mortgages follows again the stress process model’s proposition that stressors are economically patterned (Drentea & Reynolds, 2015).

A third research objective of this study is whether mortgage debt that is acquired before retirement age and carried into older age is more or less stressful than ‘new’ mortgage debt obtained after retirement age, testing the notion of sustained, inescapable economic burden (Charles, 2010). This research question has not been addressed in the financial stress literature. It is possible that new mortgage debt obtained after age 62 is related to higher levels of financial stress than debt obtained earlier because new debt increases the monthly mortgage payment and reduces discretionary income, thus, contributing to feelings of material deprivation (Whelan, Layte, Maitre, & Nolan, 2001). However, new mortgage debt may have a low association with financial stress in the material and psychological domains if the new mortgage debt is instrumental for achieving a specific goal, such as meeting medical expenses, helping with educational expenses of children, or if the new debt reflects financial planning strategies that have positive long-run outcomes (Dwyer, McCloud, & Hodson, 2011; Marino, 2017; Tay et al., 2017). Conversely, debt obtained earlier in life that is carried into retirement may reflect lower income (Collins et al., 2018), pointing to the vulnerabilities of a life-long struggle to pay off the mortgage balance, as identified in the theoretical model of strength and vulnerability integration (Charles, 2010). We therefore expect that reducing a mortgage balance in older age is associated with lower levels of financial stress in the material and psychological domains.

Analytic strategy

We estimate separate regression models for two binary measures of financial stress, representing the material domain (‘bill-paying difficulty’) and psychological domain (‘ongoing financial strain’) of financial stress, using random-effects logit regressions of adults age 62 and older.² We chose this age threshold because it represents the age when most individuals in the United States retire and claim Social Security retirement benefits (Social Security Administration, 2019). To examine the first research objective, we regress both measures of financial stress on lagged measures of credit card debt, other non-housing consumer debt, and total mortgage debt. In the first specification, total mortgage debt combines four types of mortgage debt into one measure, including first mortgages, HELOCs, second mortgages, and other mortgages on the primary residence. For the second research objective, we split our indicators of mortgage debt into first mortgage debt and the other three types of mortgage debt.

For the third research objective, we estimate the relationship of both measures of financial stress with new mortgage debt obtained after the age 62 and before age 62. We limit the sample to respondents who turned 62 between the 1996 and the 2006 or 2008 survey waves and add to the primary specification a measure of the difference between respondents’ current mortgage balance and their mortgage balance at age 62. Next, we separate out increases in the mortgage balance since the age of 62, which indicate new mortgage debt, and decreases to the balance since the age of 62, which indicate debt repayment. We, thus, estimate a specification that splits the change in mortgage debt into positive and negative differences.

All regression specifications account for a rich set of socio-economic household and respondent characteristics. The independent variables are lagged by one survey wave (2 years) to break the obvious simultaneity that would occur if we used contemporaneous indictors of debt and financial stress. However, because it is likely that there are omitted factors that affect both financial stress and our lagged independent variables, we do not claim to identify causal effects.

Data

The Health and Retirement Study

We employ the HRS, a well-regarded survey of American adults age 50 and older, with a response rate above 80%. Respondents are surveyed every two years with new birth cohorts added to the existing sample every three waves. Each wave has around 20,000 respondents. The HRS has detailed wealth information and contains two financial stress questions. We use the seven HRS waves from 2004 to 2016 for the present study. All variables are from the RAND HRS.³ We use the responses of the ‘financial respondent’ in a household for the household-level debt, income, and asset questions (Sonnega et al., 2014). The responses of all household members are used for the individual-level questions. Both homeowners and renters are included in the sample, with 83% being homeowners.

The sample is limited to respondents who were age 62 and older when the two financial stress measures, bill-paying difficulty and ongoing financial strain, were first used, which are the survey waves 2006 and 2008. The financial stress measures are part of the supplementary leave-behind questionnaire of the HRS. About half of the sample answered the leave-behind questionnaire in 2006, 2010, and 2014. The second half answered the leave-behind questionnaire in 2008, 2012, and 2016. As a result of the panel nature of the data, the full sample consists of 5084 respondents for bill-paying difficulty and 4990 respondents for ongoing financial strain who were aged 62 and older in 2006 or 2008. They represent 12,740 and 10,311 respondent-year observations, respectively, corresponding to an average of 2.5 survey waves for bill-paying difficulty and 2.1 waves for ongoing financial strain (range: 1–3 waves).

Measures of financial stress

We use two measures of financial stress, bill-paying difficulty representing the material domain of financial stress and ongoing financial strain representing the psychological domain, following earlier studies based on HRS data (Ayyagari, Deb, Fletcher, Gallo, & Sindelar, 2013; Brown & Hargrove, 2018; Marshall & Tucker-Seeley, 2018; Tucker-Seeley, Marshall, & Yang, 2016). We focus the analysis on the dichotomous versions, which measure exposure to financial stress (Brown, Mitchell, & Ailshire, 2020). The results are similar in direction and size if the continuous measures are used as dependent variables.

Bill-paying difficulty was inquired with the question, ‘How difficult is it for you to meet monthly payments on your bills?’ Following previous HRS-based studies, the dichotomous variable is coded as 0 for the responses not at all difficult (=1) or not very difficult (=2) and 1 if paying bills was somewhat, very, or completely difficult (=3, 4, and 5) (Marshall & Tucker-Seeley, 2018).

Ongoing financial strain was inquired with the question, ‘Please read the list below and indicate whether or not any of these are current and ongoing problems that have lasted twelve months or longer. If the problem is happening to you, indicate how upsetting it has been’. Following previous studies, the dichotomous variable is coded as 0 if no strain occurred (=1) and 1 if strain occurred but was not upsetting, was somewhat upsetting, or very upsetting (=2, 3, and 4) (Birditt, Newton, Cranford, & Ryan, 2016; Marshall, Baker, Song, & Miller, 2018; Marshall & Tucker-Seeley, 2018). Ongoing financial strain was excluded in the 2008 HRS, which reduces the sample size for models with this outcome. Because bill-paying difficulty and ongoing financial strain were first asked in 2006, the regressions use lagged control variables beginning in 2004. As expected, the two financial stress measures are significantly correlated (Pearson r = 0.68, p < .05) for adults aged 62 and older.

Debt measures

We include two measures to capture non-housing consumer debt; they are credit card debt and other non-housing consumer debt. Credit card debt is measured as the balance of credit card debt that was carried over from the prior month. In addition, we create another non-housing consumer debt measure by subtracting credit card debt from the consumer debt variable. Other non-housing consumer debt includes medical debts, loans from life insurance policies, and loans from relatives. Per definitions in the HRS, this measure does not include loans for cars or other vehicles or money owed on other assets. The HRS started to collect credit card balance data in 2008. For the 2004 and 2006 waves, we use the value in 2008.

Mortgage debt on the primary residence consists of four types: first mortgages, HELOCs, second mortgages, and other mortgages. Some respondents in the HRS own a second home and hold a mortgage on their second home. Secondary residence mortgage balance is a total measure of all mortgages for the secondary residence. We use the one-wave lagged balance of each type of mortgage. To measure the change in mortgage debt, we take the difference of the amount of mortgage debt as of the current survey wave and the amount of mortgage debt when the respondent was 62-year old. We also split the change in mortgage debt into two separate indicators: decreases (mortgage debt repayment) and increases (new mortgage debt) since age 62. For renters, we control for the size of the monthly rent payment.

With regard to the quality of the debt measures, which rely on survey respondents’ recall, the HRS has instituted a reconciliation and cross-wave imputation approach for wealth measures (Hurd, Meijer, Moldoff, & Rohwedder, 2016), but not for debt measures. As a result, we carefully assessed outliers and set the top 1% of values to missing.

Measures of household and respondent characteristics

Household characteristics include the number of household members, location of residence (urban (reference group), suburban, rural area), and nine regional indicators (New England [reference group], Middle Atlantic, East North Central, West North Central, South Atlantic, East South Central, West South Central, Mountain, Pacific). Financial characteristics include the financial respondents and their spouses or partners’ income from Social Security retirement and disability income, earnings, and other income. Assets include household-level net cash assets (including cash in checking accounts and Certificates of Deposit), net investment assets (including assets held in Individual Retirement Arrangements, stocks, bonds, and trust accounts), and net other assets (including other savings, non-housing real estate, transportation, and business assets). The home values of the primary and secondary residences are based on the HRS respondents’ estimates. We drop outliers from the sample with values of financial variables at or above the 99th percentile.

Individual characteristics include age, gender (male = 1, female = 0), race (white [reference group], black, other), Hispanic origin (yes = 1, no = 0), immigrant (yes = 1, no = 0), educational attainment (less than high school education, GED, high school diploma [reference group], some college, college degree, or more), marital status (married [reference group], separated/divorced/widowed, never married), number of living children of the respondent and spouse (continuous), any health insurance, including government, employer, or other health insurance (yes = 1, no = 0), self-rated health and general memory status (poor = 1, excellent = 5). We account for respondents’ social support network with a dummy variable that indicates friends or relatives to provide help if needed.

Results

Descriptive analysis

The characteristics of the regression sample are presented in Table 1 (focal variables) and Appendix Table A1 (control variables). Overall, 26% of our sample report bill-paying difficulty and 36% report ongoing financial strain. The average credit card debt is $1088, the average non-housing other consumer debt is $986, and the average total mortgage debt on the primary residence amounts to $19,720 of which $16,793 is for the first mortgage and $2927 for secondary mortgages. The HELOC balance averages $2339. It is the largest component of secondary mortgages. Households in our sample hold average net cash assets of $38,385, investment assets of $101,140, and an average home value of $163,708. Annual household income consists of an average of $7766 in earned income, $18,318 in Social Security benefits, and $21,803 in pensions and other income sources. The average household size is two persons. About 39% of respondents are male, the average age is 74 years, 12% are black, and 7% are Hispanic. About 8% were not born in the United States. The largest group, about 37%, has a high school education. Almost two-thirds, 63%, are married or partnered and about one-third are separated, divorced, or widowed. The average number of living children is three. A total of 95% has health insurance. Few suffer from difficulty with activities of daily living (ADLs), averaging 0.2 occurrences on a scale from 0 to 5. Overall health is rated at an average of 3.2 and overall mental health at an average of 2.9 on a scale from 1 to 5. About 66% can receive help from friends or family if needed.

Table 1. Sample financial stress and mortgage characteristics.

Variables (Range)	Mean (SD) or %
Bill-paying difficulty (0/1)	25.95%
Ongoing financial strain (0/1)	36.15%
Credit card debt, lagged (in $10,000; 0–9.84)	0.11 (0.43)
Other non-housing consumer debt, lagged (in $10,000; 0–23.76)	0.10 (0.46)
PrRes Total mortgage debt, lagged (in $10,000; 0–108.39)	1.97 (5.23)
PrRes First mortgage debt lagged (in $10,000; 0–108.39)	1.68 (4.93)
PrRes All secondary mortgage debt, lagged (in $10,000; 0–27.53)	0.29 (1.50)
PrRes HELOC debt, lagged (in $10,000; 0–27.53)	0.23 (1.32)
PrRes Other secondary mortgage debt, lagged (in $10,000; 0–22.02)	0.06 (0.68)
SecRes All mortgage debt, lagged (in $10,000; 0–127.00)	0.13 (3.35)

Notes: Number of individuals: 5084; number of observations: 12,740.

Unweighted sample, descriptives reflect regression sample.

All financial variables are in $10,000 units.

PrRes: Primary residence; SecRes: Secondary residence.

Remaining sample characteristics of the household and respondent are shown in Appendix Table A1.

As a first step, we explore the characteristics of older adults relative to their financial stress. Columns (1) and (2) of Table 2 report the sample means for older adults who report bill-paying difficulty and those who do not. Columns (3) and (4) of Table 2 present the sample means for individuals who report ongoing financial strain and those who do not. All variables are lagged one HRS wave with regard to the two financial stress measures.

Table 2. Means comparison of bill-paying difficulty and ongoing financial strain with regard to mortgage characteristics.

Variables	Bill-paying difficulty				Ongoing financial strain
	(1) Yes	(2) No	Signf.		(3) Yes	(4) No	Signf.
	Mean or %	Mean or %			Mean or %	Mean or %
Bill-paying difficulty	100.00%	0.00%			58.46%	6.63%	***
Ongoing financial strain	83.31%	20.21%	***		100.00%	0.00%
Credit card debt, lagged	0.20	0.08	***		0.18	0.06	***
Other non-housing consumer debt, lagged	0.15	0.08	***		0.12	0.06	***
PrRes Total mortgage debt, lagged	2.48	1.80	***		2.58	1.54	***
PrRes First mortgage debt, lagged	2.15	1.52	***		2.22	1.28	***
PrRes All secondary mortgage debt, lagged	0.33	0.28			0.36	0.26	**
PrRes HELOC debt, lagged	0.25	0.23			0.27	0.22
PrRes Other secondary mortgage debt, lagged	0.09	0.05	*		0.09	0.04	***
SecRes All mortgage debt, lagged	0.14	0.13			0.17	0.13
Number observations	3306	9434			3698	6531

Note: *** p < .001, ** p < .01, * p < .05.

Unweighted sample, descriptives reflect regression sample.

All financial variables are in $10,000 units.

PrRes: Primary residence; SecRes: Secondary residence.

As shown in Table 2, older adults who report financial stress, either due to bill-paying difficulty or due to ongoing financial strain, differ in their debt holdings. Older adults who experience both measures of financial stress carry higher amounts of credit card debt, other non-housing consumer debt, and total mortgage debt. With regard to mortgage debt, first mortgage debt is more strongly associated with both measures of financial stress than secondary mortgage debt. The sum of secondary mortgages as well as the particular amount held in HELOC debt is not different among those who report bill-paying difficulty and those who do not. With regard to older adults who report ongoing financial strain, they hold higher amounts of secondary mortgage debt compared to those who do not report ongoing financial strain. The finding is due to higher balances in other secondary mortgage debt because there was no difference in the balance of HELOC debt. Appendix Table A2 provides the differences in means for household and respondent characteristics.

Research objective 1: Financial stress and debt type

We begin with a random-effects logit model of both measures of financial stress. The results are presented in Table 3 (focal variables) and Appendix Table A3 (full specification). Columns (1), (3), and (5) present results for bill-paying difficulty; Columns (2), (4), and (6) for ongoing financial strain.

Table 3. Odds ratios of random-effects logit regression of binary measure of bill-paying difficulty and ongoing financial strain on credit card debt, mortgage debt, and other explanatory variables (focal variables).

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	Bill-paying difficulty	Ongoing financial strain	Bill-paying difficulty	Ongoing financial strain	Bill-paying difficulty	Ongoing financial strain
	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)
Credit card debt, lagged	1.646***	1.929***	1.644***	1.928***	1.641***	1.925***
	(0.158)	(0.199)	(0.157)	(0.199)	(0.156)	(0.198)
Other non-housing	1.125*	1.201**	1.123*	1.200**	1.123*	1.199**
consumer debt, lagged	(0.067)	(0.081)	(0.067)	(0.081)	(0.067)	(0.081)
PrRes Total mortgage	1.048***	1.056***
debt, lagged	(0.009)	(0.011)
PrRes First mortgage			1.045***	1.055***	1.045***	1.055***
debt, lagged			(0.009)	(0.011)	(0.009)	(0.011)
PrRes All secondary			1.086***	1.067**
mortgage debt, lagged			(0.024)	(0.024)
PrRes HELOC					1.069**	1.047
debt, lagged					(0.027)	(0.026)
PrRes Other secondary					1.139**	1.145*
mortgage debt, lagged					(0.050)	(0.061)
SecRes All mortgage	1.002	1.004	1.002	1.004	1.002	1.004
debt, lagged	(0.010)	(0.008)	(0.010)	(0.008)	(0.010)	(0.008)

Notes: *** p < .001, ** p < .01, * p < .05.

Robust standard errors in parentheses.

All financial variables are in $10,000 units.

PrRes: Primary residence; SecRes: Secondary residence.

The credit card balance is the strongest predictor of both measures of financial stress, stronger than other non-housing consumer debt and mortgage debt. A $10,000 higher amount of credit card debt is related to 64.6% higher odds of reporting bill-paying difficulty and almost doubles the odds of reporting ongoing financial strain (92.9%).

Total debt from mortgages held on the primary residence is a significant predictor of both measures of financial stress but its role is much smaller compared to credit card debt (bill-paying difficulty: chi2(1)=21.52, p < .001; ongoing financial strain: chi2(1)=32.86, p < .001). This result confirms earlier literature. A $10,000 greater amount of total mortgage debt is related to 4.8% higher odds of reporting bill-paying difficulty and 5.6% higher odds of reporting ongoing financial strain.

As shown in Appendix Table A3, other financial variables are also significantly associated with the two measures of financial stress. Among household assets, higher levels of cash assets and investment assets are associated with lower odds of bill-paying difficulty and ongoing financial strain, with cash assets being associated with the largest reduction in the odds of the two financial stress measures. Higher house values are also associated with lower odds of bill-paying difficulty and ongoing financial strain. Turning to income, higher levels of earned income and other income, which includes pension income, are associated with lower odds of bill-paying difficulty. Social Security income is not significantly associated with either measure of financial stress. This result holds when measuring financial stress with scales. While the odds ratios of all three income measures are similar in size and direction, the descriptive statistics in Appendix Table A1 indicate that the standard deviation of the Social Security income variable is much smaller (SD/mean = 0.51) compared to earned income and other income (SD/mean = 2.92 and 2.61, respectively), indicating low variation in this measure.

All models control for a vector of household characteristics. A larger household size is associated with higher odds of both measures of financial stress. Among respondent characteristics, those who are younger, with a greater number of ADL difficulties, lower self-rated health, and lower self-rated memory are associated with higher odds of both measures of financial stress.

Respondents who are female, no longer married (separated, divorced, or widowed), and do not have a college degree have higher odds of bill-paying difficulty. Hispanic ethnicity is associated with lower odds of ongoing financial strain. Unrelated to both measures of financial stress are race, immigrant status, number of living children, health insurance, and having friends or family nearby. For a robustness check, we repeated the regressions by omitting renters; the results are similar in direction and size.

Research objective 2: Financial stress and mortgage types

Among the types of mortgage debt, both first mortgages and secondary mortgages are positively and significantly related to the two measures of financial stress. The association with secondary mortgages is slightly stronger than the association with a first mortgage. Each additional $10,000 of first mortgage debt is related to 4.5% higher odds of reporting bill-paying difficulty, whereas each additional $10,000 on secondary mortgages is related to 8.6% higher odds (chi2(1)=2.68, p = .10). Similarly, an additional $10,000 of first mortgage debt is related to 5.5% higher odds of reporting ongoing financial strain, whereas $10,000 in secondary mortgages increases the odds by 6.7% (chi2(1)=0.24, p = .62).

Among the types of secondary mortgages, both HELOCs and other secondary mortgages significantly increase the odds of both measures of financial stress (p < .05). The odds of financial stress from other secondary mortgages are about twice as high as from HELOCs, 13.9% for bill-paying difficulty (HELOCs: 6.9%) and 14.5% for ongoing financial strain (HELOCs: 4.7%). Mortgage debt held on second residences is not associated with older adults reporting either bill-paying difficulty or ongoing financial strain.

Research objective 3: Old vs. new mortgage debt

In an alternative set of specifications, we limit the sample to those age 62 and older and measure amount of mortgage debt held as of age 62. We then add indicators for the change in mortgage debt since the age of 62 and whether it is an increasing amount or decreasing amount. The average amount of changes in the mortgage balance is –$21,549 (SD=$155,143) with new mortgage debt amounting to an average of $8907 (SD=$35,776) and mortgage debt repayment to an average of –$30,456 (SD=$149,154).

Table 4 and Appendix Table A4 present the findings about the role of changes in mortgage debt since turning 62 for bill-paying difficulty and ongoing financial strain. Columns 1 and 2 in Table 4 show that the mortgage balance at age 62 as well as the difference in mortgage balance since age 62 are both associated with 6% to 7% higher odds of bill-paying difficulty and ongoing financial strain. We then split the mortgage-difference variable into mortgage increases (‘new mortgage debt’) and decreases (‘mortgage debt repayment’), see Columns 3 and 4 of Table 4. A $10,000 decrease in the mortgage balance since age 62 due to repayment is associated with 6.5% lower odds of bill-paying difficulty and 11.3% lower odds of ongoing financial strain.⁴ In contrast, a $10,000 increase in the mortgage balance since age 62 due to new mortgage debt is associated with 4.7% higher odds of bill-paying difficulty. The association with ongoing financial strain is insignificant at the 5% level, indicating that old mortgage debt carried into retirement can be associated with higher odds of ongoing financial strain, while new mortgage debt after age 62 cannot.

Table 4. Odds ratios of random-effects logit regression of binary measures bill-paying difficulty and ongoing financial strain on credit card debt, mortgage debt at age 62, new mortgage debt since age 62, and other explanatory variables (focal variables).

Variables	(1)	(2)	(3)	(4)
	Bill-paying difficulty	Ongoing financial strain	Bill-paying difficulty	Ongoing financial strain
	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)
Credit card debt, lagged	1.494***	1.856***	1.486***	1.811***
	(0.169)	(0.240)	(0.168)	(0.229)
Other non-housing consumer debt, lagged	1.169*	1.354***	1.167*	1.350***
	(0.088)	(0.116)	(0.088)	(0.114)
PrRes Total mortgage debt at 62, lagged	1.058***	1.071***	1.067***	1.116***
	(0.012)	(0.014)	(0.015)	(0.018)
PrRes Total mortgage difference at 62 to t – 1	1.056***	1.066***
	(0.012)	(0.014)
PrRes New mortgage debt			1.047**	1.020
(Positive mortgage difference at 62 to t – 1)			(0.017)	(0.015)
PrRes Mortgage debt repayment			1.065***	1.113***
(Negative mortgage difference at 62 to t – 1)			(0.015)	(0.018)
SecRes All mortgage balances at 62, lagged	0.980	0.986	0.980	0.987
	(0.011)	(0.011)	(0.011)	(0.011)

Notes: *** p < .001, ** p < .01, * p < .05.

Robust standard errors in parentheses.

All financial variables are in $10,000 units.

In summary, the increase in new mortgage debt, when compared to the total balance of old mortgage debt at age 62 is less stressful for both outcome measures based on the size of the coefficients. The difference in the coefficients for mortgage increase and decrease since age 62 is statistically significant for ongoing financial strain (chi2(1)=15.76, p < .001), indicating an asymmetric relationship between ongoing financial strain and the change in mortgage debt after age 62. While mortgage debt repayment reduces ongoing financial strain, taking out additional mortgage debt after age 62 does not significantly increase ongoing financial strain.

Discussion

The current study contributes new insights about the association of credit card debt, other non-housing consumer debt, and mortgage debt with two measures of financial stress, bill-paying difficulty and ongoing financial strain. First, the results indicate that credit card debt is the strongest predictor of both measures of financial stress in both the material and psychological domains among older adults, compared to other non-housing consumer debt and mortgage debt. These findings support the direct association of debt and financial stress, as first proposed by Drentea and Reynolds (2015) on the basis of the stress process model (Pearlin, 1999). We contribute new evidence that this association extends to both material and psychological domains of financial stress and that the type of debt matters within this theoretical framework. Second, we find that an increase in the mortgage balance after age 62 is associated with higher odds of bill-paying difficulty but is not related to ongoing financial strain. It could be that new mortgage debt originated after age 62 allows for additional liquidity (in the form of cash for consumption) or is used for discretionary purchases, and thus may not be associated with the same levels of stress as debt carried into retirement (Marino, 2017). However, because new mortgage debt still needs to be repaid, it makes sense that it is associated with an increase in difficulty paying bills. In contrast, a reduction in mortgage debt since age 62 is associated with lower bill-paying difficulty and ongoing financial strain. The findings reflect the strengths and vulnerabilities of aging and identify pathways of stress experiences in older age (Charles, 2010).

There are limitations to our analysis. Both financial stress measures are based on study participants’ recall over a two-year period and rely on their accurately remembering stressful events (for detailed discussion, see Brown et al., 2020). A second limitation is our use of binary debt stress measures instead of their continuous form. This decision was made because the response options for the ongoing financial strain measure follows a binary pattern of ‘no, didn’t happen’ (codes as 1) and ‘yes, did happen’ (coded as 2 if not upsetting, 3 if somewhat upsetting, and 4 if very upsetting). To parallel both stress measures and to align with common procedures in the literature, we coded both as binary variables (Marshall & Tucker-Seeley, 2018).

Conclusion

The present study adds to a small body of literature that documents the relationship between financial debt and older households’ mental health. The results of our analyses suggest to researchers and policymakers a need for conceptualizing debt in retirement in more nuanced ways. It is common in the health literature to measure net worth or total monthly debt divided by income as a single construct, which misses the differential stress contributed by housing debt relative to other forms of debt. Our findings thus have important implications for practice. There is increasing awareness among health practitioners that financial stress can negatively affect older adults’ mental health (Marshall, Kahana, Gallo, Stansbury, & Thielke, 2020). Our study highlights the importance of not simply measuring net wealth or income when considering an individual’s financial health, but rather paying attention to the different sources of debt held by older adults, as debt held in the form of credit cards or personal loans is likely to contribute more stress than debt held in the form of a mortgage. In addition, our study highlights the importance of financial products that provide affordable access to housing wealth in retirement, such as the Federal Housing Administration’s Home Equity Conversion Mortgage, which is a reverse mortgage for adults age 62 and older. Future research is needed that examines how borrowing from home equity in retirement may positively or negatively affect the mental health of older adults.

Acknowledgements

The opinions and conclusions expressed are solely those of the authors and do not represent the opinions or policy of U.S. Social Security Administration or any agency of the Federal Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of the contents of this report. The substance and findings of the work are dedicated to the public. The author and publisher are solely responsible for the accuracy of the statements and interpretations contained in this publication. Such interpretations do not necessarily reflect the view of the Government.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported as part of the Retirement and Disability Consortium through the University of Wisconsin Retirement Research Center under grant RRC08098401.

Notes

1 Authors’ calculations using the 2016 Survey of Consumer Finances data

2 We use the xtlogit procedure in Stata.

3 Information on RAND HRS: https://www.rand.org/well-being/social-and-behavioral-policy/centers/aging/dataprod.html.

4 Mortgage debt is splined with both increases and decreases measured as positive values.

 

Appendix

Table A1. Sample household and respondent characteristics

Variables (Range)	Mean (SD) or %
Net cash assets, lagged (in $10,000; 0–189.75)	3.84 (8.33)
Net investment assets, lagged (in $10,000; 0–397.12)	10.11 (23.14)
Net other assets, lagged (in $10,000; −2.32–1104.19)	6.13 (23.38)
House value, lagged (in $10,000; 0–594.00)	16.37 (15.85)
Secondary residence house value, lagged (in $10,000; 0–96.30)	1.23 (5.30)
Annual rent, lagged (in $10,000; 0–19.24)	0.09 (0.38)
Social security income, lagged (in $10,000; 0–12.10)	1.83 (0.93)
Earned income, lagged (in $10,000; 0–47.12)	0.78 (2.28)
Other income, lagged (in $10,000; 0–257.17)	2.18 (5.69)
Number household members, lagged (1–12)	2.01 (0.95)
Urban residence, lagged (0/1)	43.72
Suburban residence, lagged (0/1)	24.60
Rural residence, lagged (0/1)	31.68
Male (0/1)	38.97
Age, lagged (62–102)	73.93 (6.65)
White (0/1)	84.69
Black (0/1)	12.32
Other race (0/1)	2.98
Hispanic (0/1)	6.94
Not born in United States (0/1)	7.74
Less than high school (0/1)	20.86
High school (0/1)	36.77
GED (0/1)	5.48
Some college (0/1)	21.33
College or more (0/1)	15.56
Married or partnered, lagged (0/1)	63.05
Separated, divorced, widowed, lagged (0/1)	34.36
Never married, lagged (0/1)	2.59
Number of living children, lagged (0–20)	3.42 (2.16)
Health insurance, lagged (0/1)	95.19
Activities of daily living (ADL), lagged (0–5)	0.20 (0.66)
Self-rated health, lagged (1–5)	3.18 (1.02)
Self-rated memory, lagged (1–5)	2.93 (0.88)
Receive help, lagged (0/1)	66.25

Notes: Number of individuals: 5084; number of observations: 12,740.

All financial variables are in $10,000 units.

Unweighted sample, descriptives reflect regression sample.

Table A2. Means comparison of those who experience financial stress: Household and respondent characteristics.

Variables	Bill-paying difficulty				Ongoing financial strain
	(1) Yes	(2) No	Signf.		(3) Yes	(4) No	Signf.
	Mean or %	Mean or %			Mean or %	Mean or %
Annual rent, lagged (in $10,000)	0.13	0.08	***		0.13	0.07	***
Net cash assets, lagged (in $10,000)	1.41	4.69	***		2.04	4.98	***
Net investment assets, lagged (in $10,000)	2.75	12.69	***		4.02	13.75	***
Other assets, lagged (in $10,000)	3.29	7.12	***		3.80	7.32	***
House value, lagged (in $10,000)	12.21	17.83	***		13.12	17.58	***
Secondary residence house value, lagged (in $10,000)	0.77	1.39	***		0.90	1.34	***
Household social security income, lagged (in $10,000)	1.69	1.88	***		1.79	1.91	***
Household earned income, lagged (in $10,000)	0.67	0.82	***		0.68	0.75
Household other income, lagged (in $10,000)	1.12	2.55	***		1.45	2.61	***
Number household members, lagged	2.13	1.97	***		2.09	1.96	***
Urban residence, lagged	43.13	43.93			43.89	43.78
Suburban residence, lagged	25.32	24.35			25.20	23.96
Rural residence, lagged	31.55	31.73			30.91	32.26
Male	33.55	40.87	***		37.02	41.08	***
Age, lagged	73.00	74.25	***		73.92	74.90	***
White	77.53	87.21	***		80.64	87.49	***
Black	18.78	10.06	***		16.06	9.88	***
Other race	3.69	2.73	**		3.30	2.63
Hispanic	9.89	5.90	***		7.27	6.43
Not born in United States	9.92	6.97	***		7.76	7.23
Less than high school	28.98	18.01	***		22.66	18.86	***
High school	36.03	37.04			35.80	37.59
GED	7.02	4.94	***		6.41	4.85	**
Some college	19.42	22.01	**		22.85	20.73	*
College or more	8.56	18.01	***		12.28	17.96	***
Married or partnered, lagged	55.96	65.54	***		58.90	65.61	***
Separated, divorced, widowed, lagged	41.44	31.87	***		38.43	32.09	***
Never married lag	2.60	2.59			2.68	2.30
Number of living children, lagged	3.63	3.34	***		3.52	3.34	***
Health insurance lag	94.43	95.45	*		96.19	96.29
Activities of daily living (ADLs), lagged	0.36	0.15	***		0.30	0.15	***
Self-rated health, lagged	2.85	3.29	***		2.95	3.31	***
Self-rated memory, lagged	2.78	2.98	***		2.83	2.99	***
Receive help, lagged	66.72	66.10			66.18	65.97
Number observations	3306	9434			3698	6531

Note: *** p < .001, ** p < .01, * p < .05.

Unweighted sample, descriptive reflect regression sample.

All financial variables are in $10,000 units.

Table A3. Odds ratios of random-effects logit regression of a binary measure of financial stress on credit card debt, mortgage debt, and other explanatory variables (full specification).

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	Bill-paying difficulty	Ongoing financial strain	Bill-paying difficulty	Ongoing financial strain	Bill-paying difficulty	Ongoing financial strain
	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)
Credit card balance,	1.646***	1.929***	1.644***	1.928***	1.641***	1.925***
lagged	(0.158)	(0.199)	(0.157)	(0.199)	(0.156)	(0.198)
Other debt, lagged	1.125*	1.201**	1.123*	1.200**	1.123*	1.199**
	(0.067)	(0.081)	(0.067)	(0.081)	(0.067)	(0.081)
PrRes Total mortgage	1.048***	1.056***
balance, lagged	(0.009)	(0.011)
PrRes First mortgage			1.045***	1.055***	1.045***	1.055***
balance, lagged			(0.009)	(0.011)	(0.009)	(0.011)
PrRes All secondary			1.086***	1.067**
mortgage balances, lagged			(0.024)	(0.024)
PrRes HELOC balance,					1.069**	1.047
lagged					(0.027)	(0.026)
PrRes Other secondary					1.139**	1.145*
mortgage balances, lagged					(0.050)	(0.061)
SecRes All mortgage	1.002	1.004	1.002	1.004	1.002	1.004
balances, lagged	(0.010)	(0.008)	(0.010)	(0.008)	(0.010)	(0.008)
Annual rent, lagged	1.281*	1.309**	1.280*	1.308**	1.280*	1.309**
	(0.138)	(0.135)	(0.138)	(0.135)	(0.138)	(0.135)
Net cash assets, lagged	0.922***	0.952***	0.922***	0.952***	0.922***	0.952***
	(0.010)	(0.011)	(0.010)	(0.011)	(0.010)	(0.011)
Net investment assets,	0.963***	0.965***	0.963***	0.965***	0.963***	0.965***
lagged	(0.005)	(0.003)	(0.005)	(0.003)	(0.005)	(0.003)
Net other assets, lagged	0.991*	0.992*	0.991*	0.992*	0.991*	0.992*
	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)
House value, lagged	0.988***	0.984***	0.988***	0.984***	0.988***	0.984***
	(0.004)	(0.003)	(0.004)	(0.003)	(0.004)	(0.003)
Secondary residence	1.003	1.002	1.002	1.001	1.003	1.001
house value, lagged	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)
Social security income,	0.962	0.999	0.961	0.999	0.962	1.001
lagged	(0.046)	(0.044)	(0.046)	(0.044)	(0.046)	(0.044)
Earned income, lagged	0.925***	0.968	0.924***	0.968	0.924***	0.968
	(0.018)	(0.021)	(0.018)	(0.021)	(0.018)	(0.021)
Other income, lagged	0.934**	0.969	0.934**	0.969	0.934**	0.970
	(0.021)	(0.024)	(0.021)	(0.024)	(0.021)	(0.024)
Number household	1.138***	1.164***	1.137***	1.164***	1.137***	1.164***
members, lagged	(0.044)	(0.047)	(0.044)	(0.047)	(0.044)	(0.047)
Suburban residence,	0.961	1.033	0.964	1.034	0.964	1.034
lagged	(0.103)	(0.107)	(0.103)	(0.107)	(0.103)	(0.107)
Rural residence, lagged	0.931	0.869	0.933	0.870	0.932	0.870
	(0.097)	(0.088)	(0.098)	(0.088)	(0.098)	(0.088)
9 Census regions	Yes	Yes	Yes	Yes	Yes	Yes
Male	0.795*	0.906	0.795**	0.906	0.794**	0.905
	(0.071)	(0.078)	(0.071)	(0.078)	(0.071)	(0.077)
Age, lagged	0.948***	0.952***	0.948***	0.952***	0.948***	0.952***
	(0.007)	(0.007)	(0.007)	(0.007)	(0.007)	(0.007)
Black	1.192	0.997	1.191	0.997	1.191	0.996
	(0.148)	(0.129)	(0.147)	(0.129)	(0.148)	(0.129)
Other race	0.926	1.047	0.926	1.047	0.927	1.047
	(0.221)	(0.257)	(0.221)	(0.257)	(0.221)	(0.257)
Hispanic	0.937	0.513***	0.932	0.512***	0.932	0.513***
	(0.171)	(0.098)	(0.171)	(0.097)	(0.171)	(0.098)
Not born in United States	1.097	1.005	1.102	1.006	1.104	1.009
	(0.181)	(0.167)	(0.182)	(0.167)	(0.182)	(0.168)
Less than high school	1.184	0.929	1.188	0.929	1.190	0.932
	(0.132)	(0.104)	(0.132)	(0.104)	(0.133)	(0.105)
GED	1.257	1.174	1.259	1.175	1.259	1.176
	(0.225)	(0.210)	(0.225)	(0.210)	(0.225)	(0.210)
Some college	0.999	1.386**	1.000	1.387**	1.002	1.390**
	(0.114)	(0.152)	(0.114)	(0.152)	(0.115)	(0.153)
College or more	0.686*	1.076	0.687*	1.076	0.686*	1.076
	(0.101)	(0.145)	(0.101)	(0.145)	(0.101)	(0.145)
Separated, divorced,	1.233*	1.091	1.234*	1.092	1.234*	1.093
widowed, lagged	(0.126)	(0.111)	(0.126)	(0.111)	(0.126)	(0.111)
Never married, lagged	0.938	1.171	0.936	1.172	0.933	1.172
	(0.238)	(0.327)	(0.238)	(0.328)	(0.238)	(0.327)
Number of living	1.003	0.995	1.004	0.995	1.003	0.995
children, lagged	(0.019)	(0.019)	(0.019)	(0.019)	(0.019)	(0.019)
Number of living	1.395	1.084	1.394	1.083	1.388	1.074
children missing, lagged	(0.379)	(0.294)	(0.379)	(0.294)	(0.378)	(0.293)
Health insurance, lagged	1.109	0.982	1.106	0.981	1.107	0.980
	(0.159)	(0.164)	(0.159)	(0.164)	(0.159)	(0.163)
ADL, lagged	1.288***	1.176**	1.287***	1.176**	1.287***	1.175**
	(0.069)	(0.066)	(0.069)	(0.066)	(0.069)	(0.066)
Self-rated health, lagged	0.760***	0.751***	0.759***	0.751***	0.759***	0.751***
	(0.029)	(0.029)	(0.029)	(0.029)	(0.029)	(0.029)
Self-rated memory,	0.894**	0.892**	0.895**	0.892**	0.895**	0.893**
lagged	(0.037)	(0.038)	(0.037)	(0.038)	(0.038)	(0.038)
Receive help, lagged	0.906	0.919	0.906	0.919	0.906	0.918
	(0.071)	(0.073)	(0.071)	(0.074)	(0.071)	(0.073)
Receive help missing,	1.219	1.186	1.220	1.187	1.221	1.187
lagged	(0.135)	(0.135)	(0.135)	(0.135)	(0.135)	(0.135)
Year = 2008	0.988		0.991		0.991
	(0.106)		(0.107)		(0.107)
Year = 2010	1.154	1.531***	1.154	1.531***	1.153	1.529***
	(0.106)	(0.132)	(0.106)	(0.132)	(0.106)	(0.132)
Year = 2012	0.917	1.357**	0.921	1.359**	0.919	1.355**
	(0.104)	(0.144)	(0.105)	(0.144)	(0.105)	(0.144)
Year = 2014	0.759*	1.366**	0.759*	1.367**	0.759*	1.366**
	(0.086)	(0.145)	(0.086)	(0.145)	(0.086)	(0.145)
Year = 2016	1.142	1.695***	1.143	1.695***	1.142	1.692***
	(0.152)	(0.212)	(0.152)	(0.212)	(0.152)	(0.212)
Constant	52.260***	48.258***	51.814***	47.949***	51.914***	47.650***
	(33.811)	(30.007)	(33.548)	(29.807)	(33.625)	(29.610)
Wald Chi2 (df)	715*** (49)	585*** (48)	716*** (50)	586*** (49)	718*** (51)	587*** (50)
Log likelihood	–5816	–5640	–5815	–5640	–5814	–5033
Observations	12,740	10,311	12,740	10,311	12,740	10,311
Number of individuals	5084	4990	5084	4990	5084	4990

Notes: *** p < .001, ** p < .01, * p < .05.

Robust standard errors in parentheses.

All financial variables are in $10,000 units.

Table A4. Odds ratios of random-effects logit regression of a binary measure of financial stress on credit card debt, mortgage debt at age 62, new mortgage debt since age 62, and other explanatory variables (full specification).

Variables	(1)	(2)	(3)	(4)
	Bill-paying difficulty	Ongoing financial strain	Bill-paying difficulty	Ongoing financial strain
	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)	Odds ratio (SE)
PrRes Total mortgage balance at 62, lagged	1.058***	1.071***	1.067***	1.116***
	(0.012)	(0.014)	(0.015)	(0.018)
PrRes Total mortgage difference at 62 to t – 1	1.056***	1.066***
	(0.012)	(0.014)
PrRes New mortgage debt			1.047**	1.020
(Positive mortgage difference at 62 to t – 1)			(0.017)	(0.015)
PrRes Mortgage debt repayment			1.065***	1.113***
(Negative mortgage difference at 62 to t – 1)			(0.015)	(0.018)
SecRes All mortgage balances at 62, lagged	0.980	0.986	0.980	0.987
	(0.011)	(0.011)	(0.011)	(0.011)
Credit card balance, lagged	1.494***	1.856***	1.486***	1.811***
	(0.169)	(0.240)	(0.168)	(0.229)
Other debt, lagged	1.169*	1.354***	1.167*	1.350***
	(0.088)	(0.116)	(0.088)	(0.114)
Annual rent, lagged	1.443	1.882**	1.447*	1.887**
	(0.270)	(0.395)	(0.270)	(0.394)
Net cash assets, lagged	0.930***	0.963*	0.930***	0.964
	(0.015)	(0.019)	(0.015)	(0.019)
Net investment assets, lagged	0.960***	0.970***	0.960***	0.970***
	(0.007)	(0.004)	(0.007)	(0.004)
Net other assets, lagged	0.989*	0.997	0.989*	0.997
	(0.005)	(0.004)	(0.005)	(0.004)
House value, lagged	0.990*	0.984**	0.990*	0.983***
	(0.005)	(0.005)	(0.005)	(0.005)
Secondary residence house value, lagged	1.002	1.000	1.002	1.001
	(0.010)	(0.010)	(0.010)	(0.009)
Social security income, lagged	0.935	0.932	0.937	0.938
	(0.059)	(0.053)	(0.059)	(0.053)
Earned income, lagged	0.909***	0.968	0.908***	0.963
	(0.021)	(0.024)	(0.021)	(0.025)
Other income, lagged	0.900**	0.928**	0.899**	0.924**
	(0.030)	(0.023)	(0.030)	(0.023)
Number household members, lagged	1.098	1.110*	1.098	1.105
	(0.053)	(0.058)	(0.053)	(0.058)
Suburban residence, lagged	1.125	1.018	1.126	1.024
	(0.165)	(0.149)	(0.165)	(0.150)
Rural residence, lagged	0.974	0.929	0.980	0.950
	(0.139)	(0.129)	(0.140)	(0.131)
9 Census regions	Yes	Yes	Yes	Yes
Male	0.744*	0.884	0.741*	0.868
	(0.090)	(0.104)	(0.090)	(0.102)
Age, lagged	0.947*	0.970	0.949*	0.975
	(0.021)	(0.021)	(0.021)	(0.021)
Black	1.316	1.135	1.318	1.141
	(0.220)	(0.199)	(0.220)	(0.199)
Other race	1.135	1.418	1.136	1.445
	(0.348)	(0.451)	(0.347)	(0.459)
Hispanic	1.015	0.687	1.012	0.678
	(0.235)	(0.167)	(0.235)	(0.166)
Not born in United States	0.957	0.900	0.961	0.921
	(0.214)	(0.209)	(0.215)	(0.213)
Less than high school	1.042	0.794	1.040	0.783
	(0.163)	(0.127)	(0.162)	(0.125)
GED	1.571	1.426	1.570	1.424
	(0.382)	(0.349)	(0.382)	(0.346)
Some college	0.957	1.476**	0.953	1.449*
	(0.148)	(0.222)	(0.147)	(0.217)
College or more	0.578**	1.008	0.575**	0.985
	(0.117)	(0.187)	(0.116)	(0.181)
Separated, divorced, widowed, lagged	1.035	0.989	1.037	0.995
	(0.141)	(0.138)	(0.141)	(0.139)
Never married, lagged	0.768	0.923	0.769	0.924
	(0.239)	(0.339)	(0.239)	(0.338)
Number of living children, lagged	0.980	1.007	0.981	1.008
	(0.026)	(0.028)	(0.026)	(0.028)
Number of living children missing, lagged	1.121	1.376	1.126	1.411
	(0.401)	(0.506)	(0.403)	(0.518)
Health insurance, lagged	1.022	0.876	1.027	0.896
	(0.164)	(0.161)	(0.165)	(0.166)
Activities of daily living (ADL), lagged	1.239**	1.254**	1.237**	1.243*
	(0.097)	(0.109)	(0.097)	(0.109)
Self-rated health, lagged	0.778***	0.740***	0.777***	0.739***
	(0.039)	(0.040)	(0.039)	(0.040)
Self-rated memory, lagged	0.904	0.948	0.904	0.947
	(0.051)	(0.056)	(0.051)	(0.056)
Receive help, lagged	0.922	1.014	0.921	1.015
	(0.096)	(0.110)	(0.096)	(0.110)
Receive help missing, lagged	1.331	1.230	1.331	1.242
	(0.214)	(0.206)	(0.214)	(0.208)
Year = 2008	1.021		1.023
	(0.152)		(0.153)
Year = 2010	1.090	1.544**	1.095	1.583**
	(0.163)	(0.222)	(0.164)	(0.228)
Year = 2012	0.897	1.368	0.899	1.389
	(0.174)	(0.248)	(0.175)	(0.251)
Year = 2014	0.732	1.328	0.730	1.332
	(0.163)	(0.280)	(0.163)	(0.280)
Year = 2016	1.260	1.807*	1.256	1.808*
	(0.320)	(0.437)	(0.319)	(0.437)
Constant	46.895*	9.862	42.188*	7.009
	(73.841)	(14.909)	(66.513)	(10.584)
Wald Chi2 (df)	410*** (50)	341*** (49)	413*** (51)	355*** (50)
Log likelihood	–3194	–3009	–3194	–3000
Observations	6823	5470	6823	5470
Number of respondents	2646	2599	2646	2599

Notes: *** p < .001, ** p < .01, * p < .05.

Robust standard errors in parentheses.

All financial variables are in $10,000 units.