Household Saving and Portfolio Change: Evidence from the 1983-89 SCF Panel

Household Saving and Portfolio Change: Evidence from the 1983-89 SCF Panel Arthur B. Kennickell and Martha Starr-McCluer Federal Reserve Board of Governors April 1996 Abstract There are very few sources of high-quality data on the dynamics of wealth accumulation. This paper uses newly-available data from the 1983-89 panel of the Survey of Consumer Finances to examine household saving and portfolio change over the 1980s. The 1983 SCF collected detailed information on households’ assets, liabilities, income and other characteristics for a sample of 4,103 families. In 1989, 1,479 of these families were re-interviewed using a similar questionnaire. After describing the sample and methodology of the panel survey, we analyze changes in household wealth over the 1983-89 period. We also investigate changes in the structure of households’ assets and debts. Please address correspondence to: Martha Starr-McCluer, Federal Reserve Board, Stop 180, Washington, D.C. 20551, phone (202) 452-3587 We are grateful to Carol Bertaut, Gerhard Fries and Annika Sunden for valuable comments, and to Kevin Moore and Diane Whitmore for excellent research assistance. The views expressed in this paper do not necessarily reflect those of the Board of Governors or its staff. /

Household Saving and Portfolio Change: Evidence from the 1983-89 SCF Panel I. Introduction This paper uses data from the 1983-1989 panel of the Survey of Consumer Finances to examine household saving and portfolio change over the 1980s. This survey provides unique information for studying these issues. There are very few sources of high-quality data on the dynamics of wealth accumulation. Some of the major household surveys provide valuable information on the savings behavior of typical households--for example, the Survey of Income and Program Participation (SIPP) and the 1984 and 1989 waves of the Panel Survey of Income Dynamics (PSID) . But these surveys are not specifically intended to measure wealth, and provide little detail on changes in the composition of household portfolios over time. Some recent panel studies have begun collecting detailed data on household wealth, notably the Health and Retirement Survey (HRS) and Asset and Health Dynamics Among the Oldest Old (AHEAD). However, the samples are confined to households in particular age ranges, and thus provide only limited insights into the ways in which savings behavior may change over the life cycle. Moreover, because there is evidence that survey nonresponse is correlated with wealth (Kennickell and McManus 1994) , estimates of many important wealth characteristics will be biased unless specific measures can be taken to adjust for such nonresponse. The Survey of Consumer Finances (SCF) collects detailed information on households’ assets, liabilities, income and other characteristics , and has a special sample designed specifically to support wealth estimation. The main goal of the SCF is to provide accurate cross-sectional information on families’ financial situations . However, a subset of the families who were interviewed for the 1983 SCF were re-interviewed in 1989, providing an opportunity for studying the dynamics of wealth accumulation with a nationallyrepresentative sample.

-2- The 1983-89 period is interesting for several reasons. First, various legal changes encouraged the growth of tax-deferred savings vehicles such as Individual Retirement Accounts (IRAS) and 401(k) -type retirement accounts. While such accounts came to represent a sizable share of households’ financial assets, it is unclear whether the growth represented higher saving by households, or simply a shift of savings into tax-deferred forms (Poterba, Venti and Wise 1992; Engen, Gale and Scholz 1994). Second, household debt rose substantially over the 1980s, while tax changes eliminated deductions for non-mortgage interest payments (Canner, Kennickell and Luckett 1995) . Third, real wages were virtually flat for those with high school education or less, but they rose considerably for workers with college educations (Levy and Murnane 1992) . Fourth, overall stock prices rose substantially over the period, despite a sharp decline during the stock market crash of October 1987. Finally, although median wealth increased over the 1980s, there appeared to be a substantial increase in the inequality of the distribution of household wealth (Kennickell and Woodburn 1992). This paper is organized as follows. Section II discusses the background, sample and methodology of the 1983-89 SCF panel. Section III uses the data to describe changes in wealth and the distribution of those changes. Section IV examines the determinants of saving, estimating models of changes in net worth as a function of a set of explanatory variables. Section V analyzes changes in the composition of households’ portfolios. The final section offers some conclusions and points toward future research. II. Back~round, Sam~le and Methodolo~v Background. The current series of SCFS has been conducted every three years since 1983, under the sponsorship of the Federal Reserve Board with the cooperation of Statistics of Income (S01) at the Internal Revenue Service. To represent accurately the full distribution of household wealth, the 1983 SCF had a dual-frame sample design.1 The first part consisted of a standard multi-stage area- 1. The 1983 sample is described in Avery, Elliehausen and Kennickell (1988). For a general description of the 1983 SCF, see Avery, Elliehausen and Canner (1984).

-3probability (AP) sample, intended to provide good coverage of assets and liabilities that are broadly distributed in the population (such as vehicles and mortgages) . The second part was a list sample drawn from a file of taxpayers maintained by SOI, using a procedure that 2 oversampled households with relatively high incomes. The list sample considerably improves the precision of estimates of assets and liabilities that are held by relatively wealthy households, such as 3 stocks and bonds. Altogether, the sample interviewed for the 1983 SCF included 4,103 households, of whom 438 came from the list sample. In addition, the information from the SOI file provided a way to make systematic nonresponse adjustments, mitigating the problem of bias due to differential nonresponse. Panel sample. The 1989 SCF had a complex sample design 4 intended to provide both cross-section and panel data. The panel part of the survey was based on a subsample of the respondents to the 1983 SCF. If a person living at one of a subset of AP addresses was a respondent to the 1983 SCF, or was the spouse or partner of that person, an attempt was made to obtain an interview.5 A subsample 6 of original AP respondents who had moved was also pursued. Efforts were made to secure interviews with all list sample respondents . Divorced, widowed and separated spouses of list sample 2. According to the conditions for access to the list sample, to be included in the 1983 SCF, the potential list respondents were asked to return a postcard if they were willing to be interviewed. Only about 10 percent of these cases returned postcards. Since the 1989 SCF, list sample cases have been asked to return a postcard only if they did not wish to be interviewed, and unsurprisingly, the response rate improved. 3. See Avery, Elliehausen and Kennickell (1988) and Curtin, Juster and Morgan (1989) for comparisons of the 1983 SCF with other sources of wealth data. 4. Briefly, the 1989 survey consisted of three major parts: (1) a completely new cross-section survey, with list and area-probability samples, (2) list sample respondents to the 1983 survey, who have only panel representation in the 1989 survey, and (3) a subsample of the names and addresses of area-probability cases from the 1983 SCF (see text) . There was also a small additional sample included to account for new construction. For details, see Heeringa, Connor and Woodburn (1994), and Kennickell and McManus (1994). 5. The geographic scope of the AP was limited to the “half sample” of PSUS maintained by the Survey Research Center at the University of Michigan. 6. Among the AP households, all households with heads over age 45 in 1983 were followed, but only 25 percent of younger movers were followed.

-4respondents were also eligible to be included in the panel. In total, 1,479 households--including 361 list sample cases--were reinterviewed in 1989. Panel weights. Weights are a critical link between the data and their interpretation. Because some important assumptions have been made in the construction of the panel weights, we provide a brief 7 summary here. There are four main steps in the calculation of the panel weights . First, the “FRB final merged sample weight” from the 1983 SCF (B3016) is adjusted for the systematic part of the panel sample 8 selection. Had the adjustment been made for all cases selected from the 1983 sample for the panel, the sum of the weights would be very close to the number of households with heads aged 22 and over in 1983. Since data are available only for people who completed interviews , a second adjustment attempts to account for the implicit selection process imposed by nonresponse. Weights are resealed using a set of nonresponse factors that condition on a number of important financial and demographic characteristics of respondents.9 At this stage, the estimates of key financial variables in the panel are different from estimates for comparable age groups in the 1983 and 1989 cross-sections to a degree that could not be explained by simple sampling error or other such sources. In particular, panel households generally look much wealthier in 1983 than the comparable age subset of the 1983 cross-section. Moreover, the difference actually widens when the 1989 wealth of the panel is compared to the 1989 cross-section, suggesting a “success bias” in the retention of households in the sample. Since one of the most 7. Kennickell and Woodburn (1996) discuss the construction of the panel weights in detail. 8. Avery, Elliehausen and Kennickell (1988) explain the construction of the 1983 weight. 9. Heeringa (1993) describes the choice of adjustment cells. In light of the other adjustments to the weights that we discuss below, it might seem desirable to try other approaches to nonresponse adjustment at this stage. Unfortunately, the complete list of cases (respondents and nonrespondents) selected for the panel was not available to us. Thus , it is not known which 1983 cases are absent from the panel because they were not sampled, and which are absent because they declined to participate.

-5interesting uses of this panel sample is to examine changes in wealth, such a problem raises critical questions. Because the nonresponse adjustments discussed above already condition on observable data within the sample, an appropriate recourse for dealing with the unobserved factors that drive this implicit selection is to use information outside the panel. The 1983 and 1989 cross-sections are a natural place to turn for such data. Ignoring sampling error (and problems induced by the death of 1983 respondents), the panel and the 1983 cross-section should produce the same estimates. Although the 1989 cross-section is more independent of the panel sample, there is some overlap in terms of interviews, and all interviews in 1989 were conducted with questionnaires and procedures that were almost identical. At the third stage of adjustment, the weights are post-stratified in turn by 1983 income, 1983 homeownership, 1989 age, and 1983 and 1989 net 10 worth. The net worth adjustments are applied iteratively (raked) to ensure that the percentage of the population in different wealth groups in each year is approximately the same as that implied by the 11 cross-sections. Editing and imputation. Survey data on wealth typically contain a fair amount of missing or incomplete information. Some survey respondents are unable or unwilling to report details of their financial situations. Sometimes respondents’ answers are recorded or processed with error. Traditionally, the SCF has addressed these problems through systematic data editing and statistical imputation of missing data values. Processing of panel data involves some additional complications. There is an immense amount of information in the 1983 and 1989 surveys that might be used to edit the data, as 12 well as some information from a brief 1986 reinterview. Moreover, the panel interviews provide information that was not used in the 10. See Little (1993) for a discussion of post-stratification. 11. The weights are also slightly smoothed: 12. In 1986, some 2,822 of the households from the 1983 SCF were reinterviewed by phone, and asked to provide summary information on their assets and liabilities. Analysis of the 1986 data suggested that holdings are reported much less accurately when information is collected in summary form (Avery and Kennickell 1991) .

-6original cross-section imputations, so the missing data could be reimputed using a broader set of conditioning variables. To keep the processing of the data manageable, the editing and imputation of the panel data had to be limited in significant 13 ways . First, rather than use all of the detailed information in the 1983 SCF, variables summarizing holdings of the main types of assets and liabilities were used for editing and imputation. Second, questions asked in 1989 about changes in assets or liabilities between 1983 and 1989 were not used in editing or imputation, because this information was too often inconsistent with the information on current holdings (Kennickell and Starr-McCluer 1995). Finally, the data from the 1986 re-interview were used only for minor editing. The missing values in the panel data were imputed, using the multiple imputation technique developed for the SCF (Kennickell 1991).14 This method preserves the first and second moments of the data and allows estimation of the uncertainty of the imputations. The imputed values in the panel data may differ from imputations in the cross-section data, because a broader set of variables is used to condition the imputations. Representativeness of the data. It is important to emphasize the nature of the population included in the panel sample. The sample design specifically excluded households with heads under the age of 22 in 1983. Because many people below this age are in college or the military, or live with their families, the independent households with heads who were in this age group in 1983 do not represent very well the set of independent households with heads aged less than 28 in 1989.15 Also, the panel does not include individuals who immigrated 13. A detailed description of the panel processing is contained in Kennickell and McManus (1994). 14. See Rubin (1987) for an explanation of multiple imputation. The method used to impute the SCF data is based on a Gibbs sampling technique. The panel data are imputed three times. 15. The cut-off of age 22 is somewhat arbitrary, since some fraction of every age group may be temporarily in an institution or living as a secondary family member. However, the fraction declines substantially after the early 20s. A small number of under-28 households appear in the panel sample, because the sample design followed both halves of couples that separated since 1983.

-7to the U.S. during the 1983-89 period, unless they became a part of an 16 existing sample family. There are other, more complex reasons that statistics computed using the panel data may differ from comparable estimates using the 1983 and 1989 cross-section data. Because the panel is smaller than either of the two cross-sections, it provides less efficient estimates of population statistics. Various other factors, such as differing imputations may explain some small part of differences . However, the most important factor may be that the panel consists of people who were systematically sampled from the respondents to the 1983 cross-section, who could be located in 1989 and who were willing to be reinterviewed. Adjustments to weights can remove the effects of systematic selection and part of the implicit selection induced by the other observable factors. But the ability to find and reinterview households may also depend on unobservable factors, such as employment or marital stability, or financial success , which may well be correlated with changes in wealth. III. Changes in Household Wealth. 1983-89 A. chan~es in the Incomes and Net Worth of Individua1 Families Because income is an important determinant of wealth, it is useful to describe first the changes in real family income for the panel sample that occurred over the period. Table 1 shows mean and median income in 1983 and 1989 in terms of a number of family characteristics . In the panel sample as a whole, mean income rose from $33,400 in 1983 to $36,800 in 1989, while the median increased slightly from $24,900 to $25,100, largely reflecting the effects of the aging of the panel (all values are given in 1989 dollars) . Similar panel aging effects occur throughout the analysis that follows. 16. Additionally, some respondents may have died between 1983 and 1989. A response code indicates cases for which this fact is known. However, this method is unlikely to identify all deceased respondents, because the interviewer may not have spoken with someone who knew that the respondent had died--a situation more likely for lower-income respondents , or those who had not lived a long time in a given community.

-8- When families are grouped by their 1983 incomes, mean income increased for all groups except the group with income between $50,000 and $100,000, for whom mean income declined. Median income rose for the groups with incomes below $50,000, while declining for those with incomes above that level. Consistent with the tendency for labor income to rise into middle age, mean and median income rose over the period for families headed by persons below age 45. For other age groups , changes were more mixed, with median income declining for all groups with heads age 45 and over. In terms of education, mean and median income declined for families headed by a person without a high school degree, while increasing slightly for families in which the head was a high school graduate, and increasing more appreciably for families where the head had a college degree. This is consistent with evidence of rising returns to education over the 1980s (Levy and Murnane 1992). When families are grouped by their 1983 net worth, mean income increased most strongly for the bottom and top groups. However, median income rose for families in the lower two quartiles of the wealth distribution, but declined for families in other groups. Table 2 shows how mean and median levels of net worth changed for the panel sample between 1983 and 1989. Net worth is defined as the sum of a family’s financial and nonfinancial assets, minus all of its debts. Among the panel sample, mean net worth increased from $142,600 in 1983 to $187,600 in 1989, while median net worth rose from $43,300 to $56,600. The fact that the mean and median both increased by around 30 percent suggests little change in the concentration of wealth ownership over the period. We return to the issue of wealth concentration in the next section of the paper. When grouped by 1983 income, the age or education of the head, or 1983 net worth, mean net worth rose strongly for all groups. A similar pattern also arises for the median, but with some significant exceptions. Median wealth declined for both the 55-64 and the 75-and-over age groups, as well as the top decile of the 1983 net worth distribution. Increases in net worth were particularly noteworthy for families headed by persons with college degrees. The means also increased strikingly in the bottom half of the 1983 wealth distribution.

-9- Table 3 shows some direct measures of changes in wealth for the panel sample. The first two columns show the mean and median change in net worth between 1983 and 1989. By mathematical identity, the mean changes exactly parallel the differences in the mean levels given in table 2. On average, families’ net worth rose by $45,000 over the period. The median change of $7,600 was much lower than the mean change, and also much lower than the change in the overall median ($13,300). While median changes in net worth ranged between $10,000 and $15,000 for families in the under-55 age groups) median changes were negligible for families in older age groups. Families in the top 10 percent of the 1983 wealth distribution saw a median change in net worth of -$75,300. To provide a sense of each group’s contribution to total household saving, column (4) shows each group’s share of the total change in net worth in the panel sample. Despite the median decline in net worth for the top 10 percent of the 1983 wealth distribution, that group contributed nearly 25 percent of all saving by families between 1983 and 1989. The top two income groups contributed over 60 percent of total saving. Families with college-educated heads accounted for 67.0 percent of all saving, although they represented only 24.4 percent of all families. Families with heads in the 45-54 age range also accounted for a disproportionate share of saving. Finally, it is also interesting to examine saving relative to income. To estimate saving rates, one would ideally want information on consumption spending, as well as on net worth and income over the period. However, the panel data contain no information on consumption spending. As an alternative, it is possible to approximate saving rates from information on net worth and income in 1983 and 1989, using some assumptions about interest rates and income growth. Specifically, assuming that the real interest rate is constant and equal to r, family i’s 1989 wealth may be expressed as 5 ‘i,1989 = ‘i,1983 [l+r16 + X S ~983+t[l+r] 6-t [1] i t=o ‘

-1owhere W 1983 is the family’s 1983 wealth, and S ~983+t is saving i i, during y~ar (1983+t). If the family’s saving rate is constant and equal to ~i, its saving in period T would be s =aiYiT [2] i,T , where Y T is family income in year T. Assuming that Y t grew at a i i constant’annual rate of gi between 1983 and 1989, then i~come for the years between 1984 and 1988 could be estimated as Y . Y [31 i,T i,1983[1+gi]T 1/6 where (l+gi) = [ Y ~989/ Y 1983 1 . Then cti could be calculated P i , i as - w ‘i,1989 i,1983(1+r)6 ai = [4] 5 X (l+r)6-t(l+g)t ‘i ,1983 t=() Table 4 presents estimates of annual saving rates based on [4], assuming a real interest rate of 2 percent.17 Columns (1) and (2) show the mean and median savings rate for families within each group. For the panel sample as a whole, the median annual savings rate over the six-year period is 3.1 percent. Both mean and median saving rates rise strongly with income. For example, families with incomes below $10,000 in 1983 had a median savings rate of 0.1 percent, versus 16.1 percent for families with 1983 incomes above $100,000. There is a similar positive relationship between education and saving rates. At least in terms of medians, there is some evidence of life cycle saving, with families below age 55 having positive saving rates and those above that age having negative rates. In terms of 1983 wealth, median saving rates were highest for families 17. Note that unrealized capital gains and losses are not included in the income measure.

-11in the middle of the wealth distribution. The median rate was negative for families who had been in the top two groups in 1983. Column (3) shows an overall saving rate for each group, using the group’s total income and total wealth for the computation. The aggregate rates show qualitatively similar patterns, although the levels tend to be much higher. B. Changes in the Distribution of Net Worth Previous research using cross-section data has documented an increase in the concentration of wealth over the 1980s (see Kennickell and Woodburn 1992, or Wolff 1995) . Some analysts take the trend to imply that those who were wealthy at the outset of the period had become even wealthier by the end. However, this interpretation may or may not be correct: concentration could rise either because initially wealthy families tended to become wealthier, or because families who became wealthy amassed very high levels of wealth. It is not possible to distinguish between these two stories without information on the wealth changes of individual families. Table 5 presents information from the SCF panel on changes in the distribution of net worth by selected family characteristics. According to the panel data, the net worth distribution by age and 1983 income was fairly stable between 1983 and 1989. In both years, more than a third of total net worth was held by the 2.6 percent of families with incomes above $100,000, while over 45 percent of total net worth was held by the 31.1 percent of families with heads age 55 and older. In contrast, the share of wealth held by families with college-educated heads rose over the period, from 45.7 percent in 1983 to 50.8 percent in 1989. According to the panel data, the share of wealth held by families in the top 1 percent of the 1983 wealth distribution declined from 30.5 percent in 1983 to 25.4 percent in 1989, while the share held by families in the bottom half rose from 3.7 percent to 9.4 percent. The declining share of the top 1 percent seems largely due

-12to the declining values of businesses and real estate held by this group (see below). At first glance, this shift in the wealth distribution appears to be at odds with other evidence that wealth inequality rose over the 1983-89 period. Some part of the apparent discrepancy results from analyzing wealth shares in terms of families’ initial wealth, rather than current wealth. In terms of current wealth, the share of 1989 wealth held by families in the top 1 percent of the wealth distribution in 1989 was 31.6 percent--slightly higher than the 30.5 percent of 1983 wealth held by families in the top 1 percent in 1983 (table 6). This suggests a need for caution in using cross-section data to draw inferences about changes in wealth of. 18 individual families. The SCF panel can be used to look directly at the shifts of 19 individual families within the wealth distribution. Table 7 presents a transition matrix, showing where families fell in the 1989 wealth distribution relative to their position in 1983. The percentiles are defined using the weighted panel data. The data suggest a fair amount of persistence in the distribution of net worth over the period. Of the families in the lowest quartile of the wealth distribution in 1983, 67.2 percent were still in the lowest quartile in 1989; another 24.6 percent had moved into the second quartile; and only 8.2 percent shifted into the top half. Of those in the top 1 percent in 1983, 59.3 percent were still there in 1989: another 24.7 percent were still in the top 2-5 percent; and only 16.0 percent had moved down below that. There was somewhat more downward mobility among families originally in the top 2 to 5 percent; of these, 29.6 percent were no longer in the top 10 percent of the wealth distribution in 1989. But generally, a family’s 1983 wealth was highly correlated with its 1989 wealth, as indicated by a Spearman 18. Using the SCF cross-section data (for families in the panel age range) , the share of wealth held by the top 1 percent of families rose more appreciably, from 31.6 percent in 1983 to 33.1 percent in 1989. Conceivably, the more modest increase for the panel may reflect attrition related to wealth changes not fully accounted for in the panel weights. 19. The adjustments made to the weights described above might seem to impose the changes in wealth over the period. However, only the 1983 and 1989 marginal distributions are imposed, while the sample determines transitions, conditional on the marginal distributions.

-13correlation coefficient of 0.90.20 IV. Factors Ex~lainin~ Familv S aving Standard theories of saving behavior emphasize the roles of lifetime income, lifecycle factors and precautionary motives in explaining savings decisions. Prior inheritance or the desire to leave a bequest may also be involved. Some previous research examines the extent to which observed changes in wealth seem to reflect the concerns of standard theories, with fairly mixed results. For example, using data from the 1983-86 SCF and a comprehensive set of explanatory variables, Avery and Kennickell (1991) are only able to explain 7 to 8 percent of variation in saving in the 1983-86 sample. However, because wealth is likely to be measured with considerable error in the 1986 SCF, it is not clear whether this low explanatory power reflects noise in the data, or idiosyncratic factors in wealth accumulation, or some combination of the two. The 1983-89 SCF panel provides a good opportunity to reinvestigate this question. In particular, these surveys’ detailed questions on assets and liabilities, along with the careful cleaning and editing of the data, are likely to make problems with measurement error less influential than they might be in other surveys. This section presents regressions describing saving in terms of a comprehensive set of explanatory variables, intended to reflect the main concerns of standard theories of savings behavior. Detailed descriptions of the explanatory variables are given in table 8. The variables include measures of 1983 income; a measure of 1983-89 income growth; the age, education and race of the family head; 1983 wealth percentile; several variables indicating household composition and changes therein; indicators of events occurring between 1983 and 1989 that might be expected to affect wealth (a residential move, inheritance, deterioration in health, and expected or unexpected 20. The degree of income mobility was slightly greater than the degree of wealth mobility. For example, of families in the lowest quartile of the income distribution in 1983, 15 percent had shifted into the upper half of the income distribution by 1989; for wealth, the comparable share was 8.2 percent. The Spearman correlation coefficient for income was 0.86.

-14retirement) ; and a self-described measure of whether the family saves regularly. For the dependent variable, we use three different measures of family saving. The first measure (CHNW) is the absolute change in the family’s net worth between 1983 and 1989, expressed in 1989 dollars. The second measure (SAVRAT) is the estimated savings rate described in 111.A above, expressed as a ratio. The third measure (PCHNW) is the estimated percent change in wealth. PCHNW is calculated as the ratio of the change in net worth relative to the average of 1983 and 1989 net worth. Because these variables are highly skewed, and likely contain substantial measurement error in the tails of the distribution, it is important to use regression techniques that are not overly sensitive to influential 21 observations. Thus, we use median and robust regression to estimate the effects of the explanatory variables on these measures of 22 saving. Table 9 presents the results of the regression analysis. Regardless of the measure of saving used, saving was greater for families with higher levels of 1983 income, other things being equal. It was also positively related to income growth over the 1983-89 period. The age, education and race of the family head had few significant effects on saving, after controlling for income, initial wealth and other factors. Under some but not all measures of saving, having 1983 wealth in the bottom half of the distribution had a positive effect on saving, ceteris paribus. Having 1983 wealth in the top 10 percent was associated with significantly lower saving in all specifications . The variables indicating household composition had few significant effects of saving. An exception is families where the head and spouse were continuously married to each other; in most specifications , such families had significantly higher saving than 21. For example, in an OLS regression with CHNW as the dependent variable, 27 observations (of 1,479) have studentized residuals greater than 2 in absolute value. Deleting these observations affects the magnitude and significance of estimated coefficients. 22. The robust technique is that of Rousseeuw-Leroy, as implemented in Stata 4.0.

-15- 23 other types of families, other things being equal. Not surprisingly, families receiving an inheritance in the 1983-89 period had significantly higher levels of saving than others. In addition, families who reported that they saved regularly had significantly higher saving than others.24 Interestingly. saving did not differ by racelethnicity, once other characteristics are taken into account. Despite the significance of many coefficients in the saving regressions, the models explain only a small part of the variation in saving in the panel sample. The pseudo-R-squareds for the median regressions range between 0.05 and 0.11. The R-squareds from comparable OLS regressions range between 0.03 and 0.05. (There is no straightforward goodness-of-fit measure for the robust regressions) . Two factors may account for the low explanatory power of the regression models. First , even if problems with the measurement error are smaller in the 1983-89 panel than in other panel data sets, the level of noise in the wealth data may still be substantial. Second, the explanatory variables included in our regressions, while comprehensive, may provide noisy measures of some of the shocks and sources of heterogeneity that affected family saving over the period. Factors likely to be measured poorly include changes in expected lifetime income, more complex changes in family composition, unexpected events like accident or illness, and so forth. Iv. Changes in Household Portfolios. 1983-89 In addition to changes in wealth, the panel data also provide information on changes in its composition over time. Data on portfolio changes are valuable for several reasons. They may shed light on the question of how actively households manage their assets and liabilities. They provide information relevant to the question of whether tax incentives for saving have the intended effect of increasing saving, or just encourage portfolio restructuring to exploit tax breaks. They may also provide a basis for exploring dynamic relationships between assets, debt and income. 23. Smith (1995) documents a strong relationship between marriage and wealth, using the Health and Retirement Survey. 24. Kennickell (1995) finds a similar result.

-16- Financial assets. Table 10 shows the shares of households owning financial assets in 1983 and 1989, and the mean and median values among households with holdings. Financial assets include liquid assets (checking, savings, money market and call accounts); retirement accounts (IRA and Keogh accounts, and 401(k) -type accounts permitting withdrawals or loans); securities (stocks, bonds, mutual funds , and other managed assets such as trusts) : and ‘other’ financial assets (certificates of deposit, savings bonds, and cash value life insurance) . In both years, about 90 percent of the panel households had some type of financial asset. Mean and median holdings rose substantially over the period for most of the groups shown in the table. As an exception, median financial assets declined for families in the top 10 percent of the 1983 wealth distribution. Some important shifts in the composition of financial assets occurred over the period (table 11). For households as a whole, the share of financial assets in retirement accounts rose markedly, from 9.7 percent in 1983 to 21.6 percent in 1989. This share increased substantially for all types of households, except those where the head was 65 or older in 1983. Interestingly, the increased share in retirement accounts mostly came at the expense of securities, for which the share declined from 51.6 to 41.1 between 1983 and 1989. While it has been argued that growth in retirement accounts represented a shift of assets into tax-preferred forms, rather than new savings, the panel data suggest that much of the shift was between assets with differing degrees of tax preference, since investments in securities are also favored by the lack of a tax on unrealized capital gains. The panel data also provide some perspective on turnover in ownership of financial assets. Table 12 divides households into those owning an asset in both 1983 and 1989, those apparently selling off their holdings between 1989 and 1989, those acquiring holdings of the asset between 1983 and 1989, and those without holdings in either 25 year. For example, in 1983, 12.2 percent of households reported 25. It is possible that some acquisitions and sell-offs measured in the data are spurious, reflecting differences in the way assets were reported in the 1983 and 1989 surveys or other measurement problems.

-17- 26 that they did not have liquid assets of any kind. By 1989, 42.6 percent of this group (5.2/12.2) had acquired liquid assets. Conversely, 5.8 percent of households that had liquid assets in 1983 [5.1/(82.7+5.1)] no longer had liquid assets in 1989. Over 7 percent had no liquid assets in either year. These findings suggest that it is misleading to view households without assets at a given time as never having assets, since many will acquire them over time. A second point of interest concerns ownership of securities. Previous studies suggest that the low rate of stock ownership reflects information costs associated with getting started in stock investment (King and Leape 1987, Haliassos and Bertaut 1995). The panel data show a net inflow into securities among households with heads in the age groups between 35 and 65, with the share of households acquiring securities generally exceeding the share selling off holdings. There was a net outflow from securities among households with heads in older age ranges. However, the overall gross outflow from securities was relatively large, with 41.5 percent of households that owned securities in 1983 [9.0/(9.0+12.9)] no longer having holdings in 1989” Some part of this outflow may reflect portfolio adjustments following the stock market crash of 1987. Nonfinancial assets. Table 13 shows information on the shares of households owning nonfinancial assets in 1983 and 1989, and the mean and median values among households with holdings. Nonfinancial assets include a primary residence; business equity and investment real estate; vehicles; and other assets such as art and precious metals. The share of households owning nonfinancial assets rose from 90.6 percent in 1983 to 93.1 percent in 1989, with large increases for households whose incomes were below $10,000, or whose heads were under 35, in 1983. The share of households owning such assets declined noticeably in the group aged 75 and older. Much of the increase in ownership of nonfinancial assets is associated with an increase in homeownership among the panel households. As shown in table 14, the homeownership rate increased 26. Note that the SCF does not collect information on holdings of cash.

-18from 63.1 percent to 69.1 between 1983 and 1989. As might be expected, the largest increases in homeownership occurred among households with heads under 35 in 1983. The increase was also large among households with college-educated heads, and among those with relatively low net worth in 1983, While the mean home value rose substantially over the period, from $82,100 in 1983 to $101,000 in 1989, the median value rose only modestly, from $62,300 to $65,000. The largest increase in the median occurred among households with 1983 incomes exceeding $100,000. In both years, homes accounted for around 44 percent of the total value of nonfinancial assets held by households (table 15). The share of business equity and investment real estate fell from 50 to 45 percent, reflecting a substantial decline in holdings for the wealthiest group, partially offset by increases for the rest of the population. As shown in table 16, there are some distinct life-cycle patterns in the acquisition and sale of nonfinancial assets. The share of families becoming homeowners in the 1983-89 period was highest among families where the head was under 35 in 1983. While transitions out of homeownership were quite uncommon for families with heads between the ages of 45 and 74, almost 20 percent of older families who owned homes in 1983 no longer owned homes in 1989. Similarly, there was a net inflow into ownership of business and real estate interests among families with heads in the under-55 age groups, but a net outflow among families with heads in the older age groups. Interestingly, the wealthier a household was in 1983, the more likely it was to move out of business and investment real estate. Debts. Table 17 provides information on the share of households having debt of any kind in 1983 and 1989, and the mean and median values among households with debts. Debts include mortgages; installment loans (loans for vehicles, consumer durables, and home improvement) ; credit card debts; and other debts (loans for investment real estate, lines of credit, and miscellaneous other debts) . Reflecting the general aging of the panel, the share of households with debts declined slightly from 75.1 percent in 1983 to 73.4 percent in 1989. Mean and median debts rose considerably over the period, with the median rising from $11,800 in 1983 to $19,800 in 1989. The

-19composition of household borrowing was fairly constant, with mortgages accounting for around 62 percent of total borrowing by families in both years (table 18). As in the case of nonfinancial assets, there are some clear life-cycle patterns in debt holdings (table 19). The share of families acquiring mortgages between 1983 and 1989 was highest for the under-35 age group. In all other age groups, the share of households getting rid of mortgage debt over the period exceeded the share acquiring it. In the under-35 age group, the share of families acquiring installment loans exceeded the share getting rid of them, while the inflows largely offset the outflows for families in the 35- 54 age groups. There is also evidence of a life-cycle pattern for credit card debt, with a large net inflow into credit card debt for the under-35 age group and net outflows for families in the older age groups. This suggests that, while analyses of credit card borrowing often distinguish between ‘revolvers’ and ‘convenience users, ‘ in fact the likelihood of having credit card debt often changes over time. v. JSummarv and conclusions This paper analyzed saving and portfolio changes using the 1983-89 panel of the SCF, and had four major findings. First, there was a modest increase in median wealth over the period, partly reflecting the aging of the panel sample. Second, while overall wealth inequality rose over the period, families in the top 1 percent of the wealth distribution in 1983 saw their share of total wealth decline between 1983 and 1989. Third, regression analysis showed significant effects of age, income and initial wealth on saving over the 1983-89 period, as standard models of saving behavior would predict. However, the analysis still left a large part of total variation in saving unexplained. This may be due to measurement error in wealth, as well as problems measuring the myriad of factors that explain saving behavior. Finally, there are some clear life-cycle patterns in the portfolios of assets and liabilities held by families, with younger families acquiring homes, businesses and all types of debts , and older families getting rid of them.

-20- References Avery, Robert B., Gregory E. Elliehausen and Glenn B. Canner (1984), “Survey of Consumer Finances, 1983,” Federal Reserve Bulletin, vol. 70, pp. 679-692. Avery, Robert B. and Arthur B. Kennickell (1991). “Household Saving in the U.S. ,“ Review of Income and Wealth, 37(4), December. PP. 409-432. Avery, Robert B., Gregory E. Elliehausen and Arthur B. Kennickell (1988) . “Measuring Wealth with Survey Data: An Evaluation of the 1983 Survey of Consumer Finances, ” Review of Income and Wealth , 34(4), December, pp. 339-369. Canner, Glenn, Arthur Kennickell and Charles Luckett (1995). “Household Sector Borrowing and the Burden of Debt,” Federal Reserve Bulletin, Vol. 81, pp. 323-338. Curtin, Richard, F. Thomas Juster, and James Morgan (1989). “Survey Estimates of Wealth: An Assessment of Quality.” In Robert E. Lipsey and Helen Stone Tice, eds. , The Measurement of Saving. Investment and Wealth . Chicago: University of Chicago Press for the NBER, pp. 473-551. Engen, Eric, William Gale and John Karl Scholz (1994). “Do Saving Incentives Work?” Brookin~s Papers on Economic Activity , No. 1, PP. 85-180. Haliassos, Michael and Carol Bertaut (1995). “Why Do So Few Hold Stocks?” Economic Journal , Vol 105, No. 432, pp. 111O-1129. Heeringa, Stephen G. (199x). Heeringa, Stephen G., Judith H. Connor and R. Louise Woodburn (1994). “The 1989 Surveys of Consumer Finances: Sample Design and Weighting Documentation.” Mimeo , University of Michigan, Institute for Survey Research.

-21- Kennickell, Arthur B. and Douglas McManus (1994) . “Multiple Imputation of the 1983 and 1989 Waves of the SCF.” In American Statistical Association, 1994 Proceedin~s of the Section on Survev Research Methods , Vol. I. Alexandria, VA: American Statistical Association> PP. 523-528. Kennickell, Arthur B. and Janice Shack-Marquez (1992) . “Changes in Family Finances from 1983 to 1989: Evidence from the Survey of Consumer Finances, ” Federal Reserve Bulletin, Vol. 78, pp. 1-18. King, Mervyn A. and Jonathan I. Leape (1987). “Asset Accumulation, Information and the Life Cycle.” NBER Working Paper No. 2392. Levy, Frank and Richard J. Murnane (1992). “U.S. Earnings Levels and Earnings Inequality: A Review of Recent Trends and Proposed Explanations ,“ Jol.1rnal of Economic Literature , Vol. 30, No. 3, pp. 1333-1381. 4- Little, Roderick A. (1993). “Post-Stratification: A Modeler’s Perspective ,“ Journal of the American Statistical Association, Vol. 88, No. 423 (September), pp. 237-250. Poterba, James, Stephen Venti and David Wise (1992). “401(k) Plans and Tax-Deferred Saving.” In David Wise, cd., Topics in the Economics of Aging. Chicago: University of Chicago Press for the . NBER . Rubin, Donald B. (1987). Multiple Imputation for Nonresponse in Survevs. New York: John Wiley and Sons. . Smith, James P. (1995). “Marriage, Assets and Savings.” Rand Labor and populatio~program Working Paper 95-08 (March) . wolff, Edward N. (1995). Top Heavy: A Study of the Increasing Inequality of Wealth in America. New York: Twentieth Century Fund Press .

Table 1. Mean and Median Income. 1983 and 1989 Thousands of 1989 dollars Share of Mean income Median income families Jx3& 1 9 89 1983 1989 All families 100.0 33.4 36.8 24.9 25.1 By 1983 income (’89 $) Below $10,000 17.3 6.3 11.1 6.2 7.9 $10,000-24,999 33.5 17.1 22.8 16.8 18.8 $25,000-49,999 32.9 36.3 40.7 35.9 36.0 $50,000-99,999 13.8 66.2 57.3 62.6 55.0 $100,000+ 2.6 209.5 229.6 148.9 135.0 Bv 1983 ae,e of head (yrs) Under 33 33.1 26.0 37.2 22.5 28.0 35-44 19.7 41.9 44.8 34.9 36.9 45-54 16.2 41.7 43.7 34.5 32.0 55-65 15.8 36.4 29.5 22.2 16.0 k 65-74 11.6 28.8 29.2 17.1 13.0 N 75 and over 3.7 18.5 15.2 12.9 10.2 1 Bv education of head Below high school 26.8 18.7 16.1 13.7 12.0 High school diploma 48.8 30.9 34.5 25.7 27.0 College degree 24.4 54.3 64.3 40.5 44.0 By 1983 net Worth Bottom 25 percent 25.0 16.1 20.7 12.7 18.8 25 -49 per~ent 25.0 24.5 28.0 21.0 25.0 50 -74 percent 25.0 33.5 35.3 31.1 30.0 75 -89 percent 15.0 41.3 40.3 38.0 33.0 Top 10 percent 10.0 86.4 97,4 49.9 47+7

Table 2. Mean and Median Net Worth. 1983-1989 Thousands of 1989 dollsrs Mean net worth Median net Worth -L2&!- J-9fML 1983 L2B_9 All families 142.6 187.6 43.3 56.6 Bv 1983 income (’89 $) Below $10,000 24.4 27.8 4.9 7.3 $10,000-24,999 61.4 69.6 24.3 35.9 $25,000-49,999 120.0 159.3 57.4 98.5 $50,000-99,999 203.2 298.1 111.3 163.4 $100,000+ 1923.7 2524.5 933.5 1223.7 BV 1983 age of head (y rs) Under 35 49.4 77.6 9.2 30.5 35-44 129.1 172.0 48.9 73.7 45-54 191.0 268.0 68.6 89.3 55-64 229.5 275.4 69.7 63.6 (J 65-74 246.3 295.3 76.9 81.9 L.J 75 and over 139.6 190.0 62.4 50.9 By education of head Below high school 66.9 71.6 23.7 29.9 High school diploma 121.8 149.8 39.4 54.8 College degree 267,3 390.9 81.9 128.6 Bv 1983 net Worth Bottom 25 percent 1.0 16.5 1.2 4.0 25-49 percent 20.2 54.1 17.7 33.0 50-74 percent 74.2 126.6 72.8 93.3 75-89 percent 168.8 224.5 164.0 176.0 Top 10 percent 932.0 1044.3 476,8 416.6

Table 3. Measures of Changes in Wealth. 1983-1989 Change in net worth Percent change Group’s share Group’s share th ’89 in group’s of change in of all Mean Median total wealth total wealth families (1) (2) (3) (4) (5) All households 45.0 7.6 31.6 100.0 100.0 Bv 1983 income Below $10,000 3.5 0.2 14.2 1.3 17.3 $10,000-24,999 8.2 3.1 13.3 6.1 33.5 $25,000-49,999 39.3 24.4 32.7 28.6 32.9 $50,000-99,999 94.9 33.9 46.7 29.0 13.8 $100,000+ 600.9 236.1 31.2 34.9 2.6 BV 1983 age of head (vrs) Below 35 28.2 9.8 57.1 20.7 33.1 35-44 43.0 14.0 33.3 18.8 19.7 45-54 77.0 15.5 40.3 27.7 16.2 55-64 45.9 0.2 20.0 16.1 15.8 65-74 49.0 0.0 19.9 12.6 11.6 75 and over 50.3 -0.9 36.0 4.1 3.7 Bv education of head Below high school 4.7 0.2 7.0 2.8 26.8 High school diploma 27.9 8.3 22.9 30.2 48.8 College degree 123.6 29.6 46.3 67.0 24.4 By 1983 net Worth Bottom 25 percent 15.5 2.9 1579.2 8.6 25.0 25-49 percent 33.9 14.2 167.8 18.8 25.0 50-74 percent 52.3 17.7 70.5 29.1 25.0 75-89 percent 55.6 23.1 33.0 18.5 15.0 Top 10 percent 112.3 -75.3 12.1 25.0 10.0

Table 4. Estimated Savings Rates . 1983-1989 Savin~ as a percent of income Mean M-edQm Aggregate All families 2.2 3.1 12.2 Bv 1983 Income (’89 $): Below $10,000 0.4 0.1 0.8 $10,000-24,999 -5.2 1.2 0.3 $25,000-49,999 6.2 7.0 9.9 $50,000-99,999 11.9 5.9 17.1 $100,000+ 7.6 16.1 25.7 Bv aze of head (yrs) : Under 35 9.2 11.5 35-44 -4.2 5.4 9.7 45-54 20.3 5.1 19.4 55-64 -16.8 -1.5 7.8 (J 65-74 -11.0 -5.2 9.8 U 75 and over 17.0 -5.3 29.5 By education of head: Below high school -10.5 0.0 -3.3 High school diploma 3.5 3.3 6.1 College degree 13.6 10.1 24.2 By 1983 net worth: Bottom 25 Dercent 12.4 3.1 13.5 25-49 perc>nt 15.0 7.1 18.9 50-74 percent 17.3 5.2 19.6 75-89 percent -14.0 -4.7 13.0 Top 10 percent -68.6 -44.1 -0.9 Note : Saving rates are estimated using the method described in Section 111.A.

Table 5. Distribution of Net Worth. 1983 and 1989 Group’s share of aggregate Group’s net Worth share of all families All families 100.0 100.0 100.0 , . Bv 1983 co ( 89 $) . Below i;O,~~O 3.0 2.6 17.3 $10,000-24,999 14.4 12.4 33.5 $25,000-49,999 27.7 27.9 32.9 $50,000-99,999 19.6 21.9 13.8 $100,000+ 35.3 35.2 2.6 Bv age of head (yrs) : Below 35 11.5 13.7 33.1 35-44 17.8 18.1 19.7 45-54 21.7 23.1 16.2 55-64 25.4 23.2 15.8 65-74 20.0 18.2 11.6 75 and over 3.6 3.7 3.7 Bv educ ation of head: Below high school 12.6 10.2 26.8 High school diploma 41.7 38.9 48.8 College degree 45.7 50.8 24.4 ~V 1983 net Worth: Bottom 25 percent 0.2 2.2 25.0 25-49.9 percent 3.5 7.2 25.0 50-74.9 percent 13.0 16.9 25.0 75-89.9 percent 17.8 17.9 15.0 90-94.9 percent 12.4 10.1 5.0 95-98.9 percent 22.7 20.3 4.0 Top 1 percent 30.5 25.4 1.0

Table 6. Distribution of Net Worth. by 1983 and 1989 Wealth Percentile Share of aggre~ate net worth Net worth Share of 1983 net worth, Share of 1989 net worth, Share of 1989 net worth, percentile: based on based on based on 1983 percentile 1983 percentile 1989 percentile Bottom 25 percent 0.2 2.2 0.3 25-49.9 percent 3.5 7.2 4.3 50-74.9 percent 13.0 16.9 12.7 75-89.9 percent 17.8 17.9 17.7 90-94.9 percent 12.4 10.1 11.4 95-98.9 percent 22.7 20.3 21.9 Top 1 percent 30.5 25.4 31.6 Total 100.0 100.0 100.0

Table 7. Transition Matrix for Net Worth. 1983 to 1989 1989 wealth r)ercentile 1983 wealth rcentile Pe Bottom I 25 I 25-49 I 50-74 I 75-89 I 90-94 I Top 2-5 / TOP 1 I Total ---- ---- ---- -- -+--------+--_-_----+- - ____---+------_-_+_- - - - - - --+---_____-+--- __----+------- ---- ---- ---- -- -+- -------+-------- -+- --------+------- - -+- ------ --+------ - - -+---------+----- -- Bottom 25 I 67.2 I 24.6 I 6.3 I 1.8 I 0.1 I 0.0 I 0.0 I 100.0 ---- ---- ---- . - -+_ -------+-------- -+- - ---_---+_________ -+- __----__+- ------- -+- - --_--__+--- ---- 25-49 I 24.6 [ 49.5 \ 19.0 I 4.2 I 1.9 I 0.7 I 0.0 I 100. O ---- ---- ---- - --+__------+-------- _+_ --------+-------- _+_ -----_ _-+-------- -+- ----____+___ ____ 50-74 [ 6.6 I 19.2 I 48.0 I 20.8 I 3.7 / 1.6 I o.o I loo, o ---- ---- ---- - - -+- -------+-------- -+- --------+-- ------ -+- --------+-------- -+- --------+--- ---- 75-89 I 2.1 / 8.2 I 32.9 I 41,8 I 11.3 I 3.6 / 0.2 I loo. o ---- ---- -- -+- -------+-------- -+- --------+------- - -+- --------+------ - - -+-- -------+--- -- -- 90-94 I 1.1 I 7.1 I 21.2 I 30.1 I 22.5 I 17.7 I 0.4 i 100.0 ---- ---- ---- -- -+--------+---------+- --------+---------+- --------+---------+- --------+------- Top 2-5 I 0.0 I 2.8 I 16.4 I 10.4 I 18.0 I 43.0 I 9.4 / 100.0 ---- ---- ---- -- -+--_-_-__+---------+- -__-__-_+_--------+- _ _ - -_---+- --------+- ---___-_+_------ Top 1 percent I 0.0 I 3.1 I 2.4 I 6.1 I 4.5 I 24.7 I 59.3 I 100.0 ---- ---- ---- -- _+- - - ----_+_-_-___--+- ----__--+_____----+_ _ _ _ __ _ __+_ __- -----+_ ------__+____ ___ ---- ---- ---- -- _+- - _ _ - _ __+_- _..-,--.._+- -_--_---+.----____+- _ _ _ __ - __+_ _. __-_--+_ ---____ .+_______ Total I 25.0 I 25.0 I 25.0 I 15.0 I 5.0 I 4.0 I 1.0 I 100. O Note : Wealth percentiles are calculated using the weighted panel data,

Table 8. ~ons of Var~d In Regression Vsis Means (unweighted) ~eDendent variable CHNW Chanuesin net worth. 1983-1989, in 1989 dollars 609467.40 SAVRAT Defi;ed as in text 0.21 PCHNW Change in net worth, 1983-1989, divided by average net worth 1983 & 1989 .05 INCB1O Dummy variable for 1983 income below $1OK .09 INC1025 $10-24.9K .24 INC2550 $25-49.9K (omitted) .25 INC501OO $50-99.9K .14 INCA1OO $1OOK and above 28 PCHINC Change in income, 1983-1989, divided by average income 1983 and 1989 :08 AGEB35 Dummy variable for head under age 35 in 1983 .15 AGE3544 35-44 .19 AGE45-54 45-54 (omitted) .23 AGE5564 55-64 .23 AGE6574 65-74 .15 AGE75A 75 and over .05 EDUC Education of head (in years) 13.20 NONWHITE Dummy variable for head nonwhite or Hispanic .16 NWB25 Dummy variable for 1983 net worth in bottom 25% of weighted distribution .13 NW2549 25-49% .18 NW5074 50-74% (omitted) .22 NW7589 75-89% .14 NW901OC Top 10% .32 ALLMAR Dummy variable for head married to the same person, 1983-1989 .58 GOTWID Widowed, 1983-89 .03 GOTMAR Got married, 1983-89 .04 CHMAR Other change in marital status, 1983-89 .05 UNMAR Unmarried, 1983-89 (omitted) .30 KIDS83 tlumber of children of head and spouse living in the household, 1983 92 CHKIDS Change in number of children living in the household, 1583-89 :18 IAS83 Number of adults in the household beside head, spouse and children .10 CHIAS Change in number of adults, 1983-89 .02 MOVED Dummy variable, moved between 1983 and 1989 .16 INHERIT Dummy variable for an inheritance or trust received between 1983 and 1989 14 HEALDO Dummy variable, head’s health was good/excellent in 1983, fair/poor in 1989 :11 RETEXP Dummy variable, head retired between 1983 and 1989, expected to in 1983 .10 RETUNEXP Dummy variable, head retired between 1983 and 1989, did not expect to in 1983 .03 SAVREG Dummy variable, save regularly by putting money aside each month (1989) .30 MAJMET Dummy variable, family lives in major metropolitan area .51 NONMET Dummy variable, family lives in nonmetropolitan area .23 LIST Dummy variable, case was in list sample in 1983 .24

.. . . . . . . . . . . . . . . . . . . ... Table 9. esslon Measures CHNW SAVRAT PCHNW ---- ---- ---- ---- ---- ---- . . . . ---- ---- ---- --- ---- --- ------ ----- . . . . . . ------ ------ ----- ----.- . ----- ------ ----- --. ..— I Median Robust I Median Robust I Median Robust - -- --------+------- -------------------------------------- -------------------------------------- -------------------- INC<1OK -53436* (11040) -45990’ (12651) -.231’ .063 -.254* .057 -.596’ .099) -.325* (.094 INC1025 -33048’ ( 6902) -30518” ( 7900) -.141* .040 -.125* .035 -.264’ .062) -.234* (.058 INC501OO 42616* ( 7713) 33757’ ( 8830) .090’ .045 .089’ .039 .242* .069) .205’ (.065 INC>1OO 420512’ (11698) 166182* (13338) .535’ ,068 .470’ .1359 .621’ .103) .553* (.099 PCHINC 56462’ ( 4121) 39186’ ( 4685) 229* .024 .224* .021 .374’ -:;;;* (.035 AGE<35 -26657* ( 8449) -18239+ ( 9635) - :089+ .049 -.062 .043 -.074 ::;;; (.071 AGE3544 -28776* ( 7312) -21623’ ( 8367) -.064 .042 -.057 .037 -.091 .066) -.010 (.062 AGE5564 3915 ( 7132) -8445 ( 8110) .047 .041 041 .036 -.002 .063) -.007 (.06C AGE6574 8317 ( 8084) 1693 ( 9224) 032 .047 -:052 .041 -.042 .073) -.008 (.068 AGE>75 9003 (11318) -2653 (12912) :070 .065 -.027 .058 .007 .101) -.025 (.096 EDUC -454 ( 940) 344 ( 1073) .000 .005 -.002 .005 .010 .008) .013 (.008 NONWHITE -158 ( 6454) -6518 ( 7387) .002 .037 -.009 .033 -.039 .058) .019 (.055 NW<25 11261 ( 9169) 9916 (10568) .167* .053 .175’ .047 1.179’ .083) 1.324* (.o78 NW2549 13184+ ( 7480) 7519 ( 8581) 094’ .043 104* .038 .398* .068) .322* (.064 NW7589 -8515 ( 7852) -14490 ( 8999) -:098’ .046 -:109” .040 -.21O* .070) -.188* (.067 NW901OO -109336’ ( 8874) -150133’ (10099) -.525* .051 -.429* .045 -.655* .079) -.638* (.075 ALLMAR 7340 ( 5948) 12329+ ( 6794) .088’ .034 077’ .030 150* .053) .174’ (.050 GOTWI D 2684 (12538) 3640 (14708) -.066 .074 -:017 .066 -:010 .116) .012 (.109 GOTMAR 9520 (12047) 12807 (13823) 055 .070 041 .062 164 .108) 221* {.102 CHMAR -1o8.48 (10847) -14459 (12422) -:009 .063 -:051 .056 -:158 .098) -:036 (.092 KIDS83 778 ( 2671) -2073 ( 3043) -.008 .015 -.012 ,014 .007 .024) .002 (.022 CHKIDS -1255 ( 3685) 4282 ( 4208) -.012 .021 .008 .019 .006 .033) .022 (.031 IAS83 6516 ( 7057) -3632 ( 8456) .043 .043 .014 .038 .086 .066) .041 (.063 CHIAS 8379 ( 8430) -1480 (10053) .037 ,1)~1 002 .045 ,Olg .079) -.062 (.074 MOVED 794.4 ( 6263) -6079 ( 7155) .007 036 -:020 .024 .056) .034 (.053 INHERIT 36435* ( 6440j 18432’ ( 7344) 114 ,037 117* :::: .058) 156* (.054 HEALDO -2374 ( 7250) -4150 ( 8319) -:033 ,042 -:048 .037 :;:;” .065) :010 (.062 RETEXP -282 ( 7874) 18096” ( 9030) .036 ,046 .031 .040 .079 .071) .122+ (.067 RETUNEXP -2196 (12749) -14616 (14554) 134+ ,074 .038 .065 .076 .114) .009 (.107 SAVREG 11534’ ( 5012) 17405’ ( 5715) :066’ ,O?Q .081’ 07.() 1~]~. ,rJ45) .l~~* (.~4~ MAJMET 16424” ( 5/13) 1/832” ( 6546) .063+ .033 068’ .025 .099+ .051) 117* (.048 NONMET -269 ( 6602) -7946 ( 7544) -.022 .038 -:044 .034 -.111+ .059) -:032 (.056 LIST 654839* (10298) l 658435* (11676) .167’ .059 105* .052 -.037 .091) .006 (.086 Constant 33999* (15562) 37944* (17760) .013 .090 :083 .079 .093 .139) -.018 (.132 - -- -----:+-:------- ------------------------------- . - ---- . . - . . . - -... .-. . . . - . . - - . . . . . . . . . . . . . . Pseudo Rz .0541 .0607 .1106 F(33,1445) 47.27 15.64 24.10 Prob > F 0.00 0.00 0.00 l Significant at 5 percent level. +Significant at 10 percent level. Note: Standard errors in parentheses. See previous table for variable definitions

Table 10. Familv holdings of financial assets, 1983-1989 Value of holdings among Share of families families with finanial assets Owning fin, assets -xxi__ 1 9 89 J9a_9- All families 90.5 90.8 67.6 6.3 11.2 Bv 1983 income (’89 $) Below $10,000 70.3 73.4 4.1 10.5 1.4 1.5 $10,000-24,999 87.7 89.6 17.7 18.6 4.2 6.7 $25,000-49,999 99.1 97.2 30.6 43.6 6.0 16.4 $50,000-99,999 100.0 98.1 67.6 101.6 28.0 54.4 $100,000+ 100.0 100.0 715.6 1022.7 217.8 396.6 BV 1983 age of head (years) Under 35 89.8 89.9 12.5 20,9 2.4 5.0 35-44 92.2 89.9 32.1 44.7 6.3 13.4 45-54 93.4 93.7 48.9 82.3 8.1 19.0 55-64 87.7 90.1 88.3 117.3 13.1 16.4 65-74 90.4 92.2 116.5 143,7 20.9 21.0 75 and over 86.5 88.7 74.3 83.1 23.7 26.8 Bv education of head Below high school 77.8 79.9 20.4 20,1 3.1 3.0 High school diploma 93.3 92.3 33.2 47.1 5.4 10.4 College degree 98.7 99.7 100.9 147.4 14.1 33.7 ~V 1983 net Worth Bottom 25 percent 72.1 77.2 2.3 6.8 1.0 1.7 25-49 percent 91.9 88.3 6.0 16.4 3.2 7.0 50-74 percent 98.6 98.4 16.9 35.7 10.7 16.2 75-89 percent 99.2 99.1 51.1 75.2 35.3 50.8 Top 10 percent 99.2 99.4 302.8 366.1 97.7 77.4 Note : See text for definition of financial assets.

Table 11. Distribution of financial assets of all families. by type of asset. 1983 and 1989 Share of each ~rouD’s total financial assets Liauid assets Retirement accts Securities Other financial J$E3i dA?.9- ~ _lQ.&9_ ~ -l&!.&L _LM3- _H8.s!- All families 18.1 18.6 9.7 21.6 51.6 41.1 20.6 18,7 Bv 1983 inc. (’89 $) Below $10,000 44.1 36.0 0.2 3.3 11.3 23.2 44.4 37.5 $10,000-24,999 28.8 29.2 2.8 12:2 21.1 11.3 47.3 47.3 $25,000-49,999 24.8 22.3 11.8 29.8 37.7 21.1 25.8 26.8 $50,000-99,999 20.5 18.7 12.3 29.8 41.7 33.0 25.6 18.5 $100,000+ 9.7 13.5 9.5 16.0 73.6 62.8 7.2 7.7 Bv age of head (yrs) Under 35 24.4 24.3 13.5 38.9 27.1 13.1 35.0 23.7 35-44 23.4 19.8 12.8 36.0 46.0 19.5 17.8 24.7 45-54 16.9 14.9 12.3 30.5 56.1 37.3 14.6 17.3 55-64 16.6 17.0 9.6 17.8 54.4 50.4 19.5 14.8 65-74 14.5 19.1 7.5 7.9 56.2 55.1 21.8 17.9 75 and over 25.1 25.2 0.2 0.1 52.3 50.3 22.5 24.4 By education of head Below high school 27.4 22.8 4.8 13.2 33.1 20.7 34.7 43.3 High school dipl. 18.7 21.0 8.4 20.6 43.0 33.4 29.9 25.0 College degree 16.1 16.6 11.4 23.3 60.3 48.1 12.2 12.0 Bv 1983 net Worth Bottom 25 percent 35.0 35.4 3.4 25.5 36.0 7.9 25.5 31.2 25-49 percent 44.4 21.8 11.3 38.7 13.3 7.5 31.0 31.9 50-74 percent 33.8 23.6 12.6 33.0 10.4 14.4 43.2 29.0 75-89 percent 30.8 21.8 10.4 25.9 19.3 22.4 39.6 29.9 Top 10 percent 11.3 15.4 9.2 15.8 67.5 57.9 12.1 11.0

. . Table 12. Turnover . . . . J,lquld assets Retir-t accounts Securltles Other fin~ial Owned in 1983 Yes Yes No No Yes Yes No No Yes Yes No No Yes Yes No No d In 1989 Yes No No Yes No Yes L J.uL -&L All families 82.7 5.1 5.2 7.0 18.2 4.0 17.9 59.9 12.9 9.0 9.9 68.2 40.8 12.8 16.0 30.4 , . BY ncow Below $1OK 50.5 12.6 13.6 23.4 0.7 2.5 96.8 1.5 0.9 4.6 93.0 13.3 16.4 11.3 59.1 $10-24.9K 80.1 4.3 6.9 8.8 ::: 3.6 18.6 73.5 7.2 7.2 5.6 80.0 36.7 13.3 17.2 32.8 $25-49.9K 94.4 3.7 1.8 0.2 22.5 5.7 25.4 46.4 14.9 12.8 12.5 59.8 47.6 10.3 19.8 22.2 $50-99.9K 98.1 1.9 0.0 0.0 54.3 4.7 18.9 22.1 25.9 13.4 21.1 39.5 64.3 13.1 10.4 12.2 Over $1OOK 99.9 0.0 0.1 0.0 71.7 3.3 12.1 12.9 65.4 14.3 9.6 10.7 65.8 10.9 13.5 9.8 By head’s a @X Under 35 80.8 4.4 5.9 12.1 4.8 24.5 58.6 6.6 8.2 7.7 77.6 34.3 9.3 17.9 38.5 35-44 83.2 8.6 3.8 ;:: 25.8 2.1 24.8 47.3 13,8 11.3 14.4 60.5 45.4 15.1 16.1 23.4 45-54 86.1 5.1 2.6 6.2 27.0 5.3 18.3 49.4 14.8 9.3 14.6 61.3 41.8 10.4 21.2 26.5 55-64 82.1 4.1 6.1 7.7 22.2 5.2 10.6 62.1 15.9 7.3 8.9 67.8 40.3 16.0 15.7 28.0 65-74 82.4 3.5 8.6 5.5 10.8 1.8 84.9 21.4 8.8 6.2 63.7 50.6 15.2 7.1 27.1 75 and over 85.5 1.1 3.2 10.3 0.0 ;:; 2.0 97.8 16.0 9.9 2.0 72.0 41.7 20.2 5.0 33.0 d’s edup Below HS 64.3 9.6 9.2 16.9 5.4 2.0 6.0 86.6 3.7 3.4 5.9 87.0 26.6 13.2 14.5 45.8 HS diploma 86.5 4.9 5.1 16.7 5.2 19.3 58.8 11.4 9.6 9.0 70.0 41.1 12.9 16.6 29.4 College degree 95.3 0.3 ::: 0.0 35.4 3.7 28.2 32.7 25.8 14.0 16.1 44.0 55.8 12.0 16.5 15.6 )3V ’83 net worth Bottom 25% 56.4 B.5 14.2 20.9 2.4 3.4 16.3 78.0 o.~ 3.6 4.1 91.7 11.4 13.3 15.6 59.7 25-49% 82.7 7.4 3.5 6.3 8.5 21.4 66.4 5.8 9.0 6.0 79.3 36.0 10.5 18.5 35.0 50-74% 95.8 1.0 2.2 1.0 23.1 ::; 19.2 53.3 11.7 9.8 14.5 64.0 57.4 14.8 15.7 12.0 75-89% 94.1 5.0 1.0 0.0 36.6 5.0 15.1 43.4 26.1 11.0 19.9 43.0 55.3 13.7 16.5 14.5 Top 10% 98.3 0.6 1.1 0.0 42.2 3.5 14.4 40.0 44.3 17.4 8.0 30.3 62.8 10.8 10.7 15.7 Note: See text for definitions.

Table 13. Familv holdings of nonfinancial assets. 1983 and 1989 Value of holdings among Share of families families with nonfinancial assets Owning nonfin. assets 12ii2-- 1 9 89 All families 90.6 93.1 131.5 166.1 57.9 67.3 Bv 1983 income (’89 $) Below $1!?.000 65.1 74.3 37,4 34.0 14.4 20.0 $10,000-24,999 90.5 93.8 61.1 74.1 37.6 49.8 $25,000-49,999 99.5 98.9 114.0 150.8 67.1 91.8 $50,000-99,999 99.7 99.7 175.2 251.6 115.6 144.5 $100,000+ 99.9 100.0 1340.2 1659.1 509.6 674.0 Bv 1983 age of head (years) Under 35 88.9 98.4 63.9 98.3 14.9 52.4 35-44 91.8 94.3 142.1 179.8 76.3 78.8 45-54 92.4 92.6 185.5 239.2 84.5 91.5 55-64 92.5 90.5 181.7 205.2 66.3 64.6 65-74 91$0 87.1 165.1 194.9 63.1 62.8 75 and over 81.7 71.6 97.6 167.8 56.5 57.6 Bv education of head Below high school 85.8 84.7 67.6 75.0 37.2 38.6 High school diploma 91.1 94.9 122.1 138.3 56.7 64.1 College degree 94.8 98.9 213.1 305.2 93.8 123.7 BV 1983 net Worth Bottom 25 percent 65.7 79.0 6.2 32.4 3.1 11.2 25-49 percent 97.8 95.7 31.4 68.6 25.7 46.3 50-74 percent 98.9 97.9 79.3 120.1 73.2 87.9 75-89 percent 100.0 99.6 142.4 183.7 141.5 123.7 Top 10 percent 100.0 100.0 693.1 748.2 382.3 283.8 Note : See text for items included in nonfinancial assets,

Table 14. Ownershi~ of primarv residence, 1983 and 1989 Value of Drimarv residence among Share of families families owhinz their home “ ownin~ their home Mean Median -19fL3- J-xi$L 1283- -L9-aL J u 3 .- 1989 All families 63.1 69.2 82.1 101.0 62.3 65.0 ~V 1983 “ co e (’89 s) Below ~;O,~OO 37.8 41.4 40.9 40.2 24.9 30.0 $10,000-24,999 55.6 65.0 58.0 66.8 49.8 53.2 $25,000-49,999 71.7 79.6 75.6 97.1 62.3 70.0 $50,000-99,999 86.7 84.0 108.0 144.7 93.4 105.0 $100,000+ 96.2 95.8 305.4 408.8 230.3 300.0 Bv age of he ad (years) Under 35 39.4 63.2 64.2 87.9 62.3 65.0 35-44 69.2 66.4 93.4 111.6 77.8 70.0 45-54 79.5 77.4 91.2 113.3 68.5 70.0 55-64 76.4 75.4 88.8 101.4 62.3 60.0 65-74 77.9 73.8 74,6 100.3 56.0 60.0 75 and over 68.7 60.1 62.3 91.9 49.8 50.0 BY education of head Below high school 61.6 60.3 54.2 62.1 43.6 44.3 High sch~ol diploma 65.6 69.6 76.8 84.8 62.3 60.0 College de~ree 59.9 77.9 12.5.3 163.0 93.4 100.0 BV 1983 net Worth Bottom 25 percent 6.5 29.0 22.0 56.9 18.7 50.0 25-49 percent 63.0 70.5 36.1 63.0 33.6 45.0 50-74 percent 89.8 86.9 68.4 88.4 62.3 70.0 75-89 percent 91.5 89.7 97.6 112.9 87.2 80.0 Top 10 percent 95.5 90.9 178.3 221.8 124.5 150.0

Table 15. Distribution of total nonfinancial assets. by type of asset, 1983 and 1989 Share of each ~rouD’s total nonfinancial assets Business interests Primary and investment Vehicles Other esidence eal estate nonfinancial k J-M&?-_ -X?& 19_&9- -Q83- 1989 -1-9-&L _L9-&SL All families 43.5 45.1 50.4 45.3 4.8 5.5 1.3 4.0 Bv 1983 inc. (’89 s) Below $10,000 63.4 65.8 28.3 17.5 8.2 11.2 0.0 5.5 $10,000-24,999 58.3 62.5 34.3 26.9 6.9 8.4 0.5 2.2 $25,000-49,999 47.8 51.8 45.0 38.3 6.1 6.7 1.1 3.2 $50,000-99,999 53.6 48.4 38.4 40.6 5.9 6.0 2.2 4.9 $100,000+ 21.9 23.6 75.4 69.6 1.1 1.7 1.5 5.1 Bv a~e of head (yrs) Below 35 44.5 57.4 45.3 30.3 8.3 8.9 1.9 3.4 35-44 49.5 43.7 43.7 45.0 5.7 6.5 1.1 4.7 45-54 42.3 39.6 51.8 52.4 4.3 4.7 1.7 3.3 55-64 40.4 41.2 55.0 50.5 3.3 3.8 1.3 4.6 65-74 38.7 43.6 57.7 48.8 3.0 3.1 0.6 4.4 75 and over 53.7 46.0 42.4 48.8 3.3 1.8 0.6 3.4 ~ 57.5 59.0 35.0 32.3 7.2 7.1 0.4 1.6 HS diploma 45.3 45.0 48.5 43.7 5.3 7.2 0.9 4.1 College degree 37.2 42.1 57.3 49.7 3.5 3.7 2.0 4.5 BY 1983 net worth Bottom 25% 35.6 64.4 2.2 15.2 61.9 16.4 0.3 4.1 25 -49% 74.2 67.5 8.5 19.4 16.1 9.9 1.2 3.1 50-74% 78.3 65.4 12.3 23.2 7.9 8.3 1.5 3.1 75-89% 62.7 55.3 30.2 35.4 5.5 6.2 1.6 3.1 Top 10% 24.6 26.9 72.6 65.9 1.6 2.3 1.2 4.9

Table 16. Turnover In no~ assets. 1983 and 1989 Business interests and uvesunas real e~ate Vehicles Oth.r assets . . ., . . Owned in 1983 Yes Yes xes JkIesLEs- No Yes Yes No No Yes Ye; No No 9wned In 1989 XLkzz Xe5- No Yes No _xQs_-tLL k No Yes All families 56.0 7.2 13.2 23.7 19.5 8.5 12.2 59.8 81.0 5.1 5.9 8.0 5.6 4.7 16.7 73.0 Bv 1983 income Below $1OK 26.6 11.2 14.9 47.4 3.4 7.6 2.6 86.3 41.6 11.3 15.5 31.6 0.0 0.4 14.5 85.1 $10-24.9K 49.5 6.1 15.5 28.9 13.9 7.3 11.1 67.8 80.6 6.0 6.6 6.7 4.2 3.7 14.3 77.9 $25-49.9K 66.4 5.2 13.2 15.1 25.8 7.5 15.4 51.2 94.6 2.2 2.9 0.4 7.1 5.1 17.4 70.5 $50.O-99.9K 76.7 10.0 7.3 6.1 27.5 14.4 20.8 37.3 96.7 0.5 0.9 11.2 10.6 19.1 59.0 $1OOK + 92.2 3.9 3.6 0.2 76.2 9.7 4.1 10.0 93.0 M 2.2 0.8 14.6 8.3 39.5 37.7 Bv aee of head Under 35 34.6 4.9 28.6 32.0 14.0 6.0 14.4 65.5 83.9 1.5 13.0 1.6 7.8 6.2 17.2 68.8 35-44 55.8 13.4 10.6 20.2 18.9 6.1 17.0 58.0 87.9 2.7 2.5 6.9 5.2 5.9 17.5 71.4 4 5 - 54 72.1 5.3 15.3 23.7 9.7 14.7 52.0 87.8 1.7 2.4 8.2 6.2 3.5 17.4 72.8 55-64 71.6 ;:; 3.8 19.7 27.3 12.5 6.4 53.8 77.2 8.7 2.6 11.6 4.7 4.0 17.7 73.6 65-74 73.8 4.2 22.0 22.8 11.5 4.1 61.6 68.1 12.5 2.7 16.7 2.7 1.5 15.0 80.8 75 + 55.9 12.8 ::; 27.1 10.2 11.2 5.6 72.9 46.0 25.8 1.1 27.0 0.1 1.7 4.7 93.6 ~ 54.2 7.4 6.1 32.3 11.2 5.2 7.9 75.7 66.8 12.2 3.5 17.5 1.6 1.5 12.3 84.5 HS diploma 57.3 12.4 22.0 19.0 9.9 13.8 57.3 85.4 6.8 5.8 4.7 4.1 19.9 71.3 College degree 55.4 ::; 22.5 17.5 29.6 9.3 13.7 47.4 88.0 ::; 7.1 1.8 11.9 9.2 15.1 63.8 ~ 42 2.3 24.8 68.7 0.3 H 10.2 88.4 59.6 3.8 15.6 21.0 1.0 1.3 11.7 86.1 25-49% 50:6 12.4 19.9 17.1 8.1 14.6 71.4 83.1 7.1 3.7 6.0 5.2 4.4 15.9 74.5 50-7-4% 83.7 6.1 3.2 7.0 20.0 7.6 17.5 54.9 87.3 6.0 2.8 4.0 6.7 7.2 14.0 72.1 75-89% 84.1 7.4 5.7 2.8 36.8 19.6 8.8 34.8 93.4 4.1 1.2 1.3 8.2 6~ 21.5 64.3 Top 10% 87.2 8.2 3.7 0.9 68.7 18.8 3.(I 5.5 95.1 2.1 2.,!, 0.’4 11.8 5.5 30.5 52.3 Note: See text for definitions,

Table 17. Familv debt holdin~s. 1983 and 1989 Total value of debts among Share of families families with ebt with debt Median -.DE!- -u82 lZZL ~ All families 75.1 73.4 26.9 38.7 11.8 19.8 , Bv 1983 “ co e ( 89 $) Below t~O,~OO 50.1 51.3 5.7 10.1 1.9 2.5 $10,000-24,999 68.3 70.4 13.8 23.5 3.3 11.9 $25,000-49,999 87.8 82.7 26.9 39.1 15.9 25.0 $50,000-99,999 91.1 86.5 42.9 60.6 36.2 40.9 $100,000+ 85.8 72.4 153.0 217.0 93.2 97.5 Bv age of head (years) Under 35 85.0 90.3 21.9 42.0 8.3 25.8 35-44 87.3 85.8 35.5 43.9 21.6 23.9 45-54 84.5 80.0 30.9 38.4 17.2 19.0 55-64 66.0 55.3 24.0 28.9 8.7 6.5 65-74 41.3 37.0 22.3 18.9 4.4 3.8 75 and over 25.6 19.6 17.0 19.9 1.2 0.5 Bv edur ation of head Below high school 59.9 56.1 11.7 14.2 3.9 4.6 High school diploma 80.3 76.5 25.4 32.6 13.4 22.2 College degree 81.6 86.4 42.2 67.0 23.6 36.1 By 1983 net worth Bottom 25 percent 69.3 71.3 6.8 20.2 1.8 25-49 percent 78.9 79.4 20.4 32.9 11.4 2;:; 50-74 percent 74.8 75.3 28.0 34.8 21.2 23.6 75-89 percent 76.0 66.8 32.0 49.5 21.4 26.8 Top 10 percent 80.0 69.0 76.9 97.9 36.8 45.0

Table 18. Distribution of debt. by ty~e of debt, 1983 and 1989 Share of each group’s total debt Mortgage Installment Credit card Other debt L9fKL Jw2L - D . u- JJzML 1983 __LMiL 1983 ~ All families 61.8 62.1 11.0 13.8 2.1 2.5 25.1 21.7 13v ’83 inc (’89 $) Below $10,000 63.6 54.6 19.8 41.2 1.6 3.4 15.0 0.8 $10,000-24,999 68.0 69.1 11.3 20.0 3.1 2.7 17.6 8.2 $25,000-49,999 67.3 69.9 12.4 13.6 2.6 3.2 17.6 13,4 $50,000-99,999 64.9 62.5 12.1 11.0 2.0 2.3 21.0 24.2 $100,000+ 38.5 33.1 4.6 5.1 0.3 0.5 56.6 61.3 Bv ape of head (yrs) Under 35 68.4 75.4 12.7 13.5 2.4 2.4 16.5 8+8 35-44 65.8 51.5 10.9 15.0 1.6 2.2 21.7 31.3 45-54 64.1 60.8 11.2 13.3 2.6 2.9 22.1 23.0 55-64 43.4 41.7 9.6 13.2 2.4 3.6 44.6 41.5 65-74 42.6 34.8 4.7 12.6 1.3 1.1 51.4 51.5 k Q 75 and over 11.1 . 7.2 10.5 0.6 1.8 81.1 87.7 1 -BV educ ation of head Below HS 64.7 56.9 17.6 32.3 3.3 5.5 14.4 5.4 HS degree 63.3 61.8 12.6 17.0 2.4 2.4 21.6 18.8 College degree 59.4 63.1 7.6 8.2 1.5 2.1 31.5 26.6 , Py 83 net worth Bottom 25% 26.3 66.6 37.7 26.3 7.3 3.8 28.7 3.2 25 -49% 72.5 71.8 14.3 15.5 3.0 3.3 10.3 9.5 50-74% 80.2 71.7 10.2 15.0 2.2 2.9 7.4 10.4 75-89% 71.2 62.4 9.0 10.7 2.1 2.5 17.8 24.4 90-100% 40.5 40.7 5.6 6.6 0.5 0.6 53.3 52.1

,. . Table 19. mnover in debt. 1983 and 1989 s h a r e o f fa m 1 1 e s t Credit Card ~ ., -., I es I es l\. 70 No Yes Yes No No Yes Yes Qwnd in 1989 1= No X.%L _NQ__xs- No Yes No Yes No Yes All families 27.4 12.0 14.2 46.4 33.3 14.8 16.4 35.6 24.1 14.9 15.2 45.8 8.9 19.5 10.0 61.6 Bv 1983 income Below $1OK 3.8 11.1 11.8 73.4 19.3 11.6 18.8 50.3 3.2 5.8 19.5 71.5 2.6 19.2 1.5 76.6 $10-24.9K 16.0 8.2 17.0 58.9 31.6 10.6 19.2 38.6 21.2 11.7 14.7 52.4 5.4 15.1 9.9 69.7 $25-49.9K 41.1 12.1 13.6 33.2 40.5 17.5 14.3 27.6 35.4 22.0 15.0 27.6 10.1 20.4 11.8 57.6 $50-99.9K 49.7 19.8 13.1 17.4 40.3 21.3 13.1 25.2 33.1 18.1 12.1 36.7 17.7 25.9 16.4 40.0 $1OOK + 40.4 26.3 6.5 26.8 18.7 19.7 7.8 53.9 10.9 9.5 11.4 68.2 32.6 31.3 12.6 23.4 Bv a~e of head Under 35 28.7 6.2 28.2 36.9 50.0 13.5 20.0 16.4 25.8 14.2 24.7 35.3 8.4 23.0 11.4 57.2 35-44 38.3 20.7 11.8 29.2 39.6 18.1 19.4 22.9 29.7 19.5 14.1 36.7 9.2 22.5 17.5 50.9 45-54 40.1 16.7 8.6 34.7 34.8 18.7 17.7 28.8 34.4 16.6 12.3 36.7 14.5 23.3 8.8 53.4 55-64 16.8 14.7 6.2 62.3 14.4 16.1 12.8 56.7 20.7 15.8 7.6 55.9 8.4 15.1 70.9 65-74 10.5 6.0 1.1 82.3 7.5 8.8 6.3 77.5 8.7 8.0 77.0 4.9 10.4 ::: 80.5 75 + 0.0 5.8 0.0 94.2 3.9 3.8 9.5 82.8 ::; -4.6 2.9 87.7 1.0 2.7 0.9 95.4 Bu&-#Ja % 13.6 9.4 5.0 71.9 24.6 14.4 13.1 47.9 11.7 12.9 14.6 60.7 2.8 13.8 3.9 79.6 1 HS diploma 30.5 14.4 14.5 40.6 39.5 14.7 17.3 28.5 27.0 16.3 14.6 42.1 8.6 21.8 12.0 57.7 College degree 36.5 10.2 23.7 29.7 30.4 15.2 18.1 36.3 32.0 14.2 17.1 36.8 16.2 21.2 13.0 49.6 ~ 28 2.8 20.9 73.4 32.8 18.0 21.7 27.5 17.5 12.0 19.6 51.0 5.7 21.0 8.1 65.1 25-49% 28:3 14.1 19.8 37.8 45.3 11.9 15.8 27.0 30.7 13.7 16.1 39.5 5.6 16.5 12.7 65.2 50-747. 42.2 12.3 7.2 38.3 30.9 12.1 17.1 39.9 29.9 16.6 15.9 37.7 6.6 14.0 10.4 69.0 75-89% 38.9 16.9 8.6 35.7 7-/, , 3 16.5 10.9 47.9 21.5 16.5 12.2 49.7 12.2 25.6 8.4 53.7 Top 10% 32.2 22.3 8.9 36.6 23.5 17.3 11.2 48.0 13.7 18.5 4.8 63.0 25.5 27.5 9.8 37.1 Note: See text for definitions.