The Savings and Credit Management of
Low-Income, Low-Wealth Black and White Families
William D. Bradford
Endowed Professor of Business and Economic Development
Professor of Finance
School of Business Administration
University of Washington
Seattle, WA 98195-3200
Phone: 206 543 4559
Fax: 206 221 6856
AUTHOR’S NOTE: I am grateful to Timothy Bates and Constance Dunham for their helpful
comments on previous versions of this article.
The Savings and Credit Management of
Low-Income, Low-Wealth Black and White Families
First, the greater reluctance of low-income, low-wealth Black families to use checking
and savings accounts than low-income, low-wealth White families is only partially explained by
regression models that consider a wide array of demographic variables. The greater reluctance
impedes Black families from getting the benefits of participating in financial markets. Second,
the success of low-income, low-wealth Black families in managing credit is not found to differ
from that of low-income, low-wealth White families. But the Black families owe less credit card
and other noncollateralized debt than the White families, implying either that Black families have
a lower demand for such debt or that lenders are biased against them. Third, Black families are
less likely to achieve wealth increases than White families because of differences in labor income
and, to a lesser degree, gifts and inheritances. This has implications for the continued wealth
disparity between Black Americans and White Americans.
The Savings and Credit Management of
Low-Income, Low-Wealth Black and White Families
This study examines the savings and credit management of U.S. families that are both
low income and low wealth. I seek to answer three questions. First, how and why do low-
income, low-wealth Black and White families differ in participation in financial markets? I look
specifically at the families’ use of transaction accounts (checking and savings accounts) and
noncollateralized debt (primarily credit cards and personal loans). Second, how do these
families differ in credit management success, that is, paying bills on time and maintaining good
relationships with creditors? Third, how and why do these families differ in success at increasing
family wealth over time?
These questions are significant for three reasons. First, public and private providers of
financial services can be more effective in working with low-income, low-wealth consumers if
they better understand these consumers. Second, Black applicants are known to be rejected more
often for mortgage loans and commercial loans than White applicants.1
After controlling for
relevant variables, if low-income, low-wealth Black families display less success at credit
management than comparable White families, then higher rejection rates are not surprising,2
the use of public resources to provide training on managing credit relationships may be as
valuable as finding and punishing acts of prejudicial lending discrimination. Third, with regard
to wealth management, the mean and median wealth of Black families are much lower than those
of White families in the United States (Altonji, Doraszelski, & Segal, 2000, Blau & Graham,
1990, Oliver & Shapiro, 1990, Wolff, 1998), and movement toward wealth equality is slow at
best (Bradford, 2000a). This lack of parity has been described as an endemic problem in the
On mortgage lending see the articles cited by Yinger (1998) and Ladd (1998). On commercial lending,
see Ando (1988), Bates (1997), Bates & Bradford (1992), Blanchflower, Levine, & Zimmerman (1998) and
Cavalluzzo & Cavalluzzo (1998).
This relationship leads to statistical discrimination (Phelps, 1972). Firms use race to infer likelihood of
default on the loan. Here, discrimination does not reflect prejudicial discrimination (as discussed by
Becker, 1971, for example) but rather an attempt to minimize costs of gathering more information.
United States associated with free-enterprise capitalism and racism (Cotton, 1998). However, if
Black families build less wealth than economically comparable White families, then the financial
decisions of Black families are at least partially responsible for sustaining the wealth gap. 3
would imply that both the financial decisions of Black families and the socioeconomic system
should be changed if more evenly distributed wealth is the goal.
In this article, I use longitudinal data from the nationally representative Panel Study of
Income Dynamics (PSID), which allows me to follow and observe the income, wealth, asset
portfolios, and demographic information of the families over time. The data extend from 1984 to
1999, although not all of the analyses cover this entire period. The findings of this study can be
summarized in reference to the three questions as follows.
1. The use of transaction accounts and noncollateralized debt of low-income, low-
wealth families. Using PSID data for 1984, 1989, 1994, and 1999, I find that a lower percentage
of Black families hold transaction accounts than do White families in each income quartile and
each wealth quartile. Logistic and ordinary least squares (OLS) regressions were conducted on
Black and White families that were below median income and wealth in the PSID data. The
regressions show that, after controlling for the impacts of demographic traits, family income, and
family wealth, Black families are less likely to hold transaction accounts than White families; and
for those families that do have transaction accounts, Black families hold lower dollar amounts
than White families. With regard to debt, Black families, compared to White families, owe more
noncollateralized debt in the top two wealth and income quartiles but less in the bottom two
quartiles. The regressions show that, controlling for the effect of the independent variables, low-
income, low-wealth Black families are less likely to owe noncollateralized debt than comparable
The focus here is on changes in wealth instead of accumulated wealth. The latter reflects the activities of
prior periods, not necessarily the wealth management over the current period. Most researchers have
developed models that observe wealth accumulation instead of wealth changes. (See Altonji et al, 2000;
Blau & Graham, 1990; Hurst, Luoh & Stafford, 1998; and Menchik & Jianakoplos, 1997.)
White families; and for families that owe this form of debt, Black families owe less than White
2. The credit management of low- income, low-wealth families. This study also
examines the credit management and wealth management of low-income, low-wealth Black and
White families from 1989 to 1994. Credit management includes the family’s difficulty in paying
bills and its experience with debt consolidation loans, creditors demanding payment, wage
attachments, liens against and repossessions of property, and bankruptcy. In univariate analyses,
the Black families and White families on balance achieve the same success rates at credit
management. However, logistic regressions show that the credit management success of low-
income, low-wealth Black families is equal to or greater than that achieved by the low-income,
low-wealth White families after controlling for demographic and other relevant variables.
3. The wealth management of low- income, low-wealth families. In univariate
comparisons, low-income, low-wealth Black families achieve lower levels of wealth management
success in terms of whether wealth increased, and the size of the increase, than comparable White
families. In the logistic regressions predicting whether wealth increases, I find that the wealth of
Black families is less likely to increase, controlling for the impact of the other independent
variables. However, when labor income, gifts and inheritances are subtracted from the change in
wealth, race becomes statistically insignificant. Thus, differences between Black families and
White families in labor income and, to a lesser extent, gifts and inheritances can explain
differences in wealth management as defined by whether their wealth increases. A second
measure of wealth management is the amount of change in wealth. In univariate comparisons,
the mean change in the wealth of White families is found to exceed that of Black families.
However, in the OLS regressions predicting the amount of change in wealth, race does not matter.
Characteristics such as age, education, number of children, receipt of a monetary gift or
inheritance, and amount of labor income dominate race in predicting the amount of change in
Some implications of these findings are as follows. First, the greater reluctance of low-
income, low-wealth Black families to use checking and savings accounts than low-income, low-
wealth White families remains largely unexplained. The reluctance may be an impediment to the
benefits of participating in financial markets. For example, holding a transaction account is
positively associated with holding noncollateralized debt. Second, the result that Black families
perform as well as or better than White families in credit management, along with the lower
likelihood of Black families owing noncollateralized debt, implies either that Black families have
a lower demand for such debt or that lenders impose prejudicial lending bias against these
families. Third, the finding that low-income, low-wealth Black families are less likely to achieve
wealth increases than low-income, low-wealth White families because of differences in labor
income and, to a lesser degree, gifts and inheritances has implications for wealth disparity
between Black Americans and White Americans. White families--even low-wealth White
families--have a greater propensity to inherit wealth, and studies have found that labor markets
discriminate against Black workers (Holzer & Neumark, 2000). To the extent that gifts,
inheritances and labor income are distributed in the favor of White families, Black families will
generate lower levels of wealth increases than will White families in the United States. Thus the
wealth gap between Black and White families may expand, but this does not result from inferior
wealth management by low-income, low-wealth Black families compared to low-income, low-
wealth White families.
In the remainder of this article, I cover background and data for the study. Next, I discuss
my hypotheses and methodology and provide descriptive statistics. Then, I discuss the results of
the tests, and finally present my conclusions.
Background and Data
The PSID is an annual longitudinal survey of a national sample of U.S. households
conducted by the Survey Research Center of the University of Michigan. The total sample is
representative of the U.S. population when sample weights provided by PSID are used. Black
and low-income families were initially oversampled in the PSID, 4
so I have used weights
throughout this article to adjust for both the differential initial sampling probabilities and the
differential nonresponse that has arisen since the beginning of the study in 1968.5
1984, at 5-year intervals, PSID has gathered wealth data on its panel families.6
This study uses
the amount and components of the families’ wealth for the years 1984, 1989, 1994, and 1999. I
also follow the changes in wealth of families with the same head of household from 1989 to 94.7
In addition to family wealth, PSID provides information on the financial experiences of the
family as reported by the head of household or spouse. In 1996, PSID included a series of
questions on the financial difficulties of the family between 1991 and 1996. This supplement
includes information about whether families have difficulty in paying bills when they were due,
and their experiences with debt consolidation loans, creditors demanding payment, wage
attachments, and liens against and repossessions of property.
Measures of Participation in Financial Markets
Families participate in financial markets through the nature of the assets they hold and
the ways in which they finance their assets and expenditures. Previous studies have examined the
These data contain essentially only Black and White families. Other ethnic groups (Latinos, Asian, and
Native Americans) are not represented in the sample.
Attrition averages about 3% from 1 year to the next. About 60% of the original sample was still being
interviewed in 1998. Researchers who have compared PSID with other data have concluded that these
weighting procedures make the study representative of the nonimmigrant U.S. population (Hill, 1992).
Wealth includes real estate (main home, second home, rental real estate, land contract holdings), cars,
trucks, motor homes, boats, farm or business, stocks, bonds, mutual funds, savings and checking accounts,
money market funds, certificates of deposit, government savings bonds, Treasury bills, IRAs, bond funds,
cash value of life insurance policies, valuable collections for investment purposes, and rights in trust or
estate, less mortgage, credit card, and other debt on such assets. This measure does not include wealth in
the form of private pensions or expected Social Security retirement benefits. I add two observations about
the PSID wealth data here. First, PSID does not capture wealth information on households at the very top
of the wealth distribution. The majority of the measurement problems in PSID occur beyond the 98th
percentile of the wealth distribution, possibly even beyond 99.5%. Juster, Smith. and Stafford (1999)
found that the PSID wealth data for 1989 lined up closely with those from the 1989 Survey of Consumer
Finances through the 99.5 percentile. Of course, a major concern in this study is the experience of African
Americans, who are oversampled by PSID and who are considered to be underrepresented in the top 99.5
percentile of wealth holders in the United States. (Hurst, Luoh, & Stafford, 1998)
Wealth data on families are available for 1999, but the information needed to connect the 1994 families
with their 1999 status is incomplete.
participation of families in purchasing homes. Here, I observe (a) the holding of transaction
accounts at financial institutions--checking and savings accounts- and (b) the use of
noncollateralized debt. Noncollateralized debt, which is reported in the PSID wealth data,
includes credit cards, student loans, medical or legal bills, and personal loans; it does not include
mortgages and automobile debt.
Measures of Credit Management
The measures of credit management follow from special supplements of the 1996 PSID.
The questions in the supplements are paraphrased below:
Between 199 and 1994 did you:
a. Find yourself unable to pay your bills when they were due?
b. Obtain a loan to consolidate or pay off your debts?
c. Have a creditor call or come to see you to demand payment?
d. Have your wages attached or garnisheed by a creditor?
e. Have a lien filed against your property because you could not pay a bill?
f. Have your home, car, or other property repossessed?
Between 1989 and 1994 did you:
g. File for bankruptcy?
The concept of credit management success is paying bills on time and maintaining good
relationships with creditors. These outcomes quantify credit management success along a
discrete continuum, based upon the seven questions posed. “Successful” is denoted as one and
“unsuccessful” as zero in the measures of credit management. The first measure of credit
management is the most conservative:
CRM1 = 1 if No to all of a through g
= 0 if Yes to at least one of a through g
When CRM1 = 1, the family had no problems with paying bills when due, and no problems with
creditors. Families for which CRM2 = 1 consists of families for which CRM1 = 1 plus families
that answered yes to a or b, but no to c through g.
CRM2 = 1 if No to all of c through g
= 0 if Yes to at least one of c through g
Success at CRM2 means no creditor called to press for payment and no legal proceedings
associated with creditors. The family may have had trouble at some point paying bills when they
were due or may have obtained a loan to consolidate or pay off debts. Presumably, the family was
able to manage through such times without difficulty with creditors. Families for which CRM3 =
1 consists of families for which CRM2 = 1 plus those who answered yes to c and no to d through
CRM3 = 1 if No to all of d through g
= 0 if Yes to at least one of d through g
For CRM3, a family is unsuccessful in credit management if creditors pursued such legal
measures as garnishments, liens and property repossession or if bankruptcy was declared.8
Measures of Wealth Management
Two categories of wealth management will be analyzed. The first category is based on
whether the family’s wealth increased between 1989 and 1994. Assume for the moment that
flows in and out of the family unit occur only at the start and end of the period. The wealth at the
end of the period is equal to the wealth at the start, 89W plus the net change in the value of the
wealth claims held over the period, 89W∆ , plus income earned, Y, plus gifts and inheritances
received, G, less expenditures, E.
94 89 89W W W Y G E= + ∆ + + −
W94 = Wealth in 1994
W89 = Wealth in 1989
G = Gifts and inheritances from others from 1989 to 94.
89W∆ = Net change in the value of assets and debts held in 1989.
Y = Income between 1989 and 1994 (except for 89W∆ )
E = Expenditures for consumption, contributions, gifts to others, medical expenses, etc
Instead of considering bankruptcy alone, I added the other measures of severe financial distress to CRM3.
This helps to avoid problems caused by the relationship between the bankruptcy decision and the
bankruptcy laws of the state in which the family resides. It is likely that the other items under CRM3
happen when the severe financial distress associated with bankruptcy occurs, even when bankruptcy is not
from 1989 to 94.
The first category of wealth management contains four measures that make different adjustments
to changes in wealth. The first measure reflects whether or not family wealth changed between
1989 and 1994. Thus WLM1 is defined as
WLM1 = 1 if 94 89 89 0W W W Y G E− = ∆ + + − >
= 0 if 94 89 89 0W W W Y G E− = ∆ + + − ≤
The second measure subtracts out gifts and inheritances from the change in wealth to adjust for
wealth changes due to gifts and inheritances.
WLM2 = 1 if 94 89 89 0W W G W Y E− − = ∆ + − >
= 0 if 94 89 89 0W W G W Y E− − = ∆ + − ≤
The third measure adjusts by subtracting out labor income from WLM1. If labor income is
systematically lower, for example, for Black families (due to current or past racial biases),
subtracting out labor income can provide a clearer picture of the ability to build wealth without
the differences that occur because of different experiences in the labor markets.9
WLM3 = 1 if 94 89 89 0L oW W Y W G Y E− − = ∆ + + − >
= 0 if 94 89 89 0L oW W Y W G Y E− − = ∆ + + − ≤
where Y0 denotes other (nonlabor) income. Finally, the fourth measure takes out both gift or
inheritance inflows and labor income:
WLM4 = 1 if 94 89 89 0L oW W G Y W Y E− − − = ∆ + − >
= 0 if 94 89 89 0L oW W G Y W Y E− − − = ∆ + − ≤
Two comments on this category of wealth management are relevant here. First, it is
likely that funds will move in and out of the categories in W89 within the 5-year period instead of
just at the endpoints. One can define 89W∆ to include such intermediate flows without a loss of
relevance for the measures here. Second, all of the measures in wealth management are in current
The degree to which Black workers receive lower wages because of racial discrimination is a topic of
ongoing controversy. See Holzer and Neumark (2000).
dollars. Current dollar measures are the most clear to the family unit as it evaluates its financial
position. The family will translate the changes in current dollar wealth to changes in constant
dollar wealth depending on the change in the cost of the goods and services it expects to
consume. This can differ for each family, and data on these differences are not available.
The second category of wealth management is the dollar change in wealth between 1989
and 1994. As I will discuss below, regression models will specify to what extent Black and
White families differ in the changes in wealth between 1989 and 1994, after controlling for
various traits that also affect wealth accumulation. In particular, the elimination of gifts,
inheritances, and labor income may exclude information that is helpful in understanding their role
in wealth accumulation. These two variables are included in the regression models that predict
the amount of change in family wealth.
Methodology, Hypotheses and Descriptive Statistics
Multivariate regression models are used to examine the relationship between transaction
accounts or noncollateralized debt and race while considering the impact of the other family
traits. For example, the logistic regression model estimates PT, the probability that a family
holds transaction accounts, as a function of the independent variables:
PT = Probability (family holds transaction accounts) = F[Race; Other Independent Variables]
In addition to the head of the family’s race, the independent variables are the head of the
family’s age, education, health, marital status, gender, employment status, the number of children
in the family and other dependents; and whether the head or spouse receives public assistance
(ADC/AFDC, SSI or other welfare payments). The descriptive variables relate to those reported
as of 1989. The gifts or inheritances received refers to the 1989-94 period. Appendix A contains
descriptions of the independent variables. Finally, I include real total family income and family
wealth, both in 1999 dollars. The same model holds for PD, the probability that the family owes
credit cards and other noncollateralized debt. For credit management, the logistic regression
model estimates PC, the probability that a family achieved success in credit management (e.g.,
CRM1 = 1) as a function of the dependent variables:
PC = Probability (CRMi = 1) = F[Race; Other Independent Variables]
Here, I follow the experience of each of the families from 1989 to 1994. In addition to
the independent variables mentioned above, I also included whether the head or spouse (for
married couples) received a gift or inheritance during the period studied, and the education level
and health status of the spouse. The model and independent variables used for wealth
management are similar to those of credit management. The probability, PW, associated with the
discrete measures WLM1-WLM4 is a function of the independent variables in the logistic
PW = Probability (WLMi = 1) = F[Race; Other Independent Variables]
The treatment of gifts and inheritances in WLM1–WLM4 is different from that in the credit
management tests. The receipt of gifts and inheritances is omitted from the independent variables
in the regressions for WLM1–WLM4, because the dollar amount of gifts and inheritances is a
component of WLM1–WLM4. In the analyses of the magnitude of the changes in wealth, the
following relationships are assumed:
94 89W W− = F[Race; Other Independent Variables]
where W represents family wealth. OLS regressions are used to determine the impact of race and
other variables on (a) the dollar amount of transaction accounts held by the family, (b) the dollar
amount of credit card and other noncollateralized debt, and (c) the change in the wealth of the
sample families. With regard to the change in wealth, the total labor income of the head of
family and spouse (if any) from 1989 to 1994 and the starting wealth (wealth in 1989) are
included as independent variables. The regressions for both categories of wealth management
include CRM1 as an independent variable. This provides an analysis of the association between
credit management and wealth management. Finally, I include a dummy variable indicating
whether the family holds a transaction account (Yes = 1, No = 0) in the analyses of
noncollateralized debt, credit management and wealth management. This allows me to pursue the
issue of how holding a transaction account is associated with these areas of family financial
It is expected that, controlling for demographic variables, the participation of low-
income, low-wealth Black and White families in financial markets are not different. As noted
earlier, the analyses focus on transaction accounts and noncollateralized debt. Thus, the specific
hypothesis is that, controlling for demographic variables, Black and White families do not differ
in the holding of transaction accounts and the use of noncollateralized debt.
With regard to credit management, I hypothesize that credit management is positively
related to age and education. Age and education should increase accumulated knowledge about
effective short-term money management. Married couples are expected to achieve more success
in credit management than families headed by single people. The former potentially have two
people discussing and participating in money matters. Several of the independent variables
describe possible positive or negative shocks to the family’s budget. Poor health is expected to
have a negative impact on credit management.10
The receipt of a gift or inheritance and the receipt
of public assistance are expected to have positive impacts on credit management. The number of
children and the number of dependents outside of the family are expected to have negative
impacts on credit management.
Based on previous studies, I hypothesize that homeownership has a negative impact on
credit management (Hurst & Stafford, 1998). Mortgage payments reduce budget flexibility, and
some families overcommit funds in purchasing a home. With regard to work status, I expect that
being unemployed relative to self-employment, earning a wage or salary both have a negative
Smith (1995) found that healthier households are wealthier households. The direction of causality is
tricky. The relationship between health and credit management is examined here.
impact on credit management. Self-employment has been found to increase wealth at a faster rate
over time than wage or salary status (Bradford, 2000b; Quadrini, 1999), and thus funds to manage
the family’s finances should be more plentiful. Retirement status is expected to have positive
effect, compared to self-employment. Retirees presumably have more time to manage finances,
and typically their budgets are simpler and (except for health) more predictable.
The relationship between a family’s credit management and accumulated wealth is
analyzed by including wealth at the start—family wealth in 1989. I hypothesize that higher
wealth has a positive impact on credit management. At lower wealth levels, the family has less to
draw on to make payments when unforeseen outflows occur. No difference is expected between
single female- and single male-headed families in credit management. A major focus of this
study is the financial management of Black families compared to White families. I hypothesize
that race has no statistically significant impact on credit management after considering the impact
of the other variables.
Prior studies have focused on predicting wealth accumulation instead of changes in
wealth. This is because most sources of wealth data do not follow the same families over an
extended time period. The relationships between wealth management and the independent
variables are hypothesized to be the same as those for credit management except for two items.
First, unlike the negative relationship hypothesized between credit management and
homeownership, a positive relationship is expected between wealth management and
homeownership. The family will have a greater motivation to own a home as its wealth
increases, for tax and investment purposes as well as the mental satisfaction of homeownership.
Second, it is expected that the coefficient on race is negative for WLM1 but declines in influence
going from WLM1 to WLM4. WLM1 reflects whether the wealth of the family increased. The
differences between Black families and White families in the probability of wealth increases are
expected to be associated with differences in gifts, inheritances, and labor income. WLM2
subtracts out gifts and inheritances, WLM3 subtracts out labor income, and WLM4 subtracts out
both. To the extent that the probabilities of increases in wealth are similar for Black families and
White families except for the differences in gifts, inheritances, and labor income, then the race
variable will become less influential as one moves from WLM1 to WLM4. Finally, the dollar
amount of gifts and inheritances and labor income are included as independent variables in the
regressions that predict the dollar amount of changes in wealth. It is expected that both of these
variables have a positive impact on changes in wealth over the studied-period.
Appendix B shows the median income and wealth of the PSID families in 1984, 1989,
1994, and 1999 and the percentages of families below these medians by race and family type
(married couple, single male, and single female). Table 1 contains various statistics on the
families that were below the median income and median wealth in 1984, 1989, 1994, and 1999.
Each year is independent: A given family may be included in only 1 year of the data if it was
below median income and wealth in only 1 of the 4 years. Table 1 shows that the heads of the
Black families and White families were about the same age (40.6 years vs. 40.2 years) and the
heads of Black families had less education. For example, 40% of the heads of Black families had
less than a high school education, compared to 29% for the heads of White families. Six percent
of the heads of Black families had a college degree, compared to 15% of the heads of White
families. A lower percentage of Black families owned their own homes (24% vs. 30%) and a
higher proportion reported fair or poor health (28% vs. 21%). Black families had more children
(1.0 vs. 0.7) than did the White families. The income and wealth data show that the mean
incomes and wealth of the Black families were less than those of the White families. This same
relationship holds for each family category: White married couples had higher incomes and
wealth than Black married couples, etc. These figures combine the real incomes and wealth over
all of the 4 years. The same relationships for income and wealth hold at each of the 4 years.
( Table 1 goes here)
Results of the Tests
For many families, deposit accounts at financial institutions are the means through which
most of their large transactions occur and, correspondingly, the tools through which the family
interfaces with financial markets. Table 2 compares the patterns of ownership of transaction
accounts for all households and separately for Black families and White families. These figures
combine the results for 1984, 1989, 1994, and 1999. The table partitions transaction accounts and
noncollateralized debt by both wealth quartile and income quartile. Table 2 shows that for both
wealth and income, White families use transaction accounts more than Black families in the same
quartile. In addition, the lower the economic status, the greater the difference between Black
families and White families in the proportion of families holding transaction accounts. For
example, the difference in the number of families that hold transaction accounts is less than 10%
for the top wealth quartile, but the difference for the bottom wealth quartile is 38%. Overall, 48%
of Black families hold transaction accounts, and 86% of White families do.
(Table 2 goes here)
These univariate comparisons of transaction accounts do not control for age, education,
income, etc. I want to disentangle the relationships that the independent variables have with
holding transaction accounts, so that a clearer picture of the differences between Black families
and White families may emerge. Because this study focuses on Black and White families in the
lower economic status, in each of the four years examined, I selected those families with below
median income and below median wealth. Table 3 shows the logistic regressions and OLS
regressions on the odds of holding transaction accounts and the dollar amount of transaction
accounts, respectively, for families both below median income and below median wealth.
The logistic regression of combined Black and White families shows that, after
controlling for all of the other variables, Black families are less likely to hold transaction
accounts than are White families. The separate logistic regressions show that, for both Black and
White families, age has a nonlinear relationship with holding transaction accounts, more
education increases the odds of holding a transaction account, married couples are more likely to
hold transaction accounts than single males (the reference group), families with public assistance
are less likely to hold transaction accounts, and greater income and greater wealth increase the
odds of holding a transaction account. The unemployed are less likely to hold a transaction
account than those self-employed. Differences emerge in that, for White families, owning a
home and good health are positively associated with holding a checking or savings account,
whereas these relationships are negative for Black families. The White self-employed are more
likely to hold a transaction account than White workers and retirees, but the differences between
the Black self-employed and Black workers and retirees are not statistically significant.
Table 3 also includes OLS regressions of the dollars held in transaction accounts for
those families that have such accounts. The combined regression results show that Black families
hold $451 less than White families after controlling for the impact of the other variables. A
major difference between the Black and White families in the OLS regression coefficients is that
for White families the amount held in transaction accounts increases with education, but
education does not positively affect the size of the transaction accounts of Black families. Black
married couples and single female heads hold less than Black single males; category of the
household head has no statistical significance among White families.
(Table 3 goes here)
As discussed earlier, PSID information on the families’ financial difficulties between
1991 and 94 is utilized for this study. Data were also gathered in a separate part of the 1996
PSID on the two most recent bankruptcy filings, with details about the reasons and effects of the
proceedings. Information on any bankruptcy that occurred between the 1989 and 94 is used in
this study. Based upon the required information, included in the study are data on 1,951 families
below the sample 1989 medians in income and wealth that had the same head from 1989 to 9411
and for which data on credit management and wealth management are available. There were 831
families with a White head of household and 1,120 families with a Black head of household.
Table 4 contains the summary statistics on the 1,951 families. The first two columns
describe all of the White families and all of the Black families, respectively. The next six
columns compare White and Black families by category: married couples, single males and single
females. The credit performance of the Black families overall is comparable to that of the White
families. Although White families report a 79% success rate for CRM1 compared to 77% for
Black families, Black families report higher success rates in CRM2 (89% vs. 90%) and CRM3
(96% vs. 97%). The comparability of performance for Black and White families also holds by
family category. In terms of wealth management, for WLM1 and WLM2, the success rates of
Black families are lower than those of White families. For WLM3 and WLM4, the success rates
of the Black families are slightly higher than those of the White families. The same relationships
hold in the comparison between Black married couples and White married couples. But on
balance, Black single males perform more favorably in comparison to White single males, and
Black single females perform less favorably then White single females. Similar to the
relationships in Table 1, Table 4 also shows that the mean starting wealth and labor income of the
Black families are less than those of the White families, and this also holds for each category of
Black family in comparison to its corresponding White family. In addition, the mean change in
wealth of the Black families is less than that of the White families, and this relationship also
holds for each family category, with the largest disparity between the advantage of single White
females over single Black females. A smaller fraction of Black families report gift and
inheritance income than do White families, and this relationship holds for each category of
The requirement that the same person heads the family is a way of identifying the same family over time
and linking single years of data. The requirement also connects the credit management outcomes to the
same decision makers. Of course, the results then refer to families with the same head of household.
family. Not shown in the table is that, of those families reporting gifts and inheritances, the mean
for White families is $33,652, and the mean for Black families is $23,917.
(Table 4 goes here)
Table 5 contains the results of the logistic regressions for credit management.
The dependent variables in the credit management regressions are CRM1, CRM2, and CRM3.
The credit management regressions show that, after considering the impact of the other variables,
race is not statistically significant in predicting credit management success if credit management
is defined as CRM1. However, Black is positive and statistically significant for CRM2 and
CRM3, indicating that Black families experience more success in these measures after
considering the impacts of the other variables.
In these regressions, the single male and single female categories are negatively related to
credit management success in CRM1 and CRM2, and are positively related for CRM3. The
married couple category is the reference category, implying that the credit management success
of married couples compared to single heads of household differs across the measures of success.
Credit management success is positively related to age, but the negative and statistically
significant coefficient for age squared in CRM1 and CRM2 indicates a nonlinear relationship
between credit management and age. The reference category for education is less than high
school completion. Education also has an inconsistent impact on credit management. Although
education is positively related to credit management success for CRM1, the relationship is less
consistent for CRM2 and CRM3. Owning a home has a negative impact on credit management,
indicating that the size and inflexibility of mortgage payments and real estate taxes increase
financial pressures on families. Better health for the head and better health for the spouse are
positively related to credit management success. As hypothesized, more children and other
dependents negatively impact credit management. Receiving public assistance is positively
related to credit management for CRM1 and CRM2 but negatively related for CRM3.
Self-employment is the reference category in the variables that express work status. For
CRM1 and CRM2, self-employment is positively related to credit management success,
compared to retirement and unemployment, and has no differential impact compared to the
worker category. For CRM3, however, the negative coefficient for the three categories implies
that the self-employed have more severe financial problems such as liens and bankruptcy than
workers, retirees, and the unemployed. The impact of gifts and inheritances differs among the
measures of credit management. For CRM1, the impact is negative, but for CRM2 and CRM3,
the impact is positive. The coefficient for CRM3 is not statistically significant. The negative
impact for CRM1 may reflect family heads who had difficulty in paying bills (thus unsuccessful
in CRM1) and then requested and received funds from relatives. The receipt of those funds
enabled the families to achieve relatively better outcomes in CRM2 and CRM3. The relationship
between the spouse’s education and credit management is inconsistent for CRM1 and CRM3, but
for CRM2 credit management success is positively associated with the spouse’s education. The
association between income and credit management success is not statistically significant for
CRM1 but is positive and statistically significant for CRM2 and CRM3. More income enables
the family to avoid serious financial difficulty. With regard to the association between credit
management and accumulated wealth, as the level of family wealth increases, the odds of credit
management success improve. Credit management success increases with the family’s wealth.
Finally, holding a transaction account has a positive impact on credit management for CRM1 but
is not statistically significant for CRM2 and CRM3.
(Table 5 goes here)
This study has observed the credit management experience of Black and White families
between 1989 and 1994, and has found that race does not matter in predicting credit management
success. How do these findings match with the actual amount of noncollateralized debt owed by
Black and White families? In other words, is the amount of noncollateralized debt provided to
Black and White families also race neutral when controlling for independent variables? The
bottom section of Table 2 compares noncollateralized debt of Black and White families by
income quartile and wealth quartile. The results for consumer debt differ from those of
transaction accounts. First, for both wealth and income, in the top 2 quartiles, a higher proportion
of Black families owe this form of debt than do White families. In the bottom 2 quartiles, a lower
proportion of Black families borrow than do White families, and the gap is widest at the lowest
wealth quartile. At the lowest wealth quartile, 63% of the White families owe noncollateralized
debt, versus 40% for Black families. This difference reflects that low income and low wealth
families do not fully overlap and that wealth and income may have a different correlation with
Table 6 reports the logistic and OLS regressions that predict whether the family owes
noncollateralized debt and the dollar amount of this debt, respectively. The logistic regression
shows that Black families are less likely to hold noncollateralized debt than White families after
controlling for the other variables. The results also show that holding a transaction account is
positively related to holding noncollateralized debt. In addition, noncollateralized debt has a
nonlinear relationship with age because the coefficients for age and age squared are statistically
significant. The likelihood of owing noncollateralized debt increases with education. Married
couples and single females are more likely to owe this form of debt than single males. The use of
noncollateralized debt is negatively related to public assistance status and the number of children.
Homeownership is positively associated with having noncollateralized debt. Real income is
positively associated with owing noncollateralized debt, but wealth is negatively associated. The
individual race logistic regressions show that families of the Black self-employed owe less
noncollateralized debt than Black families in the other categories; the reverse is true for White
Table 6 also reports the results of regressions predicting the dollar amount of
noncollateralized debt families owed. The OLS regressions show that among families owing
noncollateralized debt, Black families hold $1,291 less than do White families when the impact
of the independent variables are controlled for. The results also show that the amount of
noncollateralized debt is positively associated with holding a transaction account. Age and
education are positively related to the amount of noncollateralized debt for both Black and White
families. Marriage and homeownership are positively related to families’ amount of
noncollateralized debt. Public assistance is negatively associated with noncollateralized debt for
White families, but the relationship is not statistically significant for Black families. Real income
is positively related to debt, but wealth is negatively related to the amount of debt. Although the
self-employed borrow more than the other work categories for White families, this relationship
does not hold for Black families.
(Table 6 goes here)
Table 7 contains the results of the logistic regressions utilizing the discrete measures of
wealth management: WLM1–WLM4. As hypothesized, the coefficient reflecting race is
negative and statistically significant for WLM1, remains negative and statistically significant but
is smaller in absolute size for WLM2, and is statistically insignificant for WLM3 and WLM4.
Thus, the differences between Black families and White families in the probability of increasing
wealth dissipate when gifts, inheritances, and labor income are removed. The key equalizer
between Black families and White families is eliminating of the differences produced by labor
income and its effect on wealth changes. This is obvious when comparing WLM1 to WLM3.
Holding a transaction account has a positive relationship with WLM1 and WLM2, but it
is negatively associated with WLM3 and WLM4. The latter two adjust for labor income,
demonstrating that a transaction account is not positively related to wealth management when the
impact of labor income is controlled for. Public assistance status is positively associated with
each of the measures of wealth management. The coefficients for single male and single female
are negative and statistically significant for WLM1 and WLM2, but they are positive and
statistically significant for WLM3 and WLM4. Relative to married couples, single males and
females experience lower increases in wealth before and after adjusting for gifts and inheritances.
The positive coefficients for WLM3 and WLM4 indicate that single males and females, in
comparison to married couples, achieve higher levels of nonlabor income plus gains on wealth
holdings less expenditures.
(Table 7 goes here)
Table 7 also shows that age has a small negative impact on wealth management for
WLM1 and WLM2 and a small positive impact on wealth management for WLM3 and WLM4.
Age squared and age have the same sign throughout the models. The implication is that age has a
small impact on wealth management but interpretation of the impact is not clear. Wealth
management success does not consistently increase with education and is negative for WLM4.
As expected, owning a home has a positive relationship with wealth management, indicating that
the blend of investment gain, tax benefit, and mental satisfaction of homeownership is positively
related to increases in wealth. Better health is negatively related to wealth management success
for the head of household, but good health for the spouse is positively related to wealth
management success. This may indicate that better health may reduce the precautionary motive
for accumulating wealth for the household’s head. More children is inconsistent with wealth
management success, but other dependents have a positive association with all measures of
wealth management success. The regression coefficient for the bottom wealth quartile is positive
and statistically significant. This indicates that those in the lowest wealth quartile have a greater
probability of increasing wealth. Self-employment is the reference for the work category. For
WLM1 and WLM2, the employee status has an advantage over self-employment in terms of the
probability of increasing wealth. Of course, if labor income is removed, the advantage of
employee status over self-employment becomes a disadvantage, as indicated in WLM3 and
WLM4. As expected, retirement and unemployment have a negative impact on wealth
management compared to self-employment; the two have a positive impact on wealth
management when labor income is removed.
Nevertheless, excluding differences between families in terms of gifts, inheritances and
labor income may omit information that is helpful in understanding their role in wealth
accumulation. Thus, Table 8 reports the results of four OLS regression models using change in
wealth as the dependent variable for measuring wealth management. All of the models include
gifts and inheritances as an independent variable. Model 1 excludes both total labor income and
starting wealth from the independent variables. Model 2 adds the total labor income of the family
head (and spouse, if any) from 1989 to 1994 as an independent variable. Model 3 excludes labor
income but includes family wealth in 1989 as an independent variable. Model 4 includes both
labor income and 1989 wealth as independent variables. The F-statistics show that each of the
regressions in Table 4 is statistically significant at the 0.01 level, and the adjusted coefficients of
determination range from 0.08 to 0.10.
None of the models in Table 8 indicates that race matters in predicting changes in wealth.
All of the regression coefficients for Black are positive but not statistically significant. Low-
income, low-wealth Black families build wealth at the same pace as low-income, low-wealth
White families after controlling for the other variables. Holding a transaction account is not
statistically significant when labor income is included in the model. Education is positively
related to changes in wealth, but the coefficients for age are typically not statistically significant.
The number of children has a negative effect on the change in wealth. Owning a home has a
positive relationship with changes in wealth. Being in better health for the head has a
consistently negative impact on the change in wealth, although better health for the spouse is not
statistically significant. The number of dependents outside the family unit is positively associated
with change in wealth. Public assistance is not statistically significant in predicting the dollar
amount of the change in wealth. The work status dummies are not statistically significant,
meaning that the change in wealth of the low-income, low-wealth self-employed is not different
from the other employment categories when the impacts of other variables are controlled for. As
expected, labor income is positively associated with wealth changes, although starting wealth is
not statistically significant. The low starting wealth of these families has less impact on wealth
changes during the period studied than does their labor income.
(Table 8 goes here)
Conclusions and Discussion
This study uses PSID data covering 1984 through 1999 to examine various aspects of the
financial management of low-income, low-wealth families. First, the greater reluctance of low-
income, low-wealth Black families to use checking and savings accounts than low-income, low-
wealth White families is only partially explained by regression models that consider a wide array
of demographic variables. For many families, transaction accounts are the tools through which
they begin to successfully interface with financial markets. Thus, related topics for future
research include: What are the impediments for low-income, low-wealth Black families to hold
transaction accounts at financial institutions, and to what extent are there impediments to other
minorities of lower economic status? Second, the credit management success of low-income,
low-wealth Black families is equal to or greater than that of low-income, low-wealth White
families, based on three measures. Black families, however, owe less credit card and other
noncollateralized debt than White families, implying either that Black families have a lower
demand for such debt or that lenders are biased against them. The extent of bias against Black
borrowers is an ongoing area of research, and the results here show that the issue is relevant when
comparing the credit card and other noncollateralized debt of low-income, low-wealth Black and
White families. Third, wealth increases faster among poor White families faster than among poor
Black families because White families have more income from labor, gifts, and inheritances than
do Black families. This points to related topics for future research: To what extent does this
relationship hold for Black and White families across the income and wealth spectrum, and what
are the implications for the concentration of wealth? These results show that the lower wealth
gains of poor Black families are not due to inferior wealth management. If these relationships
hold generally for Black families compared to White families, the disproportionate concentration
of wealth among White families in the United States will continue for the foreseeable future.
Altonji, J., Hayashi, F. & Kotlikoff, L. (1997). Parental altruism and inter vivos transfers:
theory and evidence. Journal of Political Economy, 105, 1121-1166
Altonji, J., U. Doraszelski & Segal, L., (2000). Black/white differences in wealth.
Economic Perspectives, Federal Reserve Bank of Chicago, 24, 30-50.
Ando, F. (1988). Capital issues and minority-owned businesses. Review of Black Political
Economy, 16, 77-109.
Bates, T. (1997). Unequal access: financial institution lending to black- and white-owned
small business start-ups. Journal of Urban Affairs, 19, 487-495.
Bates, T. & Bradford, W. (1992). Factors affecting new firm success and their use in
venture capital financing. Journal of Small Business Finance 2, 23-38.
Becker, G. (1971). The Economics of Discrimination. Chicago: University of Chicago
Blanchflower, D., Levine, P. & Zimmerman, D. (1998). Discrimination in the small
business credit market. (Working Paper No. 6840). Cambridge, Massachusetts:
National Bureau of Economic Research.
Blau, F. & Graham, J. (1990). Black-White differences in wealth and asset composition.
Quarterly Journal of Economics, 105, 331-339.
Bradford, W. (2000a). Black family wealth in 2000. In State of the Black Economy, 2000
(pp.103-145). New York: National Urban League.
Bradford, W. (2000b). The wealth dynamics of self-employment for Black and White
families, Unpublished Working Paper, University of Washington.
Cavalluzzo, K. & Cavalluzzo, L. (1998). Market structure and discrimination: the case of
small businesses. Journal of Money, Credit and Banking. 30, 771-792.
Cotton, J. (1998). On the permanence or impermanence of Black-White economic
inequality. Review of Black Political Economy. 26, 47-56.
Hill, M. (1992). The panel study of income dynamics. Newbury Park, CA: Sage.
Holzer, H. & Neumark, D. (2000). Assessing affirmative action. Journal of Economic
Literature. 38, 483-568.
Hurst, E., Luoh, M. & Stafford, F. (1998). The Wealth Dynamics of American
Families, 1984-1994 Brookings Papers on Economic Activities 1998:1, 267-329.
Hurst, Erick & Stafford, F. (1998). Grasshoppers and ants: mortgage refinancing and
bankruptcy. Unpublished Working Paper, University of Michigan.
Juster, F., J. Smith, J. & Stafford, F. (1999). The measurement and structure of household
wealth. Labor Economics. 6, 253-275.
Ladd, H. (1998). Evidence on discrimination in mortgage lending. Journal of Economic
Perspectives. 12, 41-62.
Menchik, P. & Jianakoplos, N. (1997). Black-White wealth inequality: is inheritance the
reason? Economic Inquiry. 35, 428-442.
Oliver, M. & Shapiro, T. (1990). Wealth of a nation—at least one-third of households are
asset poor. American Journal of Economics and Sociology. 49, 129-150.
Phelps, E. (1972). The statistical theory of racism and sexism. American Economic
Review. 62, 659-61.
Quadrini, V. (1999). The importance of entrepreneurship for wealth concentration and
mobility .Review of Income and Wealth. 45, 1-19.
Smith, J. (1995). Racial and ethnic differences in wealth in the health and retirement
study Journal of Human Resources. 30, s158-s183.
Wolff, E. (1998). Recent trends in the size distribution of household wealth. Journal of
Economic Perspectives. 12, 131-150.
Yinger, J. (1998). Evidence on discrimination in consumer markets. Journal of Economic
Perspectives. 12, 23-40.
Appendix A: Description of Variables
Race: Black head of household = 1; White head of household = 0
Education of household head (spouse):
Less than High School
High School Only
High School Plus College (No Degree)
Own home in 1989: yes = 1; no = 0
Health of head (spouse): excellent, very good or good = 1; fair or poor = 0
Type of Household:
Male single, divorced, or separated
Female single, divorced, or separated
Married couple not separated
Inheritances or gifts received, 1989-94: dollar amount
Children younger than 18 years old in residence: number
Number of dependents outside of family: number
Employment category: worker, retired, self-employed or unemployed
Public Assistance: If the household head or spouse received ADC/AFDC, SSI, or other welfare.
yes = 1, no = 0
Appendix B: Various Statistics on PSID Family Income and Wealth
Median Income ($) Median Wealth ($)
1984 1989 1994 1999 1984 1989 1994 1999
22,000 27,590 32,899 41,300 31,350 38,000 49,200 58,185
% of Families
Married Couples 44.8 44.1 41.8 44.0 65.9 62.4 67.4 71.9
Single Males 79.3 81.1 78.9 77.4 91.8 90.1 87.9 89.9
Single Females 91.1 88.6 90.5 88.9 87.3 87.5 83.5 87.9
Group 73.5 74.6 75.5 73.5 80.7 80.9 78.5 83.8
Married Couples 28.3 28.1 27.3 27.1 32.4 30.0 32.3 33.7
Single Males 60.8 57.5 59.5 62.4 67.9 67.6 65.8 66.9
Single Females 77.6 74.1 75.3 75.3 61.7 60.8 58.7 58.4
Group 46.5 46.0 45.9 46.3 45.5 45.0 45.3 46.3
Dollars in Current Dollars. Sample weights are used in calculations.
SOURCE: PSID Core and Supplemental Wealth Files
Descriptive Statistics for Low-Income, Low-Wealth Families
Combined Data for 1984, 1989, 1994 and 1999
White White White Black Black Black
All White All Black Married Single Single Married Single Single
Families Families Couples Males Females Couples Males Females
Number of Families 4,361 5,858 1,787 1,034 1,540 1,396 1,283 3,179
Age (Years) 40.2 40.6 40.7 34.6 43.3 44.3 36.8 41.1
Standard deviation 76.9 36.4 64.1 64.6 91.3 31.7 29.9 39.7
Age, Head (%)
Less than 35 years 48.2 42.7 45.8 62.6 41.0 33.7 52.2 41.5
35 through 54 years 30.1 37.7 31.9 28.2 30.0 37.0 38.6 37.6
More than 54 years 31.7 19.6 22.3 9.2 29.0 29.4 9.2 20.9
Education, Head (%):
Less than high schol 29.1 40.4 34.5 21.9 29.8 49.9 33.2 40.5
High school only 36.6 34.6 40.3 33.8 35.7 37.3 33.3 34.4
College, no degree 19.7 19.2 15.7 25.5 18.9 9.5 26.2 19.2
College degree 14.6 5.8 9.5 18.9 15.6 3.3 7.3 6.0
Work Status, Head (%):
Retiree 15.2 16.6 18.0 10.0 16.4 26.2 16.1 14.4
Worker 65.3 57.5 65.4 72.2 61.0 60.8 62.5 54.8
Self-Employed 6.4 1.9 9.0 7.3 4.1 2.1 3.4 1.3
Unemployed 13.1 24.0 7.7 10.5 18.5 10.9 18.0 29.5
Number of Children 0.62 0.97 1.19 0.15 0.52 1.29 0.25 1.14
Standard deviation 5.06 3.25 5.07 2.87 5.10 2.85 1.81 3.59
Number of Dependents Outside 0.17 0.18 0.17 0.27 0.13 0.12 0.49 0.10
Standard deviation 3.52 1.52 3.26 3.96 3.56 0.95 2.43 1.12
Health Fair or Poor (%) 20.9 28.0 21.2 16.8 23.2 35.9 24.5 27.9
Own Home (%) 29.5 24.1 50.7 14.7 24.0 44.5 15.9 21.8
Means (1999 dollars):
Family Income 21,103 16,651 25,003 20,582 18,712 22,503 15,855 15,548
Family Wealth 9,889 7,047 12,668 7,163 9,654 13,072 7,168 5,482
Source: Author's calculations based on data from the PSID. Proportions may not add to 1.0 because of rounding.
Percentage of Familiesa
Holding Transaction Accounts and Noncollateralized Debt
by Wealth Quartile and Income Quartile
1. Transaction Accounts
Quartile 1 Quartile 2 Quartile 3 Quartile 4 All Families
Black White Black White Black White Black White Black White
By Family Wealth
% of Families in this Group 3.2 28.5 14.9 26.5 29.8 24.1 52.1 20.9 100.0 100.0
% of Families in this Group
with Transaction Accounts 87.6 97.1 72.9 92.5 61.3 86.0 30.1 63.9 47.6 86.3
By Family Income
% of Families in this Group 7.7 28.7 16.4 27.0 27.8 24.3 48.0 20.0
% of Families in this Group
with Transaction Accounts 88.5 96.3 76.4 91.9 58.3 85.4 25.0 65.4
2. Noncollateralized Debt
Quartile 1 Quartile 2 Quartile 3 Quartile 4 All Families
Black White Black White Black White Black White Black White
By Family Wealth
% of Families in this Group 3.2 28.5 14.9 26.5 29.8 24.1 52.1 20.9 100.0 100.0
% of Families in this Group
with Noncollateralized Debt 51.9 39.1 55.9 52.1 49.0 58.6 39.8 63.3 45.3 52.3
By Family Income
% of Families in this Group 7.7 28.7 16.4 27.0 27.8 24.3 48.0 20.0
% of Families in this Group
with Noncollateralized Debt 68.2 56.1 61.2 58.2 49.4 51.9 33.8 39.5
N = 10,219
Note. The entire sample was separated into quartile rankings by either wealth or income and by Black family or
White family. The sample includes observations of individual family wealth or income in 1984, 1989, 1994, and
1999. SOURCE: PSID core and supplemental files, and the author's calculations. a
n = 10,219.
Regression Models Explaining the Holding of Transaction Accounts at Financial Institutions by
Low-Income, Low-Wealth Families
1. Logistic Regressions of Whether Family Holds a Transaction Account
Combined Families Black Families White Families
Independent Variables Coefficient Chi-Square Coefficient Chi-Square Coefficient Chi-Square
Intercept -0.2965 45.6*** -1.4969 187.4*** -0.3437 47.5***
Black -1.1635 6,568.3***
Age 0.0016 7.0*** -0.0077 32.4*** 0.0027 15.5***
Age Squared -0.2368 278.7*** -0.1798 43.9*** -0.2574 235.0***
Education= Less than high school -0.5949 1,449.5*** -0.3405 111.5*** -0.6591 1,343.5***
Education= College, no degree 0.4891 772.9*** 0.3281 99.0*** 0.5618 696.6***
Education= College degree 1.2740 2,530.8*** 1.2413 513.4*** 1.2890 2,009.6***
Married 0.2399 147.0*** 0.0562 1.7 0.2588 133.2***
Single Female 0.4647 785.5*** 0.4233 166.3*** 0.4815 618.2***
Number of Children -0.1786 692.9*** -0.2380 378.0*** -0.1494 321.9***
Public Assistance -0.6961 1,132.7*** -0.4537 136.7*** -0.7741 981.5***
Health Excellent Or Good 0.2505 216.6*** 0.0042 0.1 0.3563 310.8***
Own Home 0.2239 180.5*** -0.2122 34.9*** 0.3391 316.4***
Real Income/$10,000 0.4938 5,028.2*** 0.6703 2,426.5*** 0.4299 2,768.0***
Real Wealth/$10,000 0.0580 261.9*** 0.1947 359.5*** 0.0401 150.4***
Retired 0.1028 8.5*** 0.3716 14.5*** 0.1594 16.7***
Worker -0.2380 63.5*** 0.1410 2.5 -0.2665 68.9***
Unemployed -0.6756 417.1*** -0.6424 45.5*** -0.5915 267.1***
Year=1984 0.1831 116.3*** 0.0964 8.3*** 0.2236 126.9***
Year=1994 -0.5133 924.4*** -0.6410 363.9*** -0.4811 596.2***
Year=1999 0.0769 15.9*** 0.1647 20.9*** 0.0531 5.4**
Minus 2 Log L 149,989*** 38,301*** 110,072***
N 10,042 5,560 4,482
2. OLS Regressions of the Dollar Amount of Transaction Accounts
Independent Variables Coefficient t-value Coefficient t-value Coefficient t-value
Intercept -311.7 -0.7 1,515.4 1.7* -926.5 -1.5
Black -451.0 -2.4**
Age 67.5 10.2*** 0.7 0.1 76.4 9.3***
Age Squared -358.6 -2.3** -86.2 -0.4 -391.2 -2.0*
Education= Less than high school -403.6 -2.1** 393.0 1.4 -504.7 -2.1**
Education= College, no degree 477.0 2.7*** -441.0 -1.7* 668.2 3.0***
Education= College degree 959.8 5.0*** 85.6 0.3 1,163.4 4.9***
Married -199.3 -1.0 -1,158.7 -3.4*** -11.7 0.0
Single Female -145.9 -0.9 -690.8 -2.7*** 31.4 0.1
Number of Children -171.7 -2.2** -1.3 0.0 -206.9 -2.0**
Public Assistance -1,318.3 -4.5*** -1,186.9 -3.2*** -1,323.7 -3.4***
Health Excellent Or Good 149.3 0.7 23.4 0.1 242.4 0.9
Own Home -1,397.6 -8.5*** -2,415.8 -9.3*** -1,326.3 -6.5***
Real Income/$10,000 394.8 5.2*** 269.8 2.4** 422.8 4.4***
Real Wealth/$10,000 388.3 16.4*** 1,235.3 18.1*** 350.2 12.5***
Retired -292.0 -0.8 63.1 0.1 -287.7 -0.6
Worker -677.2 -2.3** -577.6 -0.8 -613.3 -1.7*
Unemployed -77.0 -0.2 -484.9 -0.6 23.3 0.1
Year=1984 -440.1 -2.5** -356.9 -1.3 -451.1 -2.0**