Wealth effects and the consumption of Italian households in the Great Recession

Bottazzi, Renata; Trucchi, Serena; Wakefield, Matthew Working Paper Wealth effects and the consumption of Italian households in the Great Recession Quaderni - Working Paper DSE, No. 1097 Provided in Cooperation with: University of Bologna, Department of Economics Suggested Citation: Bottazzi, Renata; Trucchi, Serena; Wakefield, Matthew (2017) : Wealth effects and the consumption of Italian households in the Great Recession, Quaderni - Working Paper DSE, No. 1097, Alma Mater Studiorum - Università di Bologna, Dipartimento di Scienze Economiche (DSE), Bologna, https://doi.org/10.6092/unibo/amsacta/5522 This Version is available at: https://hdl.handle.net/10419/177597 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/3.0/ ISSN 2282-6483 Wealth Effects and the Consumption of Italian Households in the Great Recession Renata Bottazzi Serena Trucchi Matthew Wakefield Quaderni - Working Paper DSE N°1097 1  Thisversion:August2015 WealthEffectsandthe ConsumptionofItalianHouseholds intheGreatRecession*  RenataBottazzi(UniversityofBolognaandInstituteforFiscalStudies,London) SerenaTrucchi(UniversityCollegeLondon) MatthewWakefield(UniversityofBolognaandInstituteforFiscalStudies,London)  Abstract WeestimatemarginalpropensitiestoconsumefromwealthshocksforItalianhouseholdsinthe earlypartoftheGreatRecession.Largeassetpriceshocksin2008underpinanIVestimator.A eurofallinriskyfinancialwealthresultedincutsinannualtotal(non‐durable)consumptionof8.5‐ 9(5.5‐5.7)cents.Thereisevidenceofeffectsonfoodspending.Responsesoftotalandnon‐ durablespendingtochangesinhousingwealthare0.2to0.3cents/euro.Pointestimatesofthe effectofthefinancialwealthshockarelargeriftheyoungestand/oroldesthouseholdsare excluded.Resultsindicatethatresponsestothewealthshockwerestrongerforthosewho becamepessimisticaboutthestockmarket,andforthoseownersofriskyassetswhoalsoheld mortgagedebt.Counterfactualsindicatefinancialwealtheffectswereimportant(relativetoother factors)forconsumptionfallsinItalyin2007/08. Keywords:Wealtheffects;householdconsumption;theGreatRecession JELcodes:D12, D91  *ThispapersubstantiallyimprovesonBottazzi,TrucchiandWakefield(2013)andearlierversionswere circulatedandpresentedunderthetitle“TheEffectsoftheFinancialCrisisofthelate2000sontheWealth, ConsumptionandSavingofHouseholdsinItaly”. Thenameorderofauthorsisalphabetical. Contactdetails:Bottazzi:renata.bottazzi(at)unibo.it;Trucchi:serena.trucchi(at)ucl.ac.uk; Wakefield(correspondingauthor):matthew.wakefield(at)unibo.it; DepartmentofEconomics‐UniversityofBologna,PiazzaScaravilli2,Bologna,40126(BO),Italy. TheauthorsgratefullyacknowledgefinancialsupportfromMIUR‐PRIN2010‐11,project2010T8XAXB_006 andfromMIUR‐FIRB2008,projectRBFR089QQC‐003‐J31J10000060001.Wethank:AndreaNeriattheBank ofItalyforsupportregardingtheSHIWdataandpatienceinfieldingourqueries;RichardBlundell,Tom Crossley,CarlEmmerson,LucaNunziataandLuigiPistaferriforcommentsatvariousstagesofourproject; and,seminarparticipants(includingseveraldiligentdiscussants)at:theWarsawInternationalEconomics Meetings,2012;theCERPconferenceonFinancialLiteracy,SavingandRetirementinanAgeingSociety, 2012;theRoyalEconomicSocietyAnnualConference,2013;theconferenceoftheItalianSocietyof Economists(Societa’ItalianadegliEconomisti),2013;the12thWorkshopofMacro‐Dynamics,2013;the NetsparInternationalPensionsWorkshop,2014;theInstituteForFiscalStudies,2014;andattwomeetings ofourMIUR‐FIRB2008projectgroup(bothinPadua),forconstructivefeedback.Allerrorsareourown. 2  WealthEffectsandtheConsumption ofItalianHouseholdsintheGreatRecession  1.Introduction Astrikingfeatureoftheearlypartofthe“GreatRecession”wasasuddencrashinthevalueof financialassets.Majorstock‐marketindices1intheUSandtheUKapproximatelyhalvedinvalue betweenpeaksinautumn/summer2007andlowsinMarch2009.ThedropinvalueofItaly’sFTSE‐ MIBwasevenmorepronouncedatmorethan60%betweenMay2007andMarch2009. Furthermore,alargepartofthechangesinassetvaluesoccurredduringthecentralmonthsof 2008,andsohouseholdsthatheldwealthinthestockmarketsufferedasudden,potentiallylarge andmostlyunanticipatedshocktothevalueoftheirfinancialwealth.Alongsidethesefallsinasset valuesthereweresubstantialfallsinhouseholds’consumptionexpenditures.Figure1.1shows thatforItalythepathofaggregateconsumptioncloselyshadowedthepathofthestock‐market index,witha3percentfallbetweenlate2007andmid2009that,ifanything,slightlylaggedthe fallinstockprices. Ouraimistousetheshocktoassetvaluesobservedin2008tomeasurethestrengthofthe responseofconsumptionspendingtothechangeinthevalueoffinancialwealth,fora representativesampleofItalianhouseholds.Giventhetimeperiodweareobservingwewillalso beabletocommentontheimportanceofthese“wealtheffects”,relativetootherfactors,in drivingthefallinhouseholds’consumptionduringtheearlypartoftheGreatRecessioninItaly. Amongthese“otherfactors”weconsidertheroleofchangesinhousingwealth.However,unlike theUSandUK,housevaluesinItalydidnotsufferlargefallsnearthebeginningoftheGreat Recession(AgenziadelTerritorio,2012),andsoouremphasisisontheeffectsoffinancialwealth.2 [Figure1.1abouthere] The2008shocktoassetvaluesisnotonlyusefulforusinprovidingempiricalvariationand thechancetoanalysetheimportanceofthisdriverofconsumptionintherecession,itisalso fundamentaltoourstrategyfordealingwithakeyendogeneityproblem.Allelseequal,a householdthatcuts(increases)itsconsumptionbymore,willmechanicallyaccumulatemore(less)  1TheDowJonesIndustrialAveragefortheU.S.,andtheFTSE“AllShare”fortheUK. 2Theabsenceofafallinhousevaluesalsounderpinsthelimitedeffectofwealthonconsumptioninresultsfromthe BankofItalyquarterlymodelforthe2006‐08period,reportedinRodanoandRondinelli(2014). 3  wealth.Unlessthisisproperlyaccountedforintheempiricalsetup,thiscouldleadtoa downwardsbiasin,orevenanegativeestimateof,wealtheffects.Weusetheideathatthe2008 shocktoassetvaluescanprovideasourceofvariationinwealththatisexogenoustohouseholds’ consumptionbehaviour.ApplyingamethodproposedbyBanksetal(2012),thisinsightisusedto buildaninstrumentalvariables(IV)estimator.Theprecisenatureoftheestimatorisdiscussedin section3. Ourstudyisrelatedtootherpapersthathaveaimedtoestimatetheimportanceofwealth effectsindrivingconsumptionbehaviourduringtheGreatRecession.Inaninfluentialstudyofthe U.S.,Mian,RaoandSufi(2013)estimateamarginalpropensitytoconsumeoutofhousingwealth of5–7percentfortheperiod2006‐09.Thisfindingisrobusttoinstrumentingusinggeographical constraintsonhousingsupply(whichshouldnotbecorrelatedwithotherdriversofconsumption), andtheauthorsalsoemphasizeevidencethatresponsestothewealthshockarestrongerwhere householdsarepoorerormore“levered”(indebted).3Giventhelackofafallinhousevalues mentionedabove,thewealthshockthatwelookatistofinancial,ratherthanreal,wealth.In anotherexcellentanalysisoftheUS,Christelis,GeorgarakosandJappelli(2015)havelookedat howlossesonfinancialwealth,lossesonrealwealth,andunemployment,affectedconsumption in2008‐09.Usingdatathataskhouseholdstoreportcapitallossesondifferentassetstheyfinda marginalpropensitytoconsumeoutoffinancialwealthofaround3.3percent(andsmallereffects forlossesonhousing).Inlinewitheconomictheory,theyalsofindevidencethatthosewho expectedthestock‐marketshocktobepermanentadjustedtheirconsumptionmorestronglythan thosewhoexpectedtheshocktobetransitory.Asmentionedabove,themethodologyofour studydrawsontheEnglandbasedanalysisofBanksetal(2012).Thoseauthorshaveless comprehensivedataonspendingthanwedoandasampleofagentsaged50+.Theyfindonly modesteffectsonhouseholdspendingofwealthshocksduringthecrisis,butarealsoabletofocus onwealtheffectsonotheroutcomes(includingexpectationaloutcomes)thatwedonotobserve. Toourknowledgeourpaperisthefirstattempttolookatthedriversofchangeinconsumption, includingwealtheffects,duringtheGreatRecessioninItaly.OurdatacomefromtheBankof Italy’sSurveyonHouseholdIncomeandWealth(SHIW),whichprovidesrichdataonhouseholds’ assetholdings(valuesandownership),consumptionoutcomes,anddemographicandeconomic  3IndeedthisworkispartofabroaderprogrammeofresearchinwhichMianandSufiarguefortheimportanceof “leveredlosses”fordrivingthelargeconsumptioncontractionearlyintheGreatRecession;theirargumentsand findingsarebroughttogetherinMianandSufi(2014). 4  characteristics.ThedataaredesignedtoberepresentativeoftheItalianresidentpopulationand alsohaveapanelcomponent,andthiscombinationofcharacteristicsisuniqueforItalyand impressiveevenbyinternationalstandards.ThusouranalysisoftheItalianexperienceisof broaderinterestforunderstandingtheimportanceofwealtheffectsandtheevolutionof consumptioninEuropeintheGreatRecession.4 AnalysesoftherelationshipbetweenwealtheffectsandconsumptionduringtheGreat Recessionfitinanestablishedliteratureregardingmeasuringwealtheffectsonconsumption. Therehasbeenmuchrecentemphasisonhowpropensitiestoconsumefromrealwealthdiffer frompropensitiestoconsumefromfinancialwealth5;animpressiverecentsurveyoftimeseries andmicro‐econometricevidenceonwealtheffectsisprovidedbyPaiella(2009),itselfbuildingon theequallyexcellentPoterba(2000).Thestudiesmostrelatedtothepresentpaperarethosethat provideevidenceforItaly.Paiella(2007)usespooledcross‐sectionsofdatatoestimatelong‐run marginalpropensitiestoconsumefromdifferentformsofwealthwhileCalcagno,Forneroand Rossi(2009)focusontheeffectsofrealestatewealth.Guiso,PaiellaandVisco(2005)iscloserto ourstudyinthat,inlinewithouranalysisbasedonshocks,theyaimtoestimatetheeffectsof capitalgainsaswellaslongrunrelationshipsbetweenwealthandconsumption;theyfindthaton averagea1eurogaininhousingwealthincreasesannualconsumptionbyaround2cents,while capitalgainsonfinancialassetsmayevenleadtoreductionsinconsumption.Ourkeycontribution tothisliteratureliesinourexploitationoftheassetpriceshockatthestartoftheGreatRecession asanewsourceofplausiblyexogenousvariationinassetvaluesinordertoestimatehow consumptionrespondstochangesinwealth.   4ThestudiescitedinthisparagraphthemselvesfitintoabroaderliteratureonconsumptionduringtheGreat Recession.Petev,PistaferriandSaportaEksten(2011)andDeNardi,FrenchandBenson(2012)fortheUS,and Crossley,LowandO’Dea(2013)fortheUK,providedescriptiveanalysesthatpointtounusualfeaturessuchasthe durationofthecontractioninconsumption,andthebroadrangeofconsumptioncategoriesthathavebeenaffected. ForItaly,Rondinelli,BassannettiandScoccianti(2014)showthatnationalaccountsdataandhouseholdleveldata matchquitewellduringtherecentperiodandthattherehavebeensomedifferencesinchangesinexpenditure sharesacrossagegroupsbutwithageneralshiftawayfrom“leisure”expenditures.RodanoandRondinelli(2014) comparedifferentrecentrecessions.ResultsfromtheBankofItaly’squarterlymodeldonotindicateastrongrolefor wealthinaggregateconsumptionduringtheperiod2006‐08.However,themeasureofwealthinthemodelcombines realandfinancialwealthanddoesnotshowacontractionduringtherelevantperiod;whentheauthorsdescribe microdatatheydofindthefallsinthevalueoffinancialassetsthatareourfocus. 5SeeSlacalek(2009)andCase,QuigleyandShiller(2005). 5  Preciselystated,ourresearchgoalistoestimatethemarginalpropensitytoconsume(mpc6) outoftheshocktofinancialwealththatoccurredatthestartoftheGreatRecession.Apreviewof keyresultsisasfollows.Aoneeurofallinfinancial(orriskyfinancial)wealthresultedin householdscuttingannualtotalconsumptionspendingbybetween8.5and9cents,andslightly morethan5.5centsofthiscutwasinspendingonnon‐durablegoodsandservices.Wefind effectsofaround1.5centsforfoodspending,andinsignificantresults(thoughwiththeexpected positivecoefficients)forexpenditureondurables.Wealsofindthataoneeurochangeinhousing wealthresultsintotalandnondurableconsumptionspendingmovinginthesamedirectionby aroundbetween0.2and0.3cents,butwedonotfindsignificanteffectsonfoodordurables expenditures.Wealsofindevidencethatourestimatesofwealtheffectswouldbelargerifwe excludetheoldestandyoungesthouseholdsfromourdata,andevidencethatindicatesstronger consumptionresponsestothewealthshockamonghouseholdswhoalsobecamepessimistic abouthowtheyexpectedthestockmarkettoperformandamonghouseholdswithsome mortgagedebt.Finally,counterfactualsimulationsindicatethatfinancialwealtheffectswerean importantdriver(relativetootherfactors)ofconsumptionfallsintheearlypartoftheGreat RecessioninItaly,accountingfor17to22percentofcutsinspendinginoursample.Thusour resultsindicatethatwealtheffectsonconsumptioncanbeimportantforhouseholds’welfareand foraggregateconsumptionandeconomicperformance. Thepaperisorganisedasfollows.Section2introducesthedatasetthatweuseandprovides somedatadescriptivesthatfurthermotivateouranalysis.Section3thenexplainsourresearch method,describingbothourIVestimatorandakeyvariablethatmustbeconstructedinorderto implementthisestimator.Section4thenpresentsourmainresultsonwealtheffects.Wefirst presentaveragewealtheffectsforbroadmeasuresofconsumption,thenresultsforfinerspending categories.Wethenputthesizeofourresultsincontext,includingthroughcounterfactual simulations,andsubsequentlylookatheterogeneityinwealtheffectsbetweengroupsofthe population.Finally,section5concludes.   6Weuse“mpc”indifferentlyfor“marginalpropensitytoconsume”and“marginalpropensitiestoconsume”.Context shouldrevealwhetherwehaveasingularoraplural. 6  2.Data Inthissectionwedescribethestructureofthedatasetthatweuse,andtheconsumptionand wealthvariablesthatareessentialtoouranalysis.Descriptionofthesekeyvariablesalsohelpsto motivateouranalysisofwealtheffects. 2.1TheSHIWDataset TheSurveyonHouseholdIncomeandWealth(SHIW)isarepresentativesampleoftheItalian residentpopulation.Samplingisintwostages,firstmunicipalitiesandthenhouseholds.From 1987onwardthesurveyisconductedeveryotheryear(withtheexceptionofatwo‐yeargap between1995and1998)andcoversabout24,000individualsand8,000householdsinaround300 municipalities.Ahouseholdisdefinedasagroupofindividualsrelatedbyblood,marriageor adoptionandsharingthesamedwelling.About50%ofhouseholdsinagivenyearareinterviewed atleastonceinsubsequentyears(panelcomponent). Thesurveyrecordsarichsetofhouseholdandpersoncharacteristicsaswellasinformation onincomesandsavings,andonhouseholdexpenditureandwealth.Wealthdataisrich,containing bothparticipationandvalueforarangeoffinancialassets,housingwealth,andbusinesses.For thepurposeofouranalysis,weusedatafortheyears2004‐2010.Inthiswayweareableto observechangesinwealthandconsumptionduringthe“GreatRecession”(2006–08and2008– 10)andalsotoconstructourinstrumentalvariableusinginformationonhouseholdportfolios fromthe2004and2006surveys. InthenexttwosubsectionswedescribetheSHIWvariablesthatarethemostimportantfor ouranalysis,thoseregardingconsumptionandassetholding. 2.2SHIWconsumptionvariables TheSHIWdatasetrecordsconsumptionspendingonfourdifferentcategoriesofproducts.Total consumptionisthesumoftwoothercategories,namelydurable(meansoftransport,furniture, householdappliances,etc.)andnon‐durableexpenditures.Foodconsumptionisasubclasson non‐durablespendingandincludesmealsathomeoreatenout.Inouranalyseswealways measureexpendituresannuallyandinrealterms(2010euros,basedontheHouseholdIndexof ConsumerPricesprovidedbyIstat). Descriptivestatisticsonconsumptioninoursampleareshownintable2.1.Totalconsumption decreasesbetween2004and2010,but,onaverage,thedropisstatisticallysignificantat1%only 7  between2006and2008.Thisdropislargelydrivenbynon‐durableexpenditurethatsignificantly decreasesbymorethan600eurosonaveragebetween2006and2008,withalmost400eurosof thischangecomingfromfoodconsumption.Durableconsumptiondisplaysaslightlydifferent pattern.Itsignificantlydecreasesonlyin2010,when,onaverage,durablegoodsexpenditure decreasesby300euros. [Table2.1abouthere] 2.3SHIWfinancialwealthvariables TheSHIWdatasetcollectsdetailedinformationonhouseholdportfolios.Respondentsareasked whethertheyholdeachofmanytypesofassetand,ifso,abouttheamountofwealththeyholdin eachasset.Assetsaregroupedinbroadcategories:cash(bankaccountsandsavingcertificates); Italiangovernmentbonds(withdifferentdurations);domesticbondsandinvestmentfunds;Italian shares;foreignbondsandshares;otherminorcategories.Withineachofthesebroadcategories individualsareaskedaboutadetailedsetofassets.SHIWalsoprovidesinformationonhousehold wealthinseveraltypesofmutualfunds,andthesefundscanbecategorisedaccordingtowhether ornot(andtheextenttowhich)theyexposetheholdertostock‐marketrisk. Ifsurveyrespondentsreportthattheyholdanasset,theyarethenaskedabouthowmuch wealththeyheldinthatassetatthe31stofDecemberintheyearafterwhichthesurveywaveis named(i.e.December31st2008forthe“2008SHIW”).7Respondentsarefirstaskedtoindicatein towhichofseveralbandsofvaluetheirassetfellandthentoreportapointamountforthisvalue. Failuretoreportapointamountresultsinthehouseholdbeingaskedwhetherthevalueoftheir holdingisnearertothebottom,middleortopoftheband.Sincenotallindividualsgiveapoint amountweusesomeimputedvaluesforwealth.Inimputationweusebandand bottom/middle/topinformationtoallocatevaluesbyasset.8 Sinceourmainregressionsareinfirst‐differences(seesection3)wehavetobecarefulabout thefactthatimputationcouldconsiderablyincreasenoisetosignalratio,especiallyforcases whereindividualsreportholdingsintherelativelybroadtopbandsofassetvalues.Forthisreason inoursampleselectionweexcludefromthesamplehouseholdswhodonotprovideapoint  7Havingendofyearwealthmeanswehavedataonhouseholdsatclosetothetopofthestockmarket(attheendof 2006)andatclosetothebottomofthecrash(attheendof2008). 8TohaveahomogeneousmeasureofassetvalueswedonotuseimputedvaluesprovidedbytheBankofItaly,since theyarenotavailableforthe2004wave.WeneedtorelyonimputationbytheBankofItalyfor(thesumof)three typesofdepositin2006,sinceinformationonthebandtheybelongtoisnotavailable.Resultsofsection4arenot sensitivetosubstitutingBankofItalyimputationforourimputationasfaraspossible. 14  capitallossesarepartiallyoffsetbyactivesaving.Ontheotherhand,thereportedchangeinrisky assetsismorenegativethanthe“calculatedchange”.Thismayindicatethereshufflingby householdsoftheirportfolios,toreduceexposuretostock‐marketrisk.Theideaisalsosupported byobservedexitsfromthestockmarketduringthecrisis:thestock‐marketparticipationrate decreasesfrom14%before2006to12%in2008andto10%in2010.18Aregressionofreported changesinwealthoncalculatedchangesandaconstantgivessignificantcoefficientsof0.65for overallwealthand0.88forriskywealth.Thuscalculatedchangesinwealthdohavethedesired positivecorrelationwithactualchanges,andtherelationshipiscloserforriskythanforoverall wealth.  4.Measuringwealtheffects Wenowpresentourestimatesofhowconsumptionrespondedtotheshocktofinancialwealth. OurmainestimatoristheIVestimatordescribedinsection3.Asexplainedinthatsection,we believethismethodwillprovideconsistentestimatesoftherelationshipofinterest.Bycontrast, andagainforreasonsexplainedintheprevioussection,OLSestimationwouldbelikelytoproduce anunderestimateofthetruerelationship.Forthefirstsetofresultsthatwepresent,whichare ourbaselineresults,wepresentOLSestimatesalongsidetheIVestimatesandseethatOLSdoes indeedgeneratecoefficientsthataresmallerthantheIVestimates. Inlinewithequations(2)and(3),allofourestimatesforwealtheffectscomefrom regressionsthatincludeseveralotherindependent(X)variablesalongsidethekeyfinancialwealth variables.Onevariableofparticularinterestisthechangeinthehousehold’sperceivedvaluation oftheirhousingwealth.Whilewehavegonetoagreatdealofefforttoensureexogeneityofthe financialwealthvariables,forthishousingwealthvariablewesimplyincludethechangeinthe reportedvalueofhousing.Theideahereisthatsincesurveyrespondentsareaskedwhatthey perceivetobethevalueoftheirhouse,whattheyreportshouldbethelevelofwealththat informstheirconsumptionchoices.Furthermore,since(unlikefinancialwealth)realestatewealth isnotreadilyadjustable,thereislessofaproblemofamechanicalrelationshipbetweenactive  18TheseownershipratesarecalculatedbytheauthorsusingtheSHIWdata(withsampleasusedinTableB1).Note thattheownershipratescannotbeinferredfromthenumbersinTable3.1becausethesample“withriskyassets”in thatTablearethosethathadriskyassetsatafixedpointintime(2006whentheownershipratewasapproximately 14%). 15  savinginhousingandchangesinconsumption.Onthebasisoftheseargumentsweinterpretthe coefficientonchangesinhousevaluetobethempcoutofshockstohousingwealth. Theremainingregressorsinallthereportedregressionsincludechangesin:unemployment status;retirementstatus;and,inthenumberofpeopleandearnerslivinginthehousehold.The variablessofardiscussedareallinfirstdifferences.Thereportedresultsalsoalwaysallowforthe possibilitythatthechangeinconsumptionisrelatedtothecharacteristicsofhomeownership, retirementstatusandemploymentintheprivateorpublicsector(allmeasuredattime t‐1 ),and notjusttodifferencesinsuchvariables,andwealwayscontrolforagebands,sex,educationlevels (compulsory,post‐compulsory,andsomecollegeeducation),and,tocaptureeffectscomingfrom thestateofthemacroeconomy,regiondummies,yearandtheregionalunemploymentrate. Finally,wehaveobtainedallourresultswithandwithoutacontrolforthechange(first difference)inlabourincome,andinseveralcasesreportbothregressions.Thereisaworrythat labourincomemaybeendogenous(due,forexample,toreversecausalityfromthedesireto increaseorreduceconsumptiontolaboureffortandthereforeincome)soitisreassuringthatthe inclusionofthechangeinlabourincomedoesnotnoticeablyaffecttheotherestimated coefficientsinthemodel,andparticularlytheestimatedwealtheffects.Descriptivestatisticsfor thefinancialwealthvariablesthatarecrucialforourempiricalandIVstrategywerepresentedin section3(Table3.1);descriptivestatisticsforallotherregressorsarecontainedinAppendixTable B2. 4.1Averageresponsestothewealthshock Akeyaimofourstudyistounderstandhowthewealthshockaffectedoverallconsumption spendingforthehouseholdsinoursample.Inlinewiththis,ourfirstsetofresults,inTables4.1 and4.2,respectivelyhavethechangeintotalhouseholdconsumptionspending,andchangein householdconsumptionspendingonnon‐durables,asdependentvariables.Eachofthesetables presentsresultsforasubsetofcoefficientsfrom8regressions.ThefirsttwocolumnspresentOLS regressionsofthechangeinconsumptiononthechangeinwealth(andotherregressors),first withandthenwithoutacontrolforthechangeinlabourincome,whilecolumns3and4present the“secondstage”resultsofthepreferredIVspecificationswith,thenwithout,thelabourincome control.Thedifferencebetweenthetopandbottompanelsisthefinancialwealthvariablethatis themainvariableofinterest:inthetoppanelthisvariableisthechangeinriskyfinancialwealth 16  thatisinvestedinthestockmarketeitherdirectlyorthroughawrapperproductsuchasamutual fund,whileinthebottompanelitisthechangeintotal(accessible19)householdfinancialwealth. [Table4.1abouthere] [Table4.2abouthere] Forinterpretationofthemainwealthcoefficients,itiseasiesttoconsideranexample.The coefficientonthecalculatedchangeinriskyfinancialwealthinIVregressionreportedincolumn3 ofthetoppanelofTable4.1(thisistheregressionforthechangeintotalconsumptionandthat includesthechangeinincomeasaregressor)is0.088.Sincecalculatedchangesinwealthare measuredinreal(2010)euros,andconsumptionismeasuredineurosperyear,thispointestimate indicatesthatifwealthincreases(falls)byoneeuro,annualconsumptionincreases(falls)by8.8 cents.Inotherwords,thecoefficientindicatesanmpcoutofthewealthshockof8.8percent. Othercoefficientsonwealthvariablescanbeinterpretedanalogously. Withthisinterpretationinmind,wecansummariseourmainIVresultsfortheeffectsofthe shocktofinancialwealthonconsumption,asfollows.Wereportresultsfromregressionswiththe controlforthechangeinlabourincome,andinbracketstheresultswithoutthiscontrol.Fortotal consumption(Table4.1),thempcis8.8(8.6)percentoftheshocktoriskyfinancialwealth,and10 (9.9)percentoutoftotalfinancialwealth.Fornon‐durableconsumption(Table4.2)theestimates are5.7(5.5)percentoutoftheshocktoriskyfinancialwealth,and6.2(6.1)outoftotalfinancial wealth.Itisageneralpatterninourresultsthatcontrollingforthechangeinlabourincomehas almostnoimpactontheestimatesofourmaincoefficientsofinterest.Asapointofcomparison, theestimatedmpcforchangesinfinancialwealthareconsiderablylargerthantheresultsthatwe getforthepropensitytoconsumeoutofachangeinhousingwealth,whichisrobustlyestimated tobebetween0.3percentand0.1percent. ComparingtheseIVresultstoOLSestimates(columns1and2inTables4.1and4.2),wesee thattheOLSresultsarealwayssubstantiallysmallerthantheIVestimates.Thisisinlinewiththe argumentsoftheprevioussectionthattherearegoodreasons(themechanicalrelationship betweenchangesinconsumptionandinwealththroughthebudgetconstraint,andattenuation biasduetomeasurementerror)toexpectOLSestimationtounderestimatethiscoefficient.20  19Accessiblewealthexcludeswealth“lockedaway”inpensionsorlife‐insuranceorsimilarproducts. 20Wedonotreportthe“reducedform”thatrelatesthechangeinconsumptiontotheexcludedinstrument(andthe otherregressors).Giventhatwehaveoneendogenousvariableandexactidentification,thecoefficientonthewealth variableinthisreducedformcanbeinferredastheproductofthecoefficientsontherespectivewealthvariablesin 17  Focussing(fromhereforward)onourpreferredIVestimates,wenoticethatthepoint estimatesforthempcareveryslightlysmallerbutmorepreciselyestimatedwhenthekey independentvariableisthechangeinriskyfinancialwealth(toppanelofTables4.1or4.2), comparedtowhenitisthechangeintotalfinancialwealth.Forexample,ifwetakethecasesfor totalconsumptionandwiththecontrolforthechangeinlabourincome(column3oftheTable 4.1),theestimatedcoefficientis:0.088(significantatthe10%level)whentheregressorrelatesto riskywealth;and,0.1(notsignificantatconventionallevels)whentheregressoristhechangein totalfinancialwealth.Theequivalentcoefficientsfornon‐durableconsumptionrespectivelyare 0.057(significantatthe5%level),and0.062(significantatthe10%level).Thegreaterprecision whentheregressoristhechangeinriskywealthispartlyduetothefactthatwehavegreater precision(asmallerstandarderrorontheinstrument)inthefirststageforthecaseusingrisky wealth,whichindicatesthatinstrumentingaddslessnoiseinthiscase.Thegreaterprecision,and largersize,oftheestimatedcoefficientontheexcludedvariableinthecasewithriskywealth againindicate(asnotedinsubsection3.2)thatthereisastrongerrelationshipbetweenasset pricesandthevalueofriskywealththanbetweenassetpricesandthevalueofallfinancialwealth, perhapsbecauseportfolioreshufflingandtheaccumulationofsafeassetsweakentherelationship tototalwealth.TheF‐statisticontheexcludedinstrumentalsoindicatesthatourIVstrategyis moreeffectivewhentheregressoristhechangeinriskywealth:inthatcasewedonotneedto worryaboutaweakinstrumentproblem.Fullfirst‐stageresultsarereportedinAppendixTables B3(riskyfinancialwealth)andB4(totalfinancialwealth). Asalreadynoted,thewealtheffectcoefficientsarerobusttocontrollingforthechangein labourincome.Ourresultsarealsorobusttoaseriesofothermodificationstoourspecifications. Forexample,whileourdataareforchangesinwealthandconsumptionbetween2006and2008, and2008and2010,mostofthevariationthatourestimatorexploitsisrelatedtotheassetprice shockthatoccurredbetweenthe2006and2008wavesofdata.Estimationbasedonlyon differencesforthe2006‐2008periodleftourresultsalmostunchangedrelativetothosereported. Anotherrobustnesscheckinvolvedslightlychangingthewayinwhichweconductedour  thefirstandsecondstagesoftheIVregression(thisisthereverseofindirectleastsquares).Giventhatthecoefficients inthefirststages(reportedinAppendixTablesB3andB4)are0.672(forthechangeinriskywealth),andjustbelow 0.6(forthechangeintotalfinancialwealth),thereducedformestimateswouldbesmallerthanourIVestimates. However,therearereasonstosupposethatthereducedformwouldunderstatetherelationshipofinterest.In particular:thechangeincalculatedwealthislikelytooverstatethetrueshocktowealth(andthusleadtoan understatementofeffectsmeasuredpereuroofchangeinwealth)ifhouseholdscanoffsetsomeoftheassetprice shockthroughtheirportfoliochoices;and,thereducedformwouldbeaffectedbyattenuationbiasifthereis measurementerror.TheseissueswerediscussedinmoredetailinBottazzi,TrucchiandWakefield(2013). 18  instrumentalvariablesanalysis.Asnotedinsection3,thevariationexploitedbyourinstrumentis heterogeneitybetweenhouseholdsintermsofthelevelofwealthheldindifferentassets,and differencesinreturnsbetweenassets.Wehavetriedincludingthe(twice‐lagged)levelofwealthin differentassetsandtheinteractionofthiswiththestock‐marketindex(theFTSEMIB)intheplace oftheconstructedchangeincalculatedwealthastheexcludedvariablesinthefirststageofourIV analysis.Againtheresultsarealmostunchangedrelativetothosewereport(infactthechange sometimesimprovesthesignificanceofourresults).Anothermodificationthatmakespractically nodifferencetoourestimatesistodropregressorsfor(lagged)retirementstatus,sectorof employment,andhomeownership;intheresultswereport,theselevelvariablesareincluded alongsidevariablesmeasuringthechangeinretirementstatus,housingwealthandemployment. Afinalrobustnesscheckistoruntheregressionswithdeltariskywealthasthekeyregressor,only onthesampleofthosewhohaveriskywealth.21Sincethecoefficientonthekeywealthvariable measuresthechangeinconsumptionperunitofchangeinwealth,includinghouseholdswithout riskywealth(aswedointhereportedresults)shouldnotaffectourestimatedcoefficientsand indeeddroppingthesehouseholds(andshiftingtoamuchsmallersample)doesnotsubstantially affectourpointestimates.Fullresultsfromtherobustnesschecksdescribedinthisparagraphare availablefromtheauthorsonrequest. Otherthantheestimatedwealtheffects,theothercoefficientsthatarereportedinTables4.1 and4.2arecoefficientsonthevariablesthatwemostoftenfoundtobesignificant(fullresultsfor theregressionscanbefoundinAppendixTablesB5toB8).Thepatternsofresultsareinlinewith economicintuition:becomingunemployed(butnotbecomingretired)isassociatedwithcutsin spending,whiletheadditionofextrahouseholdmembersorofanextraearnerinthehouseholdis linkedtohigherexpenditures. Tosummarise,theresultsdiscussedinthissectiongivetheaverageeffectofthewealthshock ontheconsumptionofhouseholdsinoursample.Ourfavouredestimatesindicatethataeuroloss ofriskywealthintheperiodofthestock‐marketcrashled,onaverage,toan8.8(or8.6without thecontrolforthechangeinlabourincome)centcutinconsumption,and5.7(5.5)centsofthis cutwasinspendingonnon‐durablegoods.Pointestimatesfortheresponsetothechangeintotal financialwealthareslightlylarger,butlesspreciselyestimated.  21Morepreciselythesampleisthosewhohadriskywealthintheappropriate(lagged)waveofdatasuchthatthey contributetotheestimationofthecoefficientontheinstrumentedwealthvariable. 19   4.2ResultsforCategoriesofConsumptionSpending Theresultsintheprevioussubsectionareforbroadcategoriesofconsumptionspending. Theoreticalconsiderationsthat“luxuriesareeasiertopostpone”(BrowningandCrossley,2000), andfindingsthathouseholdsintemporarilystraitenedcircumstancesmaypostponetherenewal ofdurablesratherthanimmediatelycuttingbackonallspending(BrowningandCrossley,2009), meanitisinterestingtolookatfinercategoriesofspending.Asidefromtotalconsumption spendingandspendingonnon‐durables,ourdataallowustolookatspendingondurablesand spendingonfood. Table4.3presentskeycoefficientsforourwealtheffectregressionsforspendingonfoodand durables,alongsidetheresultsfortotalspendingandspendingonnon‐durables(fullsetsof coefficientsfromtheregressionsarepresentedinAppendixTableB9).Wepresentresultsfrom ourpreferredIVspecificationandwiththekeyindependentvariablebeingthechangeinrisky wealth(sothatwedonothaveaproblemofweakinstruments22).Thus,theresultsfortotal consumptionandnon‐durableconsumptionreplicatethosepresentedinthetoppanelsofTables 4.1and4.2. [Table4.3abouthere] Thepointestimatesincolumns(c1)and(c2)ofthetableindicatethatdurablesexpenditures wereaffectedby,onaverage,3.1centsperyearforaeurochangeinriskywealth.Sincetotal consumptionspendingisthesumofspendingondurablesandspendingonnon‐durables,the changeintotalspendingpereurochangeinwealthshouldbethesumofthechangesinspending onnon‐durablesanddurables,pereurochangeinwealth.Lookingatthecoefficientsincolumns (a1),(b1)and(c1),orincolumns(a2),(b2)and(c2),wecanseethatthisrelationshipdoesindeed hold.Whilethis“addingup”isreassuringabouttheconsistencyofhouseholds’responsestothe differentconsumptionquestionsinthesurvey,theresultsfordurablesspendingarenot significant.Inaddition,coefficientsonchangesinhousevalueareinsignificantandclosetozeroin thespecificationsfordurables.Thelackofsignificancemayinpartbeduetothefactthatdurable purchaseshappenonlyinfrequentlyandsowedonotobserveenoughdurablespurchasesto identifypatternsinthedata.  22Giventhatthesampleandregressors(includingtheendogenousregressor)donotchangeacrosstheregressions reported,the“firststage”resultsarealwaysthosealreadydiscussedandpresentedinAppendixTableB3. 20  Forfoodspendingweagainfindnoevidenceofeffectsfromhousingwealth(coefficientsare verysmallandhaveverysmallstandarderrors).Forourmainvariableofinterest,aeurochangein thevalueofriskyfinancialwealthisseentoleadtoacutinfoodspendingof1.5centsperyear andthisresultissignificantatthe10%level(seecolumns(d1)or(d2)ofTable4.3).Theseresults arepotentiallystriking.Iffoodisanecessity,thenevensmallchangesinfoodspendingcouldbe potentiallyimportantforhouseholds’welfare.However,weshouldbecarefulininterpretation. Ourdataonfoodspendingarenotverydisaggregatedandwecannot,forexample,distinguish “foodin”and“foodout”.23Wenextconsiderinmoredetailtheinterpretationofthemagnitudeof ourestimatedwealtheffects.  4.3HowLargearetheseWealthEffects? OurestimatesofwealtheffectsarebasedontheearlyyearsoftheGreatRecession,andusingthis periodhelpsustohaveaplausiblyexogenoussourceofvariationinfinancialwealththatwe exploittoidentifyeffects.Thisexogeneitymaygiveestimatesthathavegeneralityoutsideour sampleperiod,oritmaybethatthetimeperiodthatweexploitisunusualintermsofaverage wealtheffects.Whilewecannotinvestigatethisdirectly,wecanatleastputourestimatesinthe contextofpreviousliterature. Findingsregardingwealtheffectsinconsumptionhaveusuallyfocussedonbroadmeasures suchastotalconsumptionornon‐durableconsumption.Ourpreferredpointestimatesforthe mpcoutofshockstofinancialwealtharebetween8.5and9percentfortotalconsumption,and aroundorjustabove5.5percentfornon‐durableconsumption.TheseeffectsdifferfromtheItaly basedfindingofGuiso,PaiellaandVisco(2005)thatconsumptionmayevenfallinresponseto capitalgainsonfinancialassets.Onecouldonlyspeculateastowhetherthisdifferencecomes fromdifferencesinsampleperiodordifferencesinthemethodandvariationusedtocapture effects.ItisslightlydifficulttomakeadirectcomparisonofourresultstothoseofBanksetal. (2012),thepaperthatisclosesttooursintermsofmethodology,sincetheydonotobservesuch comprehensivemeasuresofconsumptionspendingaswedoandtheirsampleisforarestricted (older)agerange.However,ifwetrytoextrapolateaneffectontotalconsumptionfromtheir resultsitwouldseemthatthiswouldbeweakerthanourfindings.Ourfindingsforspendingon  23Basedonadifferentdataset(aHouseholdBudgetSurvey),Rondinelli,BassanettiandScoccianti(2014)donotice,for somegroupsofthepopulation,differencesintheevolutionofexpenditureshareson“food”andon“accommodation servicesandrestaurants”duringthe2000s(until2012). 21  foodarealsostrongerthanthesumoftheirresultsforfoodinandfoodout.Comparedtoother, USbased,resultsforwealtheffectsintheGreatRecession,ourmpcoutoffinancialwealthis somewhatlargerthanthe3.3percentfoundbyChristelis,Georgarakos,andJappelli(2015),but onlyslightlylargerthanthempcoutofhousingwealthestimatedbyMian,RaoandSufi(2013). Moregenerallyourfindingsonmpcoutofshockstofinancialwealthdonotseemoutoflinewith findingsintheliterature(seeforexamplethecollectionofmicro‐databasedresultsinTable3of Paiella(2009)),althoughanestimateof0.088or0.086fortotalconsumptionisperhapsatthetop endoftherange. Ourfindingsonmpcfromchangesinhousingwealthare(fortotalandnon‐durable consumption)robustlyintherange0.001–0.004.ThisisinlinewiththefindingsofGuiso,Paiella andVisco(2005).Thusourfindingsseemtoconfirmthattheaveragemarginalpropensityof Italianhouseholdstoconsumefromchangesintheirhousingwealthisreasonablyinlinewith (perhapsatthelowerendof)therangeofinternationalestimatesofthisparameter. Anotherwayofthinkingaboutthesizeofourestimatedmpcistoconsiderwhatthesempc implyforhowmuchsmallerobservedfallsinconsumptionwouldhavebeeninourdataifthe valueoffinancialassetshadnotfallenin2008.Wecanaddressthisissuebyperforming counterfactualsimulationsbasedonourregression.Thatistosay,wefirstusetheregressionto predicttheaveragechangeinconsumptioninoursample.Wecanthen(counterfactually)setthe changeinwealthtozeroforallindividualsinoursampleandmakeanewprediction.24Comparing thetwopredictionswillgiveameasureofhowmuchoftheaveragefallinconsumptionisbeing drivenbywealtheffects.Wecanalsocomparethisinfluenceofwealtheffectstotheimpactof otherfactorsbyusingasimilartechniqueto“switchoff”theinfluenceof(say)changesinhousing wealth,unemploymentstatus,thenumberofearnersinthehousehold,orinlabourincome. [Table4.4abouthere] Table4.4displaystheresultsofthiskindofcounterfactualexercisebasedontheIV regressionforthechangeintotalconsumptiononthechangeinriskywealth(andincludingthe  24OurpreferredestimatesareIVregressions.TheeasiestwaytoperformthiscounterfactualanalysiswithintheIVset upisto“manually”computethetwostepsoftheIV.Thatis,ratherthanusingabuiltinpackageintostatistical software(inourcaseStata13)tocomputetheIV,usearegressioncommandtocomputethefirststage,then constructthe“predictedwealth”variablethatbecomesaninputintothesecondstagewhichiscomputedbya seconduseoftheregressioncommand.Sincethisprocedureinvolvesexplicitlyobtainingthe“predictedwealth” variable,itisstraightforwardtoproducepredictionsbasedoncoefficientsofthesecondstageregressionbutwiththe predictedwealthvariablesettozero. 22  changeinlabourincome),thatisreportedinthetoppanelandthirdcolumnofTable4.1.We reportresultsforthecounterfactualexercisecomputedacrossallhouseholdsinoursample(first columnofTable4.4),andonlyforthe(approximately)halfofthesamplewhosechangein consumptionismeasuredfortheperiodofparticularlylargeassetpriceshocks(2006–08,second columnofTable4.4). Inourfullsampletheaveragetwo‐yearfallinannualconsumptionis515euros.Thisfall amountstoalmost3%ofaverageconsumptionspendinginoursample,25afigurewhichis reasonablyinlinewiththefallinaggregateconsumptioninItalyoverthesameperiod(seeFigure 1.1).Sinceweareusingleast‐squaresregression,the515eurofallismatchedbytheaverage predictionofconsumptionchangesbasedonourregression.Ifwerepeatthepredictionexercise butwiththe“predictedwealth”variablefromthefirststageoftheIVsettozeroforthosewho haveriskywealth,wefindtheaveragefallinconsumptionisreducedto425euros.Thus,wealth effectsareexplainingaround90outofthe515euroaveragefall,orapproximately17%ofthefall inconsumptiononaverage.Incontrast,changesinhousingwealthonlycapturearound3%ofthe averagefallinconsumption.Thepartofthechangeinconsumptionexplainedbywealtheffectsis alsoveryslightlylargerthantheproportionscomingfromeitherchangesinlabourincome,or fromthejointimpactofchangesinthenumberofearnersandinunemploymentstatus. Itmayseemsurprisingthatthechangeinthevalueofriskywealthissopowerful,relativeto otherfactors,whenonlyaround14%ofoursampleheldriskyfinancialwealthbeforetheasset priceshock(in2004or2006).However,theshocktowealthwaslarge.Theresultsfromourfirst stageindicatethattheassetpriceshockledtoanaveragefallinthevalueorriskywealthof7130 eurosamonghouseholdswithsomeriskywealthbeforethecrisis,andcombiningthiswithour mpcestimategivesanaveragecutinconsumptionduetothewealthshockof627eurosperyear amongthesehouseholds.Averagingthesizeofthecutacrossallhouseholds(withandwithout riskywealth)givesusbackthe90euroresult. Consideringonlythe2006–08sample,theaveragefallinconsumptionisnow796euros(or around4.5percent,whichisagainquitewellinlinewithaggregatedata).26Inthiscasewesee thatthechangesinfinancialwealtharedrivingmoreofthefallinconsumption(around22%ofit) thanareanyoftheotherfactorsweconsiderthroughourcounterfactuals:inthissampleeventhe  253%iscalculatedas100*515/17454. 26Sincethereisa“2010dummy”,thisfallisagainmatchedexactlybyregressionpredictions. 23  compositeeffectofchangesinthenumberofearnersandinunemploymentandinlabourincome, isnotasstrongastheeffectoftheshocktowealth.  4.4HeterogeneityinWealthEffects Thewealtheffectsconsideredsofarareaverageeffectsinoursample.Theconceptualframework underpinningourresearchsuggeststhattheremightwellbeheterogeneityinwealtheffects.We considerheterogeneitybyage,andwhetherthestrengthofwealtheffectsisrelatedtobecoming pessimisticaboutthestockmarketorbeingexposedtomortgagedebt. Heterogeneitybyage Thesimplelife‐cyclemodelthatweappealedtoinsection3,predictsthatagentsshouldconsume aproportionoftheirwealthineachperiod,andsoshouldsmoothoutshockstowealthby spreadingthechangesinspendingthattheseshocksnecessitate,acrosstheremainingperiodsof theirlives.Giventhat,allelseequal,olderindividualshaveashorterhorizonoverwhichtheycan distributechangesinconsumptionspending,themodelsuggeststhatolderindividualsshould respondmorestronglytothewealthshocks.27Aricherversionofthemodelmakestheprediction lessclearcut.Ifhouseholdsformlinksinongoingfamilialdynasties,thenthemodelmay effectivelyhaveaninfinitehorizonthuspotentiallydecouplingthelinkbetweenageandthelikely strengthofresponsestowealthshocks.Alternatively,creditconstraintsmaymeanthatthe consumptionofyoungerhouseholdsiscloselytiedtocurrentresources,andsothesehouseholds maybeveryresponsivetoshocks.Giventheambiguity,itisanempiricalquestiontotryto establishwhetherandhowwealtheffectsvarywithage. Withourinstrumentalvariablesstrategy,preciseidentificationisquitedemanding.To mitigateproblemswithsamplesizeandpotentiallyweakinstruments,ourapproachto investigatingheterogeneitybyageistore‐estimateourmainmodelonthefollowingsubsamples: asampleexcludinghouseholdsheadedbysomeoneaged70orabove;asampleexcludingthose agedlessthan50;and,asampleexcludingboththe“young”(under50s)andthe“old”(70plus). Eachrestrictiondropsaroundaquarterofthehouseholdsfromourmainsample,sothesample withneithertheoldnortheyoungisslightlylessthanhalfthesizeofthesampleusedinTables4.1  27Thiskindofintuitionunderliesmuchworktryingtountanglewhyhousepricegrowthandaggregateconsumption aresostronglycorrelatedintheUK:seeAttanasioandWeber(1994);Attanasioetal(2009),Attanasio,Leicesterand Wakefield(2011). 30  Christelis,Dimitrios,DimitrisGeorgarakos,andTullioJappelli(2015),"WealthShocks, UnemploymentShocksandConsumptionintheWakeoftheGreatRecession,"Journalof MonetaryEconomics,vol.72,21‐41. Crawford,Rowena(2013),“Theeffectofthefinancialcrisisontheretirementplansofolder workersinEngland”,EconomicsLetters121,156‐159. Crossley,ThomasF.,HamishLowandCormacO’Dea(2013),“HouseholdConsumptionthrough RecentRecessions”,FiscalStudies34(2),203‐229. DeNardi,Mariacristina.,EricFrenchandDavidBenson(2012),“ConsumptionandtheGreat Recession”,EconomicPerspectives36(1),FederalReserveBankChicago. Dynan,KarenE.,andDeanM.Maki(2001),DoesStockMarketWealthMatterforConsumption?, BoardofGovernorsoftheFederalReserveSystems“FEDS”(FinanceandEconomics DiscussionSeries)paper2001‐23,May2001. Guiso,Luigi,MonicaPaiella,andIgnazioVisco(2005),Docapitalgainsaffectconsumption? EstimatesofwealtheffectsfromItalianhouseholds’behaviour,BankofItaly,WorkingPaper, No.555. Mian,Atif,KamaleshRao,andAmirSufi,(2013),“HouseholdBalanceSheets,Consumption,and theEconomicSlump”,TheQuarterlyJournalofEconomics128(4),1687‐1726. Mian,Atif,andAmirSufi,(2014),HouseofDebt:HowThey(andYou)CausedtheGreatRecession, andHowWeCanPreventItfromHappeningAgain,TheUniversityofChicagoPress,Chicago andLondon. Paiella,Monica,(2009),“TheStockMarket,HousingandConsumerSpending:ASurveyofthe EvidenceonWealthEffects”,JournalofEconomicSurveys23(5),947‐73. Paiella,Monica,(2007),“DoesWealthAffectConsumption?EvidenceforItaly”,Journalof Macroeconomics29,189‐205. Petev,Ivaylo,LuigiPistaferriandItaySaportaEksten(2011)“ConsumptionandtheGreat Recession”,D.Grusky,B.WesternandC.Wimer(eds.),TheGreatRecession,CUPServices. Poterba,J.(2000),“Stockmarketwealthandconsumption”,JournalofEconomicPerspectives14, 99–118. 31  Rondinelli,Concetta,AntonioBassanetti,andFilippoScoccianti,(2014)“OnthestructureofItalian Households’ConsumptionPatternsduringtheRecentCrises”,inBancad’Italia,Glieffettidella crisisulpotenzialeproduttivoesullaspesadellefamiglieinItalia,WorkshopsandConferences Report18. Rodano,Lisa,andConcettaRondinelli,(2014),“TheItalianhouseholdconsumption:acomparison amongrecessions”,inBancad’Italia,Glieffettidellacrisisulpotenzialeproduttivoesulla spesadellefamiglieinItalia,WorkshopsandConferencesReport18. Slacalek,Jiri,(2009),“WhatDrivesPersonalConsumption?TheRoleofHousingandFinancial Wealth”,TheB.E.JournalofMacroeconomics,Topics,9(1).   32  AppendixA:Constructingthe“CalculatedChangeinWealth” Sourcesforassetpriceindicesandinterestrates,andtheassetclassesthattheyareappliedtoin constructingcalculatedchangesinwealth,are:  Holdingsincurrentaccountsandcashdeposits:theannualinterestrateoncurrent accountsavailabletohouseholds(source:BankofItaly,BolletinoStatistico).  ShorttermItaliangovernmentbonds(durationlowerthan2years,assumedtobeheld tomaturity):interestratesyieldedbyBOTwith12monthsdurationandbyCTZtraded inBorsaItaliana(source:BankofItaly).  Long‐termItaliangovernmentbonds(CCTandBTP):capitalgainsbasedonpriceindices availablefromtheBankofItaly.  SharesheldinItaly:FTSEMIB(FTSEviadatastream)  Sharesheldoverseas:FTSEAll‐Worldindex(FTSEviadatastream)  Italianprivatebondsandotherforeignassets,Pfandbriefeindex. Toclassifymutualfundsaccordingtoexposuretostockmarketriskweusetheclassification providedbytheItalianassociationofsavingsproviders(Assogestioni,Guidaallaclassificazione). WethenassumetheamountinvestedinthestockmarketevolvesinlinewiththeFTSEMIBand thattheremainderofthefundisinvestedinItaliangovernmentbonds.Indetail,theshareof governmentbondsis100%formonetaryandbondfunds;15%forstockfunds;50%formixed funds;30%forbalancedstockfunds;70%forbalancedbondfunds.   AppendixB:SupplementaryTables  [AppendixTablesB1toB11abouthere]  33  FiguresandTables Figure1.1:StockpricesandAggregateConsumptionSpendinginItaly,2004–2010  Source:FTSEviadatastreamforstockprices(FTSEMIB)andIstat(databaseI.stat)forconsumption(finalconsumptionexpenditure ofhouseholdsoneconomicterritory). Notes:TheverticalaxisontheLHSmeasuresthevariationofstockpriceswithrespecttothefirstquarterof2007(thevalueof FTSEMIBinthereferenceperiodissetequalto100);theaxisontheRHSmeasuresthevariationofconsumptionwithrespecttothe firstquarterof2007.  95 96 97 98 99 100 30 40 50 60 70 80 90 100 110 2004Q1 2004Q3 2005Q1 2005Q3 2006Q1 2006Q3 2007Q1 2007Q3 2008Q1 2008Q3 2009Q1 2009Q3 2010Q1 2010Q3 Stock Prices Consumption 34  Table2.1:Descriptivesofconsumptioninoursample  Totalconsumption expenditure Non‐durables consumption expenditure Durables consumption expenditure Foodconsumption expenditure 2004 Mean 18784 16630 2154 7074 St.dev (12589) (9103) (7205) (3676) 2006 Mean 18304* 16411 1893* 6918* St.dev (11179) (8528) (5773) (3433) 2008 Mean 17541*** 15796*** 1745 6522*** St.dev (10515) (7943) (5210) (3108) 2010 Mean 17295 15859 1436*** 6391*  St.dev (9945) (8130) (4149) (3064) NotestoTable: 3047observationsin2004;3867in2006;3865in2008and3323in2010. Starsrefertothesignificanceofthetestonequalityofmeanconsumptioninthecurrentandpreviouswave(withequalvariances): *p<0.1,**p<0.05,***p<0.001.   Table3.1:Descriptivesofthechangeinwealthandthecalculatedchangeinwealth Financialwealth  Changesinreportedwealth Changesin“calculated”wealth Mean(st.dev) ‐248(56766) ‐1397(8050) 2008 ‐1259(46455) ‐2829(11431) 2010 678(64787) ‐85(909) Median 0 ‐114 25thpercentile ‐4727 ‐369 75thpercentile 5832 ‐14  Regressioncoefficient  0.649*** (0.088) Riskyfinancialwealth(hhswithriskyassetsin2006)  Changesinreportedwealth Changesin“calculated”wealth Mean(st.dev) ‐12814(57000) ‐4823(17795) 2008 ‐24957(70442) ‐10678(24269) 2010 ‐1540(37431) 612(1748) Median ‐3073 13 25thpercentile ‐20397 ‐1209 75thpercentile 0 238  Regressioncoefficient  0.882*** (0.102) NotestoTable: Thesampleisthesamethatisusedinourwealtheffectsregressions.Numberofobservations:6370observations,(3047in2008 and3323in2010)from3867families.441householdsin2008and475in2010,wereshareownersin2006. Monetaryvaluesarein2010euros.TheregressioncoefficientisobtainedbyOLSregressionofthechangeinreportedwealthon theconstructedchangeinwealth(andaconstant).  35  Table4.1:Wealtheffectsregressionsforthechangeintotalhouseholdconsumption Dependentvariable:Changeinhouseholdconsumptionexpenditure OLS IV2 ndStage IncludingΔ LabourIncome NoControlfor Income IncludingΔ LabourIncome NoControlfor Income Wealthvariable:ΔRiskyfinancialwealth  Deltariskyfinancialwealth 0.016 ** 0.016 ** 0.088* 0.086 *  (0.007)  (0.007)  (0.047) (0.047)  Deltahousevalue 0.004 *** 0.004 *** 0.003*** 0.003 ***  (0.001)  (0.001)  (0.001) (0.001)  Deltalabourincome 0.079 ***  0.079***   (0.020)   (0.020)  Deltaunemploymentstatus ‐1556.171 ** ‐1846.143 *** ‐1481.319** ‐1773.569 ***  (612.166)  (616.947)  (618.512) (623.452)  Deltaretirementstatus 425.789  392.921  569.728 532.158   (466.871)  (471.705)  (480.620) (484.088)  Deltano.ofpeopleintheHH 2026.098 *** 2263.764 *** 1930.572*** 2171.226 ***  (294.384)  (292.070)  (292.252) (292.015)  Deltano.ofearnersintheHH 1001.949 *** 1583.958 *** 1039.473*** 1619.893 ***  (297.572)  (276.589)  (297.779) (275.373)  Year2010 295.468  125.952  224.320 57.242   (575.785)  (579.205)  (580.768) (583.006)   Wealthvariable:ΔTotalaccessiblefinancialwealth  Deltafinancialwealth 0.002  0.004  0.100  0.099   (0.003)  (0.003)  (0.070)  (0.071)  Deltahousevalue 0.004 *** 0.004 *** 0.001  0.001   (0.001)  (0.001)  (0.002)  (0.003)  Deltalabourincome 0.078 ***  0.042    (0.020)   (0.037)   Deltaunemploymentstatus ‐1569.980 ** ‐1852.280 *** ‐1427.671** ‐1586.125 **  (610.928)  (616.057)  (659.231) (679.909)  Deltaretirementstatus 395.724  364.399  446.971 428.463   (469.031)  (473.835)  (733.326) (724.879)  Deltano.ofpeopleintheHH 2045.942 *** 2278.874 *** 2005.018*** 2133.766 ***  (295.844)  (293.382)  (305.885) (311.227)  Deltano.ofearnersintheHH 996.616 *** 1572.502 *** 1168.668*** 1479.847 ***  (298.533)  (276.870)  (339.568) (308.515)  Year2010 305.420  131.868  ‐30.916 ‐117.272  (576.194)  (579.429)  (682.737) (657.514) Notestotable:Numberofobservations:6370observationsfrom3867families. Significance:*p<0.1,**p<0.05,***p<0.001.Standarderrorsinparenthesisarerobusttoheteroskedasticityandtocorrelation withinthehousehold. Coefficientsinboldcanbeinterpretedasmpcoutofwealthchange.Alsoincluded:homeownership,retirementandself‐ employmentinthepreviouswave,agedummies(40‐49,50‐59,60‐69,70+),educationdummies(mediumandhigheducation), gender,regionalunemploymentrate,regionaldummies,constantterm. DetailedresultsfromthefirststageregressionsfortheIVmodelsareincludedinAppendixTableB3;F‐statisticsfrom“weak instrument”testsare:toppanel,14.61(column3)and14.51(column4);bottompanel,5.11(column3includingchangeinlabour income)and5.13(column4).   36  Table4.2:Wealtheffectsregressionsforhouseholdconsumptionofnon‐durables Dependentvariable:Changeinhouseholdexpenditureonnon‐durables. OLS IV2 ndStage IncludingΔ LabourIncome NoControlfor Income IncludingΔ LabourIncome NoControlfor Income Wealthvariable:ΔRiskyfinancialwealth  Deltariskyfinancialwealth 0.016 *** 0.016 *** 0.057** 0.055 *  (0.005)  (0.006)  (0.028) (0.029)  Deltahousevalue 0.003 *** 0.003 *** 0.002*** 0.002 ***  (0.001)  (0.001)  (0.001) (0.001)  Deltalabourincome 0.058 ***  0.058***   (0.016)   (0.016)  Deltaunemploymentstatus ‐1165.150 *** ‐1377.210 *** ‐1123.289*** ‐1337.046 ***  (396.578)  (399.493)  (401.656) (404.731)  Deltaretirementstatus ‐352.065  ‐376.102  ‐271.566 ‐299.045   (323.674)  (325.511)  (324.924) (326.027)  Deltano.ofpeopleintheHH 1784.869 *** 1958.676 *** 1731.445*** 1907.463 ***  (213.312)  (212.648)  (214.925) (215.251)  Deltano.ofearnersintheHH 940.634 *** 1366.262 *** 961.619*** 1386.149 ***  (223.564)  (198.373)  (223.897) (198.489)  Year2010 222.021  98.052  182.231 60.027   (412.706)  (416.229)  (413.681) (416.880)   Wealthvariable:ΔTotalaccessiblefinancialwealth  Deltafinancialwealth 0.003  0.005 ** 0.062* 0.061   (0.002)  (0.002)  (0.036)  (0.037)  Deltahousevalue 0.003 *** 0.003 *** 0.001  0.001   (0.001)  (0.001)  (0.001)  (0.001)  Deltalabourincome 0.057 ***  0.035    (0.016)   (0.022)   Deltaunemploymentstatus ‐1177.262 *** ‐1380.942 *** ‐1092.602** ‐1224.290 ***  (395.310)  (398.558)  (429.604) (439.629)  Deltaretirementstatus ‐382.644  ‐405.245  ‐352.157 ‐367.538   (326.928)  (328.934)  (481.610) (473.220)  Deltano.ofpeopleintheHH 1804.936 *** 1972.997 *** 1780.590*** 1887.590 ***  (213.672)  (213.050)  (222.560) (226.206)  Deltano.ofearnersintheHH 937.863 *** 1353.366 *** 1040.217*** 1298.831 ***  (223.419)  (198.625)  (240.356) (206.872)  Year2010 226.944  101.726  26.857 ‐44.912  (412.649)  (415.868)  (455.016) (445.911) Notestotable:Numberofobservations:6370observationsfrom3867families. Significance:*p<0.1,**p<0.05,***p<0.001.Standarderrorsinparenthesisarerobusttoheteroskedasticityandtocorrelation withinthehousehold. Coefficientsinboldcanbeinterpretedasmpcoutofwealthchange.Alsoincluded:homeownership,retirementandself‐ employmentinthepreviouswave,agedummies(40‐49,50‐59,60‐69,70+),educationdummies(mediumandhigheducation), gender,regionalunemploymentrate,regionaldummies,constantterm. DetailedresultsfromthefirststageregressionsfortheIVmodelsareincludedinAppendixTableB3;F‐statisticsfrom“weak instrument”testsare:toppanel,14.61(column3)and14.51(column4);bottompanel,5.11(column3includingchangeinlabour income)and5.13(column4). 37  Table4.3:Wealtheffectcoefficients:IVregressionsforcategoriesofconsumptionexpenditure Dependentvariable: ΔTotalCΔNon‐durableCΔDurablesexpenditureΔFoodexpenditure  (a1) (a2) (b1) (b2) (c1) (c2) (d1) (d2)  Wealthvariable:ΔRiskyfinancialwealth    Deltariskyfinancialwealth 0.088* 0.086 * 0.057 ** 0.055 * 0.031  0.031  0.015 * 0.015*  (0.047) (0.047)  (0.028)  (0.029)  (0.041)  (0.040)  (0.008)  (0.008) Deltahousevalue 0.003*** 0.003 *** 0.002 *** 0.002 *** 0.001  0.001  0.000  0.000  (0.001) (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  (0.000)  (0.000) Deltalabourincome 0.079***  0.058 ***  0.021 ***  0.009 **   (0.020)  (0.016)   (0.008)   (0.004)     Notestotable:Numberofobservations:6370observationsfrom3867families. Significance:*p<0.1,**p<0.05,***p<0.001. Coefficientsinboldcanbeinterpretedasmpcoutofwealthchange.Standarderrorsinparenthesisarerobusttoheteroskedasticityandtocorrelationwithinthehousehold. Alsoincluded:homeownership,retirementandself‐employmentinthepreviouswave,agedummies(40‐49,50‐59,60‐69,70+),educationdummies(mediumandhigheducation),gender,regional unemploymentrate,regionaldummies,constant. Detailedresultsfromthefirst‐stageregressionsfortheIVmodelsareincludedinAppendixTableB3.F‐statisticsfrom“weakinstrument”testsare:14.61(includingchangeinlabourincome)and 14.51(nocontrolforchangeinlabourincome);bottompanel.  38  Table4.4:CounterfactualExercises:PredictedChangesinConsumption FullSample 2006–08Sample   Averageobservedchangeintotalconsumption ‐515 (100%) ‐796 (100%) Counterfactualchanges    Δriskyfinancialwealthsetto0 ‐425 (83%) ‐619 (78%) Δhousingvaluesetto0 ‐498 (97%) ‐770 (97%) Δlabourincomesetto0 ‐437 (85%) ‐753 (95%) ΔnoearnersintheHHsetto0 ‐494 (96%) ‐781 (98%) Nounemployment ‐464 (90%) ‐748 (94%) Δnoearnerssetto0andnounemployment ‐443 (86%) ‐733 (92%) ΔnoearnersandΔlabourincomesetto0and nounemployment ‐365 (71%) ‐690 (87%) Notestotable:ThesecounterfactualsarebasedontheIVregressionreportedinthetoppanelofTable4.1,includingΔlabour income(column2). Thefullsamplesizeis6370whilethe2006‐08samplehas3047observations.Themeanlevelofconsumptionis17454inthefull sampleand17627inthe2006‐08subsample.Thepercentagesinparenthesesarethepercentageoftheaverageobservedchange.   39  Table4.5:Heterogeneityinkeyregressioncoefficientsbyage(IV2ndstage)  Age<70 Age50+ Age50‐69 Dependentvariable:ΔTotalconsumption  Deltariskyfinancialwealth 0.114 * 0.131 * 0.218***  (0.059)  (0.070)  (0.084) Deltahousevalue 0.004 ** 0.003 ** 0.002  (0.001)  (0.001)  (0.002) Deltalabourincome 0.077 *** 0.077 *** 0.073***  (0.022)  (0.019)  (0.019) Dependentvariable:ΔNon‐durablesconsumption    Deltariskyfinancialwealth 0.075 ** 0.056  0.095***  (0.032)  (0.036)  (0.035) Deltahousevalue 0.002 * 0.003 *** 0.002  (0.001)  (0.001)  (0.001) Deltalabourincome 0.054 *** 0.056 *** 0.047***  (0.016)  (0.016)  (0.016) Dependentvariable:ΔDurablesconsumption  Deltariskyfinancialwealth 0.039  0.075  0.123  (0.054)  (0.054)  (0.085) Deltahousevalue 0.002  0.000  0.000  (0.001)  (0.001)  (0.002) Deltalabourincome 0.023 *** 0.021 ** 0.026**  (0.009)  (0.010)  (0.011) Dependentvariable:ΔFoodconsumption  Deltariskyfinancialwealth 0.016  0.017  0.021  (0.010)  (0.012)  (0.019) Deltahousevalue ‐0.000  0.000  ‐0.000 (0.000) (0.000) (0.001) Deltalabourincome 0.008 ** 0.009 * 0.007  (0.004)  (0.005)  (0.004)  Notestotable:Numberofobservations:4335observationsifage<70;4885ifage50+;2850ifage50‐69.Coefficientsinboldcan beinterpretedasmpcoutofwealthchange. Significance:*p<0.1,**p<0.05,***p<0.001.Standarderrorsinparenthesisarerobusttoheteroskedasticityandtocorrelation withinthehousehold.Alsoincluded:homeownership,retirementandself‐employmentinthepreviouswave,agedummies(40‐49, 50‐59,60‐69incolumns1and2;60‐69,70+incolumns3and4;60‐69incolumns5and6),educationdummies(mediumandhigh education),gender,regionalunemploymentrate,region,2010dummy,changein:unemploymentstatus,retirementstatus,no.of peopleinthehousehold,no.ofearnersinthehouseholdandconstantterm.F‐statisticsfromweakidentificationtestsare,inorder ofcolumns:age<70:9.18and9.04;age50+:22.09and22.14;age50‐69:13.97and13.99. 46  AppendixTableB6:FullresultsforregressionsreportedinthebottompanelofTable4.1 Dependentvariable:Changeinhouseholdconsumptionexpenditure OLS IV2 ndStage  IncludingΔLabour Income NoControlfor Income IncludingΔLabour Income NoControlfor Income Deltafinancialwealth 0.002 0.004 0.100 0.099 (0.003) (0.003) (0.070) (0.071) Deltahousevalue 0.004*** 0.004 *** 0.001 0.001 (0.001) (0.001) (0.002) (0.003) Deltalabourincome  0.078*** 0.042 (0.020) (0.037) Deltaunemployment status ‐1569.980** ‐1852.280 *** ‐1427.671** ‐1586.125 ** (610.928) (616.057) (659.231) (679.909) Deltaretirementstatus 395.724 364.399 446.971 428.463 (469.031) (473.835) (733.326) (724.879) Deltano.ofpeopleinthe HH 2045.942*** 2278.874 *** 2005.018*** 2133.766 *** (295.844) (293.382) (305.885) (311.227) Deltano.ofearnersin theHH 996.616*** 1572.502 *** 1168.668*** 1479.847 *** (298.533) (276.870) (339.568) (308.515) Year2010 305.420 131.868 ‐30.916 ‐117.272  (576.194) (579.429) (682.737) (657.514) Age40‐49  ‐183.343 ‐63.027 ‐19.155 42.511 (501.452) (501.226) (524.039) (518.598) Age50‐59  ‐398.458 ‐272.016 ‐390.774 ‐321.666 (488.321) (490.100) (503.900) (509.482) Age60‐69  ‐1188.404** ‐1133.505 ** ‐954.727 ‐930.731 (525.747) (527.761) (689.408) (678.795) Age70+  ‐509.693 ‐433.666 ‐306.486 ‐270.115 (497.815) (500.194) (588.860) (578.675) Mediumeducation ‐428.612* ‐442.951 * ‐275.921 ‐287.765  (228.634) (229.772) (295.551) (295.279) Higheducation ‐42.463 ‐160.699 ‐505.516 ‐558.245  (468.235) (477.530) (659.186) (639.188) Regionalunemployment rate 180.764 246.045 257.430 291.214 (310.366) (312.239) (371.248) (360.952) Male  ‐7.218 ‐58.657 171.405 138.548 (187.410) (189.756) (233.703) (247.867) Retired(t‐1) 107.385 159.417 66.702 96.285  (264.438) (266.266) (384.998) (385.220) Publicsectoremployee (t‐1) 11.782 32.587 179.582 186.609 (263.403) (266.759) (387.998) (382.571) Homeowner(t‐1)  190.416 180.988 63.576 61.718 (209.339) (210.899) (244.676) (244.192) Regiondummies Yes Yes Yes Yes Constant ‐1316.500 ‐1748.618 ‐1691.330 ‐1918.413  (1757.985) (1770.439) (2155.759) (2084.642) Notestotable:Numberofobservations:6370observationsfrom3867families. Significance:*p<0.1,**p<0.05,***p<0.001.Standarderrorsinparenthesisarerobusttoheteroskedasticityandtocorrelation withinthehousehold. Coefficientsinboldcanbeinterpretedasmpcoutofwealthchange. DetailedresultsfromthefirststageregressionsfortheIVmodelsareincludedinAppendixTableB4;F‐statisticsfromweak identificationtestsare:5.11(column2includingchangeinlabourincome)and5.13(column4).   47  AppendixTableB7:FullresultsforregressionsreportedinthetoppanelofTable4.2 Dependentvariable:Changeinhouseholdexpenditureonnon‐durables. OLS IV2 ndStage  IncludingΔLabour Income NoControlfor Income IncludingΔLabour Income NoControlfor Income Deltariskyfinancial wealth 0.016*** 0.016 *** 0.057** 0.055 * (0.005) (0.006) (0.028) (0.029) Deltahousevalue 0.003*** 0.003 *** 0.002*** 0.002 *** (0.001) (0.001) (0.001) (0.001) Deltalabourincome  0.058*** 0.058*** (0.016) (0.016) Deltaunemployment status ‐1165.150*** ‐1377.210 *** ‐1123.289*** ‐1337.046 *** (396.578) (399.493) (401.656) (404.731) Deltaretirementstatus ‐352.065 ‐376.102 ‐271.566 ‐299.045 (323.674) (325.511) (324.924) (326.027) Deltano.ofpeopleinthe HH 1784.869*** 1958.676 *** 1731.445*** 1907.463 *** (213.312) (212.648) (214.925) (215.251) Deltano.ofearnersin theHH 940.634*** 1366.262 *** 961.619*** 1386.149 *** (223.564) (198.373) (223.897) (198.489) Year2010 222.021 98.052 182.231 60.027  (412.706) (416.229) (413.681) (416.880) Age40‐49  183.514 270.518 248.173 332.362 (343.934) (344.675) (348.224) (348.701) Age50‐59  122.799 216.729 183.567 274.847 (333.565) (334.925) (337.172) (338.412) Age60‐69  ‐611.289* ‐573.987 ‐556.063 ‐521.148 (359.123) (359.881) (364.096) (364.838) Age70+  ‐3.110 50.370 65.114 115.642 (351.374) (353.292) (355.196) (357.082) Mediumeducation ‐272.787 ‐285.728 * ‐180.239 ‐197.144  (167.019) (168.160) (176.888) (178.798) Higheducation ‐69.782 ‐150.635 36.895 ‐48.498  (403.329) (408.166) (413.807) (419.285) Regionalunemployment rate 236.447 283.807 186.308 235.797 (216.548) (219.045) (219.213) (221.485) Male  144.589 103.644 181.580 139.068 (136.186) (138.519) (138.143) (140.533) Retired(t‐1) ‐27.285 12.040 11.873 49.500  (193.719) (195.035) (198.730) (199.634) Publicsectoremployee (t‐1) 242.022 254.941 225.333 238.962 (192.013) (194.051) (192.235) (194.169) Homeowner(t‐1)  289.372* 284.292 * 305.660** 299.884 * (153.345) (155.092) (153.536) (155.075) Regiondummies Yes Yes Yes Yes Constant ‐1855.587 ‐2171.142 * ‐1654.803 ‐1978.826  (1224.313) (1239.657) (1236.205) (1249.708) Notestotable:Numberofobservations:6370observationsfrom3867families. Significance:*p<0.1,**p<0.05,***p<0.001.Standarderrorsinparenthesisarerobusttoheteroskedasticityandtocorrelation withinthehousehold. Coefficientsinboldcanbeinterpretedasmpcoutofwealthchange. DetailedresultsfromthefirststageregressionsfortheIVmodelsareincludedinAppendixTableB3;F‐statisticsfromweak identificationtestsare:14.61(column2includingchangeinlabourincome)and14.51(column4).   48  AppendixTableB8:FullresultsforregressionsreportedinthebottompanelofTable4.2 Dependentvariable:Changeinhouseholdexpenditureonnon‐durables. OLS IV2 ndStage  IncludingΔLabour Income NoControlfor Income IncludingΔLabour Income NoControlfor Income Deltafinancialwealth 0.003 0.005 ** 0.062* 0.061 (0.002) (0.002) (0.036) (0.037) Deltahousevalue 0.003*** 0.003 *** 0.001 0.001 (0.001) (0.001) (0.001) (0.001) Deltalabourincome 0.057*** 0.035 (0.016) (0.022) Deltaunemployment status ‐1177.262*** ‐1380.942 *** ‐1092.602** ‐1224.290 *** (395.310) (398.558) (429.604) (439.629) Deltaretirementstatus ‐382.644 ‐405.245 ‐352.157 ‐367.538 (326.928) (328.934) (481.610) (473.220) Deltano.ofpeopleinthe HH 1804.936*** 1972.997 *** 1780.590*** 1887.590 *** (213.672) (213.050) (222.560) (226.206) Deltano.ofearnersin theHH 937.863*** 1353.366 *** 1040.217*** 1298.831 *** (223.419) (198.625) (240.356) (206.872) Year2010 226.944 101.726 26.857 ‐44.912  (412.649) (415.868) (455.016) (445.911) Age40‐49  162.985 249.793 260.660 311.910 (343.488) (344.321) (358.345) (357.159) Age50‐59  98.699 189.927 103.271 160.704 (333.500) (334.835) (346.746) (347.991) Age60‐69  ‐625.762* ‐586.152 ‐486.747 ‐466.804 (358.873) (359.489) (452.627) (446.136) Age70+  ‐23.790 31.064 97.098 127.326 (351.182) (353.013) (404.145) (399.064) Mediumeducation ‐304.870* ‐315.216 * ‐214.034 ‐223.877  (167.472) (168.335) (203.187) (203.012) Higheducation ‐127.705 ‐213.012 ‐403.177 ‐446.999  (404.240) (409.117) (479.710) (473.066) Regionalunemployment rate 259.051 306.152 304.661 332.737 (216.690) (219.084) (244.007) (240.080) Male  135.619 98.505 241.882 214.575 (136.120) (138.241) (159.303) (167.029) Retired(t‐1) ‐44.310 ‐6.768 ‐68.511 ‐43.926  (193.741) (194.831) (257.788) (257.187) Publicsectoremployee (t‐1) 254.209 269.220 354.034 359.873 (192.961) (194.936) (253.385) (250.357) Homeowner(t‐1)  278.689* 271.886 * 203.232 201.687 (153.298) (155.014) (169.902) (170.456) Regiondummies Yes Yes Yes Yes Constant ‐1948.330 ‐2260.104 * ‐2171.317 ‐2360.042 *  (1224.689) (1239.576) (1416.872) (1386.649) Notestotable:Numberofobservations:6370observationsfrom3867families. Significance:*p<0.1,**p<0.05,***p<0.001.Standarderrorsinparenthesisarerobusttoheteroskedasticityandtocorrelation withinthehousehold. Coefficientsinboldcanbeinterpretedasmpcoutofwealthchange. DetailedresultsfromthefirststageregressionsfortheIVmodelsareincludedinAppendixTableB4;F‐statisticsfromweak identificationtestsare:5.11(column2includingchangeinlabourincome)and5.13(column4).  49  AppendixTableB9:FullresultsforregressionsreportedinTable4.3 Dependent variable: ΔTotalCΔNon‐durableC ΔDurablesexpenditures ΔFoodexpenditure (a1) (a2) (b1) (b2) (c1) (c2) (d1) (d2) Deltarisky financialwealth 0.088* 0.086 * 0.057 ** 0.055 * 0.031 0.031 0.015 * 0.015 * (0.047) (0.047) (0.028) (0.029) (0.041) (0.040) (0.008) (0.008) Deltahousevalue 0.003*** 0.003 *** 0.002 *** 0.002 *** 0.001 0.001 0.000 0.000 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.000) (0.000) Deltalabour income 0.079*** 0.058 *** 0.021*** 0.009 ** (0.020) (0.016) (0.008) (0.004) Delta unemployment status ‐1481.319** ‐1773.569 *** ‐1123.289 *** ‐1337.046 *** ‐358.030 ‐436.523 ‐209.005 ‐243.517 (618.512) (623.452) (401.656) (404.731) (462.310) (462.777) (183.093) (183.091) Deltaretirement status 569.728 532.158 ‐271.566 ‐299.045 841.294** 831.204 ** ‐117.269 ‐121.706 (480.620) (484.088) (324.924) (326.027) (367.328) (368.194) (174.683) (174.288) Deltano.of peopleintheHH 1930.572*** 2171.226 *** 1731.445 *** 1907.463 *** 199.127 263.763 971.090 *** 999.509 *** (292.252) (292.015) (214.925) (215.251) (193.044) (191.205) (101.054) (100.635) Deltano.of earnersintheHH 1039.473*** 1619.893 *** 961.619 *** 1386.149 *** 77.854 233.744 225.739 ** 294.280 *** (297.779) (275.373) (223.897) (198.489) (203.553) (206.509) (91.141) (88.480) Year2010 224.320 57.242 182.231 60.027 42.089 ‐2.785 818.169 *** 798.439 ***  (580.768) (583.006) (413.681) (416.880) (408.083) (407.970) (178.065) (177.679) Age40‐49  ‐45.521 69.583 248.173 332.362 ‐293.694 ‐262.779 ‐234.801 * ‐221.208* (509.992) (509.534) (348.224) (348.701) (376.033) (375.903) (131.555) (131.431) Age50‐59  ‐266.614 ‐141.816 183.567 274.847 ‐450.181 ‐416.662 ‐334.795 *** ‐320.058 ** (491.351) (492.696) (337.172) (338.412) (364.453) (364.787) (129.529) (129.507) Age60‐69  ‐1072.162** ‐1024.426 * ‐556.063 ‐521.148 ‐516.099 ‐503.278 ‐427.830 *** ‐422.193 *** (533.083) (535.024) (364.096) (364.838) (403.132) (403.394) (150.248) (149.960) Age70+  ‐364.744 ‐295.661 65.114 115.642 ‐429.858 ‐411.303 ‐360.494 ** ‐352.336 ** (502.262) (504.289) (355.196) (357.082) (371.172) (371.049) (140.540) (140.555) Medium education ‐230.043 ‐253.155 ‐180.239 ‐197.144 ‐49.803 ‐56.011 ‐150.156 ** ‐152.885 ** (242.037) (242.876) (176.888) (178.798) (173.796) (173.428) (72.820) (73.038) Higheducation 196.508 79.759 36.895 ‐48.498 159.613 128.257 ‐367.525 ** ‐381.312 **  (493.498) (503.352) (413.807) (419.285) (305.937) (308.428) (159.206) (159.245) 50  Regional unemployment rate 70.671 138.332 186.308 235.797 ‐115.637 ‐97.464 ‐365.621 *** ‐357.631 *** (314.687) (316.214) (219.213) (221.485) (220.921) (220.754) (91.909) (91.693) Male  70.295 12.172 181.580 139.068 ‐111.286 ‐126.896 ‐42.829 ‐49.693 (189.330) (191.626) (138.143) (140.533) (134.885) (134.884) (58.244) (58.227) Retired(t‐1) 193.063 244.507 11.873 49.500 181.190 195.007 28.906 34.981  (274.934) (276.276) (198.730) (199.634) (198.989) (199.575) (93.649) (93.622) Publicsector employee(t‐1) ‐27.110 ‐8.476 225.333 238.962 ‐252.443 ‐247.438 33.868 36.069 (266.625) (269.684) (192.235) (194.169) (189.575) (190.001) (79.650) (79.801) Homeowner(t‐1) 227.790 219.892 305.660 ** 299.884 * ‐77.871 ‐79.992 50.167 49.234 (210.334) (211.611) (153.536) (155.075) (136.246) (136.144) (65.908) (66.150) Regiondum.s Yes Yes Yes Yes Yes Yes Yes Yes Constant ‐874.557 ‐1317.563 ‐1654.803 ‐1978.826 780.246 661.262 2396.792 *** 2344.478 ***  (1780.904) (1790.618) (1236.205) (1249.708) (1269.021) (1268.633) (515.582) (514.511) Notestotable:Numberofobservations:6370observationsfrom3867families. Significance:*p<0.1,**p<0.05,***p<0.001.Standarderrorsinparenthesisarerobusttoheteroskedasticityandtocorrelationwithinthehousehold. Coefficientsinboldcanbeinterpretedasmpcoutofwealthchange. DetailedresultsfromthefirststageregressionsfortheIVmodelsareincludedinAppendixTableB3;F‐statisticsfromweakidentificationtestsare:14.61(includingchangeinlabourincome)and 14.51(nocontrolforchangeinlabourincome). 51  AppendixTableB10:Percentowningriskyassetsbyage‐band  Ageband Lessthan50 50‐69 70and above Samplesize Percentagethatownriskyassets 15.2% 16.9% 10.2% 6370  NotestoTable:Ownershipofriskyassetsisdefinedintermsofhavinganon‐zerovalueforourexcludedinstrument,sois measuredin2004or2006.   AppendixTableB11: Percentbecomingpessimisticandwithamortgage,byownershipofriskyassets  Sample All Thoseowning riskyassets Samplesize Becomepessimistic 12.3% 23.3% 3327 Havemortgage 11.1% 14.5% 5536  NotestoTable:Ownershipofriskyassetsisdefinedintermsofhavinganon‐zerovalueforourexcludedinstrument,sois measuredin2004or2006.Samplesizesarefortheregressionsreported,respectively,inTables4.6and4.7.    