Housing Activity and Consumer Spending

6 Business Economics • April 2007

Jonathan McCarthy is a researchofficer in the Macroeconomic andMonetary Studies function of theResearch and Statistics Group at theFederal Reserve Bank of New York,a post he has held since January2007. He has been at the New YorkFed since September 1992, afterreceiving his Ph.D. in economicsfrom the University of Wisconsin-Madison in May 1992. He was avisiting economist at the Bank forInternational Settlements in Basel,Switzerland, in 1997-98.

Charles Steindel is a senior vicepresident in the research and marketanalysis group of the Federal

Reserve Bank of New York. He had been a vice president inthe research area since July 1995. He joined the bank in1986 and has been an officer since 1990. He holds a B.S.degree in mathematics from Emory University and a Ph.D.in economics from Massachusetts Institute of Technology.

The current expansion has seen record-high levels oftransactions in housing, extraordinary growth in the ag-gregate value of owner-occupied housing, and large in-creases in the amount of funds realized from therefinancing of mortgage debt. Many analysts thus havepointed to the strong housing market and rising homeprices as a major pillar supporting recent economic growthand have expressed concern that a contraction in housing

activity and values could pose a significant risk to con-sumer spending and real economic growth. This paper ex-plores the channels by which the housing market mayaffect consumer spending and assesses the potential riskfrom a softening in the housing market. Our assessment isthat while a housing slowdown by itself may slow con-sumer spending some, it is probably insufficient to precipi-tate a downturn without some additional shocks outside ofthe sector.

The current expansion has seen record-high lev-els of transactions in housing, extraordinarygrowth in the aggregate value of owner-occu-pied housing, and large increases in the amountof funds realized from the refinancing of mort-

gage debt. At the same time, the macroeconomic data showthat consumer spending has played a large role in recenteconomic growth. The simultaneity of these developmentshas suggested to many analysts that the strong housing mar-ket and rising home prices have been a major (if possiblyunsustainable) pillar supporting the economy. Moreover,many of these analysts are concerned that a contraction inhousing activity and values could pose a significant risk toconsumer spending, accentuating the direct effect from afall-off in residential construction.1

This paper explores the channels by which the housingmarket may affect consumer spending and assesses the po-tential risk to aggregate activity from a softening housingmarket. The channels we examine include the following:• The direct mechanical relationship between housing

and consumption; i.e., some components of consumer

Housing Activity and Consumer Spending

Housing Activity andConsumer SpendingTHE LINK IS NOT AS STRONG AS OFTEN SUPPOSED.

By Jonathan McCarthy and Charles Steindel

The views expressed in this paper represent those of the authors alone and not necessarily those of the Federal Reserve Bank of New York or the Fed-eral Reserve System. This paper is based on data available in the middle of 2006. The implications of the subsequent contraction in housing are dis-cussed in Appendix 2.

1An extreme version of these concerns is expressed in Baker (2006).

spending are complementary to housing transactions;• The “wealth effect” from capital gains on housing;• The influence of home equity extraction (which has in-

creased greatly, stemming from less costly mortgage refi-nancing and more accessible home equity loans) onhousehold spending behavior.Our assessment is that a housing slowdown by itself may

slow consumer spending and GDP growth some; however, itis probably insufficient to precipitate a downturn withoutsome additional shocks outside of the sector. First, theamount of consumer spending more or less mechanically tiedto housing transactions is fairly limited. Second, while theprecise size of the wealth effect from gains in home prices isproblematic, plausible estimates do not suggest that con-sumer spending growth would cease if there was a declinein home values on the order of recent gains. Third, it appearsthat much of the recent home equity withdrawal has beenused to restructure household balance sheets as well as to fi-nance consumer spending decisions based on fundamentalconsumption determinants (such as expected income growth)rather than on the availability of this source of funds.

The next portion of the paper reviews the basic facts onhousing activity, aggregate home values, and the recent evo-lution of household wealth and spending. We then exam-ine and assess the consumerspending impact of housing ac-tivity through the channelslisted above. The last sectionprovides concluding remarks.

Recent Trends in Housing,Household Wealth, andSpending

The two aspects of housingactivity directly linked to real eco-nomic activity are constructionand sales.Within the national in-come accounts, the data on homeconstruction are contained in theresidential investment compo-nent. This component consists ofall outlays on home construc-tion—renovations and improve-ments as well as new building ofsingle andmultifamily structures.It also includes the value addedin the course of homesales,whichis approximatedby fees earnedbyreal estate brokers.

Residential investment hasgenerally risen briskly during

this expansion. The annual rate of increase in real residentialinvestment from the start of 2002 through 2006Q1 averaged7.9 percent,more than twice that forGDPas awhole. Nonethe-less, this growth pace is about the same as that for the corre-sponding period of the expansion of the 1990s and is wellbelow the gains seen in prior expansions. Figure 1 shows thatresidential investment on average contributed about 0.4 per-centage points to overall GDP growth over 2002Q1 to 2006Q1.This is a somewhat larger growth contribution than in the com-parable period of the 1990s expansion but less than in the eco-nomic expansions of the 1970s and 1980s.

Although the growth rate of residential investment duringthis period was not unusual, the level of housing market ac-tivity was extraordinary. This reflects the fact that the recentexpansion of residential investment occurred without a sig-nificant bust preceding it. Consequently, the share of resi-dential investment exceeded six percent of GDP in 2005 andearly 2006, its highest since early 1950s, as shown in Figure2. Other measures of single family housing activity have setnew records. Single-family housing starts exceeded their pre-vious peak (set in December 1977) several times in the pastfew years, with the January 2006 figure of 1.843 million units(annual rate) the peak. Home sales also set new highs duringthe recent expansion: the peak of new homes sales was 1.37

Business Economics • April 2007 7Housing Activity and Consumer Spending

F I G U R E 1

R E A L G D P G ROW T H A N D C O N T R I B U T I O N S

Percentage Points

Source: Bureau of Economic Analysis

Note: Shading represents NBER recessions. GDP growth and growth contributions calculated on 4-quarter basis.

9.0

6.0

3.0

0.0

-3.0

1980 1984 1988 1992 1996 2000 2004

GDP

PCE and Residential Investment

Residential Investment

million units (annual rate) inJuly 2005 and the peak of ex-isting home sales was 6.33million units in June 2005.The declines in starts andsales since these peaks haveprompted some of the concernabout a slowing housing mar-ket and its possible spilloversto the rest of the U.S. economy.

The other extraordinaryfeature of the housing marketin recent years has been therise in home prices. The four-quarter change in the repeatsales index compiled by theOffice of Federal Housing En-terprise Oversight (OFHEO)exceeded ten percent forabout two years after 2004Q2,the first time this occurredsince the high overall infla-tion of the late 1970s (Figure3). More impressively, realhome prices have increasedsteadily since 1995. Thelength of the real home priceboom and its cumulative sizeis unprecedented in the his-tory of the OFHEO series. Ona longer perspective, the datain Shiller (2006), which startsin 1890 for the United States,indicates that the recent risein real home prices is un-usual.

This rise in real homeprices has sparked consider-able debate about whetherthere is a “bubble”.2An analy-sis of this issue is beyond thescope of this paper, but it issufficient to note that for mostof the issues we address, theanswer to the bubble question

8 Business Economics • April 2007 Housing Activity and Consumer Spending

F I G U R E 3

H OME P R I C E A P P R E C I AT I O N

Percent Change - Year to Year

Source: Office of Federal Housing Enterprise Oversight and Bureau of Economic Analysis.

Note: Shading represents NBER recessions. Numbers in shaded region represent cumulative real appreciationduring that episode.

15

10

5

0

-51978 1980 1984 1988 1992 1996 2000 2004

OFHEO IndexNominal

AppreciationReal

Appreciation

14%(3.8%)

16%(2.3%)

70%(5.1%)

F I G U R E 2

R E S I D E N T I A L I N V E S TM E N T S H A R E O F G D P

Percentage Points

Source: Bureau of Economic Analysis

Note: Shading represents NBER recessions.

7.0

6.0

5.0

4.0

3.0

1947 1957 1967 1977 1987 1997

8.0

2Some papers have argued thathousing market “fundamentals”(particularly the long-term declinein interest rates) can explain the re-cent rise in real home prices; see, forexample, McCarthy and Peach(2004) and Himmelberg, Mayer, and

is not particularly relevant; that is, the source of a homeprice decline—for instance, if it is the result of the burst-ing of a “bubble”—will have only a minor differential im-pact on the effects of that decline on consumer spending,unless the home price decline is associated with a signifi-cant change in households’ longer-term attitudes.

The brisk pace of residential investment and thestrength in housing prices have swelled the value of homesin the household balance sheet. According to the FederalReserve Board’s Flow of Funds data, the aggregate value ofresidential real estate owned by U.S. households was $20.4trillion as of the end of 2006Q1, accounting for about 30percent of all the assets owned by the household sector.

Mortgage indebtedness has grown in tandem with thevalue of housing. This is one factor behind the rise in theleverage in the household sector. Nonetheless, the owners’equity share of residential real estate has been littlechanged in recent years at around 55 percent. The dollarvalue of equity in residential real estate increased by morethan $1 trillion in both 2004 and 2005 and was $11.4 tril-lion at the end of 2006Q1, over 20 percent of householdnet worth.

The strength of home values cushioned the impact ofthe 2000-2002 stockmarket decline, andafterwards—with thestock market recover-ing—the ratio ofhousehold net worthto disposable personalincome had risen toalmost 580 percent atthe end of 2006Q1.This figure was higherthan any earlier ones,except for those seenin 1998-2000 at theheight of stock marketvalues, as shown inFigure 4. Althoughmuch of the volatilityin the wealth-incomeratio reflects stockmarket swings, an im-

portant component of the longer-term rise has been thesurge in housing values. Figure 4 also shows that he ratioof owners’ equity in residential real estate to income grewfrom less than 90 percent to more than 120 percent between2000 and 2006Q1.

Along with strength in housing activity and values, thisexpansion has seen sustained momentum in consumerspending; however, it has not been the case that consumerspending has been unusually strong. Real consumer spend-ing contributed an average of about 2¼ percentage pointsto overall GDP growth in 2002-2006Q1. This contributionwas slightly smaller than in the comparable period of theexpansion of the 1990s and considerably smaller than thatin the 1980s. However, the recent strength of consumptionhas occurred in an environment of very low and decliningpersonal saving. Indeed, the personal saving rate fell belowzero in 2005 and has since remained negative. The combi-nation of well-maintained consumer spending, low personalsaving, and remarkable gains in home values (along withhigh levels of residential investment) has naturally raiseda belief that the latter has contributed to the two formerphenomena. A consequence of that belief and these obser-vations is a concern of some analysts that a decline in hous-


F I G U R E 4

N E T WOR T H R E L AT I V E T O D I S P O S A B L E P E R S O N A L I N C OME

Ratio

7

6

4

1995 2000 2005

5

4

3

1

2

0

199019801980

1996 2000 2004199219881980 1984

Stock and Real Estate

Real Estate

Total

Source: Flow of Funds, Federal Reserve Board, and Bureau of Economic Analysis


Sinai (2005). Of course, anumber of other analystshave claimed that only non-fundamental factors couldexplain the extent of therise. For recent examples,see Baker (2006) andShiller (2006).

ing activity and home values will lead to a substantial de-cline in consumer spending and increase the risks to themacroeconomic environment.

Housing and Consumer Spending: DirectLinkages

A portion of consumer spending appears complemen-tary to housing activity. This portion includes expendituresdirectly tied to transactions in housing markets, such asmoving expenses, and spending often associated with homepurchase, such as expenditures on furniture, home fur-nishings, and major appliances. We examine the relation-ship between these expenditures and home sales to assessthe implications of housing market fluctuations on thishousing-related spending.

We use the detailed data on consumer expenditures inthe national income accounts to compute the dollar and realvolume of these aggregated expenditures.3 To assess the re-lationship between this aggregate and housing activity, weuse a Granger causality test of the growth rate of these ex-penditures with the growth rate of new and existing homesales.4 The results of these tests are in Table 1; they indi-cate that home sales Granger-cause these expenditures andvice versa (upper panel). In contrast, home sales do notGranger-cause overall consumer spending nor vice versa(lower panel). One can interpret (even though any such con-clusion based on this reduced form is tentative at best) thedual causality for housing-related expenditures and home

sales as suggesting that a third factor contributes to the re-lationship. In any event, these results are consistent withour intuition.

Although the linkage between these expenditures andhome sales is reasonably firm, the potential spillover froma slowing housing market probably would not be substan-tial because these purchases are a relatively small share ofconsumer spending. In 2005, aggregate outlays on theseproducts were slightly under $200 billion, or less than 2½percent of total consumption. These expenditures haverisen robustly in real terms in recent years (5 to 10 per-cent), but their small share implies that they have had asmall contribution to real GDP growth, typically on theorder of 0.1 percentage point per quarter. As such, a re-versal in the path in these expenditures probably would re-duce real GDP growth by a small amount—maybe ¼percentage point at most. Still, this drag would reinforcethe impact of a contraction in residential investment.

The Housing Wealth EffectMuch of the emphasis on the effects of housing on eco-

nomic activity relates to the possible connections betweenhome price appreciation and consumer spending. Two in-terrelated channels have been articulated. The first is astraightforward connection between housing wealth andspending. The second is a connection between borrowingagainst home equity and spending. In this section, we con-sider the wealth channel.

A Simple Theoretical FrameworkThe standard economic framework to analyze con-

sumption has been the life-cycle/permanent-incomemodel.5 In the model, households make their spending de-cisions taking into account their expected lifetime re-sources, which include current and expected future laborincome (“human wealth”) as well as current net worth(“non-human wealth”). An implication of the model is thatwhen the expected value of their resources rises, house-holds will increase their consumption, even when currentincome has not risen. In particular, increases in net worth,including that from rising home values, will lead to higherspending.

Although the basic logic that changes in wealth mayaffect spending is fairly straightforward, the size of that ef-fect is uncertain. In this simple model, an “exogenous” in-


TA B L E 1

G R A N G E R C A U S A L I T Y T E S T S O F C O N S UME R

E X P E N D I T U R E S A N D H OME S A L E S

(a) “Housing related” personal consumption expenditures (PCE)

Home sales series Sales cause PCE PCE cause sales

New 0.003 0.014

Existing 0.011 0.049

(b) Total personal consumption expenditures (PCE)

Home sales series Sales cause PCE PCE cause sales

New 0.614 0.152

Existing 0.714 0.350

Notes: Each cell of the table contains the p-value of the test statistic for the nullhypothesis that sales does not Granger-cause PCE (conversely, PCE does notGranger-cause sales). The estimation period of the regressions is 1990Q1 – 2006Q1,and the number of lags in the regression is four.

3In particular, we include expenditures on furniture, major appliances,floor coverings, other durable house furnishings (window treatments,etc.), and moving and storage expenses.4Granger causality means that lagged movements in home sales pre-cede movements in these expenditures that are not explained by theirown lagged movements.

5Friedman (1957) introduced the permanent income model while Andoand Modigliani (1963) is an early example of the life cycle model. Al-though the models differ in some aspects, they share a basic premisethat households make their consumption decisions by maximizing util-ity subject to an intertemporal budget constraint. As such, the modelshave been used interchangeably at least since Hall (1978).

crease in current wealth (whether in housing or corporateequities) would lead households to raise their consumptionby roughly the annuity value of the wealth increase. This inturn is related to life expectancy, expected interest rates,expected future income, and preferences for current spend-ing versus future outlays. A back-of-the-envelope calcula-tion would suggest that the propensity to consume fromwealth gains would be somewhere between 0.025 and 0.05,with the higher number being consistent with householdsplanning to spend the bulk of their resources over their re-maining lifetime.6 Given recent gains in home values, suchpropensities to consume would lead to sizable effects onconsumption. Capital gains on household real estate assetsamounted to over $2.1 trillion in 2005 according to the Fed-eral Reserve Board’s Flow of Funds: a propensity to con-sume between 0.025 and 0.05 would imply thatconsumption in 2005 was about $53-$107 billion—about0.6-1.2 percent of total consumption expenditures—higherthan if there were no increases in real estate values. A re-versal of some of these gains then would imply notablylower consumption growth.

However, casual observation of wealth fluctuations andconsumption suggests that households do not adjust con-sumption immediately in response to wealth fluctuations tothe extent implied by the simple model. For example, itdoes not appear that consumption experienced a boom dur-ing the 1990s commensurate with the stock market boom,nor did spending crash as the stock market declined pre-cipitously in 2000-02 (despite the fears of such at the time).There are at least two different ways to interpret such ob-servations.

One possibility is that households adjust their con-sumption profile slowly, perhaps because of intermittentplanning periods, the costs of determining the new optimalconsumption profile, or other behavioral considerations. Insuch a case, the immediate impact of a change in wealthmay be small but the ultimate effect would be as predictedby the model.7 This scenario then would imply that includ-

ing lagged values of wealth or an error correction mecha-nism in estimated consumption functions would be neces-sary to determine the size of the ultimate wealth effect,which traditionally has been done in many macroecono-metric models.

Another possibility is that changes in wealth have dif-ferential effects on spending depending upon their ultimatesource. A statistical examination of the aggregate data sug-gests that the propensity to consume from wealth is not sta-ble (Ludvigson and Steindel, 1999; Steindel, forthcoming).For instance, it appears to have been low during the stockmarket boom and in the period thereafter. A plausible ex-planation for this instability emerges from a closer look atthe basic model of spending. Essentially, the life-cycle/per-manent-income model assumes that households take intoaccount the likely profile of resources over their planninghorizon. If they perceive a current wealth gain as not “sus-tainable” (in a sense made more precise below), then theywill not increase consumption as much as they would if theyperceived the wealth gain as “permanent” or sustainable.

The framework of Lettau and Ludvigson (2001) pro-vides a more technical explanation for these statistical re-sults. Their analysis shows that consumption, asset wealth,and labor income are cointegrated, or share a common long-run trend, which is consistent with the premises of thehousehold lifetime resource constraint. In their model,shocks to the variables of the system are of two types: “per-manent” shocks affecting the common trend of these vari-ables (these largely reflect changes in the expected growthpath of the economy; in part, these can be thought of as ex-pected changes in productivity growth) and “temporary”shocks having no effect on the common trend. Lettau andLudvigson find that consumption fluctuations are relatedmostly to permanent shocks. In contrast, wealth fluctua-tions largely are associated with temporary shocks. Conse-quently, only wealth fluctuations related to permanentshocks have any substantive effect on consumption. Wealtheffects then depend upon the source of wealth fluctuations:if permanent, then the effect is about the rule of thumb wediscussed above; if temporary (like most wealth fluctua-tions), then the effect is very small. Ludvigson (2003) doc-uments how this decomposition accounts for the lowpropensity to consume from wealth during and after thestock market boom. Essentially, consumers appeared tohave regarded only about half the run-up in wealth from1995 to 2000 as “permanent.”

Beyond these considerations about the size and speedof wealth effects, there also is considerable evidence thatsome households may not follow the prescriptions of the


6An estimate of 0.05 would appear consistent with a seemingly low lifeexpectancy of 20 years. However, the bulk of wealth is owned by olderhouseholds. Moreover, interest earned from wealth enables householdsto spend a bit more than life expectancy considerations alone wouldallow.

7Analysis consistent with this view is presented in Davis and Palumbo(2001).

Households do not adjustconsumption immediately inresponse to wealth fluctuations.

standard life-cycle/permanent income model, potentiallyimpacting estimated wealth effects. One reason for this be-havior is that some households probably face liquidity con-straints; i.e., they cannot finance spending they wouldotherwise be willing and able to do. This observation un-derlies the “buffer stock” models of consumption that havebeen used in many recent studies.8 In such models, thepropensity to consume out of current resources (includingasset wealth) is higher for liquidity-constrained households;as such, increases in the wealth of those households, nomatter what the cause, would raise consumption more thanit would for unconstrained households and conceivablyimply a larger aggregate effect than suggested by theory.

Another reason that some households may not followthe prescriptions of the standard life-cycle/permanent-income model is that they determine spending accordingto a simple rule of thumb, such as consuming a fixed por-tion of current income and a fixed portion of wealth gainsregardless of the implications of those changes for lifetimeresources (Campbell and Mankiw, 1989). Such behaviormay reflect limited ability or resources to plan the con-sumption path according to the dictates of the standardmodel. Because the rule of thumb for spending out of wealthincreases may be higher than the propensity to consumefrom wealth implied by the standard model, wealth gainsmay have a larger effect on aggregate consumption de-pending upon the proportion of such households.

Why Housing Could Be DifferentThe above discussion tacitly assumes that all changes in

household wealth have comparable effects on spending. Ifhousing is inherently similar to other forms of wealth, thenwe could appeal to the general wealth effects literature todetermine its effect on consumer spending. Implicitly, thoseanalysts that have given housing a more prominent role insustaining consumption in recent years assume that it is dif-ferent from other forms of wealth in regard to its effects onaggregate consumption. Moreover, many of these analystsalso appear to assume that developments in the housing andmortgage markets have led to the effect of housing wealth onconsumption becoming greater in recent years. Some possi-ble reasons for these views are the following.

Housing is more widely and evenly distributed. Accord-ing to survey data, including those collected during the late1990s stock market boom, for most American householdstheir home is the most important asset they own (Tracy,Shroeder, and Chan, 1999). This pattern continues in morerecent data. The 2004 Survey of Consumer Finances foundthat 69.1 percent of families owned their primary residence;

for the bottom income quintile, the percentage was 41.6percent (Bucks, Kennickell, and Moore, 2006). In contrast,48.6 percent of families held corporate equity either di-rectly or indirectly; for the bottom income quintile, this per-centage was 11.7 percent.

Clearly, the ubiquity of home ownership means thatfluctuations in home values could affect the net worth of alarger portion of the population than fluctuations in otherasset prices. Consequently, assuming that home price fluc-tuations do affect the spending decisions of homeowners,changes in aggregate home values will influence spendingacross a wide swath of the nation. However, widespread ex-posure to home price changes does not necessarily implythat the aggregate dollar or real volume of spending influ-enced by these changes will be necessarily larger than forequivalent changes in the value of other assets that are lessevenly held. Although the marginal propensity to consumeout of wealth in the standard model can in principle varyaccording to the level of wealth and income (dependingupon the exact parameterization), in most standard imple-mentations of the model it is not very sensitive to wealthlevels (i.e., consumption is approximately linear in wealth).

A differential effect of housing wealth fluctuations onconsumption would then appear to depend upon the extentof the previously described complications to the standardmodel. In particular, because housing wealth is more widelyheld, its fluctuations may affect more liquidity-constrainedhouseholds than fluctuations in other assets. If so, then thehousing wealth effect may differ from that of other assets.

In assessing the impact of housing on consumption inthe current situation, there is some basis to believe that thisavenue for a differential impact from housing may have alesser influence than suggested by the data on the holdingdistribution of the assets. In particular, recent home priceappreciation appears to have been concentrated in high-value homes in high-income areas. We illustrate this in Fig-ure 5, which plots the logarithm of state per capita personalincome in 2000 against state home price appreciation be-tween 2001Q1 and 2006Q1 according to the OFHEO index


8A couple of prominent early examples are Deaton (1991) and Carroll(1997).

Ubiquity of home ownershipmeans that fluctuations in homevalues could affect the net worth ofa larger portion of the populationthan fluctuations in other assetprices.

Business Economics • April 2007 13

It shows that most of the high-appreciation states also havehigher income: the included regression line is positively(and statistically significant) sloped.9 Therefore, the recentincrease in housing wealth has been unequally distributed,with the previously well-to-do getting a disproportionateportion.10 Correspondingly, if a correction occurs in the seg-ments of the market that have seen the greatest increases,these same individuals would experience setbacks. Anotherway to think of this is that the recent run-up in housing isto some extent comparable in its distribution to that of typ-ical stock market expansions.

Housing is primary collateral for household loans. Thesimple life-cycle/permanent income model implicitly as-sumes that households are able to borrow at the risk-freerate without having to post collateral. In reality, much ofhousehold debt is collateralized with real estate. Accordingto the Flow of Funds, over 72 percent of total household li-abilities at the end of 2005 were home mortgages (includ-ing home equity loans). Moreover, Figure 6 shows that thehome mortgage share rose rapidly over 2001-05 as morehouseholds have become homeowners and existing home-

owners extracted equityfrom their homes.

The fact that much ofhousehold borrowing isbacked by real estate col-lateral suggests that “fi-nancial accelerator” effectsmay influence the relation-ship between home valuesand consumption.11 Al-though much of the earlyliterature on the financialaccelerator had concen-trated on its possible ef-fects on firms, some recentpapers have explored howitmay affect household be-havior and consumption;see, for example, Aoki,Proudman, and Vlieghe(2004) and Iacoviello(2004, 2005).

In these models, theborrowing capacity ofagents is a function of their“collateralizable” assets.

For households, the empirical data suggest that this is largelythe value of their homes. Consequently, fluctuations in homevalues would affect the borrowing capacity of households andthe availability of loans to them. This in turn affects the con-sumption of households that are constrained; i.e., if unse-cured loans at the risk-free rate were available, they wouldprefer to borrow more to consume. If a significant portion ofspending was done by households whose access to credit canfluctuate, such shifts in credit access then could affect ag-gregate consumption.

In the environment of rising home prices during thefirst half of this decade, households may have gained froma “virtuous circle” where the increase in prices raised loanavailability, which increased the consumption of previouslycredit-constrained households, which in turn supportedcontinued economic strength and further increases in homeprices. However, this effect is dampened as fewer house-holds remain credit-constrained as home values and bor-rowing capacity increases. In contrast, if prices begin tofall, the virtuous circle becomes a vicious circle: prices fall,loan availability declines, consumption declines weakenthe economy, and home prices fall further in response. Inaddition, the vicious circle is amplified over time as more


11See Bernanke, Gertler, and Gilchrist (1999) for more on this mecha-nism within a small general equilibrium model.

F I G U R E 5

S TAT E I N C OME V S . 5 - Y R H OME P R I C E A P P R E C I AT I O N

Percent

140

100

40

0.0

0

9.9 10 10.1 10.2 10.3 10.4 10.5

60

80

120

10.6 10.7

Logarithm of State per Capita Income: 2000

Hom

eP

rice

App

reci

atio

n:2001

Q1-2

006

Q1

Source: Office of Federal Housing Enterprise Oversight and Bureau of Economic Analysis

9The slope of the line is 98.1 with a standard error of 23.2, for a t-sta-tistic of 4.24. The R2 of the regression is 0.27.10Also see Himmelberg, Meyer, and Sinai (2005) for further evidenceon this point.

14 Business Economics • April 2007

households become constrained as loan availability de-clines. Beyond that, sales by lenders of repossessed collat-eral (foreclosed homes) could further depress the market,which would further worsen the cycle.

Iacoviello (2004) presents evidence showing that thebehavior of aggregate consumption in the United States isconsistent with a version of the financial accelerator model.He estimates that the fraction of consumption by con-strained households is between 0.18 and 0.26 (dependingupon the instruments used in the estimation).12 If a simi-lar fraction of current consumption is by constrained house-holds—the developments since 2003 suggest that it issomewhat smaller—this indicates that a substantial frac-tion of consumption may be at risk because of financial ac-celerator effects from a decline in home values. Moreover,if there has been a sizable increase over recent years in thefraction of homeowners that are extremely leveraged (assuggested by anecdotal reports of 100-125 percent LTVmortgages), then the risk could be greater.

However, on a more sanguine note, the aggregate U.S.household sector retains quite substantial home equitycushions, as noted above. So even if there was a ten percentreduction in the value of residential real estate, it would re-

sult in the owners’ equityshare declining from 55 to50 percent, a level that stillappears to be substantial.Thus, while credit may be-come more restricted tosome in case of a major cor-rection in home values, inthe aggregate it appears thatmost U.S. households wouldnot necessarily move to afundamentally weaker fi-nancial condition and jeop-ardize their access to credit.

Housing has becomemore liquid. The develop-ments in the mortgage mar-ket since the early 1980shave reduced the costs ofborrowing against and in-creased the liquidity ofhome equity.13 Moreover,households increasinglyhave taken advantage ofsuch opportunities, as wediscuss in more detail below.

As such, housing has become more liquid over this period.Traditionally, one motivation for households to own liquid as-sets has been the need for precautionary savings to self-in-sure against hard times. With home equity becoming moreliquid, householdsmay have reduced previously accumulatedprecautionary savings reserves.14 The unfolding of this processcould have boosted the growth of consumer spending.

Presumably, such a transitional process would dissipate,and consumption growth would return to a “normal” paceonce households realign their portfolios. However, in an en-


F I G U R E 6

H OME MOR T G AG E S H A R E O F H O U S E H O L D L I A B I L I T I E S

Percent

70

65

60

55

75

1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 2002

Source: Flow of Funds, Federal Reserve Board


12The estimation period of his regressions is 1986:1-2002:4.

13McCarthy and Peach (2002) discuss the evolution of the mortgagemarket during the post-war period from the savings and loan-dominated“New Deal” system to the current capital market based system. To il-lustrate the decline in mortgage costs, according to the Freddie MacPrimary Mortgage Market Survey, fees and points on a 30-year fixedrate mortgage have declined from about 2.5 percent in 1985 to 0.5 per-cent in 2006. For a 15-year fixed rate mortgage (a more popular refi-nancing instrument), fees and points have declined from about 1.7percent in 1992 (the data for this begin in late 1991) to 0.5 percent in2006.14McCarthy (1995) noted that in the PSID data he examined from themid-1980s (when refinancing and home equity loans were less preva-lent) households appeared to regard home equity as illiquid and notavailable for use as precautionary savings. More recently, Hurst andStafford (2004) found that, using PSID data from the early 1990s, liq-uidity-constrained households were more likely to refinance when fac-ing an unemployment shock, which is consistent with the use of homeequity for precautionary reasons.

vironment of rising home prices and the use of home equityas precautionary assets, households may find that their pre-cautionary savings are above desired levels (particularly for“buffer stock” households). In such a case, the rise in homevalues may induce more consumption than there would beotherwise. Conversely, a decline in home prices may dropprecautionary savings below desired thresholds. This in turnmay lead households to cut spending to accumulate pre-cautionary savings balances in more traditional forms, low-ering consumption more than itwould otherwise.

As we stated above, the transi-tion of the mortgage market and thegreater liquidity of housing has beenan ongoing process for many years,suggesting that the transitional im-pact should be dissipated. Then forthis channel to have had a greaterimpact on recent spending growthone would need to argue that therehas been a recent particular increasein the ability of households to liqui-date home equity (as we note below,there has been a surge in the amountof equity liquidated, but that couldhave occurred because of favorableconditions rather than a radicalchange in access to this means of fi-nance). Beyond this, in judging therisks stemming from a potential weakening in housing, ourprevious comments about the size of home equity (which, atcurrent values, is larger than either disposable income orconsumer spending) apply: even in the wake of a large cor-rection in home values, there would still be very largeamounts of home equity available to tap as a source of pre-cautionary savings. Therefore, for a decline in home prices tohave much impact through this channel, households’ demandfor precautionary savings would have to take a form such thatthey would want to retain a large amount of home equity inthe aggregate.

Housing wealth fluctuations are more “permanent” thanother wealth fluctuations. Stock market fluctuations have tra-ditionally accounted for the bulk of the short-term changes inU.S. household net worth (Ludvigson and Steindel, 1999). Assuch, the aggregate evidence on the effects of wealth changeson consumer spending may essentially reflect the effect ofstock market fluctuations. In particular, the sharp distinctiondrawn between the effects of “permanent” and “transitory”changes in wealth that we discussed previously may onlyapply for movements in the stock market. Home price move-ments have typically been less volatile. Consumers then may

inherently regard any change in home values as having ahigher “permanent” component than an equivalent change inthe stockmarket. Therefore, the recent increases in home val-ues may have had a stronger impact on consumer spendingthan a comparable gain in stockmarket values, and the down-ward risk to consumption of a correction in housing pricesmay be larger.

Although this argument has intuitive appeal, it is diffi-cult to assess rigorously. In the aggregate data, it is prob-

lematic to differentiate “permanent” and “transitory”movements in different assets. Fluctuations in the relativevalues and portfolio allocations of different assets can re-flect changes in the choices of investors stemming from ahost of reasons, of which their linkage to fundamental long-run income and spending (the “permanent” component) isonly one.15 Therefore, even if households have regardedprevious fluctuations in home values as more “permanent,”it is certainly within the realm of possibility that, becauseof the extraordinary recent strength of home values, house-holds regard them as less “permanent” than the smallergains of past upturns and have adjusted their consumptionless in response.16


15Lustig and van Nieuwerburgh (2005) discuss some of the intercon-nections between consumer behavior, housing values, and the stockmarket.16We would like to emphasize that categorizing an asset value increaseas “transitory” does not suggest that there is something irrational aboutits current price; rather it is a statement that it is not associated with cur-rent movements in consumer spending. Lettau and Ludvigson (2001)find that household assessments that movements of aggregate wealthare transitory helps to forecast stock market fluctuations.

TA B L E 2

R E C E N T E S T I M AT E S O F H O U S I N G WE A LT H E F F E C T S

MPC from MPC relativeStudy Data home wealth to other wealthIacoviello (2004) Aggregate: 1986-2002 0.07 Higher

Belskey & Prakken (2004) Aggregate: 1960-2003 0.056 (long-run) Same0.045 (1-year) Higher

Benjamin, Chinloy, & Jud (2004) Aggregate: 1952-2001 .08-.16 Higher

Case, Quigley, & Shiller (2005) State level: 1982-1999 0.02-0.04 HigherLehnert (2004) Panel Survey of Income 0.025 N.A.

Dynamics (PSID)

Bostic, Gabriel, & Consumer Expenditure Survey, ≤0.025, lower in HigherPainter (2005) Survey of Consumer Finances volatile periods

Juster, Lupton, Smith, &Stafford (2006) PSID 0.03, but not Lower

statistically sig.

Estimates of the Propensity to Consume from HousingWealth

Table 2 summarizes the results from some recent studiesthat have estimated the wealth effect from housing. Here, wediscuss some general points that we derive from those results.

For the reasons just noted, aggregate evidence bearingon the propensity to consume from housing wealth (as dif-ferentiated from aggregate wealth) is problematic. Neverthe-less, a number of recent studies have examined aggregateconsumption sensitivity to changes in housing wealth (as op-posed to overall wealth): these include Belsky and Prakken(2004); Benjamin, Chinloy and Jud (2004); and Iacoviello(2004). Belsky and Prakken (2004) impose similar long-runconsumer responses to changes in all types of wealth butallow short-run responses to differ between housing and non-housing wealth. Iacoviello (2004) concentrates on the finan-cial accelerator effects of changes in housing wealth and sodoes not explicitly include changes in other forms of wealthin his model. Benjamin, Chinloy, and Jud (2004), in contrast,explicitly allow for differing propensities to consume from fi-nancial wealth and housing wealth.

In general, the results from these studies indicate ag-gregate consumer spending has been more sensitive tochanges in home values, at least in the short run, than tochanges in the values of other assets. Belsky and Prakkenestimate that the propensity to consume from changes inhousing wealth is over 0.05in the long run, and above0.04 within one year of achange; the short-run pro-pensity is well above that forcorporate equity. The esti-mates in Iacoviello are con-sistent with a propensity toconsume from changes inhousing prices of 0.07,which is above the typicalestimates of propensities toconsume from aggregatewealth. Benjamin, Chinloyand Jud find even strongereffects of housing wealth onconsumption; their estimatesof the propensity to consumefrom changes in home equityare between 0.08 and 0.16,which is well above their es-timate of the propensity toconsume from financial as-sets (about 0.025).

Taken at face value,

these results suggest that the recent increase in home valueshas played a major role in the expansion of consumption andthat a fall-off in home values would have significant nega-tive consequences for spending. However, these studies as-sume that wealth effects are essentially stable over time,regardless of the forces driving wealth values. On the whole,it is not surprising to find that on average, consumption hasbeen more responsive to the typical change in home valuesthan to the typical change in other forms of wealth. As wementioned previously, one would normally assume that ahigher portion of housing value changes was “permanent” inthe sense discussed above (Belsky and Prakken note thispoint). It is not altogether clear that one should apply theseaverage, or typical, responses to compute the spending ef-fects of the recent increase or of a possible correction.17 Infact, in Appendix 1, we provide evidence that the propensityto consume out of housing wealth does not differ drasticallyfrom that out of non-housing wealth. Moreover, the propen-sity to consume out of wealth depends upon household viewsof the “permanence” of wealth changes and thus is not sta-ble over time.


F I G U R E 7

MOR T G AG E R E F I N A N C E O R I G I N AT I O N S

Percent

3000

1500

1000

0

3500

1990 1993 1996 1999 2002 2005

2500

2000

500

Source: Mortgage Bankers Association


17Also, Iacoviello (2004) notes that his results are only applicable to an-ticipated changes in home prices. If one believes that the recent in-creases have been largely unanticipated, his results are not applicable.

Given the ambiguities fromthe aggregate data, examiningthe microeconomic data couldprovide some clarification. Assuch, some recent studies haveturned to cross-sectional evi-dence relating observations ofthe level of spending by house-holds to observations of theirincome and net worth, separat-ing housing from other compo-nents of the balance sheet.These studies uniformly sug-gest that the propensity to con-sume from changes in homevalues is less than 0.05. Onestudy—Bostic, Gabriel, andPainter (2005)—found that theconsumption response to homeprice movements was smallerthan normal following periodsof exceptional volatility inhome prices. One interpreta-tion of this result is that fluctu-ations in home values duringvolatile periods are considered “transitory,” leading to asmaller consumption response, an interpretation with im-plications for the current situation. Comparing housingwealth to other assets, these studies reported mixed resultsas to whether the consumer response to home price move-ments is larger than for other forms of wealth.

On balance, we see the results of these recent studiesas consistent with the notion that the propensity to consumefrom permanent changes in housing values is between0.025 and 0.05, which is consistent with the rule of thumbfor the propensity to consume out of general wealth. Thereis some probability that the propensity may have beensomewhat higher on average over the postwar period. Onthe other hand, there also is evidence that the consumptionresponse has been weaker in times of high volatility inhousing prices. With housing capital gains averaging about$2 trillion over 2004-05, and applying our interpretation ofthe central estimates of the propensity to consume out ofhousing wealth, these findings suggest that increases inhome values could have added roughly $50 to $100 billionto the level of consumer spending in 2004 and 2005.

The Refinancing ChannelAlong with the large increases in home values in re-

cent years has come an unprecedented wave of mortgagerefinancings, encouraged by both home price increases and

low interest rates. Figure 7 shows that while refinancingshave ebbed from their extraordinary pace of 2003, the grossvalue of mortgage refinance originations was still nearly$1.3 trillion in 2005. While the bulk of these funds proba-bly were used to liquidate existing prior mortgages, theretypically are considerable funds remaining for other pur-poses. One measure of such funds is based on work byGreenspan and Kennedy (2005), which includes explicitequity withdrawal through refinancings along with implicitwithdrawals through mortgage payments.18 Based on theirmethodology, Figure 8 shows that mortgage equity with-drawal by households was between $700 and $900 billionin 2005.19 In principle, these funds are available to financeconsumer spending.

As suggested above, we view the increase in refinanc-ings as connected to the secular trend of increased avail-ability of credit to U.S. households, with the favorabledevelopments in home prices and interest rates fueling therecent upsurge. It could well be the case that this increasedavailability of credit has provided some support to con-


F I G U R E 8

H OME E Q U I T Y E X T R AC T I O N

Billions of Dollars—Annual Revenue

1100

500

300

-100

1300

1990 1993 1996 1999 2002 2005

900

700

100

19961993

Gross

Net

Source: Federal Reserve Board Note: Shading represents NBER recessions.

18Specifically, they compute the amount of borrowing to finance homepurchases at average loan-to-value ratios and then take the differencebetween this amount and the change in mortgage debt outstanding todetermine gross equity extraction.19The smaller net number removes mortgage transactions fees, points,and taxes from gross mortgage equity withdrawal.

sumption growth in recent years, though we do not thinkthe dollar amount of equity extraction is a valid measure ofavailability.

A vexatious issue is the assessment of the spending im-plications of the funds extracted from home equity. At oneextreme, equity extractions can be viewed as financial de-cisions entirely independent of basic spending and savingdecisions, a view consistent with the standard life-cycle/permanent-income model. In the aggregate, equityextraction redirects interest expenses and income acrosshouseholds (McConnell, Peach, and al-Hashimi, 2003);consequently, the net effect on consumption would be quitesmall. At the other extreme, the funds raised from home eq-uity extraction could be viewed as equivalent to straighthousehold income and thus have a major effect on spend-ing. Under this latter view, if home equity extractions werereduced because of higher interest rates and/or a correc-tion in home prices, spending growth would then be at risk.

The available empirical results are subject to thesevarying interpretations. A 2002 study of the use of fundsfrom mortgage refinancings by Canner, Dynan, and Pass-more (2002) found that perhaps $18 billion in spending in2001—16 percent of the dollar volume of home equity ex-tracted through refinancings—was financed through thismeans (the balance of the equity extracted through refi-nancings was used for home repairs, alterations, and im-provements, as well as to restructure other components ofhousehold balance sheets). The gross dollar volume of re-financings in 2005 was comparable to that in 2001, so oneview is that equity extraction recently has been a relativelyminor contributor to consumer spending. More extremewould be the view that most of this additional spending stillwould have occurred even without the refinancings, whichwould imply that the refinancing has made little contribu-tion. At another extreme, if we apply the 16 percent ratio tothe 2005 estimate of home equity extraction, then between$110 billion and $150 billion of consumer spending wasfinanced through this means that year.20

However, these larger values probably would not be ac-curate estimates of the amount of spending at risk from aweakening in the housing market. First, it is not clear if the2001 ratio would be applicable for 2005. Secondly, it wouldbe rather extreme to project the outright disappearance ofthe refinance market in any circumstance.21 More substan-tively, the increased ease and lower cost of mortgage refi-nancings, along with the tax-exempt status of mortgage

interest payments, suggest that refinancings likely have dis-placed many other means of raising funds to finance spend-ing. Equity extraction—particularly through refinancings—rather than an independent source of spending, may es-sentially be a primary mechanism through which spendingarising from capital gains on homes is financed.

Adding up the RisksResidential investment consistently contributed nearly

one-half percentage point to overall GDP growth over thepast few years. A removal of this contribution and a turn tocontraction would have a noticeable impact on aggregategrowth, unless activity accelerated in some other sector.Nonetheless, housing weakness would need to spread toother sectors to have the substantial effect on aggregate ac-tivity that is incorporated in outlooks where housing is thecurrent economic linchpin. Clearly, the sector of concern inthat regard would be consumer spending.

Our read of the evidence on the linkages is that thereis some risk that a period of sustained weakness in housingactivity and values would have a noticeable impact on thegrowth of consumer spending. One linkage is direct: certaintypes of spending have clear ties to home sales and pur-chases. These spending categories have grown rapidly inrecent years, and a contraction in housing could quite likelylead to some slowdown in their growth. Nonetheless, theaggregate amount of spending in these areas, and their con-tribution to overall GDP growth, has been quite limited.

The more substantive risk arises from the possibilitythat slower growth or declines in housing prices would leadto reduced growth in consumer spending. The primarymechanism by which this could happen is a traditionalwealth effect: the view that there is a substantive connec-tion between changes in the value of household wealth, in-cluding those generated by changes in housing values, andchanges in consumer spending. Another possible mecha-nism is the equity extraction effect—the view that high val-ues of home equity facilitate household borrowing andspending. This mechanism could be one avenue through


20The lower figure refers to the “net” mortgage equity withdrawal esti-mate; the higher figure refers to the “gross” value.21As noted above, Hurst and Stafford (2004) found that some house-holds would refinance after an unemployment shock even if the refi-nancing were not financially advantageous.

The more substantive risk arisesfrom the possibility that slowergrowth or declines in housingprices would lead to reducedgrowth in consumer spending.

which the wealth effect operates, or it could be a separateeffect enabling households to consume at levels not previ-ously achievable. The risk then is that the effect of a slow-down in the growth of home prices would be two-fold:added to the direct wealth effect would be a further mech-anism by which households no longer able to borrow againsthome equity would cut back spending.

These risks look real, but their size is quite problem-atic. Our review of the evidence is that 0.025 to 0.05 maybe a plausible range of the propensity to consume from per-manent changes in home values. Recent years have seencapital gains on homes averaging close to $2 trillion a year:assuming that the propensity to consume is in the above-cited range, roughly $50 to $100 billion a year was addedto consumer spending through housing wealth effects (someof that spending has likely been in the form of expendituresdirectly tied to housing transactions). It could be that thisspending effect estimate is high, since it is possible thathouseholds regard much of the recent unusually strong in-crease in home values as transitory. Nevertheless, takingthese estimates at face value, it would be the case that acessation of this support would have a substantive impacton the growth of consumer spending.

In addition, we find no striking evidence that therehave been large recent changes in households’ ability touse home equity as a means to overcome borrowing con-straints or as a potential source of precautionary savings. Itis likely that the huge amount of funds realized through eq-uity extractions in recent years reflects—for the mostpart—the exploitation of opportunities to recalibrate house-hold balance sheets, rather than a sign of a behavioral shiftin spending associated with a changed financial environ-ment. In any event, even with the large volumes of equityextraction, the aggregate ratio of equity to home values hasremained well above 50 percent. However, because there islittle evidence bearing on the recent disposition of the largeamount of funds raised through equity extraction, it is pos-sible that a good deal of spending could have occurred dueto the availability of this means of finance.

Of course, there is always the potential for an evenlarger slowdown of consumer spending growth if householdwealth (including home values) falters. Our analysis sug-gests that the significant risks would stem from changes inunderlying perceptions of the longer-term growth prospectsof the economy, which have the potential to hamper bothspending and valuations. Fluctuations in asset values stem-ming from developments that are isolated in the housingmarket appear to have less potential to influence spending,unless they are so large as to imply major changes in theavailability of home equity as a means of finance

Appendix 1. Stability of Consumption EquationsOur interpretation of the evidence concerning the effects

of housing wealth on consumer spending is predicated in parton two premises. First, the propensity to consume out ofhousing wealth does not differ drastically from that out ofnon-housing wealth. Second, the propensity to consume outof wealth depends upon household views of the “perma-nence” of wealth changes and thus is not stable over time.

We examine these premises briefly within an Eulerequation model similar to that used in Campbell andMankiw (1989), augmented to included growth in housingand non-housing wealth. Specifically, we estimate the fol-lowing regression.

(A1) �ct = �0 + �1rt + �2�yt-1 + �3�wnt-1 + �4�wht-1 + �t

In this regression, �ct is the growth rate of real percapita consumption (defined as the chained aggregate ofnondurable and services consumption), r is a real three-month T-bill rate, �y is the growth rate of real per capitalabor income (defined similarly as in Lettau and Ludvig-son, 2001), and �wn and �wh are the growth rates of realper capita non-housing and housing wealth respectively.

Once we estimate the model, we examine our premisesusing two statistical tests. First, we test whether the coeffi-cients on non-housing and housing wealth (α3 and α4, re-


A P P E N D I X TA B L E A 1

E S T I M AT E S O F E Q U AT I O N ( A 1 )

Variable (1) (2)Constant 1.803*** 1.824***

(0.229) (0.225)Interest rate -0.088 -0.090

(0.065) (0.065)Income growth 0.132** 0.138***

(0.053) (0.048)Nonhousing wealth growth 0.048***

(0.011)Housing wealth growth 0.023

(0.022)Total wealth growth 0.060***

(0.013)R-bar squared 0.169 0.171Stability test statistic 2.656*** 2.336***Equality of non-housing and housing coefficients:Chi-squared (1) statistic: 0.871(p-value) (0.351)

Notes: The sample period is 1960Q1-2006Q1. Standard errors are shown in parenthesesbelow the parameter estimate. *** denotes statistical significance at the one percentlevel; ** denotes statistical significance at the five percent level. The distribution for thestability test statistic comes from Hansen (1992): in column (1), the one percent criticalvalue is 2.12; in column (2), the one percent critical value is 1.88.

spectively) are equal. Second, we test for the stability of themodel using a statistic developed in Hansen (1992), whichallows for the possibility of more general forms of instabilitythan do structural break tests such as Chow tests. Evidenceof instability would be consistent with our premise that thepropensity to consume out of wealth is not stable over time.

The results of estimating equation (A1) are presented inthe first column of Appendix Table A1.22 The coefficientestimates on the short-term interest rate and income growthare in rough accord with the results from Campbell andMankiw (1989). The coefficient on the interest rate is smalland statistically insignificant. The coefficient on incomegrowth is positive and statistically significant: it is some-what smaller than the Campbell and Mankiw estimates, butit still signifies that realized income growth has a substan-tial effect on consumption growth.

The coefficients on non-housing and housing wealthare both positive, although only the one on non-housingwealth is statistically significant. Even so, the null hypoth-esis that the coefficients on the two wealth variables areequal cannot be rejected: the p-value of the statistic is over0.35. In fact, when estimating equation (A1) by replacingthe growth of housing and non-housing wealth with thegrowth of total wealth, the coefficient on total wealth ishighly statistically significant and the fit of the regressionis similar, as seen in the second column of Table A1. Theseresults imply that when estimating the propensity to con-sume from a long sample, the propensity out of housingwealth does not differ drastically from that out of non-hous-ing wealth.23

Turning to the stability of the model in equation (A1),the stability test statistics indicate that the regression is notstable both in the case where the wealth variables are en-tered separately (column 1) and in the case where they en-tered in a single total wealth variable (column 2), since thetest statistic rejects the null hypothesis of stability easilyat the one percent significance level. This implies that es-timating propensities to consume from long-sample regres-sions such as this is fraught with hazard. Examining someadditional information from stability test statistics (notshown in Table A1 to conserve space) indicates that muchof the instability results from a non-constant variance of theerror term in the regression. This observation in turn im-plies that the coefficients on the variables in the regressionprobably are not stable. In particular, it is unlikely that theestimates of the propensity to consume from models likethe one presented here are stable, which is consistent withour second premise.

Appendix 2: Developments Since the Middle of2006

Since this paper was drafted, residential investment hasentered into a severe downturn. By early 2007, housingstarts were off nearly 40 percent from their peak levels. Theannualized rate of decline in real residential investmentwas between 10 and 20 percent in each of the last threequarters of 2006, on average reducing the overall real GDPgrowth rate by more than one percentage point. While thenumbers received as of early March 2007 do not show aprecipitous drop in home values, declines were evident insome home price measures (though not the OFHEO index,which still showed moderate growth through the fourthquarter of 2006); and capital gains on housing clearly havegreatly diminished as a source of increases in aggregatewealth. Moreover, the volume of home equity withdrawalhas plummeted: in the third quarter of 2006, the dollar val-ues of gross and net home equity extraction were both lessthan half their year-earlier figures, and had even fallen be-neath their 2002 averages.

Nonetheless, despite the weakness in housing activityand finance, consumer spending remained sturdy. Thegrowth rate of real consumption averaged more than threepercent over the last three quarters of 2006, contributingsubstantially more than two percentage points to real GDPgrowth. These figures were well in line with the norms of thelast few years. Thus, the consumer spending numbers havenot yet suggested any marked spillover from housing weak-ness. One might argue that the housing impact has beenmasked: strong gains in the stock market offset throughearly 2007 the weakening of gains from housing and keptaggregate household wealth on a strong upward trajectory.It also may, of course, be the case that there will be laggedeffects on consumption from the weakening of housing, pos-sibly resulting from declines in home values that might beexperienced. �

AC K N OW L E D G EM E N TWe would like to thank Cartier Stennis for his research

assistance on the figures and tables.

R E F E R E N C E SAndo, Albert and Franco Modigliani. 1963. “The ‘Life Cycle’ Hy-

pothesis of Saving: Aggregate Implications and Tests.” American Eco-nomic Review, 53:55-84.

Aoki, Kosuke, James Proudman, and Jan Vliege. 2004. “HousePrices, Consumption, and Monetary Policy: A Financial AcceleratorApproach.” Journal of Financial Intermediation, 13:414-435.

Baker, Dean. 2006. “The Menace of an Unchecked Housing Bub-ble.” The Economists’ Voice, 3(4) Article 1, http://www.bepress.com/ev/vol3/iss4/art1.

Belsky, Eric, and Joel Prakken. 2004. “Housing Wealth Effects:Housing’s Impact on Wealth Accumulation, Wealth Distribution and


22The sample period in our estimation is 1960Q1-2006Q1.23However, it is possible that the propensities may differ over shortersamples if they fluctuate over time.

Consumer Spending.” Harvard University, Joint Center for HousingStudies, W04-13.

Benjamin, John D., Peter Chinloy, and G. Donald Jud. 2004. “RealEstate Versus Financial Wealth in Consumption.” Journal of Real EstateFinance and Economics, 29:341-354.

Bernanke, Ben S., Mark Gertler, and Simon Gilchrist. 1999. “TheFinancial Accelerator in a Quantitative Business Cycle Framework.” InJohn B. Taylor and Michael Woodford (eds.), Handbook of Macroeco-nomics, Vol. 1C. Amsterdam: Elsevier, North Holland, 1341-93.

Bostic, Raphael, Stuart Gabriel, and Gary Painter. 2005. “HousingWealth, Financial Wealth, and Consumption: New Evidence from MicroData.” Lusk Center for Real Estate, University of Southern California.

Bucks, Brian K., Arthur B. Kennickell, and Kevin B. Moore. 2006.“Recent Changes in U.S. Family Finances: Evidence from the 2001 and2004 Survey of Consumer Finances.” Federal Reserve Bulletin, 92,www.federalreserve.gov/pubs/bulletin/2006/financesurvey.pdf.

Campbell, John Y., and N. Gregory Mankiw. 1989. “Consumption,Income, and Interest Rates: Reinterpreting the Time Series Evidence.”In Olivier J. Blanchard and Stanley Fischer (eds.), NBER Macroeco-nomics Annual: 1989. Cambridge, MA: MIT Press, 185-216.

Canner, Glenn, Karen Dynan, and Wayne Passmore. 2002. “Mort-gage Refinancing in 2001 and Early 2002.” Federal Reserve Bulletin,88:470-481.

Carroll, Christopher D. 1997. “Buffer-Stock Saving and the LifeCycle/Permanent Income Hypothesis.” Quarterly Journal of Economics,112:1-55.

Davis, Morris A., and Michael G. Palumbo. 2001. “A Primer on theEconomics and Time Series Econometrics of Wealth Effects.” FederalReserve Board Finance and Economics Discussion Paper No. 2001-09.

Deaton, Angus. 1991. “Saving and Liquidity Constraints.” Econo-metrica, 59:1221-48.

Friedman, Milton. 1957. A Theory of the Consumption Function.National Bureau of Economic Research

Greenspan, Alan and James Kennedy. 2005. “Estimates of HomeMortgage Originations, Repayments, and Debt on One-to-Four FamilyResidences.” Federal Reserve Board Finance and Economics Discus-sion Series 2005-41.

Hall, Robert. 1978. “Stochastic Implications of the Life Cycle-Per-manent Income Hypothesis: Theory and Evidence.” Journal of Politi-cal Economy, 86:971-987.

Hansen, Bruce. 1992. “Testing for Parameter Instability in LinearModels.” Journal of Policy Modeling, 14:517-533.

Himmelberg, Charles, Christopher Mayer, and Todd Sinai. 2005.“Assessing High House Prices: Bubbles, Fundamentals and Misper-ceptions.” Journal of Economic Perspectives, 19(4):67-92.

Hurst, Erik, and Frank Stafford. 2004. “Home is where the Equityis: Mortgage Refinancing and Household Consumption.” Journal ofMoney, Credit, and Banking, 36:985-1014.

Iacoviello, Matteo. 2004. “Consumption, House Prices, and Col-lateral Constraints: A Structural Econometric Analysis.” Journal ofHousing Economics, 13:304-320.

________. 2005. “House Prices, Borrowing Constraints, and Mon-etary Policy in the Business Cycle.” American Economic Review. 95:739-64.

Lettau, Martin, and Sydney Ludvigson. 2001. “Consumption, Ag-gregate Wealth, and Expected Stock Returns.” Journal of Finance,56(3):815-49.

Ludvigson, Sydney. 2003. “Analyzing the Linkage Between Con-sumption and Wealth.” Presented to the Federal Reserve Bank ofPhiladelphia Policy Forum, November 14, 2003, http://www.phil.frb.org/econ/conf/forum2003/Ludvigson_final.pdf.

Ludvigson, Sydney and Charles Steindel. 1999. “How Important isthe Stock Market Effect on Consumption?” Federal Reserve Bank of NewYork Economic Policy Review, 5(2): 29-51.

Lustig, Hanno, and Stijn van Nieuwerburg. 2005. “Housing Col-lateral, Consumption Insurance and Risk Premia: An Empirical Per-spective.” Journal of Finance, 60: 1167-1219.

McCarthy, Jonathan. 1995. “Imperfect Insurance and DifferingPropensities to Consume Across Households.” Journal of Monetary Eco-nomics. 36: 301-327.

McCarthy, Jonathan and Richard W. Peach. 2002. “Monetary Pol-icy Transmission to Residential Investment.” Federal Reserve Bank ofNew York Economic Policy Review, 8(1):139-58.

———. 2004. “Are Home Prices the Next ‘Bubble’?” Federal Re-serve Bank of New York Economic Policy Review, 10(3):1-17.

McConnell, Margaret, Richard Peach, and Alex Al-Hashimi. 2003.“After the Refinancing Boom: Will Consumers Scale Back Their Spend-ing?” Federal Reserve Bank of New York Current Issues in Economicsand Finance, 9(12).

Shiller, Robert J. 2006. “Long-Term Perspectives on the CurrentBoom in Home Prices.” The Economists’ Voice, 3(4) Article 4,http://www.bepress.com/ev/vol3/iss4/art4.

Steindel, Charles. Forthcoming. “The Consumption Wealth Effect.”In Critical Perspectives on Recent Developments in Macroeconomics (PerGunnar Berglund and Leanne J. Ussher, eds), Routledge.

Tracy, Joseph, Henry Schneider, and Sewin Chan. 1999. “AreStocks Overtaking Real Estate in Household Portfolios?” Federal Re-serve Bank of New York Current Issues in Economics and Finance, 5(5).

Housing Activity and Consumer Spending Business Economics • April 2007 21

Housing Activity and Consumer Spending

Documents

Transcript of Housing Activity and Consumer Spending