The Finances of American Households in the Past Three Recessions: Evidence from the Survey of Consumer Finances

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. The Finances of American Households in the Past Three Recessions: Evidence from the Survey of Consumer Finances Kevin B. Moore and Michael G. Palumbo 2010-06 NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

The Finances of American Households in the Past Three Recessions: Evidence from the Survey of Consumer Finances Kevin B. Moore∗ and Michael G. Palumbo† December 10, 2009 Abstract Abstract: The downturn in economic activity in the U.S. that began in December 2007 (as determined by researchers with the National Bureau of Economic Research) has been noticeably deeper and has already lasted considerably longer than the prior two recessions—those beginning in July 1990 and in March 2001. In addition, a key difference between the current and the past two recessions is the extent to which consumer spending and residential investment have dropped since late 2007—that is, the extent to which the household sector appears to have “led” the drop in aggregate economic activityinthisrecession. Thispaperuseshousehold-leveldatafromtheFederalReserve Board’s series of Surveys of Consumer Finances to document three factors that appear to have contributed to greater financial stress in the household sector in the current downturn compared with the prior two: 1) substantial and widespread reductions in home values that resulted in sizable erosions of home equity and net worth for many homeowners; 2) markedly expanded holdings of corporate equity among middle-income households which lost significant market value, on net, as stock prices sunk; and, 3) greater debt on household balance sheets and overall financial vulnerability around the onset of the 2008-09 recession, particularly for those in the middle of the income distribution. JEL classification: E21, E32, D32. Keywords: Household net worth; consumer finances; business cycle; recession. ∗Division of Research and Statistics, Federal Reserve Board, kevin.b.moore@frb.gov. †DivisionofResearchandStatistics,FederalReserveBoard,michael.g.palumbo@frb.gov. Weappreciate comments from Eric Engen and Arthur Kennickell, as well as those from Ernesto Villaneuva and other participants in a conference on “Household Finance and Macroeconomics,” sponsored by Banco de Espan˜a in October 2009. Alycia Chin provided excellent research assistance with the macroeconomic background sectionofthepaper, andBrianBucksgenerouslyhelpeduswiththeLoanPerformancehousepriceindexes. The views expressed herein are those of the authors and do not necessarily reflect those of the Board of Governors or the staff of the Federal Reserve System. 1

1 Introduction The downturn in economic activity in the U.S. since December 2007 has been noticeably deeper and has already lasted considerably longer than the prior two recessions—those beginning in July 1990 and in March 2001. In addition, a key difference between the currentandthepasttworecessionsistheextenttowhichdeclinesinconsumerspendingand residential investment contributed to the drop in aggregate demand since late 2007—that is, the extent to which the current episode has, thus far, been more of a “consumer-led” than a “business-led” recession. This paper uses household-level data from the Federal Reserve Board’s series of Surveys of Consumer Finances (SCFs) to document three specific factors that appear to have contributed to greater financial stress in the household sector in the current downturn compared with the prior two: 1) substantial and widespread reductions in home values that resultedinsizableerosionsofhomeequityandnetworthformanyhomeowners; 2)markedly expanded holdings of corporate equity among middle-income households which lost significant value, on net, as stock prices sunk; and, 3) greater debt on household balance sheets, andoverallfinancialvulnerability,aroundtheonsetofthe2008-09recession,particularlyfor those in the middle part of the income and wealth distributions. These latter developments appear to have left many households without significant financial buffers and, therefore, particularly vulnerable to drops in assets values and to job losses. AkeymethodologicalissueouranalysisfacesisthattheBoard’scross-sectionsurveys areconductedonlyeverythreeyears,withnoneofthoseconductedsince1989corresponding precisely with business cycle peaks and troughs. Thus, many of our balance sheet tabulations are based on market values for real estate and corporate equity holdings that have been adjusted to approximately measure changes in the year-and-a-half following the past three business cycle peaks (July 1990, March 2001, and December 2007). 2 The Macroeconomic Backdrop: A Brief Comparison of the Past Three U.S. Recessions Timing and duration of the past 3 business cycles: OnDecember11, 2008, theBusiness Cycle Dating Committee of the National Bureau of Economic Research (NBER) “determined that a peak in economic activity occurred in the U.S. economy in December 2007,” marking “the end of the expansion that began in November 2001 and the beginning of a recession.”1 The available data—and consensus forecasts—suggest that it will likely be quite a while before the NBER Dating Committee determines the timing of the trough for the current business cycle. In general, the Dating Committee tends to put off its announcements of 1See http://www.nber.org/cycles/main.html; accessed 5 May 2009. 2

the key business cycle dates until a range of macroeconomic indicators suggest that a peak or trough in activity has clearly occurred. Thus, it was not until July 17, 2003, that the NBER Dating Committee announced its determination that a trough in economic activity had occurred in November 2001—almost two years before. The Dating Committee emphasizes four broad indicators of economic activity: industrialproduction,realmanufacturingandretailtradesales,realpersonalincomelesstransfer payments, and payroll employment. Current data indicate that industrial production and real manufacturing and retail trade sales may have bottomed out in June 2009. However, realpersonalincomeedgeddown, onnet, throughSeptember2009, andpayrollemployment continuedtocontractsharplythroughOctober2009(190,000joblosseswereestimatedthat month, while the unemployment rate rose 0.4 percentage point to 10.2 percent). Nonetheless, real GDP is estimated to have expanded at an annual rate of 3-1/2 percent in the third quarter of 2009, the first quarterly increase since the second quarter of 2008 and the largest increase since the third quarter of 2007.2 Through August 2009 (the latest available data when this paper was drafted), the 2008-09 U.S. recession had lasted 20 months—2-1/2 times as long as either of the previous two recessions. As can be seen in the table below, the prior recession lasted just 8 months, from a peak in activity in March 2001 to a trough in November 2001; the recession that occurred prior to that also lasted 8 months, from a peak in July 1990 to a trough in March 1991. In terms of the broadest measure of aggregate economic activity, real GDP contracted 3-3/4 percent, in total (not at an annual rate), from the fourth quarter of 2007 and the second quarter of 2009. In the 2001 recession, real GDP only decreased slightly in two quarters and the lowest level of real GDP was only 1/4 percent below the peak level reached in the fourth quarter of 2000. In the 1990-91 recession, the trough in real GDP (1991:Q1) was 1-1/2 percent lower than in the NBER peak quarter (1990:Q3)—much smaller than in the current downturn. Timing and Duration of the Past 3 Downturns in Overall U.S. Economic Activity Business Cycle Reference Dates Duration, in months Peak Trough Contraction Expansion Cycle Peak to Previous trough Trough from Peak from Quarterly dates in parentheses trough to this peak previous trough previous peak December 2007 (IV) through August 2009 20 73 94 81 March 2001 (I) November 2001 (IV) 8 120 128 128 July 1990 (III) March 1991 (I) 8 92 100 108 Source: NBER; as of this writing, the NBER has not determined that a trough in economic activity has occurred since the peak in December 2007. 2All of the figures in this paragraph use the current, “provisional” vintage of macroeconomic data. As emphasized by Dynan and Elmendorf (3), however, significant revisions are often made to major economic indicators, particularly around business cycle turning points. 3

Figure 1 displays the evolution of selected macroeconomic variables associated with householdsectordevelopmentsinthecurrentrecessionandinthepriortworecessions; plots areindexedsothatlevels equal100atthe peakinthe overallbusiness cycle—asdetermined by the National Bureau of Economic Research—and so that changes are measured as percentage deviations from activity at each cyclical peak. The vertical lines (placed at “0” along the x-axis) distinguish the time periods before and after the NBER-dated business cycle peaks. The primary take-away from figure 1 is that the 2008-09 recession has seen a larger contraction in consumer spending and residential investment, suggesting that the household sector has contributed more to the drop-off in aggregate demand in this cyclical downturn than in either the 1990 or the 2001 episodes.3 Employment: Cumulative employment losses over the first 19 months of the current recession have been much larger than in either the 1990 or 2001 recessions—about 5 percent, on net, compared with around 1‰ percent at similar points in the two prior episodes. Although the extent of the contraction in overall employment in the first 10 months of 2008 were very similar to the experience in the prior 2 recessions, job losses accelerated sharply late last year and into early 2009, whereas levels of employment had more or less stabilized a year after the two prior business cycle peaks. Thus, in terms of aggregate net job losses alone, the 2008 recession would seem to amount to a larger cyclical “shock” to households’ economic circumstances than either of the previous two. Personal income: Likeaggregateemployment, realpersonalincomefromwagesandsalaries has also decreased substantially in the current recession—6 percent, on net, from December 2007 through July 2009. At this point in the recessions, the size of the net decrease in real aggregatewagesisaboutthesameasoccurredinthe1990episodeandisthreetimesaslarge as was seen following the 2001 cyclical peak.4 Meanwhile, as can be seen in the middle left panel, on balance since the cyclical peak in December 2007, real disposable personal income has increased about 1 percent—not quite as much as the 2 percent net rise experienced at 3Figures1through3andAppendixAwerecreatedattheendofSeptember2009;atthattime,monthly data were available through June, July, or August of 2009 (depending on the specific series) and quarterly data were available through the second quarter of 2009. 4In figure 1, the underlying dollar-denominated variables—wage and salary income, disposable personal income, personal consumption expenditures, and residential investment —are expressed in “real” terms; wage and salary income and disposable personal income have been deflated using BEA’s chain-weighted price index for personal consumption expenditures. All data from the U.S. national income and product accounts reflect the comprehensive revision published by BEA in July 2009. 4

this point in the 1990 episode and markedly less than the 3‰ percent increase following the 2001 cyclical peak. As can be seen by the “blips,” in both 2001 and 2008, one-time stimulus payments and tax rebates were quickly enacted that temporarily raised disposable incomes, with the largest (arithmetic) effects occurring about six months into theses two recessions. Thestimuluspackagesthatwereenactedincludedlesstransitoryelementsand,importantly, automatic fiscal stabilizers contributed to holding up disposable income in recent months (relative to wages and salaries and to personal interest and dividend income, as well). In the year-and-a-half since the December 2007 cyclical peak, personal transfer payments have risen more than 15 percent, on net, and personal taxes have dropped nearly 30 percent. At the same point in the 1990 episode, personal transfers had also increased nearly 15 percent, but personal taxes had decreased just 8 percent; in the 2001 recession, transfers rose about 10 percent and personal taxes dropped about 20 percent.5 Personal consumption expenditures (PCE): On net over the first 18 months of the 2008- 09 recession, real aggregate consumer spending declined about 2 percent—a considerably weaker trajectory than occurred in either the episode in 1990 (no net change in real PCE over the similar time frame) or that in 2001 (a net increase of about 4 percent). The 2 percent peak-to-trough decline in real PCE that was reached about twelve months into the current recession is similar in magnitude to the peak-to-trough decline that occurred about six months into the 1990 episode. However, consumer spending rebounded fairly quickly in late 1990, then remained essentially flat for most of 1991. Although consumer spending appears to have more or less stabilized at a relatively low level this year, it has yet to recover at all. What really stands out in the plot, of course, is the rising trajectory for real PCE that occurred in the 2001 recession, resulting in a 4 percent net increase in real consumer spending over the year-and-a-half following the NBER-dated peak in overall economic activity. Residential investment and sales of newly built homes: It is clear from the lower two panels of figure 1 that the housing sector’s downward trajectory in the current recession has been much more severe than the contraction in the 1990 episode and the steady rise in housing activity that occurred in the 2001 episode (note the wide y-axis scaling in this panel). 5Appendix A shows cyclical comparisons for some of the major components of aggregate disposable personal income in the U.S. national accounts. 5

Indeed, the 40 percent drop in real residential investment and in new home sales since December 2007 occurred on the heels of 60 percent decreases in residential investment and new home sales in the year or two leading up to the business cycle peak. By contrast, there was no net contraction in aggregate housing activity around the 2001 cyclical peak, and the contraction seen in the 1990 episode was shallower and shorter-lived than in the current episode. Indeed, a gradual recovery in construction activity was underway only a few quarters after the 1990:Q3 peak in overall economic activity; in current episode, real residential investment fell sharply over the first six quarters of the recession. Figure 2 shows cyclical comparisons for several factors affecting aggregate net worth of the household sector. Personal saving: On net, the combination of a cumulative decrease in consumer spending in the current recession and a small cumulative increase in disposable income has led to a net rise in the aggregate personal saving rate of about 3 percentage points, from a rate of 1‰ percent in December 2007 to 4‰ percent in July 2009. By contrast, in the two previous cyclical downturns, the saving rate fluctuated in a much narrower range and was essentially unchanged, on net, over the first 18 months of those recessions. The personal saving rate was close to 7 percent heading into the 1990 recession, while it was only about 3 percent at the onset of the 2001 downturn. Householdnetworth: Aggregatehouseholdnetworthtumbledmorethan15percent,onnet, in the first six quarters of the current economic downturn, from $63.9 trillion in 2007:Q4 to $53.1 trillion in 2009:Q2. By contrast, in the six quarters following the 1990 peak in economic activity, aggregate net worth increased more than 10 percent and it was only down about 3 percent, on net, at this point in the 2001 episode. The substantial decline in household net worth in the current recession primarily reflects appreciable decreases in both equity prices—–down around 25 percent, on net, even after the past few months’ market rally—–and the drop in house prices across the country–—about 10 percent since December 2007 and 20 percent since the peak in prices in December 2006. Of course, plunging equity prices were also a big part of the story behind the 2001 recession, but the timing was different: Figure 2 shows that, in the 2001 episode, stock prices fell around 35 percent ahead of the NBER-dated peak in economic activity; indeed, the “bursting of 6

the tech bubble” is commonly viewed as having triggered that business-cycle downturn. In addition, equity prices decreased another 25 percent, on balance, between March 2001 and September 2002. In the current episode, stock prices were roughly flat over the year before the NBER-dated peak and plunged while overall economic activity was also contracting. Equity prices might be considered a relatively “neutral” factor in the 1990 recession: They were about unchanged in the year leading up to the NBER-dated peak in overall economic activity and over the following year, as well. A subsequent stock-market rally left equity prices 15 percent higher 20 months after the July 1990 cyclical peak. House prices: The aforementioned drop in aggregate house prices in the current recession is not something that played a large role in either of the prior two episodes. In the 2001 recession, home prices across the country rose at an annual rate of about 10 percent in the yearleadinguptotheNBER-datedpeakinoverallactivityandataboutthesamepaceover the year-and-a-half after that peak. In addition, although nominal home prices decreased almost 2 percent, on net, in the first six months of the 1990 recession, they recovered in the following six months and flattened out after that. Household debt: With household assets declining in value, and amid very tight credit market conditions, aggregate household debt has been essentially flat since the end of 2007, reflecting essentially steady levels of home mortgage debt outstanding (including first liens, second mortgages, and home equity loans and drawn lines of credit) and of unsecured consumer credit (credit card balances, auto loans, student loans, and personal loans). Thus, the borrowing experience in this recession differs starkly from that in the 2001 episode—– when both mortgage debt and consumer credit increased at average annual rates close to 10 percent—–and is noticeably weaker even than what occurred after the 1990 business cycle peak—–when mortgage debt expanded at an average annual rate of 7‰ percent rate, while consumer credit outstanding was about unchanged. Figure 3 documents cyclical changes in the performance of some major components of household debt and plots indexes of consumer sentiment and confidence to summarize households’ attitudes over the past 3 recessions. Household credit performance: Figure3makesclearhowmuchmoredifficultythehousehold sector as a whole has had staying current on their debt obligations in the current recession 7

compared with the prior two episodes. While overall mortgage delinquency rates edged up by just a quarter percentage point or less in the 2001 and the 1990 episodes, they have climbed about 2 percentage since 2007:Q4; as of 2009:Q1, overall mortgage delinquency rates stood almost 4 percentage points above the trough level registered at the end of 2006.6 In addition, as of June 2009, delinquency rates on credit card accounts were about 2 percentage points higher than at the business cycle peak and were 2‰ percentage points above the level registered at the end of 2006. Delinquency rates on auto loans extended to prime-rated borrowers also rose much more in the first year of the current recession than in the 2001 episode, although these delinquency rates have improved significantly over the first half of 2009. Moreover, personal bankruptcy filings have continued to trend higher in recent months—–about 125,000 more personal bankruptcy petitions were filed (per month) in the second quarter of 2009 than at the onset of the recession. At the same point in the 2001 recession, the pace of bankruptcy filings had risen only by about 25,000 per month. 3 Analysis of Household-Level Data in the Survey of Consumer Finances The remainder of the paper draws on data in the Federal Reserve’s Survey of Consumer Finances (SCF) to examine changes in balance sheets across the distribution of American households in each of the three most recent recessions. The SCF is a cross-section survey conducted every three years to gather comprehensive information on household assets, liabilities, and income. Because the core elements in the surveys were essentially unchanged from 1989 through 2007, the data collected in those waves are directly comparable. AlthoughtheunderlyingqualityoftheSCFdataisverygood,wefacetwoproblemsin using them for this study: First, because households are not re-interviewed across waves of the SCF, we cannot directly observe how they responded to changes in assets, employment, or income during the cyclical downturns or economic recoveries—the things we would most like to know. Therefore, our focus is on contrasting the severity of the “cyclical shocks” as they affected household net worth across the three recent recessions, taking into account 6Figure3plotsthe“seriousdelinquencyrate”fromtheMortgageBankersAssociation’squarterlyNational Delinquency Survey; this rate is the percentage of first-lien mortgages with payments that are 90 days or more delinquent or in foreclosure. 8

the composition of assets and liabilities at the onset of the economic downturns to gauge the vulnerability of household finances heading into each episode. Second, the timing of the SCF surveys is not intended to correspond with any particular state of the macroeconomy. The 1989 survey was in the field several months before a peak in economic activity and the 1992 data were collected about a year after the macroeconomic recovery had begun. The 2001 survey was also conducted several months into the business-cycle downturn and the 2004waveoccurredwellaftertheeconomyhadreestablishedfairlystronggrowth. Although the timing of the 2007 wave coincided closely with the peak in economic activity, the next SCF cross-section survey is not scheduled until 2010.7 In this study, we focus on household-level data in each SCF survey that is closest to a peak in overall economic activity—the 1989, 1992, 2001, and 2007 waves.8 We then use indexesofpricechangestoprojectthevalueofselectedhouseholdassets—residentialrealestate (primarily, but not exclusively, owner-occupied housing) and corporate equity (held directly and through mutual funds and retirement accounts, and privately-held businesses)— to the subsequent NBER-dated trough in economic activity. For the 2001 episode, we project the value of these household assets back to the NBER-dated peak in activity— March 2001—since the survey was actually conducted in the middle of the episode. For real estate revaluations, we use price indexes from LoanPerformance matched to each survey respondent’s state of residence;9 for corporate equity and privately-held business valuations, we use the Wilshire 5000 stock market index.10 We do not make any projections for other household assets (primarily checking accounts, saving accounts, and consumer durable goods) or for household liabilities (primarily, mortgages, auto loans, student loans, and credit card balances). Our thinking is that these other household assets are not subject to substantial revaluations, household liabilities (in the aggregate, at least) are very slow 7Aswenotebelow,theFederalReserveBoard(anditscontractor)iscurrentlyfieldingafollow-upsurvey to respondents who participated in the 2007 SCF. 8A broader set of results from the 2007 SCF can be found in Bucks, Kennickell, Mach, and Moore (1); detailed results from the earlier SCF waves are published also published in various issues of the Federal Reserve Bulletin. In addition, in a recent paper, Dynan (2) analyzes longer-run trends in the composition of household balance sheets using SCF data going back to the late 1960s, but also emphasizing financial consequencesoftherecentdropinhomevaluesandequityprices. Thus,thereisconsiderableoverlapbetween our discussion of changes in household wealth since the late 1980s and hers. 9LoanPerformance is a division of First American Core Logic that produces repeat-sales house price indexes covering a very large sample of owner-occupied residential transactions across the United States. Information about the indexes is available at http://www.loanperformance.com/loanperformance_hpi.asp. 10This index can be found at http://www.wilshire.com/Broad/Wilshire5000/. 9

moving, and household saving (again, in the aggregate) is generally too low to affect net worth very much over short time periods (that is, changes in household-sector net worth are overwhelmingly driven by revaluations as market prices of corporate equity and housing change). The following table summarizes our methods for generating household-level balance sheetdataroughlycorrespondingtoeachofthepastthreebusiness-cyclepeaksandtroughs: Our Translation of Nearest SCF Data to NBER-Dated Cyclical Peaks and Troughs SCF wave closest How we generate data corresponding to the cyclical: NBER-dated peak to NBER peak peak trough 1989 project to July 1990 using 1. July 1990 and 1992 both SCFs and average the results use the 1992 SCF 2. March 2001 2001 project to March 2001 project to November 2002 3. December 2007 2007 use the 2007 SCF project to June 2009 Finally, note that because the NBER-dated peak in economic activity of July 1990 falls almost right between the 1989 and 1992 SCF surveys, we projected household-level values to July 1990 from both of those SCFs, then reported simple average values for each of the statistics corresponding to that date. 3.1 Changes in net worth across the distribution of households in the past three recessions Figure 4 presents the evolution of household net worth at the mean (blue line; left y-axis) andthemedian(redline;righty-axis)ofthedistributionfrom1989to2009. Ourprojections of net worth between the SCF survey waves are denoted by the month-year labels along the x-axis. Generally, the two lines show similar time-series patterns for central tendencies of householdnetworth. Wealthdeclined, onnet, between1989and1992, and, by1995, wealth of the typical household had essentially returned to the 1989-level. Household net worth acceleratedbetween1995andearly2001—particularly,atthemeanofthedistribution—and it rose markedly between 2004 and 2007, before plunging from 2007 through the middle of 2009. By June 2009, we estimate that median household net worth was $84,000 (measured in constant dollars using a deflator based on the consumer price index for 2007)—a bit lower than in 1998 and a bit higher than in 1995. Indeed, we estimate that as of June 10

2009 median household net worth was about 10 percent higher (in real terms) than in the 1989 SCF. As of June 2009, we estimate that mean net worth was $402,000—between the levels registered in the 1998 and 2001 surveys and about 33 percent higher than in the 1989 survey. Table 1 focuses on changes in median and mean household net worth during the past three recessions. The columns in red in the table compare percentage changes in household net worth in the three recent recessions at the median and mean of the distributions, unconditionally (rows 1 and 2) and conditional on position in the cross-section income distribution (rows 3 through 8) and conditional on place in the age distribution (rows 9 through 14). The statistics in the table show how much more severe the cumulative—and combined—declines in equity prices and house prices have been in the current recession compared with the prior two. Overall, at the median, household net worth dropped 30 percent since December 2007 and, at the mean, net worth about the same amount, 28 percent. In the 1990 and 2001 recessions, median net worth did not decrease significantly, on net, while mean net worth declined only by about 5 percent. At the mean, the drops in networthexperiencedinthe1990and2001episodesrepresentedlossesontheorderofthree months of income (put another way, the drops in average net worth lowered wealth-income ratios by about 1/4 percentage point); in the 2008-09 episode, however, the drop in mean net worth represented 1-1/2 years’ worth of income. In the current recession, households in the three different income groups shown in table 1 experienced comparably sized drops in mean and median net worth. By contrast, in the prior two recessions, only households in the highest quintiile of the income distribution (that is, between the 80th and 100th percentiles of the income distribution) experienced significant decreases in median and mean net worth. Moreover, a distinctive feature of the current recession (as compared with the prior two) is that the youngest households (those headedbyapersonlessthan45yearsold)haveexperiencedmuchlargerdecreasesinmedian and mean net worth (42 and 34 percent, respectively) than have households in the older two age categories. In the prior two recessions, being relatively young was not nearly so large a “risk factor,” in terms of tracking the deterioration in balance sheet positions. As we document below, a key element of the increased risk of young households’ balance sheets heading into the 2008-09 recession was their greater exposure to revaluations of corporate 11

equity and their increased balance sheet leverage (primarily, from higher mortgage debt). Figure 5 documents how much more widespread very large decreases in household net worth have been in the current recession than in the previous two. The figure shows percentage changes in net worth at various percentiles of the distribution of household net worthduringthepastthreerecessions. Ascanbeseeninfigure5,followingthecyclicalpeak in July 1990, there were widespread (among households), but relatively small, decreases in net worth (that is, the red, green, and orange bars are only slightly in negative territory). Following the business cycle peak in March 2001, only households in the upper quartile of the wealth distribution, on average, experienced a decrease in wealth and the average decrease was only about 7 percent. However, from the cyclical peak to June 2009, we estimate very large average decreases in net worth accruing to households throughout the wealth distribution. Table 2 focuses on the “tails” of the distribution of changes in household net worth by comparing the frequency of particularly large decreases in household net worth across the three recent recessions, as well as the frequency of sizable increases in net worth. In the current episode, the combination of falling stock prices and home prices (in many places) resulted in 67 percent of households experiencing a decline in net worth exceeding 10 percent and in 47 percent of households seeing their net worth fall by at least 20 percent. By contrast, in the 2001 recession, only 15 percent of households experienced a larger than 10 percent decrease in net worth and just 4 percent saw their positions deteriorate by more than20percent. Inthe1990recession,thesharesofhouseholdsexperiencingrelativelylarge decreases in net worth were relatively small, as well. Moreover, a distinguishing feature of thecurrentrecessionisthewidespreadnatureofthedeclinesinhomevalues, which, inlarge part, are responsible for only 2 percent of households experiencing an increase in net worth of 10 percent or more since the 2007 SCF was fielded. By contrast, we estimate that in the 1990 and 2001 recessions, respectively, about 18 percent and 25 percent of households saw their net worth increase by 10 percent or more. The middle section of Table 2 reveals that, as in the 2001 episode, the largest (percentage) declines in household net worth were concentrated among households in the upper quintile of the income distribution. That said, in the current recession, more than half of households in the middle of the income distribution saw their net worth positions fall by 12

20 percent or more, and more than a quarter of households in the bottom of the income distributionexperiencedsuchlargedecreasesinnetworth. Thefrequencyoflargedecreases in net worth among lower- and middle-income households were negligible in the 1990 and 2001 recessions, in large part because many in the middle-income group saw the value of their homes rise in those episodes and because, as we’ll see later, their exposures to the stock market were relatively small. 3.2 Parsing the factors that have pushed down household net worth in the current recession It seems clear that falling home prices in many parts of the country and the drop in stock prices has driven down household net worth in the current recession. After documenting the arithmetic contribution of negative revaluations of these key assets to the drop in net worth across the wealth distribution, this section quantifies perhaps a lesser known channel (albeit, one of secondary importance)—the effect of greater exposure to real estate and corporate equity investments across a wide range of households. 3.2.1 The arithmetic contribution of home prices and stock prices to the drop in household net worth across the distribution of households. The financial strain on household balance sheets in the current recession can be strongly linked to the fact that corporate equity and housing were revalued sharply at (more or less) the same time. By contrast, while many household balance sheets were suffering from negativehousingrevaluationsinthe1989episode, homeownersinotherpartsofthecountry were not; in addition, stock prices rebounded fairly soon in the 1989 recession (and early stages of the recovery) and this served to offset much of the housing revaluation that did occur. In the 2001 recession (and recovery), balance sheets for many households were hit by the sharp equity revaluations, but home price appreciation across most of the country provided a substantial offset. The following table makes this point more directly by comparing estimated revaluations of corporate equity and homes in each of the three business cycles: Comparison of Corporate Equity and Housing Revaluations in the Past 3 Business Cycles 13

Business cycle: Percentage change in average: 1990-91 2001 2008-09 1. Household net worth -5 -6 -28 2. Corporate equity values +10 -21 -37 3. Home values -6 +14 -22 That is, the table shows our estimate that, overall, in the current recession, corporate equityvaluesonhouseholdbalancesheetsfell37percentowingtothenetfallinstockprices from December 2007 to June 2009, while sagging home prices cut the value of housing by 22 percent over the same period. By contrast, in the 2001 episode, stock prices fell 21 percent over a period of similar duration, but this effect on household net worth was substantially offset by a 14 percent average net increase in home values. In the 1990-91, episode, a weak housing market pushed home values down by an average of 6 percent (cumulatively), but a net 10 percent increase in the stock market provided a significant arithmetic offset for household wealth. All told, of the 23 percentage point larger net decrease in average household net worth in 2008-09 (-28 percent) compared with the 1990- 91 episode (-5 percent), we estimate that 10 percentage points can be accounted for by the larger drop in stock prices (-37 percent vs. +10 percent) and that another 10 percentage points reflects the larger drop in home prices (-22 percent vs. -6 percent). The remaining 3 percentage points greater decline in net worth during the current recession compared with the 1990-91 episode (-3 = -23 - -10 - -10) is associated with other changes in the typical household’s balance sheet between the early 1990s and the latter 2000s: In particular, as we discuss next, the key other factors appear to be: greater exposure to the stock market, greater exposure to housing revaluations, and higher balance sheet leverage (greater debt relative to assets). 3.2.2 Increased holdings of real estate and corporate equity investments across the distribution of households. Not only have the revaluations ofhousing and corporate equity been more severe inthe current recession than in the prior two, but ownership of these two important assets increased notably from the late 1980s through the late 1990s. Thus, a wider range of households were more exposed to revaluations than used to be the case. 14

Table 3 documents the broad trends. The rows highlighted in red show that between 1989 and 2001 the share of households owning corporate equity (directly or through a mutual fund or in a retirement account) climbed from 32 percent to 52 percent. Moreover, the increased incidence of corporate-equity ownership occurred among lower-, middle-, and higher-income households and was evident across all age groups. The largest increases (in absolute terms) in ownership between 1989 and 2001 occurred for middle-income (28 percentage points) and younger households (24 percentage points). Overall ownership rates for corporate equity were essentially unchanged between 2001 and 2007, although middleincome and younger households reduced their exposures to corporate equity, on net, this decade. Asashareoftotalassets,households’exposuretocorporateequitydoubled,onnet, from 9 percent in 1989 to 18 percent in 2007. Over this time frame, exposures about tripled for households in the lower- and middle- income groups, while they doubled for those in the top income quintile. From 1989 to 2007, the share of corporate equity in total household assets essentially doubled for households in all three age groups shown in table 3. Table 3 documents a trend toward rising home ownership rates between 1989 and 2007 that was also widespread among households across the income and age distributions, although the 5 percentage point overall increase in the aggregate homeownership rate was much smaller than the rise in corporate equity holdings. From 1989 to 2007, households in each of the three income groups saw their homeownership rates rise by close to 4 percentage points, while the net increase was concentrated among households in the oldest age group (7 percentage points since 1989). Table 4 reports results from some “counterfactual” scenarios intended to gauge the quantitative importance of the size of the revaluations of corporate equity and housing the current recession for the deterioration in balance sheet positions across the distribution of U.S. households. In particular, the left-hand columns of table 4 are repeated from table 2—they simply report the share of households in each of the past three recessions that are estimated to have experienced decreases in net worth exceeding 10 percent or 20 percent respectively and the share whose net worth increased 10 percent or more. The two righthand columns compute the same statistics using data from balance sheet positions from the 1989 and the 2001 SCF waves and the monthly revaluations for corporate equity and housing that occurred in the 2008-09 recession. 15

Thus, table 2 shows that while only 4 percent of households actually experienced decreases in net worth of 20 percent or larger during the 1990 recession, the share would have been 37 percent of households under a counterfactual scenario in which corporate equity and housing revaluations were assumed to follow the same trajectory as in the 2008- 09 episode. In addition, the table shows that using the current recession’s larger negative revaluations of equity and housing rather than the actual experience in the 2001 episode raises the fraction of household experiencing decreases in net worth exceeding 20 percent to 48 percent (the counterfactual calculation) from the 4 percent actually experienced in the 2001 case. The counterfactual scenarios reported in table 4 provide a way to gauge the quantitative importance of the greater exposure of households to corporate equity and housing revaluations. In particular, we can compare estimates of the incidence of sizable decreases in net worth that actually occurred in the current recession (column 3) with the counterfactual simulations based on the SCF balance sheets in 1989 and in 2001 (columns 4 and 5). Given the relatively small shifts in ownership of corporate equity and housing between 2001 and 2007, it is not surprising that the figures in columns 3 and 5 are so similar. However, given the more substantial shifts in ownership that have occurred since 1989, the figures in columns 3 and 4 differ more significantly. These indicate, for example, that had the ownership patterns for corporate equity and housing in 2007 been the same as they were in 1989, the sharp, coincident revaluations in the current episode would have resulted in 37 percent of households experiencing decreases in net worth exceeding 20 percent—this contrasts with the 47 percent of households who actually experienced such a large drop in net worth in the current recession. Lookingattheresultsintable4amonghouseholdsindifferentincomeandagegroups, the SCF data suggest that around four-fifths of the greater severity of the “shock” to net worth in the current recession compared with the prior two can be traced to the size and coincidence of the revaluations to equity and housing; about one-fifth of the greater severity can be traced to the trend toward rising ownership rates for corporate equity and housing since 1989 that left household balance sheets much more exposed to revaluations of those assets this decade than had been the case in the late 1980s. Table 4 shows that for lowerincome households, the relatively steep deterioration in balance sheets seen in this recession 16

almost entirely owes to the more severe equity and housing revaluations that have occurred. By contrast, an increased exposure to equity and home values among the oldest households accounts for almost as much of the greater drop in net worth as does the worse performance of the stock market and the housing market in this recession. 3.3 Increased Leverage Heading into the Current Recession In addition to increased exposure to revaluations associated with rising ownership of corporate equity and housing since the late 1980s, there has been a steady increase in household indebtedness. The red rows in table 5 report trends in the share of households who have any debt on their balance sheets (as well as the component categories, mortgage debt or consumer debt): The share rose from 72 percent in 1989 to 75 percent in 2001 and 77 percent in 2007. This measure of increased leverage was fully accounted for by households in the lower- and middle-income groups, where the shares rose by 8 and 5 percentage points, respectively. However, table 5 shows that the increased incidence of debt between 1989 and 2007 was smallest for households under 45 years of age and was largest for those aged 65 years and older. Although the prevalence of credit card balances and auto loans (included in the “consumer debt” category) increased somewhat, on net between 1989 and 2007—with all of the increase occurring since 2001—most of the overall increase in the incidence of household debt took the form of mortgages. Among all households, the share with a mortgage climbedfrom40percentin1989,to45percentin2001,and49percentin2007. Theshareof middle-income households with an outstanding mortgage balance increased from 47 percent in 1989 to 60 percent in 2007. Net changes in the incidence of consumer debt between 1989 and 2007 were concentrated among households in the lower four quintiles of the income distribution and among households headed by a person between 45 and 64 years old or by a person older than 64 years of age. Table 5 documents significant increases in the levels of mortgage debt and total debt relative to household income from 1989 to 2007, even as median ratios of consumer debt to income were very little changed over the period. For just about every group of households shown in table 5, the ratio of mortgage debt to income doubled, on net, since 1989, and for many of the groups shown, the ratio of total debt to income more than doubled. Thus, 17

the SCF data indicate that rising balance sheet leverage was also one of the reasons for the more-widespread and larger drops in household net worth seen in the current recession.11 Althoughmuchoftheincreaseintheincidenceofmortgagedebt—9percentagepoints acrossallhouseholdsbetween1989and2007—canbetracedtotheriseinhomeownership— 5 percentages points since 1989, table 6 presents evidence on how many more homeowners took on very large mortgages over the period. For example, at the time the 2007 SCF was conducted, about 7 percent of homeowners reported mortgage debt exceeding 90 percent of their home’s value; this fraction had been 7 percent in the 2001 and 2004 SCFs, but it was under 4 percent and under 5 percent, respectively, in the 1989 and 1992 surveys.12 Of course, the substantial declines in home values across most parts of the country since the 2007 SCF have pushed up the fraction of homeowners with relatively large mortgages: We estimate that the share with home loan-to-value ratios above 90 percent more than doubled to 17 percent by the middle of 2009 and that the share whose owe more than their homes are worth (LTV > 100 percent) has soared from just 1 percent at the time of the 2007 SCF to around 12 percent recently. There can be little doubt the swing to very low and frequently negative equity positions has been a major factor contributing to the rise in mortgage defaults over the past year-and-a-half.13 By our calculations, roughly 35 percent of homeowners younger than 45 years of age now have mortgages exceeding 90 percent of their home’s value and around 25 percent of younger homeowners appear to owe more than their homes are currently worth. Table 7 takes a different look at the trends in credit and payment problems in the SCF to gauge the vulnerability of households heading into the 2008-09 recession compared with the prior two. More specifically, table 7 reports the fraction of households reporting having been turned down for credit in the past few years, the fraction with debt payments (interestandprincipleonmortgages,creditcardbalances,andstudentloansandautoloans) exceeding 40 percent of their income, and the fraction reporting being 60 days or more late on any payment (including utilities, medical bills, or other “non-debt” payments). The 11To be more explicit, greater balance sheet leverage (that is, more debt relative to assets) means that, allalsebeingequal,agivenpercentagerevaluationinanassetwilltranslateintoalargerpercentagechange in net worth. 12In table 6, the terms “loan-to-value” and “LTV” represent the ratio of homeowners’ mortgage debt to the reported value of their homes. 13For research on this issue, see Mayer, Pence, and Sherlund (7), Sherlund (9), and Gerardi, Lehnert, Sherlund, and Willen (4). 18

table also reports the fraction of households reporting any of those indications of having been financially “over-extended” or vulnerable. Near the business cycle peak in 2007, 13 percent of households reported having been turned down for credit in recent months—a fraction that was virtually the same as in the 1989 and 2001 surveys. The fraction of households reporting being 60 days or more late on any of their payments was also virtually the same in the 1989, 2001, and 2007 surveys. At face value, these measures do not indicate substantially greater financial vulnerability heading into the current recession than in the previous two downturns. However, the 2007 SCF did show a larger fraction of households reporting debt paymentsexceeding40percentonincome—15percentofallhouseholdsin2007,comparedwith 12 percent in 2001 and 10 percent in 1989. Among households in the lower two quintiles of the income distribution, the fraction with such high debt payments relative to income was 23 percent in 2007, not very different than the 22 percent share in 2001, but more noticeably above the 19 percent share in 1989. The share of middle-income households with heavy debt payments rose to 14 percent in 2007, from 9 percent in 2001 and 8 percent in 1989. Table 7 documents that the net rise in the share of households with relatively heavy debt payments from 1989 to 2007 was evenly distributed across the age distribution. The share of households reporting any of the three financial vulnerabilities fluctuated in a fairly narrow range around 28 percent in all of the SCF surveys conducted from 1989 through 2007, indicating no strong overall trend in payment problems. Near the 2007 peak inbusinesscycleactivity,43percentofhouseholdsinthebottomtwoquintilesoftheincome distribution reported at least one of the three indicators of financial vulnerability, a share that was no different (statistically speaking) than in any of the prior four surveys. At 35 percent, the share of younger households reporting at least one of the credit or payment problems in 2007, was 3 percentage points greater than in 2001 and 4 percentage points higher than in 1989. Of course, the period leading up to 2007 is recognized for the ease in which households obtained credit, which could explain the relatively low incidence of households reporting having recently been turned down for credit and having remained “current” of their obligations (news loans generally carry very low delinquency rates). 19

3.4 Indications of substantial financial stress Table 8 presents our final calibration of how much more financially stressful the current recession has been compared with the previous two. The left-hand columns simply report thefractionofhouseholdswhoreportedatleastoneofthefinancialvulnerabilitiesqueriedin the most recent SCF survey (debt payments > 40 percent of income or an existing credit or paymentproblem)and isestimatedtohavesufferedadecreaseinnetworthof20percentor more. Ascanbeseen,inthe1990andthe2001recessions,weestimatethatonly1percentof households met this definition of substantial financial stress. By contrast, as of June 2009, at least 13 percent of households are estimated to have. We say “at least” because, no doubt, job losses and other recession-related changes in economic conditions have raised the existence of credit and payment problems well above the levels measured in the 2007 SCF. The right-hand columns repeat our counterfactual estimates that rely on the distribution of balance sheets and existing credit problems measured in the 1989 and 2001 SCFs, but that impose the changes in corporate equity and housing values measured for the 2008 and 2009. Again, the counterfactuals imply that most, but not all, of the greater incidence in our measure of substantial financial stress in the current recession reflects the confluence of sharp declines in the stock and housing markets. Had those price declines occurred with 2001’s “existing” balance sheets and credit and payment problems, we estimate that 11 percentofhouseholdswould’vecomeundersubstantialfinancialstressduringthatrecession. In 1989, our estimate is 8 percent of households, suggesting that about two-fifths of the greater incidence of substantial financial stress in the current recession reflects the greater financial vulnerability of households in 2007 compared with 1989 (2 ≈ 13−8). The 2001 5 13−1 counterfactualsuggeststhataboutone-sixthofthegreaterincidenceofsubstantialfinancial stressinthecurrentrecessionreflectsgreatervulnerabilityin2007thatin2001(1 ≈ 13−11). 6 13−1 3.5 A caveat We should emphasize that data from the 2007 SCF permit very little to be said with firm conviction about the role that balance sheet leverage may have played in the current recession. This is because any amplifying effects that may have been caused by leverage at the onset of the recession might not have occurred until after households had experienced job losses or drops in incomes to levels that could no longer support timely debt payments. 20

In other words, this is a case where specific follow-ups with more-leveraged households and withthosewhoseincomesdroppedthemostisneededtounderstandthefullramificationsof their“pre-existing”balance-sheetconditions. Thenextsectionofthepaperbrieflydescribes a Federal Reserve initiative to supplement the 2007 SCF with just that type of follow-up information. 3.6 Looking Ahead: Forthcoming Information from the 2009 SCF Panel Survey In an effort to better understand the exceptional economic circumstances in 2008 and 2009, the Federal Reserve Board sponsored follow-up interviews for households who participated inthe2007SCF.Theseinterviewspromisetoprovideimportantinformationonthedistribution of changes in economic conditions—employment and income, balance sheet positions and other measures of financial stress—and indications of how households responded to them—including possible changes in saving, spending, and rebalancing of investments. The follow-up interviews began in the summer of 2009 and will be completed before the end of the year; comprehensive data are currently expected to be ready by the middle of 2010. As suggested above, having two interviews with each households should allow a more focussed analysis of households’ various responses to the financial turmoil and sharper conclusions abouttherolethatbalancesheetleverageandotherindicatorsoffinancialstressandchanging credit conditions may have played in amplifying and propogating “shocks” to equity and home valuations. 4 Discussion This paper documents how much larger and more pervasive substantial decreases in household net worth have been in the 2008-09 U.S. recession compared with the downturns in 2001 and in 1990-91. Arithmetically, the more sizable decreases in wealth across a wide range of households in the current episode was traced primarily to large net decreases in corporate equity and home values. A key quantitative result is that in the 2001 recession, rising home values provided a significant (but incomplete) financial offset to falling corporate quity values, while in the 1990-91 episode, rising equity values offset to a considerable 21

extent the balance sheet consequences of generally falling home values. We also show that increased exposure to equity and housing revaluations and to somewhat more-leveraged positions also contributed to the greater deterioration in household net worth in the current episode, but that, quantitatively, these factors played more of a secondary role. In terms of the macroeconomic consequences of our results, one issue is how households will respond to the particularly acute net deterioration in their financial positions since the peak in overall business cycle activity late in 2007. According to much commentary by analysts and forecasters, the contour for broad macroeconomic activity seems likely to be strongly influenced by the path of consumer spending. Moreoever, forecasters seem to have quite a range of opinions and express significant uncertainty about the outlook for consumer spending and for the trajectory of the personal saving rate.14 Indeed, some experts expect a considerable rise in the saving rate fairly soon (for example, see Keene and Walker ((6)) or Glick and Lansing (5)). The context for much of the discussion surrounds the prospects for the pace of household deleveraging or “balance sheet repair,” and, indeed, our analysis of the SCF data suggests that the pervasive and substantial deterioration in household net worth in the current business cycle suggests that such a focus might well be justified. That said, our analysis of the extant cross-section data in the SCF cannot speak directly to questions regarding where household savings rates will end up—that is, whether a pronounced adjustment to a “permanently” higher saving rate could be in train—or how quickly any adjustment might take place. However, we think analysis of the forthcoming data from the 2009 SCF re-interviews will provide some important, albeit stillcircumstantial, evidence on these issues. For example, those data should help us identify what types of adjustments households have already taken in reaction to the sharp revaluations of key assets on their balance sheets. Although the re-interview will only collect indicators of spending, we hope to be able to parse to an extent some of the adjustments 14A range of fairly detailed private-sector forecasts are summarized each month in Moore (8); the Federal Reserve Bank of Philadelphia reports a Survey of Professional Forecasters near the middle of each calendar quarter (http://www.phil.frb.org/research-and-data/real-time-center/ survey-of-professional-forecasters/). In addition, the minutes from meetings of the Federal Open Market Committee (FOMC) in January, April, August, and November include a “Summary of Economic Projections”—forecasts of real GDP, the unemployment rate, and consumer price inflation provided by the BoardofGovernorsandthepresidentsofthetwelveFederalReserveBanks(http://federalreserve.gov/ monetarypolicy/fomccalendars.htm). 22

households made in response to different changes in their economic circumstances: revaluations to their holdings of corporate equity and housing, lay-offs and reductions in wages and hours of work, and problems servicing existing debt and in accessing usual sources of credit. References [1] Bucks, Brian K., Arthur B. Kennickell, Traci L. Mach, and Kevin B. Moore. 2009. “Changes in U.S. Family Finances from 2004 to 2007: Evidence from the Survey of Consumer Finances,” Federal Reserve Bulletin 95, 1st Quarter, pp. A1-55. http:// www.federalreserve.gov/pubs/bulletin/2009/pdf/scf09.pdf. [2] Dynan, Karen E. 2009. “Changing Household Financial Opportunities and Economic Security”, Journal of Economic Perspectives 23, number 3, Fall, pp. 49-68. [3] Dynan, Karen E., and Douglas W. Elmendorf, “Do Provisional Estimates of Output Miss Economic Turning Points?” Finance and Economics Discussion Series 2001-52, FederalReserveBoard.http://www.federalreserve.gov/pubs/feds/2001/200152/ 200152pap.pdf. [4] Gerardi, Kristopher, Andreas Lehnert, Shane M. Sherlund, and Paul Willen. 2008. “Making Sense of the Subprime Crisis,” Brookings Papers on Economic Activity, Fall, pp. 69-145. [5] Glick, Reuven, and Kevin J. Lansing. 2009. “U.S. Household Deleveraging and Future Consumption Growth,” FRBSF Economic Letter,” number 2009-16, May 15, 2009. http://www.frbsf.org/publications/economics/letter/2009/el2009-16.pdf. [6] Keene, Thomas, and Susanne Walker, “Pimco’s Clarida Says U.S. Savings Rate May Exceed8%,” Bloomberg.com, September28, 2009.http://www.bloomberg.com/apps/ news?pid=20601087&sid=aHM.2Uhxnnr4. [7] Mayer, Christopher, Karen Pence, and Shane M. Sherlund. 2009. “The Rise in Mortgage Defaults,” Journal of Economic Perspectives 23, number 1, Winter, pp. 27-50. 23

[8] Moore, Randall E. 2009. “Blue Chip Economic Indicators” (ISSN: 0193-4600), New York: Aspen Publishers. [9] Sherlund, Shane M. 2008. “The Past, Present, and Future of Subprime Mortgages,” Finance and Economics Discussion Series 2008-63, Federal Reserve Board. http:// www.federalreserve.gov/pubs/feds/2008/200863/200863pap.pdf. 24

1 elbaT secnaniF remusnoC fo yevruS eht ni htroW teN dlohesuoH ni segnahC fo nosirapmoC A skaeP elcyC ssenisuB REBN dnuora ot 70-voN ot 10-raM ot 09-luJ )FCS 7002( )FCS 2991( 90-nuJ 20-voN 29-voN 90-nuJ 70-voN 20-voN 10-raM 29-voN 09-luJ :nrutnwod lacilcyC )egnahc tnecrep( )srallod 7002 fo sdnasuoht( sdlohesuoh llA 03- 7 3- 48 121 601 99 27 47 htrow ten naideM 82- 6- 5- 204 855 534 364 072 582 htrow ten naeM elitnecrep emocnI 04-0 32- 11 61 61 02 52 32 91 61 htrow ten naideM 32- 2 2- 29 021 99 79 07 27 htrow ten naeM 08-04 92- 5 2- 89 931 121 511 38 58 htrow ten naideM 72- 1- 4- 312 392 062 362 571 381 htrow ten naeM 001-08 72- 0 01- 734 995 694 394 772 803 htrow ten naideM 92- 9- 6- 9931 2691 8541 6951 958 419 htrow ten naeM daeh fo egA sraey 5544 << 24- 31 3- 81 13 14 63 92 03 htrow ten naideM 43- 4- 7- 831 112 591 302 521 531 htrow ten naeM sraey 46 ot sraey 54 72- 7 9- 851 612 281 171 331 641 htrow ten naideM 82- 7- 4- 865 887 136 086 434 254 htrow ten naeM sraey 46 > 31- 4 4 191 122 002 391 331 821 htrow ten naideM 52- 6- 7- 126 628 336 076 073 793 htrow ten naeM tseraen eht ni atad level-dlohesuoh gnisu gnitcejorp neeb evah 9002 enuJ dna ,2002 rebmevoN ,1002 hcraM ,0991 yluJ rof seulaV :etoN .sexedni ecirp esuoh lanoiger dna sexedni ecirp kcots ni segnahc dna evaw FCS 25

Table 2 A Comparison of the Frequency of Large Changes in Household Net Worth in the Survey of Consumer Finances around NBER Business Cycle Peaks Cyclical downturn: Jul-90 to 92 SCF Mar-01 to Nov-02 07 SCF to Jun-09 (percent of households) All households Net worth down >= 10% 6 15 67 Net worth down >= 20% 4 4 47 Net worth up >= 10% 18 25 2 Income percentile 0-40 Net worth down >= 10% 10 13 45 Net worth down >= 20% 9 8 28 Net worth up >= 10% 16 25 5 40-80 Net worth down >= 10% 4 12 77 Net worth down >= 20% 1 2 54 Net worth up >= 10% 17 27 0 80-100 Net worth down >= 10% 4 23 94 Net worth down >= 20% 1 1 71 Net worth up >= 10% 23 23 0 Age of head < 45 years Net worth down >= 10% 8 17 60 Net worth down >= 20% 5 7 48 Net worth up >= 10% 22 27 3 45 years to 64 years Net worth down >= 10% 5 15 77 Net worth down >= 20% 4 2 53 Net worth up >= 10% 18 24 1 > 64 years Net worth down >= 10% 3 11 65 Net worth down >= 20% 3 3 36 Net worth up >= 10% 9 23 2 26

Table 3 Trends in Ownership of Selected Assets in the SCF from 1989 to 2007 SCF Wave Change from: 1989 2001 2007 1989 to 2001 1989 to 2007 (percent) (percentage points) All households Corporate equity (directly or indirectly held) Share of households with balance > 0 32 52 51 21 19 Share of total assets 9 24 18 15 10 Primary residence Share of households with balance > 0 64 68 69 4 5 Share of total assets 32 27 32 -5 0 Income percentile 0-40 Corporate equity (directly or indirectly held) Share of households with balance > 0 9 24 24 14 15 Share of total assets 3 11 9 8 6 Primary residence Share of households with balance > 0 44 49 48 5 4 Share of total assets 47 48 50 0 3 40-80 Corporate equity (directly or indirectly held) Share of households with balance > 0 36 64 60 28 24 Share of total assets 5 21 15 15 9 Primary residence Share of households with balance > 0 71 74 77 3 6 Share of total assets 43 38 46 -5 4 80-100 Corporate equity (directly or indirectly held) Share of households with balance > 0 67 86 88 19 21 Share of total assets 11 27 20 16 10 Primary residence Share of households with balance > 0 90 93 93 3 4 Share of total assets 25 20 24 -5 -1 Age of Head < 45 years Corporate equity (directly or indirectly held) Share of households with balance > 0 30 54 46 25 16 Share of total assets 6 18 11 12 5 Primary residence Share of households with balance > 0 51 54 53 3 2 Share of total assets 42 37 45 -5 3 45 years to 64 years Corporate equity (directly or indirectly held) Share of households with balance > 0 39 59 60 19 21 Share of total assets 9 25 19 16 10 Primary residence Share of households with balance > 0 78 79 79 1 1 Share of total assets 30 24 29 -5 -1 > 64 years Corporate equity (directly or indirectly held) Share of households with balance > 0 26 38 46 11 20 Share of total assets 11 26 21 16 11 Primary residence Share of households with balance > 0 74 79 81 5 7 Share of total assets 25 24 28 -2 3 27

Table 4 Actual and Counterfactual Frequencies of Large Changes in Household Net Worth around NBER Business Cycle Peaks Counterfactuals based on asset price SCF-based Estimates: changes since 2007 and: Jul-90 to 92 Mar-01 to 07 SCF to Cyclical downturn: SCF Nov-02 Jun-09 1989 SCF 2001 SCF (percent of households) All households Net worth down >= 10% 6 15 67 59 66 Net worth down >= 20% 4 4 47 37 48 Net worth up >= 10% 18 25 2 4 3 Income percentile 0-40 Net worth down >= 10% 10 13 45 41 47 Net worth down >= 20% 9 8 28 25 30 Net worth up >= 10% 16 25 5 10 7 40-80 Net worth down >= 10% 4 12 77 63 73 Net worth down >= 20% 1 2 54 40 52 Net worth up >= 10% 17 27 0 0 0 80-100 Net worth down >= 10% 4 23 94 85 92 Net worth down >= 20% 1 1 71 56 75 Net worth up >= 10% 23 23 0 0 0 Age of head < 45 years Net worth down >= 10% 8 17 60 56 64 Net worth down >= 20% 5 7 48 42 52 Net worth up >= 10% 22 27 3 5 4 45 years to 64 years Net worth down >= 10% 5 15 77 71 73 Net worth down >= 20% 4 2 53 43 50 Net worth up >= 10% 18 24 1 4 1 > 64 years Net worth down >= 10% 3 11 65 49 61 Net worth down >= 20% 3 3 36 19 34 Net worth up >= 10% 9 23 2 2 2 28

Table 5 Trends in Household Debt in the SCF from 1989 to 2007 SCF Wave Change from: 1989 2001 2007 1989 to 2001 1989 to 2007 (percent or ratio) (percentage points or change in ratio) All households Debt balance > $0 72 75 77 3 5 Mortgage debt 40 45 49 5 9 Consumer debt 63 63 66 0 3 Median ratio of debt to income 0.5 0.8 1.1 0.3 0.6 Mortgage debt/income 0.8 1.1 1.5 0.3 0.7 Consumer debt/income 0.3 0.2 0.2 -0.1 -0.1 Income percentile 0-40 Debt balance > $0 53 60 61 7 8 Mortgage debt 16 20 22 5 7 Consumer debt 48 53 52 5 5 Median ratio of debt to income 0.3 0.5 0.7 0.2 0.3 Mortgage debt/income 1.0 1.8 2.1 0.8 1.1 Consumer debt/income 0.2 0.2 0.3 0.1 0.1 40-80 Debt balance > $0 82 84 87 2 5 Mortgage debt 47 53 60 6 13 Consumer debt 74 72 77 -1 4 Median ratio of debt to income 0.6 0.9 1.4 0.3 0.8 Mortgage debt/income 0.9 1.2 1.6 0.4 0.8 Consumer debt/income 0.2 0.2 0.2 0.0 0.1 80-100 Debt balance > $0 91 88 89 -2 -2 Mortgage debt 72 76 79 4 7 Consumer debt 71 62 68 -8 -3 Median ratio of debt to income 0.6 0.9 1.2 0.2 0.6 Mortgage debt/income 0.6 0.8 1.1 0.2 0.5 Consumer debt/income 0.1 0.1 0.1 0.0 0.0 Age of Head < 45 years Debt balance > $0 84 86 85 2 1 Mortgage debt 45 48 48 3 3 Consumer debt 77 75 76 -1 -1 Median ratio of debt to income 0.6 0.9 1.3 0.3 0.7 Mortgage debt/income 0.9 1.2 1.9 0.3 1.0 Consumer debt/income 0.2 0.2 0.3 0.1 0.1 45 years to 64 years Debt balance > $0 78 81 85 3 6 Mortgage debt 48 56 61 7 13 Consumer debt 64 64 71 1 7 Median ratio of debt to income 0.5 0.8 1.1 0.3 0.6 Mortgage debt/income 0.6 1.0 1.3 0.4 0.7 Consumer debt/income 0.1 0.2 0.2 0.0 0.0 > 64 years Debt balance > $0 38 43 48 6 11 Mortgage debt 15 21 28 6 13 Consumer debt 29 33 37 4 7 Median ratio of debt to income 0.2 0.4 0.7 0.1 0.5 Mortgage debt/income 0.3 0.9 1.3 0.6 1.0 Consumer debt/income 0.1 0.1 0.1 0.0 0.1 29

6 elbaT 8002 ot 9891 morf )egareveL gnisuoH( soitaR eulaV-ot-naoL emoH ni sdnerT FCS ,daeH fo egA dna elitnecreP emocnI yb ,%09 => VTL htiw srenwoemoH FCS FCS FCS FCS FCS 90-nuJ 7002 4002 20-voN 1002 10-raM 2991 09-luJ 9891 )sdlohesuoh fo tnecrep( 71 7 7 3 7 8 5 3 4 sdlohesuoh llA elitnecrep emocnI 01 5 5 4 6 7 2 2 3 04-0 22 9 9 4 9 01 6 4 6 08-04 71 5 5 1 3 5 5 1 1 001-08 daeh fo egA 53 61 41 7 31 51 01 6 8 sraey 54 < 31 4 5 2 4 5 2 1 2 sraey 46 ot sraey 54 4 1 0 1 2 2 0 0 0 sraey 46 > FCS ,daeH fo egA dna elitnecreP emocnI yb ,%001 > VTL htiw srenwoemoH FCS FCS FCS FCS FCS 90-nuJ 7002 4002 20-voN 1002 10-raM 2991 55033 9891 )sdlohesuoh fo tnecrep( 21 1 1 1 2 3 1 1 1 sdlohesuoh llA elitnecrep emocnI 8 1 1 1 2 4 0 1 1 04-0 51 1 1 1 2 3 1 2 2 08-04 21 1 1 0 0 1 1 1 1 001-08 daeh fo egA 52 2 2 2 3 5 1 3 3 sraey 54 < 8 1 1 1 1 2 1 0 0 sraey 46 ot sraey 54 2 0 0 0 0 1 0 0 0 sraey 46 > 30

Table 7 Trends in Credit and Payment Problems from 1989 to 2007 Change, SCF wave: 1989 1992 2001 2004 2007 1989 to 2007 (percent of households who have any debt) (percentage pts) All households Turned down for credit recently 14 18 14 15 13 -1 Debt payments > 40% of income 10 11 12 12 15 5 60 days late on any payment 7 6 7 9 7 0 Any of the above problems 26 29 27 29 28 2 Income percentile 0-40 Turned down for credit recently 20 23 20 20 18 -1 Debt payments > 40% of income 19 21 22 22 23 4 60 days late on any payment 15 10 12 15 13 -2 Any of the above problems 43 43 43 44 43 -1 40-80 Turned down for credit recently 15 18 14 16 13 -2 Debt payments > 40% of income 8 9 9 10 14 5 60 days late on any payment 5 6 6 9 6 1 Any of the above problems 24 28 24 28 27 3 80-100 Turned down for credit recently 5 10 7 5 4 -1 Debt payments > 40% of income 3 3 3 2 6 3 60 days late on any payment 2 1 2 1 1 -1 Any of the above problems 9 13 10 7 11 2 Age of Head < 45 years Turned down for credit recently 19 23 20 20 19 0 Debt payments > 40% of income 10 12 11 13 14 4 60 days late on any payment 9 8 9 13 9 0 Any of the above problems 31 35 32 35 35 3 45 years to 64 years Turned down for credit recently 9 13 9 12 9 0 Debt payments > 40% of income 10 12 12 12 15 5 60 days late on any payment 6 5 7 6 6 1 Any of the above problems 20 25 22 25 24 5 45 years to 64 years Turned down for credit recently 4 6 5 3 4 0 Debt payments > 40% of income 9 9 15 11 15 6 60 days late on any payment 3 1 1 4 3 0 Any of the above problems 15 13 18 15 19 4 31

8 elbaT snoisseceR eerhT tsaP eht ni ssertS laicnaniF laitnatsbuS fo srotacidnI detroper dah dna erom ro %02 deppord htrow ten esohw sdlohesuoh fo tnecreP smelborp tnemyap ro tiderc gnitsixe ro emocni fo %04 gnideecxe stnemyap tbed ecirp tessa no desab slautcafretnuoC setamitse desab-FCS :dna 7002 ecnis segnahc dna FCS 7002 dna FCS 1002 dna FCS 9891 FCS 1002 FCS 9891 90-nuJ ot FCS 70 20-voN ot FCS 10 FCS 20 ot 09-luJ :nrutnwod lacilcyC 11 8 31 1 1 sdlohesuoH llA elitnecrep emocnI 11 7 21 1 1 04-0 31 11 61 1 8.0 08-04 7 6 8 2.0 1 001-08 daeH fo egA 41 11 51 2 1 sraey 54 < 11 8 41 3.0 7.0 sraey 46 ot sraey 54 4 2 5 0.0 2.0 sraey 46 > 32

Figure 1 Selected Measures of Economic Activity around NBER Business Cycle Peaks Macroeconomic Activity Before and After NBER Peaks Total Payroll Employment Real Income from Wages and Salary Index: 100 = Business cycle peak Index: 100 = Business cycle peak 102 104 July 1990 Peak July 1990 Peak March 2001 Peak 101 March 2001 Peak December 2007 Peak December 2007 Peak 102 100 100 99 98 98 97 96 96 94 95 94 92 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 20 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 Months from NBER peak Months from NBER peak Source. Bureau of Labor Statistics. Source. Bureau of Economic Analysis. Real Disposable Personal Income Real Personal Consumption Expenditures Index: 100 = Business cycle peak Index: 100 = Business cycle peak 109 106 July 1990 Peak M De a c rc e h m 2 b 0 e 0 r 1 2 0 P 0 e 7 a k Peak 107 104 105 102 103 100 101 98 99 July 1990 Peak March 2001 Peak 97 December 2007 Peak 96 95 94 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 Months from NBER peak Months from NBER peak Source. Bureau of Economic Analysis. Source. Bureau of Economic Analysis. Real Residential Investment Sales of New Homes Index: 100 = Business cycle peak Index: 100 = Business cycle peak 160 150 J M u a ly rc 1 h 9 2 9 0 0 0 P 1 e P a e k ak 160 December 2007 Peak 140 140 130 120 120 110 100 100 90 80 1990:Q3 Peak 2001:Q1 Peak 80 2007:Q4 Peak 60 70 60 40 −8 −6 −4 −2 peak 2 3 4 5 6 7 8 9 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 20 Quarters from NBER peak Months from NBER peak Source. Census Bureau. Source. Census Bureau. 33

Figure 2 Selected Balance Sheet Items around NBER Business Cycle Peaks Balance Sheet Items Before and After NBER Peaks Personal Saving Rate Difference relative to business cycle peak Household Net Worth (percentage points of disposable income) Index: 100 = Business cycle peak 7 120 July 1990 Peak March 2001 Peak 6 115 December 2007 Peak 5 110 4 105 3 100 2 95 1 90 0 −1 1 2 9 0 9 0 0 1 : : Q Q 3 1 P P e e a a k k 85 2007:Q4 Peak −2 80 −3 75 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 −8 −6 −4 −2 peak 2 3 4 5 6 7 8 9 Months from NBER peak Quarters from NBER peak Source. Bureau of Economic Analysis. Source. Federal Reserve Board, Flow of Funds. Stock Prices House Prices Index: 100 = Business cycle peak Index: 100 = Business cycle peak July 1990 Peak 120 July 1990 Peak 135 March 2001 Peak March 2001 Peak December 2007 Peak December 2007 Peak 125 115 110 105 95 100 85 75 90 65 55 45 80 −12 −9 −6 −3 1 3 5 7 9 11 14 17 20 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 20 Months from NBER peak Months from NBER peak Source. Dow Jones. Source. First American CoreLogic. Home Mortgage Debt Consumer Credit Index: 100 = Business cycle peak Index: 100 = Business cycle peak 135 115 1990:Q3 Peak 2 2 0 0 0 0 1 7 : : Q Q 1 4 P P e e a a k k 125 110 105 115 100 105 95 95 90 July 1990 Peak March 2001 Peak 85 December 2007 Peak 85 75 80 −8 −6 −4 −2 peak 2 3 4 5 6 7 8 9 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 Quarters from NBER peak Months from NBER peak Source. Federal Reserve Board, Flow of Funds. Source. Federal Reserve Board. 34

Figure 3 Selected Measures of Household Credit Performance around NBER Business Cycle Peaks Household Credit Performance Before and After NBER Peaks Mortgage Delinquency Rate Credit Card Delinquency Rate Difference relative to business cycle peak Difference relative to business cycle peak (percentage points) (percentage points) 2.5 1990:Q3 Peak 2.0 2001:Q1 Peak 2007:Q4 Peak 2.0 1.5 1.5 1.0 1.0 0.5 0.0 0.5 −0.5 0.0 July 1990 Peak −1.0 March 2001 Peak −0.5 December 2007 Peak −1.5 −1.0 −2.0 −8 −6 −4 −2 peak 2 3 4 5 6 7 8 9 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 20 Quarters from NBER peak Months from NBER peak Source. Mortgage Bankers Association. Source. Moody’s Investors Service. Prime Auto Loan Charge−Off Rate Household Bankruptcies Difference relative to business cycle peak Difference relative to business cycle peak (percentage points) (thousands) 1.25 125 1.00 100 0.75 75 50 0.50 25 0.25 0 0.00 −25 −0.25 −50 March 2001 Peak 2001:Q1 Peak December 2007 Peak −0.50 2007:Q4 Peak −75 −0.75 −100 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 −8 −6 −4 −2 peak 2 3 4 5 6 7 8 9 Months from NBER peak Quarters from NBER peak Source. Moody’s Investors Service. Source. Lundquist Consulting, Inc. Seasonally adjusted by Board staff. Consumer Sentiment Consumer Confidence Index: 100 = Business cycle peak Index: 100 = Business cycle peak 140 130 July 1990 Peak July 1990 Peak 120 March 2001 Peak 130 March 2001 Peak December 2007 Peak December 2007 Peak 110 120 100 110 90 80 100 70 90 60 80 50 40 70 30 60 20 −12 −9 −6 −3 1 3 5 7 9 11 14 17 20 −12 −9 −6 −3 1 3 5 7 9 11 14 17 20 Months from NBER peak Months from NBER peak Source. Reuters/University of Michigan. Source. Conference Board. 35

4 erugiF 9002-9891 ,secnaniF remusnoC fo yevruS eht ni htroW teN dlohesuoH 002 061 021 08 04 srallod 7002 fo sdnasuohT 006 084 063 042 021 9 0 − n u J 7 0 4 0 2 0 1− 0v 1 0 o N − ra M 8 9 5 9 2 9 0 9 − lu 9 J 8 srallod 7002 fo sdnasuohT raeY htrow ten naideM htrow ten naeM 36

5 erugiF noitubirtsiD htroW teN eht ssorcA htroW teN ni segnahC 01 0 01− 02− 03− 04− egnahc egatnecreP )FCS 2991( 29 voN ot 09 luJ 01 0 01− 02− 03− 04− ht59 ht57 ht05 ht52 selitnecrep htrow teN egnahc egatnecreP 20 voN ot 10 raM 01 0 01− 02− 03− 04− ht59 ht57 ht05 ht52 selitnecrep htrow teN egnahc egatnecreP 90 nuJ ot )FCS 7002( 70 voN ht59 ht57 ht05 ht52 selitnecrep htrow teN 37

Appendix A Changes in Selected Components of Disposable Personal Income Chanagreos uinn EdcoNnoBmEic RActBivuitys iBneefosrse aCndy cAlfeterP NeBaEkRs Peaks Components of Personal Income Transfer Payments Rental Income Index: 100 = Business cycle peak Index: 100 = Business cycle peak 130 175 150 120 125 110 100 100 July 1990 Peak July 1990 Peak 75 March 2001 Peak March 2001 Peak December 2007 Peak December 2007 Peak 90 50 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 Months from NBER peak Months from NBER peak Interest Income Personal Dividend Income Index: 100 = Business cycle peak Index: 100 = Business cycle peak 105 110 100 100 90 95 80 July 1990 Peak 90 July 1990 Peak March 2001 Peak March 2001 Peak 70 December 2007 Peak December 2007 Peak 85 60 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 Months from NBER peak Months from NBER peak Contributions for Social Insurance Personal Current Taxes Index: 100 = Business cycle peak Index: 100 = Business cycle peak 110 110 July 1990 Peak July 1990 Peak March 2001 Peak March 2001 Peak December 2007 Peak December 2007 Peak 100 105 90 100 80 95 70 90 60 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 −12 −9 −6 −3 peak 3 5 7 9 11 14 17 Months from NBER peak Months from NBER peak Source: Bureau of Economic Analysis. 38

B xidneppA eht ni htroW teN dlohesuoH ni egnahC fo nosirapmoC A skaeP elcyC ssenisuB REBN dnuora secnaniF remusnoC fo yevruS ot FCS 70 ot 20-voN ot 10-raM ot 10-raM 29 ot 09-luJ ot FCS 98 FCS FCS FCS FCS FCS 90-nuJ FCS 40 20-voN FCS 10 FCS 09-luJ 90-nuJ 7002 4002 20-voN 1002 10-raM 2991 09-luJ 9891 :doirep emiT )egnahc tnecrep( )srallod 7002 fo sdnasuoht( sdlohesuoh llA 03- 3- 7 3 3- 3- 48 121 201 601 101 99 27 47 67 htrow ten naideM 82- 31 6- 0 5- 5- 204 855 494 534 564 364 072 582 992 htrow ten naeM elitnecrep emocnI 04-0 32- 92- 11 3 61 91 61 02 81 52 42 32 91 61 41 htrow ten naideM 32- 8 2 2 2- 2- 29 021 701 99 89 79 07 27 37 htrow ten naeM 08-04 92- 8 5 2 2- 2- 89 931 131 121 711 511 38 58 78 htrow ten naideM 72- 31 1- 2 4- 4- 312 392 492 062 762 362 571 381 191 htrow ten naeM 001-08 72- 41 0 2 01- 9- 734 995 365 694 505 394 772 803 933 htrow ten naideM 92- 41 9p 6- 6- 9931 2691 5661 8541 3951 6951 958 419 969 htrow ten naeM daeh fo egA sraey 54 < 24- 31- 31 5 3- 3- 81 13 63 14 83 63 92 03 13 htrow ten naideM 43- 3 4- 0 7- 7- 831 112 102 591 402 302 521 531 641 htrow ten naeM sraey 46 ot sraey 54 72- 7 7 3 9- 8- 851 612 691 281 671 171 331 641 851 htrow ten naideM 82- 71 7- 0 4- 4- 865 887 937 136 286 086 434 254 964 htrow ten naeM sraey +56 3311- 22- 44 22 44 44 119911 112222 559911 000022 779911 339911 333311 882211 332211 hhttrrooww tteenn nnaaiiddeeMM 52- 6 6- 0 7- 6- 126 628 866 336 376 076 073 793 524 htrow ten naeM tseraen eht ni atad level-dlohesuoh gnisu gnitcejorp neeb evah 9002 enuJ dna ,2002 rebmevoN ,1002 hcraM ,0991 yluJ rof seulaV :etoN .sexedni ecirp esuoh lanoiger dna sexedni ecirp kcots ni segnahc dna evaw FCS 39