Twenty studies in the literature were reviewed to synthesize available information on the assets and liabilities of low-income households. Exhibit 1 lists the major studies used in this report and summarizes their key findings. It also summarizes the data sources and methods used to generate the findings. Both the data sources and methods used in the literature are discussed in detail in the Appendix.
|Method||Outcomes Analyzed||Key Explanatory Variables||Findings||Authors Principal Conclusions|
|Aizcorbe, Kennickell, & Moore (2003)||1998, 2001 SCF.||U.S. families ("primary economic units" that include families and other persons in the household).||Descriptive.||Net worth, assets, liabilities, credit card balances, & debt to income ratio.||Percentile of income, age, education, race/ethnicity, work status, region, & home ownership.||Assets and net worth skewed toward the wealthy. Growth in assets and net worth flat during early 1990s, accelerated during late 1990s.||(1) Median and mean net worth of families grew substantially between 1998 and 2001. (2) Ownership of homes and financial assets grew. (3) Debt holding and debt levels increased.|
|Badu, Daniels, & Salandro (1999)||1992 SCF.||Families with a black or white head.||(1) Instrumental variables regression the model measures the difference in net worth between blacks and whites. (2) Canonical correlation analysis measures the differences in asset and liability holdings of blacks versus whites.||Net worth, financial assets, portfolio composition, & use of credit.||Control variables capture income, age, education, employment, race, marital status, children health, and pension.||(1) White households have significantly greater net worth and financial assets than black households. (2) The authors find no evidence that the net worth of black households is constrained by barriers to obtaining credit. (3) There is evidence that blacks are more risk averse in their asset choices, and that they pay higher interest rates.||(1) Both blacks and whites rely a great deal on credit cards. (2) According to the canonical correlation analysis, blacks do not on average have any assets that are independent of liabilities. (3) Whites are more risk tolerant than blacks. (4) On average both whites and blacks' principal asset is their vehicle.|
|Belsky & Calder (2004)||2001 SCF.||U.S. families.||Descriptive.||Financial assets, non-financial tangible assets, debt.||Age, race, income.||(1) Median net wealth of the lowest income quintile is less than one-quarter of the median wealth of the second income quintile. (2) The most commonly held financial asset for individuals in the bottom income quintile is the transaction account. (3) The most commonly held non-financial asset for individuals in the bottom income quintile is the automobile.||(1) Minorities in the lowest income quintile have much lower rates of asset ownership than whites in the same income quintile. (2) A large share of families in the lowest income quintile are operating in a cash economy, preventing them from accessing mainstream, long-term credit. (3) Access to debt is similar for whites and nonwhites in the lowest income quintile.|
|Browning & Lusardi (1996)||Literature review includes SCF, CES, PSID, HRS, AHEAD, SIPP, NLS, RHS.||U.S. Households||Descriptive.||Household saving.||Various demographic variables in many different surveys; models used to describe household saving behavior.||(1) A "standard optimizing framework," integrating the CEQ model and standard additive model for consumption decisions can provide an adequate framework for examining household saving. (2) Though current datasets provide an accurate portrait of who saves over time, they are less effective in explaining why people save.||(1) More and "better" data are needed to examine the question of why households save, including more information on health status, perception of mortality risk, the situation of children, and liquidity constraints. (2) Simulation models that explore both observed and difficult-to-measure heterogeneity variables across populations could shed more light on the savings decision. These models could also be linked to the standard model, which could integrate a larger range of life-cycle decisions.|
|Bucks, Kennickell, & Moore (2006)||1995, 1998, 2001, 2004 SCF.||U.S. families ("primary economic units" that include families and other persons in the household).||Descriptive.||Net worth, assets, liabilities, credit card balances, & debt to income ratio.||Percentile of income, age, education, race/ethnicity, work status, region, & home ownership.||Assets and net worth skewed toward the wealthy. Financial assets comprised a smaller proportion of portfolios than in 2001. Despite lower interest rates in 2004 than in 2001, there were moderate increases in debt burden.||(1) Despite small changes in income between 2001 and 2004, there were some increases in mean (6.3%) and median (1.5%) net worth that pale in comparison to the increases between 1998 and 2001. (2) Real estate values increased sharply between 2001 and 2004. (3) The rise in debt use is attributable to the increased use of real estate debt.|
|Caner & Wolff (2004)||1984-99 PSID.||U.S. families.||Descriptive.||Asset poverty rates.||Race, age, education, tenure, family type.||(1) Nonwhites are more than twice as likely as whites to be asset poor, though the nonwhite asset poverty rate declined from 1984-1999. (2) From 1994-99, asset poverty rates increased for most age groups. (3) Asset poverty rates decrease with higher education levels. (4) Changes in race/ethnicity and family type had a negligible effect on the overall poverty rate.||Though the traditional income poverty measure decreased over the 1984-99 period that the authors examined, the asset poverty rate barely changed and the severity of poverty increased.|
|Carasso, Bell, Olsen, & Steuerle (2005)||CPS, SCF, various years.||U.S. households and families.||Descriptive.||Homeownership rates, homeownership subsidies and tax expenditures.||Income, education, race, children in family, marital status.||(1) Homeownership rates increased from 1990-2003, but this is mostly just catch up from losses in the 80s. (2) Homeownership rates for minorities and less educated individuals are significantly lower than for whites and more educated groups. (3) Federal housing subsidies and tax expenditures are unevenly distributed in a "U" shape.||Future policies need to smooth the "U"-shaped distribution by changing ownership incentives for low- and middle-class families.|
|Domowitz & Sartain (1999)||Proprietary data of 827 bankruptcy filers and the 1983 SCF.||U.S. bankruptcy filers and matched non-filers.||Nested logit model relating the probability of filing for bankruptcy to household demographic and financial characteristics.||Bankruptcy choice (yes/no).||Medical debts above 2% of annual income, homeownership, marital status, debts/assets, credit card balance/income, secured and unsecured debts/income.||(1) The effect of marriage on bankruptcy is statistically negligible (p. 410). (2) Debtors without homes are almost seven times more likely to file than the average homeowner (p. 413). (3) An increase in credit card debt to the level of an average chapter 7 debtor is predicted to cause a 624% increase in the probability of filing for bankruptcy (p. 414).||(1) Substantial medical debt is found to be the most important factor in assessing the impact of household conditions (p. 419). (2) On the margin, the largest single contribution to bankruptcy is credit card debt (p. 419). (3) Homeownership discourages bankruptcy by a substantial amount (p. 410).|
|Fay, Hurst, & White (2002)||PSID.||U.S. households.||Probit regressions- measuring the determinants of filing for bankruptcy.||Bankruptcy.||Financial benefits of applying for bankruptcy, debts, nonexempt assets, local bankruptcy rate, income, education, family size.||(1) Far fewer households file for bankruptcy than the number that would actually benefit from filing (p. 712). (2) Debts, when financial benefits from filing for bankruptcy are greater than zero, are a more important contributors for the filing decision than foregone nonexempt assets (p. 716). (3) Education, income, and homeownership (marginally) have the predicted effects on the decision to file for bankruptcy (p. 713). (4) Surprisingly, business owners are less likely to file for bankruptcy (p. 713). (5) Finally, individuals living in districts where bankruptcy is more prevalent are more likely to file for bankruptcy (p. 706).||(1) The authors find support for the notion that households are more likely to file for bankruptcy when their financial benefit from filing is higher (p. 706). (2) Their model predicts that a $1,000 increase in the benefit from filing for bankruptcy results in a 7% increase in the number of bankruptcy filings (p. 715). (3) They also conclude that the argument that households file for bankruptcy in times of extreme duress is not supported by the data (p. 706).|
|Gross & Souleles (2002)||Panel dataset of credit card accounts from several credit card issuers. Representative of all open accounts in 1995. Contains other financial information and spans from 1995-1997.||U.S. credit card holders in 1995.||Duration models.||Default and bankruptcy.||Credit score, employment, credit use, age, lack of health insurance, & regional home prices.||Risk controls (credit scores) were highly significant in predicting bankruptcy and delinquency. Even controlling for credit scores, accounts with larger balances and purchase or smaller payments were also more likely to default. Unemployment and lack of health insurance also increased default, but these variables only explain a small part of the change in bankruptcy and delinquency rates over the sample period (p. 345).||Ceteris paribus, a credit card holder in 1997 was 1% more likely to declare bankruptcy and 3% more likely to be delinquent on payments than cardholders with an identical profile in 1995. This magnitude is almost as large as if the whole population of cardholders became one standard deviation riskier in terms of credit scores. These results are consistent with a demand effect. That is, it is likely that the costs of defaulting or filing for bankruptcy decreased between 1995 and 1997 (p. 322).|
|Himmelstein et al. (2005)||Dataset of 1,771 bankruptcy filers in five federal courts in 2001.||U.S. bankruptcy filers.||Descriptive.||Bankruptcy.||Medical reasons for filing bankruptcy, such as illness or injury, uncovered medical bills, lapse of insurance coverage, or death of a family member.||(1) The average debtor tends to be middle-aged, working or middle class, with children and at least some college education. (2) Nearly half the debtors surveyed had a "major medical bankruptcy." (3) A lapse in health insurance is a strong predictor of a medical bankruptcy.||Four major deficiencies in the "financial safety net" for families confronting illness (1) lapses in insurance; (2) underinsurance in the face of serious illness; (3) lack of comprehensive employment-based coverage; and (4) lack of adequate disability insurance and paid sick leave for most families warrant broad policy reforms regarding medical and social insurance.|
|Kennickell (2000)||1989-1998 SCF.||SCF families and related response variables.||Descriptive.||Comparisons across different SCF datasets.||Net worth, response rates, sample types.||(1) Complex tax situations, varying rates of return, omitted asset income, and geographic distribution detrimentally affect the SCF list sample. (2) Unit nonresponse in the survey has become worse since at least 1992. (3) Interviewer and respondent errors indicate that survey instruments warrant further improvement.||(1) A "less noisy stratification" of the list sample can be achieved using multiple years of SOI data. (2) New research tools, interviewer incentives, and survey cooperation should be used to address the problem of unit nonresponse. (3) Further research should be dedicated to the area of instrument design and the imputation of missing data.|
|Kennickell (2003)||1989-2001 SCF.||U.S. families.||Descriptive.||Distribution of and shifts of wealth holdings.||Net worth, financial and non-financial assets, debts, equity.||(1) Roughly 1/3 of total wealth is held by the highest 1%, the next 9%, and the remaining 90% of the wealth distribution. (2) Wealth levels of the baby boomers trended upward during the 1989-2001 period. (3) Median wealth of white non-Hispanics was 6.4 times that of African Americans in 2001. This has decreased from a factor of 18.5 in 1989. A substantial portion of African American families had "middle class" net worth in 2001.||(1) Changes in financial services and economic structures have contributed to a variety of changes in the wealth distribution. (2) Leverage declines sharply with wealth. (3) Analysis of portfolio structure and institutional relationships would be helpful given the changes in the available set of financial services.|
|Kennickell & Sundén (1997)||1989 and 1992 SCF.||U.S. families.||(1) Descriptive. (2) OLS and robust regressions.||The effect of Social Security and other pension wealth on non-pension net worth.||Pension wealth, Social Security wealth, net worth.||(1) Including Social Security and pension wealth in the SCF measurement of net worth makes the net worth distribution more even. (2) Defined benefit plan coverage has a negative effect on non-pension net worth, while the effect of defined contribution plans and Social Security wealth is insignificant.||(1) While there has been a shift over time from defined benefit to defined contribution plans, workers do not seem to be contributing as much to the latter. (2) The insignificant effect of Social Security on saving may reflect households' uncertainty concerning the level of future Social Security benefits.|
|Lupton & Smith (1999)||HRS and 1984, 1989, and 1994 PSID.||U.S. households.||(1) Descriptive. (2) Multivariate OLS estimates of household savings.||Household saving.||Marital status, composition of wealth, sex, net worth, family income, race, age.||(1) Wealth distribution is skewed more than income distribution. (2) Net worth varies across marital status, with marital disparities being much larger among minorities. (3) The typical married couple has more social security wealth than personal net worth. (4) Savings between PSID waves are significantly lower among not married households. Savings differences between married and not married households are largest in the "earliest duration" in marital states and then tend to converge. (5) Children do not explain why married families save more.||There is a "quantitatively large" relationship between saving and marriage. The duration of the marriage, furthermore, positively affects wealth. The initial savings of married households contributes to the large wealth gap between married and not married households.|
|McKernan & Chen (2005)||1998 SCF.||Business owners.||Descriptive.||Small business and microenterprise programs and participation.||Age, firm size, net worth, race, sex, education, household income.||American small business owners tend to be male, white, have at least a high school education, and have moderate incomes. While anecdotal evidence suggests that small business grants and microenterprise programs increase entrepreneurship, studies that applied more stringent controls (for selection bias, for example) found no increase in wage rates or employment due to grants for these programs, but did find decreases in the length of unemployment spells.||(1) More research is needed to evaluate whether small business and microenterprise programs increase self-sufficiency. (2) Policies should consider self-sufficiency and economic development goals separately. (3) Standards and accreditation for programs should be created by foundations and policymakers. (4) Metrics for measuring success should be further developed. (5) Barriers that limit at-risk groups should be considered. (6) Small business and microenterprise should be evaluated against other programs with similar objectives.|
|Shapiro (2004)||Qualitative data from in-depth interviews, SIPP, PSID.||In-depth interview sample of 200 poor to middle-class families with school-age children in Boston, LA, and St. Louis.||Descriptive.||Receipt of transfer or financial assistance, effects of transfer/financial assistance.||Race.||(1) Sizable inheritances and inter vivos gifts can give young families a "head start" (ex: Allows home purchase in neighborhood with good schools). (2) Whites are more likely than blacks to receive sizable transfers. (3) Families with assets are able to acquire high-quality education for their children, and their education can transfer their economic advantages to their children.||Transfer of "transformative assets" perpetuates inequality.|
|Smith (1995)||HRS||U.S. households.||(1) Descriptive. (2) Multivariate OLS estimates of net worth using the HRS.||Comparisons of major asset surveys; racial, ethnic and other disparities in the distribution of net worth.||Race, age, education, income, region, health status, and other demographic variables.||(1) Missing values in surveys are much larger among the financial and real asset categories. (2) In the HRS, the distribution of wealth is severely skewed mean net worth is 2.4 times the median. (3) The average middle-aged black or Hispanic household has no liquid assets at their disposal.||(1) Compared to other surveys, the quality of HRS asset data is high, although nonresponses remain a problem. (2) Racial disparities in the HRS are mainly due to differential inheritances across generations, lower minority incomes, poorer health, and a very narrow definition of wealth that systematically excludes Social Security and employer pensions.|
|Smith (1999)||PSID, AHEAD, HRS.||U.S. households.||(1) Descriptive. (2) Ordered probit models of self-reported health status by income and wealth. (3) OLS models of new chronic health problems on household wealth.||Health and economic status.||Self-reported health status, income, wealth (models included also control for race, sex, age, marital status and education).||(1) Across all age groups, those in excellent health have more wealth than other respondents. Changes in health over time are also associated with wealth changes. (2) Out-of-pocket medical costs are relatively small for the average person. Despite these relatively low costs, the impact of new severe health problems on savings produces a mean wealth reduction of 7%. (3) There is a steep inverse relationship between employment grade and poor health outcomes.||(1) For middle to older age groups, there are significant effects of new health events on income and wealth, though this may not be true for earlier ages. (2) Though economic resources impact health outcomes, this direction of causality may be most pronounced during childhood and early adulthood.|
|Stavins (2000)||1998 SCF.||U.S. families.||(1) Descriptive. (2) Logit regressions measuring the likelihood of bill payment delinquency or prior bankruptcy filing.||Bankruptcy, bill payment delinquency.||Age, homeownership, income, net worth, unemployment, family size, marital status, number of credit cards, credit card balance, debt/income ratio.||(1) The strongest factors increasing the probability of being behind on bill payments were previously filing for bankruptcy and being unemployed at any time in the last 12 months. Past bankruptcy filing increases the probability of having delinquent loans whether or not bankruptcy filing was recent or 10+ years ago (p. 24). (2) Households with higher unpaid credit card balances are more likely to have filed for bankruptcy in the past (p. 25).|
|Sullivan (2004)||SIPP||Single women.||Descriptive||Vehicle ownership & vehicle equity.||Children-Yes/No.||(1) Families with a high probability of participating in welfare are more likely to have vehicle equity than any other type of asset. (2) 48% of all single mothers without a high school diploma own a car.||(1) Asset restrictions do have an effect on vehicle ownership. (2) Asset limits have no effect on liquid asset holdings.|
|Wolff (2004)||1983, 1989, 1992, 1995, 1998 SCF.||U.S. families.||Descriptive.||Distribution of net worth, mean/median net worth, & elements of wealth.||Income, age, children in family, female-headed family.||(1) The bottom income quintile does not have the assets to sustain consumption for any amount of time given an emergency versus 25 months for those in the highest income quintile. (2) In 1998, poor blacks, on average, had 1/4 the net worth of poor whites. (3) For the poor, mean net worth fell by 5% between 1983 and 1998.||(1) Only the richest 20% experienced large wealth gains between 1983 and 1998. (2) Wealth inequality continued to rise. (3) Overall indebtedness continues to rise. (4) There continue to be large wealth disparities across race.|
Data sources. Most of the wealth data in this report come from tables produced by Bucks, Kennickell, and Moore (2006) using the 19952004 Survey of Consumer Finances (SCF), Lerman (2005) using the 2001 Survey of Income and Program Participation (SIPP), Caner and Wolff (2004) using the 1984-99 panels of the Panel Study on Income Dynamics (PSID), and Lupton and Smith (1999) using the 1992 Health and Retirement Study (HRS).
These are all high quality surveys, but it is still important to bear in mind some of the data limitations of these surveys such as imputations for missing asset and liability data, 50-98 percent survey response rates, and the fact that some assets are difficult to measure and so are not included in many surveys. For example, Social Security benefits and defined-benefit pensions may be particularly important assets for low-income households, yet national surveys generally do not capture them (Ratcliffe et al. 2006). Vehicles are often captured, but other consumer durables such as appliances are missed (the SIPP is an exception). As this report depends in large part on national household surveys for the portraits of assets and liabilities, little can be said about Social Security Benefits, defined-benefit pensions, and holdings of durable goods other than vehicles. Social Security and defined-benefit pensions are also discussed in the Net Worth section and Appendix of this report.
Units of analysis. The units of analyses for the report are: the SCF family, SIPP household, HRS household, and PSID family. The appropriate definitions are described in detail in the Appendix. In general, households are often larger than families and thus are likely to have more assets and liabilities than families because of the greater number of people included in the household total. This report focuses on comparisons within surveys rather than across surveys, so the slight differences in unit of analysis across surveys will not affect comparisons. In describing findings, the term household or family is used as appropriate for the survey and research cited. For most surveys, a cohabiting partner who shares income and assets will be included in the unit of analysis.
Measures. The analysis focuses on median, rather than mean, holdings of assets, debt, and net worth due to the skewed distribution of wealth and for the sake of brevity. The median (and where applicable, mean) asset and debt values reported in the following exhibits apply only to families that hold the particular asset or debt and must be weighed against the likelihood of holding the item, also provided. This conditional mean provides a sense of the typical holding. One disadvantage of focusing on the median is that it is not additive. For example, median assets minus median liabilities does not necessarily equal median net worth. To allow comparisons over time and across data sets from different time periods, all values are converted to 2004 dollars using the consumer price index for all urban consumers.
Classifiers. To paint a portrait of the assets and liabilities of low-income households and compare how assets and liabilities are distributed among households, the following key classifiers are used: income, education, age, race or ethnicity, family structure, and housing status.
While current income is often used to determine who is low-income, this study also uses education because it may be a better predictor of potential lifetime income. Education helps to differentiate the long-term poor, who have low-paying jobs, from the short-term poor, who may be pursuing education.
Age is one of the most important classifiers because life cycle patterns are evident in the accumulation of assets and liabilities. For example, it would be unreasonable to expect meaningful accumulations of assets like home equity or pension wealth among younger populations. For younger households, holding an asset that is, purchasing an asset like a home or contributing to an asset like a pension may be more crucial than the current value of that asset. It is not necessarily that age per se is increasing assets a persons assets do not increase just because he or she gets older. Instead, getting older is usually associated with rising income, perhaps lower expenses (children move out), and possibly increased concern about retirement, all of which are associated with increasing assets.