Net Worth Poverty and Adult Health

Abstract

This study broadens the traditional focus on income as the primary measure of economic deprivation by providing the first analysis of wealth deprivation, or net worth poverty (NWP), and adult health. Net worth poverty—having wealth (assets minus debts) less than one-fourth of the federal poverty line—likely exacerbates the negative effects of income poverty (IP). In 2019, one-third of US households were net worth poor, with substantially higher rates among Black (60%) relative to White (25%) households. We estimate longitudinal growth curve (i.e., linear mixed effects) models to test how NWP, IP, and the interaction of the two to predict a diverse set of health measures. We also consider whether NWP resulting from either low assets or high debts is more predictive of health outcomes and test for heterogeneous associations by race. Data come from Panel Study of Income Dynamics on 8,962 individuals ages 25 to 64, observed between 2011 and 2019 (n=26,776). Adjusting for income poverty, net worth poverty, relative to no poverty, was associated with a one-quarter to one-third increase in the likelihood of reporting poor self-rated health, psychological distress, and work limitations. Simultaneously experiencing both NWP and IP was associated with the largest deficits. Both asset-driven (low asset) and debt-driven (high debt) NWP reduced health, but asset-driven NWP had stronger associations (e.g., a 5-percentage point increase of being in poor health, twice that of debt-driven). White, relative to Black, adults exhibited statistically larger associations for psychological distress (4.3 vs 1.1 percentage points) and work limitations (3.7 vs. 1.5 percentage points). White and Black adults who were jointly net worth and income poor exhibited the most disadvantage. Findings underscore how wealth is a critical component of financial deprivation and that wealth deprivation, particularly the lack of assets, merits attention in socioeconomic studies of health inequalities.

Introduction

Economic resources promote health and well-being in numerous ways, and those with insufficient resources are at heightened risk of poor physical and mental health. Income poverty (IP) has been the primary focus of previous research on economic deprivation and health outcomes, and it is abundantly clear that IP increases mortality, chronic and acute diseases, depression and anxiety, and mental health disorders (Adler and Rehkopf 2008, Chetty et al. 2016, Chokshi 2018, Khullar and Chokshi 2018). Although IP is clearly an important driver of health, a growing body of research further identifies various dimensions of wealth—including its component assets and debts—as important, though less studied, determinants of health (Houle and Light 2014, Drentea and Reynolds 2015, Addo 2017, Sun and Houle 2020, Wolfe 2022).

We propose that wealth scarcity, or net worth poverty (NWP), is a critical correlate of health outcomes that complements and exacerbates the effects of income poverty. As a robust body of research demonstrates, wealth contributes to physical and mental health, beyond any linkages between income and health (Maskileyson 2014, Boen and Yang 2016, Makaroun et al. 2017, Ettman, Cohen and Galea 2020). Associations between wealth and health persist across a range of outcomes, including body mass index (BMI) (Hajat et al. 2010, Boen 2016, Cesarini, Lindqvist, Östling and Wallace 2016, Boen, Keister and Graetz 2021); biomarkers of physiological function including systolic blood pressure and C-reactive protein (Boen and Yang 2016, Yang et al. 2020); subjective ratings of health status (Hajat et al. 2010, Boen 2016); and all-cause mortality (Hajat et al. 2010, Pool et al. 2018). Evidence from quasi-experimental designs suggest that the links between wealth and health may be causal (Cesarini et al. 2016, Pool et al. 2018).

NWP indicates that a household’s net worth is less than one fourth of the federal poverty line, adjusted for family size and composition (Brandolini, Magri and Smeeding 2010, Gibson-Davis, Keister and Gennetian 2021). In 2021, a household with two adults and two children would be NWP if their net worth was less than $6,870. Approximately one-third of US households were NWP in 2019; the majority of these households were not IP, suggesting that NWP households represent a largely separate group of households that are conventionally overlooked in assessments of economically fragile populations (Gibson-Davis et al. 2021). Importantly, NWP is more than an indicator of low wealth. Rather, NWP signals that a household’s financial resources put it in a particularly vulnerable, and often persistent, state of deprivation that is likely to lead to both short- and long-term health risks. We propose that NWP complements measures of IP by acknowledging that having a stock of resources provides security that income alone cannot offer. Distinguishing NWP from IP also incorporates evidence that wealth inequality is much more extreme than income inequality and that NWP has been growing in recent decades even as IP declined (Gibson-Davis et al. 2021).

NWP is likely to be associated with health outcomes, net of and jointly with measures of IP, through three interconnected pathways: spending, time use, and stress. First, NWP may shape household spending in ways that affect health. Saved assets allow households to purchase health-promoting goods and services such as healthy food, safe and adequate shelter, leisure and recreation, medical care, and health insurance, even when individuals exit the labor market or wages are suspended. By contrast, NWP may reduce spending on health-related goods and services in ways that negatively affect health (Rose 1999, Lusardi, Schneider and Tufano 2010, Charron-Chénier, Fink and Keister 2017). In times of economic crisis (e.g., the loss of a primary breadwinner’s income, medical emergency, or other financial shock), NWP households may not be able to meet their spending needs, and this may negatively affect health. Second, NWP may pattern time use in ways that shape health (Catalano et al. 2011, Burgard, Ailshire and Kalousova 2013, Mani, Mullainathan, Shafir and Zhao 2013). Higher-wealth households generally have more time than lower-wealth households for activities that promote health such as leisure, socializing, and physical activity. By contrast, lower-wealth households may need to allocate more time to the participating in the paid labor market to cover the cost of basic needs such as rent, utilities, and food. Third, stress and anxiety may link NWP and health. Low wealth can produce financial pressure, anxiety, and strain that diminish health through a variety of biophysiological and psychological processes (Houle and Light 2014, Drentea and Reynolds 2015). In response to stress, the hypothalamic-pituitary-adrenal (HPA) axis and sympathetic nervous system (SNS) secrete hormones to upregulate functioning across bodily systems, affecting immune, cardiovascular, metabolic, and inflammatory functioning (McEwen 2007); this process is often referred to as fight or flight. Chronic or repeated activation of stress response systems—as when people respond to financial shocks or ongoing financial strain—can erode health over time (Cohen et al. 2012). Indeed, there is evidence that the stress resulting from financial challenges, including low wealth, is associated with a variety of physiological pathologies, including increased metabolic risks and elevated inflammation (Boen 2020, Boen et al. 2021), as well as increased psychological and mental health risks (Houle and Light 2014, Sun and Houle 2020)

We also propose that enormous racial differences in NWP may contribute to racial health disparities. Black households in the U.S. have significantly less wealth than White households (Wolff 2018, Gibson-Davis and Hill 2021), reflecting institutionalized racism and structural inequalities that reduce Black wealth accumulation and the intergenerational transfer of assets (Oliver and Shapiro 1995, Massey 2015, Darity Jr and Mullen 2020). As a result, NWP rates for Black households are more than double that of White households, and racial differences in NWP are considerably higher than racial differences in IP (Gibson-Davis et al. 2021). The health implications of these differences are likely wide ranging, affecting both physical and psychological well-being. On the other hand, it is possible that there are heterogeneous associations between NWP and health by race, given evidence that Black households may receive fewer health benefits than White households from increases in either income or wealth (Shuey and Willson 2008, Boen 2016, Boen, Keister and Aronson 2020). Exposure to the many stressors associated with contextual deprivation and exposure to racial discrimination are some of the factors that may weaken the associations between indicators like income and wealth and health among Black individuals (Pearson 2008, Assari 2018, Gaydosh et al. 2018). We explore the association between NWP, IP, and health separately for Black and White respondents in order to assess the degree to which economic deprivation differentially patterns health by race.

Finally, we propose that NWP is related to health through each of the two components of net worth: assets and debts. A household may be NWP because they have few assets (asset-poor NWP) or because their debts are high (debt-poor NWP). To illustrate the difference, consider two families of four, each with wealth levels of $6,000. Both families are NWP. Family A has total assets of $6,000, meaning that regardless of their level of debt, their assets are insufficient to bring them above the net worth poverty line (they are asset poor). Family B has total assets of $20,000 and total debts of $14,000. Family B’s assets place them above the poverty line, but their debts bring them below the poverty line (they are debt poor). The financial situations of these two households are different, and it is likely that this experience plays out in different ways in the health of household members. We anticipate that both types of net worth poverty are likely to be associated with declines in health: low assets have clear limits for spending during an income crisis, and high debt has been linked with declines in health and psychological well-being through stress and the other financial constraints the liabilities entail (Sweet, Nandi, Adam and McDade 2013, Hojman, Miranda and Ruiz-Tagle 2016, Boen et al. 2020). However, we anticipate that asset-poor households are likely to have worse health than debt-poor households because asset-poor households have so few financial resources that a financial shock can quickly lead to a cascade of other financial crises, including those that affect the immediate safety and well-being of the family such as food insecurity, housing eviction, or homelessness. This heightened level of financial precarity has both tangible (e.g., running out of food or being unable to pay the rent) and more covert (e.g., the stress that accompanies these potential problems) links to health. However, those who are debt poor may be able to use their assets to leverage goods for more health.

We assess wealth deprivation thorough dichotomous measures of NWP for both substantive and analytic reasons. Substantively, we seek to broaden the traditional conceptualization of poverty to encompass wealth as well as income. By conceptualizing wealth deprivation in a way that parallels income deprivation (e.g., assessing resources relative to a threshold), we hope that scholars will more readily incorporate NWP in their studies of economic well-being and health. Moreover, by considering both asset and debt poverty, we can offer advantages over measures that consider only debt (such as debt-to-income ratios) and does so in a way that can be readily compared to income poverty. Analytically, the advantage of using NWP is that it provides a standardize definition of wealth deprivation that does not depend on the wealth distribution captured by a particular data set. Wealth amounts captured in the bottom quintile (for example) will vary depending on the data set and will be difficult to compare across populations and samples. Definitions of wealth scarcity based on percentile measures are also sensitive to the portion of the wealth distribution that the data capture.

Despite the potential for NWP to be a driver of health outcomes, this relationship has not been explored in previous research. We fill this gap by examining whether and how NWP is associated with three measures of physical and mental health: self-rated health, psychological distress, and work limitations. We focus on these outcomes because they provide broad coverage of the outcomes that are likely to reflect the non-specific pathways linking social exposures like NWP and health (Aneshensel, Rutter and Lachenbruch 1991) and because they are outcomes have been the focus of previous research on wealth and health (e.g., Boen, Keister, and Aronson 2020). As we describe later, in additional supplemental analyses, we further consider the associations between NWP and several other outcomes, including asthma, high blood pressure, diabetes, heart attack, and stroke. Given striking racial inequities in both NWP and IP (Gibson-Davis et al. 2021) as well as evidence that the links between socioeconomic indicators and health may vary by race (Boen 2016, Boen et al. 2020), we also consider whether the links between NWP and IP and health vary between Black and White respondents. Finally, we study whether there are differences in health outcomes for those whose NWP is asset-driven versus debt-driven to better understand how NWP functions.

We use data from the 2011–2019 waves of the Panel Study of Income Dynamics (PSID), the longest running nationally-representative study of households in the U.S. The PSID is well-suited for this research because it contains sufficient detail to construct NWP and IP measures and to connect them to measures of physical and mental health. We use person-year data and growth curve (i.e., linear mixed effects) models of 26,776 observations of 8,982 individuals. In order to understand how NWP and IP jointly and uniquely relate to health, we classify respondents into one of four mutually exclusive poverty classifications: (1) neither NWP nor IP, (2) NWP but not IP, (3) IP but not NWP, and (4) both NWP and IP. We also categorize respondents into three NWP categories: (1) not NWP, (2) asset-poor NWP, and (3) debt-poor NWP to examine how different manifestations of wealth deprivation may drive health outcomes. We present results for the full sample and then separately for Black and White respondents.

Findings show that NWP does, indeed, shape health risks for the outcomes we study and that the NWP association with health persists beyond the association between IP and health. Findings also show that both asset deprivation and excessive debt are associated with poor health but that the associations between asset-poor NWP and health are generally stronger than the associations between debt-poor NWP and health. Taken together, results from this study contribute new understanding of the links between wealth, poverty, and health with implications for research, policy, and intervention on population health inequality.

Data and Methods

Data and Samples

We use data from the 2011 to 2019 waves of the Panel Study of Income Dynamics (PSID), the longest-running longitudinal survey of individuals and families in the United States. The PSID sample includes individuals who were originally selected as part of a probability sample in 1968, with an oversample of low-income families. It also includes individuals who were born to or adopted by these original sample members and those included in immigrant refreshers that were conducted in 1997 and 2017. The PSID is ideally suited for this analysis as it is the most comprehensive large, longitudinal data set that has sufficient detail on both wealth and health outcomes for a sample of adults to explore the relationships we address above.

Individuals were eligible for our analytic sample if they were designated as sample persons by the PSID, the head/reference person or spouse/partner of the PSID household, between the ages of 25 and 64, and observed at least twice in consecutive waves between 2013 and 2019. We limit our analytic sample to sample persons because these are the individuals for which the PSID has known information they can apply to the calculation of longitudinal weights, which we use in our analysis (see Chang et al. 2021 for discussion of PSID weights). We limit the sample to heads/reference persons and spouses/partners because the PSID only collects health information on these household members. We also limit the sample to individuals ages 25 through 64 to include respondents who are adults and have had an opportunity to start accumulating wealth but who are not yet at retirement age (when they may begin drawing on their wealth for living expenses). Finally, we limit the sample to respondents who were observed at least twice in consecutive waves so that we can use a mixed effects modeling approach and apply lagged measures of wealth and income in our analysis.

Our analytic sample includes 26,776 observations of 8,982 individuals, 71% of whom were heads/reference persons of the household and 29% were spouses/partners. On average, individuals were observed in three out of the four waves of the PSID in our study. Over half of the sample was observed in all four waves, and only 20% of the sample were observed in only one wave of the survey. Our main analytic sample includes respondents who identify as White, Black, Hispanic, and other races. We also estimate separate models for White and Black respondents to explore differences by race in fully interacted models; we do not estimate separate models for Hispanic respondents because there are not enough Hispanic respondents in the PSID to make this possible.

Measures

We use the PSID’s imputed household net worth variable to measure NWP, consistent with previous research on wealth using the PSID (Pfeffer 2018, Pfeffer and Killewald 2018). Our net worth variable is time-varying and includes assets such as vehicles, stocks, real estate, and checking and savings accounts net of a broad range of debt, including medical, legal, student, family loan, and credit card debt. We converted net worth to 2019 dollars using the Consumer Price Index.

To measure IP, we use the PSID’s imputed total family income measure, which includes wages, salaries, social security, and other income from the reference person and spouse/partner, taxable and social security income from other family unit members, and transfer income. Transfer income includes cash transfers from welfare programs, such as Temporary Assistance for Needy Families, unemployment compensation, worker compensation, child support, and financial support from relatives. We also converted income to 2019 dollars using the Consumer Price Index. The measure of IP is time-varying.

We define NWP and IP relative to the 2019 poverty line (US Census Bureau 2021). Following previous work (Haveman and Wolff 2004, Haveman and Wolff 2005, Brandolini et al. 2010), we consider households to be net worth poor if their net worth (total assets minus total debts) is less than 25% of the federal poverty line, adjusted for family composition. In 2019, the poverty line for a household of four was $25,926 (US Census Bureau 2021); a household would be net worth poor if their net worth was less than $6,482. Note that because we use constant dollars, we adjust for over-time variation in wealth, income, and poverty thresholds. Sensitivity analyses (described later) examine the robustness of our results to alternative definitions and thresholds of NWP.

Our first set of analysis uses a joint measure of poverty status, where households are divided into one of four mutually exclusive categories: not income or net worth poor, income poor only, net worth poor only, or both income and net worth poor. The second set of measures is a mutually exclusive, three-category division of net worth poverty status: not poor, asset-poor, or debt-poor. Asset-poor households are those households whose levels of assets are insufficient to bring them above the net worth poverty line (these households may or may not have debts). Debt-poor households refers to households whose assets exceed the net worth poverty threshold, but their debt levels bring them below the threshold. Both sets of poverty measures are time-varying and, to reduce concerns about reverse causality, were lagged by one wave.¹

We focus on three time-varying dependent variables, each of which is coded as a dichotomous variable. Poor self-rated health is coded 1 if respondents indicated that their health was fair or poor. High psychological distress is measured via the Kessler Psychological Distress Scale, where 1 indicates a total score of 11 or higher (Kessler et al. 2002). Treating psychological distress as a continuous measure produced substantially the same results as is reported below. Finally, work limitations are defined by whether a respondent reported that they had a physical or nervous health condition that limited the amount of work they could do. Sample sizes are smaller for psychological distress than for the other two outcomes because the PSID only collected information on psychological distress for the survey respondent, whereas the information on self-rated health and work limitations was collected for both the survey respondent and their spouse/partner.

Covariates included respondent’s sex (male or female), age, and age squared. We also controlled for marital status using a series of dichotomous variables including married [omitted], unmarried, divorced/separated, or widowed. We control for education using four variables: no high school diploma [omitted], high school diploma, some college, and a bachelor’s degree or more. Finally, we include a dichotomous indicator that the household has any children under the age of 18. When results were not stratified by race, we use four categorical measures of race and ethnicity: non-Hispanic White, non-Hispanic Black, Hispanic, and other race/ethnicity. All covariates except sex and race/ethnicity are time-varying. The proportion of missing data on our covariates was low. Only 243 observations fitting our sample selection criteria (<1%) had at least one missing value among our covariates. Because of this low missingness, observations with missing data on covariates were deleted listwise from the study.

Appendix Table 1 displays weighted descriptive statistics for the outcome variables and covariates. Across years, the most commonly reported health problem was poor self-rated health (16%), followed closely by having a work limitation (15%). Relatively few adults reported being in high psychological distress (7%). Roughly two-thirds of the weighted sample was White, with Black and Hispanic adults each constituting about one-sixth of the sample.² The average age was in the mid-40s, with a slight majority of the sample reporting being female. More than half of the sample had at least some post-secondary education, with 38% reporting having at least a Bachelor’s degree. The modal relationship status was married (65%), with similar fractions of the sample reporting being single (18%) or divorced, separated, or widowed (17%). Variation across time in the prevalence of the health outcomes or the demographic composition of the sample was minimal.

Analysis Strategy

We use growth curve (also known as linear mixed effects) models to estimate the longitudinal associations between NWP and IP status and outcome measures. This approach is ideal for the analysis of nested observations of individuals in unbalanced panels. In these models, individuals are observed at multiple time periods (i.e., observations at Level 1 are nested within individuals at Level 2). The main advantage of the mixed-effects method over other hierarchical methods is that it contains both fixed and random effects, allowing between-individual analysis while adjusting for within-individual variation. The model is:

$$Y_{\text{it}}={\text{X}}_{\text{it}}\beta +{\text{Poverty}}_{it}\delta +{\text{Z}}_{it}{\mu }_{it}+{\eta }_t+{\epsilon }_{it},$$ (1)

where X is a matrix of the covariates described earlier; poverty is a matrix capturing poverty status; Z is a matrix for random effects at the individual level; and η is a vector of time-fixed effects. The subscript i corresponds to individuals, and the subscript t corresponds to time. Standard errors are clustered at the family level. For some estimates, we use Eq. (1) to generate predicted values in health outcomes while holding covariates at their means.

Results

Prevalence of Poverty

Our descriptive estimates illustrate the prevalence of net worth and income poverty, their joint distribution, and the incidences of asset and debt poverty (Table 1). Results are presented for the full sample and by race. We present results for 2019; estimates did not vary substantially over the period and estimates for other years are available upon request. All estimates are weighted using individual, longitudinal weights calculated by the PSID.

Table 1:

Poverty Status by Race, 2019

	Total	White	Black
Net worth poor (NWP)	34.2	25.7	60.2
Income poor (IP)	9.5	6.2	20.0
Joint Poverty Status
No Poverty	63.5	72.6	36.4
IP only	2.3	1.8	3.4
NWP only	27.0	21.2	43.6
IP and NWP	7.2	4.4	16.6
Asset and debt poverty
No NWP	65.7	74.3	39.8
Asset NWP	27.7	18.5	54.0
Debt NWP	6.5	7.2	6.2
Individuals	7,138	3,031	2,721
Percentage	100.0	42.5	38.1

Notes: Numbers represent column percentages. Descriptive statistics are weighted using individual, longitudinal weights from the PSID.

In 2019, roughly one in three adults was net worth poor, a rate that was three and half times as high as that of income poverty.³ A high prevalence of net worth poverty was also evident when net worth poverty was considered jointly with income poverty: 27% of the weighted sample experienced net worth poverty alone, with an additional 7% experiencing it in conjunction with income poverty. Very few (2%) of respondents experienced income poverty by itself. Conditional on being poor, then, net worth poverty was the dominant form of poverty, as nearly 94% of those who experienced poverty were net worth poor, either singly or with income poverty. Asset poverty (28%) was more common than debt poverty (7%). Among those who were net worth poor, 81% experienced asset poverty, whereas the other 19% were wealth poor because of their debts.

Black adults were more likely than white adults to experience poverty. Only one-third of Black respondents avoided poverty, whereas three-fourths of White respondents were not poor. NWP rates were exceptionally high for Black adults, as 60% experienced net worth poverty (relative to 20% for income poverty). Rates of NWP were substantially lower for white adults; still, just over one quarter of White adults were net worth poor. For both racial groups, net worth poverty was the modal form of poverty, with very few households experiencing income poverty by itself. For both groups, asset poverty was more common than debt poverty. Rates of asset poverty were particularly high among Black adults, with the majority of the sample being asset poor.

Association Between NWP, IP, and Health

Our first set of regression results provides support for our expectation that both NWP and IP are associated with worse health. Figure 1 illustrates the association between lagged joint poverty status and the three outcomes, with results presented as predicted probabilities.⁴ Coefficients underlying probabilities are presented in Appendix Table 3. As expected, poverty in any form was associated with worse outcomes for all three outcomes, with income poverty and income poverty experienced in conjunction with net worth poverty associated with the highest (or worst) predicted values.

Figure 1:

Estimated Margins from Regressions of Poverty Status on Health Outcomes

Notes: Black bars represent 95% confidence intervals. Covariates in regression models include race-ethnicity, marital status, age, age squared, sex, educational attainment, presence of children in household, and year of survey. Estimates weighted. Sample size: self-rate health n=26,765; psychological distress n=18,718, and work limitation n=26,754 observations.

Net worth poverty was associated with worse health when compared to those not in poverty. After adjusting for demographic covariates and income poverty, the predicted value of having high psychological distress for those who were net worth poor was 0.08, relative to 0.06 for not being poor (p<0.001 for difference). The predicted values for poor self-rated health for net worth poor and not poor individuals was, respectively, 0.19 and 0.15; for health condition limiting work, predicted probabilities were 0.15 and 0.12, respectively, (p<0.001 for difference). These estimates indicate that, relative to those not in poverty, net worth poverty was associated a one-quarter to one-third increase in the likelihood of reporting these outcomes. Predicted margins associated with net worth poverty were smaller than, but statistically indistinguishable from, those of income poverty for poor health and psychological distress. Income poor adults had a slightly higher risk of work limitations that was statistically different from that for the NWP group.

The positive association between NWP and health was particularly pronounced when it was experienced in conjunction with IP. Those who experienced both types of poverty had the worst outcomes: they had the highest predicted values for poor health (0.23), psychological distress (0.11), and work limitations (0.19). Based on these estimates, experiencing both types of poverty, relative to not being poor, was associated with a 50 to 83% increase in the likelihood of reporting these health problems. For all three outcomes, these predicted values were significantly higher than those observed for those not in poverty or for those experiencing net worth poverty by itself. These values did not differ at conventional levels of statistical significance from those associated with income poverty.

These results suggest that the experience of net worth poverty likely has deleterious associations with health, particularly if it is experienced jointly with income poverty. Net worth poverty by itself was negatively associated with our outcomes. Relative associations for NWP were smaller than the negative associations of income poverty, but this difference was only statistically significant for work limitations.

Net Worth Poverty and Health: Race Differences

Our next set of results document race differences in the associations between the joint poverty status measure and our three health outcomes (Table 2). For both Black and White adults, results suggest that net worth poverty, experienced alone or in conjunction with income poverty, was associated with worse outcomes. Consistent with expectations, however, coefficient estimates were generally larger for White than for Black adults.

Table 2.

Regressions on Health Outcomes by Race and Poverty Status

Panel A: White respondents
	Poor Health	Psychological Distress	Work Limitation
Not poor	-	-	-
IPa only	0.081*** (0.024)	0.038 (0.028)	0.116*** (0.029)
NWPb only	0.040*** (0.007)	0.028*** (0.007)	0.038*** (0.007)
NWP and IP	0.084*** (0.019)	0.077*** (0.019)	0.103*** (0.019)
Observations	12,420	7,886	12,416
Panel B: Black respondents
	Poor Health	Psychological Distress	Work Limitation
Not poor	-	-	-
IP only	0.040 (0.022)	0.033 (0.018)	0.064** (0.022)
NWP o nly	0.037*** (0.009)	0.0 10 (0.007)	0.016* (0.008)
NWP and IP	0.073*** (0.015)	0.040*** (0.011)	0.061*** (0.012)
Observations	10,242	8,278	10,238

^aIP=Income poverty

^bNWP=net worth poverty

Notes: Covariates in regression models include marital status, age, age squared, sex, educational attainment, presence of children in household, and year of survey. Estimates weighted. Standard errors are clustered by family and included in parentheses.

^*p<0.05

^**p<0.01

^***p<0.001

White adults who were net worth but not income poor experienced small but statistically significant increases in reporting of poor heath (4.0 percentage points), experiencing psychological distress (2.8 percentage points), and having a work limitation (3.8 percentage points; p<0.001 for all three) relative to those who were not poor. Experiencing both income and net worth poverty was associated with an 8.4 percentage point increase in poor health, a 7.7 percentage point increase in psychological distress, and 10.3 percentage point increase in work limitation (p<0.001 for all). Income poverty was associated with increased risks of poor health (8.1 percentage points) and having a work limitation (11.6 percentage points, p<0.001 for both). Wald tests indicated that the coefficients for joint poverty status, for all three outcomes, was larger and statistically different from the same coefficients for net worth poverty status by itself. However, these coefficients were not significantly different from the coefficients for income poverty alone.

Results for net worth poverty were more muted for Black respondents. Net worth poverty by itself was associated with a statistically significant increase in the likelihood of poor health (3.7 percentage points, p<0.001) and work limitation (1.6 percentage points, p<0.05), but no statistically significant association with psychological distress. The coefficients for net worth poverty by itself for psychological distress and work limitation, based on statistical tests, were significantly smaller than the coefficients for the White subsample. Like the White subsample, however, experiencing both types of poverty were associated with especially adverse consequences for the Black subsample: Black adults who were both net worth poor and income poor, relative to those who were not poor, reported significant increases in all outcomes, ranging from a 4 percentage point increase in distress to a 7.3 percentage point increase in poor health (p<0.001 for both). Among Black respondents, income poverty alone was associated with increased risk of having a work limitation (6.4 percentage points, p<0.05), but not the other outcomes.

Asset and Debt Poverty: Descriptive Differences

Our next set of results explore potential differences in outcomes by the type of net worth poverty experienced – asset poverty versus debt poverty. We first present descriptive associations between patterns of home and asset and debt ownership (Table 3) and then present predicted scores for regression models (Figure 2; coefficients for models underlying Figure 2 are presented in Appendix Table 4). For asset and debt ownership, we describe the share of households having any asset or debt and then present results for two common types of assets (home equity and checking/savings accounts) and three types of debt (credit card, medical, and student). Patterns of debt and asset ownership are only presented for 2019, but results for other years were similar.

Table 3.

Asset and Debt Portfolios by Net Worth Poverty Status, 2019

	Not NWP		Asset Poor NWP		Debt Poor NWP
	Owned (%)	Median ($)	Owned (%)	Median ($)	Owned (%)	Median ($)
Home equity	81.9	100,000	4.2	0	48.7	29,000
Assets	100.0	136,000	51.3	800	100.0	20,000
Checking/Savings	89.4	10,000	50.1	800	95.4	6,000
Debts	54.0	14,000	44.0	12,000	100.0	62,000
Credit Card	37.1	5,000	23.6	3,050	64.2	6,000
Medical	6.7	3,000	10.6	5,000	13.2	5,000
Student Loan	20.5	20,000	23.8	20,000	84.9	59,000
Individuals	3,909		2,668		561
%	54.8		37.4		7.9

Notes: Medians are calculated only among individuals who own that type of asset or debt.

Figure 2:

Estimated Margins from Regressions of Poverty Status on Health Outcomes

Notes: Black bars represent 95% confidence intervals. Covariates in regression models include race-ethnicity, marital status, age, age squared, sex, educational attainment, presence of children in household, and year of survey. Estimates weighted. Sample size: self-rate health n=26,765; psychological distress n=18,718, and work limitation n=26,754 observations.

Patterns of asset and debt ownership suggest that asset-poor verses debt-poor NWP households represent qualitatively different economic experiences (Table 3). Asset-poor NWP adults owned very few assets (only 51% report owning any asset), and the median conditional value of assets was very low ($800). Homeownership was rare among asset-poor adults (4%). In fact, the median value of home equity for these adults was zero, because many owed more on their home than its value. Slightly more than half of asset-poor adults reported a checking or savings account, with median amounts of $800. However, asset-poor households mostly avoided debt; more than one-half reported no debt, and among those who did, median debt levels ($12,000) were lower than the same estimates for households who are not net worth poor ($14,000). In contrast, 100% of debt-poor NWP adults reported owning at least one asset, with the median value of their assets being 25 times higher than that of the asset-poor group. Debt-poor adults, relative to asset-poor adults, were more likely to own homes and checking and savings accounts and had more money in each asset group conditional on owning that asset.

However, the debt-poor group was burdened by high levels and amounts of debt. Relative to either the asset-poor or not NWP groups, the debt-poor group was more likely to have debts and owe more money on those debts. Relative to the asset-poor, debt-poor adults were nearly three times as likely to owe on a credit card and nearly four times more likely to have a student loan. Owing on medical debt was roughly equivalent. Median amounts of debt were nearly six times higher ($62,000 vs $12,000) in the debt-poor group relative to the asset-poor group. The debt-poor was particularly burdened by student loans; 85% reported having a student loan, with median loan amounts of $59,000.

Asset and Debt Poverty: Associations with Health

Predicted values from regression models (Figure 2) suggest that the associations between net worth poverty and health outcomes may have been more pronounced for asset poverty compared to debt poverty. Asset poverty, but not debt poverty, was also associated with statistically significantly higher probabilities of experiencing high psychological distress. Though the estimates for asset poverty for all three outcomes were substantively larger than the estimates for debt poverty, the predicted values for these poverty statuses only differed from each other at conventional levels of statistical significance for poor self-rated health and psychological distress. Estimates of the predicted values across the two poverty status measures for work conditions were statistically equivalent. Those in either asset or debt poverty, relative to those not in net worth poverty, had statistically significantly higher predicted values on the poor health and work limitations measures. Those in debt poverty did not have statistically significantly higher levels of psychological distress than those not in net worth poverty.

Variation in health status by net worth poverty status was more pronounced for White relative to Black adults, as seen in Table 4 (estimates only presented for poverty status; for estimates for all covariates, see Appendix Table 5). Asset poverty among White adults was associated with a 5-percentage point increase in poor health, a 4.3 percentage point increase in high distress, and a 3.7 percentage point increase in work limitations (p<0.001 for all) relative to no poverty. Debt poverty was associated with significant increases in both poor health (2.0 percentage points) and work limitations (3.3 percentage points) for White adults. Debt poverty was not significantly predictive of psychological distress (with an estimate that was close to zero), and the debt poverty estimate for poor health was statistically significantly smaller than the asset poverty estimate (p<0.01). For Black adults, asset poverty was associated only with increased likelihood of reporting being in poor health (0.040, p<0.001). Estimates for the other two outcomes were small (0.015 or less) and did not meet conventional levels of statistical significance. The debt poverty estimates did not approach conventional levels of statistical significance for Black adults. For work limitation, the debt poverty estimate was smaller and statistically significantly different than that found for White adults.

Table 4.

Regressions on Health Outcomes by Race and Net Worth Poverty Status

Panel A: White respondents
	Poor Health	Psychological Distress	Work Limitation
Not net worth poor	-	-	-
Asset poor NWPa	0.049*** (0.009)	0.043*** (0.009)	0.037*** (0.009)
Debt poor NWP	0.020* (0.009)	−0.000 (0.008)	0.033*** (0.009)
Observations	12420	7886	12416
Panel B: Black respondents
	Poor Health	Psychological Distress	Work Limitation
Not net worth poor	-	-	-
Asset poor NWP	0.040*** (0.010)	0.011 (0.007)	0.015 (0.008)
Debt poor NWP	0.018 (0.014)	−0.003 (0.009)	0.005 (0.011)
Observations	10242	8278	10238

^aNWP=net worth poverty

Notes: Covariates in regression models include marital status, age, age squared, sex, educational attainment, presence of children in household, and year of survey. Estimates are weighted. Standard errors are clustered by family and included in parentheses.

^*p<0.05

^**p<0.01

^***p<0.001

Supplementary Analysis

We have concentrated on three markers of health (self-rated health, psychological distress, and work limitations) because they are broad in scope and, unlike other disease outcomes, none of them is contingent on provider diagnosis or health care access. While measures of disease status may be most useful at older ages, markers of self-rated health, work limitations, and psychological distress can be used in estimating trajectories of health across early-, mid-, and later adulthood (Deaton and Paxson 1998). Further, diagnosis-based outcomes may be subject to misclassification error, whereby individuals who do not yet have an official diagnosis or who are otherwise unable to get a diagnosis are classified as “well” despite being in poor health (Aneshensel et al. 1991). This misclassification of “unhealthy” individuals as “healthy” could result in underestimating the roles of NWP and IP in shaping health risks. Given that NWP and/or IP likely curtails access to health care, we worried about the potential bias that relying on diagnosable conditions could introduce.

Nevertheless, the PSID does collect information on a number of other health markers, and in the interest of stimulating further research into how NWP may affect diagnosable health conditions, we conducted exploratory models on five other outcomes (all measured dichotomously): asthma, high blood pressure, diabetes, heart attack, and stroke. Each outcome was regressed as a function of either joint poverty status (Panel A) or net worth poverty status (Panel B), using the same set of covariates as was used previously (results for covariates not presented, but are available upon request). Across our non-elderly sample, prevalences for these conditions ranged from 2% for heart attack and stroke, 10 to 12% for asthma and diabetes, and 25% for high blood pressure.

We note that the mechanisms linking these outcomes may differ from our primary outcomes. For example, diabetes, depending on whether it is Type 1 or Type II, may be a function of social, behavioral, environmental, or genetic factors, or some combination of these influences. These models, therefore, should be interpreted as providing preliminary evidence as to how NWP and IP shape these conditions, with additional research needed.

Results for the joint poverty status models (Table 5; Panel A) were broadly consistent with the findings presented above. Both NWP and IP were associated with an increased likelihood of having asthma, diabetes, high blood pressure, heart attack, or stroke. Three of NWP estimates (asthma, diabetes, and stroke) were statistically significant. Also consistent with earlier findings, those who were both net worth and income poor may be particular risk. Estimates for all outcomes except diabetes were larger in magnitude than the estimates for either net worth or income poverty, with effects ranging from 1.5 percentage points for stroke (p<..001) to 2.2 percentage points for asthma (p<.01). As for net worth poverty status (Table 5; Panel B), models generally accord with previous findings, insofar as asset poverty, relative to debt poverty, had more robust associations with outcomes. Estimates for asset poverty were always in the expected direction (e.g., increased likelihood) and were statistically significant for asthma (b=0.013, p<.01) and stroke (b=0.06, p<.01). Estimates for debt poverty varied in direction (e.g., a 0 estimate for diabetes, a negative finding for debt poverty) and did not approach statistical significance. The net worth poverty models, though, were likely underpowered, given the relative lack of variance in the outcomes and the relatively small number of people who experienced debt poverty.

Table 5.

Regressions on Other Health Conditions, by Poverty Status

Panel A: Poverty status
	Asthma	High Blood Pressure	Diabetes	Heart Attack	Stroke
Not poor	-	-
IPa only	0.010 (0.011)	0.016 (0.013)	0.013 (0.009)	0.008 (0.006)	0.012 (0.007)
NWPb only	0.009* (0.004)	0.009 (0.005)	0.007* (0.003)	0.003 (0.002)	0.004* (0.002)
NWP and IP	0.022** (0.007)	0.019* (0.009)	0.007 (0.005)	0.009* (0.004)	0.015*** (0.004)
Observations	26,766	26,756	26,763	26,762	26,763
Panel B: Wealth poverty status
	Asthma	High Blood Pressure	Diabetes	Heart Attack	Stroke
Not poor	-	-	-	-	-
Asset poverty	0.013** (0.004)	0.006 (0.006)	0.005 (0.004)	0.003 (0.002)	0.006** (0.002)
Debt poverty	0.000 (0.005)	0.012 (0.007)	0.008 (0.004)	0.002 (0.002)	−0.001 (0.002)
Observations	26,766	26,756	26,763	26,762	26,763

^aIP=Income poverty

^bNWP=net worth poverty

Notes: Covariates in regression models include marital status, age, age squared, sex, educational attainment, presence of children in household, and year of survey. Estimates weighted. Standard errors are clustered by family, and included in parentheses. Model in Panel B includes an additional control for income poverty.

^*p<0.05

^**p<0.01

^***p<0.001

Robustness Checks

We explored the robustness of our estimated associations between NWP and the three health outcomes using a variety of different specifications for NWP. In alternative specifications of NWP, we assigned NWP to individuals if the levels of various subcategories of their complete set of assets did not meet our NWP threshold. These subcategories included using only liquid wealth (value in checking and savings accounts, stocks, annuities, and other assets), only nonliquid wealth (value of farms, businesses, home equity, and other real estate), only home equity, and all assets aside from home equity. The negative association between health and wealth poverty remained substantively similar and highly statistically significant for all three health outcomes, indicating that the association between NWP and poor health is not driven by any particular subset of assets in particular.

We also tested the robustness of our NWP threshold by assigning individuals to NWP if their levels of wealth were below 50% or below 200% of the poverty line. Regardless of whether the threshold was set at 50% or 200% of our conventional NWP threshold, we found that the relationship between NWP and poor health outcomes remained highly significant and substantively similar to our main models reported in Figure 1 and Appendix Table 3.

Conclusion

This study expanded understanding of the link between economic deprivation and health by studying how wealth deprivation, measured as net worth poverty, relates to three diverse health measures. Building on a well-established literature that demonstrates the salience of wealth for health (Boen and Yang 2016, Addo 2017, Sun and Houle 2020), we found that respondents who experienced NWP—either alone or in conjunction with IP—had an increased likelihood of reporting being in poor health, having high psychological distress, and or having a health condition that limited work. Estimated associations were modest but substantially larger when adults experienced NWP and IP jointly. Those who experienced both NWP and IP had the worst predicted levels of health across the three outcomes, being 50 to 80% more likely than those not in poverty to report having poor health, high psychological distress, or a work-limiting condition. Importantly, levels of net worth poverty in the U.S. are high; our descriptive analyses showed that roughly one in three U.S. adults aged 25–64 were NWP in 2019. Our results highlight that these Americans are at increased risks of adverse health outcomes.

Consistent with our theoretical and empirical expectations, the associations between NWP and health were statistically independent from those between IP and the outcomes. Wealth and income are likely complements, but not substitutes, for one another and may operate along distinct channels to affect health. Insofar as NWP represents the sufficiency of a household’s cumulative stock of resources, it likely shapes health through a variety of pathways—including by shaping stress levels, time use, and consumption behaviors—in ways that may not be captured by assessing income flows alone. At the same time, inadequacy in either wealth or income is cause for concern, as those who experience insufficiency in either wealth or income are at increased health risk. Indeed, results suggest that even though relatively few individuals experienced income poverty alone without net worth poverty, those that did were still at risk for adverse outcomes. Given that both IP and NWP are independent risk factors for health, it is perhaps not surprising that those who experienced both were at highest risk of compromised health.

Another contribution of our study was to provide the first analyses of the experiences of asset poverty versus debt poverty. Asset poverty was four times as common as was debt poverty, with less than 7% of respondents experiencing debt poverty. And as descriptive statistics indicated, the asset poor had a qualitatively different economic profile than the debt poor, with extremely low levels of asset ($800 in median assets, relative to $20,000 for the debt poor). Regression results provide suggestive evidence that these differences translated into different associations with health. Though our estimated associations between asset poverty and debt poverty and the outcomes did not always differ from each other at conventional levels of statistical significance, asset poverty was associated with the qualitatively worst outcomes, with estimates that were substantively larger than those of debt poverty. Results should not be interpreted as suggesting that debt poverty was not predictive of health; consistent with past studies, we found that debt poverty was associated with worse health relative to not being poor. But the comparison group mattered, insofar as those who were asset poor were particularly vulnerable.

Results by race suggest that, as expected, Black respondents had higher rates of NWP than did White respondents, with the majority of Black adults (60%) being net worth poor. These high levels of NWP result from historical legacies and contemporary manifestations of racism, discrimination, and exclusion (Darity and Mullen, 2020; Oliver and Shapiro, 1995). Still, these high rates of net worth poverty did not necessarily translate into adverse health outcomes for Black adults in the same ways as for White adults. We found that the associations between NWP and outcomes were generally larger for White, rather than Black, adults. Similarly, the associations between asset and debt poverty were also more robust for White versus Black respondents. Though we cannot explain why these racial differences exist, we suspect, as others have found (Pearson 2008, Shuey and Willson 2008, Boen 2016, Assari 2018), that differences in community contexts, experiences of racism discrimination, and other systematic inequalities contribute to the poor health of Black individuals and weaken the associations between NWP and asset poverty and health.

Limitations to our study should be noted. Results cannot establish causality, and unmeasured or unobserved characteristics could bias results. We lagged our poverty measures in part to address these concerns. Adding in additional years of data that cover a event like the Great Recession cannot fully address this problem, as unobserved or unmeasured characteristics could still bias estimates between wealth and health over the Great Recession. Other studies that have used pre- and post-Recession waves (e.g., Boen, Keister, and Aronson 2020) found that wealth-health linkages were robust across periods. Future work should consider additional ways to account for potential endogeneity and reverse causality. We are also unable to analyze the mechanisms that underlie our associations. Our proposed mechanisms are consistent with previous literature on wealth and health, but a full exploration of how and why net worth poverty predicts health is beyond the scope of this study. Net worth poverty, as a measure of wealth deprivation, also poses challenges. A three-month period to define net worth poverty is standard in the literature (Haveman and Wolff 2004, Haveman and Wolff 2005), but this is an arbitrary time frame (though analyses suggest that that a three-month time frame may correspond to the amount of savings displaced when a household suffers an income shock; Brandolini et al. 2010). Moreover, our preliminary explorations suggested that other time frames did not significantly alter the results. Nonetheless, future research might usefully explore alternative definitions of net worth poverty in more depth then we are able to do in this paper. Finally, net worth poverty, like income poverty, is also an absolute measure, and a relative measure may be preferred (Brady 2003).

Despite its limitations, our results highlight the importance of capturing wealth deprivation in future studies of health, morbidity, and mortality. Use of a standard definition of wealth deprivation that can be used across data sets will encourage further research into net worth poverty and encourage scholars to specifically consider wealth deprivation. As results have shown here, net worth and income poverty are both independently predictive of adverse health, with heterogeneous associations for different subgroups. Insofar as wealth and income deprivation may compromise health—and those that experience scarcities of both wealth and income may be at particular risk—a fuller understanding of net worth poverty’s effects will advance knowledge on how economic scarcity impedes health.

Our results speak to the importance of policy and intervention efforts to redress wealth poverty’s pernicious impacts on health. For one, policies aimed at building assets and reducing debts—particularly among structurally disadvantaged and racially subjugated households that have largely been excluded from accessing the instruments of wealth accumulation—would likely have substantial health-promoting benefits on a population level. Further, social welfare programs—including those that expand access to health care, food, and housing—may allow families to build wealth thus weakening the association between poverty and health. Extending eligibility for these benefits to those who are net worth poor (in addition to those who are income poor) could expand the potential health benefits of these programs.

Highlights.

• Net worth poverty-wealth less than 1/4 of poverty line-negatively predicts health.

• Net worth poverty related to poor health, psychological distress, and work limits.

• Being both net worth and income poor put Black and White adults at the most risk.

• Lack of assets may be more consequential for poor health than high debt.