ContentslistsavailableatScienceDirect
Journal of Economic Behavior & Organization
jo u r n al h om ep age :w w w . e l s e v i e r . c o m / l o c a t e / j e b o
House price responses to a national property tax reform
夽Mikael Elinder
a,b,c, Lovisa Persson
c,baDepartmentofEconomics,UppsalaUniversity,P.O.Box513,SE-75120Uppsala,Sweden
bUppsalaCenterforFiscalStudies(UCFS),UppsalaUniversity,Sweden
cResearchInstituteofIndustrialEconomics(IFN),P.O.Box55665,SE-10215Stockholm,Sweden
a r t i c l e i n f o
Articlehistory:
Received23November2016 Receivedinrevisedform29May2017 Accepted17September2017 Availableonline22September2017
Keywords:
Announcementeffects Capitalization Boundedrationality Housingmarket Inattention Saliency JELclassification:
D1 D3 D4 H2 R3
a b s t r a c t
Weshowthathousepricesingeneraldidnotrespondtoasubstantialcutinthenational propertytaxinSweden.Theestimatesarebasedonrichregisterdatacoveringmorethan 100,000salesoveratimeperiodoftwoandahalfyears.BecausetheSwedishpropertytaxis nationalandthusunrelatedtolocalpublicgoods,oursettingisidealforcausalidentification ofthepropertytaxonhouseprices.Weobservepriceincreasesonlyinasmallsegmentof themarketcontainingpropertieswithveryhightaxvalues.Wediscuss,butcanadmittedly notempiricallydiscriminatebetween,severalpotentialexplanationsforwhywefindno evidenceofcapitalizationexceptforthetopsegmentofthemarket.
©2017TheAuthor(s).PublishedbyElsevierB.V.Thisisanopenaccessarticleunderthe CCBYlicense(http://creativecommons.org/licenses/by/4.0/).
1. Introduction
Supposeyouhavedecidedtobuyahouse.Twohousesappeartobenearlyidenticalbutoneofthemisslightlyless expensivetoownbecausetheannualpropertytaxburdenisD100lowercomparedwiththeotherhouse.Howmuchmore wouldyoubewillingtopayforthelowertaxedhouse?Thisquestionisattheheartofthestandardcapitalizationtheory (Oates,1969;Yinger,1982)wherethepriceofahouseisdeterminedbythetotalstreamofhousingservicesminusthenet presentvalueofallcostsofowningthehouse.Whenthepropertytaxdecreases,buyersrealizethatthecostoflivingalso decreasesandtheyarethuswillingtopayahigherpriceforthehouse.Ifthesupplyoflandandhousingisfixed,themarket pricewillincreasewiththefullnetpresentvalueofthetaxreduction.Furthermore,ifthehousingmarketisefficientand
夽 WewouldliketothankAronBergandNinaÖhrnKarlssonforexcellentresearchassistance,NiclasBerggren,SebastienBradley,GabriellaChirico, PerEngström,JonFiva,HenrikJordahl,Che-YuanLiang,HeléneLundqvist,EvaMörk,HåkanSelin,ErikSpector,OskarTysklindandseminarpartici- pantsattheInternationalInstituteofPublicFinance’sannualcongress,theMinistryofFinance,theResearchInstituteofIndustrialEconomics(IFN),the SwedishConferenceinEconomics,andUppsalaUniversityforvaluablecomments,andtheSwedishResearchCouncilforfinancialsupport(GrantNumber:
217531001).
E-mailaddresses:mikael.elinder@nek.uu.se(M.Elinder),lovisa.persson@ifn.se(L.Persson).
http://dx.doi.org/10.1016/j.jebo.2017.09.017
0167-2681/©2017TheAuthor(s).PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/
by/4.0/).
individualsuseallrelevantinformation,priceswillchangeimmediatelywheninformationaboutfuturetaxchangesismade public(PalmonandSmith,1998;RossandYinger,1999).
Mostearlierstudiesonpropertytaxcapitalizationfocusonlocalpropertytaxesandsupportthepredictionthatlower propertytaxesleadtohigherhouseprices(seee.g.thereviewsbyRossandYinger,1999;Sirmansetal.,2008;Hilber, 2015).However,recentcontributions,usingricherdataandmorecredibleidentificationmethodsaddimportantinsights totheliterature.Bradley(2015)findsthatatemporarytaxrebateleadstopriceincreasesthataremuchlargerthanthe net-present-valueoftherebate.Heinterpretsthisfindingasbeinginconsistentwithstandardcapitalizationtheoryandas evidenceofboundedrationalityamonghousebuyers.Moreover,Lutz(2015)andHilberandVermeulen(2016)findthat capitalizationdegreesvarywiththeelasticityofthehousingsupply.Theimportanceofpayingattentiontothesupplyside inempiricaltestsofcapitalizationtheoryisextensivelydiscussedinHilber(2015).
InthispaperweempiricallyanalyzehowhousepricesrespondedtoareformthatsubstantiallyreducedtheSwedish nationalpropertytaxonowneroccupiedresidentialproperties(referredtoasjusthousesorproperties).Thereformis frequentlyreferredtoasan“abolition”inSwedishpolicydiscussionsduetotheextensivenessofthereduction.Thereform isremarkablealsointhesensethatownersofveryexpensivepropertiesgotdisproportionatelylargetaxreductions.The taxwasreducedintwosteps;apreliminaryreformincludingamedium-sizedtaxcutwasintroducedimmediatelyafter thecenter-rightcoalitionhadwonthe2006parliamentaryelection.Apermanentandfinalreformincludingalargertaxcut wasimplementedon1January2008.Priortoanyofthetwosteps,theyearlytaxpaymentequaled1percentoftheproperty valueasassessedbytheSwedishTaxAgency(herebyreferredtoasthe“taxvalue”).Thefinalreformpackageincludeda decreaseofthetaxratefrom1percentto0.75percent.Butthemostsweepingpartofthereformwastheintroductionofa caponyearlytaxpaymentsatSEK6000($710orD630).1Thecapwasbindingforroughlyhalfofallproperties.Thecapping ofpropertytaxpaymentsimpliesadisproportionatelylargedecreaseinthetaxliabilityforpropertiesabovethecap.
We utilizethisdifferentialtreatmentin adifference-in-difference(DiD)approach, withuncappedproperties inthe controlgroupandcappedpropertiesinthetreatmentgroup.Undertheassumptionofparalleltrendsinpricedevelopments forhousesofdifferenttaxvalues,weestimatethecausaleffectofthetaxcaponhouseprices.Ourdataisobtainedfrom theofficialhomeownershipregisterandcoversall(roughly100,000)salesofsingle-familyhouses,mediatedthroughareal estateagentconnectedtoSvenskMäklarstatistikAB(Ourtranslation:SwedishRealEstateAgentStatistics,Inc.)inSweden duringthethreeyearsthatspanthereformperiod:2006,2007and2008.
Mostoftheearlierstudiesonpropertytaxcapitalizationanalyzelocalorregionalcross-sectionalvariationinproperty taxrates.Therearetwofundamentalidentificationproblemswiththatapproach,andtheseproblemshavebeenknownand discussedsincetheseminalpaperbyOates(1969).First,ahighertaxrateimplieshighertaxrevenuesandconsequently higherqualityofpublicgoods.Higherqualityofpublicgoodsputsupwardpressureonhouseprices,makingitdifficultto isolatetheeffectofthetaxseparately.Controllingforpublicgoodsqualityhasbeenthemainconcernsoastoavoidbiased estimates,butthistaskhasprovendifficult.Second,whenlocalgovernmentssettheirtaxrate,areaswithhigherhouse prices,allelseequal,areabletosetalowertaxratetocollectagivenamountoftaxrevenues.Thiscreatesasimultaneity biasbetweenthepropertytaxrateandhouseprices.AkeyadvantageofourstudyisthattheSwedishpropertytaxrate issetatthenationallevel,withoutconcernforlocalhousepricesorqualityoflocalpublicgoods.Withourapproach,we avoidthetwoidentificationproblemsexplainedabove.Usingvariationstemmingfromanationalreformhasanadditional advantage,namelythatwecantracktheeffectsofallreformevents,includingpolicyannouncements.2
Inthiscontextitisimportanttonotethatresponsestolocalandnationalpropertytaxchangesneednotbeidentical.
Whenalocalpropertytaxisraisedinonejurisdiction,itisinprinciplepossibleforabuyertobuyanidenticalproperty inanotherjurisdictionwithalowertaxburden.Thisdecreasesdemandinthejurisdictionwherethepropertytaxwas raisedandincreasesdemandintheotherjurisdiction,withcorrespondingpriceadjustments.Whenanationalpropertytax ischanged,allidenticalpropertiesareaffectedbythetaxchange.Withakinkednationalpropertytaxschedule,asinthe post-reformregimeinSweden,demandforpropertieswithhightaxvaluesincreasesrelativetodemandforpropertieswith taxvaluesbelowthekink.However,sincetwopropertieswithdifferenttaxvaluesarearguablynotidentical,andhencenot perfectsubstitutes,itispossiblethatthehousingdemandelasticitywithrespecttonationalpropertytaxislowerthanfor alocalpropertytax.Ifhousingsupplyisperfectlyinelastic,thedemandelasticitydoesnotaffectcapitalizationdegrees,but withelasticsupplythecapitalizationdegreemightwellbelowerforanationalpropertytaxthanforalocalpropertytax.It shouldalsobenotedthatwhilemostoftheearlierliteraturehasusedcross-sectionalvariationinpropertytaxesandhouse prices,andhenceimplicitlycomparedlong-runequilibriumoutcomes,ourreformevaluationapproachmakesitpossibleto studydynamicadjustmenttochangesinthetaxes.Ourstudycontributestothepreviousliteratureestimatingcapitalization ofpropertytaxesinthatwemakeuseofanationalreformandalargenationwideregisterbaseddatasettoidentifyprice responses.
Fig.1.Timingofeventsandpricedevelopmentofhousesincontrolandtreatmentgroups.Note:Thedataiscollapsedonmonthlylevelandseasonally adjustedwithgroupspecificcalendarmonthdummies.
Fig.1showsthetimingofreformeventsandthedevelopmentofhousepricesforpropertieswithtaxvaluesaboveand belowthetaxcap.Notethattaxvaluesweresetin2006andwerenotchangeduntil2009,andarethusnotendogenousto pricechangesinducedbythereform.Thesolidthicklinerepresentssalespricesinthecontrolgroup,i.e.propertieswith taxvalueslowerthanthecut-offwhichisimpliedbythecap.Thepricedevelopmentinthisgroupservesasacounter- factualforthepropertiesinthethreetreatedgroups,whichcontainshighervaluedproperties.InFig.1wedisplaythree treatmentgroups;propertieswiththetop1percenthighesttaxvalues,thetop5percent(excludingthetop1percent), andtheremainingtreatedpropertiesseparately.Upuntil1January2007,thepricedevelopmentisfairlysimilarinallfour
1 $1≈SEK8.45andD1≈SEK9.51(September2016).
2 Arelated,andquitelarge,literatureinvestigateshowotherfiscalvariablesliketransfertaxes(e.g.Besleyetal.,2014;KopczukandMunroe,2015;Best andKleven,2016;Slemrodetal.,2016),centralgovernmentgrants(Hilberetal.,2011),localincometaxes(Boije,1997),specificsubsidies(Bergeretal., 2000;HilberandTurner,2014)orneighborhoodcharacteristicssuchasschoolquality(e.g.Black,1999;FiglioandLucas,2004;FivaandKirkebøen,2011), airpollution(ChayandGreenstone,2005),oravailabilityofreligiousamenities(BlindandDahlberg,2015)affectshouseprices.
groups.Fromthatdate,whichcoincideswiththefirstpartofthereform,thepricesofthepropertiesinthetwohighest valuedgroupsdivergesupwards,andthetop1percentmoreso.3Thelargestofthetreatmentgroups,whichcontainsthe remainingpropertiesbelowthetop5percentdonotrespondatalltoanyofthereformevents.Ourinterpretationofthe patternshowninFig.1,whichissupportedbyalltheeconometricanalyzes,isthattheintroductionofacaponproperty taxesdidnotleadtoanyhousepriceincreasesforthevastmajorityofpropertiesthatwereboundbythiscap.Mosttreated housessharethesametrendasthecontrolgroupduringthetimeperiodthatwestudy,whichcontainsalltheimportant reformevents.However,thetaxcapseemstohavehadaneffectforveryhighlyvaluedproperties.Thesehousesaretypically situatedinareasinoraroundthethreegreatercitiesofSweden,mostnotablyaroundthecapitalofStockholm.Theeffect sizeofthetop1percentisa8.4percentagepointsincreaserelativetothecontrolgroup.Althoughasubstantialeffect,the estimateissmallerthantheexpectedpriceresponse(undertheassumptionsofrationalexpectationsandfixedsupply) whichis16.0percentagepoints.
OurresultsareconsistentwithcapitalizationtheoryonlyifthesupplyofhousingisperfectlyelasticinallareasinSweden excepttheareasinwhichthemostexclusivehousesarelocated.However,wefindtheregulatory,politicalandeconomic restrictionstobetooserioustowarrantsuchanassumptionabouttheSwedishhousingmarketasawhole.Thisstatement issupportedbystudiesoftheSwedishhousingmarketthathavecharacterizedthesupplyelasticityaslowingeneral(see e.g.HüfnerandLundsgaard,2007).Instead,wehavetolookforcomplementaryexplanationsforourzero-results.Besides variationsinsupplyelasticity,weproposeheterogeneityinsaliency,financialliteracy,frictionsandexpectationsofpolicy reversalaspotentialmechanisms.Furtherstudiesoncapitalizationofpropertytaxesareneededtoconfirmorrejectthe relevanceofeachofthesemechanisms.
2. TheSwedishpropertytaxandthereform
Between1985and2008,theSwedishpropertytaxwasuniformacrossthecountryandprovidedrevenuestothenational government.4TheguidingprincipleofpropertytaxationinSwedenisthatpropertyownersareliabletopayapercentageof thetaxvalueofthepropertyonayearlybasis.From2001until2008thetaxratewas1percentofthetaxvalue.Thetaxvalue isreassessedbytheSwedishTaxAgency(Swedish:Skatteverket)everythirdyearandisapproximately75percentofthe propertymarketvalueatthetimeofassessment.Severalvariablesenterintothealgorithmthatdeterminesthetaxvalue, suchas:theaveragemarketpriceofsurroundingproperties,andindividualpropertycharacteristicssuchashousesize.The totaltaxvalueconsistsofthesumoftheassessedvalueofbuildingsandthevalueoftheland.Amajoradvantageforour purposesisthatanationalassessmenttookplacein2006.5 Allthetaxvaluesinourdatawereconsequentlyestablished beforethereformcouldhavehadanyeffectonmarketprices,andthereforeonthetaxvaluesthemselves.Thenextnational assessmenttookplacein2009.Thepropertytaxispaidbytheownerofthehouseinconnectionwiththeannualtax settlement,whichlistsalloftheindividualtaxliabilitiespertainingtothepreviousyear.Theindividualdoesnothaveto reporthouseownershipandvaluesincethetaxauthorityretrievesthisinformationfromthepropertyownershipregister heldbytheSwedishlandsurveyingoffice,whichisthesameregisterfromwhichweretrieveddataforthispaper.
2.1. Thepropertytaxreform
Beforetheelectiontothenationalparliamentin2006fouroppositionparties(theModerateParty,theLiberalParty, theCenterPartyandtheChristianDemocrats)formedthecoalition“AllianceforSweden”(henceforth:theAlliance).In thecampaigntheypromisedthattheywouldsubstantiallycutthepropertytax,wouldtheywintheelection.Immediately afterthevictory,theAlliancedeclaredthattheywoulddeliverontheirelectionpromisetosubstantiallylowerthenational propertytax.6Measuresweretakenalreadyin2007withtheimplementationofatemporarytaxcutwhilethedetailsof apermanentreformwerediscussed.Thefinalreform,implementedon1January2008,impliedareductioninthetaxrate from1percentto0.75percentofthetaxvalueandacaponyearlytaxliabilitiesatSEK6000.Thecapwasbindingforroughly halfofallproperties.TheaverageyearlyreductionintaxliabilitiesfromthereformamountedtoSEK4900andtheaverage netpresentvalueofthetaxreductionwasSEK243,000;14percentoftheaveragemarketprice.7
In2006,beforethereform,propertytaxrevenuesconstitutedslightlylessthan3percentofalltaxrevenuesinSweden.
Afterthereform,thesharewasreducedto2.3percent.Thesepropertytaxrevenuesincludealsopropertytaxesoncommercial andindustrialbuildings,aswellasmulti-familyhousing.ComparedwiththeOECDaverageof5.4percent(ortheUSorUK whichbothcollectmorethan10percentoftheirtaxrevenuesfrompropertytaxes),relianceonpropertytaxeswas,andstill
3Welimittheanalysistothefirsthalfof2008sincethefinancialcrisishitthemarketinthefallof2008andcausedahugedropinthenumberofsold properties.SeeAppendixA.2formorediscussiononthefinancialcrisisandhowitaffectedtheSwedishhousingmarket.
4ForahistoricalreviewofthepropertytaxinSwedenseeStenkula(2015)
5In2005,homeownershadtodeclarethecharacteristicsoftheirhousebyfillinginaformandsendittothetaxagency.Homeownerswerethen informedinthesummerof2006aboutthefinaltaxvalue.
6LikeintheU.S.andelsewhere,thepropertytaxisunpopularandregularlydiscussedalsoinSweden.Houseownerslobbyforacompleteremoval, economistssupportthetaxforefficiencyreasons,andpoliticianswanttherevenueitbringsbutalsothesupportofthevoters.SeeCabralandHoxby(2012) foradiscussionontheunpopularityoftheU.Spropertytax.
7DetailsabouthowthenetpresentvalueiscalculatedispresentedinSection4.2.
Fig.2. Yearlytaxpayments.Note:TheunitisSEKthousand.
Fig.3. Yearlyreformgain.Note:TheunitisSEKthousand.
is,veryweakinSweden.Arecurringsetupindevelopedcountriesisthatlocalgovernmentsorcitiessettheirownproperty taxesaccordingtothelocalpreferencesforpublicgoodssuchaseducation.InSwedenhowever,thepropertytaxrateis determinedatthenationallevel.Asaresultofthereformstudiedinthispaper,theactualrevenuesfromthepropertytaxon owneroccupiedresidentialhousingwereshiftedfromthenationalgovernmenttothemunicipalgovernments.However, duringthefirstyearofimplementationthereformwasfinanciallyneutraltoeachmunicipalitysinceintergovernmental grantsweredecreasedbytheexactamountofthepropertytaxrevenue.Thereafter,therevenuesfromthepropertytaxmade apositivenetcontributiontomunicipalfinancesaccordingtotheirjurisdictionalpropertytaxbase,althoughtheeffecton themunicipalbudgetwasverylimited.Therevenuefromthepropertytaxconstitutesonly2–3percentoftotalmunicipal revenues,whichconsistprimarilyofincometaxrevenuesandintergovernmentalgrants.8Alsonotethatmunicipalities– beforeaswellasafterthereform–cannotaffectthepropertytaxschedule,buthavetoaccepttherate(andcap)decidedat thenationallevel.
Thereformwaspartlyfinancedwithanincreaseinthetaxrateontherealizedgainsfromahousesalefrom20percent to22percent,andpartlybytheintroductionofaninterestrateondelaysonthesetaxpayments.Whilethereductioninthe propertytaxbenefitedownersofhighlyvaluedhousesdisproportionatelymorethantheownersoflowervaluedhouses, thetaxonrealizedgainswasunrelatedtothetaxvalueofthehouse.Thisdifferenceisessentialwhenweestimatetheeffect ofthecutinthepropertytax.InAppendixC.3,wediscussindetailwhythefinancingofthereformisnotlikelytoconfound ourestimatesoftheeffectofthepropertytaxmuch.
InFig.2weshowhowyearlypropertytaxpaymentsdependonthetaxvaluebeforeandafterthereform.Thedashed straightlinewithagradientof0.01isthepre-reformproportionaltaxschedule.Thepost-reformtaxschedulehasagradientof 0.0075andakinkatataxvalueofSEK800,000,whichillustratesthatthefinalpartofthereformcontainedtwocomponents:
theproportionaltaxdecreaseof0.25percentagepointsandthecapatSEK6000.WecombinethesetwoschedulesinFig.3 inordertoillustratehowthegainfromthereformisrelatedtothetaxvalue.Clearly,thetaxreductionwaslargerforhighly valuedhousesbothinrelativeandabsoluteterms,althoughallhouseownersbenefitedfromtheratecut.
8 FormoredetailsonthereformseeGovernmentProposal,2007/08:27.
2.2. Frompromisetoimplementation–whentoexpectresponses?
Asdiscussedbye.g.Blundelletal.(2011),whenevaluatingreformsitisimportanttotrackannouncementsthatareleading uptoactualimplementation.Iftherearedynamicincentivestoreacttothereformalreadyatthetimeofannouncement, individualsmaytimetheirresponseaccordingly,andanestimationoftheeffectofthereformatthetimeofimplementation couldbebiaseddownwards.Forus,theimportantquestioniswhenexpectationsabouttheimplementationofaproperty taxreform–andaboutthedetailsofthatreform–changedduetonewlyreleasedinformation.
BeforetheAlliancepartiesdecidedtomakethepropertytaxreformoneofitsmostsalientelectionpromisesduringthe electioncampaignin2006,itwasonlytheChristianDemocrats,asmallright-wingparty,whohadpropagatedforsubstantial reductionsofthepropertytax.9However,theAllianceannouncedinthesummerof2006thattheyintendedtomakeacross theboarddecreasesinthepropertytaxiftheywouldwintheelection.Asafirststep,theAlliancepromisedthattheyearly taxliabilityonthelandpartshouldnotexceedSEK5000.Whilethepropertiesthatwereaffectedbythefirstpartofthe reformwaslargelythesamepropertiesthatwereaffectedalsobythefinalpartofthereform,themaindifferencewasthat thefirstpartimpliedasmallercutthanthefinalreform.10Thisfirstpartwasintendedtobetemporaryandeventually replacedwitha“low”propertytaxcollectedbythemunicipalitiesin2008.Howevernofurtherdetailsaboutthesecondand finalpartofthereformwerementioned.TheAlliancealsostatedthatthetaxdecreasewouldbepartlyfinancedwithinthe housingsector.RepresentativesfromtherulingSocialDemocraticpartydeemedtheproposalasirresponsibleduetothe lackofaclearerfinancingplan,andalsopointedoutthatthetaxreductionwouldpredominantlybenefitownersofhighly valuedproperties.TheSocialDemocraticresponsetotheAllianceproposalwasthatthesystemshouldstaymainlyasitwas, althoughasmallreductionofthetaxratewasonthetable.
InSeptember2006theAlliancewontheelectionandformedamajoritygovernment.Thenewgovernmentimplemented thepromisedfirstpartofthereforminJanuary2007.During2007therewasapublicdiscussionaboutthedetailsofthe finalreform.Differentplacementsofthecapweresuggested.InSeptemberthefinalproposalofacapatSEK6000anda taxdecreaseto0.75percentwaspubliclyannounced.ThepropositionwasputforwardtotheparliamentinOctoberand wasacceptedaslawinDecember.Fig.1showedthetimingofannouncementsandimplementationofdifferentpartsofthe reform.
Whatcouldprospectivehousebuyersandsellersexpectatdifferentpointsonintime?Inthesummerof2006,when theAllianceannouncedthattheywouldsubstantiallyreducethepropertytax,itwasaclearbreakfromvoters’earlier expectations.Beforethiselectionpromise,amajorreductioninthepropertytaxwasonlygivenprioritypubliclybythe ChristianDemocrats.However,itwasexpectedthatthecomingelectionwouldbeacloseraceandeveniftheAlliance wouldwin,itwasstillnotcertainthatthepromisewouldbedeliveredonandwhatthedetailsofthereformwouldbe.In thepresenceofsuchhighuncertaintywewouldnotexpecttoseemuchofaresponseinhousepricesinthesummerof 2006.AftertheAlliancewontheelection,andquicklythereafterre-announcedthattheywouldreducethepropertytax, theprobabilityofasubstantialandpermanentcutgreatlyincreased.Theirintentionswerefurthersubstantiatedasthey implementedthefirstpartofthereformgoinginto2007.Althoughthedetailsofthefinalreformwereunknownuntilthe secondhalfof2007,ownersofhighlyvaluedpropertieslikelysuspectedasubstantialcutintheirpropertytaxpayments.
Accordingly,weexpectpricestohavefullyadjustedtothereformbytheendof2007.
Finally,from1January2008,whenthenewtaxschedulewasimplemented,theonlyuncertaintythatremainedwas whetherthetaxwouldbechangedagain,e.g.ifSocialDemocratswouldrevertthepropertytax,weretheytowinthe electionin2010.However,theSocialDemocratsquicklyadjustedtothenewtaxschedule,andin2009theyadmittedto onlywantingtoincreasethepropertytaxforextremelyexpensivehouses.11Houseownersandprospectivebuyers,therefore, hadgoodreasontobelievethatthepropertytaxreductionwouldnotberevertedwiththeSocialDemocratsingovernment.
Furthermore,itturnedoutthattheAlliancewasabletomaintaingovernmentpowerafterthe2010election.Theminority coalitiongovernmentconsistingoftheSocialDemocratsandtheGreenPartythatcametopowerin2014hasyettomake anysuggestionsofchangesinthepropertytax.In2014theSocialDemocraticfinanceministersaidthatthepropertytax wouldnotberaised.
3. Data
WehaveatourdisposalacombineddatasetoriginatingfromtheSwedishlandsurveyingofficeandSvenskMäklarstatistik AB(Ourtranslation:SwedishRealEstateAgentStatistics,Inc.).Thedatasetcontainsallinstanceswhereahouse–intended forpermanentorsummerliving–hasswitchedownersduringthetimeperiod2006to2008,forallofSweden.Itcontains informationontaxliability,thedatewhenthesalescontractwassigned,thefinalprice,characteristicsoftheproperty,and preciselocation.
9TheChristianDemocratsreceivedroughly9and6percentofthevotesinthe2002and2006parliamentaryelections.
10DetailsaboutthedistributionoftaxdecreasesfromthispartofthereformarepresentedinAppendixA.1.
11InaleadingSwedishnewspapertheSocialDemocratswrotethattheywouldliketocomplementthenewlyintroducedtaxcapwithaproportional propertytaxamountingto1percentonhouseswithtaxvaluesaboveSEK4.5million.Thereareonly160observationswithsuchhightaxvaluesinour dataset,whichis0.16percentofallobservations.
Fromthisextensivedatasetweselectobservationstobeincludedinarelevantpopulationbasedonanumberofcriteria.
Weleaveoutownerchangesduetoinheritance,distributionofmaritalproperties,premaritalsettlements,gifts,purchases betweenfamilymembersandthelike.Second,weonlyfocusonobservationsthatrefertoonesingulartaxvalueunit,and whichcontainsexactlyonehouseandexactlyonelandunit.Third,onlypropertiesthathadtheirlatesttaxvalueassessedin themandatorytaxvalueassessmentin2006areincludedinthestudypopulationsincetaxvaluesasaruledependonmarket pricesinthevaluearea,andtaxvalueassessmentsafter2006willthereforebeendogenoustothetaxreform.Assessments in2007and2008wereonlyconductedforrebuilthouses.Weleaveoutallsalesduringthesecondhalfof2008whenvery fewhousesweresold,especiallyattheveryendoftheyear,duetotheextraordinaryeventofthefinancialcrisis.12Finally, wemissinformationonvaluearea,valueyear,taxvalue,livingortotalareaforafewhundredsales.
Fromthispopulation,weanalyzeallsalesmediatedthrougharealestateagentsconnectedtoSvenskMäklarstatistik AB,whichconstitute52percentofallsalesintherelevantpopulation.Thereasonwefocusonthisgroupisthatweonly knowthecontractsigningdate,i.e.thedatewhenthesalespricewasagreedupon,forsalesthataremediatedwitha realestateagentthatisconnectedtoSvenskMäklarstatistikAB.Fortunately,itismainlysmallerrealestateagentsthat arenotconnectedtoSvenskMäklarstatistikAB.However,inTable3inAppendixA.3weshowthatthesampleusedinthe estimationsislargelyrepresentativeoftheentirepopulationofsales.Asexpected,propertiesforwhichsalesarenotreported toSvenskMäklarstatistikABaremoreoftenlocatedinnon-urbanareasandhencehaveslightlylowermarketvalue.The aboverestrictionsleaveuswithasamplecontaining101,449observations.
4. Empiricalstrategy
AswasshowninFigs.2and3,theyearlytaxliabilityisadeterministicfunctionofthetaxvaluebothbeforeandafter thereform;andthetaxreductionwasproportionallylargerformoreexpensivehousesbecauseoftheintroductionofatax capforhouseswithtaxvaluesoverSEK800,000.Asimplecross-sectionalregressionofmarketpriceontaxreductionwill thereforeyieldapositivecoefficientasadirectconsequenceoftheendogenousconstructionofthereform.Sincewehave dataonhousesalesbothbeforeandafterthereformwedealwiththisendogeneityissueinaDifferences-in-Differences (DiD)framework,wherethepricedevelopmentofpropertiesbelowthetaxcapserversasacounter-factualforhouseswith taxvaluesabovethecap.13WeestimatethefollowingempiricalmodelinEq.(1).
yijt=Pt+Tij+Dijt+Xijt+Vj+εijt, (1)
whereyijtisthenaturallogarithmofthemarketpriceforhousei,inmunicipalityj,andinperiodt.Weincludeperiod specificdummiesinPtandcontrolandtreatmentgroupspecificdummiesinTij.Unlessotherwiseisstated,Ptisequalto oneduring2007and2008,andzeroduring2006.Insomespecifications,wedividethehousesabovethecapintothree mutuallyexclusivetreatmentgroupsasinFig.1.However,wealsoshowresultswhereweincludeallhousesabovethecap inonesingletreatmentgroup.Thetreatmentgroupspecificdummiesareinteractedwiththeperiodspecificdummiesto generatethevariablesofinterestinDijt.Ifthereformtakesplaceintimeperiodt=k,andtherearenopriorannouncements, theDiD-estimateofthereformeffectisfoundinDijk.Inmostspecificationsweuseonetimeperioddummytoindicatepre- andpost-reform,butwealsoshowmoredynamiceffectsinSection6.3.Herewedividethepre-andpost-dummiesinto halfyearlongperiodsinordertostudymorespecificallytimedresponses.Standarderrorsareclusteredatthemunicipality level(288clusters).
IntheresulttableswerefertoModel1isabaselineDiD-modelthatonlycontainstreatment-Tijandtimeperioddummies Pt,andinteractionsbetweenthetwoDijt.InModel2weaddcontrolvariablesXijt,whicharethefollowing:thesizeofthe house(livingarea),thesizeofthelandproperty(totalarea),ataxauthoritymeasureofhousestandard(standardpoints), andthelatestyearwhenextensivechangesweremadetothehouse,normallytheyearofconstruction,(constructionyear).
InModel3wefurtheraddmunicipalityfixedeffectsVj,andinModel4weexchangethemunicipalityfixedeffectsforthe smaller(exclusivelyreligious)jurisdictionofparishes.
TheDiD-approachidentifiestheeffectofthepropertytaxcutunderthecrucialassumptionthathousepricesinthe treatmentgroupwouldevolvesimilarlytothecontrolgroupiftherewasnotreatment.Sinceourtreatmentandcontrol groupshavegenerallydifferentlevelsofpricesweargue thattheparallel trendassumptionismorelikelytoholdfor logarithmsthanforlevels.Accordingtoourempiricalspecification,wheretheoutcomevariableislogarithmic,housesin thecontrolandtreatmentgroupsshouldhavesimilarproportionalpricechangesiftherewasnoreform.14Theassumption isconsistentwiththeideathathousesarecapitalinvestmentsthatyieldthesameequilibriumrateofreturnwithinthe fixedbordersofacountry.AscouldbeenseeninFig.1,thecontrolandtreatmentgroupsfollowasimilarproportional trendduringthepre-treatmentperiodof2006,whichsupportsthevalidityoftheparalleltrendassumption.Theparallel
12 Thecollapseofthemarketintheendof2008canbeseeninFigs.10and11inAppendixA.2.
13 GiventhatthereisakinkintreatmentattaxvalueSEK800,000,itisnaturaltosuggesttheuseofRegressionKinkDesign(RKD)toidentifythecausal effectofthetaxreduction.Wehavetriedsuchastrategybutobtainedestimatesthatweretooimprecisetobeinformative.Thisisnotsurprisinggiven thatthedifferenceinthetaxreduction,andhencetheexpecteddifferenceinpriceresponse,isquitesmallaround,andcloseto,thekink.Theresultsare presentedinAppendixD.
14 Ortobemoreaccurate,ifownersofpropertieswithtaxvaluesabovethecapalsoonlyreceivedtheproportionaltaxratedecrease.
Table1
Theoreticaleffectsunderfullcapitalization.
(1) (2)
Firstdifference DiD-estimate
Control 0.074 Controlgroup
Treated 0.154 0.080
T<p(95) 0.146 0.072
Tp(95−98) 0.215 0.141
T<p(99) 0.233 0.160
Note:Column1showsthefirstdifferencebetweenthemarketpricewithandwithouttheNPVofthereformforeachgroupseparately.Column2isthe seconddifferencenetofthefirstdifferenceforthecontrolgroup.
trendassumptionisviolatedifthereareotherpolicychangesordemand/supplyshocksin2007or2008thataffecttheprice developmentofhousesinthecontrolandtreatmentgroupsdifferently.InAppendixC.1weprovideadditionalsupportfor theparalleltrendassumption.InAppendixC.2weshowthattherewerenosubstantialendogenoussupplyresponsesto thereform.AndinAppendixC.3,wethoroughlydiscussthreepotentiallyconfoundingpolicychanges:theincreasedcapital gainstaxation,introductionofanearnedincometaxcreditandtheabolitionofthewealthtax.Tosummarize,theconclusion isthateachofthesepolicychangesislikelytoonlymarginallybiasourestimates.Oneexceptionistheabolitionofthewealth tax,whichmightbiastheestimatesinthetopsegmentsupwards.However,itisdifficulttoquantifythemagnitudeofthat potentialbias.
4.1. Expectedpriceresponsesunderfullcapitalization
Ifthepropertytaxreductionisfullycapitalized–ofwhatorderofmagnitudedoweexpecttheestimatestobe?How largedifferencesinpricedevelopmentsdoweexpecttoseebetweenourcontrolandtreatmentgroups?Inthissection wecalculateexpectedpriceresponseswhenthefullvalueofthetaxreductioniscapitalized.Werefertotheseestimates as“theoreticalestimates”sincethesearethebenchmarkestimatesthatwederivefromthestandardcapitalizationmodel.
ThesetheoreticalestimatesareobtainedbyperformingsimulatedDiDestimationsinthepre-treatmentperiodsincethe marketpricesofthesepropertiesarenotaffectedbythereform.
Webeginbycalculatingthenetpresentvalue(NPV)ofthefinalreformtaxreductionforeachindividualpropertythat wassoldinthefirsthalfof2006.15WeaddtheNPVtotheactualmarketpriceofeachpropertyinordertogetamarketprice underfullcapitalization;asifthe2008reformwasimplementedalreadyforthehousessoldinthebeginningof2006.We thenperformaDiDestimationcomparingthedifferencesinpricesbetweengroupsbeforeandafterthevalueofthefully capitalizedreformhasbeenaddedtothemarketprice.
ThetheoreticalDiD-estimatesunderfullcapitalizationandfixedsupplyarepresentedinTable1.Thefirstcolumndisplays thetheoreticalreformeffectforeachgroupseparately,i.e.adifferencebetweentheactuallogpriceandthelogpricewiththe addedreformNPV.Underfullcapitalization,pricesinthecontrolgroupareexpectedtoincreaseonaveragebyapproximately 7.4percent(usingdifferencesinlogasanapproximationforpercentagechange).Thecorrespondingpriceincreaseforall treated(capped) propertiesis15.4percent.Thegainfromthereformisclearly largerinhigher valuedsegments.The priceresponseinthethreetreatedgroupsofpropertiesis14.6,21.5,and23.3percentrespectively.Column2displaysthe theoreticalDiDestimatesforthethreetreatmentgroupscomparedwiththecontrolgroupusingModel1.Thesetheoretical DiD-estimatesimplyrelativepriceincreasesamountingto8.0,7.2,14.1,and16.0percentagepointsrespectively.These theoreticalestimatesareanimportantpointofcomparisonwhenweanalyzeourempiricalresultsinSection6.Insome specifications,thesampleandtheidentifyingvariationdiffersfromwhatwehaveusedwhenwecalculatedthetheoretical effectspresentedinTable1.Whenwechangethespecificationorsample,wethereforecalculatenewtheoreticaleffectsand presentthemtogetherwiththeresults.
4.2. NPVcalculations
Thetheoreticalestimatesdependonassumptionsabouttheagents’expectationsaboutthefutureofthepropertytax, timehorizonsanddiscountrates.Ashortreflectiononeachoftheseassumptionsisneeded.First,wenotethatthenet presentvalueofastreamofincomescanbewrittenas:
NPV=
T t=0
It
(1+r+)t, (2)
15Wealsotakeintoaccountthatnewlybuilthouseswereexemptedfromthepropertytax,bothbeforeandafterthereform,forfiveyearsanda50 percentreductionforanotherfiveyears.However,thishasaverysmalleffectsinceonly4507housesinthesample(4.4percent)areexemptedfromthe tax(arebuiltafter1996).
whereItdenotetheincomeinperiodt,orinourcasethenominalannualtaxreduction,rtherealinterestrateandthe inflationrate.ThetrickypartistocalculateIt.Accordingtothedetailsinthepropertytaxbilleffectivefrom1January2008, thecap(initiallyatSEK6000)wouldbeadjustedonayearlybasisaccordingtobothinflationandtherealgrowthinwage incomes(g).Assumingthattaxvalues(andhencetaxpaymentsundertheoldpropertytaxregime)alsoincreaseattherate
+g,wegetIt+1=It×(1+g+).Withthesimplifyingassumptionthattherealinterestrateequalsthegrowthinrealwages wegetthat:
NPV=I0×T (3)
ThevalueofthetaxreductionthusdependscriticallyonthetimehorizonT.Withnofuturechangesinthetaxpolicy, thetimehorizonshouldbeverylong.Instandardnetpresentvaluecalculations,theflowofincomesiscertain.Howeverin ourcase,thepropertytaxislikelytochangeagainatsomepointintime.Bothfurtherreductionsandincreasesarepossible.
Thismeansthatarationalagentmaydiscountthegainfromthereformwithariskpremium.Weassumeariskpremium of2percentandaninfinitediscounthorizon.Thisisequivalenttoacasewithnouncertaintyandthatthepropertytaxis revertedafter50years.
OurNPVcalculationsassumethattheagentsunderstandthedetailsabouthowthetaxpaymentswouldevolve,both undertheoldandnewpolicy,andthattheyadjustforthepoliticaluncertainty.Ifweinsteadassumethatagentssimplify theproblemandexpectthefirstyear’sreductiontobeconstantinrealterms,wewouldobtainexactlythesameNPV,with arealinterestrateof2percentandwithaninfinitehorizon.
Fromthisdiscussion,itshouldbeclearthatthetheoreticalbenchmarksaresensitivetoseveralassumptions.Importantly, theywouldbesmallerwithashortertimehorizonand/orwithahigherrealinterestrate,andviceverse.Moreoverifhouse buyerswouldexpectthepropertytaxtosoonbeincreasedagain,thetheoreticalestimatescouldbetoohigh.Ontheother hand,furthercutsarealsopossible,andifhousebuyersputsomeprobabilityonthatscenario,thetheoreticalestimatesmay betoosmall.
5. Descriptiveanalysis
ThedistributionofpropertiesaccordingtotheirtaxvaluesisshowninFig.4.Thedistributionispositivelyskewed,with atail(cutclosetothe99thpercentileforaestheticreasons)ofhighlyvaluedproperties.Amajorityofpropertieshaveatax valuebelowSEK800,000(52percent),andwereaffectedbythereformonlytotheextentthatthetaxratewasdecreased from1percentto0.75percentofthetaxvalue.
InTable2weshowpropertycharacteristicsusingthevariablesinthedataset.Weshowvaluesforthefullsample ofproperties,butalsoforthecontrolgroup,thetreatmentgroupandthreesegmentedtreatmentgroups.Aspreviously mentioned,theaveragegainfromthereformisSEK4907intermsofyearlytaxpayments,andSEK243,000intermsof NPV,seedescriptivestatisticsforthefullsampleinColumn1.Forownersofhousesinthetop1percent,theNPVofthe reformwassubstantiallylarger;roughlySEK1.7million,seeColumn5.Ownersofthesehouseswentfrompayingalmost SEK40,000everyyeartoonlySEK6000inpropertytaxes.TheNPVgainforthetwoothersegmentedtreatmentgroupsis SEK994,000andSEK364,000.Housesingroupswithhighertaxvaluesarelarger,ofhigherquality(morestandardpoints), andarelocatedinmoreattractivegeographicareas;thelandshareofthetaxvalueishighereventhoughthetotalareais typicallysmallerthaninthecontrolgroup.
Fig.4. Distributionoftaxvalues.Note:TheunitisSEKthousand,andthebinwidthis100.Distributionistruncatedattaxvalue3200.
Table2
Propertycharacteristicsbytreatmentstatus.
(1) (2) (3) (4) (5) (6)
Full Control Treated T<p(95) Tp(95−98) Tp(99)
Marketprice 1778 903 2726 2429 4721 7481
Taxvalue 959 462 1497 1336 2617 3963
NPVgain 243 58 444 364 994 1658
Yearlygainfromreform 4.9 1.2 9.0 7.4 20.2 33.6
Yearlytaxpre-reform 9.6 4.6 15.0 13.4 26.0 39.6
Yearlytaxpost-reform 4.7 3.4 6.0 6.0 6.0 6.0
Effectivetaxpre-reform% 0.54 0.51 0.57 0.57 0.58 0.56
Effectivetaxpost-reform% 0.26 0.38 0.26 0.27 0.13 0.087
Livingarea,m2 115 101 131 126 159 201
Totalarea,m2 1473 1826 1091 1075 1162 1501
Standardpoints 29 27 31 30 32 34
Constructionyear 1963 1958 1969 1969 1969 1960
Landvalueshare 0.37 0.36 0.38 0.37 0.44 0.48
Observations 101,449 52,750 48,699 43,617 4067 1015
Note:Prices,values,gainsandtaxesareexpressedinSEKthousandnominalprices.Meanvalues.
6. Results 6.1. Baselineresults
WestartbypresentingtheresultsfromDiD-estimationswhendividingintotwotimeperiods.Thepre-reformperiod coversyear2006,andtheremainingtimeperiod(Jan2007–Jun2008)isthepost-reformperiodsincethefirstandfinalsteps ofthereformwereimplementedduringthisperiod.Inthissectionweonlyuseonecontrolgroupofuncappedproperties andonetreatmentgroupofcappedproperties(propertycharacteristicsofthesetwogroupsareshowninColumn2and 3inTable2).Weshowresultsfromthistwo-by-twoDiD-estmationusingallfourempiricalmodelsthatwerepreviously discussed.Aswecontrolformorepotentiallyconfoundingfactors,thevariationusedforidentificationdecreasesandthe theoreticallyexpectedpriceresponsesdecrease.Fig.5showstheestimatedeffectsaswellasthetheoreticallyexpected effectsforeachofthefourmodels(detailedregressionresultsarepresentedinTable5inAppendixB).
Thepointestimatesineachofthefourspecificationsareslightlynegativebutverysmall.Noneofthemsuggestevena onepercentagepointresponseinhousepricesinthetreatedgroupcomparedwiththecontrolgroup.Withpointestimates assmallasthesewecannotstatisticallyrejectazeroeffect.Wecan,however,rejectfullcapitalizationalongwithanyother economicallysignificantdegreeofcapitalization.Contrarytowhatwewouldexpectunderstandardcapitalizationtheory, wethereforeconcludethatthereisnosubstantialdifferenceinpricedevelopmentsinthegroupthatwastreatedwiththe propertytaxcapcomparedwiththecontrolgroup.
BecausethepointestimatesappearstobestableacrossthespecificationswewillrefertoModel1,whichisthemost parsimoniousandhasthelargestexpectedtheoreticaleffects,asourmainspecification.Thefollowingsubsectionswill thereforefocusonresultsobtainedfromvariationsofModel1.ResultsfortheotherspecificationsarepresentedinAppendix B.
Fig.5.Effectsizesandconfidenceintervals:onetreatmentgroup.Note:Rangesindicate95%confidenceintervals.Pointestimatesfromlefttoright:
−0.00535,−0.00616,−0.00119and−0.00429,seeTable5inAppendixB.Model1isabaselineDiD-model,Model2addscontrolvariables,Model3further addsmunicipalityfixedeffects,andModel4addsparish,insteadofmunicipality,fixedeffects.
Fig.6. Effectsizesandconfidenceintervals:threetreatedgroups.Note:Rangesindicate95%confidenceintervals.Pointestimatesfromlefttoright:
−0.00412,−0.00505,0.00985,seeModel1(middlepanel)inTable5inAppendixB.
6.2. Evidencefromdifferentsegmentsofthemarket
Weshowedearlierthatthegainfromthereformintermsofdecreasedtaxliabilityincreaseswiththetaxvalueofthe property.Wewouldthereforeexpectpricesofhouseswithhighertaxvaluestoincreasemorethanhouseswithlower taxvalueswithinthetreatmentgroup.Wethereforedividethetreatmentgroupintothreeequallysizedgroupsintermsof numberofobservations.InthefirstgroupT1taxvaluesrangefromSEK800,000to1,130,000.InthesecondgroupT2therange isSEK1,130,000to1,595,000andthethirdgroupT3containshouseswithtaxvaluesaboveSEK1,595,000.Toestimateprice responsesseparatelyforthethreegroups,weadjustModel1,bysplittingthetreatmentindicatorintoseparatedummiesfor eachofthethreetreatmentgroups.TheresultsarepresentedinFig.6(DetailedregressionresultsarepresentedinTable5 inAppendixB).
Contrarytoourexpectations,wefindnoevidenceofcapitalizationofthetaxcutforanyofthethreegroups.Thisis quitenoteworthy,aswewouldexpectpricestoincreasebyabout13percentagepoints,relativetothecontrolgroup,inthe treatmentgroupcontainingthehouseswiththehighesttaxvalues.Inallthreetreatmentgroups,thepointestimatesare quitepreciselyestimatedandclosetozero.Itisonlyinthegroupcontainingthehouseswiththehighesttaxvaluesthat wecanseeapositive,howeververysmall,pointestimate.Toinvestigateifthetaxcutiscapitalizedinhousepricesonly amongtheverymostexpensiveproperties,wesplittheoriginaltreatmentgroupintothreenewtreatmentgroupsinthe followingway.Weextractthetop1percentexpensiveproperties(accordingtotaxvalue)asonegroupTp(99).Thesecond groupTp(95−98)consistsofthetop5percent,excludingthetop1percent.ThefinalgroupT<p(95)consistsoftheremaining treatedproperties,i.e.treatedpropertieswithtaxvaluesbelowthe95thpercentile.The95thpercentileisSEK2,271,000 andthe99thpercentileisSEK3,280,000.Pricedevelopmentsinthesethreegroupsandthecontrolgroupwerepreviously showninFig.1,thetheoreticaleffectswereshowninTable1descriptivestatisticsinTable2.
Again,inFig.7weseenoevidenceofcapitalizationuptothe95thpercentileofthemarket.However,inthemost expensivegroupsofhouseswedofindevidenceofpartialcapitalization.Amongthehouseswithtaxvaluesinthe95thto the98thpercentileweestimatepriceincreasesofabout4percentagepoints.Whilethiseffectisbothstatisticallysignificant
Fig.7. Effectsizesandconfidenceintervals:topexpensivegroups.Note:Rangesindicate95%confidenceintervals.Pointestimatesfromlefttoright:
−0.00488,0.0412,0.0842,seeModel1(bottompanel)inTable5inAppendixB.
Fig.8. Effectsizesandconfidenceintervals:topexpensivegroups.Note:Rangesindicate95%confidenceintervals.Pointestimatesfromlefttoright:
0.00573,0.0181,0.0601;0.0122,0.0639,0.117;−0.0019,0.0583,0.132;−0.0176,0.0255,0.0952,seeModel1inTable6inAppendixB.
andeconomicallyrelevant,itisstillaquitesmalleffectcomparedwiththetheoreticallyexpectedincreaseof14percentage points.Forthetoponepercent,weestimateastatisticallysignificantpriceincreaseof8percentagepoints.Aresponseofthis magnitudeisclearlyeconomicallyimportant,butstillonlyhalfofthetheoreticallyexpected16percentagepoints.However, asthissegmentofthemarketonlycontains1percentofthehousesthegroupisnaturallyquitesmall,andhencethestandard errorsquitelarge.WhilewecanrejectfullcapitalizationgivencurrentassumptionsunderlyingtheNPVcalculations,minor changesinintheassumptionscouldchangethisconclusion.Forinstance,theobtainedestimateforthemostexpensive groupisconsistentwitha25yeartimehorizonasopposedtoourdefaultof50years.
6.3. Dynamicresponses
Wenowturntotheresultsfroma morecarefulinvestigationofwhenpriceresponsesmayhaveoccurred. Ifprices increasedalreadybythetimeoftheelectionpromise(secondhalfof2006),thentheestimatedpriceresponsesinFigs.5–7 couldbedownwardbiased.Theestimatescouldalsobedownwardbiasedifpricesincreasedonlyaftertheimplementation ofthefinalpartofthereform(firsthalfof2008).
AvisualinspectionofFig.1revealsthatthegroupsevolvesimilarlyupuntilJanuary2007,whenthepricesofthemost highlyvaluedhousessuddenlyincreasesmorethanthepricesintherestofthemarket.Thissuggestthatpricesdidnot respondatthetimeoftheelectionpromiseandforthetopsegmentthepriceincreaseshowsupalreadyinJanuary2007 andappearstostayatahigherlevelthroughoutthestudyperiod.
Inordertoformallyestimateseparateresponsesindifferenttimeperiods,wesplitthestudyperiodoftwoandahalf yearsintofivetimeperiods,eachofsixmonths:
• Controlperiod(Jan2006–Jun2006)
• Electionpromise(Jul2006–Dec2006)
• Firstpart(Jan2007–Jun2007)
• Finalannouncement(July2007–Dec2007)
• Finalreform(Jan2008–Jun2008)
Thissplitallowsustostudypriceresponsesinallimportantstagesofthetotalreform:fromthefirstelectionpromise inthesummerof2006,totheimplementationofthefinalpartofthereformin2008.Ifhousebuyersanticipatedthefinal detailsofthereforminthesummerof2006,weexpecttoseeareactioninthehousepriceseriesofhighlyvaluedhouses alreadybeforetheimplementationofthefirstpartofthereform.Iftheimplementationofthefirstpartcameasasurprise, orifindividualsestimatedthattheriskwashighthattheAlliancewouldrenegeontheirelectionpromise,weexpecttosee responsesonlyinthelaterperiods.
Theresultsoftheformaltests,showninFig.8,containnosurprisesandconfirmstheconclusionsfromthevisualinspection ofFig.1.Wefindnoevidenceofcapitalizationinanysegmentofthemarketduringtheelectionpromiseperiod.Moreover, wefindevidenceofpriceincreasesinallpost-periodsforhouseswithtaxvaluesinthe95thtothe98thandinthe99th percentile.Theestimatedpriceresponseinthosetwogroupsare,however,smallerinthelastperiod.Thisisentirelydueto afallinpricesinthelastmonth,whichisvisiblealsoinFig.1.Thiscouldbeasignthatthefallinhousepricesduringthe secondhalfof2008startedslightlyearlierinthetopsegmentofthemarket.Fortheothergroupswefindnoevidenceof capitalizationinanyperiod.DetailedregressionresultsareprovidedinTable6inAppendixB.Wethereforeconcludethat theanalysisofthedynamicsoftheresponseslendnosupportfortheworriesthattheresultsfromthepreviousanalyzes wouldcontainimportantbiases.
7. Couldtheelasticityofsupplyexplainourfindings?
Whencalculatingexpectedpriceresponsesfromthereform(theoreticaleffects)weassumeafixedsupplyofhousing.The fixedsupplyassumptionisasimplification,anditgivesusaninformativeupperboundonwhattoexpectintermsofprice responses.Ourestimatedpriceresponses,however,seemtoimplytheotherextremeintermsofsupply,i.e.thathousingin Swedenisperfectlyelastic.Ahorizontalsupplycurveistheonlyshapethatiscompatiblewiththezeroresultweobserve, giventhatactorsonthehousingmarketarerational.SoisitreallythecasethathousinginSwedenisperfectlyelastic?At facevalue,landisabundantinSweden.Butlandavailabilityaside,therearecrucialaspectsoftheSwedishhousingmarket thatputintodoubtthathousingsupplyisperfectlyelastic.
Thequotientbetweenthemarketvalueofnewlybuilthousesandthecorrespondingconstructioncosts(Tobin’sQ)isa commonlyusedmeasuretoanalyzeifanincreaseindemandwouldleadtoincreasedinvestmentsinhousing(Brainardand Tobin,1968).Ifthemarketvalueislowerthantheconstructioncosts(Q<1),itissimplynotprofitabletobuildnewhouses.
CalculationsbySwedisheconomistsattheInstituteforHousingandUrbanResearchatUppsalaUniversityshowthatfor mostpartsofSwedenthequotientislowerthanone(inyear2006).ThismeansthatforlargepartsofSwedensupplyis essentiallyfixedduetohighconstructioncostsrelativetohousevalues.Inoursample,wehave77percentofhousesinthe controlgroup,and15percentofthetreatmentgroup,aresoldinmunicipalitieswherenewconstructionisnotprofitable.
Inmostareaswherehousesinthetreatmentgrouparelocated,financialincentivesdoesnotseemtohindersupply responses.Butevenifconstructionisprofitable,supplyisrestrictedbyplanningandregulation,theresponsibilityofwhich fallsonthemunicipalities.Considerthefactthatexistingmunicipalresidentsfacenegativeexternaleffectsofnewconstruc- tion;intermsofhousepricedecreasesandcongestion.Municipalitiesthatarepoliticallyaccountabletoexistingresidents thereforehaveincentivestokeepconstructiontoaminimum.Moreover,mostoftheadditionalincometaxesthatcouldbe gainedbyexpandingthehousingstockandthepopulationisredistributedtoothermunicipalitiesinthesocalled“municipal equalizationsystem”(HüfnerandLundsgaard,2007).
Forlegislative,politicalandeconomic reasons,supplyofhousingismoreorlessinelasticacrossSweden.16Around medium-sizedorsmallercities,wherehousinginvestmentsareprofitableandwherelandismoreavailable,housingsupply islikelymoreelasticthaninhighlyvaluedareasaroundlargercities,wherelandismorephysicallyscarce,andresidentsin particularlyaffluentareaspayhighpricesforlivingingreenareasclosetothecitycenter.Thisexclusivelifestyleismaintained bymunicipalplanningandlegislationthatprotectscertainareaswithaestheticandculturalvalues.Differencesinsupply elasticitycouldthereforeexplainahigherdegreeofcapitalizationinhighlyvaluedareas.Butadditionalexplanationsare neededtoexplainthezeroeffectformosttreatedproperties,sinceperfectelasticityofsupplyishighlyunlikelyonthe Swedishhousingmarketforthereasonsdiscussedabove.
8. Concludingremarks
FormostofthepropertiesontheSwedishmarket,wedocumentazeroeffectonpricesfromasubstantialdecreaseinthe propertytax.Althoughthezeroresultsareconsistentwithcapitalizationtheory–ifthesupplyofhousingisperfectlyelastic –wecanconcludethatperfectelasticityisveryunlikelyontheSwedishhousingmarket,primarilybecauseofrestrictive legislationandthelackofmunicipalincentivestoincreaseconstruction.
Whilethereareseveralmechanismsthatcouldexplainwhycapitalizationratesarelowerthanourtheoreticalbenchmark, afullysatisfactoryexplanationshouldbeabletoexplainbothazeroeffectformostofthemarket,aswellasthepartial capitalizationresultwefindinthemostexclusivesegmentofthehousingmarket.Asinglesuchexplanationisadmittedly hardtofind.Forinstance,thedemandresponsetoanationaltaxcutmightbeweakerthantoalocaltaxcut,butarguably stillexistent.Similarly,whileitispossiblethatsomehousebuyersexpectedthetaxtobeincreasedagaininthenearfuture, wedonotthinkitislikelythatmanyexpectedanimmediatepolicyreversal,whichisrequiredtoexplainthezeroeffects.
Imperfectionsonthefinancialmarkets,suchascrudebanklendingpoliciesthatimposeliquidityconstraintsonhousebuyers, areotherpotentialexplanationforlowercapitalization.Whilesupplyanddemandelasticities,expectedpolicyreversal,and liquidityconstraintscouldexplainlowercapitalizationrates,eachexplanationcannotaloneexplainthezeroresultorthe differentialeffectsfordifferentsegmentsofthemarket.
Whatcharacterizesthemostexclusivesegmentofthemarket,forwhichwefindpositivecapitalization,isthat:(1)the housesaresituatedinplaceswherelandisphysicallyscarce,(2)theownersreceivedalargetaxreduction,and(3)the buyersofthesehousesarelikelytobemorefinanciallyliterate.Inotherwords,restrictionsonnewconstructioninexclusive areas,thesaliencyofthereform,andtheabilitytocalculatethenet-present-valuefromadecreaseinthepropertytaxcould explainthedifferentialeffects.Sincethesefactorsarehighlycorrelated,wecannotempiricallyseparatethemandhencenot concludewhichismostimportant.Moreover,thefactorsmightinteract,orpartiallycontributetotheresults.
Finally,itisworthcommentingonthetimeframeofourinvestigation.Whileanyinformationaboutachangeinthe propertytaxisexpectedbytheorytobeimmediatelycapitalizedintohousepricesonanefficienthousingmarket,manyof thepotentialexplanationsforlowercapitalizationdiscussedabovemaysimplycausepriceadjustmentstobeslower,rather thannottomaterializeatall.Therefore,wecannotruleoutthepossibilitythatthepropertytaxwaseventuallycapitalized,
16 HilberandVermeulen(2016)findthatcapitalizationvarybecauseoflocaldifferencesinregulation,landscarcityandgeography.
althoughextendingtheanalysisbeyondthefirsthalfof2008isveryproblematicsincethefinancialcrisissubstantially affectedthenumberofpropertiesonthemarketandthesalespricesoftheseproperties.
AppendixA. Thefirstpartofthereform
A.1Thefirstpartofthereform
Thetemporarylandtaxreform wasintroducedin2007asasteppingstonetowardsthemorecomprehensivefinal reform.Inthissectionweshowthatthepropertiesthatweregenerouslyaffectedbythefirstpartofthereformwerethe samepropertiesthatweregenerouslyaffectedbythefinalreform.
ThefirstpartofthereformimpliedthatthetaxonthelandpartwasreducedtoSEK5000orSEK2perm2depending onwhichoneyieldedthelowesttaxpayment.Thetaxstayedthesameifthenewrulesdidnotyieldareductioninthe tax.Ofallthepropertiesinoursample,56percentweretreatedwithataxreductionduetothetemporarylandtaxreform.
Beforethefirstpartwasimplemented,thetreatedpropertiespaidayearlytaxofSEK13,322onaverage,andafterthefirst parttheyearlytaxwasdecreasedtoSEK9620;theaverageyearlygainfromthefirstpartofthereformwasthusSEK3702.
ThesamepropertieswentontohaveanadditionaldecreaseinyearlytaxliabilitiesofSEK3950whenthefinalreformwas implemented.
InFig.9weshowhowtheyearlygainofthefirstpartofthereformcorrelateswiththepropertytaxvalueandcompare itwiththegainfromthefinalreform.ItisclearfromFig.9thatthegainoffirstpartwasdirectedtothesamekindof (expensive)propertiesasthefinalreform,onlytoasmallerextent.
A.2The2008financialcrisisandthehousingmarket
Inthissectionweillustratehowthefinancialcrisisof2008affectedthehousingmarket.InFig.10weshowtheevolution oftheaggregatepriceindexinoursampleoverthreeyears,includingthelastpartof2008whenthefinancialcrisishit.In theautumnandwinterof2008pricesreachedtheirlowestlevelduringthetimeperiodthatwestudy.InFig.11weshow
Fig.9. Yearlygain:comparisonbetweentwopartsofthereform.Note:TheunitisSEKthousand.
Fig.10.Marketprice.