Limitations of Combining Hispanics and African Americans for Analysis of Credit Problems

Article excerpt

This study uses a combination of six Survey of Consumer Finances data sets to examine whether factors affecting credit delinquency differ by the racial/ethnic identity of households. Hispanic households are less likely than white households and white households are less likely than African American households to be delinquent. Our full model with interaction terms shows that the effects of financially adverse events, financial buffers and debt burden on the debt delinquency differ across racial/ethnic groups. Combining African American and Hispanic households into one racial/ethnic minority group as previous studies have done can be problematic.

**********

Attention to racial/ethnic minorities' credit acquisition dates back at least to the 1970s (Canner and Smith 1991). Public policy goals have included increasing racial/ethnic minorities' access to credit by eliminating discrimination (Anderson and VanderHoff 1999), and redlining and racial/ethnic preference in credit lending have been investigated (Canner, Gabriel, and Woolley 1991). Racial/ethnic differences in borrowing opportunities and access to credit have been also discussed in empirical literature (e.g., Anderson and VanderHoff 1999; Munnell et al. 1996; Scharfer and Ladd 1981; Tootell 1996). Financial innovation has relaxed credit constraints of households and has provided affordable and easily accessible credit to most Americans (Greenspan 1997). As a result, it has been much easier for households traditionally constrained by the credit market, including racial/ethnic minorities, to take on debt and consume more than income (Lyons 2003).

With increases in credit usage of racial/ethnic minorities, higher rates of debt payment problems by African Americans and Hispanics have continued and constitute the most consistent finding across previous studies (e.g., Anderson and VanderHoff 1999; Berkovec and Gabriel 1995; Canner, Gabriel, and Wolley 1991; Getter 2003; Godwin 1999; Sullivan and Fisher 1988; Volkwein et al. 1998). The default rate for African American households was significantly higher than that for white households, even after controlling for household characteristics (Anderson and VanderHoff 1999). African Americans were also twice as likely to lose their homes due to foreclosures as whites (Warren and Tyagi 2004). More recently, African Americans and Hispanics suffered excessively in the subprime crisis (Rivera et al. 2008).

Rising household debt payment problems are a concern since they influence other aspects of household finances negatively. Late debt payment is recorded on credit reports, often resulting in lower credit scores. These credit reports are then used to determine the level of risk associated with most loans or insurance. Household debt payment problems can affect access to credit or homeownership. Racial/ethnic minorities without the economic cushion provided by household wealth face severe economic hardships due to household debt problems. In addition, prospective employers may use credit reports in choosing between job candidates (Manning 2001; Scott 2005).

The household credit literature has long suggested that there exist racial/ethnic disparities in debt payment performance. Nevertheless, a rigorous analysis of payment performance with attention to borrowers' race/ethnicity is lacking. Previous studies on this issue are limited in two aspects. First, most previous studies combined racial/ethnic groups, especially African American and Hispanic households and then compared these to white households or perhaps a combination of white and "other" households. Combining African American and Hispanic households has been done presumably because the number of credit problems of each racial/ethnic group for any single survey year used in those studies is relatively low, although as Lindamood, Hanna, and Bi (2007) noted, many authors have not offered explanations for combining racial/ethnic groups. Second, many previous studies used only those who had already been screened by lenders who had removed those at greatest risk of default (Greene 1998; Jacobson and Roszbach 2003). Consequently, these studies could examine factors associated with missed payments of people with credit scores sufficient to obtain credit. Therefore, their credit payment models could be biased by latent variables related to credit acquisition (Jacobson and Roszbach 2003).

This study extends previous studies in two important ways to provide a more complete picture of debt payment of households. First, this study uses six survey years (1992 through 2007) of the Survey of Consumer Finances (SCF) to obtain more robust estimates of the effects of race/ethnicity on debt payment performance. This study delineates three racial/ethnic categories (white, African American and Hispanic) rather than combining African Americans and Hispanics as done in most previous studies. Second, this study uses a logit selection framework, which is a modification of Heckman's (1979) two-step procedure. This study considers both those who have debt and those who do not to address possible sample selection bias. Addressing the sample selection issue will be an improvement from previous studies on debt payment performance, which derive data from a sample of only those who have been granted credit.

This study examines whether the racial/ethnic identity of households is related to their debt delinquency and shows how the effects of contributors to debt delinquency differ across racial/ethnic groups. The results of this study provide insight into the debt payment performance of American households and have important implications for financial education and research.

LITERATURE REVIEW

Racial/Ethnic Disparities in Obtaining Credit

Racial/ethnic disparities in credit acquisition have been a controversial issue in the United States dating back to the 1970s (Boehm and Schlottmann 2002). The debate was especially strong with the release of mortgage application data that was collected as part of the Home Mortgage Disclosure Act, which was intended to monitor minority and low income access to the mortgage market (Canner and Smith 1991). There has been evidence of racial/ethnic differences in credit acquisition, even though there has been no consensus on the causes of the differences. One stream of studies argues that racial/ethnic differences in credit acquisition have been due to debt applicants' creditworthiness. African American and Hispanic households had lower creditworthiness than white households. Therefore, they were more likely to be discouraged from applying for a loan (Crook 1996). Therefore, debt acquisition and debt payment problems were inextricably linked (Bostic and Lampani 1999). Another stream of studies attributes racial/ethnic disparities in credit acquisition to discrimination, which might exist in credit markets. Munnell et al. (1996) concluded that minority applicants were more likely to be rejected for mortgage loans than similarly creditworthy white applicants. Racial/ethnic minorities were significantly less likely to have debt, even after accounting for a large number of variables related to the creditworthiness of borrowers (Cox and Jappelli 1993; Crook 1996; Jappelli 1990).

Racial/Ethnic Disparities in Debt Payment

Literature presenting the relationship between race/ethnicity and debt payment has arisen out of a spectrum of motivations, but almost all authors concluded that racial/ethnic minority borrowers are more likely to miss debt payments or default than otherwise equivalent white borrowers. Canner, Gabriel, and Woolley (1991) estimated the probability of borrowers' delinquency in loan repayment using the 1983 SCF. They found that payment delinquency was more likely for a combined group of African American, Hispanic, and other households than for white households. Berkovec, Canner, and Hanna (1994) found that mortgages in neighborhoods with higher proportions of African American households had a higher likelihood of default than mortgages in neighborhoods with higher proportions of white households.

Several studies demonstrated possible causes of racial/ethnic disparities in debt payment problems. Volkwein et al. (1998) examined the similarities and differences among whites, African Americans and Hispanics using a nationally representative database of student loan borrowers. They found that the variables that influenced student loan defaults held constant across white and minority populations, but the extent of these variables' influence differed in accordance with race/ethnicity. Ross and Yinger (2002) attributed higher average default rates among minorities to the fact that minority borrowers were more likely to have larger debt burdens and higher loan-to-value ratios than white applicants on average. Getter (2003) showed that a combined group of African American and Hispanic households was more likely to miss a loan payment than were white households. He attributed the higher delinquency risk for African American and Hispanic borrowers to negative income stressors, which neither the lender nor the borrower could have anticipated at the time of the credit application. The majority of previous studies combined African American and Hispanic households (e.g., Canner, Gabriel, and Woolley 1991; Canner and Smith 1991; Crook 1996; Munnell et al. 1996). In other studies, it was not clear whether African American and Hispanic households were being compared with white households or to a combination of white and other households (e.g., Getter 2003). However, in a study that separated racial/ethnic groups, Lee and Hanna (2008) analyzed factors related to getting behind or missing payments on household debt using the 2004 SCF. They found that conditional upon having debt, African American households had much higher predicted debt payment delinquency rates than white households, while Hispanic households had lower predicted delinquency rates than white households. Lee and Hanna demonstrated that African American and Hispanic households had different incidences of debt payment delinquency, even though the two groups had similar characteristics in terms of low homeownership rates, income and net worth. Therefore, they suggested more in-depth research on debt payment problems of the two groups.

METHOD

Theoretical Framework

In perfect financial markets, households would be able to insure against income disrupting events (Weinberg 2006). An important implicit assumption of basic Life Cycle Savings models without uncertainty is that households' borrowing and debt payment occur as planned. Lawrence (1995) noted that most models of consumption ignore default despite the potential importance of the default option to households. She extended the Life Cycle Savings model to explain why some households choose to default when they experience unexpected income loss.

Unexpected changes in the borrower's circumstances can trigger debt payment problems (Elmer and Seelig 1998; Gross and Souleles 2002; Reeder 2004; Springer and Waller 1993). Even though borrowers may intend to repay debt, they may find that they are unable to pay debt if they experience adverse events. These events include health problems or changes in income. Health problems can reduce income (Sullivan, Warren, and Lawrence 2000) and also can create large, unanticipated medical expense and resultant debt, especially if the household does not have adequate health insurance coverage (Zywicki 2005). Changes in economic circumstances such as having unexpectedly low income significantly increase the probability of late payment (Getter 2003). Coles (1992) reported that 40% of mortgage arrears and repossessions are associated with income shocks such as business failure and unemployment. Therefore, we expect that having household income drop or having household members with poor health will increase the likelihood of debt delinquency.

Financial resources that reduce the repercussions of financially adverse events might reduce the likelihood of debt payment problems (Black and Morgan 1999; Elmer and Seelig 1998; Getter 2003; Godwin 1999; Reeder 2004). These financial buffers might protect households against unexpected changes in income, reducing the impact on their ability to keep up with payments. In some cases, households can successfully adapt to negative financial events and continue to make regular debt payments by relying on financial buffers. For example, a lapse in employment may result in only a minor disruption for a household with sufficient precautionary financial assets. Similarly, health expenditures related to poor health status may be mitigated by current health insurance coverage. If, however, a household is unable to draw on household wealth or use health insurance, the impact of a drop in income or the presence of health problems may be much more devastating. Therefore, we expect that financial buffers such as household wealth and health insurance reduce the repercussions of financially adverse events and reduce the likelihood of debt delinquency.

Households might have difficulties in paying their debt because they excessively spend and become overburdened. In this case, a high household debt burden will be positively related to payment problems (Sullivan and Fisher 1988). With high debt burdens, even small income decreases can lead to financial distress (Black and Morgan 1999).

Empirical Models

Two-Step Estimation Previous studies on household debt payment have used those who had already been screened by lenders who had presumably rejected those at greatest risk of default (Greene 1998; Jacobson and Roszbach 2003). Consequently, the debt payment models might suffer from a sample selection bias because they are estimated from a sample of households who had debt. This deficiency might be due to the lack of publicly available data on rejected debt applicants.

To avoid this type of bias, Heckman (1979) proposed a two-step estimation procedure. Our debt delinquency model uses the basic approach with two equations. One equation is for the binary decision to have household debt, [y.sub.1i]. The other equation is for the binary outcome to be delinquent on household debt payment, [y.sub.2]. Following Greene's (1998) formulation of the model, we let the superscript * denote an unobserved variable:

[y.sup.*.sub.1i] = b x [X.sub.1i] - [[epsilon].sub.1i] (1)

[y.sup.*.sub.2i] = g x [X.sub.2i] - [[epsilon].sub.2i], for i = 1,2, ..., N. (2)

where [X.sub.ji], j = 1, 2, are 1 x [k.sub.j] vectors of independent variables and the disturbances are assumed to be zero-mean, bivariate normally distributed with unit variance and a correlation coefficient [rho]. The binary choice variable, [y.sub.1i], takes value 1 if the household had debt and 0 if the household did not:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (3)

The second binary variable, [y.sub.2i], takes the value 1 if the household was delinquent on their debt payment and 0 if not:

1 if the household was delinquent on debt payment ([y.sup.*.sub.2i] [greater than or equal to] 0), [y.sub.2i] 0 if the household was not delinquent on debt payment 2i

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (4)

But we only observe y such that

[y.sub.1i] = 1 if [y.sup.*.sub.1i] > 0 and 0 else.

[y.sub.2i] [y.sup.*.sub.1i] if [y.sub.1i] 1[y.sub.1i] not observed if [y.sub.1i] = 0,

[y.sub.2i] and [X.sub.2i] are only observed if [y.sub.1i] = 1,

[y.sub.1i] and [X.sub.1i] are observed for all households.

Selectivity [[epsilon].sub.1i][[epsilon].sub.2i]] ~ [N.sub.2][0, 0, 1, 1[[rho].sub.[epsilon]1[epsilon]2]].

The attributes in the vector [beta] are the factors used in the household's debt-holding equation and the attributes in the vector [gamma] are the factors used in the household's debt delinquency. The probability of debt delinquency given that a household had debt is estimated by a logistic regression.

First-Step Estimation: Debt Holding Model

The empirical model of this study is built upon two parts. The first part analyzes the holding of household debt. The second part focuses on the payment of household debt. The probability of holding of household debt is expected to be the joint outcome of supply of and demand for household debt. A complete explanation of the factors leading to the holding of household debt would need to take into account the supply of debt by lenders with complex decisions involving risk and uncertainty. However, the probability of a debt holding is expected to depend primarily on the demand for debt since household survey data typically do not contain all the information considered by creditors. The implications of the first-step estimation are integrated into the second-step estimation. We expect that the probability of having debt is determined by

DEBT = f(X)(5)

DEBT refers to debt holding. Vector X includes regressors reflecting demographics and life cycle effects.

Second-Step Estimation: Debt Delinquency

We expect that debt delinquency is determined by financially adverse events, financial buffers, household debt burden and household characteristics. Debt delinquency is a binary dependent variable that is equal to 1 if the household reports that during the past year it was behind in the debt payments by two months or more and 0 otherwise. Therefore, the relationship is assumed to be as follows for the second model:

DELINQUENCY = f (FE, FB, DB, X, inverse Mills ratio), (6)

where DELINQUENCY refers to debt delinquency, FE refers to financially adverse events, FB refers to financial buffers, DB refers to debt burden and X refers to other control variables (see the Data section for details on the variables and how they were constructed from the data). The inverse Mills ratio is computed from the output of the first analysis and is included as an explanatory variable in the second-step analysis to estimate the probability of debt delinquency. We pool six data sets of the SCF and therefore include dummy variables for survey year to test for possible shifts over time.

To compare the impact of these selected factors across racial/ethnic group, we first estimate the probability of debt delinquency based on vectors of these selected factors as well as indicator variables of race/ethnicity with all households. The null hypothesis is that racial/ethnic groups are not different from each other in the probabilities of debt delinquency when other variables are controlled. If we reject this hypothesis, then an important question is whether the effects of household characteristics on debt delinquency differ between racial/ethnic groups.

We need to separately estimate the impact of these selected factors on debt delinquency for each racial/ethnic group to provide adequate results for each racial/ethnic group. That is, we examine whether and how the impact of financially adverse events (FE), financial buffers (FB) and debt burden (DB) differ across different racial/ethnic groups using the following equation:

Prob(DELINQUENCY) = [[alpha].sub.k] + [[beta].sub.k] + FB + [[gamma].sub.k] DB + [[delta].sub.k] X + inverse Mills ratio + [epsilon], (7)

where [alpha] is the coefficient of financially adverse events, [beta] the coefficient of financial buffers, [gamma] the coefficient of debt burden, [delta] the coefficient of other controls, k the racial/ethnic groups and [epsilon] the error term. Finally, this study examines whether the impact of selected independent variables on debt delinquency differs across different racial/ethnic groups. Because race/ethnicity is a categorical variable, it is represented by two dummy variables representing African American and white, with Hispanic being treated as the reference group. The racial/ethnic dummy variables are conceptualized as the moderator variable and the qualitative/continuous variables are focal independent variables. Interaction terms are generated between each of these dummy variables and focal independent variables. Demographic characteristics other than racial/ethnic status are used as control variables.

Racial/ethnic status has two possible effects. The first could be a constant effect that is an overall or homogenous effect of race/ethnicity on debt delinquency. The second possible type of effect, a coefficient effect, implies that determinants of debt delinquency differ between racial/ethnic groups. These effects would be identified if the effects of an explanatory variable differ across racial/ethnic groups. The cause of such differences may have been explained by differences in financially adverse events, financial buffers and household debt burden tied to these coefficient effects.

Data

To estimate our debt delinquency models, we use data from the SCF. The SCF is a triennial cross-sectional survey sponsored by the Board of Governors of the Federal Reserve System (Bucks et al. 2009). It provides detailed information on the finances of the US households. In particular, it contains information on the demand for debt with rich detail on household characteristics (Zinman 2004). It allows for the examination of household debt issues by drawing from its extensive reporting of both household balance sheets and households' financial services usage.

A number of previous studies compared households with white respondents to households with minority respondents, combining several racial/ethnic groups rather than distinguishing between data on African Americans, Hispanic Americans, and white households. They broadly compared minority groups, which were not separated according to race or ethnicity, to the white group. These studies did not include a large subset of any one racial or ethnic minority group, because the number of debt delinquencies for any single survey year is relatively low, so that econometric estimates for a separate group other than white based on a single survey would lack robustness. Therefore, this study uses six SCF data sets (1992-2007) to obtain an adequate sample size and increase the reliability of descriptive and multivariate tests.

Analytical Sample

The 1992, 1995, 1998, 2001, 2004 and 2007 SCF data sets have a total of 25,889 households. However, the racial/ethnic identification of the respondent is different across implicates in some households (Hanna and Lindamood 2008). Since the racial/ethnic identifications are critical in this study, this study deletes 71 households that do not have the same racial/ethnic identification in all five implicates. The weighted proportions of racial/ethnic identification of household respondents are 75% white, 13% African American, 8% Hispanic and 4% others.

We use the repeated-imputation inference (RII) method to correct for underestimation of variances associated with parameter estimates due to imputation of missing data (Montalto and Sung 1996). Weights provided for use by the Federal Reserve Board are used for descriptive analyses to adjust for systematic differences in response rates by demographic groups, as well as to adjust for the sample design to produce point estimates that are nationally representative. The probit and logit estimates are not weighted to avoid possible endogeneity bias (Lindamood, Hanna, and Bi 2007).

The overall debt delinquency rate is low, so in each survey year in the SCF, there are small numbers of households reporting they were two months or more late in debt payments. The weighted percent of all households in debt delinquency for the combined sample is 5.6%, with the rate at 4.4% in 1992, 5.3% in 1995, 5.9% in 1998, 5.3% in 2001, 6.9% in 2004 and 5.5% in 2007. Table 1 shows the actual (unweighted) total numbers and the numbers in each racial/ethnic group who were delinquent in each survey year. The actual number of Hispanic households in delinquency in each survey year in the SCF data sets is very small, so separate analysis of delinquency of Hispanic households in a particular survey year would not produce robust estimates of effects.

Variable Identification

Racial/Ethnic Classification

This study uses the same classification system of racial/ethnic categories as the public release data sets of the SCF. Each respondent is asked, "which of these categories do you feel best describe you ?" when presented with six racial/ethnic categories. The public data set, however, combines Asian, American Indian, Alaska Native, Native Hawaiian, Other Pacific Islander and Other into a single category and presents four racial/ethnic categories (i.e., whites, African Americans, Hispanics, and Others). On the basis of the Census reports, a majority of respondents choosing one of the "Others" categories are of Asians or Pacific Islander ancestry (Hanna and Lindamood 2008). Even though the "Others" category in the public versions of the SCF data sets plausibly has a majority of respondents who are Asian Americans (Hanna and Lindamood 2008), the category also includes Native Americans, who have very different household characteristics in terms of education and income. Because of the diversity of this group in the public SCF data sets, we exclude those households from our analyses. This study delineates three racial/ethnic categories: white, African American, and Hispanic. The SCF changed its method of asking racial/ethnic questions in the 2004 survey, asking a separate Hispanic question in addition to the categories offered in the 1989 through 2001 surveys. Researchers using the 2004 or more recent versions of the SCF could use the separate question to determine whether a respondent choosing a non-Hispanic category in the old question chose Hispanic in the separate question. However, this study uses multiple SCF data sets from 1992 to 2007. Therefore, this study uses the old racial/ethnic categories instead of employing a new racial/ethnic category. As mentioned previously, we exclude households with different respondent racial/ethnic identifications across implicates. Our analysis uses the remaining 24,861 households from the 1992 to 2007 SCF data sets.

Debt Holding and Debt Delinquency

We construct two dependent variables with the consideration of sample selection effects. The first dependent variable is whether a household had debt. The second dependent variable is whether a household was behind in the debt payments by two months or more. These dependent variables are measured by the following two SCF questions.

X3004: Now thinking of all the various loan or mortgage payments you made during the last year, were all the payments made the way they were scheduled or were payments on any of the loans sometimes made later or missed?

1. All paid as scheduled or ahead of schedule

5. Sometimes got behind or missed payments

0. Inapplicable

X3005: Were you ever behind in your payments by two months or more?

1. Yes

5. No

0. Inapplicable

Using these two SCF questions, each household is grouped into one of the four payment performance tiers: no household debt, scheduled payment, having missed payments by less than two months, and having missed payments by two months or more. This categorization allows for distinction between those who experienced serious payment delinquency by two months and those who only missed a payment occasionally due to temporary financial mishaps or forgetting to mail a payment. We define debt delinquency as having missed payments by two months or more.

Independent Variables for Debt Holding

The following variables are included as controls for demographics, wealth effects or life cycle effects on the probability of debt holding. We include age of the household head, race/ethnicity of the respondent, family composition, household income, expected income for next year and employment status and education of the head. We also include a set of year dummy variables to control for any time trends.

Independent Variables for Debt Delinquency

Racial/ethnic differences in debt delinquency are confounded by racial/ethnic differences in other factors that influence the debt delinquency. To account for the racial/ethnic difference in debt payment, we

control for covariates that mediate the racial/ethnic differences in debt delinquency.

We measure financially adverse events using negative transitory income and poor health status. Studies using cross-sectional data have been criticized for being conducted in static or comparative static contexts (Oxford Economic Research Associates 2004). A snapshot from a particular point in time cannot adequately address debt issues that may arise over an extended period (Bridges and Disney 2004). Getter (2003), however, used the SCF question asking whether last year's income was at the usual level or higher or lower as a key measure of financial shock occurrence. We use a binary variable of transitory income change that indicates whether respondent's income one year before the survey year was unusually low. Our health status variable is based on the answer to the following question: "Would you say your health is excellent, good, fair or poor?" The SCF codes the answers to this question on a 1-4 scale, with 1 representing excellent health and 4 representing poor health. Although the four-category scale of health status is presented, a dichotomous measure reflecting a response of "poor" health has been found to be both reliable and highly correlated with household debt performance (Lyons and Yilmazer 2005; Yilmazer and DeVaney 2005). We create a dichotomous variable that for noncouple households takes a value of 1 if the respondent reported health as poor and 0 otherwise. For couple households, if either the respondent or the spouse/partner reported having poor health, the variable is coded as 1, and if neither had poor health, it is coded as 0, allowing the estimation of the impact of both spouses' health problems on debt payment performance.

We include two variables, net worth and health insurance coverage, to control the impact of financial buffers on debt delinquency. The log of net worth is used to allow a nonlinear relationship between debt delinquency and net worth since money-based variables are more prone to give rise to nonlinear relationships than any other variable (Cohen et al. 2003). All net worth and income variables for the six survey years are converted to 2007 dollars using the coding provided on the SCF Web site. We include a binary variable of whether everybody in the household was covered by health insurance.

The ratio of monthly payments on debt to monthly household income is used as a proxy for the household debt burden. The ratio assesses the level of indebtedness of the US households and provides a view of the economic health of the overall household sector (Greenspan 2004).

The ratio measures the share of income committed by households for paying interest and principal on their debts. It results in a more accurate comparison of the immediate financial stress that households experience than the total-debt-to-total-income ratio, because the latter ratio can be misleading due to differing lengths of maturity and interest rates (Getter 2003). A high ratio reflects that households have less money available to purchase goods or services. This study includes independent variables on the basis of the theoretical relevance to the research literature discussed above. The variables are included as controls for household characteristics, wealth effects and life cycle effects on the probability of debt delinquency. We include age and age squared to allow for nonlinear effects of age on debt holding and debt delinquency. We also include the log of household income, dummy variables for education level, family composition, expected household income growth and employment status.

In addition, dummy variables for survey year are included to test for any time trends. Some variables that might have an impact on delinquency, such as the interest rate, could not be used because of the inconsistency across households and across survey years, (1) Other variables such as gender were considered and tested but are not presented. (2)

RESULTS

Descriptive Results

Table 2 presents descriptive statistics of debt holding and debt delinquency. Overall, 70% of households reported they had debt. The pattern of debt holding differed significantly by race/ethnicity, with 72% of white households, 64% of African American households and 65% of Hispanic households reporting that they had debt. African Americans and Hispanics were significantly less likely to have debt than whites.

For those who had debt, 8% answered that they were delinquent on their debt payment by two months or more. Among households with debt, 15% of African American households, 11% of Hispanic households, and 7% of white households were behind in their payments by two months or more. African Americans and Hispanics were more likely to be delinquent on debt payment than whites, and African American households were significantly more likely to be delinquent than Hispanic households.

Descriptive statistics of racial/ethnic groups are presented in Table 3.

The percentages in each row represent the proportion of each racial/ethnic group in the category. The adverse event variables (lower than normal income and poor health) all have significantly lower rates for white households than for African American and Hispanic households, and there are small differences in poor health between African American and Hispanic households. For instance, the proportion reporting that current income was lower than normal was higher for African American households (22%) and Hispanic households (24%) than for white households (17%). There are small differences in poor health between African American and Hispanic households.

Financial buffers that are supposed to keep households out of economic troubles also have significantly higher rates for white households than for African American and Hispanic households. The percentage of households with all family members covered by health insurance was relatively high for white households (85%) compared with African American households (74%) and Hispanic households (55%), and the rate for Hispanic households was significantly lower than the rate for African American households.

African American and Hispanic households were in weaker economic positions, as the percentages of those who were in the lowest net worth quartiles were 46% and 50%, respectively, which was much higher than the percentage of white households (19%). Hispanics were more likely to be young than African Americans, who were more likely to be age 60 and over than Hispanics. Hispanics were the most likely to have less than a high school diploma as the highest level of education of the head, with 44%, compared with only 25% for African Americans. Hispanics were more likely to be in a couple household with children (36%) than African Americans (15%). The distribution of current household income was somewhat similar for Hispanics and African American households, but employment status differed, with Hispanics less likely to be retired and more likely to be self-employed than African Americans.

Our primary focus is on differences between African American and Hispanic households, and even though African American and Hispanic households had somewhat similar patterns of current household income levels, transitory income and health, African American households were more likely to have health insurance, be older, have a high school degree, have a college degree and be retired than Hispanic households.

Multivariate Results

Combined Estimations

The first multivariate procedure (Table 4) is a probit analysis performed both to identify factors affecting the likelihood of having debt and to estimate the inverse Mills ratio reflecting the selection bias in the estimation of debt delinquency. The second multivariate procedure (Table 4) is a logistic regression for the likelihood of debt delinquency.

The probit analysis shows that whites and African Americans were more likely to have debt than otherwise similar Hispanics. On the basis of the results from the same statistical procedure with white as the reference group (not shown), whites were more likely to have debt than African Americans. Married couples with children were more likely to have debt than other types of households. On the basis of the combined effect of age and age squared, the likelihood of having debt increased with age until age 32, then decreased. Households with some college degree were more likely to have debt than those with less than a high school degree. As household income and net worth increased, the likelihood of having debt decreased. Households in 1992 had the lowest likelihood of having debt. Households with heads employed were more likely to have debt than those with self-employed, retired or not working heads.

The logistic regression for the likelihood that a household was seriously delinquent on debt payment shows that whites and African Americans had a significantly higher likelihood of delinquency than otherwise similar Hispanics. On the basis of results with white as the reference group, African Americans were more likely to be delinquent on debt than were whites (not shown).

On the basis of the combined effect of age and age squared, the likelihood of debt delinquency increased with age until age 41, then decreased. Couples without children were less likely to be delinquent than other types of households. Consumers who reported poor health status were more likely to have debt delinquency than those who reported excellent, good or fair health status. Households with current income lower than normal income were more likely to be delinquent than those with current income the same as or higher than normal. Financial buffers decreased the likelihood of debt payment delinquency. Having all household members covered by health insurance and higher net worth were associated with a lower likelihood of delinquency. The likelihood of being delinquent was lower in 1992 than in other years. Those with some college degree had a higher likelihood of delinquency than did those with no high school diploma. Employees had the highest likelihood of payment delinquency compared with all other employment statuses.

The effect of the Mills ratio was significant in the logistic regression results, which is evidence that the probability of having debt and the probability of debt payment delinquency are correlated. This result indicates that controlling for sample selection is critical to obtain unbiased estimates in the payment delinquency model.

Separate Estimations

Table 5 presents the coefficients, standard errors and significance levels of the probit regression analyses of the debt holding and logistic regression analyses of debt delinquency for each racial/ethnic group. The effects of most independent variables on the probability of debt holding were similar across racial/ethnic groups. The combined effect of age and age squared implies that the likelihood of having debt increased with age until age 31 for whites, age 39 for African Americans and age 31 for Hispanics, and then decreased. Homeownership was positively related to the probability of having debt. Couple households with children under 19 were more likely to have debt than any other group. The likelihood of having household debt was higher in 2007 than in 1992.

The results of the second multivariate procedure showed whether the effects of financial buffers were similar across the three racial/ethnic groups. As net worth increased, the probability of debt delinquency decreased for each racial/ethnic group. Having all household members covered by health insurance decreased the probability of debt delinquency for each racial/ethnic group. The effects of trigger events differed somewhat across groups. Poor health status increased the probability of debt delinquency for white and African American households, whereas the impact of poor health status was not significant for Hispanic households. Having negative transitory income (having income lower than normal) increased the probability of debt delinquency for each racial/ethnic group. The debt-to-income ratio did not have a significant effect on payment delinquency for any of the groups.

The probability of debt delinquency was higher for households with children than households without children regardless of their marital status for each racial/ethnic group. As household income increased, the probability of debt delinquency decreased for white households, but income did not have a significant effect for African American and Hispanic households. The combined effect of age and age squared implies that the likelihood of debt delinquency increased with age until age 40 for white households and until age 53 for African American households, and then decreased. (The pattern for Hispanic households was somewhat similar but neither age nor age squared effects are significant.)

The effect of the Mills ratio was significant for white and African American households, which means that controlling for sample selection bias is critical to arrive at unbiased estimates in the debt delinquency model.

Combined Model with Interaction Effects

Interaction terms between race/ethnicity and selected independent variables provide evidence of racial/ethnic differences of the magnitude of impact of the independent variables on debt delinquency. We first tested all possible interaction effects of independent variables with racial/ethnic categories, but with all 48 interaction terms in the logistic regression, 42 of the interaction terms had large variance inflation factors, suggesting that multicollinearity is a severe problem if all interaction terms are included. Therefore, we present a model with interaction terms only for adverse events (poor health and low transitory income), financial buffers (health insurance and net worth) and the debt burden. Table 6 shows the results of a logistic model with interaction terms for the combined sample of white, African American, and Hispanic households. (This is the second stage of a selection estimation, but we do not show the probit results, as they are generally similar to the results shown in Table 4.) The significant effects for the dummy variables for African American and for white indicate that African American and white households with characteristics matching the reference categories for adverse events, financial buffers, debt burden and any other variable were more likely than comparable Hispanic households to be delinquent.

The interaction terms for poor health imply that poor health increased the debt delinquency rate for African American and white households but not for Hispanic households. The interaction terms for negative transitory income are not significant, implying that having lower than normal income increased the debt delinquency rate for each group similarly.

The interaction terms for racial/ethnic group with the log of net worth and health insurance imply that having these buffers reduced the chance of debt delinquency more for white households than for Hispanic households, but the effect of these buffers were similar for African American and Hispanic households.

Having a higher debt burden decreased the debt delinquency rate for Hispanic households, but the effect for African American and white households is small, considering the combination of the main effect and the effect for each interaction term.

DISCUSSION

Summary of Results

An important public policy goal for several decades has been to increase access to credit for racial/ethnic minorities. Lyons (2003) suggested that financial innovation has relaxed credit constraints for racial/ethnic minorities, making it easier for racial/ethnic minorities to obtain credit. Most previous studies have found higher rates of debt payment problems for combined groups of African American and Hispanic households than for white households. Measuring racial/ethnic disparities in credit problems and finding important factors that are correlated with credit problems across each race/ethnicity is a prerequisite for addressing the disparities. This study is the first to examine the different impacts of financially adverse events, financial buffers and debt burden on debt delinquency across different racial/ethnic groups.

The descriptive patterns (Table 2) are that white households were more likely to have debt than African American and Hispanic households, and among those who had debt, the proportion of those who were delinquent on their payment was higher for African American and Hispanic households than for white households. These descriptive findings are consistent with the findings from previous studies (e.g., Berkovec, Canner, and Hanna 1994; Canner, Gabriel, and Wolley 1991; Getter 2003; Godwin 1999). One possible explanation for the higher debt delinquency rates of African American and Hispanic households is their generally weaker economic position and lower education level. In general, individuals with lower income might be less able to afford the out-of-pocket costs of health care, even if they have health insurance coverage. Less education may impair their familiarity with complex financial terms, their ability to communicate with creditors and to understand credit providers' instructions.

The combined multivariate analysis of debt delinquency (Table 4) shows among households with debt, African American and white households were more likely than Hispanic households to be delinquent. Despite some similarities between African American and Hispanic households in terms of income and net worth, with Hispanics being more disadvantaged in terms of health insurance coverage and education (Table 3), Hispanics are less likely than African American households to have debt delinquency (Table 4).

In the separate logistic regression models (Table 5), as net worth increased, the probabilities of debt delinquency decreased for each racial/ethnic group. Having all household members covered by health insurance decreased the probabilities of debt delinquency for each racial/ethnic group. Having negative transitory income increased the probabilities of delinquency for each group, whereas poor health status increased the probabilities of debt delinquency for white and African American households. Household debt burden did not affect the probability of debt delinquency for any racial/ethnic group.

The full model with interaction terms showed that the magnitudes of effects of financially adverse events, financial buffers and debt burden on the debt delinquency differed across race/ethnic groups. Financially adverse events increased the probability of debt delinquency for all racial/ethnic groups, but the magnitudes of impact of financially adverse events on likelihood of debt delinquency differed across racial/ethnic groups. Poor health affected debt delinquency more for white and African American households than for Hispanic households. Even though financial buffers such as net worth and health insurance lowered the likelihood of debt delinquency for all racial/ethnic groups, the magnitude of impact of the two proxy variables on the likelihood of debt delinquency were greater for white households than for the other groups. In other words, white households were even more vulnerable to lack of health insurance than African American households. The interaction model showed that the magnitudes of household debt burden for white and Hispanics were not different between the two groups.

In sum, the selected variables were significant but the effects were larger for white households than for African American and Hispanic households. Assuming homogenous effects of these factors on debt payment for all racial/ethnic groups is not a reasonable assumption. Even though African American and Hispanic households have been combined for analysis in previous studies, the impacts of several factors were quite different from the two groups. Therefore, combining African American and Hispanic households into one racial/ethnic minority group as previous studies have done can be problematic.

Implication for Policymakers

Our research shows that disadvantaged racial/ethnic minorities have different credit patterns, and therefore, policies for information and analysis of lending patterns should separate Hispanic and African American households. In particular, efforts to identify discriminatory lending patterns should analyze these groups separately, as combining these groups may distort underlying patterns. Our research did not have the objective of identifying racial/ethnic discrimination, but our results have implications for future research on lending discrimination. Phelps (1972) identified discrimination as being statistical or taste. Becker (1971) suggested that profit-maximizing firms would not engage in taste discrimination. However, statistical discrimination might exist if lenders have limited information about consumers and so they find it efficient to assume that individual borrowers have characteristics of their racial/ethnic group. Both statistical and taste discrimination are illegal under the US federal law, but presumably competition would reduce taste discrimination, and plausibly the increase in credit information available to lenders would reduce the need for statistical discrimination. There is mixed evidence about the extent of lending discrimination (e.g., Dymski 2006) but if lenders engaged in taste discrimination, the group discriminated against should have lower default rates than the other groups (Han 2004). Our finding that Hispanic households have lower predicted delinquency rates than white households, and African American households have higher predicted delinquency rates than white households, does not directly provide evidence about discrimination but does suggest the importance of analyzing these two groups separately.

Implications for Researchers

The United States is a multicultural country where the population of racial/ethnic minorities is becoming more visible (Watchravesringkan 2008). However, the majority of studies regarding consequences of financial management have focused on the dominant US racial/ethnic group, whites. Research on racial/ethnic minorities has been limited, with most studies simply comparing whites to a combined group of all other racial/ethnic groups. This study used six data sets from SCF from 1992 to 2007 because the number of debt delinquency of each racial/ethnic group in a single survey year was relatively low for our econometric estimates. Data sets that have higher numbers of African American and Hispanic households in each survey year should be used to obtain more robust estimates in racial/ethnic differences in factors related to debt delinquency.

Implications for Educators

Financial education can help borrowers better manage their debt and avoid payment problems. Financial educators and professionals need to better understand how borrowers' characteristics can affect their debt payment performance. The financial management and control of money within a household is influenced not only by a family's socioeconomic status but by its marriage patterns and the ideology of family (Singh 1997; Waseem 2004). Our results show racial/ethnic differences in debt payment problems, so financial education programs need to be more sensitive to differences between racial/ethnic groups.

Our results are similar to those of Getter (2003) in terms of the impact of adverse events (poor health and lower than normal income) and financial buffers (health insurance and net worth) on the likelihood of debt payment problems. While the impacts of these factors differs somewhat between racial/ethnic groups, general personal financial advice about considering risks before taking on credit seems applicable to all households.

DOI: 10.1111/j.1745-6606.2012.01237.x

REFERENCES

Anderson, Richard and James VanderHoff. 1999. Mortgage Default Rates and Borrower Race. Journal of Real Estate Research, 18 (2): 279-289.

Becker, Gary. 1971. The Economics of Discrimination. Chicago, IL: The University of Chicago Press.

Berkovec, James A., Glenn B. Canner, and Timothy H. Hanna. 1994. Race, Redlining, and Residential Mortgage Loan Performance. Journal of Real Estate Finance and Economics, 9: 263-294.

Berkovec, James A. and Stuart A. Gabriel. 1995. Discrimination and the Performance of FHA-Insured Mortgage Loans. In Fair Lending Analysis, edited by Anthony M. Yezer (95-104). Washington, DC: American Bankers Association.

Black, Sandra E. and Donald P. Morgan. 1999. Meet the New Borrowers. Current Issues in Economics and Finance, 5 (3): 1-6.

Boehm, Thomas P. and Alan M. Schlottmann. 2002. Housing and Wealth Accumulation: Intergenerational Impacts. In Low Income Homeownership: Examining the Unexamined Goal, edited by Nicolas P. Retsinas and Eric S. Belsky (407-426). Cambridge, MA: Joint Center for Housing Studies.

Bostic, Raphael W. and Patrick K. Lampani. 1999. Racial Differences in Patterns of Small Business Finance: The Impact of Local Geography, Business Access to Capital and Credit. A Federal Reserve System Research Conference: Proceedings of a Conference Held in Arlington, VA, March. Chicago, IL: Federal Reserve Bank of Chicago, pp. 149-179.

Bridges, Sarah and Richard Disney. 2004. Use of Credit and Arrears on Debt among Low-Income Families in the United Kingdom. Fiscal Studies, 25: 1-25.

Bucks, Brian K., Arthur B. Kennickell, Traci L. Mach, and Kevin B. Moore. 2009. Changes in U.S. Family Finances from 2004 to 2007: Evidence from the Survey of Consumer Finances. Federal Reserve Bulletin, 95 (February): A1-A55.

Canner, Glenn B. and Dolores S. Smith. 1991. Home Mortgage Disclosure Act: Expanded Data on Residential Lending. Federal Reserve Bulletin, 77 (November): 859-881.

Canner, Glenn B., Stuart A. Gabriel, and Michael J. Woolley. 1991. Race, Default Risk and Mortgage Lending: A Study of the FHA and Conventional Loan Markets. Southern Economic Journal, 58 (1): 249-262.

Cohen, Jacob, Patricia Cohen, Stephen G. West, and Leona S. Aiken. 2003. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences. Mahwah, N J: Lawrence Erlbaum Associates.

Coles, Adrian. 1992. Causes and Characteristics of Arrears and Possessions. Council of Mortgage Lenders Housing Finance, 13: 10-12.

Cox, Donald and Tullio Japelli. 1993. The Effect of Borrowing Constraints on Consumer Liabilities. Journal of Money, Credit, and Banking, 25: 197-213.

Crook, Jonathan. 1996. Credit Constraints and US Households. Applied Financial Economics, 6 (6): 477-485.

Dymski, Gary A. 2006. Discrimination in the Credit and Housing Markets: Findings and Challenges. In Handbook on the Economics of Discrimination, edited by William Rodgers (215-259). Cheltenham, UK: Edward Elgar.

Elmer, Peter J. and Steven A. Seelig. 1998. Insolvency, Trigger Events, and Consumer Risk Posture in the Theory of Single-Family Mortgage Default. Journal of Housing Research, 10 (1): 1-25.

Getter, Darryl E. 2003. Contributing to the Delinquency of Borrowers. The Journal of Consumer Affairs, 37 (1): 86-100.

Godwin, Deborah D. 1999. Predictors of Households' Debt Repayment Difficulties. Financial Counseling and Planning, 10 (1): 67-78.

Greene, William H. 1998. Sample Selection in Credit-scoring Models. Japan and the World Economy, 10: 299-316.

Greenspan, Alan. 1997. Consumer Credit and Financial Modernization. Remarks at the Economic Development Conference of the Greenlining Institute, San Francisco, California, October 11.

Greenspan, Alan. 2004. Understanding Household Debt Obligations. Remarks at the Credit Union National Association 2004 Governmental Affairs Conference. Washington, DC, February 23.

Gross, David B. and Nicholas S. Souleles. 2002. An Empirical Analysis of Personal Bankruptcy and Delinquency. The Review of Financial Studies, 15 (1): 319-347.

Han, Song. 2004. Discrimination in Lending: Theory and Evidence. Journal of Real Estate Finance and Economics, 29 (1): 5-46.

Hanna, Sherman D. and Suzanne Lindamood. 2008. The Decrease in Stock Ownership by Minority Households. Journal of Financial Counseling and Planning, 19 (2): 46-58.

Heckman, James J. 1979. Sample Selection Bias as a Specification Error. Econometrica, 47 (1): 153-162.

Jacobson, Tor and Kasper Roszbach. 2003. Bank Lending Policy, Credit Scoring and Value-at-Risk. Journal of Banking & Finance, 27: 615-633.

Jappelli. Tullio. 1990. Who Is Credit Constrained in the U.S. Economy? Quarterly Journal of Economics, 105 (1): 219-234.

Lawrence, Emily C. 1995. Consumer Default and the Life Cycle Model. Journal of Money, Credit and Banking, 27 (4): 939-954.

Lee, Jonghee and Sherman D. Hanna. 2008. Delinquency Patterns by Racial/Ethnic Status: A Selection Model. Proceedings of the Academy of Financial Services.

Lindamood, Suzanne, Sherman D. Hanna, and Lan Bi. 2007. Using the Survey of Consumer Finances: Methodological Considerations and Issues. Journal of Consumer Affairs, 41 (2): 195-214.

Lyons, Angela C. 2003. How Credit Access Has Changed Over Time for U.S. Households. The Journal of Consumer Affairs, 37 (2): 231-255.

Lyons, Angela C. and Tansel Yilmazer. 2005. Financial Strain and Health: Evidence from The Survey of Consumer Finances. Southern Economic Journal, 71 (4): 873-890.

Manning, Robert D. 2001. Hearing on 'Consumer Bankruptcy,' Before the Committee on Judiciary. House of Representative. http://judiciary.senate.gov/oldsite/te02080lrdm.htm&e=6251.

Montalto, Catherine P. and Jaimie Sung. 1996. Multiple Imputation in the 1992 Survey of Consumer Finances. Financial Counseling and Planning, 7: 133-146.

Munnell, Alicia H., Lynn E. Browne, James McEneaney, and Geoffrey M.B. Tootell. 1996. Mortgage Lending in Boston: Interpreting HMDA Data. American Economic Review, 86 (1): 25-53.

Oxford Economic Research Associates. 2004. Are UK Households Over-Indebted? Oxford: Oxera Consulting Limited.

Phelps, Edmund S. 1972. The Statistical Theory of Racism and Sexism. American Economic Review, 62 (4): 659-661.

Reeder, William J. 2004. The Role of Trigger Events in Ending Homeownership Spells: A Literature Review and Suggestions for Further Research. Abt Associates Inc., pp. 1-23. http://www.abtassociates.com/reports/triggers_final_report.pdf.

Rivera, Amaad, Brenda Cotto-Escalera, Anisha Desai, Jeannette Huezo, and Dedrick Muhammad. 2008. Foreclosed: State of the Dream. Boston, MA: United for a Fair Economy.

Ross, Stephen and John Yinger. 2002. The Color of Credit: Mortgage Discrimination, Research Methodology, and Fair-Lending Enforcement. Cambridge, MA: MIT Press.

Scharfer, Robert and Helen F. Ladd. 1981. Discrimination in Mortgage Lending. Cambridge, MA: MIT Press.

Scott, Robert H., III. 2005. The Determinants of Default on Credit Card Debt. PhD dissertation, University of Missouri, Kansas City.

Singh, Supriya. 1997. Marriage Money: The Social Shaping of Money in Marriage and Banking. St. Leonards, NSW: Allen & Unwin.

Springer, Thomas M. and Neil G. Waller. 1993. Lender Forbearance: Evidence from Mortgage Delinquency Patterns. Journal of the American Real Estate and Urban Economics Association, 21 (1): 27-46.

Sullivan, Charlene A. and Robert M. Fisher. 1988. Consumer Credit Delinquency Risk: Characteristics of Consumers Who Fall Behind. Journal of Retail Banking, 10: 53-64.

Sullivan, Teresa A., Elizabeth Warren, and Westbrook J. Lawrence. 2000. The Fragile Middle Class: Americans in Debt. New Haven, CT: Yale University Press.

Tootell, Geoffrey M.B. 1996. Redlining in Boston: Do Mortgage Lenders Discriminate Against Neighborhoods. Quarterly Journal of Economics, 111 (4): 1049-1079.

Volkwein, Fredericks J., Bruce P. Szelest, Alberto F. Cabrera, and Michelle R. Napierski-Prancl. 1998. Factors Associated with Student Loan Default among Different Racial and Ethnic Groups. Journal of Higher Education, 69: 206-237.

Warren, Elizabeth and Amelia W. Tyagi. 2004. The Two-Income Trap: Why Middle-Class Parents Are Going Broke. New York: Basic Books.

Waseem, Saba. 2004. Household Monies and Decision-Making (Policy Research Paper 23). Canberra, Australia: FACS.

Watchravesringkan, Kittichai. 2008. Financial Behavior of Hispanic Americans. In Handbook of Consumer Finance Research, edited by Jing J. Xiao (271-285). New York: Springer.

Weinberg, John A. 2006. Borrowing by U.S. Households. Economic Quarterly, Federal Reserve Bank of Richmond, 92 (3): 177-194.

Yilmazer, Tansel and Sharon A. DeVaney. 2005. Household Debt Over the Life Cycle. Financial Services Review, 14: 285-304.

Zinman, Jonathan. 2004. Why Use Debit Instead of Credit? Consumer Choice in a Trillion-Dollar Market. Federal Reserve Bank of New York Staff Reports. Staff Report No. 191, July 2004. http://www.newyorkfed.org/banking/debit_or_credit.pdf.

Zywicki, Todd J. 2005. An Economic Analysis of the Consumer Bankruptcy Crisis. Northwestern University Law Review, 99 (4): 1463-1541.

(1.) During the 1992 2007 period, the credit industry shifted from denying credit to awarding credit at higher rates to previously credit-constrained households, a trend discussed by Lyons (2003). This shift probably changed the interest rates charged to different racial/ethnic groups and these differences could have affected the likelihood of delinquency. Unfortunately, SCF questions about interest rates differed across survey years and the variety of types of interest rates makes a standard measure of interest rates across households and across survey years impractical. The debt-income ratio variable captures some of the possible effects of interest rate differences.

(2.) We included a gender variable in our models and found that the gender variable was not significantly related with the probability of debt delinquency. Given that household heads are defined as male in mixed-sex couple households in the SCF (Lindamood, Hanna, and Bi 2007), and single-head households with children are likely to be female, adding gender would confound the interpretation of family type. Adding gender does not change the effects of the racial/ethnic variables. Therefore, we do not present results with the gender variable in Tables 4-6.

Jonghee Lee (jay.jongheelee@gmail.com) is an Instructor at the Department of Home Economics, Korea University. Sherman D. Hanna (hanna.1@osu.edu or sdhanna@gmail.com) is a Professor at the Consumer Sciences Department, Ohio State University.

TABLE 1

Total Number of Households and the Number of Households in Each
Racial/Ethnic Group  and Numbers of Delinquent Households

                                           Total               African
Survey Year                               Sample     White    American

1992 total number                          3,890      3,138       355
1992 number in delinquency                   141         91        26
1995 total number                          4,283      3,551       379
1995 number in delinquency                   173        123        31
1998 total number                          4,298      3,495       413
1998 number in delinquency                   188        126        45
2001 total number                          4,429      3,576       461
2001 number in delinquency                   197        126        38
2004 total number                          4,506      3,511       482
2004 number in delinquency                   245        148        70
2007 total number                          4,412      3,514       409
2007 number in delinquency                   184        127        35
Total number in combined sample           25,818     20,785     2,499
Number in delinquency, combined sample     1,128        741       245

Survey Year                               Hispanic    Others

1992 total number                             217       180
1992 number in delinquency                     17         7
1995 total number                             176       177
1995 number in delinquency                     13         6
1998 total number                             249       141
1998 number in delinquency                     10         7
2001 total number                             275       117
2001 number in delinquency                     24         9
2004 total number                             347       166
2004 number in delinquency                     21         7
2007 total number                             313       176
2007 number in delinquency                     19         3
Total number in combined sample             1,577       957
Number in delinquency, combined sample        104        39

Note: Unweighted numbers of households, excluding households with
different racial/ethnic  identification in different implicates.

TABLE 2

Racial/Ethnic Comparison in Debt Holding and Debt Delinquency

                     Debt-Holding
Race/Ethnicity         Rate (%)       Significance

White                   72.12             (bc)
African American        64.23             (a)
Hispanic                65.26             (a)
All                     70.41

                   Debt Delinquency
Race/Ethnicity         Rate (%)       Significance

White                    6.60             (bc)
African American        15.16             (ac)
Hispanic                10.53             (ab)
All                      7.97

Note: Weighted analyses of all five implicates of the 1992-2007 SCF,
excluding households  with differences in racial-ethnic
identification across implicates, and excluding households with
respondents identifying themselves as "other," for a sample size of
24,861 households. Significance  tests are based on RII means tests.
The percentages are based on the SCF population weight variable.  The
debt delinquency rate is the percentage of debt-holding households
that were behind in their  debt payments by two months or more.

(a) Significantly different from white rate at p < .01.

(b) Significantly different from African American rate at p < .01.

(c) Significantly different from Hispanic rate at p < .01.

TABLE 3

Selected Independent Variables by Race/Ethnicity ([dagger])

                                                 Race / Ethnicity

                                                    African
Variables                 Categories        White   American   Hispanic

Adverse events
  Transitory         Higher than usual       9.43     7.52      10.52
  Income             Lower than usual       16.66    22.45      24.47
                     Same as usual          73.91    70.03      65.00
  Health             Excellent              31.09    23.01      25.99
                     Good                   46.49    44.53      41.84
                     Fair                   16.48    24.23      25.89
                     Poor                    5.95     8.23       6.29
Financial buffers
  Health insurance   Yes                    85.00    73.62      55.11
                     No                     15.00    26.38      44.89
  Net worth          NW <$13,368            19.00    45.64      49.52
                     $13,368 [less than     24.66    31.89      27.25
                       or equal to] NW
                       <$92,164
                     $92,164 [less than     28.03    16.54      16.39
                       or equal to] NW
                       <$285,732
                     $285,732 [less than    28.31     5.93       6.84
                       or equal to] NW
Household debt burden
  Debt to income     DIR <10%               74.11    75.80      72.87
  Ratio              10% [less than or      18.07    16.95      19.50
                       equal to] DIR
                       <25%
                     25% [less than or       4.85     4.14       4.48
                       equal to] DIR
                       <40%
                     40% [less than or       2.96     3.12       3.15
                       equal to] DIR
Demographics
  Age                20-29                  13.48    17.20      25.32
                     30-39                  19.55    24.77      28.11
                     40-49                  21.53    22.45      22.77
                     50-59                  15.93    14.81      12.91
                     60 and over            29.51    20.76      10.89
  Education          Less than high         13.39    24.79      43.72
                       school
                     High school            31.16    35.12      28.77
                     Some college           18.55    19.85      14.13
                     Bachelor degree        36.89    20.24      13.37
  Family type        Couple w/o child       29.59    12.69      17.83
                     Couple with child      25.87    15.00      36.19
                     Single w/o child       36.22    44.13      25.61
                     Single with child       8.32    28.18      20.36
  Homeownership      Yes                    72.60    46.02      44.13
                     No                     27.40    53.98      55.87
  Current income     Income <22,150.90      20.78    43.61      37.16
                     22,150.90 [less        24.19    25.24      30.71
                       than or equal to]
                       income < 43,287.50
                     43,287.50 [less        26.32    19.66      21.08
                       than or equal to]
                       income < 77,457.20
                     77,457.20 [less        28.7     11.48      11.05
                       than or equal to]
                       income
  Employment         Self-employed          11.62     4.19       7.20
                     Not working            14.86    24.02      10.60
                     Retired                19.19    13.76       2.54
                     Salary earner          54.32    58.03      63.35
  Expected income    Sure the same          34.58    25.52      25.37
                     Sure increase          14.14    12.17      14.33
                     Sure decrease          22.80    14.94      11.33
                     Not sure               28.48    47.36      48.98
Environmental variable
  Survey year        1992                   14.98    14.87      14.39
                     1995                   17.02    16.68      12.05
                     1998                   17.09    15.49      15.24
                     2001                   17.28    17.55      17.35
                     2004                   16.96    18.56      20.44
                     2007                   16.67    16.85      20.54

                                               Difference
Variables                 Categories        ([double dagger])

Adverse events
  Transitory         Higher than usual            (ab)
  Income             Lower than usual             (ab)
                     Same as usual                 (ab
  Health             Excellent                    (abc)
                     Good                         (abc)
                     Fair                         (abc)
                     Poor                         (abc)
Financial buffers
  Health insurance   Yes                          (abc)
                     No                           (abc)
  Net worth          NW <$13,368                  (abc)
                     $13,368 [less than           (abc)
                       or equal to] NW
                       <$92,164
                     $92,164 [less than            (ab
                       or equal to] NW
                       <$285,732
                     $285,732 [less than          (abc)
                       or equal to] NW
Household debt burden
  Debt to income     DIR <10%                     (abc)
  Ratio              10% [less than or            (bc)
                       equal to] DIR
                       <25%
                     25% [less than or             NS
                       equal to] DIR
                       <40%
                     40% [less than or             NS
                       equal to] DIR
Demographics
  Age                20-29                        (abc)
                     30-39                        (abc)
                     40-49                         (b)
                     50-59                        (abc)
                     60 and over                  (abc)
  Education          Less than high               (abc)
                       school
                     High school                  (abc)
                     Some college                 (abc)
                     Bachelor degree              (abc)
  Family type        Couple w/o child             (abc)
                     Couple with child            (abc)
                     Single w/o child             (abc)
                     Single with child            (abc)
  Homeownership      Yes                          (ab)
                     No                           (abc)
  Current income     Income <22,150.90            (abc)
                     22,150.90 [less              (abc)
                       than or equal to]
                       income < 43,287.50
                     43,287.50 [less              (abc)
                       than or equal to]
                       income < 77,457.20
                     77,457.20 [less              (ab)
                       than or equal to]
                       income
  Employment         Self-employed                (abc)
                     Not working                  (ab)
                     Retired                      (abc)
                     Salary earner                (abc)
  Expected income    Sure the same                (abc)
                     Sure increase                (ac)
                     Sure decrease                (abc)
                     Not sure                     (ab)
Environmental variable
  Survey year        1992                          NS
                     1995
                     1998                          NS
                     2001                          NS
                     2004                         (abc)
                     2007                         (bc)

NS = no differences are significant.

([dagger]) Using weighted analyses of all five implicates of the
1992-2007 SCF.

([double dagger]) RII means tests are used to test for a relationship
between race/ethnicity and each independent  variable.

(a) White rate is significantly different from African American rate
at p < .01.

(b) White rate is significantly different from Hispanic rate at
p < .01.

(c) African American rate is significantly different from Hispanic
rate at p < .01.

TABLE 4

Probit Regression of Probability of Debt Holding and Logistic
Regression of Probability  of Debt Delinquency

                                                 Debt Holding

Variable              Categories          Estimate SE     Significance

Age              Age                      0.0388 0.0016       ***
                 Age squared             -0.0006 0.0001       ***
Education        Less than high
                   school#
                 High school              0.2237 0.0144       ***
                 Some college             0.4014 0.0162       ***
                 Bachelor degree          0.3459 0.0148       ***
Family type      Couple with child        0.2396 0.0132       ***
                 Single w/o child        -0.1477 0.0112       ***
                 Single with child       -0.1692 0.0171       ***
                 Couple w/o child#
Income           Log of income           -0.0088 0.0024       ***
Net worth        Log of net worth        -0.0306 0.0011       ***
Employment       Self-employed           -0.0753 0.0124       ***
                 Not working             -0.4861 0.0125       ***
                 Retired                 -0.4699 0.0152       ***
                 Salary earner#
Expected         Sure the same            0.0020 0.0124
Income           Sure increase            0.0378 0.0148        *
                 Not sure                -0.1016 0.0124       ***
                 Sure decrease#
Year             1995                     0.2182 0.0143       ***
                 1998                     0.5410 0.0148       ***
                 2001                     0.5532 0.0147       ***
                 2004                     0.5572 0.0147       ***
                 2007                     0.5962 0.0148       ***
                 1992#
Race/ethnicity   African American         0.1109 0.0208       ***
                 White                    0.1928 0.0178       ***
                 Hispanic#
Homeowner        Yes                      0.8769 0.0125       ***
                 No#
Insurance        Yes
                 No#
Debt burden      Debt-to-income ratio
Health           Poor health
                 Excellent, fair or
                   good#
Transitory       Lower than usual
  Income           income
                 Same as usual or
                   higher than usual
                   income#
Mills ratio
Intercept                                -0.4526 0.0504       ***

                                               Debt Delinquency

Variable              Categories          Estimate SE     Significance

Age              Age                      0.1134 0.0082       ***
                 Age squared             -0.0014 0.0001       ***
Education        Less than high
                   school#
                 High school              0.2143 0.0517       ***
                 Some college             0.3821 0.0573       ***
                 Bachelor degree          0.0053 0.0584
Family type      Couple with child        0.6661 0.0522       ***
                 Single w/o child         0.2099 0.0535       ***
                 Single with child        0.5935 0.0584       ***
                 Couple w/o child#
Income           Log of income           -0.2778 0.0187       ***
Net worth        Log of net worth        -0.0863 0.0025       ***
Employment       Self-employed           -0.2305 0.0501       ***
                 Not working             -0.2510 0.0473       ***
                 Retired                 -0.6288 0.1083       ***
                 Salary earner#
Expected         Sure the same           -0.2287 0.0472       ***
Income           Sure increase           -0.0269 0.0535
                 Not sure                 0.0140 0.0438
                 Sure decrease#
Year             1995                     0.3019 0.0588       ***
                 1998                     0.2928 0.0613       ***
                 2001                     0.3485 0.0611       ***
                 2004                     0.5195 0.0596       ***
                 2007                     0.3062 0.0628       ***
                 1992#
Race/ethnicity   African American         0.5055 0.0620       ***
                 White                    0.2088 0.0560       ***
                 Hispanic#
Homeowner        Yes
                 No#
Insurance        Yes                     -0.0863 0.0025       ***
                 No#
Debt burden      Debt-to-income ratio     0.0149 0.0081
Health           Poor health              0.7910 0.0613       ***
                 Excellent, fair or
                   good#
Transitory       Lower than usual         0.5109 0.0347       ***
  Income           income
                 Same as usual or
                   higher than usual
                   income#
Mills ratio                               1.2072 0.0997       ***
Intercept                                -1.8014 0.2671       ***

Note: The values marked in bold have been taken as reference category
used in the probit analysis and/or in the logistic analysis.

* p < .05, *** p < .001.

Note: The category used in the probit analysis and/or in the logistic
analysis are indicated with #.

TABLE 5

Probit Regression Estimates of Debt Holding and Logistic Regression
Estimates of Debt Delinquency

                                                        White

                                                    Debt holding

Variable             Categories                Estimate   Significance

Age                  Age                        0.0369        ***
                     Age squared               -0.0006        ***
Education            High school                0.2203        ***
                     Some college               0.3705        ***
                     Bachelor degree            0.3332        ***
                     Less than high school#
Family type          Couple with child          0.1969        ***
                     Single w/o child          -0.2015        ***
                     Single with child         -0.1124        ***
                     Couple w/o child#
Current income       Log of income             -0.0112        ***
Net worth            Log of net worth          -0.0449        ***
Employment           Self-employed             -0.0170
                     Not working               -0.4723        ***
                     Retired                   -0.4524        ***
                     Salary earner#
Expected income      Sure the same             -0.0108
                     Sure increase              0.0461        ***
                     Not sure                  -0.0720        ***
                     Sure decrease#

Year                 1995                       0.2343        ***
                     1998                       0.5643        ***
                     2001                       0.5601        ***
                     2004                       0.5705        ***
                     2007                       0.6128        ***
                     1992#
Homeownership        Yes                        0.8409        ***
                     No#
Health               Poor health
                     Excellent, fair or
                       good#
Transitory income    Lower than usual
                       income
                     Same as usual or
                       higher than usual
                       income#
Net worth            Log of net worth
Health insurance     Yes
                     No#
Debt-to-income
  ratio
Intercept
Mills ratio
Percent concordant

                                                        White

                                                   Debt delinquency

Variable             Categories                Estimate   Significance

Age                  Age                        0.1371        ***
                     Age squared               -0.0017        ***
Education            High school                0.0759
                     Some college               0.1706         *
                     Bachelor degree           -0.2092         **
                     Less than high school#
Family type          Couple with child          0.5730        ***
                     Single w/o child           0.0183
                     Single with child          0.5696        ***
                     Couple w/o child#
Current income       Log of income             -0.3460        ***
Net worth            Log of net worth
Employment           Self-employed             -0.1589         **
                     Not working               -0.2788        ***
                     Retired                   -0.6006        ***
                     Salary earner#
Expected income      Sure the same             -0.1768         **
                     Sure increase             -0.0445
                     Not sure                   0.1741        ***
                     Sure decrease#

Year                 1995                       0.3631        ***
                     1998                       0.4019        ***
                     2001                       0.4581        ***
                     2004                       0.6019        ***
                     2007                       0.4997        ***
                     1992#
Homeownership        Yes
                     No#
Health               Poor health                0.8214        ***
                     Excellent, fair or
                       good#
Transitory income    Lower than usual           0.5501        ***
                       income
                     Same as usual or
                       higher than usual
                       income#
Net worth            Log of net worth          -0.0972        ***
Health insurance     Yes                       -0.6245        ***
                     No#
Debt-to-income                                 -0.0448
  ratio
Intercept                                       -1.302        ***
Mills ratio                                      1.370        ***
Percent concordant                              83.9

                                                   African American

                                                    Debt holding

Variable             Categories                Estimate   Significance

Age                  Age                        0.0232        ***
                     Age squared               -0.0003        ***
Education            High school                0.1833        ***
                     Some college               0.4663        ***
                     Bachelor degree            0.6905        ***
                     Less than high school#
Family type          Couple with child          0.5188        ***
                     Single w/o child          -0.0446
                     Single with child         -0.0702
                     Couple w/o child#
Current income       Log of income              0.0750        ***
Net worth            Log of net worth          -0.0183        ***
Employment           Self-employed             -0.2987        ***
                     Not working               -0.5091        ***
                     Retired                   -0.6765        ***
                     Salary earner#
Expected income      Sure the same              0.1382         **
                     Sure increase              0.2143        ***
                     Not sure                  -0.0565
                     Sure decrease#

Year                 1995                       0.2630        ***
                     1998                       0.6042        ***
                     2001                       0.7208        ***
                     2004                       0.6227        ***
                     2007                       0.7129        ***
                     1992#
Homeownership        Yes                        1.1219        ***
                     No#
Health               Poor health
                     Excellent, fair or
                       good#
Transitory income    Lower than usual
                       income
                     Same as usual or
                       higher than usual
                       income#
Net worth            Log of net worth
Health insurance     Yes
                     No#
Debt-to-income
  ratio
Intercept
Mills ratio
Percent concordant

                                                   African American

                                                   Debt delinquency

Variable             Categories                Estimate   Significance

Age                  Age                        0.0424         **
                     Age squared               -0.0004         *
Education            High school                0.2783         *
                     Some college               0.5749        ***
                     Bachelor degree            0.5361        ***
                     Less than high school#
Family type          Couple with child          0.8439        ***
                     Single w/o child           0.6914        ***
                     Single with child          0.8179        ***
                     Couple w/o child#
Current income       Log of income             -0.0065
Net worth            Log of net worth
Employment           Self-employed             -0.3632         *
                     Not working                0.0517
                     Retired                   -0.7369         **
                     Salary earner#
Expected income      Sure the same             -0.3537         **
                     Sure increase              0.2615         *
                     Not sure                  -0.2457         *
                     Sure decrease#

Year                 1995                       0.1300
                     1998                       0.3127         *
                     2001                       0.0628
                     2004                       0.5638        ***
                     2007                      -0.1542
                     1992#
Homeownership        Yes
                     No#
Health               Poor health                0.6296        ***
                     Excellent, fair or
                       good#
Transitory income    Lower than usual           0.4090        ***
                       income
                     Same as usual or
                       higher than usual
                       income#
Net worth            Log of net worth          -0.0766        ***
Health insurance     Yes                       -0.5797        ***
                     No#
Debt-to-income                                  0.0302
  ratio
Intercept                                      -3.3747        ***
Mills ratio                                     0.9345        ***
Percent concordant                             74.9

                                                      Hispanic

                                                    Debt holding

Variable             Categories                Estimate   Significance

Age                  Age                        0.0183         **
                     Age squared               -0.0003         **
Education            High school                0.3067        ***
                     Some college               0.6339        ***
                     Bachelor degree            0.5914        ***
                     Less than high school#
Family type          Couple with child          0.3477        ***
                     Single w/o child          -0.0069
                     Single with child         -0.0001
                     Couple w/o child#
Current income       Log of income              0.0810        ***
Net worth            Log of net worth           0.0038
Employment           Self-employed             -0.0905
                     Not working               -0.2512        ***
                     Retired                   -0.4935        ***
                     Salary earner#
Expected income      Sure the same              0.0564
                     Sure increase              0.0729
                     Not sure                  -0.0905
                     Sure decrease#

Year                 1995                      -0.0059
                     1998                       0.4542        ***
                     2001                       0.4728        ***
                     2004                       0.4417        ***
                     2007                       0.4348        ***
                     1992#
Homeownership        Yes                        1.0841        ***
                     No#
Health               Poor health
                     Excellent, fair or
                       good#
Transitory income    Lower than usual
                       income
                     Same as usual or
                       higher than usual
                       income#
Net worth            Log of net worth
Health insurance     Yes
                     No#
Debt-to-income
  ratio
Intercept
Mills ratio
Percent concordant

                                                       Hispanic

                                                   Debt delinquency

Variable             Categories                Estimate   Significance

Age                  Age                        0.0509
                     Age squared               -0.0005
Education            High school                0.2215
                     Some college               0.6221         **
                     Bachelor degree            0.2371
                     Less than high school#
Family type          Couple with child          0.8204        ***
                     Single w/o child           0.3893
                     Single with child          1.1539        ***
                     Couple w/o child#
Current income       Log of income             -0.0640
Net worth            Log of net worth
Employment           Self-employed             -0.3206
                     Not working               -0.4403         **
                     Retired                   -1.8503
                     Salary earner#
Expected income      Sure the same             -0.4205
                     Sure increase             -0.3879
                     Not sure                  -0.3060
                     Sure decrease#

Year                 1995                       0.1426
                     1998                      -0.6801         **
                     2001                       0.1553
                     2004                      -0.4765         **
                     2007                      -0.3194
                     1992#
Homeownership        Yes
                     No#
Health               Poor health                0.4122
                     Excellent, fair or
                       good#
Transitory income    Lower than usual           0.4807        ***
                       income
                     Same as usual or
                       higher than usual
                       income#
Net worth            Log of net worth          -0.0559        ***
Health insurance     Yes                       -0.3308         **
                     No#
Debt-to-income                                 -1.6437
  ratio
Intercept                                      -2.6481
Mills ratio                                     0.4911
Percent concordant                             73.7

Note: The values marked in bold have been taken as reference category
used in the probit analysis and/or in the logistic analysis.

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

Note: The category used in the probit analysis and/or in the logistic
analysis are indicated with #.

TABLE 6

Logistic Regression Estimates of Debt Delinquency with Interaction
Terms (a)

                                                     Main effect

                          Parameter            Estimates       SE

Race/ethnicity            African American       0.6096     0.1099 ***
                          White                  0.4994     0.0965 ***
                          Hispanic
Adverse events
  Health                  Poor health            0.1850     0.1851
                          Excellent, fair or good
Transitory                Lower than usual       0.5118     0.0893 ***
                            income
  Income                  Same as usual or higher than usual income
Financial buffers
  Health insurance        Yes                   -0.2642     0.0850 **
                          No
  Net worth               Log net worth         -0.0701     0.00717 ***
Household debt burden
  Debt-to-income ratio                          -2.7996     0.9138 **
Demographics
  Age                     Age                    0.1137     0.0092 ***
                          Age squared           -0.0014     0.0001 ***
  Education               High school            0.1728     0.0571 **
                          Some college           0.3543     0.0702 ***
                          Bachelor degree       -0.0345     0.0676
                          Less than high school
                          Couple with child      0.6969     0.0526 ***
                          Single w/o child       0.2248     0.0546 ***
                          Single with child      0.6013     0.0597 ***
                          Couple w/o child
  Income                  Log income            -0.2719     0.0190 ***
  Employment              Self-employed         -0.2066     0.0496 ***
                          Not working           -0.2462     0.0658 ***
                          Retired               -0.6715     0.1188 ***
                          Salary earner
Expected income           Sure the same         -0.2283     0.0466 ***
                          Sure increase         -0.0358     0.0527
                          Not sure               0.0028     0.0443

                          Sure decrease
Environmental variable
  Year                    1995                   0.3160     0.0624 ***
                          1998                   0.3123     0.0801 ***
                          2001                   0.3849     0.0803 ***
                          2004                   0.5305     0.0800 ***
                          2007                   0.3074     0.0844 ***
                          1992
  Mills ratio                                    1.3040     0.2380 ***
  Intercept                                     -2.3238     0.3566 ***

                                                  Interaction of
                                                 Hispanic, African
                                                   American (b)

                          Parameter            Estimates      SE

Race/ethnicity            African American
                          White
                          Hispanic
Adverse events
  Health                  Poor health            0.4573     0.2222 *
                          Excellent, fair or good
Transitory                Lower than usual      -0.2279     0.1174
                            income
  Income                  Same as usual or higher than usual income
Financial buffers
  Health insurance        Yes                   -0.1748     0.1103
                          No
  Net worth               Log net worth         -0.0056     0.0079
Household debt burden
  Debt-to-income ratio                           2.8223     0.9141 **
Demographics
  Age                     Age
                          Age squared
  Education               High school
                          Some college
                          Bachelor degree
                          Less than high school
                          Couple with child
                          Single w/o child
                          Single with child
                          Couple w/o child
  Income                  Log income
  Employment              Self-employed
                          Not working
                          Retired
                          Salary earner
Expected income           Sure the same
                          Sure increase
                          Not sure
                          Sure decrease
Environmental variable
  Year                    1995
                          1998
                          2001
                          2004
                          2007
                          1992
  Mills ratio
  Intercept

                                                   Interaction of
                                                Hispanic, White (c)

                          Parameter            Estimates       SE

Race/ethnicity            African American
                          White
                          Hispanic
Adverse events
  Health                  Poor health            0.6778    0.1970 ***
                          Excellent, fair or good
Transitory                Lower than usual       0.0894    0.0971
                            income
  Income                  Same as usual or higher than usual income
Financial buffers
  Health insurance        Yes                   -0.4421    0.0930 ***
                          No
  Net worth               Log net worth         -0.0254    0.0068 ***
Household debt burden
  Debt-to-income ratio                           2.7810    0.9139 **
Demographics
  Age                     Age
                          Age squared
  Education               High school
                          Some college
                          Bachelor degree
                          Less than high school
                          Couple with child
                          Single w/o child
                          Single with child
                          Couple w/o child
  Income                  Log income
  Employment              Self-employed
                          Not working
                          Retired
                          Salary earner
Expected income           Sure the same
                          Sure increase
                          Not sure
                          Sure decrease
Environmental variable
  Year                    1995
                          1998
                          2001
                          2004
                          2007
                          1992
  Mills ratio
  Intercept

(a) Interaction terms between racial/ethnic dummies and selected
independent variables are included for the analyses.

(b) Reference group is Hispanic.

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

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.