Full text of Working Papers (Federal Reserve Bank of Chicago) : Savings of Young Parents, Working Paper 2000-23

View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Federal Reserve Bank of Chicago

Savings of Young Parents

By: Annamaria Lusardi, Ricardo Cossa
and Erin L. Krupka

WP 2000-23

Savings of Young Parents

December 2000

Annamaria Lusardi∗∗
(Dartmouth College)
Ricardo Cossa
(Chicago Partners)
Erin L. Krupka
(Federal Reserve Bank of Chicago)

∗∗

Corresponding author. Address: Department of Economics, Dartmouth College, Rockefeller Hall,
Hanover, NH 03755. Tel: (603) 646-2099. Fax: (603) 646-2122. E-mail:
Annamaria.Lusardi@Dartmouth.edu.

We wish to thank Rob Alessie, Gary Engelhardt, Mike Hurd, Erik Hurst, Arthur Kennickell, Bob Michael,
Randy Olsen, Jim Smith, Jim Walker, an anonymous referee, and participants at the Conference of Early
Results from the National Longitudinal Survey of Youth, 1997 Cohort, for suggestions and comments.
Annamaria Lusardi acknowledges with thanks the financial support of the Spencer Foundation through its
grant to the Harris School in support of the Conference of Early Results from the National Longitudinal
Survey of Youth. The opinions expressed in this paper are those of the authors and do not necessarily
represent the opinions of the Federal Reserve Bank of Chicago or of Chicago Partners. All errors are our
own.

Abstract
In this paper, we examine household savings using data from the National
Longitudinal Survey, Cohort 1997 (NLSY97). This data set provides detailed information
about assets and liabilities of parents with teen-age children and allows researchers to
examine patterns of accumulation at early stages of the life cycle.
In our empirical work, we have first to deal with several problems in measuring
wealth. While many respondents report owning assets and liabilities, they often do not
report their values. This problem is severe, in particular among financial assets. It is also
difficult to devise an appropriate measure of accumulation when examining young
parents, since assets and liabilities display different degrees of liquidity.
To get around the non-response problem, we impute the missing values for assets
and liabilities. This allows us to calculate household wealth for the whole sample. We
examine household wealth holdings by considering several measures of accumulation:
total (non-pension) net worth, financial net worth, and retirement savings. We study their
distribution across different demographic groups and show that many households, in
particular those headed by young parents (younger than 35), minorities, and individuals
with low educational attainment, display very little accumulation. Many have no financial
assets and their total net worth is also low. Housing equity is the main asset in many
household portfolios and often the only asset families own. Overall, there is much
heterogeneity in wealth holdings not only across but also within demographic groups.
This suggests that many factors are at play in shaping the wealth accumulation of parents
with young children.

JEL classification: D31, D91.
Key Words: Wealth, Motives to save; Parents with teen-age children.

1. Introduction
In this paper, we examine household wealth holdings using data from the National
Longitudinal Survey, Cohort 1997 (NLSY97). This survey provides detailed information
about assets and liabilities of parents with teen-age children and allows researchers to
investigate patterns of wealth accumulation at early stages of the life cycle. There are at
least four motives to save that can be relevant for this demographic group. First and most
importantly, parents may be saving for their children’s education and, in particular, for
sending children to college. Second, parents, particularly those at the beginning of their
career, may be saving to insure against shocks to income, such as unemployment, job
loss, and other unexpected events. Third, as the simple life-cycle model predicts, they
may save for their retirement. Fourth, parents may save to buy a house or start a business.
Given the age range of children and their parents, the first two motives should
feature prominently in the data. Children are only a few years away from college and, for
many, the event is imminent. Additionally, a very large proportion of children have
expressed high expectations of completing a college education.1 Many of these parents do
not have long tenures at their jobs and are likely to face much uncertainty about their
earnings.2 Thus, these data allow us to shed some light on how many resources parents
have to buffer shocks to income as well as to meet future financial obligations, such as
those involved in sending children to college.3 We may also gain insights on the
relevance of other motives to save. Even though there has been much development in
financial markets and opportunities for borrowing, a down-payment is required to buy a
house. Similarly, starting capital is often required to become an entrepreneur. Studies
have shown that families overcome these potential liquidity constraints by saving more.4
The analysis in this paper is simply descriptive and aims to highlight some of the major
empirical facts about the patterns of accumulation of parents with teen-age children.
Even though the main objectives of the NLSY97 are not concerned with
household savings, this information can be useful for many empirical studies. First, in
1

See Pemberton and Reynolds (2000).
Empirical studies show that precautionary savings are high among young workers. See the review of this
work in Browning and Lusardi (1996).
3
For a detailed discussion of the motives to save and models of saving behavior, see the survey by
Browning and Lusardi (1996).
2

many studies, it is often necessary to account for household economic status, and income
alone is often not a good proxy for the economic resources available to a family. For
example, current income can be temporarily low, while permanent income of the
household may be high. In addition, families in the same income group may have rather
different wealth holdings. Families with real assets, such as home equity, other real
estates, cars, and other vehicles, may also be able to borrow in periods of low income or
income shocks. This suggests that information on household wealth as well as
composition of household portfolios can enhance empirical analyses using this data set.
Even for researchers interested in savings, there are several advantages in using
the NLSY97. First, few data sets report extensive information on families with teen–age
children, and it is important to study this group of the population. Second, this data set
provides a richness of information on household characteristics (both on parents and on
children), that can prove important in explaining the wide heterogeneity that we observe
in saving behavior. Third, the information about savings is extensive (data is collected on
more than ten asset components and five debt components) and follow-up brackets after
each component of wealth allow researchers to measure wealth holdings with some
accuracy.
Measurement issues are critically important when examining wealth and we
devote extensive discussion to this issue. While many respondents report owning assets
and liabilities, they often fail to report their values. Consequently, we have to deal with
non-responses when constructing household total net worth. Another potential issue
concerns the appropriate measure of accumulation to consider when examining young
and middle-aged parents since assets and liabilities display different degrees of liquidity.
In our empirical analysis, we consider several measures of accumulation: total (nonpension) net worth, financial net worth, and retirement savings. We examine the
distribution of these different measures of wealth as well as of ownership of assets and
liabilities across different demographic groups.
Our major findings are that households, in particular those headed by young
parents (younger than 35), minorities, and individuals with low educational attainment
display very little accumulation. Many families hold little or no financial wealth and also
4

See, among others, Engelhardt (1994, 1996) and Gentry and Hubbard (1998).

their total net worth is low. The most important asset in many portfolios is home equity
and many own little in anything other than their home equity. Overall, there is a great
deal of heterogeneity in wealth holdings and patterns of accumulation vary widely not
just across but also within demographic groups. This suggests that many factors are at
play in explaining the differences in wealth holdings among parents with teen-age
children.
The paper is organized as follows: In section 2, we examine many of the issues
associated with measuring wealth, and we compare data in the NLSY97 with other data
sets. In section 3, we examine the distribution of wealth and the ownership of assets and
liabilities, and we provide a discussion of the main empirical findings. In section 4, we
provide a brief conclusion.

2.1 The National Longitudinal Survey of Youth, 1997 Cohort

The NLSY97 is a nationally representative sample of the U.S. population aged 12
to16 in 1997, and hence born during the years 1980 through 1984. The sample consists of
9,022 respondents from 75,291 pre-identified households in 147 non-overlapping primary
sampling units containing residents age 12 through 16 as of December 31, 1996. Two
samples were drawn. The first was a nationally representative sample of youths born
between 1980 and 1984. Additionally, Black and Hispanic youths for that age group
were oversampled. The sample also included those who usually reside at home, but were
away at school or college and those in hospitals, correction facilities, or other types of
institutions.
The survey is designed to document young adults’ transition from school to work
and to identify defining characteristics of that transition. Thus, it contains extensive
information on respondents’ labor market behavior, educational experiences, and the
respondents’ family and community. In addition to the youth interview, the NLSY97
contains a separate interview conducted with a responding parent. This interview is
designed to provide information about the home environment and detailed parent
characteristics. Potential responding parents were limited to those that lived in the

household. They were selected according to a pre-ordered list.5 The responding parent
was asked extensive questions about personal background information and questions
about the responding youth’s life. The questions most relevant to our studies are those
regarding parent’s wealth.

2.2 The Measurement of Wealth

In the NLSY97, the respondent’s parent is asked to report information on a list of
assets and liabilities aimed to measure household total net worth. Specifically, the
respondent is asked to report information on the following asset components:
1) Housing (distinguished into ranch or farm, mobile home, and house or apartment);
2) Other real estate;
3) Business equity (business partnership or professional practice);
4) Retirement savings (thrift/savings plans, 401(k)s, profit sharing or stock ownership
plans, IRA or Keogh plans, and other types of plans);
5) Educational IRA accounts or other pre-paid tuition savings accounts established to
pay college costs;
6) Checking and saving accounts, money market accounts or funds, accounts held in
investment trusts;
7) Certificates of deposit, government savings bonds, Treasury bills, corporate,
municipal, government or other types of bonds and bills including any CD’s, bonds or
bills held in investment trusts (bonds hereafter);
8) Shares in publicly held corporations or mutual funds, including any stocks or mutual
funds held in investment trusts (stocks hereafter);
9) Cars, vans, trucks and other vehicles including boats or airplanes;
10) Other assets, such as money owed to you by others, the cash value of any whole or
straight life insurance policies, future proceeds from a lawsuit or estate that is being

The order in which the responding youth’s parent was chosen is as follows; Biological mother, biological
father, adoptive mother, adoptive father, stepmother, stepfather, guardian (relative), foster parent (youth has
lived with for 2 years or more), other non-relative (youth has lived with for 2 years of more), mother-figure
(relative), mother-figure (non-relative youth has lived with for 2 years or more), father-figure (non-relative
youth has lived with for 2 years or more).

settled, assets in a trust, annuity, or managed investment accounts, art work, precious
metals, antiques, oil and gas leases, future contracts, royalties or something else;
11) Household furnishings including furniture, major appliances, and home electronic
items.
The respondent is also asked to report information on the following debt components:
1) Mortgage or land contracts on housing;
2) Second mortgages, home equity loans, or any outstanding loans against a home equity
line of credit;
3) Debt owed on vehicles;
4) Loans for children’s educational expenses;
5) Any other debt currently owed, including store bills, credit cards (if respondent
carries a balance), loans obtained through a bank or credit union, margin loans
through a stock broker and other installment loans.

As the previous list shows, the information on assets and liabilities is extensive in
the NLSY97 and it encompasses major components of household wealth. It is important
to note that, with respect to previous NLSY waves, the information has become more
detailed. For example, in the NLSY79 assets and liabilities were aggregated rather
broadly,6 and this could lead to less accurate reports.
To perform the analysis, we consider responses at the household level (a
household can have multiple children interviewed in the survey). There are a total of
7,973 youth respondents in the NLSY97 for which information from a parent interview is
available for a total of 6,113 families. Note that there were no parent interviews for 811
families. We examined the characteristics of these non-respondents and found that they
concentrate among low education groups, Black or Hispanic families, young respondents
and families in the West regions of the country. This suggests that some caution must be
used in interpreting the values for wealth since the selected sample shows some evidence
of selectivity.
The responding parent is first asked whether s/he owns the assets and liabilities
listed above, then to provide a value. The latter refers to the market value, i.e., what the
6

See Engelhardt (1998) and Zagorsky (1999).

respondent would obtain if s/he were to liquidate the asset or liability. One important
feature of these data is that many responding parents do not know the value of their assets
and liabilities or refuse to report a value. While there are only few respondents who,
when questioned about the ownership of assets and liabilities, answered with a ‘refusal’
or ‘do not know,’ a large fraction of respondents were not able or willing to report values
for their assets and liabilities.
Table 1 reports the ownership of assets and liabilities and the fraction of
respondents who refused to report a value or responded they ‘do not know’ the value.7
Note that the majority of non-responses are due to an inability to report a value rather
than a refusal. Additionally, non-responses vary substantially across assets and liabilities.
Non-responses are particularly high for financial assets. For example, the proportion of
‘do not knows’ is high for stocks and for retirement savings. It is also relatively high for
business equity, bonds, and educational IRAs. Undoubtedly, these questions are complex.
In particular, reporting the market value of assets implies some knowledge of current
market quotes. Accurate reports become further complicated when different assets have
to be evaluated and added together. This raises concerns about the accuracy of reports
even when respondents report values.
This is one of the major problems of collecting wealth data and one that is critical
for the correct evaluation of household resources. This problem is common to other data
sets on wealth and was present in previous waves of the NLSY that collected information
about wealth. Smith (1995) compares non-responses about wealth across four different
data sets: the Health and Retirement Study (HRS), the Panel Study on Income Dynamics
(PSID), the Survey of Income and Program Participation (SIPP), and the Survey of
Consumer Finances (SCF). Similar to our findings in the NLSY97, he finds that nonresponses about ownership are very small, usually less than 1%. However, non-responses
about the values of the items owned are pervasive. For example, more than 30% of
respondents do not report the value of stocks or bonds in the HRS. Percentages of nonresponses are high for these assets even in the SCF, a survey specifically designed to
measure household wealth. Other assets as well are affected by non-responses. For

In our empirical work, we consider the cross-sectional sample as well as the supplemental sample (which
oversamples Blacks and Hispanics) and always use household weights.

example, the proportion of respondents that do not report the value of their businesses
range from 37% in the SCF to 24% in the PSID. The fraction of non-responses in
retirement assets, such as IRAs and Keoghs, is approximately 27% in the HRS. Recent
work by Gustman and Steinmeier (1999) shows that respondents are also poorly informed
about their pensions; many do not know the type of pensions they have (a defined benefit
or defined contribution plan) or the benefits associated with it.
Consistent with other surveys,8 in the NLSY97 non-responses are less pervasive
concerning home values (non-responses range from 2% to 7%). Usually, respondents are
not only more willing to report the value of their home, but they also seem well informed
about that value. Alessie, Lusardi and Aldershof (1997) compared self-reports of housing
equity with other micro data sets that collect detailed information on housing and found
that household reports compare well across different sources of data.9 They also compare
well with current market values. This is a useful feature of the data, since housing is one
of the major assets in household portfolios and, as will be shown below, frequently the
only asset people own.
Non-responses may be rather severe in the NLSY97 because these questions were
not asked to the respondent most knowledgeable about the household’s financial
situation. Non-responses may also be affected by the degree of the aggregation of assets
and liabilities. In previous NLSY waves, which collected information about wealth, there
were many assets and liabilities that displayed high non-responses. Engelhardt (1998)
reports that non-responses for stocks and bonds in the NLSY79 range from 16 to 23%.
Non-responses were present, but more limited, on house values and mortgage debt.
Contrary to previous waves, respondents in the NLSY97 who do not report a
value are asked a follow-up question where they have to indicate their best estimate of
the value by picking among a range of values (brackets). Many of the non-respondents
were able, and willing, to report information on the bracketed amounts. In fact, the
percentage of non-responses drops dramatically when the information reported in
brackets is used. This procedure represents an important innovation in the collection of
wealth data and one that is worth emphasizing. A similar procedure had been

8
9

See also Juster and Smith (1997).
See, however, Goodman and Ittner (1992) for a description of the biases in home-owners reports.

implemented in the HRS10 and to a different extent in the PSID and the SCF. Smith
(1995) and Juster and Smith (1997) provide a careful and thorough evaluation of this
procedure and show that it leads to major improvements in the collection and
measurement of wealth data. As the authors report, non-responses are hardly random.
More importantly, estimates of aggregate wealth change considerably when the value of
assets and liabilities reported in brackets are included.11
We have used the information reported in the brackets to impute the value of
assets and liabilities. After this imputation, the proportion of non-responses drops
dramatically.12 We also imputed the values for the remaining non-responses using the
procedure reported in the data appendix. Since there are non-responses for ownership as
well, we also impute ownership, even though it only affects a small percentage of
respondents.
Before defining wealth, in Table 2, we report the conditional means and medians
of all assets and liabilities reported in the NLSY97.13 Many components of wealth, and in
particular many asset components, show a distribution greatly skewed to the right. This is
particularly the case for assets such as stocks, business equity, and retirement savings.
Contrary to previous public releases of NLSY waves, asset and liability values were not
truncated at the top and consequently, we do not have to worry about this problem.14
Overall, there is wide heterogeneity in the holdings of assets and liabilities. It is important
to look at medians in addition to means, since the former may be better representative of
the typical household in the population.

In the HRS, there is a set of unfolding brackets. See Juster and Smith (1997) for detail.
See, also, Hurd et. al. (1997).
12
We are able to impute the value of housing for almost every respondent and ‘non-responses’ for other
homes and mortgages are also reduced substantially. However, ‘non-responses’ are still sizable for stocks
and retirement savings. Approximately 17% of stock-owners have not reported any value, either explicitly
or in bracket amounts, and 13% of respondents reporting retirement savings have not indicated the amounts
invested in these assets. Non-responses are also still present in bonds and business equity; 14.5% of
business owners and 14% of bond-holders have not reported any values in brackets. Note, however, that
percentages of bond and stock holders as well as entrepreneurs in this sample of young families is relatively
small. See the data appendix for a detailed description of the procedure used in the imputation.
13
Figures differ between Table 1 and Table 2 since in the latter table we impute the missing data for assets
and liabilities (ownership as well as values).
14
For example, in the NLSY79 the value of the house was top-coded at $150,000; farms, businesses and
other real estate assets at $500,00, and stocks and bonds at $100,000. See Engelhardt (1998) and Zagorsky
(1999) for details.
11

It is clear that the house is one of the major assets in many household portfolios.
The conditional median of house values is $95,000 and the conditional mean is
approximately $125,000. However, the large majority of home-owners have a mortgage
and the median and mean housing equity in this sample are $45,000 and $70,000
respectively. As reported in Table 1, in the NLSY97, it is possible to distinguish among
those who own homes or apartments and those who own mobile homes, ranches or farms.
However, only a small fraction of households own mobile homes or ranches;
consequently, in our analysis we combined all these categories together (from Table 2 on)
into the variable ‘housing.’ In addition to home equity, some households have other real
estate, and the value of this asset is also sizable.
Even though a small fraction of the sample report owning business equity, the
actual values reported in this asset are often huge. The conditional mean is more than
$487,000 and for a few households the reported value is above $1,000,000. Consistent
with the evidence in other data sets, we also find that business equity accounts for a
disproportionate share of total wealth. It is not obvious, however, how to treat wealth
invested in business equity, since in this case the enterprise motive is mixed with other
savings motives. In addition, it is not obvious how easy it is to liquidate business equity
in case one needs to have access to those resources, or how easy it is to borrow against
business equity. Consequently, in our empirical analysis we examine different measures
of accumulation that exclude and include business equity.
Another important wealth component is retirement savings. Many parents
accumulate wealth in IRAs and 401(k)s. However, there are constraints and penalties in
accessing these assets and, given these limitations, we examine them separately. We do
not have information on pension wealth in the NLSY97 and we may end up treating
respondents that have defined contributions versus defined benefits plans differently. In
addition, as mentioned before, a large proportion of households were unable to report the
values of retirement savings and we had to impute many of those values.
Other variables to consider in the analysis of wealth, given the age group in our
sample, are educational IRA accounts and other pre-paid tuition saving accounts
established to help pay college costs, as well as loans for children’s educational expenses.
Approximately 9% of the households report having educational IRAs (Table 1). The

conditional median and means are approximately $16,600 and $27,000 respectively
(Table 2). As discussed below, having educational IRAs is also strongly correlated with
the education and race of the respondent. A small proportion of households also report
having educational loans. The amount owed on these loans is on average $6,600.
The amount invested in assets such as checking and saving accounts, bonds and
stocks, varies widely across the sample. In particular, the distribution of stock is very
skewed to the right. While the median stock-owner reports $10,000 in stocks, the mean is
more than $52,000 and there is a small proportion of respondents who report very large
amounts in stocks. Given the behavior of the stock market and the large appreciation in
the value of stocks in the 1990s, this component of wealth is likely to play an important
role both in explaining wealth accumulation and the wide disparity of wealth holdings
across the population. As reported before, however, the values of financial assets, such as
stocks and bonds, have frequently been imputed.
We examine household wealth holdings by considering several measures of
accumulation. We consider financial net worth, total net worth, and retirement savings. In
the first measure, we sum the values of checking and saving accounts, bonds, stocks,
other assets, and the value of educational accounts, and we subtract short-term debt and
debt on educational loans. In the second measure, we also add the value of homes, other
real estate, cars and other vehicles, business equity and we subtract all mortgages and
other debts on homes or cars.15 The first measure represents an indicator of all liquid
assets (or assets easy to liquidate). This could provide some measure of the ability of
households to buffer short-term shocks and short-term expenses. Total net worth is a
more comprehensive measure of accumulation even though it includes assets such as
homes and cars that have consumption purposes in addition to investment purposes and
may not be liquid or easy to liquidate. Note that the sample we have is only
representative of the population of parents with teen-age children (children who are 12 to
16 years old), not simply of young parents with children.

In our measure of wealth we also do not include the value of furniture, which is reported in bracketed
amounts only. First, even when using brackets, there are many non-responses. Second, there is not a welldeveloped second-hand market for this type of asset and it is not clear how households assess the value of
their furniture.

2.3 Comparisons with Other Data Sets.

To provide an evaluation of the quality of the data, we compare wealth holdings
in the NLSY97 with other data sets that report data on household wealth. The SCF is one
of the best and most thorough data sets concerning wealth. It is a triennial survey of U.S.
families sponsored by the Board of Governors of the Federal Reserve System and is
designed to provide detailed information on U.S. families’ balance sheets and their use of
financial services. To that effect, the data set is organized to collect information on assets
and liabilities at a very detailed level. In addition, to accurately measure wealth
accumulation, the SCF oversamples high income households. Data on assets and
liabilities are much more disaggregated in the SCF with respect to the NLSY97. For our
work, we consider the 1995 wave (SCF95 hereafter) which surveyed a total of 4,299
households.16
To make the data sets comparable, in the SCF95 we consider only the households
that have teen-age children (12-16 years old). This restricts the sample to a total of 625
households.We always use the household weights to account for the fact that the SCF95
oversamples rich households, and we use data that have already been adjusted to take
care of non-responses.
Comparisons across surveys suffer from several difficulties. First, it is not always
possible to match the exact definition of assets and liabilities across data sets. Also, the
years when the data are collected are different (1997 for the NLSY and 1995 for the
SCF). Second, differences in methods of data collection are going to inevitably generate
discrepancies across surveys. For example, while in the NSF97 there is only one question
concerning the amount invested in stocks, the information about stocks in the SCF is
collected by going through a long set of detailed questions concerning several categories
of stocks. To gauge the importance of these differences, we compare the NLSY97 data

For a thorough description of the SCF95 and many descriptive statistics, see Kennickell and Starr
McCluer (1997).

with another data set, the PSID, that collects wealth data in a similar fashion than the
NLSY97.
The PSID is a panel data set reporting extensive information about household
income. It started in 1968 and interviewed approximately 5,000 households. Similarly to
the NLSY97 that oversamples Black and Hispanics, the PSID oversamples low income
people. Starting in 1984, special supplemental surveys have been administered on assets
and liabilities and these data are collected in 5-year intervals. In our work, we use the
1994 wave (PSID94 hereafter).
Unfortunately, we do not have very detailed information on the age of children in
the PSID94, but by using data in previous waves we can identify families with children in
the age range 8 to 19. Thus, we have taken the sample of all parents with those children
in the PSID94 and also distinguished between married and non-married ones (which
include those parents who were never married, separated, divorced or widowed). We
make this distinction in the SCF95 sample as well. The total number of observations in
the PSID sample is 2,327.
Note that even though the PSID was not designed to collect wealth data, its
measures are rather accurate. Juster, Smith, and Stafford (1998) report an evaluation of
wealth data across data sets and find that the PSID and the SCF compare very well in
their estimates, up to the top 1% of the wealth distribution where estimates diverge.
Similar findings regarding the accuracy of wealth data in the PSID were also reported in
an earlier study by Curtin, Juster, and Morgan (1989).
Table 3a reports the value of assets and liabilities that can be compared between
the NLSY97 and the SCF95. The amount in checking accounts and bonds, and the
amount in stocks is much lower in the NLSY97 than in the SCF95, especially for married
couples. This may be due to the high level of aggregation at which these assets are
collected in the NLSY97 and to the difficulties respondents have in following closely the
behavior of financial markets.
Business equity is also different, but there are very large values in the amount of
business equity in the NLSY97 and they have an influence on the mean. Retirement
assets are also somewhat different and again there are some influential observations in
the NLSY97.

Overall real assets in the NLSY97, such as housing equity and cars (net of debts),
compare relatively well with data from the SCF95. As mentioned before, these assets also
suffer relatively less from the problem of non-responses. This is an important finding
since housing and cars are the major assets young families have. Financial net worth and
total net worth (which is inclusive of retirement savings) are usually lower in the
NLSY97 than in the SCF95.
In order to better understand what is driving these differences, we have also
compared data on ownership and values conditional on ownership (Tables 3b and 3c).
We find that the ownership of financial assets, such as checking accounts and bonds as
well as stocks are under-reported in the NLSY97 compared with the SCF95. There is a
tendency to under-report business equity as well. However, ownership of real assets,
such as housing and vehicles, compare well across the two surveys. Conditional values
compare well for real assets, but there are often under-reports for financial assets. The
data reveal again the importance of some influential observations for retirement assets
and business equity in the NLSY97.
As far as the PSID is concerned (Table 4a), real assets in the NLSY97, such as
housing equity and cars (net of debts), compare relatively well with data from the
PSID94. As mentioned before, these assets also suffer relatively little from the problem
of non-responses. On the other hand, financial assets show some differences between the
two surveys. The amounts in checking accounts and bonds and in stocks are lower in the
NLSY97 than in the PSID94. While some of the differences may be due to differences in
asset definitions (in the PSID94, IRAs are included in both bonds and stocks, while they
are listed among retirement assets in the NLSY97), the amounts invested in these assets,
and particularly in stocks, are rather different in the two samples. This led to lower values
of total net worth in the NLSY97 as compared to the PSID94.
Comparisons of asset and liability ownership show similar findings as previously,
i.e., the ownership of financial assets is under-reported in the NLSY97. Business
ownership is under-reported as well. However, ownership of real assets compares well
across the two data sets. Values, conditional on ownership, continue to be lower for
financial assets and the presence of influential observations in the NLSY97 in business
equity, retirement assets, as well as stocks persists.

Note that these findings are similar to the results of Engelhardt (1998), who
compares previous NLSY wealth data with data from SIPP. He finds that the largest
discrepancies are concentrated among financial assets, while housing equity is reported
rather well.

3.1 The Distribution of Wealth Across Demographic Groups

In the following section, we examine the distribution of financial and total net
worth across demographic groups.17 We also examine retirement savings. Further, to
complete the analysis we look at the ownership of assets and liabilities in addition to
values. As mentioned before, data on ownership is useful per se and, in addition, it is less
affected by measurement error. All characteristics refer to the mother (biological, stepmother, adopted or foster mother or mother figure) of the children interviewed in the
NLSY97. This analysis serves to illustrate the main features of patterns of accumulation
as well as shed some light on the determinants of household savings.
We first consider the distribution of wealth across age groups (Table 5a).18One
important finding is that families with young mothers (younger than 35) hold very small
amounts of wealth. These families have almost nothing in terms of financial net worth
and their total net worth is very small. However, wealth increases strongly with age. For
example, we find that families with parents in their late thirties or forties have sizable
amounts of total net worth. While it is not possible to disentangle age and cohort effects
in a single cross-section and it is clear that we are not following the same family overtime, this fact has been documented in other studies as well.
Families with an older mother are also more likely to be home-owners or have a
business. In fact, home-ownership and business ownership are particularly low for young
families (mother younger than 35). Older families are also 2 or 3 times more likely to
hold stocks and bonds. They are also much more likely to hold educational IRAs (Table
17

For an analysis of the distribution of saving and wealth in other data sets, see the survey by Browning
and Lusardi (1996).
18
There are, however, several pitfalls at simply looking at age of the mother. First, this may be a bad proxy
for the age of the main earner of the family, which can be for example much older than the mother is.
Additionally, in particular in the case of non-biological mothers, among the older age group we may have

7a), and to accumulate sizable amounts in these accounts. In Tables 7a-d, we also report
the total number of assets and liabilities of households and the percentage of families
with zero financial assets, which is defined as the percentage of families that do not have
any checking and saving accounts, bonds, stocks, and educational IRAs. Overall, the
majority of families hold their wealth in 2 or 3 assets, which are mainly their house and
some liquid assets. A sizable proportion of families, however, do not have any financial
assets. For example, more than 40 percent of young families have zero financial assets.
The simple life-cycle/permanent-income model predicts that parents facing an
upward sloping age-earnings profile should borrow to smooth consumption over their life
cycle. While it is not surprising to see low wealth holdings at young ages, it is an issue
how early fertility affects family formation and performance in the labor market. Of
equal importance is how young parents deal with the financial consequences of sending
children to college and buffering shocks to income.
The lower panels of Table 5 report the distribution of wealth across race and
ethnicity, education, and marital status. These characteristics can serve as proxies for
permanent income and allow us to examine more closely the distribution of wealth across
classes of income. Wealth varies widely across education groups. Families where the
mother has a college education have approximately 4 times more total net worth
(considering medians) than families where the mother has a high school education.
Differences become particularly large when considering lower levels of education;
families where the mother has a college education have approximately 30 times the total
net worth of families with less than a high school education. Differences in wealth
become particularly large when considering financial wealth. Many parents with low
levels of education have almost nothing in financial wealth. Many other studies report
huge differences in wealth holdings across education groups in the population.19 Thus,
these differences are present at the beginning of the life cycle of young parents and tend
to persist at an older age.
Differences in wealth holdings are large not only across education, but also within
education groups. Looking at both financial and total net worth, families differ
grand-parents that take care of children and this may also distort the statistics of wealth across age groups.
These figures should therefore be examined with caution.
19
See, among others, Bernheim and Scholz (1993), and Hubbard, Skinner and Zeldes (1995).

substantially in their wealth holdings even in the same education group. This suggests
that other factors, in addition to income, play a role in explaining wealth accumulation.
An examination of ownership rather than values provides additional information
on patterns of accumulation (Table 7b). Families whose responding parent has less than a
high school education are very unlikely to hold any bonds or stocks, as well as basic
assets, such as saving and checking accounts. Less than 45% of families without a high
school degree hold checking and saving accounts. Only 2% of families in this education
group hold educational IRAs versus 20% of families where mother has a college degree.
More than 50% of families with less than a high school education have zero financial
assets. Overall, these households hold all of their wealth in one or two assets. A possible
explanation for these findings is the lack of financial literacy among these households,
which can provide obstacles to accumulation, in addition to low income. Given the
behavior of the stock market and the housing market, as well as the booming of starts-ups
and business opportunities, we can expect divergences of wealth across education groups
to continue growing, given the small percentage of families with low education that hold
those types of assets.20
Table 5c reports the distribution of wealth across race and ethnicity. Differences
in wealth holdings are huge. Both Black and Hispanic parents report a very low amount
of total net worth. The differences in wealth with respect to White households are large,
perhaps more than differences in labor income can rationalize. White households report
ten times more total net worth (in the median) than Blacks or Hispanics. Differences are
particularly large in financial net worth where, again, Blacks and Hispanics hold very low
amounts of financial wealth. This is due not only to the fact that the amount invested in
financial assets is low, but also to the fact that 50% of Blacks and Hispanics hold no
financial assets at all (Table 7c).
The distribution of assets and liabilities across race and ethnicity shows that less
than 50% of Blacks and Hispanics hold a checking or saving account, and very few hold
stocks or bonds.21 A disproportionately low fraction of Black households have any
20

Wolff (1994) documents that the distribution of wealth has become more unequal. See also Bernheim
(1996) for a discussion of financial literacy.
21
These findings are confirmed in the study by Caskey and Peterson (1994). These authors show that the
percentage of households without checking and/or saving accounts is concentrated among racial and ethnic

business equity. This finding has been reported in many other studies, but there are no
convincing explanations yet for why there are so few black entrepreneurs.22 Whites are
more than twice as likely to have educational IRAs. They are also substantially more
likely to be home owners (77% of Whites own a home compared with 46% and 49% of
Blacks and Hispanics respectively). Overall, with respect to White households, Blacks
and Hispanics are less likely to own any assets and be in debt.
Other studies report similar findings for other age groups. Using data from the
1995 SCF, Kennickell and Starr-McCluer (1997) show that net worth of White, nonHispanic households is more than 4 times larger (in the median) than net worth of nonWhites or Hispanics. Smith (1995), Lusardi (1999), and Venti and Wise (1998) report the
distribution of total net worth in the HRS which considers households whose respondents
were 51 to 61 years old in 1992. Wealth differences are large not only at the beginning of
the life cycle, but they magnify at later stages of the life cycle. For every dollar of wealth
a middle-aged White household has, a Black household has 21 cents and an Hispanic
household has 26 cents (in medians). Additionally, at the median, a middle-aged Black or
Hispanic household has no liquid assets. Thus, for some demographic groups, low
accumulation of financial wealth persists over the life-cycle.
The last panel of Table 5 reports the distribution of wealth across marital status.
Differences in total net worth are striking. Divorced or separated mothers report very low
amounts of total net worth. Marital disruption has a strong effect on financial wealth too
and in particular, separated parents have little or no financial assets.23 Although there are
difficulties in assessing the direction of causality, it is clear that family break-ups are
strongly associated with the accumulation of wealth. Wealth holdings are also low for
mothers who never got married. These findings become even more apparent when
looking at asset and debt ownership. Only one third of mothers who never got married
own a home and only 42% own checking and saving accounts.
These findings are relevant. A large proportion of children grow up with only one
biological parent. McLanahan and Sandefur (1994) examine the role of single parenthood

minorities, and among families headed by an individual who is unmarried, female, or has not completed
high-school.
22
See Meyer (1995) for a review.
23
See also Smith (1994).

on children. They present evidence that suggests that children from two parent homes are
more successful at transitioning in school, finding a job, and starting families. Children
who grow up with only one parent face a higher risk, than those that have two biological
parents, of dropping out of high school. Further, they present evidence that the
disadvantages associated with family disruption persist beyond the high school years. It is
useful to know whether some of these disadvantages were due to the lack of financial
resources.
Tables 6a-d report the distribution of retirement savings across demographic
groups. The heterogeneity in this type of assets is particularly high and, as before,
differences are substantial not only across, but also within demographic groups.24 The
pattern of assets earmarked for retirement mirrors the pattern of accumulation of other
components of total net worth. While these assets may be strongly correlated with
earnings and the types of jobs held by parents, they also vary widely across households.
As for financial and total net worth, some demographic groups simply have little or no
retirement savings. In particular, a large share of parents with low educational attainment,
and Black and Hispanic parents have no retirement savings. Retirement savings are also
low for families that experienced a break-up (divorced or separated parents) and are even
lower for the never married.
On the other hand, there is also a group of households that have already
accumulated a great deal of retirement savings. Thus, at least for parents with a college
degree, accumulation for retirement is present and relevant even at early stages of the life
cycle. Even some young households (older than 35) invest high amounts in retirement
savings and while, as expected, retirement savings are strongly correlated with age, they
also vary widely within age groups.

3.2 Discussion

The patterns of accumulation highlighted in the previous sections raise several
questions. As mentioned before, several demographic groups, and in particular young
24

There are few households that reported very large amounts in retirement assets. These observations have
effects on both means and standard deviations reported in Tables 6. We were unable to determine, however,
whether these potential outliers were due to measurement error and decided to keep them in our sample.

mothers, mothers with low educational attainment, and Blacks or Hispanics have
basically no wealth. This raises concerns about how these families will be able to deal
with potential shocks to income (periods of unemployment, illnesses, etc.) and the
financial burden of sending children to college. It also raises the issue of whether periods
of financial strain affect children’s behavior; for example, affect children’s expectation of
going to school and entering the labor market in the future.
The composition of household portfolios shows that many young families do not
invest in high return assets, such as stocks, bonds and real estate, and many do not even
have basic assets, such as saving and checking accounts. Returns on portfolios and assets’
allocation may be another important reason why wealth differs and continues to differ so
much across households of similar characteristics and economic status. They may also
explain why differences become larger at older ages. This factor may also play a bigger
role in the current economy if the stock market continues to deliver returns different than
other financial markets. The re-valuation in the housing market may also be at play to
explain difference in wealth accumulation across households.
Also, note that wealth can be low because families have been hit by shocks that
depleted their resources. While income shocks can be a cause of these low wealth
holdings, family break-ups can also drain resources. The data reported in the previous
tables indicate that families which are intact have much more wealth than families that
experienced a break-up.
Are low wealth holdings, in particular among poor families, a puzzle?
Unfortunately, there exist several tax incentives for poor families to hold low wealth, in
particular, little or no financial assets. Many welfare programs are means-tested and they
provide strong incentives against accumulation. As Hubbard, Skinner, and Zeldes (1995)
document, these programs have a disproportionate impact on the saving behavior of
lower income households. The implicit tax on wealth and saving for these families can be
as high as 100%. Gruber and Yelowitz (1999) also find that the extension of social
insurance programs over the period 1984-1993 had a sizeable and significant negative
effect on the wealth holdings of poor families.
Similarly, college scholarship rules provide many disincentives to accumulate
wealth. As Feldstein (1995) shows, families that are eligible for college scholarships face

very steep “education tax rates;” scholarship rules implicitly levy taxes on capital income
and on accumulated assets that range from 30 to 50%. Such taxes are a strong incentive
not to save for college expenses and instead rely on financial assistance as well as on
market borrowing. His calculations show that these taxes can reduce accumulation of
financial assets by as much as 50%.25 In addition, since any funds saved for retirement
are also subject to education levies, scholarship rules discourage saving for other motives
as well. This may explain why households that have little in total net worth have also
little in retirement savings.
One has also to be cautious in making assessments about household wealth
holdings by looking at private wealth only. Families accumulate wealth in pensions and
Social Security as well. It is hard if not impossible to gauge household wealth and, in
particular, savings for retirement without information on pension and Social Security
wealth. Recent studies on the HRS show that many of the families that have little private
wealth have a large accumulation in pension and Social Security. Additionally, the
evaluation of total wealth rather than private wealth leads often to different results
concerning the adequacy of savings for retirement.26
The previous analysis provides some insights into the reasons for the sharp
differences in wealth holdings across race and ethnicity. A striking finding of previous
tables is the low percentage of Black households that have any business equity. Given
how much wealth entrepreneurs hold and the upward mobility associated with
entrepreneurship,27 the analysis of who becomes an entrepreneur and whether or not there
exist financial constraints in starting a business can resolve some of the difficulties of
explaining Black-White wealth differences.28
Another consistent finding throughout the analysis is that there is a wide amount
of heterogeneity in household wealth holdings. Many studies have reported this finding
across the U.S. population and among older households.29 However, this is present even
in earlier stages of the life cycle and even among similar demographic groups (such as
25

For further examination of the effects of implicit taxes from college financial aid on incentives to save,
see Dick and Edlin (1997).
26
See Gustman and Steinmeier (1998).
27
See Quadrini (1999).
28
For a discussion see, among others, Blau and Graham (1990) and Meyer (1990).
29
See, in particular, Venti and Wise (1998).

parents with teen-age children). This suggests that, in addition to permanent income,
shocks (such as family break-ups) as well as preferences can be important determinants
of household accumulation.

4. Concluding Remarks and Further Work

In this paper, we examine the wealth holdings of parents with teen-age children.
We find that there is much heterogeneity in household wealth holdings, even among
families in early stages of the life-cycle. In addition, we find that a sizable proportion of
these families have little financial and total net worth. The major asset in their portfolios
is home equity and the vast majority of these households hold no financial assets.
We plan to pursue this work in several directions. First, we plan to examine
whether household resources, not just income but also savings, have any effects on
household behavior. More importantly, we are interested in examining whether wealth
affects children’s expectations of completing a college education. In the NLSY97,
children are asked to report their subjective expectations of completing a college
education by the time they turn 30 and we can examine whether family resources, in
addition to other variables that are predicted to affect children’s behavior, play a role in
shaping expectations about the future.
We also plan to use the richness of information provided in the NLSY97 on
household characteristics to explain White/Black differences in household wealth
holdings. In this respect, we plan to investigate the role and importance of entrepreneurial
wealth and explore the factors that lead to becoming an entrepreneur.

Data Appendix

Treatment of missing values for assets and liabilities
As described in the text, there are several cases where respondents reported
owning a certain asset or liability but did not report its value. However, in most of these
cases, they identified a bracket in which the asset or liability value would fall. This
scheme of responses leads to two different types of missing observations for the values of
each asset and liability, i.e., those not reporting the value but identifying a bracket, and
those not reporting anything at all. In addition, some observations contained missing
values for the variable indicating ownership of a certain asset or liability.
In order to fill in missing ownership indicators and missing values for each asset
and liability, we used a hot-deck imputation method.30 This method consists of replacing
each missing value with a randomly picked observed value extracted from a pool of
respondents that are similar to the non-respondents according to a set of observed
characteristics. Ideally, one would like to use a large number of observed characteristics
and use a fine grid for each characteristic in order to make the matching more precise.
However, the number and diversity of reported values limits the feasible extent of such
precision.
In order to determine what variables to use to do the matching when imputing
ownership, we examined the results of probit regressions of asset (and liability)
ownership on a set of variables indicating mother’s age, race, and marital status, mother’s
and father’s education, region of residence, family size, and income. For each asset and
liability that required ownership imputation, we based the matching procedure of such
imputation on those variables with higher predictive power to assess ownership (See table
A1).
Once we had all observations either with reported or imputed ownership
indicators, we proceeded to impute brackets to those cases with missing values and
brackets. Finally, we imputed the values of assets and liabilities.

The procedure we followed is the one proposed by Juster and Smith (1997).

When imputing brackets and values for each asset and liability, we selected the
variables used for the matching by examining results of OLS regressions of asset (and
liability) values on a set of characteristics that included mother’s age, race, and marital
status, mother’s and father’s education, region of residence, and income. In each case,
again, we picked the variables with the most predictive power. Table A2 shows the set of
variables that, in addition to ‘bracket’, we used to do the matching for the imputation of
missing values for each asset and liability.
In those observations where neither the value nor the bracket was reported, before
imputing the asset or liability value, we first imputed the bracket. We used the same hot
deck imputation method and matched the observations with missing brackets to those
who identified a bracket for the value of their asset or liability but did not report a precise
value. We did not include in this matching process those who reported exact values of
their assets or liabilities since it is reasonable to expect that those who did not identify a
bracket are more similar to those who only identified a bracket than to those who
reported a precise value. For the imputation of brackets, we also differentiated missing
values in brackets due to ‘Do not know’ from ‘Refusals’ and we required that the
observation selected to fill in the imputation had to be of the same type in their reason for
not reporting a value for the asset or liability. The variables used for this matching are
the same reported in Table A2.
The discrete values that each of the matching variables (used either for the
imputation of ownership, bracket or values) could assume are given in Table A3. To
minimize the effect of imputation estimation error, we computed all hot-deck imputations
for brackets and values across 25 independent trials.

Table A1
Asset/Liability
Home or apartment
Other homes
Business equity
Educational IRA
Retirement savings
Checking/savings accounts
Stocks
Bonds
Other savings
Car/vehicles
Mortgages
Other mortgages
Educational loans
Loans on cars
Other loans

Variables used to do the matching when imputing ownership
Income, marital status, race, mother’s age
Income, marital status, race, mother’s age
Income, Father’s education, race, marital status
Income, father’s education, mother’s education, family size
Income, marital status, race, mother’s education
Income, race, mother’s education, marital status
Income, race, father’s education, marital status
Income, race, marital status, mother’s education
Income, mother’s education, race, region
Region, marital status, race, income
Income, marital status, race, mother’s education
Income, marital status, race, region
Region, race, mother’s age
Marital status, race, region, income
Mother’s education, marital status, race, mother’s age

Table A2
Asset/Liability
Ranch
Mobile home and site
Home or apartment
Other homes
Business equity
Educational IRA
Retirement savings
Checking/savings accounts
Stocks
Bonds
Other savings
Car/vehicles
Mortgages
Other mortgages
Educational loans
Loans on cars
Other loans

Variables used to do the matching when imputing brackets (1)
and values (2)
Mother’s education
Income
Income, race
Income, mother’s education
Father’s education
Income, father’s education
Income, father’s education
Income, father’s education
Income, mother’s age
Income
Income, race
Income, race
Income, race
Income, race
Income
Income, mother’s education
Income, father’s education

Notes: (1) For the imputation of brackets, the reason of non-response (i.e. ‘do not know’
vs. ‘refusal’ was also considered.
(2) For the imputation of values, ‘bracket’ was also added to the list of matching
variables.

Table A3
Variable
Brackets

Grid points
7

Income
Father’s education

5
5

Family Size
Marital Status

5
5

Mother’s age
Mother’s education

3
5

Race
Region

4
4

Grid values
Brackets specified for each asset in the NLSY
questionnaire
First through fourth income quartile, and missing values
High school dropout or less – High school graduate –
College dropout – College graduate or more – Missing v.
2-3 – 4-5 – 6-7 – 8+ – Missing values
Married – Divorced or separated – Widow – Never
married – Missing values
Less than 35 – 35 to 45 – More than 45
High school dropout or less – High school graduate –
College dropout – College graduate or more – Missing
values
White – Black – Hispanic – Other race
East – Central – South – West

References

Alessie, Rob, Annamaria Lusardi and Trea Aldershof (1997), “Income and Wealth Over
the Life Cycle: Evidence from Panel Data,” Review of Income and Wealth, 43, pp. 1-32.

Blau, Francine and John Graham (1990), “Black-White Differences in Wealth and Asset
Composition,” Quarterly Journal of Economics, 105, pp. 321-339.

Bernheim, Douglas(1996), “Personal Saving, Information, and Economic Literacy: New
Directions for Public Policy,” in Tax Policy for Economic Growth in the 1990s,
Washington, D.C.: American Council for Capital Formation, pp. 53-78.

Bernheim, Douglas and John Karl Scholz (1993), “Private Saving and Public Policy”, in
J. Poterba (ed.) Tax Policy and the Economy, MIT Press, pp. 73 –110.

Browning, Martin and Annamaria Lusardi (1996), “Household Saving: Micro Theories
and Micro Facts,” Journal of Economic Literature, 34, pp. 1797-1855.

Caskey, John and Andrew Peterson (1994), “Who Has a Bank Account and Who
Doesn’t: 1977 and 1989,” Eastern Economic Journal, 20, pp. 61-73.

Curtin, Richard, Thomas Juster and James Morgan (1989), “Survey Estimates of Wealth:
An Assessment of Quality,” in The Measurement of Saving, Investment, and Wealth, in
Robert Lipsey and Helen Stone Tice (eds.), Chicago: The University of Chicago Press,
pp. 473-548.

Dick, Andrew and Aaron Edlin (1997), “The Implicit Taxes from College Financial Aid,”
Journal of Public Economics, 65, pp. 295-322.

Engelhardt, Gary (1998), “Income and Wealth in the NLSY79,” mimeo, Dartmouth
College.

Engelhardt, Gary (1996), “Consumption, Down Payments, and Liquidity Constraints,”
Journal of Money, Credit and Banking, 29, pp. 255-271.

Engelhardt, Gary (1994), “House Prices and the Decision to Save for Down Payments,”
Journal of Urban Economics, 36, pp. 209-237.

Feldstein, Martin (1995), “College Scholarship Rules and Private saving,” American
Economic Review, 85, pp. 552-566.

Gentry, William and Glenn Hubbard (1998), “Entrepreneurship and Household Saving,”
mimeo, Columbia Business School.

Goodman, John and John Ittner (1992), “The Accuracy of Home Owners’ Estimates of
House Value,” Journal of Housing Economics, 2, pp. 339-357.

Gruber, Jonathan and Aaron Yelowitz (1999), “Public Health Insurance and Private
Savings,” Journal of Political Economy, 107, pp. 1249-1274.

Gustman, Alan and Tom Steinmeier (1999), “What People Don’t Know About Their
Pensions and Social Security: An Analysis Using Linked Data from the Health and
Retirement Study,” mimeo, Dartmouth College.

Gustman, Alan and Tom Steinmeier (1998), “Effects of Pensions on Saving: Analysis
with Data from the Health and Retirement Study,” NBER Working Paper n. 6681.

Hubbard, Glenn, Jonathan Skinner and Steve Zeldes (1995), “Precautionary Saving and
Social Insurance”, Journal of Political Economy, 103, pp. 360 – 399.

Hurd, Michael, Daniel McFadden, Harish Chand, Li Gan, Angela Merrill and Michael
Roberts (1997), “Consumption and Savings Balances of the Elderly: Experimental
Evidence on Survey Response Bias,” mimeo, RAND.

Juster, Thomas, James Smith and Frank Stafford (1998), “The Measurement and
Structure of Household Wealth,” mimeo, University of Michigan.

Juster, Thomas and James Smith (1997), “Improving the Quality of Economic Data:
Lessons From the HRS and AHEAD,” Journal of the American Statistical Association,
92, pp. 1268 – 1278.

Kennickell Arthur and Martha Starr-McCluer (1997), “Family Finances in the U.S.:
Recent Evidence from the Survey of Consumer Finances,” Federal Reserve Bulletin,
January, pp. 1-24.

Lusardi, Annamaria (1999),“Information Expectation and Savings for Retirement,” in H.
Aaron (ed.) Behavioral Dimensions of Retirement Economics, Washington, D.C.:
Brookings Institution and Russell Sage Foundation, pp. 81 – 115.

Meyer, Bruce (1990), “Why Are There So Few Black Entrepreneurs?” NBER Working
Paper n. 3537.

Pemberton, Jennifer and John Reynolds (2000), “Rising College Expectations Among
Youth in the U.S.: A Comparison of 15 and 16 Years Old in the 1979 and 1997 NLSY,”
mimeo, Florida State University.

Quadrini, Vincenzo (1999), “The Importance of Entrepreneurship for Wealth
Concentration and Mobility,“ Review of Income and Wealth, 45, pp. 1-19.

Smith, James (1994), “Marriage, Assets, and Savings,” mimeo, RAND.

Smith, James (1995), “Racial and Ethnic Differences in Wealth in the Health and
Retirement Study,“ Journal of Human Resources, 30, Supplement, pp. 158-183.

Venti, Steven and David Wise (1998), “The Cause of Wealth Dispersion at Retirement:
Choice or Chance?,” American Economic Review papers and Proceedings, 88, pp. 185191.

Zagorsky, Jay (1999), “Young Baby Boomers’ Wealth”, Review of Income and Wealth,
45, pp. 135 – 155.

Wolff, Edwards (1994), “Trends in Household Wealth in the United States: 1962-1983,
and 1983-1989,” Review of Income and Wealth, 40, 143-174.

Table 1
Non-Responses on Assets and Liabilities

Assets
Check. Acct.
Bonds
Stocks
Edu. IRAs
Retirement Savings
Other Savings
Houses & Apts.
Ranches & Farms
Mobile Home
Other Real Estate
Business Equity
Cars
Liabilities
Mortgages
Other Mortgages
Car Debt
Edu. Loans
Other Debt

Proportion
Proportion of
of Ownership Owners who
Don’t Know

Proportion of
Owners who
Refuse to Answer

67.21
17.36
17.11
9.04
54.66
13.73
63.26
0.02
5.20
13.10
11.89
89.26

14.63
24.05
37.06
23.55
35.45
24.57
6.70
19.38
13.39
14.06
28.87
10.54

5.98
6.19
5.06
3.79
3.41
2.53
1.50
4.02
0.71
2.02
4.16
1.59

56.42
11.95
47.32
4.50
56.28

11.77
10.30
9.98
5.48
9.02

3.42
1.48
3.53
1.72
1.83

Note: This table reports the proportion of NLSY97 respondents reporting ownership of assets and
liabilities. The second and third columns report the proportion of those who refuse to report the
value or report that they do not know the value of their assets and liabilities. All values are weighted
using household weights. The total number of observations is 6,113.

Table 2
The Distribution of Assets and Liabilities in the NLSY97

Assets

Ownership First
(% own) Quartile

Median

Third
Quartile

Mean

Stand.
Dev.

Max. Value

Check. Acct.

67.22

500

2,000

7,224

10,584

38,688

900,000

Bonds

17.49

1,000

5,000

19,840

21,529

63,237

1,000,000

Stocks

17.21

2,200

10,000

40,000

52,147

183,636

3,000,000

Edu. IRAs

9.14

5,168

16,600

30,000

26,951

41,276

400,000

Retirement Savings

54.67

8,000

25,000

66,336

78,308

978,789

50,000,000

Other Savings

13.74

5,000

15,000

53,570

61,838

197,374

4,000,000

Property

68.74

60,000

95,160

150,000

125,100

127,030

2,000,000

Other Real Estate

13.09

10,000

33,000

80,000

79,847

157,223

2,000,000

Business Equity

11.93

2,000

30,000

250,200

487,550

Cars

89.28

5,000

12,000

20,740

16,086

45,923

4,000,000

Mortgages

56.37

32,000

57,000

90,000

69,723

59,601

600,000

Other Mortgages

12.04

7,200

14,000

21,880

18,305

19,179

170,000

Car Debt

47.31

4,000

8,000

14,832

10,396

11,589

370,000

Edu. Loans

4.50

600

4,000

7,567

6,653

12,380

150,000

Other Debt

56.28

1,800

4,300

10,000

8,550

19,316

500,000

2,649,019 40,000,000

Liabilities

Note: This table reports the conditional distribution of assets and liabilities. All values are weighted using household weights. The
total number of observations is 6,113.

Table 3a
Comparison of Wealth between the NLSY97 and the SCF95

Assets and
Liabilities
Checking and bonds
Stocks
Short-term debt
Business
Retirement Savings
Housing equity
Vehicles
Financial net worth
Total net worth incl.
retirement savings

Marital
Status
Married
Non-Married
Married
Non-Married
Married
Non-Married
Married
Non-Married
Married
Non-Married
Married
Non-Married
Married
Non-Married
Married
Non-Married
Married
Non-Married

Medians

Means

NLSY 97

SCF95

NLSY 97

SCF95

1,464
45
0
0
1,000
550
0
0
8,220
0
30,000
0
7,600
1,500
500
0
78,262
6,256

3,465
682
0
0
1,564
966
0
0
5,040
0
33,600
0
8,400
2,625
4,042
105
77,448
12,022

13,593
4,573
11,291
3,742
5,522
4,213
72,993
25,321
43,038
43,507
55,820
17,702
11,436
4,896
32,452
10,309
228,675
105,320

28,186
6,624
21,971
3,696
6,053
3,242
78,351
8,219
39,511
4,955
61,525
28,779
12,132
4,139
65,923
26,638
281,777
81,402

Note: This table reports a comparison of assets and liabilities in the NLSY97 and the SCF95 by marital status. All values
are in 1997 dollars. The total number of observations is 6,113 and 625 in the NLSY97 and the SCF95 respectively.

Table 3b
Comparison of Asset Ownership in the NLSY97 and the SCF95
Ownership of Assets and Liabilities

Assets and
Marital
Liabilities
Status
Checking and bonds Married
Non-Married
Stocks
Married
Non-Married
Short-term debt
Married
Non-Married
Business
Married
Non-Married
Retirement Savings Married
Non-Married
Housing equity
Married
Non-Married
Vehicles
Married
Non-Married
Financial net worth Married
Non-Married
Total net worth incl. Married
Retirement savings Non-Married

NLSY 97

SCF95

0.74
0.54
0.21
0.08
0.57
0.56
0.15
0.06
0.64
0.32
0.80
0.44
0.95
0.75
0.54
0.34
0.92
0.70

0.94
0.77
0.28
0.10
0.74
0.68
0.19
0.08
0.60
0.27
0.83
0.50
0.93
0.74
0.68
0.52
0.95
0.80

Note: This table reports a comparison of the ownership of assets and liabilities in the NLSY97 and the SCF95
by marital status. The values in the last two rows refer to the percentage of families that report strictly positive
financial and total net worth (including retirement savings). The number of observations is 6,113 and 625 in

Table 3c
Comparison of Wealth between the NLSY97 and the SCF95
conditional on Ownership
Assets and
Marital
Liabilities
Status
Checking and bonds Married
Non-Married
Stocks
Married
Non-Married
Short-term debt
Married
Non-Married
Business
Married
Non-Married
Retirement Savings Married
Non-Married
Housing equity
Married
Non-Married
Vehicles
Married
Non-Married
Financial net worth
Married
Non-Married
Total net worth
Married
Non-Married

Medians

Means

NLSY 97

SCF95

NLSY 97

SCF95

3,000
1,000
15,000
4,000
5,000
3,955
40,404
10,000
30,000
10,000
45,000
25,000
8,000
3,000
16,700
5,000
91,415
22,760

3675
1,186
12,600
5,250
3,150
2,415
84,000
5,250
24,150
4,200
46,200
32,550
8,820
3,832
16,170
6,090
81,931
30,870

18,416
8,474
52,951
48,642
9,588
7,550
497,652
436,336
66,990
133,714
70,145
40,566
11,979
6,519
65,842
39,787
250,397
153,881

29,955
8,621
78,139
35,459
8,194
4,744
417,673
101,364
65,465
18,338
73,899
58,800
13,033
5,590
100,260
54,096
294,547
101,543

Note: This table reports a comparison of assets and liabilities conditional on ownership in the NLSY97 and the SCF95 by
marital status. All values are in 1997 dollars. The values in the last two rows refer to strictly positive amounts of financial
and total net worth (including retirement savings). The number of observations is 6,113 and 625 in the NLSY97 and the

Table 4a
Comparison of Wealth between the NLSY97 and the PSID94

Assets and
Marital
Liabilities
Status
Checking and bonds Married
Non-Married
Stocks
Married
Non-Married
Short-term debt
Married
Non-Married
Business
Married
Non-Married
Housing equity
Married
Non-Married
Vehicles
Married
Non-Married
Total net worth
Married
Non-Married

Medians

Means

NLSY 97

PSID94

NLSY 97

PSID94

1,464
45
0
0
1,000
550
0
0
30,000
0
7,600
1,500
54,500
4,000

3,240
54
0
0
1,620
0
0
0
37,800
0
10,800
3,240
78,300
9,180

13,593
4,573
11,291
3,742
5,522
4,231
72,992
25,321
55,821
17,702
11,436
4,896
185,616
61,788

16,473
6,825
29,498
5,167
7,977
3,323
32,857
4,778
58,430
45,180
14,407
6,818
178,486
49,257

Note: This table reports a comparison of assets and liabilities in the NLSY97 and the PSID94 by marital status. All
values are in 1997 dollars. The number of observations is 6,113 and 2,327 in the NLSY97and the PSID94
respectively.

Table 4b
Comparison of Asset Ownership between the NLSY97 and the
PSID94

Assets and
Liabilities
Checking and
bonds

Marital
Status
Married

Non-Married
Stocks
Married
Non-Married
Short-term debt Married
Non-Married
Business
Married
Non-Married
Housing equity Married
Non-Married
Vehicles
Married
Non-Married
Total net worth Married
Non-Married

Ownership of Assets and Liabilities
NLSY 97

PSID94

0.74

0.83

0.53
0.21
0.07
0.57
0.56
0.15
0.06
0.80
0.43
0.95
0.75
0.90
0.67

0.54
0.42
0.15
0.60
0.46
0.20
0.07
0.82
0.41
0.95
0.70
0.95
0.73

Note: This table reports a comparison of assets and liabilities in the NLSY97 and the PSID94 by marital
status. All values are in 1997 dollars. The last row refers to the percentage of families that report strictly
positive total net worth. The number of observations is 6,113 and 2,327 in the NLSY97 and the PSID94
respectively.

Table 4c
Comparison of Wealth between the NLSY97 and the PSID94
conditional on Ownership

Medians

Means

NLSY 97

PSID94

NLSY 97

PSID94

3,000
1,000
15,000
4,000
5,000
3,955
40,404
10,000
45,000
25,000
8,000
3,000
67,000
18,290

5,400
3,240
21,600
10,800
5,400
3,240
48,600
16,200
48,600
33,480
10,800
5,400
84,240
29,160

18,416
8,474
52,950
48,641
9,588
7,550
497,652
436,335
70,145
40,566
11,979
6,518
207,891
95,902

20,287
12,707
70,260
33,542
11,729
7,272
162,744
63,297
71,202
50,721
15,189
9,798
189,733
68,739

Note: This table reports a comparison of assets and liabilities conditional on ownership in the NLSY97 and the PSID
by marital status. All values are in 1997 dollars. The values in the last row refer to strictly positive amounts of total
net worth. The number of observations is 6,113 and 2,327 in the NLSY97 and the PSID94 respectively.

Table 5a
Wealth Across Age
Net Worth

Financial Net Worth
Age
Less 35
36-39
40-45
Over 45

N. Obs.
1,403
1,947
1,568
1,186

Median
0
0
968
887

Mean
5,569
12,492
31,481
60,763

St. Dev.
43,227
51,779
107,367
247,235

N. Obs.
1,403
1,945
1,566
1,186

Median
7,000
28,600
60,204
72,880

Mean
45,032
108,858
224,149
216,732

St. Dev.
208,333
1,025,125
1,411,615
507,833

Table 5b
Wealth Across Education
Financial Net Worth
Education Level
Less HS
HS
Some College
College
More than College

N. Obs.
1,256
2,135
1,426
692
392

Median
0
0
100
9,000
16,264

Mean
2,631
12,147
28,725
64,068
67,913

Net Worth
St. Dev.
32,363
78,997
175,234
192,170
137,662

N. Obs.
1,256
2,134
1,425
690
392

Median
3,500
26,000
40,500
99,674
135,260

Mean
70,177
89,520
196,852
257,949
252,802

St. Dev.
974,773
292,222
1,653,001
633,191
386,959

Table 5c
Wealth Across Race and Ethnicity
Financial Net Worth
Race
White
Black
Hispanic
Other

N. Obs.
3,277
1,515
1,073
148

Median
576
0
0
500

Mean
32,656
7,192
8,389
36,151

Net Worth
St. Dev.
148,466
83,834
42,681
92,684

N. Obs
3,275
1,514
1,072
148

Median
50,562
4,100
8,500
27,500

Mean
189,505
33,407
57,870
149,883

St. Dev.
1,159,847
116,376
206,765
319,838

Table 5d
Wealth Across Marital Status
Net Worth

Financial Net Worth

Marital Status
Married
Divorced
Separated
Widowed
Never Married

N. Obs.
3,976
929
355
135
618

Median
500
0
0
0
54,500

Mean
32,452
12,249
3,189
38,826
185,616

St. Dev.
138,191
117,221
75,129
221,531
1,056,482

N. Obs
3,974
928
355
135
617

Median
54,500
7,300
1,800
12,850
528

Mean
185,616
85,935
22,847
105,320
22,955

St. Dev.
1,056,482
1,013,990
90,547
340,249
92,770

Note: These tables report the distribution of financial net worth and total net worth across age, education, race and ethnicity, and marital
status. All characteristics refer to the mother of the responding youth. All values are weighted using household weights.
40

Table 6a
Retirement Savings Across Age
Retirement Savings
Age
Less 35
36-39
40-45
Over 45

N. Obs.
1,403
1,947
1,568
1,186

Median
0
1,800
8,336
8,856

Mean
10,013
55,905
42,854
56,631

St. Dev
31,806
1,267,538
152,210
142,410

Table 6b
Retirement Savings Across Education
Retirement Savings
Education Level N. Obs.
Less HS
1,256
HS
2,135
Some College
1,426
College
692
More than College
392

Median
0
800
4,000
20,000
40,000

Mean
7,169
54,014
30,024
58,820
87,459

St. Dev.
31,200
1,218,261
67,931
102,985
132,547

Table 6c
Retirement Savings Across Race and Ethnicity
Retirement Savings
Race
White
Black
Hispanic
Other

N. Obs.
3,277
1,515
1,073
148

Median
7,500
0
0
6,000

Mean
55,865
11,980
10,066
40,609

St. Dev.
872,481
40,765
35,162
84,938

Table 6d
Retirement Savings Across Marital Status
Retirement Savings

Marital Status
Married
Divorced
Separated
Widowed
Never Married

N. Obs.
3,976
929
355
135
618

Median
8,220
0
0
0
0

Mean
43,038
76,852
4,668
13,462
6,002

St. Dev.
134,782
1,829,361
18,520
35,000
22,835

Note: These tables report the distribution of retirement savings across age, education, race and ethnicity,
and marital status. All characteristics refer to the mother of the responding youth. All values are
weighted using household weights.

Table 7a
Ownership of Assets and Liabilities Across Age

Assets
Check. Acct.
Bonds
Stocks
Edu. IRAs
Retirement Savings
Other Savings
Housing
Other Real Estate
Business Equity
Cars
% w/ zero Fin. Assets
N. of Assets
Liabilities
Mortgages
Other Mortgages
Car Debt
Edu. Loans
Other Debt
N. of Debts
Num. Obs.

Age
<35
(% own)
56.20
9.09
7.75
3.73
37.43
10.11
50.55
6.13
6.80
84.35
40.74
2.35

Age
35-39
(% own)
68.80
15.56
15.56
8.36
55.70
13.08
68.13
9.97
10.3
90.10
28.55
2.99

Age
40-45
(% own)
72.40
23.97
21.91
12.53
62.74
16.19
77.68
17.18
15.15
91.34
22.15
3.50

Age
>45
(% own)
70.99
20.72
23.48
11.54
60.38
15.32
76.93
19.98
15.05
90.41
23.62
3.44

39.57
6.21
46.92
2.00
54.70
1.49
1403

59.05
10.16
50.00
2.80
59.12
1.81
1947

64.32
15.75
49.92
5.88
55.99
1.88
1568

59.07
16.24
43.80
8.25
54.65
1.82
1186

Note: This table reports the ownership of assets and liabilities across the responding youth’s mother’s age. All values
are weighted using household weights.

Table 7b

Less
HS
(% own)
43.24
4.54
4.14
2.38
25.00
6.27
46.77
4.62
5.34
77.22
54.78
1.95

High
School
(% own)
66.52
14.07
12.62
5.41
51.98
11.96
68.70
10.32
9.94
89.35
30.21
2.88

Some
College
(% own)
75.46
20.30
19.09
10.17
60.16
17.58
71.35
15.09
12.49
93.03
20.16
3.34

College
(% own)
78.86
28.66
31.55
20.25
77.12
16.72
83.43
21.28
19.43
95.47
14.68
3.95

More Than
College
(% own)
84.36
35.84
37.86
19.80
83.07
23.85
87.41
26.37
21.99
95.32
8.58
4.33

32.60
4.91
35.50
1.47
42.10
1.16
1256

55.78
11.26
48.96
4.71
58.17
1.78
2135

61.16
14.24
52.11
5.37
65.41
1.98
1426

70.46
18.33
50.30
6.34
55.29
2.01
692

76.92
15.91
47.93
4.15
55.59
2.00
392

Note: This table reports the ownership of assets and liabilities across the responding youth’s mother’s education level. All
values are weighted using household weights.

Table 7c
Ownership of Assets and Liabilities Across Race and Ethnicity
Assets
Check. Acct.
Bonds
Stocks
Edu. IRAs
Retirement Savings
Other Savings
Housing
Other Real Estate
Business Equity
Cars
% w/ zero Fin. Assets
N. of Assets
Liabilities
Mortgages
Other Mortgages
Car Debt
Edu. Loans
Other Debt
N. of Debts
Num. Obs.

White
(% own)
75.38
20.04
21.38
10.44
63.94
16.58
77.22
15.61
14.67
94.53
20.15
3.47

Black
(% own)
49.69
8.82
7.11
5.54
32.52
5.17
46.19
6.53
3.92
71.11
50.92
2.01

Hispanic
(% own)
46.71
8.36
6.63
5.49
29.68
8.14
49.71
6.53
6.44
80.21
50.54
2.18

Other
(% own)
66.66
16.27
16.21
12.55
57.77
17.45
66.01
12.64
12.01
92.72
26.31
3.12

64.09
14.81
51.76
4.54
59.47
1.95
3277

36.77
5.33
40.35
4.88
50.79
1.38
1515

38.20
5.05
33.17
3.30
45.76
1.25
1073

55.38
11.03
41.51
5.57
54.56
1.68
148

Note: This table reports the ownership of assets and liabilities across the responding mother’s race or ethnicity.
All values are weighted using household weights.

Table 7d
Ownership of Assets and Liabilities Across Marital Status
Married
(% own)

Divorced
(% own)

Separated
(% own)

Widow
(% own)

Assets
Check. Acct.
Bonds
Stocks
Edu. IRAs
Retirement Savings
Other Savings
Housing
Other Real Estate
Business Equity
Cars
% w/ zero Fin. Assets
N. of Assets

72.73
21.31
21.32
10.97
64.25
15.16
79.59
16.05
14.66
95.46
22.50
3.47

62.68
10.58
10.18
5.19
38.92
12.78
47.96
7.53
7.40
82.76
34.79
2.47

49.48
3.80
3.89
4.03
27.00
10.35
42.13
5.95
5.70
73.61
48.89
1.98

56.84
14.24
9.12
10.13
30.85
10.94
56.25
10.34
2.06
74.91
37.65
2.44

Never
Married
(% own)
41.46
5.86
4.42
3.31
22.84
5.84
31.65
2.51
3.46
59.39
57.29
1.58

Liabilities
Mortgages
Other Mortgages
Car Debt
Edu. Loans
Other Debt
N. of Debts
Num. Obs.

66.33
14.85
53.55
5.00
56.60
1.96
3976

38.60
5.54
38.60
3.22
61.11
1.47
929

31.56
8.02
28.98
4.58
58.15
1.31
355

39.32
5.45
31.42
2.54
59.03
1.37
135

22.10
3.93
24.65
2.62
41.95
0.95
618

Note: This table reports the ownership of assets and liabilities across the responding youth’s mother’s marital status.
All values are weighted using household weights.

Full text of Working Papers (Federal Reserve Bank of Chicago) : Savings of Young Parents, Working Paper 2000-23

FRASER