Changes in family finances from 1983 to 1989: evidence from the Survey of Consumer Finances.

Arthur Kennickell and Janice Shack-Marquez, of the Board's Division of Research and Statistics, prepared this article.

Between 1983 and 1989, family finances in the United States were affected by many factors. Financial deregulation altered the availability and the cost of financial services to consumers. Banks moved to explicit pricing of checking services and the payment of interest on transactions accounts, and money market accounts and other mutual funds became more available. The progressive elimination of tax deductions for consumer interest other than the deduction on home mortgages influenced the effective price of borrowing. Other tax changes, such as the elimination of general deductions for individual retirement accounts, altered the return to various forms of saving.

Macroeconomic and demographic changes also affected the financial opportunities for families. In 1983, the U.S. economy was at the end of a recession; in 1989, it was near the end of a long expansion. Over the intervening six years, aggregate real disposable personal income, as measured in the National Income and Product Accounts, increased by 21 percent; the price level, as measured by the consumer price index, rose about 25 percent; and, as measured in nominal terms by the Federal Reserve flow of funds accounts, total assets and net worth for the household sector grew 61 percent and 56 percent respectively. The overall population of households grew 9 percent; the largest growth was in families headed by individuals between the ages of 35 and 45 years, a group that tends to have a relatively high rate of saving. Other important demographic changes were an increase in the proportion of dual-earner families from 26 1/2 percent of all families in 1983 to 29 percent in 1989 and a small increase in average family size.

With these changes came alterations in the income, assets, and liabilities of U.S. families. Using data from the 1983 and 1989 Surveys of Consumer Finances, this article looks at those alterations and analyzes them according to a variety of economic and demographic characteristics. Several findings are noteworthy. The small rise in the median values of income and net worth and the simultaneous substantial rise in the mean values indicate that the distributions of income and net worth became more concentrated between 1983 and 1989.(1) Also, the use of debt increased; much of that increase was in families reporting the most financial assets.(2)

THE SURVEYS

Widely regarded as a reliable source of data on family finances, the Survey of Consumer Finances (SCF) is designed specifically to gather detailed and comprehensive information on assets, liabilites, and in come flows from a representative sample of the population of U.S. families. Because the ownership of some assets, such as corporate stocks, is relatively concentrated in a small numver of families, the survey makes a special effort to ensure proper representation of such assets by systematically oversampling wealthier families.

Surveys of consumer finances were conducted regularly with support from the Federal Reserve from 1946 through 1970. Another such survey was conducted in 1977 to gather information on the use of consumer credit. The current series of surveys has been ongoing on a triennial basis since 1983. However, because the SCF conducted in 1986 was limited in scope, this article uses data only from the 1983 and 1989 surveys.

The data for both these surveys were collected by the Survey Research Center at the University of Michigan. The 1983 SCF was sponsored by the Federal Reserve in cooperation with several other agencies. This survey was described and information from it was reported in previous issues of the Federal

Reserve Bulletin (September 1984, December 1984, and March 1986). The data reported here for 1983 may differ from the figures reported in the earlier articles because of revisions of the data and of the sample weights. The 1989 SCF was sponsored by the Federal Reserve in cooperation with the Department of the Treasury, the Department of Health and Human Services, the National Institute on Aging, the Small Business Administration, the General Accounting Office, the Comptroller of the Currency, and the Congressional Joint Committee on Taxation. A technical description of this survey appears in the appendix to this article.

FAMILY INCOME

According to the SCF, real median pre-tax income for families was virtually unchanged between 1983 and 1989 (table 1). This finding is supported by data from the Current Population Survey conducted by the Bureau of the Census. Over the same period, mean real family income rose from $33,400 to $35,700. These findings suggest that incomes above the median grew faster than those below the median and that the distribution of family income became somewhat more concentrated among families with higher income.

Although in the aggregate median family income did not change, a breakdown of the population by demographic groups shows mixed changes. The median income for families headed by persons with at least some college experience rose, but this increase was offset by declines in all other education categories. A moderate increase for white families was offset by a decline for nonwhite and Hispanic families. The difference by racial group may partly reflect changes in the way race was ascertained in the two surveys. In the 1983 SCF, race was based on the survey interviewer's observation, whereas in the 1989 SCF, the survey respondent reported his or her own race, the procedure routinely followed by the Census Bureau. Consequently, the 1989 SCF race classification matches Census estimates, whereas the 1983 data overstate the proportion of nonwhites and Hispanics. This difference may cast some doubt on comparisons between 1983 and 1989 SCF results based on race classifications.

Consistent with aggregate income and population data, the SCF shows that the aggregate mean family income rose. Increases for families headed by persons less than 55 years old were partly offset by declines for families with heads between 55 and 74 years of age. Workers in highly skilled occupations fared better than other workers. Mean family income increased for managers; technical workers; and precision production, craft, and repair workers. Mean income of homeowners and of whites rose, whereas that of renters and of nonwhites and Hispanics fell. In the age group under 55 years, mean income rose most rapidly for married couples; among older families, mean income rose for families with retired heads and declined for other families.(3)

NET WORTH

Changes in the overall real net worth of families - the difference between families' total assets and their total debts-were more dramatic than changes in family income. Mean real net worth rose more than 23 percent, whereas median real net worth rose only 11 percent (table 2) As with income, the contrast between the mean and median suggests an increase in the concentration of net worth among wealthy families.

While small increases in median net worth were widespread, some marked declines also occurred. The decline was especially sharp for single parents and for nonwhites and Hispanics. The median net worth of farm families fell sharply as prices for farmland plummeted in the early 1980s. (Even with an increase in the price of farmland after 1985, the price level of farmland in 1989 was about 30 percent below the level in 1983.) Among families headed by persons 55 years of age and over, those whose heads were retired had the only increase in median net worth. Families with heads in this age group who remained in the labor force experienced declines in median net worth.

The change in mean net worth varied widely across demographic groups. Families headed by persons having some higher education and married couples headed by persons under 55 years of age showed particularly large increases. In contrast to the decline in median net worth for nonwhites and Hispanics, the mean net worth of this group rose; indeed, the percentage increase exceeded that of white families.

Changes in net worth may be attributed to changes in the value of the holdings of assets or in the amount of debt carried. The coverage of assets and liabilities in the survey is highly detailed; the aggregations presented in this article have been chosen to illustrate trends in the data.

ASSETS

The composition of assets held by all families, as recorded in the SCF, changed slightly between 1983 and 1989 (table 3). The small shift away from nonfinancial assets reflects declines in business and real estate assets as a share of total assets.(5)

Financial Assets

Despite the increase in financial assets as a proportion of total family assets, the ownership of several types of financial assets fell sharply (tables 3 and 4). In 1983, nearly 79 percent of families owned checking accounts-including conventional checking accounts, NOW accounts, and money market accounts used for checking - whereas by 1989 about 75 percent of families owned them. The decline was spread over virtually all groups shown here. Over this period, the movement of financial institutions toward more explicit prices for checking services may have made these accounts less attractive to many families. Nevertheless, data from the SCF indicate that only 16 percent of families without checking accounts in 1989 reported that service fees or balance requirements deterred them from maintaining an account. The proportion of families owning savings accounts also fell dramatically over this period, partly because families shifted assets to other investment vehicles. Again, the decline was pervasive. While the proportion of families holding checking and savings accounts declined over the period, the families that held these accounts in 1989 maintained them at about the same real level as that in 1983, with the median value of checking accounts rising only from $600 to $900 (tables 4 and 5).

The proportion of families owning money market accounts, including money market mutual funds and money market deposit accounts not used as a checking account, rose. With the increasing ownership of money market accounts, the median size of account declined from $1 1,000 to $5,000, whereas the overall mean rose slightly from 27,400 to $28,300. The decrease in the median amount held in these accounts, with the increase in the rate of ownership, indicates that the number of small money market accounts increased.

The proportion of families owning retirement accounts rose; these accounts include individual retirement and Keogh accounts as well as employer-sponsored pension accounts from which withdrawals could be made. The median size of retirement accounts doubled. In part, this growth reflected the overall shift in employer-provided pensions from defined-benefit plans, in which the employer pays a regular benefit beginning at retirement, to defined-contribution and 401(k)-type plans, in which employees have rights to a specific pool of funds at retirement.(6) While the median retirement account grew in all categories shown, the increase for families with incomes of 50,000 and more was particularly dramatic. The ownership of other financial assets, which include savings bonds and the cash value of life insurance, also grew.

The financial assets discussed thus far were held in insured depository institutions-which include banks, savings banks, savings and loans, and credit unions - or in uninsured institutions. The percentage of families owning any type of account at a depository institution changed little over the period (not shown in table). The holdings of any account in an insured depository institution fell slightly; those of families with income of less than $ 1 0,000 showed large declines. Nevertheless, for families maintaining accounts, the median value of holdings at depository institutions increased for virtually all groups.

Stock ownership, which includes holdings of publicly traded corporate stock and of equity mutual funds, declined slightly. The median value of holdings of stocks rose overall, with a noticeable drop both in ownership and in median holdings for families with incomes of 50,000 and more. Median stock holdings also declined for families headed by persons between 35 and 44 years of age. Aggregate mean holdings of stock fell sharply from $75,600 to $57,200. For all families, mean holdings of both taxable and nontaxable bonds, which include bonds held directly and bonds held through mutual funds, fell. However, median holdings of taxable bonds increased, and median holdings of tax-exempt bonds decreased. These changes in median holdings may reflect changes in the tax code that have lessened the relative advantage of nontaxable bonds.

Overall, the proportion of families owning any type of financial asset fell slightly. Nevertheless, for families still holding financial assets, median holdings more than doubled, whereas mean holdings increased only about 40 percent. These changes suggest a decline in concentration of financial assets.

Nonfinancial Assets

The largest single asset of most families is their principal residence. Rates of homeownership for all families were little changed between 1983 and 1989 (table 6). However, there were sizable declines in ownership for the lowest income group and for single parents. Rates of homeownership for childless married couples increased substantially as did rates for families with heads between 55 and 64 years of age. For all families, the median real home value for homeowners increased from $64,700 to $70,000, an 8 percent rise, whereas die mean increased from $87,500 to $107,400, a 23 percent rise (table 7). The increase occurred largely in the group of families with incomes of 50,000 and more.

The ownership of investment real estate, including all properties from vacation cabins to shopping centers, also changed little over the period, whereas the median holdings fell about I 1 percent. By income group, changes in ownership were mixed, with a notable decline for families with incomes of 50,000 and more. For this group, the median size of holdings in investment real estate also declined. The median investment in real estate increased for families without children headed by persons under age 55.

Ownership of an equity interest in a business, including limited partnerships, other partnerships, corporations, sole proprietorships, and other private businesses, fell substantially. The value of holdings fell for families in the top and the bottom income categories. Particularly striking was the decline of the real value of median holdings of businesses by nonwhites and Hispanics from $49,000 to $8,400. This decline, with a decrease in the rates of business ownership, reflects the exit of minority investors from the market.

While the overall ownership of vehicles, including automobiles, trucks, campers, boats, and other vehicles, was virtually unchanged, some individual categories did exhibit relatively large shifts. Two groups showing marked declines in rates of ownership were families with younger heads and nonwhites and Hispanics. Nevertheless, real median gross holdings for those with vehicles rose from $5, 100 to $6,900: This rise reflects the increase in the real cost of automobiles.

This comparison of mean and median values for individual categories of nonfinancial assets suggests that the concentration of all such nonfinancial assets increased. For homes and vehicles, the most widely held nonfinancial assets, both mean and median holdings grew. The median values for homes and vehicles grew 8 percent and 35 percent respectively, while the mean values grew faster - 23 percent and 43 percent respectively. This difference suggests that the top half of the distribution shifted upward faster than the bottom half. In other words, nonfinancial assets became more concentrated at the top end of the distribution. For the remaining nonfinancial assets, the median values actually fell while the mean values increased, a finding that suggests even greater increases in concentration. Overall, both the median and mean values of total nonfinancial assets increased. This fact indicates that the increases in home and vehicle assets more than offset the declines in the median holdings of less widely held assets.

The largest percentage increase in the median value of holdings was for the group with incomes of $50,000 and more. The only other sizable increase was reported for married couples with no children. Large declines were observed for single parents.

LIABILITIES

Families hold many types of debt, including mortgage and credit-card debt, lines of credit, automobile loans, and closed-end and other debts. While the largest nonfinancial asset for most families is their principal residence, their largest financial obligation is their mortgage for it. In light of changes in the tax treatment of consumer interest payments, one might expect families to increase the proportion of their debt that is secured by their principal residence. However, the real size of mortgages for principal residences plus home equity lines of credit actually declined as a proportion of total family debt between 1983 and 1989 (table 8). This decline occurred because, while the level of real total home equity debt increased, other forms of debt increased even more (tables 9 and 10). In particular, credit-card debt and automobile loans grew substantially over this period.

The proportion of families owning homes rose slightly, but the proportion of families with mortgages rose faster, from 37 percent to almost 39 percent (tables 6 and 9). The median size of a mortgage rose from $27,000 to $32,000 table 10). The two lowest income groups reduced the median value of their debt outstanding. In contrast, the three highest income groups increased the median size of their mortgage debt. The sizable decline in ownership of mortgage debt among single parents stands out from a picture of mortgage holding that otherwise shows little change.

The use of home equity lines of credit, another form of mortgage debt, rose strongly over this period; it was concentrated in the high-income groups. The median value of debt of this type increased substantially whereas the mean value of debt rose only slightly. This difference suggests that large home-equity lines of credit shrank as a proportion of afl such debt. While families may have substituted home equity lines of credit for other types of credit, the overall proportion of families using any line of credit fell from 11 3/4 percent in 1983 to 6 1/2 percent in 1989.

The ownership of investment real estate declined, and the proportion of families in virtually all income groups with investment real estate debt fell. At the same time, the median debt secured by investment real estate rose; it increased in all income groups but the middle. Median debt declined substantially for families with heads over 55 years of age.

For most of the groups examined here, the proportion of families holding credit-card debt increased. A notable exception was the category of nonwhite and Hispanic families, for whom the incidence and the median holding of credit-card debt declined. In the life-cycle categories, only for single parents did the incidence of credit-card debt decrease noticeably, whereas for all groups the median level of debt grew.

Despite a tax treatment of automobile loans that was less favorable in 1989 than in 1983, with the increasing real cost of purchasing automobiles both the incidence and the median amount of car loans increased for most groups. As with car loans, the median amount of other debt, including closed-end consumer debt other than auto loans and loans against pension plans and insurance, grew between 1983 and 1989. The increases were widespread.

The proportion of families holding any type of debt rose about 3 percentage points; the median real amount of debt held rose 13 percent; and the mean real debt rose by 42 percent, from 31,700 to $45,000. While the median debt rose for all income groups, the rise was most marked for those with incomes of 20,000-$29,000, a group for which debt increased nearly 50 percent. The median debt for childless couples with heads under 55 years of age more than doubled.

The 1989 SCF collected data on the purpose of all loans except credit card loans and first mortgages. The 1983 SCF has no comparable data. In 1989, 44 percent of the dollar amount of this debt category was associated with purchases of vehicles; about 20 percent was used for investment. Of the remaining debt, 9 percent was for outlays associated with remodeling and repairs, and 9 percent was for education expenses.

Ability To PAY

Given the magnitude of the increase in debt and the absence of change in median income between 1983 and 1989, questions arise about the ability of families to repay their loans. Indeed, data from the SCF indicate that the proportion of families reporting that they were a month or more late in making loan payments rose from 13 percent to 15 percent between 1983 and 1989.

One measure of consumers' ability to repay loans is the ratio of nonmortgage debt payments to income; nomnortgage debt refers to loans other than those on the principal residence (table 11).(7) For all income and demographic groups reported here, the median payment-to-income ratio shifted upward between 1983 and 1989. The shift seems fairly uniform across groups, except those regarding income.

Historically, the payment-to-income ratio has tended to be lower for families with higher income. Indeed, for families in the highest third of the payment-to-income distribution, those with incomes of less than $10,000 had a payment-to-income ratio of 32 percent in 1983 whereas those with incomes of $50,000 and more had a ratio of only 14 percent. By 1989, however, the ratio for the lowest income group had risen to nearly 40 percent and that for the highest income group had risen to 19 percent. The relation of family income to debt payments suggests that the payment burden grew most rapidly among those families with the greatest ability to pay. Nevertheless, the aggregate risk of default could rise if the incomes of these families fell.

Consumers may repay their loans out of their financial assets. In both 1983 and 1989, only about 4 percent of all families were both among the largest nonmortgage debt holders and among the smallest financial asset holders (table 12). In addition, in 1983, only 8.6 percent of all families were both among the largest nonmortgage debt holders and among the largest financial asset holders. By 1989, this proportion had risen to 9.4 percent. Comparing the distribution of total nonmortgage debt and financial assets in 1983 with that in 1989 (not shown in the table) indicates that the largest rise in such debt occurred for families with the greatest holdings of financial assets. This group held 56 percent of all nonmortgage debt in 1983 and 67 percent in 1989.

APPENDIX: DESIGN OF THE 1989 SURVEY OF Consumer FINANCES

The 1989 SCF was designed to gather family-level information closely comparable to that obtained in the 1983 SCF. It collected detailed data on the composition of family balance sheets, the terms of loans, and relationships with financial institutions. It also gathered information on the employment history and pension rights of the survey respondent and the spouse or partner of the respondent and on other demographic, economic, and attitudinal variables.

The survey attempts to provide an accurate representation of the distribution of these variables across U. S. families. To that end, it is distinguished from other household surveys by virtue of its sample design. Research has indicated that the distributions of income and net worth are skewed, with a relatively small proportion of families having a disproportionately large share of both income and net worth. To obtain a sufficiently large and unbiased sample of wealthier families, the SCF employs a two-part strategy for sampling families. Of the 3,143 families in the sample used in this article, 2,277 were selected by standard multistage area-probability sampling methods from the forty-eight contiguous states. The remaining 866 families in the survey were selected using tax data under the strict rules governing confidentiality and the rights of potential respondents to refuse participation. This second group of families was selected specifically to oversample wealthier families.(8)

The definition of "family" used in the SCF differs from that commonly used by the Bureau of the Census. Census families exclude single individuals, who are separately classified as unrelated individuals."

In the SCF, a given household is divided into a primary economic unit and other economic units. The primary economic unit, which may be a single individual, is generally chosen as the unit that contains the person who either holds the title to the home or is the first person listed on the lease. The primary unit is used as the reference family.

Data for the survey were collected by the Survey Research Center at the University of Michigan between August 1989 and March 1990. The achieved sample of 3,143 families represents a response rate of about 69 percent in the area-probability sample and a rate of about 34 percent in the tax-data sample. Although the response rate for the tax-data sample may appear low by conventional standards, analysis of separate data shows that the sample provides an unbiased representation of key income variables.

In any survey, there are several potential sources of error, including inaccurate survey responses, nonresponses to the entire survey or to particular questions in the survey, and errors due to sampling. Response errors, which are difficult to measure, are best controlled through the proper training of interviewers and the careful design of questionnaires. Most nonresponse errors arise either because a respondent is uncomfortable about providing information or because the respondent does not know the information being asked. In the SCF, adjustments for nonresponse errors are made through systematic imputation of unanswered questions and through weighting adjustments based on data used in the sample design for families that refused participation.(9) Sampling error arises any time survey estimates are based on a sample rather than on a census. Because of the complex design of the SCF, the estimation of potential sampling errors is not straightforward.(10)

The sampling weights used in the calculations reported in this article were produced by the Survey Research Center at the University of Michigan and were based on the probability design of the original sample. These weights have been adjusted with data from the Bureau of the Census to reflect aggregate information available on the age and geographic distribution and on homeownership patterns of the U.S. population. These weights were further adjusted to minimize the influence of extreme cases on the estimation of net worth.

The SCF data are available to the public. Copies of the survey are available on magnetic tape from the National Technical Information Service, Federal Computer Products Center, 5285 Port Royal Road, Springfield, VA 22161 or [phone].

1. When the obserbations in a set of data are arranged in order of magnitude from lowest to highest, the middle value is the median. The sum of all observed values divided by the number of observations in a set of data is the mean. The mean and the median each have advantages for describing distributions of income and other financial variables. Because of the focus on the ranking of values, the median is not influenced by extremely large values; thus it is a good indicator of the position of the "typical" family. However, the mean is a better indicator of the dollar amount of income or assets held by all families because it considers all values, the extreme as well as the typical. If the median of a distribution drops, the mean may rise or fall, as may the means of the upper and lower halves of the distribution. For the variables examined in this article, the mean of the lower half of each distribution generally drops when the median does. Thus, with before-tax family income, the near-constancy of the median plus the rise in the overall mean implies that the mean value of the distribution above the median increased - that is, the concentration of income increased in the top half of the distribution.

2. All dollar figures reported in the article are adjusted to 1989 levels using the consumer price index (CPI) for all urban consumers. There are problems with the use of any deflator to compare income, assets, and liabilities over time, particularly where changes in the interest rate implicitly change the costs of debts. The CPI is used for two reasons. First, it is widely known and applied. Second, because it is intended to reflect the relative costs of a standardized collection of goods and services, it seems a natural indicator of the importance of price level changes for most families.

3. Calculations not reported in the tables indicate that the increase in mean income for dual-earner couples more than accounted for the overall mean increase in income for married couples.

4. Data from the 1984 and 1988 Surveys of Income and Program Participation conducted by the Bureau of the Census show a similar change in real median net worth.

5. Any division of assets into financial and nonfinancial categories is somewhat arbitrary. The classifications used here may differ from others, such as those in the flow of funds accounts, which are published by the Federal Reserve.

6. The SCF collects financial data on all types of pensions. However, for the analysis of family finances, we include only retirement accounts from which funds can be withdrawn before retirement, even if with a penalty. These are generally defined-contribution or 401(k)-type plans.

7. Mortgage debt is excluded because those families not making mortgage payments generally pay rent. Median payments for housing (both mortgage and rent) increased from 1,900 per year in 1983 to $3,000 per year in 1989; these payments relative to income rose from 9 percent to 13 percent over the period.

8. A detailed description of the sample design is given in S. Heeringa and L. Woodbum, Sample Design Documentation, 1989 Survey of Consumer Finances,' memo, Institute for Social Research, University of Michigan, 1991.

9. A description of the imputation procedures is given in A. Kennickell, "Imputation of the 1989 Survey of Consumer Finances: Stochastic Relaxation and Multiple Imputation," Proceedings of the Section on Survey Research Methods, American Statistical Association, 1991. See L. Woodburn, "Using Auxiliary Information to Investigate Nonresponse Bias,' unpublished paper(1991), Statistics of Income Division, Internal Revenue Service, for a description of response bias in the SOI sample; and Heeringa and Woodburn for a description of overall weighting adjustments.

10. Selected estimates of sampling errors are given in A. Kennickell and L. Woodburn, 'Estimates of the Variance of Changes in the Distribution of Net Worth," memo, Federal Reserve Board, 1991.