wealth and asset price effects on economic activity - european

OCCAS IONAL PAPER SER IESNO. 29 / JUNE 2005

WEALTH AND ASSETPRICE EFFECTS ONECONOMIC ACTIVITY

by Filippo Altissimo,Evaggelia Georgiou,Teresa Sastre,Maria Teresa Valderrama,Gabriel Sterne, Marc Stocker,Mark Weth, Karl Whelan and Alpo Willman

In 2005 all ECB publications will feature

a motif taken from the

€50 banknote.

OCCAS IONAL PAPER S ER I E SNO. 29 / J UNE 2005

This paper can be downloaded from the ECB’s website (http://www.ecb.int).

1 This is the joint work of a team composed of Filippo Altissimo (ECB), Evaggelia Georgiou (Bank of Greece),Teresa Sastre (Banco de España), Maria Teresa Valderrama (Österreichische Nationalbank),Gabriel Sterne (Bank of England), Marc Stocker (ECB), Mark Weth (Deutsche Bundesbank),

Karl Whelan (Central Bank of Ireland) and Alpo Willman (ECB).

WEALTH AND ASSETPRICE EFFECTS ON

ECONOMIC ACTIVITY 1

by Filippo Altissimo,Evaggelia Georgiou,

Teresa Sastre,Maria Teresa Valderrama,

Gabriel Sterne, Marc Stocker,Mark Weth, Karl Whelan

and Alpo Willman

© European Central Bank, 2005

AddressKaiserstrasse 2960311 Frankfurt am MainGermany

Postal addressPostfach 16 03 1960066 Frankfurt am MainGermany

Telephone+49 69 1344 0

Websitehttp://www.ecb.int

Fax+49 69 1344 6000

Telex411 144 ecb d

All rights reserved. Reproduction foreducational and non-commercialpurposes is permitted provided thatthe source is acknowledged.

The views expressed in this paper donot necessarily reflect those of theEuropean Central Bank.

ISSN 1607-1484 (print)ISSN 1725-6534 (online)

3ECB

Occasional Paper No. 29June 2005

C O N T E N T S1 INTRODUCTION 4

2 WEALTH CHANNEL 4

2.1 A theoretical review 5

2.1.1 Implications of the lifetimebudget constraint 5

2.1.2 Implications of riskaversion 7

2.1.3 Implications of incomeuncertainty 8

2.1.4 Do marginal propensities toconsume out of differentwealth components differ? 8

2.1.4.1 Implications of marketimperfections, taxation anddiversified bequest motives 9

2.1.4.2 Implications of the(un)predictability of assetprices 9

2.1.4.3 Marginal propensities toconsume out of non-equityfinancial wealth, equitywealth and housing wealth 11

2.2 Some conclusions 12

3 HOUSEHOLD WEALTH:A STATISTICAL REVIEW 13

3.1 Decomposition of household wealthinto financial and housing wealth 13

3.2 Decomposition of householdfinancial wealth to components 14

3.3 Equity wealth effect, incomedistribution and demographics 15

4 ASSET PRICE EFFECTS VIA THE WEALTHCHANNEL: EMPIRICAL EVIDENCE 18

4.1 Micro data evidence 18

4.2 The MPC out of wealth in somecalibrated theory-based models 19

4.3 Aggregate time series estimates ofMPC out of wealth 20

4.3.1 MPC out of wealth basedon the estimated deepparameters of theconsumption function 20

4.3.2 Aggregate time-seriesestimate of MPC out ofwealth: direct data-basedapproach 21

4.3.2.1 Evidence based oncross-country studies 21

4.3.2.2 Evidence based onsingle-country studies 24

4.4 Overview and provisos toestablished results 26

4.5 How significant are cross-countrydifferences in estimated MPC out ofwealth? 27

4.6 Conclusions 28

5 ASSET PRICES AND INVESTMENT 29

5.1 Equity prices and corporateinvestment 29

5.2 Tobin’s Q or cost-of-capitalchannel 30

5.3 House prices and housinginvestment 33

6 CONCLUSIONS 34

APPENDIXES

1 Comparability of equity wealthdata across countries 37

2 The US evidence on wealth effectson consumption 40

3 Balance sheet channel 45

REFERENCES 52

4ECBOccasional Paper No. 29June 2005

1 INTRODUCTION

Do asset prices affect real activity? Thisquestion has taken on a new importance inrecent years, as asset values first surged at theend of 1990s and, thereafter, dramaticallyretreated. The macroeconomic importance ofthis phenomenon is illustrated by Figure 1,which presents stock market capitalisation as apercentage of GDP in some majorindustrialised countries. From the 1970s to thelate 1990s this ratio rose six to eightfold in allcountries (except in Japan). However,subsequently it fell markedly, and in 2002 ithad declined to around half of its late 1990speak. In 2003 the stock market capitalisationratio began to recover moderately. It istherefore natural to ask how much of thisvariation has been transmitted into economicactivity?

There are four channels through which stockprices can be considered to affect activity:(1) the wealth effect on consumption, (2) theTobin’s Q effect on investment, (3) the balancesheet effect on private spending (via the creditchannel) and (4) the confidence effect onprivate spending. The approach that we follow

Figure 1 Stock market capital isat ion insome large economies

(as a percentage of GDP)

in this report is that of partial equilibriumrather than general equilibrium, because thefocus is on the transmission of asset priceeffects on activity via these channels. Our aimis to review the evidence contained in therecent and fast growing literature. The maintext concentrates on the first two channels, asdistinguishing the balance sheet andconfidence effects from those transmitted viathe wealth and Tobin’s Q channel is, at least inthe macro data, very difficult. However,balance sheet and confidence channels arediscussed in Appendix 3 of this report. Section2 presents the theoretical background of thewealth channel. A statistical review of thewealth composition across countries is given inSection 3. The empirical evidence on theeffects of wealth and its main components onconsumption is reviewed in Section 4. Theeffect of asset prices on investment via Tobin’sQ channel is reviewed in Section 5 and Section6 concludes.

2 WEALTH CHANNEL

The wealth channel has sound theoreticalfoundations. Consumers intertemporal utilitymaximisation under a lifetime resourceconstraint, states that current consumption isproportional to total wealth,

(2.1) ( )[ ]

YmpcAmpcYHAmpcC

YW

W

+≈+=

where C is consumption, A is real non-humanwealth and H is real human wealth, i.e. thepresent value of expected labour income (net oftaxes) Y. The proportionality coefficients mpcmeasure the marginal propensity to consume(MPC) out of wealth and income, respectively.As, by definition, labour income and humanwealth are co-integrated, a long-run relationfor consumption can also be presented as theweighted average of non-human wealth andlabour income, both of which are observablevariables. From the point of view of themacroeconomic importance of the wealthchannel, it is illuminating to transform (2.1)into an elasticity form as follows,

0

50

100

150

200

0

50

100

150

200

1973 1978 1983 1988 1993 1998 2003

France ItalyUS

GermanyJapanUK

5ECB


2 WEALTHCHANNEL

(2.2)

YY

CYmpc

AA

CA

mpc

YY

CYmpc

AA

CAmpc

C∆C

Yj

j

e

jw

e

Y

e

W

jW

WW

∆⎥⎦⎤

⎢⎣⎡+

∆=

∆⎥⎦⎤

⎢⎣⎡+∆

⎥⎦⎤

⎢⎣⎡=

∑

−

43421

4342143421

,

1

The relation (2.2) shows that the size of wealthelasticity of consumption e

W depends, in

addition to mpcW

, on the size of the wealthconsumption ratio or, if differentiated acrosswealth components, on the wealth consumptionratio of each wealth component j. We can alsosee that the elasticity e

W (or e

Wj), is constant

only if the wealth (or the wealth component j)to consumption ratio is constant. Given thewide variations in the wealth to income ratioindicated by Figure 1 and theoretical priorssuggesting a constant or stationary MPC, thisimplies that the logarithmic specification inestimation may be functionally misspecified,especially when wealth is disaggregated intocomponents. Hence, the direct estimation ofMPCs should be preferred to the estimation ofelasticities. Unfortunately this contrasts withgeneral practice, which has favouredlogarithmic transformation in estimation.

In the following we first concentrate on theMPC out of wealth. We start by reviewing whattheory suggests concerning the determinantsand appropriate size of the MPC out of wealth.In addition, we consider whether the size of theMPC out of wealth should be the same acrosscountries and across wealth components and, ifnot, then why not? Section 3 compares thecomposition of wealth across countries. InSection 4 we shift to the evidence given bystudies based on the use of both micro andmacro data. These studies can be classified intothree approaches: theory-based calibration,estimated consumption functions withparameter constraints imposed by economictheory and direct estimation without theory-imposed constraints. Our conclusion is thatthere is considerably more variation in theestimates presented in these studies than would

be expected in the light of underlying economictheory. Therefore, the quality of the data usedin the alternative studies is assessed. Finally asummarised view on the size of the MPC out ofwealth is proposed based on the aggregatepanel data evidence.

2.1 A THEORETIC REVIEW

In the following we take a closer look at thewealth effect on consumer expenditure. Wefirst study the implications of the householdintertemporal budget constraint without anyreference to risk aversion. Thereafter we studythe impact of risk-aversion on the MPC out ofwealth in the framework of the finitely livingoverlapping generation model by Blanchard(1985). Then the effects of income uncertaintyare discussed and, more heuristically, thelikely effects of market imperfections anddistortionary taxation are discussed. Finally,we consider the implications of dividing assetprice movements into permanent and transitorycomponents for the asset price effects onconsumption.

2.1.1 IMPLICATIONS OF THE LIFETIME BUDGETCONSTRAINT

The logic of budget constraints dictates that,when an individual’s wealth rises, theindividual must either spend that wealth whileliving, or leave the money to other individuals,charities or the government after death.Nevertheless, the central issue in analysingwealth effects is timing. If the lag between afavourable shock to the household balancesheet and an increase in consumption spendingtakes many years to develop, then marketfluctuations may have a limited impact onaggregate spending. However, if the link fromnet worth to consumption is powerful andimmediate, then sharp changes in asset valuesmay translate into sharp changes in consumerspending. The following more formalpresentation tries to deepen this insight.


Conventional budget constraints faced byhouseholds can be written as follows:

(2.3) [ ]ttttat CYARA −+= ++ 1,1 ,

where At is the aggregate level of real and

financial assets at the end of period t, Yt is non-

property income (i.e. labour income), Ct is

consumption and Ra,t+1

(=1 + ra,t+1

) is a time-varying return on total assets. If A

t contains

both riskless and risky assets, then Ra,t

can beinterpreted as a weighted average of the returnsof these assets. By solving forward andimposing the transversality condition that atthe end of a finite horizon the limit ofdiscounted future wealth is zero, we obtain:

T

iit

i

jjtatat CRRE ∑ ∏

=+

−

=+

−⎟⎟

⎠

⎞

⎜⎜

⎝

⎛

1

1

0,

1,

(2.4)

44444 344444 21tH

T

iit

i

jjtatatt YRREA ∑ ∏

=+

−

=+

−⎟⎟

⎠

⎞

⎜⎜

⎝

⎛+=

1

1

0,

1,

Equation (2.4) tells us that today’s total wealth,which is the sum of real and financial assets andthe discounted sum of expected future labourincome (W

t= A

t+ H

t), equals the discounted

value of the planned future consumption.Furthermore, (2.4) implies that as a response toa permanent unanticipated wealth shock thediscounted sum of future consumption mustrise by an equal amount, i.e.

(2.5) =∆ tA CRRET

iit

i

jjtatat ∆⎟⎟

⎠

⎞

⎜⎜

⎝

⎛

=+

−

=+

−∑ ∏1

1

0,

1,

CRR

a

Ta ∆⎟

⎟⎠

⎞⎜⎜⎝

⎛

−−

≈ −

−

111

⇔

( )( )

∆⎥⎥⎦

⎤

⎢⎢⎣

⎡

−++

+=∆

111

1A

rr

rr

C t

mpc

Ta

Ta

a

a

W

444 3444 21

where ∆ indicates the difference between thepost-shock and pre-shock values, ∆C refers tothe average level shift in consumption over thehorizon from t to T and 1+r

a= R

a is the average

return on non-human wealth over the horizonfrom t to T. Equation (2.5) defines the long-runMPC out of wealth mpc

W conditional on the

positive wealth shock being permanent. Inaddition to being dependent on the expectedaverage return on non-human wealth it isaffected by the horizon of the representativeconsumer.

To anchor the empirical analysis of how wealthaffects consumer spending, Table 2.1 followsPoterba (2000), presenting some benchmarkestimates of the amount by which a householdmight increase its consumption if it received afavourable wealth shock. Table 2.1 reportsvalues for a feasible annual permanent increasein consumption (MPC out of wealth),expressed in cents per one euro (dollar)increase in wealth, for the ranges of alternativeplanning horizons and the average expectedreturns on non-human wealth.

The length of planning horizon: For example,a household with a 30-year planning horizonthat faces a 3% real return of wealth can raiseconsumption outlays by five cents for each oneeuro increase in wealth. In addition, Table 2.1shows that the length of planning horizon isimportant; the shorter the horizon the higher

Horizon in years

R–a = 1 + r–a 15 30 50

1.03 8.1 5.0 3.8 2.91.05 9.2 6.4 5.4 4.81.07 10.3 7.5 6.8 6.5

Table 2.1 Long-run MPCs out of wealth withalternative planning horizons and returns

8

where

mpcw =

⎪⎩

⎪⎨

⎧

∆

∆⎟⎟⎠

⎞⎜⎜⎝

⎛+

1 1 rr

a

a

→

∞→

1 when

when

T

T

7ECB


2 WEALTHCHANNEL

the MPC out of wealth. This relation is non-linear so that MPC out of wealth valuescorresponding to the 30-year horizon are closerto MPC out of wealth values implied by aninfinite horizon than those implied by a 15-yearhorizon.

The following two factors, among others, mayaffect the length of the planning horizon andhence the aggregate level of the MPC out ofwealth.

– It is natural to associate the length ofplanning horizon with the expectedremaining lifetime. At a macro level animplication might be that the moreconcentrated wealth is in older generationsthe shorter the planning horizon of the“average” wealth owner and the higher theMPC out of wealth

.

– On the other hand, a strong bequest motiveaffects the MPC out of wealth largely in thesame way as the lengthening of the planninghorizon (see, for instance, the discussion byCampbell and Viceira, 1999).

The rate of return on real and financialassets: Another determinant of the MPC out ofwealth is the expected rate of return on assets.The higher it is, the higher is the MPC out ofwealth. However, the required rates of returnmay differ across wealth components,reflecting the risks associated with them.Therefore, as the relevant discounting rate ofthe consumption stream is the weightedaverage of the rates of return of individualwealth components, the allocation of wealthacross risky and less risky assets matters. Morespecifically, an increase in the portfolio shareof risky assets should increase the MPC out ofwealth. In addition, if risk premiums vary overtime, a standard assumption in the currentfinancial literature, then the MPC out of wealthshould also be time-varying. This empiricallytestable hypothesis is discussed in Section 4and Appendix 2 of this report. However, timevariation of the MPC out of wealth may bereduced if the rate of return used in discounting

the planned consumption stream is not thecurrent period rate of return, but the expectedaverage over the planning horizon.

2.1.2 IMPLICATIONS OF RISK AVERSIONThe above analysis, based on the lifetimebudget constraint, was not able to account forthe possible effects of risk aversion onconsumption. Therefore, let us assume thefollowing instantaneous constant relative risk-

aversion (CRRA) utility function ⎟⎠⎞

⎜⎝⎛ −

−σ

σ

σσ 1

1 tC ,where σ ≥ 0 is the intertemporal elasticity ofsubstitution (the inverse of relative riskaversion). Now, in Blanchard’s (1985) finitelyliving overlapping generations model with theconstant probability of death π, intertemporalutility maximisation yields the following MPCout of total wealth:

(2.6) ( )( ) ππ

π +≈−+

+= WW

WW r

rrmpc

11

with the risk-adjusted return rW

defined by

(2.7)( ) ( ) ( ) aaW rrr ⋅−+⋅≈−++= − σρσρ σσ 1111 1

and where ρ is the subjective discount rate.According to (2.6), mpc

W is positively related

to the “risk-adjusted” return rW as well as to the

probability of death (i.e. the inverse of theexpected lifetime Te =1/π). Equation (2.7)defines r

W as a weighted average of the

subjective discount rate ρ and the rate of returnon wealth r

a. When varying the “risk-adjusted”

return rW

and the expected lifetime Te =1/π inthe same range as we varied r

a and T in Table

2.1, relation (2.6) gives somewhat higherfigures than those in Table 2.1. For instance,with Te = 30 the MPCs corresponding to the r

W

values of 3%, 5% and 7% would be 6.3, 8.6 and10.3 cents per one euro permanent increase inwealth, respectively. These figures aresomewhat higher than corresponding figures inTable 2.1 (i.e. 5, 6.4 and 7.5 cents). On theother hand, as shown in Figure 2.1, if 10 ≤< σ ,which implies high risk-aversion, r

W and mpc

W

are only partially dependent on the real rate ofreturn r

a. However, with low risk-aversion,

σ > 1, the long-run real risk-free interest rate


(or the subjective discount rate) is no longer thelower bound for mpc

W, as the sign of the

dependence of mpcW

on ra changes from

positive to negative. Combining very highintertemporal elasticity of substitution (i.e.very low risk aversion) with the plausibleassumption that the real return of wealth r

a > ρ,

even negative values for mpcW

cannot be ruledout.

Accordingly, the analysis in the previoussection must be extended to considerhouseholds’ attitudes to risk. When the realrate of return on wealth r

a is higher than the

subjective discount rate, equations (2.6) and

(2.7) imply that 0<∂

∂σ

Wmpc, i.e. a decrease in

risk aversion decreases the MPC out of wealth.1

What then is a plausible the size of mpcW

? Thislargely depends on whether σ is below or aboveunity. Estimates of σ both below and aboveunity can be found in the literature, althoughmost are below unity. For example, much of thereal business cycle (RBC) literature assumesan elasticity of intertemporal substitutionbetween a half and one. Recently, Smets andWouters (2002) estimated the elasticity ofsubstitution at 0.6 with euro area data.Applying this estimate and the assumption that

the subjective discount rate is around theaverage long-term government bond rate(around 3-4%), relation (2.6) produces, withthe same values for r

a and Te, a range of mpc

W

values similar to that presented in Table 2.1.

2.1.3 IMPLICATIONS OF INCOME UNCERTAINTYWhat about uncertainty concerning labourincome? Does it affect the MPC out of wealth?As closed form decision rules for optimalconsumption in the presence of uncertainlabour income cannot in general be derived,Zeldes (1989) used numerical methods toclosely approximate the optimal consumptionfunction and the corresponding value functionfor some multi-period problems with uncertainlabour income. One of the most strikingfeatures for his numerical solutions for theconsumption rule under uncertainty was theintroduction of concavity to consumption rulesthat were linear in the absence of uncertainty.Income uncertainty makes the MPC out ofwealth state dependent, i.e. the MPC out ofwealth of a consumer with low wealth wasmarkedly higher than that of a consumer withhigh wealth. Later Carroll and Kimball (1996)have confirmed that result analytically. Theyshowed that, except under very specialconditions, adding income uncertainty to thestandard optimisation problem induces aconcave consumption function in which theMPC out of wealth declines along with theincreasing level of wealth.

2.1.4 DO MARGINAL PROPENSITIES TO CONSUMEOUT OF DIFFERENT WEALTH COMPONENTSDIFFER?

Total wealth is not homogenous, but consists ofseveral components with different risk,collateral and liquidity properties. In addition,the services associated with each wealthcomponent vary widely. Therefore, the MPCout of each wealth component may reflect thesedifferences. In this section we briefly discussthis issue, with special attention paid to theMPC out of the three main components of

1 This effect is strengthened by the reasonable assumption thatthe risk premium ρ−ar also depends positively on the rate ofrisk aversion.

Figure 2.1

Risk adjusted rate of return (risk aversion = 1/sigma)

rw = sigma * rho + (1-sigma) * ra

sigma (intertemp. elasicity of substituiton)1

ra

rho

9ECB


2 WEALTHCHANNEL

household wealth, i.e. non-equity financialwealth, equity wealth and tangible (mainlyhousing) wealth.

2.1.4.1 Implications of market imperfections,taxation and diversified bequest motives

Typically the behaviour of utility maximisingconsumers is analysed assuming unboundedrationality, perfect capital markets and theabsence of distorting taxes or rigidities. Inthese ideal circumstances, where thecomposition of wealth can be changed withoutfriction or cost, the MPCs out of differentwealth components should naturally be equal.However, this is no longer true when theseassumptions are relaxed. In imperfect capitalmarkets with distorting taxes and onlyboundedly rational consumers, the propertiesassociated with different wealth componentsmay well affect their respective MPC. Forinstance,

– The liquidity properties of alternativewealth components may be very differentreflecting, for instance, the fact that thecollateral markets for household borrowingare underdeveloped in most countries, or atleast have been until recently. Now, if theonly way for a house owner to increaseconsumption, as a response to a rise inhousing value, is first to sell the house, it isquite evident that most house owners willnot react to the increase in the market valueof housing. Moreover, even in well-developed financial markets, the requiredcapital adequacy ratios of banks may inducethe banks to evaluate collateral values ofassets below their market values. Thesearguments suggest that the MPC out of eachwealth component depends positively onthe liquidity of the asset in question andnegatively on the ratio of the market valueof the asset to its collateral value.

– From the point of view of the bequestmotive, wealth components are not similar.For instance, property (including shares)coupled with entrepreneurship (e.g. the“family business”) may be associated with a

much stronger bequest motive and, hence, alower MPC out of wealth, than other wealthcomponents. In addition, fiscal distortions(e.g. estate taxes) may favour some assetsfor consumption and others for bequests.

– Similarly, Thaler (1994) argued that“mental accounting”, which dictates thatsome assets are more appropriate forcurrent consumption while others are betterfor long-term savings, may result indifferent MPCs out of wealth across wealthcomponents. Many behavioural studiessupport this view. As saving requireswillpower, it is easier for many consumersto save from lump-sum payments or fromcapital gains rather than from “regularincome”.

2.1.4.2 Implications of the (un)predictability ofasset prices

Efficient financial markets are the underlyingassumption in most of the macroeconomicmodelling work. This hypothesis implies thatasset prices follow a random walk so that allmovements are unpredictable or, equivalently,all asset price movements are expected to bepermanent. Under this interpretation, theexpected rate of return (discount rate) isconstant and changes in asset prices generatechanges in consumption.

However, the underlying hypothesis ofefficient financial markets with a constantdiscount rate (or constant expected rate ofreturn) was challenged by Shiller (1981) andLeRoy and Porter (1981), who argued thatstock returns were too volatile to be accountedfor by variation in future dividend growth withconstant discount rates. There is now a largeand growing body of empirical literature infinancial economics which finds that aggregatestock market returns are predictable over longhorizons (see e.g. Shiller, 1984, Campbell andShiller, 1988, Fama and French, 1988,Hodrick, 1992, Lamont, 1998, Lettau andLudvigson, 2001, Campbell, 2001, Lewellen,2001, and Campbell and Goto, 2002). Thissuggests that conditional expected returns, or


discount rates, vary over time, possibly drivenby cyclical variation in risk aversion (see e.g.Sundaresan, 1989, Constantinides, 1990,Campbell and Cochrane, 1999, and Cochrane,2001, Chapter 20).

Campbell and Cochrane (1999) present aconsumption-based model that explains,among a wide variety of dynamic asset-pricingphenomena, the long-run predictability ofexcess stock returns (mean reversion). The keyelement of their explanation is the time-varying habit, or subsistence, level ofconsumption added to the constant risk-aversion utility function. However, incalibrating their model to actual data, theyfound that the process defining the habit levelof consumption must be slow-moving and non-linear to be able to generate mean reversion ofasset prices. Although the question of whetherthese requirements are compatible with actualdata is largely untested, Fuhrer (2000), forinstance, found that one period lag ofconsumption, instead of its long lagdistribution, was sufficient to capture the habitformation effect on consumption.

Anyway, independently from possibletheoretical explanations, the division of assetprice movements into the expected permanentand transitory components may affect the MPCout of equity wealth. To clarify this issueLettau and Ludvigson (2001) used a more data-oriented approach. They examined theimplications of the intertemporal budgetconstraint to the long-run cointegrationrelationship between consumption, income andwealth. Taking as their basis the work ofCampbell and Mankiw (1989) and Campbell(1996), Lettau and Ludvigson (2001) showedthat the log approximation of the budgetconstraint (2.4) can be presented in thefollowing form (lowercase letters denote logvariables):

(2.8) ( ) ttt yac αα =−−⋅− 1

( )[ ]{ } ( ) ti

itithitaiwt zcrrE αααρ −+∑ ∆−⋅−+⋅

∞

=+++ 11

1,,

( )[ ]{ } ( ) t

iitithit

iwt zcraE αααρ −+∑ ∆−⋅−+∆⋅≈

∞

=+++ 11

1,

where ithr +, is real net return on human wealthand itita ar ++ ∆≈, , see Campbell, Lo andMacKinlay (1997). Parameter α is the averageshare of non-human wealth in total wealth(equalling approximately the GDP share ofnon-labour income) and z

t is the residual of the

cointegrating relationship between humanwealth h

t and labour income y

t. If all the right-

hand-side variables of (2.8) are presumedstationary, then c, a and y must be cointegrated,and the left-hand side of (2.8) gives thedeviation in the common trend of c

t, a

t and y

t.

It follows that the real effects of changes in themarket valuation of wealth depend on how theadjustment burden of the cointegration residualimplied by the left-hand side of (2.8) is fed backto consumption, labour income and the marketvalue of non-financial wealth, reflectingfluctuations in expected returns on wealth. Ifthe mean reversion tendency of asset prices isstrong, which implies a strong transitorycomponent in asset prices, dynamic effects onconsumption may remain modest although thedynamic equation estimated for consumptionwould contain a large and significantcoefficient for the lagged cointegrationresidual. This aspect is stressed by Lettau andLudvigson (2002) and, especially, by Lettau,Ludvigson and Barczi (2001) in their commenton Davis and Palumbo (2001).

As a better summary statistic, which wouldaccount for the finding that much of thevariation in wealth is transitory anduncorrelated with consumer spending, Lettauand Ludvigson (2002) propose the followingcorrected measure:

(2.9) 10 ; 1

≤≤⎟⎟⎠

⎞⎜⎜⎝

⎛

−+= γ

γγγ

Wcorr mpcmpc

where γ is the is the share of the transitorycomponent in the variance of wealth.

11ECB


2 WEALTHCHANNEL

However, as mean reversion in asset pricesreflects fluctuations in expected returns onequity wealth, this issue is also related to thepossible time-variance of the MPC out ofwealth as is discussed by Palumbo et al. (2002)and in Appendix 2 of this report.

2.1.4.3 Marginal propensities to consume out ofnon-equity f inancial wealth, equitywealth and housing wealth

We now briefly review the discussion of therelative sizes of the MPC out of the three maincomponents of household wealth: non-equityfinancial wealth, equity wealth and housingwealth. In line with our discussion in twoprevious sub-sections, there appears to begeneral agreement that, if the MPC out ofwealth differs across these wealth components,then the MPC out of non-equity financialwealth is likely to be highest. However, therelative sizes of the MPCs out of equity wealthand out of housing wealth is morecontroversial. For both of these wealthcomponents the main variation in their marketvalues comes from the price rather than thevolume component. Hence, the sensitivity ofconsumption to these price movements maydepend on how liquid these asset are and howpermanent the price changes are expected to be,i.e. whether the mean reversion tendency isassociated with asset price changes. Thefollowing arguments have been presented infavour of a higher MPC out of housing wealth:

– Since equity prices are more volatile thanhouse prices, households may find it moredifficult to assess whether a change in theequity wealth is permanent or temporary.Therefore, they are likely to be morecautious in adapting consumption plans tochanges in equity wealth than in housingwealth.

– Since house purchases are generallyfinanced with borrowed money, increasesin property values result in a higher netreturn on this investment than on otherassets, implying that the MPC out of

housing wealth may be larger than for assetswith lower expected returns.

– In most countries housing wealth is moreevenly distributed than equity holdings,which are concentrated in the upper tail ofincome distribution. If, as some studiesindicate, the MPC is lower among therichest households, then, at the aggregatelevel, the wealth effect from housing maybe expected to be more important than theeffect from equity wealth.

– In some countries, differential tax-treatment of equity holdings and residentialproperty may lower the MPC out of equitywealth because stock holdings have to beretained to receive a more favourable tax-treatment.

On the other hand, a partial equilibriumanalysis may suggest a less clear-cut housingwealth effect on consumption because of thedual effects of house prices on non-housingconsumption, i.e. a positive wealth effect,which depend on the liquidity of houses, and anegative relative price effect.2 Since houses arepart of household wealth, changes in theirvalue are expected to have a positive effect onthe consumption of homeowners. Conversely,for tenants, an increase in house prices leads tohigher rental prices and a negative incomeimpact on their consumption. However, thewealth effect for landlords or pension fundsand similar institutional investors owningrental housing has also to be considered. Thiswealth effect, at least to the extent that thelandlords belong to the household sector,neutralises the negative consumption effects ofthe rent paid by tenants. However, typically aproportion of tenants are also futurehomeowners, who with limited access to themortgage market have to increase savings inresponse to rising house prices in order to beable to purchase a house later. Hence, if

2 See Muellbauer and Lattimore (1995), for a clear expositionof this dual role.


households are divided into house owners(including landlords) and tenants, a positiverelation between consumption and house priceswould be expected for house owners and a closeto zero (slightly negative) relation for tenants.In addition, the wealth effect is the smaller theless liquid the housing property is. Accordingto Maclennan, Muellbauer and Stephens (1998)the perceived liquidity of housing assets variesa lot across Europe. This is due to, among otherthings, differences in transaction costs (e.g.stamp duties), the collateral role of the assetsand inheritance taxes, particularly whenhousing is sheltered from inheritance taxes,which can be translated into a restriction onresale. Under these circumstances, other thingsbeing equal, they conclude:

– The higher the proportion of owner-occupiers and the lower the proportion ofhouseholds in the rented market, the largerwill be the consumption response to a rise inhouse prices.

– The larger the proportion of rental housingwhich is owned by pension funds andsimilar institutional investors, the smallerwill be the consumption response to a rise inhouse prices. The reason is that (future)pensions are rather illiquid component ofconsumers’ portfolios. 3

2.2 SOME CONCLUSIONS

On the basis of the discussion in Section 2 thefollowing conclusions can be drawn:

– The long-run MPC out of different wealthcomponents should be quite stable overtime. The main determinants are the lengthof the planning horizon, the bequest motiveand a weighted average of the expected rateof return on wealth (over the planninghorizon) and the subjective discount rate.

– Risk aversion associated with uncertaintywidens the range of feasible values for theMPC out of wealth mpc

W. With the plausible

values of risk aversion (1/σ ≥ 1) mpcW

increases with the rate of return ra, but the

opposite is true, if risk aversion is low, i.e.0 ≤ 1/σ <1. Hence, low relative riskaversion allows for low values of mpc

W.

– With typical estimates of risk aversion(1 ≤ 1/σ ≤ 3), the MPC out of wealth shouldbe above the subjective discount rate.

– Income uncertainty implies that the MPCout of wealth of wealthy consumers may bemarkedly lower than that of consumers withlow wealth. For instance, if equity wealth isconcentrated in the upper end of incomedistribution, then at the aggregate level theMPC out of equity wealth may be lowerthan the MPC out of other wealthcomponents.

– The MPC out of wealth may differ acrosswealth components as a reflection ofdifferences in liquidity properties andbequest motives associated with them.These cross asset differences may bestrongly affected by developments incollateral markets and taxation.

– The transmission of asset price changes toconsumption is largely determined by howpermanent asset price changes are expectedto be. Changes that are expected to betransitory have less effect on consumption.This issue is linked to the predictability ofequity market returns.

– If the MPC out of wealth differs acrosswealth components, then the MPC out ofnon-equity financial wealth is likely to behighest. However, there are divergingviews regarding the relative sizes of theMPCs out of equity wealth and out ofhousing wealth.

3 This is in line with Poterba (2000), who argued that pensionfunds and houses are long-term “locked” assets, the MPC outof which is lower than out of “unlocked” assets.

13ECB


3 HOUSEHOLDWEALTH:

A STATISTICALREVIEW

3 HOUSEHOLD WEALTH: A STATISTICALREVIEW

To give statistical background to ourinternational comparison of estimated wealthand asset market effects on consumption wefirst make an international comparison ofhousehold wealth and wealth composition. Thetheoretical review in Section 2 revealed that thecomposition of wealth, its distribution acrossincome classes and age groups, anddemographics may have importance in theinternational comparison of wealth effects.Therefore, in sub-section 3.1, we first comparehousehold wealth and its decomposition intofinancial and housing wealth across countries.4

Then the decomposition of financial wealthacross counties is presented in sub-section 3.2.Since planning horizons and bequest motivescan vary across consumers and across wealthcomponents, sub-section 3.3 reports cross-country comparisons of wealth distributionsacross income classes and age groups. Specialattention is also paid to housing wealth andhouse prices and to equity wealth.

3.1 DECOMPOSITION OF HOUSEHOLD WEALTHINTO FINANCIAL AND HOUSING WEALTH

Table 3.1 reports the decomposition of totalhousehold wealth into its two maincomponents, i.e. financial and housing wealth,in 1995 and 2000 for six EU countries and the

United States.5 Each wealth component ispresented as a percentage of GDP and apercentage of total wealth. As can be seen, theratios to GDP of both total household wealthand its components vary considerably bothover time and across countries. In 2000 totalhousehold wealth was highest in Spain, at521% of GDP, and lowest in Germany, at 371%of GDP.

Regarding the composition of total householdwealth, financial wealth dominates in theNetherlands, the United Kingdom and theUnited States with financial wealth sharesabove 60%. In 2000 the financial assets to GDPratio was highest in the United States (341%),the United Kingdom and the Netherlands (botharound 300%). In continental Europe this ratiowas lower: somewhat above 200% in Franceand Italy and somewhat below 200% in Spainand Germany. In the period 1995-2000financial wealth rose as a percentage of GDP inall countries as a result of strong equity priceincreases. The increase in this ratio wasparticularly strong in France (69 percentagepoints) and lowest in Germany (31 percentagepoints).

Table 3.1 Total household wealth decomposed into f inancial and housing wealth

United UnitedFrance Germany Italy Spain Netherlands Kingdom States

1995 2000 1995 2000 1995 2000 1995 2000 1995 2000 1995 2000 1995 2000

Financial assets% of GDP 165 234 149 180 189 227 150 187 254 297 261 299 292 341% of wealth 50 53 44 49 45 51 33 36 69 62 64 61 71 73

Housing wealth% of GDP 170 206 191 191 234 220 298 334 112 182 146 191 120 130% of wealth 50 47 56 51 55 49 67 64 31 38 36 39 29 27

Total wealth% of GDP 335 440 340 371 423 447 448 521 366 479 407 490 412 471% of wealth 100 100 100 100 100 100 100 100 100 100 100 100 100 100

4 For a comprehensive presentation of EU housing markets, seeECB (2003).

5 The statistical source for f inancial assets is same as in Table3.2. Housing wealth is drawn from OECD sources except forSpain and the Netherlands, which is from Arnold et al. (2002).Therefore, the f igures in Table 3.1 are not fully comparableand should be only taken as indicative.


However, in spite of this development housingwealth continues to play an important role in allcountries but, especially, in France, Germany,Italy and Spain where the housing wealth shareis close to or above 50%. Reflecting cross-country differences in house pricedevelopments, the ratio of housing wealth toGDP and the share of housing in total wealthdecreased in Italy and Germany by about 5percentage points in 1995-2000. In all othercountries, real house prices rose, increasing theratio of housing wealth to GDP. In the UnitedKingdom house prices rose so much that theshare of housing in total household wealth alsoincreased. Hence, housing prices appear to bemarkedly less synchronised and their volatilityvaries more across countries than is the casewith stock prices.

3. 2 DECOMPOSITION OF HOUSEHOLDFINANCIAL WEALTH INTO COMPONENTS6

Over the past decade financial markets havemoved towards greater integration as well asproduct innovation. In Europe, the removal ofcapital controls, the privatisation of publicutilities and recent pension reforms are some ofthe policy interventions which havetransformed financial markets andconsequently affected the allocation ofhousehold wealth.

This can be seen in Table 3.2, which shows theallocation of the financial wealth of householdsamong its six main components in 1995 and2000 in the same countries as in Table 3.1.7

Pension funds are the major factor inexplaining the differences in the financialwealth to GDP ratio among the three countrieswith the highest ratio, i.e. the United States, theUnited Kingdom and the Netherlands, wherethe financing of pensions is mainly fund-based,and four countries with markedly lower ratios,i.e. France, Germany, Italy and Spain, wherethe financing of pensions is based mainly on“pay-as-you-go”.

The share of stock market related wealth(stocks plus mutual funds in Table 3.2) ishighest in the United States (46%) and it isalmost as high in Italy and Spain (about 43.5%in 2000). All in all, the increase in the share ofthis risky wealth component was notably fasterin France, Germany, Italy, and Spain than in theNetherlands, the United Kingdom and theUnited States during the later half of 1990s.

United UnitedFrance Germany Italy Spain Netherlands Kingdom States

1995 2000 1995 2000 1995 2000 1995 2000 1995 2000 1995 2000 1995 2000

Currency and deposits 39.1 30.4 41.9 36.2 40.1 24.1 51.8 36.2 22.5 18.1 24.2 22.2 14.1 11.1

Securities otherthan stocks 5.5 2.7 13.5 10.1 30.7 18.7 4.2 1.9 3.3 2.3 1.6 1.3 9.4 6.4

Stocks 15.8 24.8 10.9 15.6 14.5 27.6 19.0 33.7 15.4 16.8 15.7 17.4 32.0 33.1

Mutual funds 8.6 9.0 5.9 10.5 3.1 15.9 6.2 9.9 3.6 5.5 3.8 5.5 9.6 12.9

Life insurance 17.9 23.3 11.8 13.6 3.3 6.2 4.5 6.2 15.3 15.4 27.7 27.5 7.1 7.1

Pension funds 1.8 1.5 5.6 5.2 0.8 1.2 4.5 5.3 35.5 37.7 22.3 22.1 23.4 23.8

Financial assetsas % of GDP 165 234 149 180 189 227 n.a. 187 n.a. 297 261 299 292 341

Table 3.2 Composit ion of f inancial assets of households and total household wealth

(as % of financial assets, unless otherwise indicated)

6 This section is mainly based on Babeau and Sbano (2002) andGuiso et al. (2002).

7 The “household” in Table 3.2 includes non-prof it institutionsserving households (e.g. churches, consumer associations,labour unions) and the self-employed.

15ECB


3 HOUSEHOLDWEALTH:

A STATISTICALREVIEW

The shift towards riskier assets is divideddifferently across countries between directholdings of equities and investments in mutualfunds. The portion of wealth held in mutualfunds is rising everywhere, but especially inItaly and Germany. The tendency to takegreater risks is also apparent in the proportionof equity funds within investment funds as awhole: they have increased their share in five ofthe six countries.

By contrast, euro area countries haveexperienced a generalised decline in bankingintermediation. This trend is particularlynoticeable in Spain, Italy and France and to alesser extent in Germany. The decline has lessof an impact on sight/overnight deposits thanon term and savings deposits. There has alsobeen a sharp decline in the portion of interest-bearing securities (bonds and loans), mainlydue to the decline in long-term interest ratesand efforts to improve public finances. Thistrend is particularly visible in Italy, where theshare of such investments was above 30% in theearly 1990s (the highest ratio in the euro area)and fell to 18.7% in 2000.

Household claims on insurance companiesgrew over the period considered in France and,to lesser extent, in Germany, Italy and Spain. In

Chart 3.1 Ratios of gross f inancial wealth tonominal consumption

0

1

2

3

4

5

6

7

0

1

2

3

4

5

6

7

Ratio

NL BG US UK JP FR IT PT SP GE

the United Kingdom and the Netherlands,where life insurance assets were already veryhigh at the beginning of the period, theyremained roughly stable. For saving purposes,life insurance policies are often perceived to below risk. However this changed recently withthe spectacular growth of unit-linked policiesin Italy, France and Spain. Since the late 1980ssuch policies have already played an importantrole in the United Kingdom and theNetherlands. Only in Germany was there noevidence of this move towards riskier assets.

Available financial wealth data indicatesurprisingly large variations across countriesin the ratio of wealth to consumption. Based onnational sources, Chart 3.1 shows that the ratioof total wealth to consumption is highest in theNetherlands, at around twice the level inGermany, where this ratio is lowest.8 If theMPC out of wealth were equal across countries,then relation (2.2) would imply that the wealthelasticity of consumption in the Netherlandswould also be around twice that of Germany.Although the role of pension funds explains alarge part of these cross-country differences inthe ratio of financial wealth to consumption,remaining differences are still important anddifficult to explain. It may be that, rather thantrue differences, they reflect different methodsof measuring wealth and, therefore, problemsof data comparability. We return this issue inthe sub-section 4.4.

3.3 EQUITY WEALTH EFFECT, INCOMEDISTRIBUTION AND DEMOGRAPHICS

In Section 2 we showed that the distribution ofwealth may affect aggregate MPCs out ofdifferent wealth components. Coupled withdemographic factors, this may help explaincross-country differences in wealth effects. Inthis section we study the distribution of wealth,especially equity wealth, across incomeearners and age cohorts in different countries.

8 Financial wealth data in Tables 3.1 and 3.2 are not fullycompatible with data used in Chart 3.1.


Table 3.3 reports recent equity holdings acrossdifferent income brackets in Germany, France,Italy, the United Kingdom and the UnitedStates. Table 3.4 presents the distribution ofequity ownership across different age groups inthe same countries.

The last column of both tables reveals thatshare ownership is much more concentrated inthe three euro area countries than in the UnitedStates and the United Kingdom. While in theUnited States around 20% of the populationown shares, this proportion falls to around 8%in Italy (in 1998), 10% in Germany (in 2000)and 13% in France (in 2000).9 In the UnitedKingdom, more than 20% of the populationowns shares.

France AllIncome group (EUR) 0-1500 1500-2300 2300-3050 3050-3800 3800-Income group by percentile 0-32 32-65 65-83 83-91 91-100Percentage owning equity:

in 1997 6.1 10.1 15.5 19.1 32.6 12.0in 2000 7.4 11.2 14.3 21.1 31.4 12.7

GermanyIncome group (EUR) 0-1300 1300-2050 2050-3050 3050-4100 4100-Income group by percentile 0-21 21-51 51-84 84-93 93-100Percentage owning equity:

in 1997 1.7 3.9 8.3 14.6 18.7 6.2in 2000 3.0 5.8 11.4 20.4 25.9 9.8

ItalyIncome group (EUR) 0-850 850-1700 1700-2600 2600-3450 3450-Income group by percentile 0-18 18-51 51-74 74-87 87-100Percentage owning equity:

in 1995 0.2 2.0 5.0 10.3 21.5 5.0in 1998 0.6 2.4 5.7 11.9 31.7 7.8

United KingdomIncome group by percentile 0-25 25-50 50-75 75-100Percentage owning equity:

in 1993 8.2 14.8 27.0 41.3 22.8in 1996 13.4 15.6 26.5 37.9 23.3

United StatesIncome group (USD) 0-850 850-2100 2100-4150 4150-8350 8350-Income group by percentile 0-13 13-37 37-66 66-91 91-100Percentage owning equity:

in 1995 2.3 8.4 13.9 24.7 43.6 15.2in 1998 3.8 7.2 17.7 27.7 56.6 19.2

Sources: “Equity wealth and consumption – the experience of Germany, France and Italy in an international context”, Bank ofEngland Quarterly Bulletin, Spring 2002, p. 81. The original sources are provided there.Notes: The table shows the proportion of each income group holding direct equities excluding mutual funds. Numbers are not fullycomparable across countries due to different def initions of households.

Table 3.3 Percentage of households with direct equity holdings, by income group

Common features across countries are that (1)the percentage of households holding equitieshas increased and (2) the share of equity ownersin each income bracket is positively related tothe income level. In Table 3.3 the proportion ofhouseholds holding shares in the highestincome bracket is between 2 and 4 times largerthan the average proportion in the population.The proportion of share ownership in highincome brackets is larger in the United States

9 Figures reported here are not fully comparable acrosscountries because of differences in the def inition ofhouseholds. In particular, the figures for Germany mayunderestimate the percentage of households holding equitiesas they refer to the population above 14 years of age, while forItaly and the United States a household is def ined as aneconomic unit that includes more than one person on average.

17ECB


3 HOUSEHOLDWEALTH:

A STATISTICALREVIEW

and the United Kingdom than in the three euroarea countries, although in this respect thedifferences between countries appear lessmarked than for the population averages. Inaddition, Table 3.3 shows that the percentageof middle-to-low income earners holdingshares is larger in the United States and theUnited Kingdom than in the euro areacountries. If coupled with the higher MPC outof wealth of the poor relative to the wealthy,this may justify, at least qualitatively, a higherMPC out of equity wealth in the UnitedKingdom and the United States than inGermany, France and Italy.

France AllAge group 15-24 25-34 35-44 45-54 55-64 65-Percentage of population 13.0 14.5 14.6 13.9 9.2 15.9Percentage owning equity:

in 1997 3.7 7.9 10.2 14.6 17.9 19.2 12.0in 2000 3.3 9.1 11.6 17.6 16.9 18.3 12.7

GermanyAge group 14-19 20-29 30-39 40-49 50-59 60-Percentage of population 6.7 11.9 17.2 14.5 12.2 23.0Percentage owning equity:

in 1997 1.0 4.4 7.3 8.7 8.7 4.9 6.2in 2000 2.4 7.6 13.5 11.7 13.1 7.2 9.2

ItalyAge group 15-30 31-40 41-50 51-65 65-Percentage of population 22.2 15.5 13.5 17.9 16.3Percentage owning equity:

in 1995 2.1 5.7 5.6 6.6 2.9 5.0in 1998 3.9 9.8 7.9 9.5 5.5 7.8

United KingdomAge group 25-34 35-49 50-64 65-Percentage of population 16.0 20.7 15.6 15.7Percentage owning equity:

in 1993 12.8 24.1 30.4 24.7 n.a.in 1996 12.9 21.6 30.2 31.2 n.a.

United StatesAge group 0-34 35-44 45-54 55-64 65-74 75-Percentage of population 23.3 23.3 19.2 12.8 11.2 10.2Percentage owning equity:

in 1995 10.8 14.6 17.7 15.0 18.6 19.7 15.2in 1998 13.1 18.9 22.6 25.0 21.0 18.0 19.2

Table 3.4 Percentage of households with direct equity holdings, by age group

Sources: “Equity wealth and consumption – the experience of Germany, France and Italy in an international context”, Bank ofEngland Quarterly Bulletin, Spring 2002, p. 81. The original sources are provided there.Notes: The table shows the proportion of each age group holding equities directly, excluding mutual funds. Numbers are not fullycomparable across countries due to different def initions of households.

Table 3.4 shows that equity ownership is moreevenly distributed across age groups thanacross income brackets, except for theyoungest age group. However, a positiverelation between equity holdings and age isapparent, especially, in the United Kingdomand the United States. In Germany and Italythis relation turns to negative among retiredage groups. Therefore, if the MPC out of wealthis higher for older generations, reflecting ashorter planning horizon, then this may alsojustify a higher MPC out of equity wealth in theUnited States and the United Kingdom than inGermany, France and Italy.


4 ASSET PRICE EFFECTS VIA THE WEALTHCHANNEL: EMPIRICAL EVIDENCE

This section reviews cross-country empiricalevidence of wealth effects on consumption.Sub-section 4.1 presents evidence based onmicro-data studies, which are most closelylinked to data characteristics examined inSection 3. The merit of the micro-dataapproach is its ability to separate direct andindirect wealth effects. Sub-section 4.2presents MPCs out of wealth in theory-basedcalibrated multi-country models. Sub-section4.3.1 discusses MPC out of wealth estimatesbased on time-series estimation of the deepparameters of theory-based consumptionfunctions. In sub-section 4.3.2 we reviewresults based on direct (unconstrained)aggregate time-series estimations. Althoughthe available evidence mostly suggest animportant wealth channel, our generalconclusion is that the variation of MPC out ofwealth estimates across studies, acrosscountries and across estimation methods isdisconcertingly large. Such differences intypical MPC out of wealth estimates acrosscountries are implausible in the light ofeconomic theory. In sub-section 4.4, after acritical assessment of the wealth data used incross-country comparisons, we propose thatincomparable measures of wealth acrosscountries and across studies may be animportant source of differences in MPC out ofwealth estimates. To reduce these problems inwealth data, sub-section 4.5 presents somerecent estimation results based on dynamicpanel approach. Sub-section 4.6 concludes.

4.1 MICRO DATA EVIDENCE

This section briefly surveys the evidence frommicro-level studies. Unfortunately most ofthese studies are limited to US data. Recentexceptions are studies by Paiella (2003) andGuiso, Paiella and Visco (2004) using Italiandata. In addition, all these studies use datasetswith limited coverage of consumption and/orasset holdings. Thus, key parameter estimatesmay not always be directly comparable to their

macroeconomic counterparts. However,micro-level studies provide a useful check onwhether theories of the wealth effect areconsistent with individual behaviour.

There are a large number of micro-data studiesof various aspects of the wealth effect and, as isoften the case with such literature, thesestudies have frequently come to differentconclusions. However, a reasonableinterpretation of these studies is that they tendto back up the macroeconomic evidence on therole of the wealth effect.10

A number of studies of have analysed therelationship between equity wealth andconsumption within the consumption-basedcapital asset pricing model (C-CAPM). Forexample Mankiw and Zeldes (1990), Attanasio,Banks and Tanner (1998), Vissing-Jørgensen,(1999), and Brav, Constantinides and Geczy(1999) find that the spending of stockholders ismore highly correlated with stock marketreturns than that of non-stockholders, whichsupports a direct effect. Less positively, Parker(1999b) and Juster, Lupton, Smith and Stafford(1999) both find that spending responds towealth at the household level, but neither papercan pinpoint responses within the time framenecessary to explain the macro relationship.

A comprehensive study by Maki and Palumbo(2001) provides important evidence in favourof the wealth effect on US consumer spendingduring the 1990s. Using data from the Surveyof Consumer Finances, they find that thegroups of families whose portfolios wereboosted most by the exceptional stock marketperformance over the latter half of the 1990sare the same groups whose net saving flowsdeclined most sharply between 1995 and 2000.

10 Of course, one can f ind some evidence to the contrary. Forexample, Starr-McCluer’s (2000) analysis of qualitativeevidence from the University of Michigan’s SRC Survey ofConsumers suggests that the spending of stockholders is onlymodestly affected by changes in wealth, and Otto (1999) f indsthat the correlation between stock prices and consumersentiment does not vary by stock ownership.

19ECB


4 ASSET PRICEEFFECTS VIATHE WEALTH

CHANNEL:EMPIRICALEVIDENCE

Families who owned relatively modestamounts of equity wealth – the vast majority ofUS households – experienced relatively smallgains in net worth to income ratios over the1990s and continued to save at about the samesteady pace over the decade. In addition, Makiand Palumbo’s study verifies that essentiallyall of the increased spending apparent in theaggregate data can be attributed to an increasein the MPC out of income of the richest UShouseholds. Finally, they present neweconometric estimates of the MPC out ofhousehold net worth based on cohort-leveltime-series data that fall in the same range – 3to 5 cents per dollar – as typicalmacroeconomic estimates (although somewhatlower than the preferred estimates in Table 2above). These econometric results bolster theconclusion that the direct wealth effect onhousehold spending is of the right size toexplain the sharp decline in the aggregatesaving rate observed in the 1990s.

These results are also compatible with those ofDynan and Maki (2001) who use data from theConsumer Expenditure Survey and imputedvalues of equity holdings to estimate the effectsof stock market wealth on consumption. Theyfind that direct wealth effects begin to show uprelatively quickly and continue to boostconsumption growth for a number of quarters,which is in line with aggregate estimates. Incontrast to many earlier studies, they find thatthe indirect wealth channel is not an importantdeterminant of consumption growth. This studyalso reports estimates of the MPC out of stockmarket wealth of between 5 and 15 cents perdollar, but notes that its sample excludes high-income households likely to have lower MPCs.Indeed, both Parker (1999a) and Dynan et al.(2000) show that the average propensity toconsume declines with permanent income.11

Evidence found by Paiella (2003) and,especially, by Guiso, Paiella and Visco (2004)for Italy, resemble in many respects the USevidence described above, although theestimated size of the wealth effect is smallerthan typical US estimates. The estimates of

MPC were around 2 cents per euro out of totalwealth and out of housing wealth, respectively,and slightly below 4 cents per euro out of netfinancial wealth. However, when dividing theirdata into homeowners and tenants Guiso,Paiella and Visco (2004) found that the MPCout of housing wealth (excluding currentperiod capital gains on real estate) and the MPCout of capital gains on real estate were 2 and 3.5cents per euro, respectively. For tenants theydid not find any statistically significantrelation between consumption and capitalgains on real estate (or house prices), but theestimated MPC out of net wealth was in therange of 3.7 to 6.2 cents per euro. All theseresults support direct effects from wealth andits components on consumption. Regardingequity wealth effects on consumption, bothstudies found hardly any evidence for the MPCout of equity wealth. However, it is quite likelythat this result contains a downward bias,because households at the top of the incomeand wealth distribution were underrepresentedin the survey data.

4.2 THE MPC OUT OF WEALTH IN SOMECALIBRATED THEORY-BASED MODELS

In theoretically-founded models, such as theIMF’s MULTIMOD, deep structuralparameters such as the rates of return and theplanning horizon are calibrated. The calibratedMPCs out of wealth of MULTIMOD provide atheory-consistent guide to reasonablebenchmark for the MPC out of wealth to beexpected from empirical research.

Chart 4.1 shows the MPC out of wealth impliedby MULTIMOD. These MPCs out of wealth arebased on the Mark III version, as this providesseparate MPCs out of wealth for the countriesof the euro area. These MPCs out of wealthrange between 5.4 and 8.2 per cent. The highestvalues are those for the United States and

11 It is worth noting, however, that while the average propensityto consume depends on wealth, this does not imply that thesame must be true for the MPC out of wealth.


Canada, followed closely by those for theUnited Kingdom and Japan; the values for theeuro area economies are somewhat below 6.Overall, the MPC out of wealth tends to bequite similar across countries. In fact, three ofthe four parameters determining the MPC outof wealth in MULTIMOD – the intertemporalelasticity of substitution in consumption, thereal interest rate, the probability of death – arethe same across countries in the long run, as aresult of calibration or pooled estimation (seeFaruquee, Isard, Laxton, Prasad andTurtelboom (1998)).

4.3 AGGREGATE TIME SERIES ESTIMATES OFMPC OUT OF WEALTH:

The following sub-section reviews estimatesbased on time-series estimation of the deepparameters of theory-based consumptionfunctions. Sub-section 4.3.2 examines theevidence based on aggregate time seriesestimation without theory-based parameterconstraints.

4.3.1 MPC OUT OF WEALTH BASED ON THEESTIMATED DEEP PARAMETERS OF THECONSUMPTION FUNCTION

Not many studies estimate the consumptionfunction explicitly in terms of deep parameters.Some exceptions are Hayashi (1982b), Darbyet al. (1999), Sefton and in’t Veld (1999),

Chart 4.1 Theoretical ly cal ibrated MPC outof wealth in MULTIMOD mark I I I

0

2

4

6

8

10

0

2

4

6

8

10

US Canada Japan UK France Germany Italy

Fuhrer (2000) and Willman (2003). Throughquasi-differing human wealth in the stochasticpermanent income hypothesis consumptionfunction, Hayashi (1982b) estimates thesolved-out consumption function (1982) for theUnited States and Darby et al. (1999) for theUnited Kingdom. Sefton and in’t Veld (1999)apply the Blanchard (1985) overlappinggenerations model with positive probability ofdeath. Willman (2003), in the framework of theoverlapping generation model, also account forhabit formation. Hayashi (1982b) and Darby etal. (1999) directly estimate the MPC out ofwealth. In Sefton and in’t Veld (1999) and inWillman (2003) the estimated length ofplanning horizon (probability of death) is thecrucial determinant of the MPC out of wealth,as shown by equation (2.6).

Hayashi’s (1982b) estimates of MPC out ofwealth were in the theoretically reasonablerange of 3.3 to 8.6 with the preferred pointestimate at 6.8 cents per dollar, which is veryclose to the Darby et al. (1999) point estimateof 7.1 pence per pound for the UK. In Seftonand in’t Veld (1999) estimates of MPC out ofwealth were in the range of 3 to 16 cents perdollar for the United States, 9 cents per dollarfor Canada, 6 to 7 pence per pound for theUnited Kingdom and 4 to 9 pfennigs perDeutsche Mark for Germany. Willman (2003)estimated the consumption function usingaggregated quarterly euro area data and foundan MPC out of wealth in the range of 7-10 centsper euro for the planning horizon in the range of12-32 years. Estimates for the habit persistenceparameter defining the adjustment speed fromshocks to consumption was in the range of 0.7to 0.8. Apart from Sefton and in’t Veld’soutlying estimates in the upper limit of therange, the ranges of these estimates were quitereasonable and broadly in line with thosecalibrated for MULTIMOD. Further, theseresults indicate no clear evidence for the MPCout of wealth being lower in continental Europethan in the English-speaking countries.

21ECB




Euro area Germany United Kingdom United States Canada

Hayashi(1982b) 3.3-8.6Darby et al.(1999) 7.1Sefton and in’t Veld (1999) 4-9 6-7 3-16 9Willman(2003) 7-10

Table 4.1 MPC out of wealth est imates based on the solved-out consumption functions

4.3.2 AGGREGATE TIME SERIES ESTIMATE OFMPC OUT OF WEALTH: DIRECT DATA-BASEDAPPROACH

This sub-section reviews the availableevidence on the long-long run effects of wealthand its components on consumption. First wereview some recent comparative internationalstudies on the long-run effects of wealth and itscomponents on consumption. Then we presentestimation results based on country-specificstudies.

4.3.2.1 Evidence based on cross-country studiesInternational comparisons include: Ludwigand Sløk (2002), Bertaut (2002), Boone et al.(1998), Labhard et al. (2005) Case et al. (2001)and Edison and Sløk (2001).

Ludwig and Sløk (2002) use a panel data of 16OECD countries, where countries areclassified (on ad hoc bases) as market-basedand bank-based economies. As shown in Table4.2, pooled results for market-based economies(Australia, Canada, the Netherlands, Ireland,Sweden, the United Kingdom and the UnitedStates) imply that a 10% increase in equityprices increases long-run consumption bynearly 1%, while in bank-based economies(Finland, France, Germany, Italy, Japan,Norway and Spain) the estimated elasticity ismuch smaller at 0.4%.

Ludwig and Sløk provide MPCs out of equitywealth for a subset of countries. ReportedMPCs are in a range of between 1.4 and 4.9,

Table 4.2 Est imates of MPC and elast ic it ies out of wealth components

BE FR DE IT NL PT SP SE UK US CA JP

Ludwig and Sløk (2002)mpc

W1.4 2.0 3.0 - - 4.9 4.0 4.0 4.0

eW-H

0.11 0.11 0.11 0.03 0.03 0.03 0.03 0.03 0.11e

W-EQ0.04 0.04 0.04 0.10 0.10 0.10 0.10 0.10 0.04

Bertaut (2002)mpc

W- - - - - 4.3 5.4 8.3 -

mpcW-F

2.7 - - 4.2 5.9 9.7 10.6mpc

W-H- - - 9.7 - -

mpcW-EQ

- - - 6.2 8.7 -e

W- - - - - 0.20 0.29 0.41 -

eW-F

0.10 - - - - 0.09 0.23 0.16 0.29e

W-H- - - - - 0.09 0.14 0.16 -

eW-EQ

- - - - - - 0.10 0.14 -(e

EQ-Price) (0.02) (0.01) (0.02) (0.03) (0.03) (0.04) (0.06) (0.06) (0.02)

Boone et al. (1998)(e

EQ-Price) (0.01) (0.02) (0.01) (0.05) (0.06) (0.02) (0.02)

Labhard, Sterne andYoung1) (2005)mpc

W0.7 0.8 7.8 2.8 1.3 -1.0 3.6 5.6 3.7 7.8 4.2

eW

0.03 0.10 0.13 0.08 0.06 -0.02 0.07 0.16 0.12 0.19 0.16

Legend: mpcW-j

and elasticity, ew—j,

out of the wealth component j, with j= equities (EQ), f inancial wealth (F), non-equities (N-EQ)and housing (H).1) Figures are averages of the estimates given by Ludvigson, Steindel and Lettau (2002) using 5 and 6-variable VARs in levels and indifferences.


broadly consistent with the lifetime budgetconstraint discussed in Section 2. AmongstEuropean countries, they are lowest forGermany (2.0) and France (1.4) and highest forthe United Kingdom (4.9) and Italy (3.0). Thereare few indications as to why the MPCs varyacross countries. This appears to be a result ofthe pooled estimation techniques, which fixesthe elasticity, instead of the MPC out of wealthcomponents mpc

W-j, to be equal within groups.

Therefore the common elasticity estimated forFrance and Italy, for instance, inevitablyimplies an MPC out of equity wealth mpc

W-EQ in

France that is half that in Italy, since the stockmarket capitalisation ratio in France is twicethat in Italy. Ludwig and Sløk’s results maytherefore be more useful for comparing MPCsacross groups than across countries within thegroups. Ludwig and Sløk find clear evidencethat the impact of changes in stock prices onconsumption is bigger in economies withmarket-based financial systems than ineconomies with bank-based financial systems.This impact from stock markets to consumptionhas increased over time for both groups ofcountries.

The effect of housing prices on consumptionhas also become more important over time. Theauthors consider this change to be a strikingevidence for relevant changes in the mortgagemarkets. In the sample period 1985-2000 theeffect of housing prices is significantlypositive and the estimated elasticity is abouttwice as large as the elasticity of stock marketprices for the combined sample of all countriesand for the group of market-based economies.Ludwig and Sløk do not provide an impliedMPC out of housing wealth mpc

W-H. Utilising

the information contained in Table 3.1 andChart 4.1, the corresponding MPC out ofhousing wealth would be around 7.5 cents pereuro for Germany and around 5 cents per eurofor France and Italy. In market-basedeconomies these estimates would be smaller,i.e. 1.3 cents per euro for the Netherlands, 1.7pence per pound for the United Kingdom and 2cents per dollar for the United States. As wellas grouping countries according to financial

structure Ludwig and Sløk also group their dataaccording to home-ownership rates. They findthat an increase in the home-ownership rateincreases the probability that a positive wealtheffect will outweigh negative income andsubstitution effects on consumption. Thisresult supports our discussion in sub-section4.3.2.2.

Bertaut (2002) estimates individual countryregressions for various forms of wealth. Heused both stock price data and disaggregatedhousehold wealth data, to the extent available.Accordingly, the estimations based on properwealth data cover fewer countries, i.e. France,the United Kingdom, the United States, Canadaand Japan. The MPC out of wealth estimates forwealth and its components are roughly in linewith those of Ludwig and Sløk (2001), beingslightly lower for the United Kingdom (4.3pence per pound), but higher for the UnitedStates (5.4 cents per dollar), France (2.7 centsper euro) and, especially, Canada (8.3 cents perdollar) and Japan (10.6 yen per 100 yen). TheUnited States was the only country for which itwas possible to disaggregate wealth into equityand housing components. The estimated MPCout of housing wealth was markedly higher (9.7cents per dollar) than that estimated by Ludwigand Sløk (2001) and also higher than that out ofequity wealth (6.2 cents per dollar). Reduced-form equations using equity prices gavemarkedly lower elasticities for equity wealththan those of Ludwig and Sløk (2001) (seefigures in brackets in Table 4.2). Even in theUnited States and Canada, where adequatedisaggregated wealth data were available,these reduced-form estimates imply markedlysmaller impacts of changes in equity wealth onconsumption than corresponding estimatesbased on wealth data. This evidence suggeststhat estimates based on stock price data containa downward bias.

Boone et al. (1998) estimates long-runconsumption function for seven major OECDcountries (the United States, Japan, Canada,Germany, France, Italy and the UnitedKingdom). Due to the lack of adequate wealth

23ECB




data they also use only equity price and housingprice data in estimating consumption responsesto changes in asset values. These estimationresults are well in line with correspondingestimation results presented by Bertaut (2002).Accordingly, they may contain a downwardbias.

Labhard, Sterne and Young (2005) presentsestimates of the MPC out of wealth usingestimates of the elasticity of total consumptionwith respect to net financial wealth obtainedfrom simple s tructural VAR (SVAR) models,including the VAR of Ludvigson, Steindel andLettau (2002, LSL in chart 4.2). As Table 4.2and Charts 4.2 show, their estimates suggestthat the MPC out of wealth is between 1 and 4cents per euro for most European countriesunder review, but with markedly higher valuesfor Germany (around 8 cents per euro onaverage) and negative values for Portugal.12 AsChart 4.2 shows, the relative size of thesecross-country estimates remains broadlysimilar across different specifications of theVARs.13 Compared to other studies, theresponse in Germany seems rather strong.However, the elasticities for equity wealth willbe significantly modified when the results arere-scaled for the relative holdings of equitywealth across countries. Estimated MPCs forthe United Kingdom, the United States, Canada

Chart 4.2 SVAR-based est imates of the long-run MPC out of net f inancial wealth

and Japan are higher than in Europe being in therange of 4 to 8 cents per dollar (or pence perpound, or yen per 100 yen).

Case et al. (2001) use extensive panel data toinvestigate the link between consumption andincreases in housing and equity wealth. Theyrely on two bodies of data: (1) a panel of annualobservations covering 14 countries over thelast 25 years and including aggregateconsumption, stock market capitalisation, andaggregate housing wealth and (2) an analogouspanel of quarterly observations of US statesincluding consumption, stock ownership, andaggregate housing wealth. Their main findingin both data sets is that the housing marketseems to be more important than the stockmarket in influencing consumption. The factthat Case et al. (2001) define their variables inper-capita terms might help to explain whystock market wealth is not significant.Although the stock market had played animportant role in explaining aggregateconsumption, it may not have been assignificant in explaining the behaviour of anaverage consumer, as most consumers do nothold equities and, hence, stock owners areunderrepresented in the sample. Even thoughequity ownership has become morewidespread, the distribution of equity holdingsis still fairly skewed (see Table 3.2.3), whilehousing wealth is much more evenlydistributed across households.

The estimated elasticity from housing wealth toconsumption is 0.11 to 0.14 in Europe and0.062 in the United States, when the panel dataof the US states are used. By using theinformation contained in Tables 3.1 and Chart4.1, these elasticities imply that the MPCs outof housing wealth would be 7.5 to 9.5 cents pereuro for Germany, 5 to 7 cents per euro for

12 The high estimates for Germany may be due to the structuralbreaks in the series associated with unif ication.

13 The variables included are aggregate consumption, income,inflation, net f inancial wealth, interest rates, and, in the caseof the 6-variable VAR, commodity prices. Restrictions areimposed on the short-run responses of variables. In Chart 4.2we refer to the SVARs as “LSL”, which comes from the namesof the authors who originally used the restrictions in theseSVARs (Ludvigson, Steindel and Lettau, 2002).

-4

0

4

8

12

16

20

-4

0

4

8

12

16

20

LSL 5 variablesLSL 6 variablesLevels VARDiffs VAR

USCanada

JapanUK

FranceGermany

ItalySpain

PortugalBelgium

Netherland


France and Italy and around 4 cents per dollarfor the United States.

Edison and Sløk (2001) investigate stock priceimpacts on consumption with a special focus onthe different effect of changes intelecommunications, media and informationtechnology (TMT) and non-TMT stocks.Estimating a reduced-form VAR for sevenOECD countries for 1990s, the empiricalanalysis suggest that both TMT and non-TMTstocks may have a significant impact onconsumption in Canada, the United Kingdom,and Japan. In the United States the effect onconsumption of a change in non-TMT marketcapitalisation was larger than for the TMTsector, while in continental Europe the impactof TMT equities on consumption was clearlydominant.

4.3.2.2 Evidence based on single-country studiesThis sub-section reviews single-countrystudies. We start with the US evidence becausethe bulk of empirical work has been done withUS data.

The US evidenceThe role of wealth effects on consumption hasbeen most extensively studied for the UnitedStates, and this field of research has grownsignificantly in recent years. Commonly cited

TMT Non-TMT

Canada 7.3 7.2

United States 1.7 5.1

United Kingdom 2.9 3.4

North America andUnited Kingdom average 4.0 5.2

France 4.1 1.9

Germany 2.8 0.1

Netherlands 4.5 1.1

Continental Europe average 3.8 1.0

Japan 6.4 13.1

Table 4.3 MPC out of TMT and non-TMTequit ies

estimates of the MPC out of wealth in theUnited States are typically in the range of 2 to 5cents increased spending from a dollar increasein wealth.14 Among other estimates using time-series consumption, income, and equity wealthdata, Reifschneider, Tetlow, and Williams(1999) note that in the Board of Governor’sFRB-US Model, wealth effects from higherstock prices generate on average about 3½cents additional consumer spending in the longrun. As shown in Table 4.4, the results reportedby Ludvigson and Steindel (1999) and Mehra(2001) are in line with earlier estimates.However, the estimates of Davies and Palumbo(2001) and, especially, Palumbo et al. (2002)are somehow higher.

According to these studies the MPC out of totalwealth and its components are markedly higherthan conventional estimates if the incomevariable in the estimated equation is measuredin terms of labour income instead of disposableincome. In addition, Palumbo et al. (2002)remarked that the conventional practice ofdeflating income and wealth by a price indexfor total consumption expenditures, but using a

mpcW-EQ

mpcW-NEQ

mpcW

Ludvigson & SteindalDisposable income:– non-durables 3.8 4.0 4.6MehraLabour income– non-durables 3 3 3– total consumption 4 5 4Davis & PalumboLabour income 5.7 8.0 6.3Disposable income - - 3.9Palumbo et al.Labour income:– scaled non-durables 6.0 11.3 6.9– total consumption 5.4 10.2 6.1Disposable income:– total consumption - - 4.0

Table 4.4 MPC out of equity wealth (W-EQ),non-equity wealth (W-NEQ) and totalwealth (W)

14 In his February 2000 Monetary Policy Report testimony,Federal Reserve Board of Governors Chairman AlanGreenspan noted that “historical evidence suggests thatperhaps three to four cents of every additional dollar of stockmarket wealth eventually is reflected in increased consumerpurchases.”

25ECB




scaled version of real consumption of non-durables and services, is inconsistent. A moredetailed discussion is presented in theAppendix of this work. Correcting thisincompatibility, which also improved the co-integration properties of the estimated long-runequation, resulted in markedly higher estimatesfor the MPC out of equity wealth (6.0), non-equity net worth (11.3) and total wealth (6.9).

However, as discussed in the previous sub-section, Lettau and Ludvigson (2002) indicatedthat if much of the variation in wealth istransitory and uncorrelated with consumerspending, then the MPC out of wealth mpc

W is

not a good summary statistic for the wealtheffects on consumption. They estimated that asmuch as 88% of the variance in wealth istransitory. If applied to their mpc

W value of 4.6,

the implied corrected summary statistic mpccorr

,as shown by equation (2.9), would be 1.2.However, Palumbo et al. (2002) challenged thisevidence. They argue that the MPC out ofwealth should vary, reflecting variations in theexpected rate of return. Recursive estimationsover the past thirty years indicate that this hasnot been the case. Hence the size of thetransitory (mean-reverting) component inwealth changes remains controversial.

The European evidenceTable 4.5 summarises the estimation results ofindividual country studies as well as providingsome information on methodology and dataissues. Depending on the estimation approach,long-run coefficients are reported for differentwealth components. For Germany, according tothe Deutsche Bundsbank’s macroeconometricmulti-country model (BbkM), the long-run netfinancial wealth elasticity is 0.14. The wealthelasticity reported by the Banque de Francewas half of that, at 0.07. Sveriges Riksbankreports an elasticity for non-human net wealthof 0.12 (in real terms) in an unpublished study.This estimate is compatible with the Claphamet al. (2002) estimates for Sweden, which are inthe range of 0.06 to 0.15, and for Finland, in therange of 0.07 to 0.014. For Spain, Willman andEstrada (2002) estimated a wealth elasticity of

0.26, which is broadly in line with Balmasedaand Tello (2002), with wealth disaggregatedinto equity, non-equity financial and housingwealth. However, a more recent estimate ofwealth elasticity by Estrada et al. (2004) wasmuch lower at 0.04.17 For the UnitedKingdom, the Bank of England reports anelasticity of 0.06 (single equation response)and 0.11 (full model response) for net financialwealth. Belgium estimates a markedly lowerfinancial wealth elasticity of 0.02 to 0.04.

According to an unpublished analysis, thelong-run elasticity of equity wealth is 0.02 forFinland, 0.06 for Spain and zero for Sweden,where equity wealth is limited to market valuesof listed shares for Sweden, Spain and Finland.

Reported MPCs out of equity wealth and non-equity financial wealth were 2 cents per euroand 7 to 9 cents per euro respectively forGermany. Reported (or implied bycorresponding elasticity estimates) MPCs outof total physical capital and financial wealthwere 3 cents per euro for Portugal (in 1997 ), 2to 4 cents per euro for Finland, 4 to 5 öre perkrona for Sweden and 2 pence per pound for theUK (in 2001). The UK disaggregatedestimations for the MPCs out of net financialwealth and housing wealth are 4 pence and 2pence per pound, respectively (March 2002figures). Italy reported an MPC out of totalwealth in the range of 1.5 to 2.0. Also in theBanca d’Italia Quarterly Model, where theMPC out of total wealth depends on the realinterest rate, a value of 1.5 is obtained for a realinterest rate of 2%. With these estimates thereported MPC was consistently somewhathigher out of financial wealth (2.7) andsomewhat lower out of housing wealth (1.4).

15 Estrada et al. (2004), in def ining nominal total wealth, applythe Madrid stock index to the volume of total productivecapital, while Willman and Estrada (2000) def ine this wealthcomponent at repurchasing prices. The former practice resultsin a markedly wider cyclical variation in the wealth-to-consumption ratio and, hence, in a lower elasticity estimatethan latter practice. However, only a proportion of companiesare quoted in the stock market and only some of the companiesare owned directly by households. Hence, the former practicetends to result in a downwardly biased and the latter practicean upwardly biased elasticity estimate.


4.4 OVERVIEW OF RESULTS

Some finding based on micro-data evidencesupport the existence of a direct wealth effecton consumption. For instance, Maki andPalumbo (2001) find that essentially all of theincreased spending observed in the aggregatedata can be attributed to an increase in the MPC

out of income of the richest households.However, there is no consensus regarding thesize of the MPC out of wealth and itscomponents. The range of MPC out of wealthestimates across countries and often also acrossstudies within the same country is wider thantheory would suggest on the bases of cross-country differences in wealth composition and

Greece Finland1) Belgium France Germany 7) Italy

Estimation period 61-97 79-98 81-01 78-01 80-03 80-01Data frequency Q Q Q Q Q QLong-run elasticity of current income 0.7-0.94 0.96-0.98 0.74(proxy for human wealth) 0.4Long-run elasticity of total non-human net wealth 0.07-0.14 0.07 0.31(incl. housing)Long-run elasticity of net financial wealth 0.14Long-run elasticity of financial wealth 0.02-0.04Long-run elasticity of equity wealth 0.02 0.01Long-run elasticity of non-equity net financial wealth 0.08 0.10Long-run elasticity of non-equity financial wealthLong-run elasticity of housing wealth 0.05MPC out of total net wealth 2.1 - 4.1 1.6 4.4 1.5 - 2MPC out of financial wealth 2.7MPC out of net financial wealthMPC out of non-equity net financial wealth 7-9MPC out of equity wealth 2MPC out of housing wealth/physical capital 6.8 1.4Dependent (endogenous) variables6) C C C VECM C

Portugal Nether- Spain Sweden UK UK (fulllands model)

Estimation period 58-97 ? 80-01 79-98 75-01 75-01Data frequency A Q Q Q Q QLong-run elasticity of current income 0.94 0.682) 0.4-0.61)

(proxy for human wealth) 0.744) 0.161)

Long-run elasticity of total non-human net wealth 0.03 0.043) 0.06-0.171)

(incl. housing) 0.264) 0.125)

Long-run elasticity of net financial wealth 0.06 0.11Long-run elasticity of financial wealth 0.141)

Long-run elasticity of equity wealth 0.062)

Long-run elasticity of non-equity net financial wealthLong-run elasticity of non-equity financial wealth 0.112)

Long-run elasticity of housing wealth 0.152) 0.051)

MPC out of total net wealth 3 2 7.74) 4 - 55) 2MPC out of financial wealthMPC out of net financial wealth 3.7MPC out of non-equity net financial wealthMPC out of equity wealthMPC out of housing wealth/physical capital 1.9Dependent (endogenous) variables6) C C4) C C VAR

1) Clapham, Hyytinen and Takala (2002).2) Balmaseda and Tello (2002).3) Estrada et al (2004).4) Willman and Estrada (2002).5) Sveriges Riksbank.6) C = private consumption. VAR/VECM = multivariate approach including income and wealth7) Hamburg, Hoffmann and Keller (2005).

Table 4.5 Individual country (central bank and other) est imates of long-run ef fects off inancial wealth on consumption

27ECB




demographics. Estimates for the United Statesare an exception in that both micro and macrostudies consistently confirm the existence of awealth effect on consumption. The range ofMPC out of wealth estimates using US data isalso theoretically feasible. Further, some USstudies find somewhat lower MPC out of equitywealth than out of non-equity wealth. InEurope, the German and Italian (panel data)evidence supports this view.

Typically, MPC out of wealth estimates ininternational comparisons have been smallerfor continental EU countries than for theEnglish-speaking countries. Both cross-country and single-country studies broadlysupport this view. However, this is notuncontroversial, especially since estimates forEuropean countries are often smaller thanwould seem admissible on theoretical grounds.Also unlike in most other studies, the MPC outof wealth estimates presented by Labhard et al.(2005) stand out as being quite high forGermany, exceeding their estimates for theUnited States. Likewise a reported estimate forthe MPC out of non-equity net financial wealthwas in line with the corresponding USestimates. In addition, the panel data studies ofLudwig and Sløk (2002) and Case et al. (2001),which try to separate equity and housing wealtheffects on consumption, found quite high MPCout of wealth estimates for housing wealth inEuropean countries. Furthermore, Edison andSløk (2001) found that consumption inEuropean countries was more sensitive tovariation in the market value of “NewEconomy” stocks than was consumption in theUnited States.

Data problems may be a potentially importantsource of wide cross-country variation inestimated MPCs. Data problems are apparent inmany of the studies referred to in this report. Ininternational comparisons in particular, stockmarket capitalisation, stock prices and housingprices have often been used as proxies for thelevel of their respective wealth components.There are also differences across countries in

the measurement of wealth. Although agreedguidelines are generally followed, the specificmethodology applied in each country dependson data availability. As the more detailedpresentation in Appendix 1 shows, thetreatment of sole proprietorships, for instance,differs in European countries and in the UnitedStates. Methods of estimating the value ofunlisted companies also vary across countries.

4.5 HOW SIGNIFICANT ARE CROSS-COUNTRYDIFFERENCES IN ESTIMATED MPC OUT OFWEALTH?

Our analysis suggests that cross-countrycomparisons based on empirical estimates ofthe MPC out of wealth may be unreliable. Wefind it implausible that demographic andstructural differences between countries couldexplain cross-country differences in MPC outof wealth as large as those suggested in theempirical literature. In contrast, we find goodreasons to believe that data mismeasurementand structural shocks that vary across time andacross countries may lead to estimates of theMPC out of wealth that differ widely acrosscountries. Therefore, we might also expectempirical estimates of the MPC out of wealth tobe imprecise. Thus despite the apparently largedifferences in point estimates of MPC out ofwealth across countries, we may be unable toreject the hypothesis that such differences arenot significant. Labhard et al. (2005) giverecent evidence on this issue. They apply adynamic panel approach recently suggested byPesaran et al. (1999). Subject to a specificationin which the parameters can be interpreted asMPC out of wealth, their pooled mean groupestimator (PMGE) can test cross-countryrestrictions on the long-run MPC out of wealth,while taking into account possible cross-country differences in the adjustment to thiscommon long-run MPC out of wealth.Therefore, a PMGE-based direct estimate ofthe MPC out of wealth is likely to provide atheory-consistent guide to wealth effects onconsumption, and a better guide than estimatesobtained from either traditional panel methods


or single equation techniques.16 In this case,traditional panel methods, including the fixedand random effect estimators, are overlyrestrictive, since they force all coefficientsexcept the intercepts to be identical acrosscountries. The PMGE-specification is writtenas follows:

⎥⎦

⎤⎢⎣

⎡β−α+⎟⎟

⎠

⎞⎜⎜⎝

⎛∆γ+⎟⎟

⎠

⎞⎜⎜⎝

⎛∆γ+γ=⎟⎟

⎠

⎞⎜⎜⎝

⎛∆

−

−

−

−

−

−

1t

1t

1t

1ti

t

ti,2

1t

1ti,1i,0

it

t

Y

W

Y

C

Y

W

Y

C

Y

C

All parameters except the MPC out of wealth(β ) are country-specific. Therefore, the PMGEprovides a means to test the hypothesis that theMPC out of wealth is identical acrosscountries. Labhard et al. (2005) estimated theabove equation for the panel of 11 countries.Their estimate for β is 6.8, which is somewhathigher than the average of the individualcountry estimates (4.8). In addition,specification test supported the hypothesis thatthe MPC out of wealth is equal for all countriesin the panel, although some of the single-country estimates yielded results which werenot economically sensible.17

4.6 CONCLUSIONS

A number of studies have estimated significantwealth effects across a very broad range of EUand non-EU economies.

Wealth-consumption ratios differ markedlyacross countries, and, in principle, may providean indication of the relative strength of wealtheffects in aggregate and across a range ofassets. However, an important contribution ofthis study is to suggest that methodologicaldifferences in compiling household balancesheet data still remain across countries and theymay have exaggerated the cross-countrydifferences in wealth-consumption ratios. Theeffect of equity price changes on householdwealth in Germany, for example, may berelatively understated, while those in theUnited States may be overstated. Any suchmeasurement discrepancies will inevitably bebuilt into model-based estimates of wealth

16 Moreover, due to its flexibility, the PMGE is particularlysuitable for the “large N, large T” case, i.e. when the cross-sectional and time-series dimensions are both large andsimilar in magnitude, as is usually the case in cross-countrystudies.

17 Pesaran et al. (1999) have argued that this is one of thesituations in which the PMGE is particularly useful.

effects since models re-value equitiesaccording to published data.

Calibrated theory-based models give a goodbenchmark for an expected MPC out of wealth.For instance, the cross-country estimates ofMULTIMOD are in the range of 5.4 to 8.2 centsper euro (cents per dollar). Although theavailable evidence is rather sparse, MPC out ofwealth estimates based on the direct estimationof the theory-based equations are well in linewith calibrated estimates, being in the range of3 to 10 cents per euro (cents per dollar), if thesome highest outlying estimates are excluded.On the bases of these estimation results, noclear evidence can be found for MPC out ofwealth being lower in the continental Europethan in the English-speaking countries.

Individual country studies find elasticities tototal non-human wealth in the range 0.03 to0.17 for European countries, while the MPC isin the range of 1.5 to 7.5. Studies which providepooled estimates of MPC tend to show highervalues for the United States and the UnitedKingdom than for other European countries. Onthe other hand, estimates from VAR modelsshow smaller differences between Europeancountries, especially Germany, and the UnitedStates. Though many existing empirical studiesestimated plausible MPCs out of financialwealth, the differences across countries andacross studies for the same country are ratherlarge. They can only be explained bycombination of factors (different timehorizons, demography, distribution of wealth,time-varying rates of return) includinginconsistencies in the data that are used.Interestingly, the panel data estimationtechnique, developed by Pesaran (1999), whichis less sensitive to cross-countryincompatibilities in data construction,

29ECB


5 ASSET PRICESAND

INVESTMENTprovided the MPC out of wealth estimate of 6.8cents per euro, which is comparable with atypical estimate for the United States. Moreimportantly, cross-country differences in theMPC out of wealth estimates were notstatistically significant.

Finally, there is some empirical support for thehypothesis that housing wealth may havestronger impact on consumption than equitywealth in some EU countries, although theevidence is scarce and inconclusive. Butrelatively little cross-country research has beendone to try to study the simultaneousrelationship between asset price andconsumption changes, or how the consumptionresponse might differ according to the cause ofthe shock to equity prices.

5 ASSET PRICES AND INVESTMENT

In addition the effects described above, assetprices may also have an effect on economicactivity through its effect on investment. Inparticular, equity prices may affect corporatesector investment and house prices may affectresidential investment. This section reviewsthe available evidence of asset price effects onboth of these investment components.

5.1 STOCK PRICES AND CORPORATEINVESTMENT

Share prices may also affect economic activityby changing the cost of equity capital to financecorporate investment. Changes in share pricesmay also alter the financial structure of thecompany, affecting the cost and the availabilityof external funds. Finally, changes in shareprices that reflect information about the futureprospects in the economy, may affectexpectations and therefore investment plans.

More formally, the mechanisms describedabove are usually classified in the followingthree channels:

1. Tobin’s Q or cost-of-capital channel2. Balance sheet channel3. Confidence channel

In general, empirical evidence linking shareprices and investment is limited. When theinvestment equation accounts for alternativeexplanatory variables, share prices still play arole in explaining investment, but the impact ofcash flow and sales suggest that the balancesheet channel and the confidence channel maybe more important than the cost of capitalchannel.

There are several possible reasons for thisrather weak link between investment and shareprices. First, when share prices are lowcompanies may substitute debt financing forequity financing and when share prices are highcompanies may use funds obtained by equityfinancing to improve their financial structureor for mergers and acquisitions instead ofinvesting. Second, response of investment maydepend on the source of a given change in stockprices. Temporary changes in share prices maynot have the same impact on investment aschanges perceived as permanent. Empirically,however, it is difficult to establish howcompanies perceive movements in the stockmarket. In other words, it is difficult todetermine whether share prices reflect thefundamental value of the company or carry aspeculative component.

Moreover, all these factors will affectinvestment simultaneously through the threechannels listed above. While speculativemovements in share prices should not affectinvestment through the cost-of-capitalchannel, they may have an impact through theconfidence channel. Similarly, while anincrease in share prices may not stimulateinvestment through the reduction in the cost ofcapital, companies may use equity issues toimprove their financial position, affectinginvestment through the balance sheet channel.

In practice it is very difficult to disentangle theeffect of each of these channels on investment,


so in the following we concentrate on Tobin’sQ channel. A separate review on the balancesheet and confidence channels given inAppendix 3 suggests that the balance sheeteffect through cash flow may be veryimportant, in particular for small (and young)companies. However, given that thesecompanies are unlikely to be listed on the stockexchange, it is difficult to identify a directeffect of stock prices on investment through thebalance sheet. Changes in stock prices may alsobe transmitted to investment through the bank-lending channel. When stock price movementslead to a deterioration in bank balance sheets,this reduces their willingness/capability tolend. Finally, although there seems to be aclose relationship between stock prices andconfidence in the euro area, empiricalestimation of a separate confidence channel oninvestment is rather difficult.

5.2 TOBIN’S Q OR COST-OF-CAPITAL CHANNEL

In theory a change in share prices implies achange in the cost of raising equity capital tofinance corporate investment. This has beenformalised theoretically in Q theory (Tobin,1969) which describes how movements in stockprices can affect investment. For example, acompany may perceive the cost of equitycapital to be rather high in periods when thecompany’s stock price is low compared toearnings per share. As a result, companies mayreduce investment in projects previouslyperceived as profitable.

According to Tobin’s Q theory, the companyshould invest as long as the market value of aninvestment exceeds the replacement cost ofthat investment. Marginal Q is defined as thechange in the market value of a companydivided by the change in its capital stock(investment) that caused the change in marketvalue. If marginal Q is larger than one, themarket price of companies is high relative tothe replacement cost of capital, and new plantand equipment capital is cheap relative to themarket value of the company. The company canthen issue stocks at a high price relative to the

cost of the facilities and equipment they arebuying. In this case, investment spending willrise because companies can now buy more newinvestment goods with only a small issue ofstock.

Although in practice marginal Q is notobservable, average Q can be expressed as themarket value of the company divided by thereplacement cost of capital. Hayashi (1982a)proves that under certain conditions average Qis equal to marginal Q. Therefore, mostempirical studies use average Q and set themarket value of the company at the share price.An increase in share prices makes it cheaper forcompanies to finance their investment becauseeach share issued yields more funds. Thus, anincrease in stock prices lowers the cost ofcapital, stimulating investment. In this theory,no other variables are needed to explaininvestment because all expectations includingfuture revenue are captured in share prices.

Thus, the investment equation can be writtenas:18

( ) tt

Kt

t

t

t Qb

aKp

Vb

aKI 11

11

1+=⎟

⎟⎠

⎞⎜⎜⎝

⎛−

−+=⎟⎟

⎠

⎞⎜⎜⎝

⎛

−δ (5.1)

where tt KI is the investment ratio; tV is thefundamental value of the company, given bythe discounted present value of the cash-flowstream expected from the existing capitalstock; the expression ( ) 11 −− t

Kt Kpδ is the

replacement cost in period t of the capital stockthat the company inherits from the previousperiod. As mentioned before, tV is measured bythe company’s share price.

The theory assumes that stock markets arestrongly efficient, in the sense that thefundamental value of the company can bemeasured by its stock market valuation. Thus,stock prices should reflect the value of thecompany given by the discounted present valueof the cash flow and this should contain allavailable information about the company.Given this assumption, the only information

18 See e.g. Bond and Cummins (2001).

31ECB


5 ASSET PRICESAND

INVESTMENTnecessary to estimate a company’s investmentdemand is given by Q. Since information aboutthe financial structure or expectations aboutfuture revenue should be known, these shouldbe reflected in share prices.19

However, empirical studies using multivariateregressions have almost unanimously failed tofind evidence of an effect of share price oninvestment expenditure. For example, Tease(1993) concludes: “When other determinantsof investment are controlled for, share pricesdo not seem to explain much of the variation ininvestment in any of the G7 countries.”20

Blanchard et al. (1993) and Morck et al. (1990)also examine the effect of share prices oninvestment and report similar results. While(non-residential) investment expenditureremains hard to explain, the relativeimportance of other key factors is confirmed byvarious studies: Chirinko (1993) concludes:“...output (or sales) is clearly the dominantdeterminant of investment spending with usercost having a modest effect.” Moreover, “theusefulness of the Q-theory is called intoquestion by its generally disappointingempirical performance.”21

Some authors attribute these disappointingresults to the use of share prices as indicators ofthe fundamental value of the company whenstock markets are not efficient. There areseveral reasons why the valuation of thecompany by managers and the market maydiverge:

1. the market may have less information

2. even if the information is the same, themarket may not value assets at theirfundamental value, and the share price mayinclude a rational speculative bubble

3. the market may be subject to fads that causemarket valuation to deviate fromfundamentals for long periods.

In theory, companies will react only topermanent changes in share prices and will tend

to ignore movements in share prices that areconsidered fads or bubbles.

Thus, some authors have attempted todistinguish between the components of theshare price that reflect the fundamental valueof the company and the residual speculativecomponent. Thus the fundamental value of thecompany may be expressed as the market valueless the speculative or “valuation” component:

Stt

Ft VVV −= (5.2)

In the literature, fundamentals have beenproxied by sales, cash flows, dividends, andanalysts’ earnings forecasts, among othervariables.

Blanchard et al. (1990) made one of the firstattempts to perform this decomposition. Theyinvestigated whether managers ignore shareprices when they deviate from fundamentals,but could not reach a definite conclusion.While there is no clear evidence of a bubble in1929, the evidence in 1987 suggests managersignored market developments.

Bond and Cummins (2001) constructed ameasure of average Q for a group of UScompanies using analysts’ forecasts instead ofshare prices. They conclude that stock marketvaluations deviate significantly fromfundamental values, rendering share pricesuseless to explain investment.

A similar approach was followed by Andersonand Subbaraman (1996) for Australia. Theyestimate the fundamental and speculativecomponent of share prices and find that thefundamental component has a strongerrelationship with investment than real share

19 This means that the information provided by the financialstructure of the company and expectations about futurerevenue, which are of central importance for the balancesheet and conf idence channels, should also be captured byaverage Q.

20 Tease (1993), p. 58.21 Chirinko (1993).


prices, while the speculative component doesnot seem to have any significant effect onbusiness investment. Evidence for Japan is alsomixed. Goyal and Yamada (2004) find that thespeculative component affects investmentduring asset prices booms, while fundamentalsmatter more after asset price collapses.Chirinko and Schaller (2001) use the Q modelto test whether there was a stock market bubblein Japan in the late 1980’s. Their resultsconfirm the existence of the bubble and alsothat developments in stock prices during thisperiod affected business investment.

Some authors have also tried to explain the lackof supporting evidence for the cost of capitalchannel by arguing that companies with accessto other sources of funds will not react to stockprices in the same way as equity-dependentcompanies (Baker et al, 2002). However, Goyaland Yamada (2004) found that bank-dependentcompanies in Japan show a stronger reaction toshare prices.

More promising results have been found whenother variables are included in the investmentequation. Even when there is no separationbetween speculative and fundamentalcomponents of share prices, the significance ofadditional variables in the investment equationsuggests that share prices do not reflect allcompanies’ investment opportunities as statedin the original theory. In other words, thisprovides evidence that stock markets are nothighly efficient, or that there are measurementproblems in Q. Some studies using thismethodology are described below.

Ashworth and Davis (2001) estimateinvestment equations for the G7 countriesincluding average Q as well as other financialvariables. Average Q is significant only forJapan and France.

Audretsch and Elston (2002) estimate aninvestment equation for German companies inwhich both Q and some financial constraintsare included. Their results show that Q issignificant for two groups of companies: the

smallest and the second-largest companies intheir sample. This partly confirms the findingsof Baker et al (2002) since the smallestcompanies may have less access to externalfinance from banks. However, it is puzzlingthat the largest companies do not react to shareprices while the second-largest do.

Alonso and Bentolila (1992) find that Tobin’sQ is only partially significant for explainingfixed capital investment by Spanish industrialcompanies. The elasticity of investment tochanges in the stock market is rather low, aresult compatible with Spain’s poorlydeveloped financial markets at that time.Sensitivity analyses suggest that estimatedcoefficients are not stable, although thequalitative conclusions are maintained. As isoften found in other studies, the significance offinancial variables, such as internal finance orcash flow rejects the simplest version of the Qmodel – namely that Q is a sufficient statisticfor investment.

Van Ees and Garretsen (1994) consider Tobin’sQ in studying the effects of liquidity oninvestment in a group of Dutch companies.They also include financial variables such ascash flow, the stock of liquid assets, and theworking capital. Their evidence corroboratesprevious findings that once sales and financialvariables are included in the equation, theeffect of Tobin’s Q on investment is very smallor insignificant. Moreover, these results do notchange when the sample is split according tocompany size or dividend pay-out. While theeffect of financial variables depends onwhether companies have close ties with banks,the small effect of Tobin’s Q on investment isnot affected by the existence of bank lendingrelationships.

Assarsson et al (2004) use a neoclassicalinvestment model extended with installationcosts for capital, agency costs for investmentfinancing, and the possibility of the companybeing output constrained as a framework for anempirical analysis of investment behaviour inthe Swedish manufacturing industry. Using a

33ECB


5 ASSET PRICESAND

INVESTMENTmultivariate error-correction model on datacovering the time period 1951 to 1995 theirresults show that Tobin’s average Q is not thesole determinant of investment, neither in theshort nor in the long run, and other variableslike real output and capital gearing also affectinvestment activity.

5.3 HOUSE PRICES AND HOUSING INVESTMENT

Exactly as in the case of corporate investment,there should also be a relationship betweenTobin’s Q and residential investment.According to Tobin’s Q theory, the supply ofhouses will increase when it becomesprofitable for a construction company to buildmore houses. This will happen when the priceat which the house can be sold is higher than itsconstructions costs. This is when Tobin’s Q,which is the ratio between house prices andconstruction costs, is larger than 1. Thus, risinghouse prices will be accompanied by anincrease in the supply of houses, i.e. housinginvestment. In the long run, however, theincreased supply of houses forces real houseprices to fall and Tobin’s Q to converge back tounity.

As in the case of corporate investment, Tobin’sQ for residential investment is not observableand its calculation poses a number of problems.The price of houses should take into accountimprovements in quality, which are in generalvery difficult to measure. In empirical studies

construction cost also poses a problem, so it isproxied by the deflator for private residentialinvestment. This measure, however, does nottake into account the cost of land, which mayaccount between 20 and 40 percent of totalbuilding costs and which has increased evenfaster than house prices in some countries.22

Due to the lack of data there are few countriesfor which studies have attempted to relatehousing or residential investment to a measureof Tobin’s Q. Examples of two Europeanmacromodels, which apply the Tobin’s Qapproach are the models of Suomen Pankki,

BOF5 (1999), and Danmarks Nationalbank,MONA (2003). In both models, but especiallyin BOF5, housing investments reactions to realhouse prices play a central role in balancinghousing demand and supply. BOF5 simulationsindicate that the elasticity of housinginvestment with respect to real housing pricesis about 0.9. These results are close to thoseestimated by Kenny (2003) for Ireland, whofound long-run elasticity of housinginvestment with respect to real house prices ofaround 1.0.

Based on the data of some selected OECDcountries in the period 1980-1999, Girouardand Blöndal (2001) find strong correlation(above 0.8) between housing investment andreal house prices in Belgium, Denmark, theNetherlands and Spain. Only very weakcorrelation is found for the larger countries(United States, Japan, France) and for Norway.Evidence for Germany is tainted by the lack ofgood data for the new Länder. However, thisevidence does not confirm whether there is along-term relationship between Tobin’s Q andprivate residential investment.

Barot and Yang (2002) analyse house pricesand housing investment in Sweden and the UK.Their results show that Tobin’s Q “Granger-causes” housing investment in both countries.Further, based on the error correction approachfor the sample period 1970-1998, their resultsindicate that in the short run Tobin’s Q issignificant only in the United Kingdom, but inthe long run is significant in both countries.However, as discussed by Muellbauer (2004),it is a generally accepted view is that the UnitedKingdom is a country, where the priceelasticity of the supply of new houses is small,being close to zero and probably lower than 0.5.

We may conclude that the available evidenceindicates that in many countries housinginvestment depend positively on the real priceof housing. However, it is also likely that the

22 OECD (2001).


strength of these supply responses varymarkedly across countries.

6 CONCLUSIONS

This report reviews the available theoreticaland empirical evidence regarding asset priceand wealth effects in Europe and some othermajor economies. For the most part, the work isof a stock-taking nature, although some newand unpublished material is also included in theanalysis. The main focus of this report is onconsumption effects via the wealth channel,reflecting the bulk of literature on the effects ofasset prices. However, asset price effects oninvestment via the Tobin’s-Q channel are alsoreviewed. In addition, Appendix 3 reviewspossible asset price effects on private spendingvia balance sheet and confidence channels. Theavailable evidence supports the view that thewealth channel is the most important of variouschannels. There is little empirical evidenceindicating that the Tobin’s-Q, balance-sheetand confidence channels play any majorindependent role in the transmission of assetprice effects to economic activity. The keyfindings in more detail are as follows:

– A baseline theoretical analysis revealedthat the size of the wealth effect onconsumption should depend on thefollowing elements. First, the wealth effectdepends inversely on the length of planninghorizons (expected remaining lifetime) orthe strength of the bequest motive. Second,the wealth effect depends positively on theexpected risk-adjusted rate of return onwealth, which in turn depends on theexpected rate of return on wealth and thesubjective discount rate. Higher riskaversion will move the risk-adjusted rate ofreturn closer to the expected rate of returnon wealth. Third, if expected returns onassets vary over time, then the response ofconsumption to changes in wealth maydepend on how permanent these changes areexpected to be. For example, an increase instock prices that is expected to be followed

by a period of poor returns may have littleeffect because a decline in the MPC maypartially offset the higher level of wealth.Finally, labour income uncertainty impliesthat consumers in the upper tail of theincome distribution may have markedlylower MPC out of wealth than those in thelower tail of the income distribution.

– Computed examples based on theintertemporal budget constraint suggestthat reasonable estimates of the MPC out ofwealth might be in the range of 3 to 10 centsfor each one euro permanent increase inwealth. A natural lower bound for the MPCout of wealth is the equilibrium risk-freeinterest rate (or the subjective discountrate), but values below that are feasible, ifcombined with low (i.e. below one) relativerisk aversion.

– Based on national data sources, the ratio ofgross financial wealth to consumptionexpenditure varies markedly acrosscountries. However, these differences alsoreflect differences in statistical andaccounting practices, which weaken theinternational comparability of results.Moreover, stock market capitalisation datatypically used in international comparisonsare subject to even more severe statisticalproblems.

– Comparing wealth composition changesthrough time and across countries, the1990s saw a striking increase in the share ofequity holding by households and, ingeneral, a reallocation of householdfinancial wealth towards more risky assetsin most countries. However, there aremarked cross-country differences in equityownership. In the United Kingdom 23% ofthe population own shares (in 2000), in theUnited States this proportion is 19% (in1998), in France it is 13% and in Germanyand Italy it is 10%. There are also strikingcross-country differences in the ratio ofdirectly held equity wealth to consumption.In 2000 this ratio was around 200% in the

35ECB


6 CONCLUSIONS

United States, 150% in France, 100-130%in the Netherlands, Belgium, Italy, Spainand Sweden, 90% in the United Kingdom,50% in Germany and 40% in Austria.However, in all countries the majority ofequities are owned by households in thehighest income brackets. Moreover, equityownership is more common among olderage groups, although more so in the UnitedStates than in continental Europe.

– International comparisons generally reveala statistically significant link betweenwealth and consumption spending.However, the implied MPC out of wealthestimates for EU countries are typicallysmaller than those reported for the UnitedStates, although this is not anuncontroversial result. In particular, paneldata estimation yields a MPC out of wealthestimate of 6.8 cents per euro, which iscomparable with a typical estimate for theUnited States. More importantly, cross-country differences in MPC out of wealthestimates were not statistically significant.

– Panel data studies report higher MPC out ofhousing wealth in European countries thanin the United States, while the situation isreversed for the MPC out of equity wealth.

– Regarding country-specific studies theavailable econometric evidence on the linkbetween equity wealth and consumption issparse in Europe. Documented, but quitetentative, elasticities with respect to equitywealth (with implied MPC out of wealth inbrackets, where available) are 0.01 forGermany (2 cents per euro), 0.02 forFinland (1 cent per euro) and around zerofor Sweden. On this basis, a permanent 50%drop in stock prices would imply adownward shift in the level of consumptionexpenditure of 0.5% in Germany and 1% inFinland.

– The link between housing wealth andconsumption also seems to be quitecountry-specific. For some EU countries

there appears to be little evidence of aneffect on consumption (e.g. Austria,France, Germany and Italy). This mayreflect cross-country liquidity differencesin housing wealth, but also the fact thatAustria, Germany and Italy are countrieswhere house prices have been most stable,which may have made it difficult to identifythe impact of housing wealth onconsumption. In countries where houseprices have been more volatile, astatistically significant positive effect ofhousing wealth on consumption hastypically been found (Finland, Greece, theNetherlands, Spain and the UnitedKingdom). There is some evidence that inthese countries the effects of housingwealth (and house prices) on consumptionmay be stronger than most reported effectsfrom equity wealth and from financialwealth more generally.

– The United States provides an importantbenchmark because it has high-quality datasources and has been the subject of mostexisting econometric studies. US evidenceprovides robust support for the importanceof the wealth channel. In particular, thepreferred estimates of the long-run MPCout of wealth were in the range of 6 to 7cents per dollar. The MPC appears to besomewhat higher for non-equity wealth(about 10 cents), and somewhat lower forequity wealth (about 5 cents). Time-variation in expected returns would implyvariation in the MPC out of wealth.However, on the basis of the availableevidence it appears that the US wealtheffect on consumption has been relativelystable over time.

– Regarding the equity price effects oninvestment via the Tobin’s-Q channel, ageneral conclusion seems to be that equityprices have no additional explanatorypower once other determinants ofinvestment, such as sales or cash flow, areincluded. Nonetheless, a link from houseprices to residential investment has been


identified in many countries. However, theavailable empirical evidence is quite sparseand it may be that the impact of house priceson residential investment is quite country-specific, reflecting, for example, cross-country differences in the transmission ofhouse prices to construction costs (landprices) and /or the availability of mortgageloans.

– Most studies find that the balance sheeteffect, operating through cash flow, seemsto be a very important determinant ofinvestment, in particular for small (andyoung) companies. Given that thesecompanies are very unlikely to be listed onthe stock exchange, it is difficult to identifya direct effect of stock prices on investmentthrough the balance sheet. However,changes in stock prices may be transmittedto investment through the bank-lendingchannel. This may happen when bankbalance sheets deteriorate as a result ofstock price movements, affecting theirwillingness or capacity to lend. Finally,although there seems to be a closerelationship between stock prices andconfidence in the euro area, the separationand empirical estimation of the effects via aconfidence channel on investment is ratherdifficult.

37ECB


APPENDIX 1

APPENDIX 1

COMPARABILITY OF EQUITY WEALTH DATAACROSS COUNTRIES

While much progress has been made inharmonising methodologies to calculatefinancial accounts – at the present time all EUcountries apply criteria laid down in theEuropean System of Accounts (ESA 95) –wealth data for international comparisonpurposes are still fairly heterogeneous in termsof the compilation criteria.23 Differences arisefrom the concepts themselves, that are notexactly the same across countries, from theways they are measured and from the fact thatthe information necessary for applying generalguidelines is not available.

There are many differences in the practicalimplementation of the recommendations ofESA 95 in each country, which affect thepublished figures. However, the differences inthe criteria used to obtain the value ofhousehold equity holdings have a greaterinfluence on wealth figures. The maindifferences are the following:

The treatment of sole proprietorships andpartnerships in European countries and theUnited States is different. In the US financialaccounts they are considered non-financialcorporations, while in the European countriesthey are included in the household sector. As aconsequence, the whole net worth of soleproprietorships is regarded as householdfinancial assets in the US methodology, whilein the other countries, part of this amount(professional buildings, plant and machinery)is regarded as non-financial assets. Therefore,it can be said that, in the United States, the ratioof “shares and other equities” to total assets isoverestimated relative to European countriesand the total amount of financial assets is alsolikely to be larger than that which reallypertains to households.

The valuation of equities follows the principlesof ESA 95, which means that marketable

financial instruments are valued according totheir market value. This poses a big problem inthe case of unlisted shares, which are not tradedon organised markets. The general principle ofESA 95 is to estimate the value of unlistedshares with reference to the value of listedshares and to also take into account the sectorand the differences between them due to theirdifferent liquidity and reserves accumulated.However, the specific estimation methodapplied by each country very much depends onthe basic statistics available. Table A.1summarises these methods in a very syntheticway for a few countries.

Valuation changes due to exchange ratemovements pose an additional problem forestimating the value of shares and participationin foreign companies. Although most of themare held by domestic firms and not directly byhouseholds, the estimate of their value alsoaffects the estimated value of domesticcompanies, which is part of household equityholdings. There is no unified criterionregarding the specific exchange rate to beapplied so each country applies what itconsiders to be appropriate. However, theapplication of different exchange rate indicesis not neutral and may give rise to verysignificant differences in the value of equityholdings in financial accounts.24

According to Table A.1, France and Spain arecountries that apply market price criteria (more

23 Comments in this appendix are based on Babeau and Sbano(2002).

24 For example, in Spain there were important revisions of theestimated market value of shares when the index used as areference to estimate value changes due to exchange ratemovements was changed from an effective exchange rate toseveral individual currency indices.


closely related to stock markets) to a largerextent. When analysing equity to consumptionratios (Table A.2), France and Spain haveexperienced a relatively large increase in thisratio in the period 1995-2000, althoughincreasing equity to consumption ratios are acommon feature across countries. Therefore,the method applied to estimate the value ofequities across countries appears to have asignificant influence on the changes of thewealth to consumption ratio over time, thoughit is difficult to disentangle this effect fromother effects on the data, such as privatisation.

Section i: Various practices in the valuation method of unlisted companiesItaly The ESA 95 general principle (capitalisation/own funds ratio of quoted companies of a similar size and activity)

is applied for large non-listed companies. For non-listed small to medium-sized companies the valuation appliedis according to their book-value.

Spain Unlisted shares issued by banks are valued by applying the ESA 95 general principle. The value of unquotednon-financial companies is estimated as the discounted value of the flow of expected ordinary profits. Inpractice, the discount rate used is inferred from the valuation of quoted corporations.

France The ratio between the market value of listed companies and their own funds is multiplied by the own funds ofunlisted companies. This procedure is currently under revision.

United Kingdom ESA 95 is applied, but work is still being done on estimates.

Netherlands Unlisted shares are evaluated at own resources.

Germany Unlisted shares (public limited companies) are valued using the market capitalisation of listed companies, withconsiderable discounts reflecting the lower ratio of own funds. Unlisted private limited companies are valuedaccording to their book-value and applying a 30% increase, in order to take into account reserves.

Section ii: Other examples of measurement issuesBelgium Large stocks of government debt may have a relatively large impact on the valuation of household wealth after

adjusting for expected future tax payments.

Germany Data availability of mutual fund equity holdings, insurance company accounting practices.

United States Different sector definition in US flow of funds accounts: 1. narrow definition: household sector does not includeself-employed;1) 2. broad definition: household sector includes self-employed and non-corporate firms.

Table A.1 Data issues in measuring wealth

1) In contrast to this sector def inition, the German household sector includes households, non-prof it organisations and the self-employed.

1995 2000 Change

Italy 1.07Spain 0.46 1.08 0.62France 0.64 1.54 0.91United Kingdom 0.60 0.90 0.29Netherlands 0.96 1.29 0.33Germany 0.29 0.48 0.19

1) Equity is def ined here as shares and other equity directlyheld by the household sector.

Table A.2 Equity to consumption ratios 1)

39ECB


APPENDIX 1

AT BE FI FR DE IT NL PT SP SE UK US JP AU CA

National sources 2000 2001 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000

Gross financial Wealth 236 539 302 414 304 386 596 312 320 311 502 525 493Directly held equities 40 121 189 154 48 107 129 73 108 97 90 196 42Of whichquoted 44 37 52 126 39unquoted 77 11 38 70 3Mutual funds 9 80 7 36 34 65 26 40 40 26 49 0Pensions and lifeInsurance 72 90 84 47 321 43 41 265 150 139Cash and Deposits 140 110 104 93 112 145 116 100 64 264Other 125 23 35 73 35 25 16 21 66 48Financial liabilities (Eurostat) 70 83 66 84 126 50 170 127 100 102 116

Measures of total financial wealth (FW) from other studiesByrne and Davis (2003)Gross financial wealth as percentageof GDP in 2000 229 168 225 300Bertaut (2002)Gross financial wealth in 1998 340 240 370 600 230 480 520 460 250 380NIGEMNet financial wealth in 2002 Q3 250 150 270 270 370

Measures of stock market wealth from other studiesByrne and Davis (2003)Household sector equity wealth 110 45 69 159Bertaut (2002)Stock market capitalisationas percentage of GDP in 2001 89 59 48 148 114 153 130 62 103 90

Table A.3 Financial assets held by household sector, measured as percentage of consumption(unless otherwise indicated)


APPENDIX 2

THE US EVIDENCE ON WEALTH EFFECTS ONCONSUMPTION

This section reviews the evidence of the effectof movements in asset wealth on USconsumption. The US economy provides aparticularly good “laboratory” for testinghypotheses about the relationships betweenasset prices and consumer spending because ofthe existence of the detailed quarterly Flow ofFunds Accounts which provide acomprehensive picture of the aggregatebalance sheet of the US household sector; theUnited States also has a number of usefulsources of detailed micro-level information onhousehold patterns of consumption and assetholdings.

This appendix first presents some estimatesfrom aggregate regressions designed toestimate the MPC out of financial wealth and toreview why various empirical studies havetended to arrive at different estimates of thisparameter. Evidence from some studies basedon micro-data is then briefly reviewed.

A2.1. Aggregate time series evidence

Empirical estimates of macroeconomic wealtheffects have generally taken the aggregatebudget constraint as their starting point, andthen made a couple of simplifying assumptionsregarding preferences and asset returns. Forexample, the most common applications of thepermanent income hypothesis (PIH) usuallyassume that the rate of return on assets isconstant. This assumption simplifies theaggregate budget constraint as

(A2.1) [ ]tttt CYArA −++=+ )1(1 .

Repeated substitution of this expression yieldsa standard expression for the intertemporalbudget constraint:

(A2.2) ∑∑∞

=

+∞

=

+

++=

+ 00 )1()1( kkktt

tk

kktt

rYE

ArCE

.

Finally, a structural “consumption function”linking consumption to wealth and income isderived by making some assumptions about theexpected evolution of consumption. Forexample, Robert Hall’s (1978) seminalcontribution assumed that consumers hadquadratic utility and a discount rate equal to therate of return on assets. This yielded theprediction that consumption should follow arandom walk. Replacing all future expectedvalues of consumption with today’s value andre-arranging we get the most common“structural” form of the PIH:

(A2.3) ∑∞

=

+

+++

+=

okkktt

tt r

YErr

Arr

C )1(11 .

This specification implies that householdsconsume a constant proportion, r

r+1 , of their

total wealth, defined as the sum of currentfinancial assets plus the discounted sum ofcurrent and expected future labour income.25

A re-formulation of this equation that has beenused to estimate the MPC out of householdwealth expresses the relationship in ratio formas

(A2.4) εα tt

t

t

t

YA

rr

YC +

++=

1 ,

where t is a mean-zero term reflecting current

expectations about deviations of future labourincome growth from its average levels. Notethat we are referring to Y as labour income tobe consistent with the underlying budgetconstraint. Capital income is reflected in thebudget constraint as the rate of return times thestock of invested assets, i.e. as [ ]ttt CYAr −+ .

Figure A2.1 shows the fit of a simple ordinaryleast squares (OLS) regression for US quarterlydata based on equation (A2.4), i.e. of aregression of the consumption to labour incomeratio on a constant and the wealth-to-labourincome ratio. The figure shows that movements

25 It should be noted that the prediction that households consumea constant proportion of total wealth is more robust to otherassumptions about preferences than quadratic utility,although the exact formula here for the MPC out of wealth isspecif ic to these preferences.

41ECB


APPENDIX 2

in wealth appear to explain most of the long-runswings in the consumption to income ratio: theadjusted R2 for this regression is 0.77. Charts ofthis sort have been used in a wide range ofstudies to illustrate how the growth in wealth inthe 1990s is a likely candidate for thesubstantial decline in the US saving rate overthis period.

Table A2.1 reports some more details behindthis regression, along with estimates from otherimplementations of this regression equation.All of the regressions employ an updatedversion of the quarterly dataset used inPalumbo, Rudd, and Whelan (2002). In eachcase the measure of asset wealth is totalhousehold net worth from the Flow of FundsAccounts (excluding the stock of consumerdurables) as measured at the end of the previousquarter.26

26 The standard assumption about the budget constraint made inderiving consumption functions is that the consumption seriesin the budget constraint corresponds to total consumptionexpenditure. This implies that expenditure on durables is notconsidered asset accumulation in this case, and so theoreticalconsistency suggests that it should be excluded from themeasure of wealth.

Fig. A2.1 The ef fect of wealth on USconsumption

Dependent variable is consumption to labour income ratio

1.05

1.1

1.15

1.2

1.25

1.3

1.35

1.05

1.1

1.15

1.2

1.25

1.3

1.35

DataFitted Values

1952 1962 1972 1982 1992 2002

The first row of the table shows that theestimate of the MPC out of wealth generated bythe regression illustrated in Figure A2.1 is 6.1cents per dollar of wealth. The second row usesa slightly different methodology: It replacestotal consumption with a scaled-up version ofconsumption of non-durables and services.This approach is commonly taken in empiricalconsumption studies on the grounds thattheories of short-run consumption dynamicsapply best to a flow measure of consumption.This approach yields a relatively similar fit tothe regression using total consumption, but hasa slightly higher estimate of the MPC out ofwealth of 6.9 cents per dollar. A number ofaspects of these results are worth noting.

Deflation: The budget constraints – equations(A2.1) and (A2.4) – are each understood to beformulated in terms of real consumption, assets,income, and asset returns, i.e. each of thenominal variables in the nominal budgetconstraint have been deflated by the price indexfor total consumption. Because of this deflationby a common price index, the ratios of realvariables in equation (A2.4) are identical to theratios of the corresponding nominal variables.The regression in the second row (labelled“Method 1”) follows this logic by replacingnominal consumption expenditures with ascaled version of nominal consumption of non-durables and services. This is the approach taken

Standard errors in parentheses

Income Measure Consumption measure ααααα µµµµµ Adjusted R2

Labour income Total consumption 0.84(0.03) 0.061 (0.005) 0.77Labour income Scaled non-durables and services (Method 1) 0.79(0.03) 0.069 (0.005) 0.74Labour income Scaled non-durables and services (Method 2) 0.98(0.07) 0.041 (0.012) 0.26Disposable income Total consumption 0.72(0.01) 0.040 (0.003) 0.70

Table A2.1 Empir ical est imates of for US data

εµα tt

t

t

t

YA

YC ++=

Note: Sample is 1952:2 to 2002:2. See text for descriptions of Method 1 and Method 2.


by Palumbo, Rudd, and Whelan (2002). Thismethod will provide an accurate approximation ofthe true relationship as long as the ratio of nominalconsumption of non-durables and services to totalnominal consumption expenditures is stable;Palumbo, Rudd, and Whelan (2002) show thatthis is the case for US data.

In contrast, however, many studies estimateconsumption functions by continuing to deflateincome and wealth by a price index for totalconsumption expenditure, but using a scaledversion of real consumption of non-durablesand services, so that consumption, income andassets are no longer all deflated by a commonprice index. Palumbo, Rudd, and Whelan showthat this approach will only be accurate if theratio of real consumption of non-durables toreal consumption expenditure is stable, andthat this assumption provides a very poordescription of US data. They show that the totalreal US consumption expenditure has risensharply over the post-war period relative to realconsumption of non-durables and services.This implies that an empirical implementationof equation (A2.4) using this approach willsuffer from significant approximation error andthat this error will tend to increase over time,probably resulting in downward bias andinstability in the estimated parameters. Thethird row of Table A2.1 (labelled “Method 2”)shows that this approximation method doesindeed result in a much poorer fit than themethod based on nominal ratios, with theadjusted R2 dropping to 0.26.

Size of the MPC: Another interesting aspect ofthe regressions reported in the first two rows ofTable A2.1 is that the size of the estimatedMPC out of wealth, in the 0.06 to 0.07 range, issomewhat larger than has been reported insome other studies. For example, in a briefreview of macroeconomic estimates, Poterba(2000) concludes that “values as high as 0.05probably represent the upper range of thecurrent estimates.”

There appear to be two explanations for thehigher estimates of the MPC reported here and in

Palumbo, Rudd, and Whelan (2002). The first isthat many other studies have replaced totalconsumption expenditure with consumption ofnon-durables and services using the Method 2approach just described. Because this approachadds a highly inaccurate approximation error toequation (A2.4), it is hardly surprising that itresults in downward-biased estimates of theMPC: Table A2.1 reports that this methodproduces an MPC of 0.041.

The second explanation is that some otherstudies have used disposable income instead oflabour income in their empirical estimation. Asnoted above, this approach is inconsistent withthe theory underlying the relationshipsbetween consumption, income and wealth:Essentially, it represents a double-counting ofthe effects of capital income. And, as shown inthe fourth row of Table A2.1, this approach alsoresults in a lower estimate of the MPC ofexactly 0.04. Note that if we use the slightlyshorter sample employed in Davis and Palumbo(2001), this exercise can exactly replicate theirpreferred estimate of the MPC of 0.039.

Upward bias (direct/indirect effects):Another issue that has been discussed in theliterature on wealth effects concerns thepotential correlation of wealth with the errorterm in the consumption regression. Poterbaand Samwick (1995) phrase this debate asrelating to whether estimated MPCs out ofwealth reflect a direct effect of wealth onconsumption or indirect effects resulting fromthe correlation of wealth with otherunobservable variables influencingconsumption. However, in terms of theframework underlying equation (A2.4), it doesnot appear that this problem represents a majorsource of bias in the estimates on the first tworows of Table A2.1. Theoretical considerationsimply that the error in this regression relates tocurrent expectations concerning future growthin labour income. This implies that estimates ofthe wealth effect will tend to be biased upwardsif the wealth-to-income ratio,

YAt

t, tends to

positively predict future increases in labourincome. However, regressions with our US

43ECB


APPENDIX 2

quarterly dataset failed to uncover evidencethat the wealth to income ratio “Granger-causes” labour income growth, so concernsover upward bias and “indirect effects” may besomewhat overstated.

Levels versus logs: Another specificationissue is that the regressions in Table A2.1estimate the MPC out of wealth, whereas anumber of other studies (for example, Braytonand Tinsley, 1996, and Ludvigson andSteindel, 1999) estimate log-linearrelationships between consumption, income,and wealth. These studies often translate theirelasticity coefficients into estimated MPCsusing sample averages of the wealth-to-incomeratio. However, given the wide variations overtime in the wealth-to-income ratio, and thetheoretical priors indicating the superiority ofdirect estimation of the MPC, these log-linearestimates should probably be consideredsomewhat unsatisfactory.

Variable expected rates of return: We notedin deriving equation (A2.4) that thisformulation, in which households consume aconstant proportion, r

r+1 , of their wealth

depended on the assumption that the rate ofreturn is constant. As Lettau and Ludvigson(2001) have stressed a more general approach,in which expected returns on assets are allowedto vary over time, predicts that the MPC out ofwealth should move positively over time withexpected future rates of return on assets. Thus,for example, if stock prices fall today becauseof a rise in the required equity premium,consumption need not decline because anincrease in the MPC out of wealth may offsetthe decline in wealth itself.

While these ideas – that the MPC out of wealthmay vary over time, and that consumption mayreact differently to asset price movementsdepending on their source – are intriguing andtheoretically well-founded, at this point theempirical evidence for their relevance for theaggregate wealth effect on consumption isunclear. One interesting piece of evidence isthe finding of Lettau and Ludvigson (2001) that

a proxy for the (unobservable) ratio ofconsumption to total wealth (human plustangible) appears to have forecasting powerboth for stock returns and for the excess returnsof stocks over short-term bonds.

This result, however, provides us with littleclue as to how important potential variations inexpected returns may be in generating changesover time in the MPC out of wealth. On thisissue, Figure A2.2 suggests that (despite thetheoretical likelihood of variations over time inthe MPC) the estimated wealth effect isactually quite stable. The figure showsrecursive estimates of the regression reportedon the second row of Table A2.1, i.e. it reportsestimates from rolling regressions which add adata point for each period. Such regressions areoften used as a diagnostic technique foruncovering parameter instability. The chartshows, however, that the estimated MPC fromaggregate US data does not appear to havechanged greatly over the past thirty years.27

Figure A2.2 Recursive est imates of the MPCout of wealth

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

Estimated MPC +1.96 Std. Error -1.96 Std. Error

1967 1969 1972 1974 1977 1979 1982 1984 1987 1989 1992 1994 1997 1999 2002

27 This stability over time of the estimated MPC contrastsmarkedly with the conclusions of Ludvigson and Steindel(1999), who argue that the estimated wealth effect variessubstantially over time. The principal reason for thisdifference is Ludvigson and Steindel’s use of realconsumption of non-durables together with measures of realincome and wealth obtained by deflating by a price index foraggregate consumption. As discussed in the text, thisapproach adds a large and trending approximation error and soshould be expected to produce unstable parameter estimates.Their study also adopts a log-linear specif ication in contrastto the linear specif ication estimated here.


Estimates of the MPC using shorter moving-sample windows display somewhat morevariation over time, as would be expected, butthere is limited evidence that these variationsare related to changes in expected returns onfinancial assets. For example, Whelan (2003)shows that the MPC out of wealth does notappear to vary significantly over time due tofluctuations in dividend-price ratios, earnings-price ratios, yields spreads or other variablesthat have been considered useful predictors offinancial asset returns. On balance, then, itappears that the jury is still out on the empiricalimportance of fluctuations in expected returnsfor the wealth effect.

Differences in MPCs across assets: Finally,Table A2.2 reports estimates from regressionsthat allow the MPC out of equity wealth to bedifferent from that out of other wealth, i.e. itreports estimates of

(A2.5) εµµαt

t

ot

ot

et

et

t

YA

YA

YC +++=

where e denotes equity wealth and o denotesother wealth. The table repeats a result reportedpreviously in a number of studies (for instance,Brayton and Tinsley, 1996, and Davis andPalumbo, 2001) which is that the MPC out ofequity wealth appears to be smaller than thatout of other wealth. In fact, these resultssuggest that the MPC out of equity wealth isonly about half that out of other wealth. Such aprediction cannot be derived from the type of“representative agent” models usually used inmacroeconomics. However, it seems possiblethat these differences in the MPC across assetclasses may reflect distribution issues. Equitywealth is more likely to be held by richer

Standard errors in parentheses

Income measure Consumption measure µµµµµe

µµµµµo

Adjusted R2

Labour income Total consumption 0.054 (0.002) 0.102 (0.004) 0.87Labour income Scaled non-durables and services (Method 1) 0.060 (0.002) 0.113 (0.005) 0.83

Table A2.2 Empir ical est imates of equation (A2.5) for US data

Note: Sample is 1952:2 to 2002:2. See text for descriptions of Method 1 and Method 2.

households, and studies have traditionallyshown that such households have higher savingrates.

45ECB


APPENDIX 3

APPENDIX 3

BALANCE SHEET CHANNEL

A3.1 THEORETICAL CONSIDERATIONSThe balance sheet channel stresses thepotential impact that the quality of a company’sbalance sheet may have on investmentspending decisions. A rise/fall in asset priceswill improve/weaken the balance sheetposition of a company, thus increasing/reducing its ability to directly fund projects orprovide collateral for external finance. In bothcases, the risk that lenders face is reduced/increased and the availability of creditincreases/reduces. The importance of the linkbetween balance sheet positions andinvestment can be seen in the slow recovery ofthe economy in many OECD countries from the1990/91 recession. A number of analysts28 haveattributed the slow recovery, in part, to thecorporate sector’s attempt to improve weakbalance sheet positions coming from heavycorporate debt burden.

The recent literature on this channel builds onthe theory of capital market imperfections suchas asymmetric information and agency costs.The basic argument is that asymmetricinformation between lenders and borrowersinduces a cost premium for funds raisedexternally in comparison to internal funds. Thispremium, called the external finance premium,reflects lenders’ expected costs of monitoringand evaluating the quality of the company’sinvestment opportunities, costs resulting fromthe fact that the borrower inevitably has betterinformation than the lender or distortions in theborrower’s behaviour coming from moralhazard. As a result, the cost of external finance(debt, equity) may be substantially higher thanthe cost of internal finance (cash flows,liquidity).

The balance sheet channel (often called the networth channel) is based on the idea that theborrower’s external finance premium dependson his financial position. In particular, manyrelated studies focus on the borrower’s net

worth – defined as the borrower’s liquid assetsplus collateral value of illiquid assets – as amain state variable: the greater the level of networth of the potential borrower (or,equivalently, the level of company’s internalfinancing) the lower will be the premium onexternal finance. Shocks that affect net worthcan be an initiating source of real fluctuations,through two transmission mechanisms:

i) One mechanism arises mainly from changesin cash flows. A negative shock reducescurrent output, lowers cash flows, andraises the company’s need for externalfunds. Consequently, the expected agencycost and external finance premium rise, thusdeferring investment and reducing outputand cash flows in subsequent periods.

ii) The other mechanism operates throughasset prices and the value of the collateral.The availability of credit for companiesdepends on the value of their assets. Here,an adverse shock reduces asset prices andthe value of the collateral provided forloans, thus raising external financepremium and reducing investment.

An extensive part of the literature refers to theamplification and propagation of initial real ormonetary shocks brought about by worseningcredit market conditions (as reflected inborrowers’ balance sheet positions), aphenomenon called the financial accelerator.

Bernanke and Gertler (1989) develop an RBCmodel in which the evolution of companies’ networth plays a critical role in macroeconomicdynamics, opening a channel through which theborrower’s balance sheet position affectsinvestment and output. A key virtue of thismodel is the inverse relationship between thepotential borrower’s wealth and the expectedagency costs of the lender-borrowerrelationship, as described earlier. In thismodel, shocks to the economy are amplified

28 Bernanke and Lown (1991), pp. 206 and 211-212, and Tease(1993), p. 47.


and propagated through their impact onborrowers’ cash flows. An adverse exogenousproductivity disturbance, for example, lowerscurrent cash flows (income), reducing theability of the company to finance investmentwith internal funds. Lower entrepreneurialwealth raises the external finance premium,thus impairing investment. Declininginvestment lowers economic activity and cashflows in subsequent periods, thus inducing apersistent investment downturn. Animplication of this model is that the dynamiceffects of such shocks may be asymmetric;specifically, the weaker the initial financialcondition of the borrower, the more powerful isthe propagation effect through cash flows.Additionally, during recessions, there is areallocation of credit funds from low net worthto high net worth borrowers.

Apart from the changes in cash flows,variations in real and financial asset prices area major channel for shocks to the net worth. Inthis direction, Kiyotaki and Moore (1997)develop a dynamic equilibrium model in whichdurable assets (land) are not only factors ofproduction, but they also serve as collateral forloans. In their model, a small negativetemporary shock to productivity lowers thevalue of the collateralised asset. This leads totightened borrowing constraints and cuttingbacks on investment spending. The fall inspending lowers the value of existing assetsand leads to a new round of reduced borrowingand spending, thus propagating the initialshock further through time. A key point of thisstudy is the dynamic interaction between creditlimits and asset prices, which turns out to be apowerful transmission mechanism by whichthe effects of shocks persist, amplify, andspread out.

Bernanke et al. (1999) develop a dynamicgeneral equilibrium model, which exhibits afinancial accelerator framework. The keyinnovation on investment is the relationship:

(A3.1) { } ( ) 1111 ++++ = ttttKt RKQNsRE

Where { }KtRE 1+ is the expected return on capital;

( )s is a function of the ratio of the cost ofexternal finance to the riskless rate, or externalfinance premium; 1+tN represents net worth;

1+tt KQ are total financing requirements; and

1+tR shows the opportunity cost of capital.

This expressions state that, in the presence ofcapital market imperfections, the premium onexternal funds depends inversely on the shareof the company’s capital investment financedby its own net worth, as measured by the ratio( )11 ++ ttt KQN .

Figure A3.129 relates the cost of funds to thelevel of investment of the company,demonstrating these effects. For financingrequirements, up to point F, the company canuse its own net worth (internal funds). The costof these funds is r

1, which is the sum of the risk-

free rate and a risk adjustment factor for thecompany. Without asymmetric information,external funds would also be available for thecompany at the rate r

1. However, asymmetric

information means that, beyond F, the companyacquires external funds at a premium above r

1.

29 Oliner and Rudebusch (1996).

Figure A3.1

Cost of funds

D

S2

S1

S1’

r2

r1

F I2 I1’ I1

Internal funds Investment

47ECB


APPENDIX 3

The premium increases with the level of totalfinance raised externally, since higher debtmeans increased moral hazard and defaultrisks. This link accounts for the upward slopeof S

1. In addition, the external premium

increases with the level of the risk-free rate, asan increase in the interest rate lowers thepresent discounted value of the borrower’scollateral and reduces the current cash flow,worsening the moral hazard problem. In FigureA3.1, an increase in the interest rate shifts thefinance supply schedule from S

1 to S

2 (which is

steeper than S1’). The consequent fall in

investment (from I1 to I

2 rather than I

1’) is

magnified by the increase in the externalfinance premium. This example illustrates thepotential link between investment demand andchanges in net worth, implying that companiesthat can provide a higher collateral or finance agreater proportion of investment with internalfunds will be likely to face a lower externalfinance premium.

A3.2 EXISTING EMPIRICAL EVIDENCE

A large part of the literature provides evidencethat company investment is a function ofliquidity and the quality of a company’sbalance sheet, as implied by the financialaccelerator mechanism. A common perceptionis that differences in investment behaviourbetween companies should come from theirrespective access to capital markets.Generally, it has been found that investment isquite sensitive to financial variables like cashflows or the stock of liquid assets for thosecompanies that face financial constraints. Inthese studies, the size of the company is widelyconsidered as a proxy for capital market access,though other criteria such as the age of thecompany, the dividend pay-out or the coverageratio are also used for this reason. Smallcompanies are considered more likely to beaffected by the balance sheet channel giventheir overall lower level of assets and thehigher collateral requirements for loans (due tohigher informational asymmetries).Consequently, a balance sheet weakening (e.g.due to a monetary policy tightening) might

lower significantly the value of their collateral,thus leading to restricted credit access.

Fazzari et al. (1988) use data from USmanufacturing companies for the period 1970-1984 to study the effect of cash flows oninvestment spending. They divide their samplebased on the dividends to income ratio.Consistently with their perception, their resultsalso show that the investment of companieswhich are believed a priori to be creditconstrained (low-dividend companies) is moresensitive to fluctuations in their cash flow thanhigh-dividend companies.

Gertler and Gilchrist (1994) used quarterlytime series data for all US manufacturingcompanies, classified by size, to examine theresponse of small versus large companies tomonetary tightening. They find that smallcompanies account for a significant part of theconsequent total manufacturing decline andthat they play a prominent role in the slowdownof inventory investment. Looking forasymmetric effects, they find that the effects ofmonetary policy changes are greater on small-company variables during recessions (low-growth states). Their evidence also suggeststhat the coverage ratio (cash flows to interestexpenses, an indicator of balance sheet quality)is positively related to inventory investmentfor small companies, but not for largecompanies.

Oliner and Rudebusch (1996) use QuarterlyFinancial Report data for the US manufacturingsector and divide companies into two classes:small and large companies. Their results showthat after a monetary contraction, theassociation between cash flow (internal funds)and investment tightens significantly for smallcompanies. In contrast, for large companiesthere was no change in this relationship.

In another paper, Guariglia (1998) uses datafrom a large panel of UK manufacturingcompanies for the period 1968-1991 to studythe linkage between internal finance andinventory investment that appears to play an


important role in business cycles. Companiesare classified into more likely and less likely tobe financially constrained in terms of thecoverage ratio that is used as a proxy of theexternal finance premium. The authorestimates several equations for inventoryinvestment growth that include the coverageratio or cash flows as explanatory variables foreach group of companies. The results suggestthat there is a significant link between thefinancial variables and inventory investment,stronger for companies with weak balancesheets, during recessions and periods of tightmonetary policy.

Vermeulen (2000) uses the BACH-database for10 manufacturing industries for the period1983-1997 to study the presence of a financialaccelerator in Germany, France, Italy, andSpain. He proxies the “weak balance sheet”term by using four different indicators: totaldebt as a fraction of total assets (leverage),short-term debt to current assets (liquidity),short-term debt as a fraction of total debt(market access) and the coverage ratio (creditworthiness). The results confirm theexpectation that small company investmentshows greater sensitivity to weak balance sheetpositions, though there is no evidence that thiseffect increases during a downturn. Medium-sized companies are found to be influenced bybalance sheet variables during downturns,rather than outside downturns. For largecompanies there is no evidence of anaccelerator at all. Regarding countrydifferences in the strength of the financialaccelerator, the estimations show that balancesheet effects in downturns are stronger inFrance and Italy than in Germany and Spain.

Chatelain et al. (2001) use large companydatabases for Germany, France, Italy, andSpain for the period 1985-1999 to investigatethe interest rate channel and the broad creditchannel in the four larger euro-area countries.Standardised neo-classical investmentregressions were used, in which investment is afunction of sales, cash flow, and the user cost ofcapital. The authors find that in all these

countries investment is quite sensitive to usercost, sales, and cash flow movements. Theyalso find that only in Italy do smallercompanies show greater sensitivity to cashflow movements (balance sheet conditions),although they argue that classification by sizemight not be a sufficient indicator for allcountries.

Van Ees and Garretsen (1994) study the linkbetween liquidity (internal funds) and businessinvestment for a number of Dutch non-financial companies. Their estimations suggestthat liquidity is of significant importance forDutch business investment and that this impactis stronger than that found in related studies forthe United States, the United Kingdom andJapan. With respect to the criteria used to dealwith asymmetric information, the dividendpolicy, the size and the age of the companies donot seem to be important. However, theexistence of close ties between banks andcompanies, which mitigates informationproblems, turns out to be very significant forthe investment of Dutch companies. In a laterstudy, Van Ees et al. (1998) also examine howdebt constraints influence corporateinvestment of a panel of Dutch companies.They provide evidence that financialconstraints influence the investment behaviourof these companies and show that suchconstraints are especially relevant forcompanies with low dividend pay-outs, whilethere was no similar evidence for smallercompanies and the leverage criterion gavemixed results.

A number of related papers30 provide evidencefor single euro area countries regarding theexistence of a balance sheet (credit) channel aswell as distribution effects across companies.For France (Chatelain and Tiomo, 2001), theexistence of this channel is confirmed forcompanies facing a high risk of bankruptcy,companies in the equipment goods sector andcompanies using trade credit extensively, all of

30 Research conducted within the Monetary TransmissionNetwork.

49ECB


APPENDIX 3

which present high cash-flow sensitivity. ForItaly (Gaiotti and Generale, 2001), the resultsalso show that all financial variables (cashflow, stock of liquidity) affect investmentdecisions. This impact is found to be strongerfor small companies and for companies with ahigher share of intangible assets, which aredifficult to evaluate and cannot be used ascollateral. For Germany (von Kalckreuth,2001), financially constrained companies,proxied in terms of lower-rated companies, arefound to be more sensitive to financialvariables. However, the sample split accordingto size presents no such sensitivity, possiblyindicating the role of the house-bankingsystem31 in Germany. Here, the balance sheetchannel seems to be of secondary importancecompared to the interest rate channel. ForAustria (Valderrama, 2001), the results supportthe existence of a credit channel as well as ofheterogeneity across groups of firms. Theliquidity ratio appears to be the mostsignificant determinant of investment, thoughcompanies with a close relationship with ahouse bank are likely to become less dependenton internal funds. For Belgium (Butzen et al.,2001), the evidence supports the hypothesis ofa credit channel as well as of distributioneffects among companies of different size andsector. For Luxembourg (Lünnemann andMathä, 2001), the idea that the strength of acompany’s balance sheet influencesinvestment behaviour is also confirmed. Youngcompanies seem to be more financiallyconstrained, as their investment is moresensitive to changes in internally generatedliquidity.

A3.3 CONFIDENCE CHANNEL

Stock market volatility may have indirecteffects on the real economy through its impacton consumer and business confidence. A fall(rise) in stock prices may signal poor (good)economic prospects, altering marketparticipants’ assessments of economic risks.Consequently, their incentive to spend will bereduced (increased), thus affectingconsumption and corporate investment. For

example, a large fall in stock prices may forcecompanies, even those that have not issuedstocks, to revise their profit expectations andreduce or postpone investment. In addition,falling stock prices and weak businessconfidence may increase investors’ riskaversion. Investors may subsequently reducetheir holdings in risky assets and ask for higherrisk premiums, which may be passed on tobusiness in terms of higher interest rates andlower equity prices, with a negative impact oninvestment.32

31 This refers to the establishment of a long-term lendingrelationship between a company and a bank, called the housebank. This practice may help companies overcome liquidityconstraints, thus weakening the balance sheet channel.

32 IMF, World Economic Outlook, December 2001.

Figure A3.2 Industrial production 1 andindustr ia l conf idence2 in the euro area

Sources: European Commission Business and ConsumerSurveys and ECB.1) Annual percentage changes, excluding construction.2) Percentage balances, seasonally adjusted.

This channel is consistent with the forward-looking nature of stock prices and their leadingindicator property for real economic activitythat has been empirically found for the euroarea. As stock prices reflect marketparticipants’ expectations for future corporateearnings, which are closely related to economicactivity, stock prices help predict futureeconomic growth.

While the potential mechanism of a confidenceeffect on the economy is comprehensible, the

-20

-15

-10

-5

0

5

10

1999 2000 2001 2002-6

-4

-2

0

2

4

6

8

10

Industrial confidence (left-hand scale)Industrial production (right-hand scale)


assessment of this effect is rather complicated.This is because confidence is a feeling and notan action, which is difficult to assessquantitatively. Generally, confidence ismeasured by surveys that include questions onthe assessment of current conditions and futureexpectations. Several of the constructedindicators have been found to possess goodleading indicator properties for the businesscycle and can be considered as adequate forresearch on macroeconomic issues.

Figure A3.2 depicts industrial confidence andindustrial production for the period 1999-2002in the euro area. In particular, industrialconfidence is found to be closely linked to thebusiness cycle component (industrialproduction) (correlation = 0.93).

Figure A3.3 depicts the development of theEuroStoxx broad index, industrial confidenceand consumer confidence in the euro area forthe same period. In several periods confidencefollows similar patterns to the stock pricemeasure (with high positive correlation in bothcases). In particular, the turning points of thestock market index tend to lead the turningpoints of confidence, showing that stock pricesand confidence co-move quite closely

Figure A3.3 Dow Jones EURO STOXX Broad1,industr ia l and consumer conf idence2 in theeuro area

Sources: European Commission Business and ConsumerSurvey and ECB.1) Monthly averages.2) Percentage balances, seasonally adjusted.

(especially in the case of industrialconfidence).

Arnold et al. (2002) and Jansen and Nahuis(2003) confirm the predictive power of shareprices for confidence indicators. They find thatequity prices lead confidence indicators in alarge number of European countries and theUnited States. This link, however, may simplyreflect more direct wealth and cost of capitaleffects and does not necessarily imply a causalrelationship between stock prices andconfidence.

Jansen and Nahuis (2003) find empiricalsupport for causality from stock market toconfidence. They also suggest that thisrelationship is driven by expectations abouteconomy-wide conditions rather than personalfinances and that the confidence channel is aseparate transmission channel and not anadjunct to conventional wealth effects. Thisconclusion may hold for studies based onmicro-level data, although, in the case ofaggregate time-series data, the estimation ofwealth effects also reflects expectations aboutthe future.

In this context, confidence could be a catalystfor an impact from stock prices on consumptionand investment spending. It is, however,difficult to disentangle empirically the relativeindirect impact of confidence on actualspending behaviour. In this regard, Figure A3.4depicts industrial confidence in relation toinvestment growth, suggesting that there wereperiods when investment growth did not followsignificant movements in industrialconfidence.

Impulse response analysis (by the ECB) of therelationship between stock prices andconfidence indicators, indicates that the effectof stock price changes is reflected after sometime in confidence indicators and dies out onlyafter several years. The effects on industrialconfidence appear and die out faster than thoseon consumer confidence. Generally, in this

-20

-15

-10

-5

10

200110

150

200

250

300

350

450

500

-5

0

5

10

1999 2000 2002100

400

Industrial confidence (left-hand scale)Consuner confidence (left-hand scale)Euro Stoxx broad (right-hand scale)

51ECB


APPENDIX 3

Fall in Impact onconfidence real GDP

One Two Aug.-Oct. if fall inCurrent quarter quarter Cumulative (standard confidencequarter lag lag impact deviations) is sustained

Impact of a one standard deviation fall in the monthly change in business confidence on real GDPUnited States– Conference Board -0.18*** -0.08* -0.07* -0.31 3.2 -1.0– NAPM -0.10** -0.04 -0.09** -0.23 4.2 -1.0Japan -0.14* -0.07 -0.06 -0.27 n.a. n.a.Germany -0.10*** -0.05 -0.05 -0.20 3.7 -0.7France -0.12*** -0.02 -0.01 -0.15 1.9 -0.3Italy -0.11** -0.09* -0.02 -0.22 3.0 -0.7United Kingdom -0.11** -0.08* -0.04* -0.23 0.5 -0.1Canada -0.05 -0.11* -0.06 -0.22 n.a. n.a.

Impact of a one standard deviation fall in the monthly change in consumer confidence on real GDPUnited States– Conference Board -0.14*** -0.08* -0.15*** -0.37 5.0 -1.9– Univ. of Michigan -0.09* -0.09* -0.09** -0.27 2.8 -0.8

Table A3.1 Impact of conf idence indices on the growth in real GDP

(% of GDP, unless otherwise indicated)

Source: IMF, World Economic Outlook, December 2001.Notes: Three asterisks indicate the results are signif icant at the 1% level, two asterisks at the 5% level, and one asterisk the 10%level. The coeff icients were calculated in such a way that they would be cumulated over time.

analysis, the adjustment of confidence topermanent stock price changes seems ratherslow and limited.

Several analyses of the impact of consumerconfidence on spending in the United Statesand the United Kingdom find that it has someexplanatory power, but the available evidencefor the euro area and business confidence arerather scarce. Table A3.1 summarises theresults of an analysis undertaken by the staff ofthe IMF on the business and consumerconfidence indices. The exercise looked intothe impact of confidence indices on growth inreal GDP for Canada, France, Italy, Japan, theUnited Kingdom and the United States. Theresults suggest that in the United States bothbusiness and consumer confidence indicatorsprovide useful information about output of thecurrent and the following two quarters. For theother countries (except Canada), only thebusiness confidence series seem to providesuch information. It was also shown that, if thefall in confidence was sustained, output couldfall by one percentage point in the UnitedStates, 0.7 in Germany and Italy, 0.3 in France

and 0.1 in the United Kingdom. In addition, itwas found that when other variables (stockprice change, interest rate movements, and lastperiod’s growth) were taken intoconsideration, the confidence effect on outputgrowth was generally reduced.


REFERENCES

Alessie, R.J.M., S. Hochguertel and A. Van Soest (2002), “Household portfolios in theNetherlands”, in L. Guiso, M. Haliassos and T. Jappelli, Household Portfolios, The MITPress, Cambridge, MA, pp. 341-388.

Alonso, C. and S. Bentolila (1992), “The relationship between investment and Tobin’s Q inSpanish industrial firms”, Documento de Trabajo nº 9203, Banco de España.

Andersen, M. and R. Subbaraman (1996), “Share prices and investment”, Reserve Bank ofAustralia Research Discussion Paper No. 9610.

Aoki, K., J. Proudman and G. Vlieghe (2002), “Houses as collateral: Has the link between thehouse prices and consumption in the UK changed”, Economic Policy Review, 8(1) FederalReserve Bank of New York, May.

Arnold, I., P. van Els and J. de Haan (2002), “Wealth effects and monetary policy”, DeNederlandsche Bank Research Memorandum No. 719.

Assarsson, B., C. Berg and P. Jansson, (2004) “Investment in Swedish Manufacturing: Analysisand Forecasts”, Empirical-Economics, 29(2), pp. 261-80.

Attanasio, O. (1995), “Intertemporal Allocation of Consumption: Theory and Evidence”,Carnegie Rochester Conference Series on Public Policy, 42(0), pp. 39-89.

Attanasio, O., J. Banks, C. Meghir and G. Weber (1999), “Humps and bumps in lifetimeconsumption”, Journal of Business and Economic Statistics, 17(1), pp. 22-35.

Attanasio, O., J. Banks and S. Tanner (1998), “Asset Holdings and Consumption Volatility”,NBER Working Paper No. 6567, May.

Audretsch, D.B. and J.A. Elston (2002), “Does Firm Size Matter? Evidence on the Impacts ofLiquidity Constraints on Firm Investment Behavior in Germany”, International Journal ofIndustrial Organization, 20(1), pp. 1-17.

Babeau, A. and T. Sbano (2002), “Household wealth in the national accounts of Europe, theUnited States and Japan”, presented at the OECD Meeting of National Accounts Experts,October.

Baker, M., J. Stein and J. Wurgler (2002), “When does the market matter? Stock prices and theinvestment of equity-dependent firms”, NBER Working Paper No. 8750.

Balmaseda, M. and P. Tello (2002), “¿Han cambiado los determinantes del consumo privado enEspaña?” (Have determinants of consumption changed in Spain?), Situación Española,BBVA, July, pp. 33-39.

Banco do Portugal (2003), “Request for information on research on wealth effects onconsumption in EU countries. 2000 Household Wealth Survey”, January.

53ECB


REFERENCES

Barot, B. and Z. Yang (2002), “House Prices and Housing Investment in Sweden and the UK:Econometric Analysis for the Period 1970–1998”, Review of Urban & Regional DevelopmentStudies, 14, pp. 189-216.

Bayoumi, T. and H. Edison (2002), “Is Wealth Increasingly Driving Consumption?”, IMFWorking Paper, October.

Bernanke, B. and M. Gertler (1989), “Agency Costs, Net Worth and Business Fluctuations”,American Economic Review, 79, pp. 14-31.

Bernanke, B.S. and C.S. Lown (1991), “The Credit Crunch”, Brookings Papers on EconomicActivity, 0(2), pp. 204-39.

Bernanke, B., M. Gertler and S. Gilchrist (1999), “The Financial Accelerator in a QuantitativeBusiness Cycle Framework”, in J.B. Taylor and M. Woodford (eds.), Handbook ofMacroeconomics, Volume 1C, Handbooks in Economics, vol. 15., Amsterdam, New York andOxford: Elsevier Science, North-Holland, pp. 1341-93.

Bertaut, C. (2002), “Equity Prices, Household Wealth, and Consumption Growth in ForeignIndustrial Countries: Wealth Effects in the 1990s”, Federal Reserve Board InternationalFinance Discussion Paper No. 724.

Blanchard, O.J. (1985), “Debt, deficits, and finite horizons”, Journal of Political Economy, 93,pp. 223-247.

Blanchard, O.J, C. Rhee and L.H. Summers (1993), “The stock market, profit and investment”,Quarterly Journal of Economics, February, pp. 115-136.

Blundell, R., M. Browning and C. Meghir (1994), “Consumer Demand and the Life-CycleAllocation of Household Expenditures”, Review of Economic Studies, 61(1), pp. 57-80.

Bond, S. and J. Cummins (2001), “Noisy share prices and the Q model of investment”, IFSWorking Papers W01/22, Institute for Fiscal Studies.

Boone, L., C. Giorno and P. Richardson (1998), “Stock Market Fluctuations and ConsumptionBehaviour: Some Recent Evidence”, OECD Economic Department Working Papers No. 208.

Boone L., N. Girouard and I. Wanner (2001), “Financial market liberalisation, wealth andconsumption”, OECD Economics Department Working Papers No. 308.

Brady, P.J., G.B. Canner and D.M. Maki (2000), “The Effects of Recent Mortgage Refinancing”,Federal-Reserve-Bulletin, 86(7), pp. 441-50.

Brav, A., G.M. Constantinides and C.C. Geczy (1999), “Asset Pricing with HeterogeneousConsumers and Limited Participation: Empirical Evidence”, Rodney L. White Center forFinancial Research Working Paper No. 23.

Brayton, F. and P. Tinsley (1996), “A Guide to FRB/US: A Macroeconomic Model of the UnitedStates”, Federal Reserve Board Finance and Economics Discussion Series No. 42.


Butzen, P., C. Fuss and P. Vermeulen (2001), “The interest rate and credit channel in Belgium: aninvestigation with micro-level firm data”, ECB Working Paper No. 107.

Butzen, P., C. Fuss and P. Vermeulen (2003), “Business investment and monetary transmission inBelgium”, in Angeloni I., A. Kashyap and B. Mojon (eds.), “Monetary Policy Transmission inthe Euro Area”, United Kingdom, pp. 162-172.

Byrne, J. and E. Philip Davis (2003), “Disaggregate Wealth and Aggregate Consumption – Aninvestigation of empirical relationships for the G7”, Oxford Bulletin of Economics andStatistics, 65(2), pp. 197-220(24).

Campbell, J.Y. (1996), “Understanding Risk and Return”, Journal of Political Economy, 104, pp.298-345.

Campbell, J.Y. (2001), “Consumption-Based Asset Pricing”, unpublished paper, HarvardUniversity.

Campbell, J.Y. and J.H. Cochrane (1999), “By force of habit: A consumption-base explanation ofaggregate stock market behaviour”, Journal of Political Economy, 107, pp. 205-251.

Campbell, J.Y. and M. Goto (2002), “Efficient Tests of Stock Return Predictability”, unpublishedpaper, Harvard University.

Campbell, J.Y., A.W. Lo and C. MacKinlay (1997), The Econometrics of Financial Markets,Princeton University Press, Princeton, NJ.

Campbell, J.Y. and N.G. Mankiw (1989), “Consumption, Income, and Interest Rates:Reinterpreting the Time Series Evidence”, in O.J. Blanchard and S. Fisher (eds.), NBERMacroeconomic Annual, pp. 185-216.

Campbell, J.Y. and R.J. Shiller (1988), “The Dividend-Price Ratio and Expectations of FutureDividends and Discount Factors”, Review of Financial Studies, 1, pp. 195-227.

Campbell, J.Y. and L.M. Viceira (1999), “Consumption and portfolio decisions when expectedreturns are time varying”, The Quarterly Journal of Economics, May, pp. 433-495.

Carroll, C. and M. Kimball (1996), “On the Concavity of the Consumption Function”,Econometrica, 64, pp. 981-992.

Case, K.E., J.M. Quigley and R.J. Shiller (2001), “Comparing Wealth Effects: The Stock MarketVersus The Housing Market”, NBER Working Paper No. 8606.

Chatelain J.B., A. Generale, I. Hernando, U. von Kalckreuth and P.Vermeulen (2003), “Firminvestment and monetary transmission in the euro area”, in Angeloni I., A. Kashyap and B.Mojon (eds.), “Monetary Policy Transmission in the Euro Area”, Cambridge University Press,United Kingdom, pp. 133-161.

Chatelain, J.B. and A. Tiomo (2001), “Investment, the cost of capital, and monetary policy in thenineties in France: a panel data investigation”, ECB Working Paper No. 106.

55ECB


REFERENCES

Chatelain, J.B. and A. Tiomo (2003), “Monetary policy and corporate investment in France”, inAngeloni I., A. Kashyap and B. Mojon (eds.), “Monetary Policy Transmission in the EuroArea”, Cambridge University Press, United Kingdom, pp. 187-197.

Chirinko, R.S. (1993), “Business Fixed Investment Spending: Modeling Strategies, EmpiricalResults, and Policy Implications”, Journal of Economic Literature, 31(4), pp. 1875-1911.

Chirinko, R. S and H. Schaller (2001), “Business Fixed Investment and “Bubbles”: The JapaneseCase”, American Economic Review, 91(3), pp. 663-80.

Choi, J.J., S. Hauser, and K.J. Kopecky (1999), “Does the Stock Market Predict Real Activity?Time Series Evidence from the G-7 Countries”, Journal of Banking and Finance, 23, pp. 1771-1792.

Clapham, E., A. Hyytinen and K. Takala (2002), “Household Wealth, Credit and Consumption:Evidence from Finland and Sweden”, mimeo, Bank of Finland.

Cochrane, J.H. (2001), Asset Pricing, Princeton University Press, Princeton, NJ.

Constantinides, G. (1990), “Habit formation: A solution to equity premium puzzle”, Journal ofPolitical Economy, 98, pp. 519-543.

Darby, J., J. Ireland, C. Leith and S. Wren-Lewis (1999), “COMPACT: a rational expectations,intertemporal model of the United Kingdom economy”, Economic Modelling, 16, pp. 1-52.

Davis, M., and M.G. Palumbo (2001), “A Primer on the Economics and Time Series Econometricsof Wealth Effects”, Federal Reserve Board Finance and Economics Discussion Series No. 09.

De Nederlandsche Bank (1999), “The Dutch Housing and Mortgage Markets: A risk analysis”Boletín Económico, November, Quarterly Bulletin, September.

Danmarks Nationalbank (2003), MONA – A quarterly model of the Danish Economy.

Deutsche Bundesbank (2002), “The impact of stock prices on private consumption – wealtheffects in Germany”, mimeo.

Dynan, K. and D.M. Maki (2001), “Does Stock Market Wealth Matter for Consumption?”,Federal Reserve Board Finance and Economics Discussion Series No. 23.

Dynan, K., J. Skinner and S. Zeldes (2000), “Do the rich save more?”, NBER Working PaperNo. 7906.

Edison, H. and T. Sløk (2001), “Wealth Effects and the New Economy”, IMF Working PaperNo. 77.

Elder R. and J. Power (2002), “The role of housing wealth in the consumption equation”, mimeo,Bank of England.


van Ees, H. and H. Garretsen (1994), “Liquidity and Business Investment: Evidence from DutchPanel Data”, Journal of Macroeconomics, 16, pp. 613-627.

van Ees, H., H. Garretsen, L. de Haan and E. Sterken (1998), “Investment and Debt Constraints:Evidence from Dutch Panel Data”, in S. Brakman, H. van Ees and S.K. Kuipers (eds.), MarketBehaviour and Macroeconomic Modelling, MacMillan, London, pp. 159-179.

Engelhardt, G.V. (1995), “House Prices and Home Owner Saving Behavior”, NBER WorkingPaper No. 5183.

Estrada, A., J.L. Fernández, E. Moral and A.V. Regil (2004), “A Quarterly MacroeconometricModel of the Spanish Economy”, Banco de España, Documentos de Trabajo no

0413.

European Central Bank (2003), Structural Factors in the EU Housing Markets.

Fama, E.F. (1990), “Stock Returns, Expected Returns, and Real Activity”, The Journal ofFinance, 45, pp. 1089-1108.

Fama, E.F. and K.R. French (1988), “Dividend Yields and Expected Stock Returns”, Journal ofFinancial Economics, 22, pp. 3-27.

Faruquee, H., P. Isard, D. Laxton, E. Prasad and B. Turtelboom (1998), “MULTIMOD Mark III,the core dynamic and steady-state models”, IMF Occasional Paper No. 164.

Fazzari, S., G. Hubbard and B. Petersen (1988), “Financing Constraints and CorporateInvestment”, Brookings Papers on Economic Activity, 1, pp. 141-195.

Fuhrer, J.C. (2000), “Habit Formation in Consumption and its Implications for Monetary-PolicyModels”, American Economic Review, 90(3), pp. 367-390.

Gaiotti, E. and A. Generale (2001) “Does monetary policy have asymmetric effects? A look at theinvestment decisions of Italian firms”, ECB Working Paper No. 110.

Gaiotti. E. and A. Generale (2003) “Monetary policy and firms’ investment in Italy” , in AngeloniI., A. Kashyap and B. Mojon (eds.), “Monetary Policy Transmission in the Euro Area”, UnitedKingdom, pp. 198-211.

Gertler, M. and S. Gilchrist (1994), “Monetary policy, business cycles and the behavior of smallmanufacturing firms”, The Quarterly Journal of Economics, 109, pp. 309-340.

Girouard, N. and S. Blöndal (2002), “House prices and economic activity”, OECD Working PaperNo. 279.

Guiso, L., M. Haliassos and T. Jappelli (2002), Household Portfolio, MIT Press.

Guiso, L., M. Paiella and I. Visco (2004), “Do capital gains affect consumption? Estimates ofwealth effects from Italian households’ behaviour”, mimeo.

57ECB


REFERENCES

Goyal, V.K. and T. Yamada (2004), “Asset prices, Financial Constraints and Investment:Evidence from Japan”, Journal of Business, 77(1), pp. 175-200.

Guariglia, A. (1998), “The effects of financial constraints on inventory investment: evidencefrom a panel from UK firms”, Economica, 66, pp. 43-62.

Hall, R. (1978), “Stochastic Implications of the Life Cycle-Permanent Income Hypothesis:Theory and Evidence”, Journal of Political Economy, 96, pp. 971-87.

Hamburg, B., M. Hoffmann and J. Keller (2005), “Consumption, wealth and business cycle: Whyis Germany different?”, Deutsche Bundesbank, Economic Research Center Discussion PaperNo. 1 16/2005.

Hayashi, F. (1982a), “Tobin’s marginal and average Q: A Neoclassical interpretation”,Econometrica, 50(1), pp. 213-244.

Hayashi, F. (1982b), “The Permanent Income Hypothesis: Estimation and Testing byInstrumental Variables”, Journal of Political Economy, 90, pp. 895-915.

Hodrick, Robert (1992), “Dividend Yields and Expected Stock Returns: Alternative Proceduresfor Inference and Measurement”, Review of Financial Studies, 5, pp. 357-386.

International Monetary Fund (2000), “Asset Prices and the Business Cycle”, World EconomicOutlook, May.

International Monetary Fund (2001), World Economic Outlook, December, Chapter II.

International Monetary Fund (2002), “Is wealth increasingly driving consumption?”, WorldEconomic Outlook, April, Chapter II.

Jansen, J.W. and J.N. Nahuis (2003), “The stock market and consumer confidence: Europeanevidence”, Economics Letters, 79, pp. 89-98.

Juster, F.T., J. Lupton, J.P. Smith and F. Stafford (1999), “Savings and wealth; then and now”,mimeo, University of Michigan, October.

Kenny, G. (2003), “Asymmetric adjustment costs and the dynamics of housing supply”,Economic Modelling, 20, pp. 1097-1111.

Kiyotaki, N. and J. Moore (1997), “Credit cycles”, Journal of Political Economy, 105(2), pp. 211-248.

von Kalckreuth, U. (2001), “Monetary transmission in Germany: new perspectives on financialconstraints and investment spending”, ECB Working Paper No. 109.

Labhard, V., G. Sterne and C. Young (2005), “The wealth effects on consumption inindustrialised countries”, Bank of England working paper, (forthcoming).


Lamont, O. (1998), “Earnings and Expected Returns”, Journal of Finance, 53, pp. 1563-1587.

LeRoy, S. and R. Porter (1981), “The Present Value Relation: Tests Based on Variance Bounds”,Econometrica, 49, pp. 555-557.

Lettau, M. and S.C. Ludvigson (2002), “Understanding Trend and Cycle in Asset Values”,revised July, (http://www.econ.nyu.edu/user/ludvigsons/eqprem1.pdf).

Lettau, M. and S.C. Ludvigson (2001), “Consumption, Aggregate Wealth and Expected StockReturns”, Journal of Finance, 56(3), pp. 815-849.

Lettau, M., S. Ludvigson and N. Barczi (2001), “A Primer on the Economics and Time SeriesEconometrics of Wealth Effects: A Comment”, Federal Reserve Bank of New York, (http://www.econ.nyu.edu/user/ludvigsons).

Lewellen, J.W. (2001), “Predicting Returns With Financial Ratios”, unpublished paper, MITSloan School of Management.

Ludvigson, S. and C. Steindel (1999), “How Important Is the Stock Market Effect onConsumption?”, Federal Reserve Bank of New York Policy Review, 2(5), pp. 29-51.

Ludvigson, S., C. Steindel and M. Lettau (2002), “Monetary Policy Transmission through theConsumption-Wealth Channel”, FRBNY Economic Review, May, pp. 117-133.

Ludwig, A. and T. Sløk (2001), “Impact of Stock Prices and House Prices on Consumption inOECD Countries”, IMF Working Paper No. 1.

Lünnemann, P. and T. Mathä (2001), “Monetary transmission: empirical evidence fromLuxembourg firm level data”, ECB Working Paper No. 111.

Lünnemann, P. and T. Mathä (2003), “Monetary transmission: empirical evidence fromLuxembourg firm level data”, in Angeloni I., A. Kashyap and B. Mojon (eds.), “MonetaryPolicy Transmission in the Euro Area”, Cambridge University Press, United Kingdom,pp. 212-220.

Maclennan, D., J. Muellbauer and M. Stephens (1998), “Asymmetries in housing and financialmarket institutions and EMU”, Oxford Review of Economic Policy, 14, pp. 54-80.

Maki, D.M. and M.G. Palumbo (2001), “Disentangling the Wealth Effect - A Cohort Analysis ofHousehold Saving in 1990s”, Federal Reserve Board Finance and Economics DiscussionSeries No. 23.

Mankiw, N.G. and S.P. Zeldes (1990), “The Consumption of Stockholders and Non-Stockholders”, NBER Working Paper No. 3402.

Mauro, P. (2000), “Stock Returns and Output growth in Emerging and Advanced Economies”,IMF Working Paper No. 89.

59ECB


REFERENCES

Mehra, Y.P. (2001), “The Wealth Effect in Empirical Life-Cycle Consumption Equations”,Federal Reserve Bank of Richmond Economic Quarterly, 87(2), pp. 45-68.

Merrigan, P. and M. Normandin (1996), “Precautionary Saving Motives: An Assessment fromUK Time Series of Cross-Sections”, Economic Journal, 106(438), pp. 1193-1208.

Millard and J. Power (2002), “The effect of shocks to equity prices on consumption”, iManageNo. 181748.

Modigliani, F. (1977), “Consumer Spending and Monetary Policy: The Linkages”, FederalReserve Bank of Boston Conference Series, 5, pp. 9-84.

Morck, R., A. Schleifer and R.W. Vishny (1990), “Stock Market and Investment: Is the Market aSideshow”, Brookings Papers on Economic Activity, 2, pp. 157-202.

Muellbauer, J. (2004), “Property and Land, Taxation and the Economy after the Barker Review”,mimeo, Nuffield College, Oxford, July.

Muellbauer, J. and R. Lattimore (1995), “The consumption function: A theoretical and empiricaloverview”, in M.H. Pesaran and M. Wickens (eds.), Handbook of Applied Econometrics:Macroeconomics, Blackwell Publishing, Oxford, pp. 221-311.

National Bank of Belgium (2003), “Answer to the request for information on research on wealtheffects on consumption in EU countries”, January.

Oliner, S. and G. Rudebusch (1996), “Is there a broad credit channel for monetary policy?”,Federal Reserve Bank of San Francisco Economic Review, 1, pp. 3-13.

Otto, M.W. (1999), “Consumer Sentiment and the Stock Market”, Federal Reserve Board Financeand Economics Discussion Series No. 60.

Palumbo, M., J. Rudd and K. Whelan (2002), “On Relationships between Real Consumption,Income, and Wealth”, Federal Reserve Board Finance and Economics Discussion SeriesNo. 38.

Paiella, M. (2003), “Does Wealth Affect Consumption? Evidence for Italy”, mimeo, Bank ofItaly, Research Department.

Parker, J.A. (1999a), “Spendthrift in America? On Two Decades of Decline in the US SavingRate”, in B. Bernanke and J. Rotemberg (eds.), NBER Macroeconomics Annual, The MITPress, Cambridge, MA.

Parker, J.A. (1999b), “The Reaction of Household Consumption to Predictable Changes in SocialSecurity Taxes”, American Economic Review, 89(4), pp. 959-973.

Pesaran, M.H, Y. Shin and R.P. Smith (1999), “Pooled mean group estimation of dynamicheterogeneous panels”, Journal of the American Statistical Association, 94, pp. 621-634.


Poterba, J. (2000), “Stock Market Wealth and Consumption”, Journal of Economic Perspectives,14(2), pp. 99-119.

Poterba, J. and A. Samwick (1995), “Stock Ownership Patterns, Stock Market Fluctuations, andConsumption”, Brookings Papers on Economic Activity, 2, pp. 295-372.

Reifschneider, D., R. Tetlow and J. Williams (1999), “Aggregate Disturbances, Monetary Policyand the Macroeconomy: The FRB/US Perspective”, Federal Reserve Bulletin, 85, pp. 1-19.

Romer, C. (1990), “The Great Crash and the Onset of the Great Depression”, Quarterly Journal ofEconomics, 105, pp. 597-624.

Rudd, J. and K. Whelan (2002), “A Note on the Cointegration of Consumption, Income, andWealth”, Federal Reserve Board Finance and Economics Discussion Series No. 53.

Sefton, J.A. and J.W. in’t Veld (1999), “Consumption and Wealth: An InternationalComparison”, Manchester-School, 67(4), pp. 525-44.

Shiller, R.J. (1981), “Do Stock Prices Move Too Much to be Justified by Subsequent Changes inDividends?”, American Economic Review, 71, pp. 421-436.

Shiller, R. (1984), “Stock Prices and Social Dynamics,” Brookings Papers on Economic Activity,2, pp. 457-498.

Starr-McCluer, M. (1998), “Stock Market Wealth and Consumer Spending”, Federal ReserveBoard Governors, April.

Smets, F. and R. Wouters (2002), “An estimated stochastic dynamic general equilibrium model ofthe euro area”, ECB Working Paper No. 171.

Sundaresan, S. (1989), “Intertemporally dependent preferences and the volatility of consumptionwealth, Review of Financial Studies 2, pp. 73-89.

Tease, W. (1993), “The Stock Market and Investment”, OECD Economic Studies, 20, pp. 41-63.

Thaler, R. (1994), “Psychology and Saving Policies”, American Economic Review, 84, pp. 186-192.

Thomas, P. (2002), “Share ownership and household spending in the United Kingdom”, I-manageNo. 160692.

Tobin, J. (1969), “A general equilibrium approach to monetary theory”, Journal of Money, Creditand Banking, 1, pp. 15–29.

Valderrama, M.T. (2001), “Credit channel and investment behavior in Austria: a micro-econometric approach”, ECB Working Paper No. 108.

61ECB


REFERENCES

Valderrama, M.-T. (2003), “The role of trade credit and bank lending relationships in thetransmission mechanism in Austria”, in Angeloni I., A. Kashyap and B. Mojon (eds.),“Monetary Policy Transmission in the Euro Area”, Cambridge University Press, UnitedKingdom, pp. 221-232.

Vermeulen, P. (2000), “Business fixed investment: Evidence of a financial accelerator inEurope”, ECB Working Paper No. 37.

Vissing-Jørgensen, A. (1999), “Limited Stock Market Participation and the Equity PremiumPuzzle”, mimeo, University of Chicago, December.

Whelan, K. (2003), “Market Valuations and the Wealth Effect on Consumption”, paper presentedat The Irish Economic Association 17th Annual Conference.

Willman, A. (2003), “Consumption, habit persistence, imperfect information and the lifetimebudget constraint”, ECB Working Paper No. 251.

Willman, A. and A. Estrada (2002), “The Spanish block of the ESCB-Multi-Country Model”,Documento de Trabajo nº 0212, Banco de España, and ECB Working Paper No. 149.

Willman, A., M. Kortelainen, H.-L. Männistö and M. Tujula (2000), “The BOF5 macroeconomicmodel of Finland, structure and dynamic microfoundations”, Economic Modelling, 17, pp.275-303.

Zeldes, S.P. (1989), “Optimal Consumption with Stochastic Income: Deviations from CertaintyEquivalence”, Quarterly Journal of Economics, 104, pp. 275-298.

63ECB


EUROPEAN CENTRAL BANKOCCASIONAL PAPER SERIES

1 “The impact of the euro on money and bond markets” by J. Santillán, M. Bayle andC. Thygesen, July 2000.

2 “The effective exchange rates of the euro” by L. Buldorini, S. Makrydakis and C. Thimann,February 2002.

3 “Estimating the trend of M3 income velocity underlying the reference value for monetarygrowth” by C. Brand, D. Gerdesmeier and B. Roffia, May 2002.

4 “Labour force developments in the euro area since the 1980s” by V. Genre andR. Gómez-Salvador, July 2002.

5 “The evolution of clearing and central counterparty services for exchange-tradedderivatives in the United States and Europe: a comparison” by D. Russo,T. L. Hart and A. Schönenberger, September 2002.

6 “Banking integration in the euro area” by I. Cabral, F. Dierick and J. Vesala,December 2002.

7 “Economic relations with regions neighbouring the euro area in the ‘Euro Time Zone’” byF. Mazzaferro, A. Mehl, M. Sturm, C. Thimann and A. Winkler, December 2002.

8 “An introduction to the ECB’s survey of professional forecasters” by J. A. Garcia,September 2003.

9 “Fiscal adjustment in 1991-2002: stylised facts and policy implications” by M. G. Briotti,February 2004.

10 “The acceding countries’ strategies towards ERM II and the adoption of the euro:an analytical review” by a staff team led by P. Backé and C. Thimann and includingO. Arratibel, O. Calvo-Gonzalez, A. Mehl and C. Nerlich, February 2004.

11 “Official dollarisation/euroisation: motives, features and policy implications of currentcases” by A. Winkler, F. Mazzaferro, C. Nerlich and C. Thimann, February 2004.

12 “Understanding the impact of the external dimension on the euro area: trade, capital flows andother international macroeconomic linkages“ by R. Anderton, F. di Mauro and F. Moneta,March 2004.

13 “Fair value accounting and financial stability” by a staff team led by A. Enria and includingL. Cappiello, F. Dierick, S. Grittini, A. Maddaloni, P. Molitor, F. Pires and P. Poloni,April 2004.

14 “Measuring Financial Integration in the Euro Area” by L. Baele, A. Ferrando, P. Hördahl,E. Krylova, C. Monnet, April 2004.


15 “Quality adjustment of European price statistics and the role for hedonics” by H. Ahnert andG. Kenny, May 2004.

16 “Market dynamics associated with credit ratings: a literature review” by F. Gonzalez, F. Haas,R. Johannes, M. Persson, L. Toledo, R. Violi, M. Wieland and C. Zins, June 2004.

17 “Corporate ‘Excesses’ and financial market dynamics” by A. Maddaloni and D. Pain, July 2004.

18 “The international role of the euro: evidence from bonds issued by non-euro area residents” byA. Geis, A. Mehl and S. Wredenborg, July 2004.

19 “Sectoral specialisation in the EU a macroeconomic perspective” by MPC task force of theESCB, July 2004.

20 “The supervision of mixed financial services groups in Europe” by F. Dierick, August 2004.

21 “Governance of securities clearing and settlement systems” by D. Russo, T. Hart,M. C. Malaguti and C. Papathanassiou, October 2004.

22 “Assessing potential output growth in the euro area: a growth accounting perspective”by A. Musso and T. Westermann, January 2005.

23 “The bank lending survey for the euro area” by J. Berg, A. van Rixtel, A. Ferrando,G. de Bondt and S. Scopel, February 2005.

24 “Wage diversity in the euro area: an overview of labour cost differentials acrossindustries” by V. Genre, D. Momferatou and G. Mourre, February 2005.

25 “Government debt management in the euro area: recent theoretical developments and changesin practices” by G. Wolswijk and J. de Haan, March 2005.

26 “The analysis of banking sector health using macro-prudential indicators” by L. Mörttinen,P. Poloni, P. Sandars and J. Vesala, March 2005.

27 “The EU budget – how much scope for institutional reform?” by H. Enderlein, J. Lindner, O.Calvo-Gonzalez, R. Ritter, April 2005.

28 “Reforms in selected EU network industries” by R. Martin, M. Roma, I. Vansteenkiste,April 2005.

29 “Wealth and asset price effects on economic activity”, by F. Altissimo, E. Georgiou, T.Sastre, M. T. Valderrama, G. Sterne, M. Stocker, M. Weth, K. Whelan, A. Willman,June 2005.

wealth and asset price effects on economic activity - european

Documents