Darmstadt Discussion Papers in ECONOMICS

Ease vs. Noise: Long-run changes in the value of transport (dis)amenities

Gabriel M. Ahlfeldt, Volker Nitsch, Nicolai Wendland

Nr. 236

Arbeitspapiere der Volkswirtschaftlichen Fachgebiete der

Technischen Universität Darmstadt

Darmstadt Discussion Papers in ECONOMICS

This text may be downloaded for personal research purposes only. Any additional reproduction for other purposes, whether in hard copy or electronically, requires the consent of the author(s), editor(s). If cited or quoted, reference should be made to the full name of the author(s), editor(s), the title, the working paper or other series, the year, and the publisher. ISSN 1438-2733 © Gabriel M. Ahlfeldt, Volker Nitsch, Nicolai Wendland 2019 Printed in Germany Technische Universität Darmstadt Department of Law and Economics D – 64289 Darmstadt Germany www.wi.tu-darmstadt.de

GabrielM.Ahlfeldt,VolkerNitsch,NicolaiWendland

Easevs.noise:Long‐runchangesinthevalueoftransport(dis)amenities

Abstract:Foracompletecost‐benefitanalysisofdurableinfrastructures,itisimportanttounderstandhowthe

valueofnon‐marketgoodssuchastransittimeandenvironmentalqualitychangesasincomesriseinthelong‐run.

Weusedifference‐in‐differencesandspatialdifferencingtoestimatethelandpricecapitalizationeffectsofmetro

railinBerlin,Germanytodayandacenturyago.Overthisperiod,thenegativeimplicithedonicpriceofrailnoise

tripled.Ourresultsimplyincomeelasticitiesofthevalueofnoisereductionandtransportaccessof2.2and1.4,

substantiallyexceedingcross‐sectionalcontingentvaluationestimates.

Keywords:Accessibility,spatialdifferencing,noise,difference‐in‐differences,incomeelasticity,landprice

Version:September,2019

JEL:R12,R14,R41,N73,N74

Conflictofinterestandfinancialdisclosurestatement

Theauthorsdeclarethattheyhavenoconflictofinterestandnomaterialfinancialintereststhatrelatetothere‐searchdescribedintheabovepaper.Allsourcesoffundingaredisclosedbelow.

Forthcomingin:JournalofEnvironmentalEconomicsandManagement

LondonSchoolofEconomicsandPoliticalSciences(LSE)&CentreforEconomicPolicyResearch(CEPR)[email protected],www.ahlfeldt.com

TechnischeUniversitä[email protected]‐darmstadt.de

[email protected]

WethanktheeditorRogervonHaefen,twoanonymousreviewers,seminarandconferenceparticipantsinBerlin(Humboldt),Bristol(RES),Barcelona(UEA),Copenhagen(UEA),Heidelberg,Miami(UEA),München(Ifo),Orléans,Toulouse(SEA),Vaduz(VfS),Venice(CESIfo),Würzburg,andZurich(KOF)andespeciallyThiloAlbers,MaximilianvonEhrlich,GustafEngsström,StephanHeblich,HansKoster,MirenLafourcade,KristofferMöller,IsmirMulalic,JosvanOmmeren,HenryOverman,MichaelPflüger,RosaSanchis‐Guarner,SimoneSchüller,SevrinWaightsandYanosZylberbergforhelpfulcommentsandsuggestions.KristofferMöllerandSevrinWaightsprovidedoutstandinghelpwiththecompilationofthedataset.SaschaMöbiusandNeeleReimann‐Phillipalsoprovidedexcellentresearchassistance.WethanktheBerlinCommitteeofValuationExpertsandtheSenateDepartmentforUrbanDevelopmentandtheEnvironmentforprovidingpropertytransactiondata.TheGermanScienceFoundation(DFGprojectcodeNI938/2‐1)andtheFritz‐Thyssen‐Foundation(projectcodeAZ.10.10.2.070)areacknowledgedforfinancialsupport.Theusualdis‐claimersapply.

Easevs.noise 2

1 Introduction

Understandinghowthevaluesoflocationalamenitiesanddisamenitieschangeasincomesriseis

crucialforoptimaldecisionsregardinginvestmentswithlong‐termconsequences.Atypicalexam‐

pleareinvestmentsintransportinfrastructure,whichareoftenundertakenpubliclyfollowingcost‐

benefitanalyses(CBA).Theevidencefromcross‐sectionalsurvey‐basedcontingentvaluationre‐

searchsuggeststhattheincomeelasticityofthevalueofnoisereductionispositive,butlessthan

unity(Wardmanetal.2005).Thevalueoftraveltimeistypicallysettoafractionofthewagerate

(Anderson2014;Parry&Small2009),whichimpliesaunityincomeelasticity,butalowerelasticity

hasbeenrecentlysuggested(Börjessonetal.2012).Itisnotclear,however,whethertheseesti‐

matedshort‐runelasticitiesgeneralizetolong‐runcomparisons.Intuitively,theinter‐temporalin‐

comeelasticityshouldbelargerthanunityiflocationalamenitiesanddisamenitiesarenon‐neces‐

sitiesastypicallyconjecturedintheliterature(Brueckneretal.1999;Glaeseretal.2001).Asreal

incomesrise,(dis)amenityvaluesshouldthenrisemorethanproportionately,implyingthatinap‐

praisalsofdurable infrastructurescostsandbenefitsneedtobe inflatedratherthandeflatedto

reflectdemandbyfuturegenerations.Todate,thereislittleevidencetosubstantiatethisintuition.

Thereisatbestindirectevidenceinthatpublicspendingtendstoincreasemorethanproportion‐

atelyinGDP,suggestingthatpublicservices,broadlydefined,areluxurygoods(Wagner’slaw,see

Lamartina&Zaghini2011;Ram1987;Wagner1890).

Inthispaper,wetakeasteptowardsfillingthisgapbyprovidingthefirstlong‐runcomparisonof

transportamenityanddisamenitycapitalizationeffects in landpricesoveraperiodas longasa

century. Theoretically, besides the amenity of offering improved access, there are a range of

transport‐relateddisamenities, including congestion, pollution, and noise,which can affect out‐

comessuchasproductivity,health,andannoyancelevels(Navrud,2002).Ourfocusonaccessibility

andnoiseeffects isdrivenbytheempiricalsettingweexploit.Wechoosetoevaluate landprice

capitalizationeffectsofmetrorail(U‐Bahn)inBerlin,Germany,duetotheavailabilityofhistorical

andcontemporarypropertydataandatransporttechnologythathasremainedapproximatelycon‐

stantsincethesystem’sinaugurationin1902.Thesystemisfullyelectrifiedandhasexclusiveright‐

of‐way,sothattheeffectsonpollutionandroadcongestionsarerathernegligible.Wefindlittle

evidenceforanegativevieweffect,sothatnoisefromtheelevatedpartsofthesystemisarguably

theprimarydisamenity.Ourpropertydatacoverscommercialandresidentialproperty;therefore,

ourestimatedcapitalizationeffectsreflectproductivityand(dis)utilityeffects.Theylikelyexclude

healtheffectsgiventhatthepublicawarenessofnoise‐inducedhealthimpactsislimited(Navrud,

2002).Inlinewiththeworldwidetrend,realincomeinGermanyhasincreasedatarateof2%per

Easevs.noise 3

yearsince1900,accumulatingtoanoverallincreaseofabout650%.1Oursetting,thus,allowsusto

comparethevaluationofrailaccessandrailnoiseonrealestatemarketsinahistoricallow‐income

scenarioandacontemporaryhigh‐incomescenario.

Ourcontributionisfacilitatedbyaratheruniquecombinationofsuitablemicro‐geographicdataat

theturnsofthe19th(1881‐1914)andthe20thcenturies(1990‐2012).Forouranalyses,wedigitize

aseriesofhistoricalmaps,compiledbythecharteredsurveyorGustavMüller,whichprovideinfor‐

mationonlandpricesasdetailedastothelevelofindividualparcels.2Wecomplementthesehis‐

toricaldatawithaconfidentialcontemporarymicrodatasetcoveringacompleterecordofproperty

transactions.Withthesedataathand,weestimatethatoverthecourseofthe20thcentury,theland

pricecapitalizationeffectofa10‐decibeldecreaseinrailnoiseincreasedfrom4.2%to13.0%.Ac‐

countingfortheincreaseintheshareoflandinthevalueofhousingoverthesameperiod,weinfer

acapitalizationeffectinhouse‐pricetermsthatincreasedfrom1%to4%.Thelandpricecapitali‐

zationeffectofaone‐kilometerreductionindistancefromthenearestmetrorailstation,ameasure

thatcapturesthevalueoftheassociatedwalkingtime(Gibbons&Machin2005),decreasedfrom

20.2%to15.5%.However,becausethelandshareincreasedsubstantiallyoverthesameperiod,this

decreaseimpliesasizableincrease,from3.6%to5.0%,intermsofhouse‐pricecapitalization.

Theseresultssuggestthatthevalueattachedtorailaccessandevenmoresotothedisamenityfrom

railnoisehasincreasedovertime.Oneinterpretationisthataccessandaquietenvironmentare

luxurygoodsonwhichrecentgenerationsarewilling tospendmoreas theyarericher.Making

admittedlystrongassumptions,weuseourestimatedcapitalizationeffects toderivenovelesti‐

matesofthelong‐runincomeelasticitiesoftheamenityvalueofaccessibilityandthedisamenity

valueofnoiseof1.4and2.2,respectively.Whileweacknowledgethatsignificantuncertaintysur‐

roundstheseestimates,onbalance,theylikelyrepresentlowerbounds.

Ontopofthesemaininsights,wecontributetotheliteratureinseveralmorespecificrespects.First,

wecontributetoavastliteratureinthetraditionofOates(1969)thathasinferredthevalueofnon‐

marketed goods from house price capitalization, including clean air (Chay&Greenstone 2005;

Hanna 2007), health risk (Currie et al. 2015; Davis 2004), proximity to hazardouswaste sites

(Greenstone&Gallagher2008)ornuclearpowerplants(Tanaka&Zabel2018),crimerisk(Linden

&Rockoff2008),publicschoolquality(Cellinietal.2010),energyefficiency(Wallsetal.2017),

1 OwncalculationsusingdatafromtheMaddisonProject(Bolt&vanZanden2014).The2%annualgrowthgeneralizestothemeanacrossasampleof170countries.Seeappendixsection3.1fordetails.

2 Toourknowledge,theonlycomparablehistoricdataarefromOlcott'slandvaluesbluebookofChicagoandsuburbs,publishedregularlybyG.C.Olcott's&Co.,Inc.fromthe1910stothe1990s.TheconstructionofthecoreofChicago’smetrorailsystem(theL),however,precedesthisperiod.

Easevs.noise 4

aircraftnoise(Boes&Nüesch2011;Ahlfeldt&Maennig2015),roadnoise(Graevenitz,2018),wind

farms(Gibbons2015)ortransportaccess(Gibbons&Machin2005).Weaddtothisliteratureby

showingthatwithinthesamespatialcontext,capitalizationeffectsofthesame(dis)amenitiescan

varysizablyinthelong‐runduetochangesinconsumerpreferences.

Second,weenrichaliteratureonrailaccesscapitalizationeffectsthathasrecentlyshiftedfromthe

useofcross‐sectionalvariationtotheuseofvariationovertimetoimproveidentification(seeDubé

etal.2013andappendixsection2forareview).Weexpandonthislineofresearchbyproposinga

novelweighteddifference‐in‐differences(DD)estimator,whichminimizestheconditionalcorrela‐

tionbetweenpre‐announcementtrendsintheoutcomevariable(landprices)andmultiplecontin‐

uoustreatmentvariables(proximitytothestationandrailnoise).Consequently,weminimizethe

riskthatunobservedtrendsinpropertypricescorrelatedwithstationaccessorrailnoiseconfound

ourestimates.

Third,wealsoaddtoaliteratureonnoisecapitalizationeffectsthat,withfewexceptionsconcerning

theanalysisofaircraftnoise(Ahlfeldt&Maennig2015;Boes&Nüesch2011),hasemployedcross‐

sectionaldesigns.Theliteratureonrailnoiseeffectsisparticularlyunderdeveloped(seeNavrud

2002andappendixsection2forareview).Ourspatiallyhighlydisaggregated,micro‐geographic

datasetsallowustoexploittherelativelysharpchangeinrailnoisethatariseswhereatrackenters

atunneltovanishbeneaththesurface,asourceofvariationthathasnotbeenpreviouslyexploited

intheliterature.Thespatialdifferencing(SD)approachusedtoassessthecausaleffectofnoiseon

thepriceofadjacentlandparcelsinourcontemporaryanalysesrepresentanimprovementinterms

ofidentificationcomparedtotheextantliterature.Ournovelestimateoftheeffectofaone‐decibel

increaseinrailnoiseonhousepricesof‐0.4%isclosetorecentestimatespointingtoanaircraft

noiseeffectof‐0.5%to‐0.6%(Ahlfeldt&Maennig2015;Boes&Nüesch2011)andaroadnoise

effectof‐0.1%to‐1.4%(Graevenitz,2018;J.P.Nelson,2008reportsacentralestimateof‐0.57%).

Fourth,weexplicitlydisentanglethepositiveeffectsofrailaccessfromthenegativeeffectsofrail

noiseinacausalanalysisofrailcapitalizationeffects.Therefore,wegobeyondmostoftheexisting

workthattypicallyfocusesontheaggregate(ornet)effectofcountervailingrailexternalities.In

doingso,wealsoexaminethedegreeofbiasthatariseswhenaccessibilityeffectsareestimated

withoutcontrollingfornoiseeffectsandviceversa.

Fifth,weprovideoneofthefewanalysesofrailcapitalizationeffectsintolandprices(e.g.Ahlfeldt,

Moeller,etal.2015;Coffman&Gregson1998),whereasmostpreviousstudieshavelookedatprice

responsesofpropertiesorhousingunits.Theanalysisoflandpricescomeswiththeadvantageof

nothavingtocontrolforstructuralcharacteristics.Inaddition,becauselandisscarceinanurban

contextandprovided(almost)inelastically,adjustmentsinlandpricescanbeassumedtobepurely

Easevs.noise 5

drivenbydemand.Theanalysisofhousepriceeffects,incontrast,maybemitigatedbysupplyre‐

sponsesifthedemandcurveislocallydownwardslopingbecauseofimperfectmobilityandidio‐

syncraticlocationpreferences(Hilber&Vermeulen2015).

Lastbutnotleast,weprovideacasestudywhichillustratesthat,duetotheincreaseinnoiseaver‐

sion,thecasefortheconstructionofundergroundmetrorailasopposedtoelevatedmetrorailis

muchstrongertodaythaninthepast.Indoingso,wealsoprovidenovelauxiliaryfindingsthatare

interestingintheirownright.Weestimatetheper‐kilometercostofanundergroundmetrolineat

thebeginningofthe20thcenturytobethreetimesthatofanelevatedline,whichissubstantially

largerthanthecontemporaryrule‐of‐thumbfactoroftwo.Wealsofindthat,overaperiodofabout

130years,theaverageannualnominallandpricegrowthratewasabout5%inBerlinand,there‐

fore,typicallywithintherangeoftheopportunitycostofcapital(centralbankinterestrates).

Theremainderofthepaperisorganizedasfollows.InSection2,wediscussthecontextofourstudy,

presentourdata,andintroduceasimpletheoreticalframeworkthatwillguidetheinterpretation

oftheparametersweestimate.Section3presentsthehistoricalanalysis,followedbythecontem‐

poraryanalysisinSection4.InSection5werelatethehistoricalandcontemporaryestimatesto

eachotheranddiscusspolicyimplications.Finally,Section6providesourconclusions.

2 Empiricalandtheoreticalcontext

2.1 MetrorailinBerlin

In1879,theGermanfounderandinventorWernervonSiemenspresentedthefirstfullyelectrified

experimental railway at the internationally renowned trade and industrial exhibition (Gewer‐

beausstellung)inBerlin.By1891,thecompanySiemens&Halskehadproposedadensenetworkof

variouslinestoconnecttheinnercoreof“oldBerlin”withitsthensurroundingmunicipalities.Ac‐

cordingtoinitialplans,thenetworkwastobebuiltentirelyonelevatedtracks,mainlybecauseof

strictregulationofundergroundactivitiesduetoconstructionworksonthenewcanalizationsys‐

temledbyJamesHobrecht.In1895,aconcessionwasgrantedforthefirstline,whichwastocon‐

necttheeasternpartsofBerlin,atthestationWarschauerBrücke,andthewealthywesterncityof

Charlottenburg,at thestationZoologischerGarten, runningexclusivelyonelevated tracks.Built

alongoneofBerlin’smajorboulevardsthisroutingdidnotrequiremajoracquisitionsoflandor

fundamentalchangestothebuildingstructure.In1897(onlyfiveyearsbeforetheinaugurationof

theline),Siemens&HalskefoundedtheElevatedRailwayCompany(Hochbahngesellschaft)inco‐

operationwiththeDeutscheBanktoguaranteethefunding.

Easevs.noise 6

Theconstructionbegan immediately,starting fromtheeasternparts.However,Berlinresidents

increasinglyexpressedconcernsaboutaviaduct’spotentiallyunpleasantappearance.Also,Berlin’s

municipalplanningandbuildingcontroloffice,withitsnewlyappointedheadFriedrichKrause,was

nolongergenerallyopposedtoplansfortheconstructionofundergroundlines.Asaresult,thecity

ofCharlottenburgmanagedtoensure,inalast‐minutemove,thatthetracksranbeneaththestreet

surfaceoncethelinereacheditscityboundaries.Eventually,thelinewasinauguratedin1902and

called“LineA”(LinieAorStammstrecke).Thefinalroutingnegotiatedbetweenvariousstakehold‐

erssuchasDeutscheBankandthecityofCharlottenburgwaslaterdescribedbyhistoriansasan

outcomeofagreementsandaccidents(Bousset1935).Theelevatedsectionofthelineconsistsof

11stations,whiletheentireline(includingtheundergroundsection)consistsof14stationswitha

totallengthofabout10km.

AsevidentfromFigure1,LineAcomplementedacommuterrailnetworkconsistingofvarioussub‐

urban lines aswell as a circular line (Ringbahn) and an east‐west connection through theCBD

(Stadtbahn).Thisnetworkwasoperatedentirelyonground‐level tracksorelevatedtracks. It is

comparabletotoday’scommuterrail(S‐Bahn)network,butthetechnologywasdifferentastrains

werepoweredbysteamandelectrificationdidnotstartbefore1924.Overtime,thesubway(U‐

Bahn)networkwascontinuouslyexpanded.Sincethere‐unificationofthecity,thecombinedsub‐

wayandcommuterrailnetworkscomprise475railkmand275stations.

Fig.1. HistoricalandcontemporarygeographyofBerlin’smetrorailnetwork

Notes: Own illustration using the Urban Environmental Information System of the Berlin Senate Department

(SenatsverwaltungfürStadtentwicklungBerlin2006).CBDisthecentralbusinessdistrict.Kurfürstendammisamajorsub‐centre.

Easevs.noise 7

2.2 HistoricalLandPricesandContemporaryPropertyPrices

Ourmainvariableofinterestarelandpriceswhichareextractedfromvariouseditions(1881,1890,

1896,1900,1904,1910,and1914)ofassessedlandvaluemapsforBerlincreatedbytherenowned

technicianGustavMüllerincooperationwithofficialplanningauthorities.Müller’smapsprovide

dataataremarkablydisaggregatedlevelofindividualplots.Thestatedobjectivewastoprovide

officialandrepresentativeguidesforbothprivateandpublicinvestorsparticipatinginBerlin’sreal

estatemarket.WhileMüllerhimselfdidnotdescribeindetailtheexactprocedureoflandvaluation,

theimperialvaluationlaw(Reichsbewertungsgesetz)oftheGermanReichcontainedastrictorder

tousecapitalvaluesfortheassessmentofcommercialplotsbasedonfairmarketprices.Inlinewith

thevaluationlawsforcommercialland,Müllerclaimsthathisassessmentreferstothepurevalue

ofland,whichisadjustedforallbuildingandevengardencharacteristics.Healsocorrectedvalues

forspecific locationcharacteristicssuchassingleanddoublecorner lots, subsoilandcourtyard

properties.

Müller’smaps are by now an established data source. They have been used, among others, by

Ahlfeldt,Moeller,etal.(2015),whoalsoprovideanextensivedataappendixthatdescribesindetail

thenatureofthedata.Morenotably,thedataaredirectlycomparabletothemorerecentBerlin

landpricedata(1928,1936,1986,2006)usedbyAhlfeldt,Redding,etal.(2015);theyalsoshare

manysimilaritiestoOlcott’sChicagolandvalues,whichhavebeenusedinstudiessuchasAhlfeldt

andMcMillen (2018), Berry (1976), Kau and Sirmans (1979), McDonald and Bowman (1979),

McMillen(1996),McMillenandMcDonald(2002),Mills(1969),andYeates(1965).

IncontrasttopreviousanalysesbasedonMüller’sdata,weexploititsfullspatialdetailattheparcel

level.Topreservethehighly‐disaggregatednatureoftheoriginaldata,wedigitizeeverysingledata

pointwithinaone‐kilometerbufferaroundthenewlybuiltelevatedtrackswithinageographical

informationsystem(GIS)environment.Aftercreatingabalancedpanelforthefinalanalyses,this

leavesuswithatotalofabout38,000observationsforsevenpointsintime.

Forthecontemporaryanalysesweutilizeaconfidentialdataset,whichisthesameasinAhlfeldt&

Maennig(2015),containingdetailed informationonmorethan70,000transactionsofbuildings

(single‐familyandmulti‐family)andthecorrespondinglandparcelsandincludingfeaturessuchas

price,transactiondate,location,andasetofparametersdescribingbuilding/plotcharacteristics.

ThedatawereobtainedfromtheCommitteeofValuationExpertsBerlin(GutachterausschussBer‐

lin).Thetransactionsaregeo‐referenced(addressesandx/ycoordinates),whichallowsthemtobe

integratedintoaGISenvironment.Thebuildingcharacteristicsincludefloorspace,parcelarea,age,

Easevs.noise 8

landuse,qualityofthebuildingstock,locationwithinablockofhouses(e.g.,acornerlot),andsev‐

eralotheramenitieslikebasements,elevators,etc.

2.3 Railnoise

Totranslatethetypicallyvolatilelevelsofrailnoiseintoastandardizedsummarystatistic,engi‐

neerscomputetheequivalentcontinuoussoundlevel,which isessentiallyasophisticatedmean

overthevaryingnoiselevelsobservedduringagivenperiod.Weuseahighlydisaggregatedmap,

containing2007estimatesofthecontinuoussoundlevelbythesourceofnoise(includingrail)ata

10x10‐metergridfromBerlin’sSenateDepartmentforUrbanDevelopmentandtheEnvironment

(2013).Thenoisemeasurereflectstheweightedaveragenoiseexposureoveroneyearandalltimes

ofaday(Lden)atareceptionpointoffourmetersabovetheground.Followingtherulesdefinedby

theEUEnvironmentalNoiseDirective,themicro‐geographicnoisemapistheresultofasimulation

usinga3Dmodelthatisfittoactualnoisemeasurements.Themodelincorporatesfeaturesofthe

trackdesign(e.g.speed,squeakingnoisesincurves,thepresenceoflubricationfacilities)andthe

terraingeography(e.g.elevationofthetrack,built‐upstructure,bridges)thataffectnoisedissemi‐

nation.Summarizingexistingresearch,Navrud(2002)concludesthat“[…]theeliminationofnoise

annoyanceoccursat37‐40db”.Thus,wemeasurerailnoiseintermsofdecibelsexceeding40deci‐

bels,i.e.45,50,and55decibelscorrespondto5,10,and15excessdecibels.Asweillustrateinan

auxiliaryanalysispresentedinappendixsection3.2,ourrailnoisemeasuresharplydeclineswith

distancefromthetrack,ishigherwheretrainsrunfaster,anddisproportionatelyaffectsthefirst

rowofbuildingsfacingthetrack.

Forourhistoricalepisode,estimatesoftherailnoiselevelunfortunatelydonotexistasthemeas‐

urementtechnologyhadnotbeendeveloped(Ampel&Uzzle1993).However,regardingthetrans‐

ferabilityofthecontemporarynoisemeasure,wenotethatthebuildingfootprintremainedlargely

thesamewithintheaffectedarea,despitesignificantdamageduringWorldWarII,asdocumented

on detailed ground plans published by the Berlin Senate Department (Senatsverwaltung für

StadtentwicklungBerlin2000).3Therefore,itseemsreasonabletoassumethatcontemporaryrail

noiselevelsalsoreflectthedisseminationofsoundabout100yearsagoinrelativeterms.Moreover,

theserviceoperatorwascontractuallyrequiredtoserveallstationsinatleastfive‐minuteintervals

duringdaytime,afrequencythatcorrespondstothecurrentservice(Lemke&Poppel1996).His‐

toricalandcontemporarytimetablesalsoreveal thattheaveragespeedremainedconstantover

time(Ahlfeldt,Redding,etal.2015).Thisisconsistentwitharollingstocktechnologythatdidnot

3 Notethatforveryfewplots,wherethebuildingstructurechanged,weimputehistoricnoiselevelsusingadjacentplots.

Easevs.noise 9

changefundamentally.Asdiscussedabove,LineAwasthefirstelectrifiedsubwaysysteminGer‐

many.Thetrains(typeA1/A2)aswellasthetrackdesignrepresentedarevolutionarytechnology.

Incomparison,thesubsequentimprovementsthatcamewiththeintroductionofnewtrainsinthe

1960s(typeA3,stillthebackboneofthefleet)wereevolutionary(Lemke&Poppel1996).

Theexactchangesinnoiselevelsfromthefirsttothesecondgenerationarenotdocumented,butit

seemslikelythattechnologicalprogressevenwithinasimilartechnologyatconstantspeedand

frequencyhasresultedinanataleastmoderatereductionofnoiselevels.Newgenerationsofroll‐

ingstocktendtoreducenoiselevelsofinter‐citytrainsbyabout10decibels(Clausenetal.2012;

Murphy&King2014),althoughasmallerreductionisexpectedforurbanrailsincetrainsoperate

atlowerspeeds.Moreover,lesstreecoverageinthepastmayhaveimpliedlessnoisemitigation.

Importantly,passivenoiseinsulationwasprobablyweakerinthepast,althoughthecharacteristic

woodendoubleboxwindows(Doppelkastenfenster)fromthelate19thcenturyhaveremainedpop‐

ularinBerlin.Allinall,itseemsreasonabletoassumethatourcontemporarynoisemeasurerep‐

resentsalower‐boundestimateofthenoiselevelsexperiencedintheearly20thcentury.

2.4 Visualdisamenity

Inadditiontoanoisedisamenity,anelevatedlinemaycauseavisualdisamenity.Theroutingof

LineAfollowsmajorroadswhichweresufficientlywidetoaccommodateaviaductinthemiddleof

thesides.Becausetheelevatedlinegenerallydoesnotobstructviewsofopenspacessuchasparks

orlakes,thevisualdisamenityislessobviousthanthenoisedisamenityinthepresentcase.More‐

over,addressingtheconcernsraisedbyBerlinresidentsmentionedabove,theelevatedtracksand

stationswereeventuallyexecutedwithsomeattentiontoarchitecture(Bohle‐Heintzenberg1980).

Toempiricallydisentangletheeffectsfromthenoisedisamenityandthevisualdisamenity,wecre‐

ateadummyvariablethattakesthevalueofoneifaparcelhasadirectviewoftheelevatedtrack

andzerootherwise.Moreover,subwayscausevibrationsthatpotentiallytransmittonearbybuild‐

ings,wheretheycanbeperceivedasadisamenity(Kurzweil1979).Becausetheeffectsarehighly

localizedandnormallyreachnofurtherthantothefirstrowofhouses(Melke1988),apotential

disamenityeffectshouldalsobecapturedbytheviewdummy.Previewingourresults,wedonot

findevidenceforadirectvieweffectconditionalonthenoiseeffectandfindsimilarnoiseeffects

whenexcludingparcelswithadirectviewfromtheanalysis.Wethereforegenerallyinterpretour

noiseestimatesasoriginatingpurelyfromnoise.

2.5 Otherspatialdata

Weutilizethecompletetransportnetworkdataforpost‐unificationBerlinprocessedbyAhlfeldt,

Redding,etal.(2015).Thenetworkdataconsistsofelectronicmaps(shapefiles)ofstreets(used

forwalkinganddriving),buses,trams,subway(U‐Bahn)andcommuterrail(S‐Bahn).In addition,

Easevs.noise 10

wedigitizetheundergroundandelevatedsectionsofLineAaswellastheotherhistoricaltrans‐

portationnetworks,includinghorse‐poweredbuses,horse‐poweredtrams(oneline),steam‐pow‐

eredtrams(oneline),electrifiedtrams(thegreatmajorityoftramlines),andcommuterrail(pow‐

eredbysteam).Tocompilethehistoricalnetworkdata(andtheassociatedspeeds)wecombinethe

contemporarytransportnetworkswithhistoricalnetworkplans.4Anillustrationofthehistorical

andcontemporarytransportnetworksisinappendixsection3.3.

We complement our keydata sets (property, access, noise)with several spatial characteristics,

whichwemergeinGIS, includingcontemporarymeasuresofdistancefromthecentralbusiness

district (still at thehistorical location), distance from theKurfürstendammsub‐center, distance

fromnearestlake,riverorcanal,distancefromnearestparkorforest,distancefromnearestland‐

markbuilding,distance fromnearestplayground,distance fromnearestmain street, andstreet

noise(excludingrailnoise).

2.6 Interpretationofestimatedimplicitprices

Ourhistoricalandcontemporaryanalysesutilizedifferenttypesofdata.Inourhistoricalanalysis,

weexploit thespatiotemporaldistributionof landprices. Inourcontemporaryanalysis, thede‐

pendentvariableistheratiooftransactionpriceofaparcelofland,includingthestructure,over

theparcelsize.Totheoreticallylinktheestimatedcoefficientsfromthesedistinctmodelstoeach

otheraswellastoavastliteratureanalyzinghouseprices,itisusefultoassumeaCobb‐Douglas

housingproductionfunctionandacompetitiveconstructionsector(Eppleetal.2010).

AssumethathousingservicesHareproducedusingtheinputscapitalKandlandLasfollows:

.Housingspaceisrentedoutatbid‐rent whilelandisacquiredatlandrentΩ.Combining

thefirstordercondition / / 1 Ω(wherethepriceofcapitalisthenumeraire)andthe

non‐profitcondition Ω gives / 1/ 1 Ω.Log‐linearizationyieldsarelation‐

shipwithaslopeofone,whichimpliesthatestimatedparametersfromourhistoricalmodels(in

whichthedependentvariablecorrespondstoln(Ω))andourcontemporarymodels(inwhichthe

dependentvariablecorrespondstoln( / ))aredirectlycomparable.Fromthefirst‐ordercondi‐

tionandthenon‐profitcondition,itisfurtherimmediatethatln 1 ln Ω ,wherecis

aconstantthatcancelsoutinfirst‐differences,i.e.,Δ ln 1 Δ ln Ω 1 Δ ln / .

Inlogterms,itis,therefore,possibletotranslatethecapitalizationeffectsfromourhistoricaland

4 Networkplansarealsoavailableonline; see, for instance,http://www.berlineruntergrundbahn.deandhttp://www.berliner‐verkehr.de.

Easevs.noise 11

contemporarymodelsintoafloorspacepricecapitalizationeffect,bymultiplyingtheformerbya

landshareparameter.

ItisimportanttonotethathousingservicesasdefinedbyEppleetal.(2010)arenotidenticalto

housingspace.Unitsofhousingservicescanbethoughtofasbundlesoffeatures,includinghousing

space,thequalityofmaterials,sophisticationofdesign,andaccesstocommunalandprivateexte‐

riorspace,thatgenerateequivalentconsumptionutility.Especiallyinplaceswherebuildingvol‐

umesaresubjecttobindingregulations,suchasincentralBerlin,supplyofhousingservicescanbe

elastic(atapricetheelasticity / / / 1 0)evenifsupplyofhousingspace

isnot,becausedeveloperschoosetoinvestinhousingquality(bettermaterialsanddesignsrequire

more / )toachievehigherrents .Infact,thebuildingfabricinthestudyareaisstilldominated

bythelate19thcenturystockandwherethebuildingshavebeenreplaced,thequantityofhousing

spacehasbeenregulatedbyfloorarearatiolimits.Yet,Hhasincreasedovertimeasthehistoric

buildingcapitalhasbeenupgraded,e.g.byretrofittingcentralheating,privatebathrooms,modern

kitchens,orbalconies(Hämer1990).Inappendixsection6.1,weshowthat / iscorrelatedwith

variousobservablefeaturesofbuildingcapital,conditionalonhousingspace.There,wealsoshow

thatvariousfeaturesthatarepresumablycorrelatedwithhousingcapitalandhousingservices,in‐

cludinghousingspace,decreasesignificantlyinstationdistanceandrailnoise,aspredictedfordis‐

amenities.

TheCobb‐Douglasformulationoftheproductionfunctionimpliesthattheelasticityofsubstitution

betweenlandandcapitalisunityatanygivenpointintime,suchthatasthepriceoflandincreases,

developersinvestincapital(viamaintenance,upgrades,orreplacements)atratesthatensurecon‐

stantfactorshares.Itdoesnotprecludethatthelandshareandthepriceelasticityofhousingser‐

viceschangeovertimeduetofactorsthatareexogenoustodevelopers’decisionsonfactorinputs.

AsdiscussedbyAhlfeldtandMcMillen(2018),theintensityofcapitalusevariesovertimeasthe

structureofdemand,regulation,orconstructiontechnologychange.Toaccountforsuchtrends,we

borrowseparatehistorical(1900)andthecontemporary(2000)estimatesoftheshareoflandin

totalhousingvalue inGermanyof1 0.18and1 0.32 fromKnoll,Schularick,

andSteger(2017).

3 Historicalestimates

3.1 Empiricalstrategy

Ourbaselineempiricalstrategyfortheestimationofhistoricalcapitalizationeffectscombineshe‐

donic (Rosen1974)anddifference‐in‐differences (DD)methods (Ashenfelter&Card1985).We

Easevs.noise 12

employthehedonicapproachtoexpressthepriceofaparceloflandasafunctionofvariousattrib‐

utes,includingrailnoiseandrailaccess,andtheirimplicitprices.TheDDmethodthenallowsusto

identifyatreatmenteffect(e.g.,ofrailaccessorrailnoise)bydifferentiatingacrossspace(with

differentdegreesofexposure)andtime(beforeandafterexposure).Ourbaselineempiricalspeci‐

ficationtakesthefollowingform:

ln , , , (1)

wherePitisthelandpriceofaparceliattimet, isaparcelfixedeffectcontrollingforunobserved

time‐invariant locational amenities such as pollution, onto which we cluster standard errors

(Bertrandetal.2004),and isayearfixedeffectcontrollingforcommonmacroeconomicshocks.

, , isatreatmentfunctionthatexpressestheeffectsofthemetrolineasafunctionofthe

straight‐linedistancetotheneareststationSi,theemittednoiseNi,andtime .

Whiletheopeningdateoftheline(1902)isknownapriori, theexacttemporalstructureofthe

capitalizationof theeffectsof the line into landprices isnot.Capitalizationwilloccurgradually

ratherthanimmediatelyiftheserviceisanexperiencegoodandittakessometimebeforetransit

ridersadjusttheirbehaviortotakefulladvantageofthenewoption.Ifthesemi‐strong(orstrong)

efficientmarkethypothesis(Fama1970)holds,marketswillrespondtoallinformationmadepub‐

liclyavailable,whichcanresultinanticipationeffectsassoonasthenewlineisannounced.Inset‐

tingupourDDmodel,webeginbyestimatingaseriesoftime‐varyingtreatmenteffectsthatreveal

thetemporaladjustmentpathinaflexiblemanner:

, , ,, ,…

(2)

where tisanindicatorvariable,whichtakesthevalueofoneiftheconditionismetandzero

otherwise.Parameters and eachrepresentanindividualDDparameterreflectinghowland

pricesforparcelsexposeddifferentlytonoiseandaccessibilityeffects(firstdifferences)changed

from1881toyearz(seconddifferences).

Wenotethat,becausetherewasnometrorailnoisepriortotheelevatedrailline,ournoisemeasure

reflectstheincreaseinnoiseduetotheelevatedrailline(suchthat ∆ ,where∆ isthebe‐

fore‐afterchangeinnoise).Therefore, providesafirst‐differenceestimateoftheeffectofrail

noiseonlandpricesthatcanbeinterpretedasahedonicimplicitprice.Incontrast, givesthe

changeinthehedonicimplicitpriceofdistancetostationlocationsfromyear1881toyearz,i.e.

,where isthehedonicimplicitpriceingivenyearz. canstillbeinterpreted

Easevs.noise 13

asthehedonicimplicitpriceofproximitytoastation sincein1881thestationscouldnotbe

anticipatedand,thus, 0.

Informedbythisanalysis,wethenestimateanextendedDDmodelwhichprovidesabefore‐and‐

aftercomparison,controllingfortheeffectsduringanidentifiedadjustmentperiod:

, , 1902 1902

,

(3)

where 1902 isanindicatorvariabletakingthevalueofoneforyearsafterthelineopening

and isthesameforavectorofyearsAduringwhichlandpricesappeartobeadjustingto

anewequilibrium.Notethatcomparedtodroppingthoseyears,controllingforadjustmenteffects

offerstheadvantageofprocessingmoreinformationforidentificationofcovariateeffects(intro‐

ducedinrobustnesschecks)andfixedeffects , .

ThecriticalandessentiallyuntestableassumptionofanyDDanalysisisthat, intheabsenceofa

treatment,allsubjects(irrespectivelyoftheintensityoftreatment)wouldhavefollowedthesame

trend.Aselectionproblemexistsifthetreatedandthenon‐treatedsubjectsdifferinobservableor

unobservable dimensions, and these differences imply heterogeneous responses to common

shocks.Inthecontextoftheanalysisoftransportinfrastructureeffects,itisanotoriousconcern

thattheplacementmaybeendogenoustolocationcharacteristicswhichmaybecorrelatedwith

trends.Avarietyoftechniqueshaveemergedtoaddressselectionproblems,manyofwhichaimat

weightingobservationsinsuchawaythatthetreatmentassignmentbecomesorthogonaltoob‐

servablecovariates.Examplesincludetheinverseprobabilityweighting(Hernánetal.2001)and

the special case of entropy balancing (Hainmueller 2012), the propensity score matching

(Rosenbaum&Rubin1983),orthesyntheticcontrolmethod(Abadie&Gardeazabal2003).The

problemwiththeapplicationofthesetoolstothepresentcaseisthattheyservethepurposeof

evaluatingsingulartreatmentsandnotmultiplecorrelatedtreatments.

Intheabsenceofasuitableoff‐the‐shelfmatchingtechnique,weuseasimplesledgehammerap‐

proachtodefiningparcelweightsthatminimizetheconditionalcorrelationsbetweenbothtreat‐

mentvariablesandthe1881‐1890trendinlandprices,aperiodforwhichweareconfidentthat

thelinehasnotbeenanticipated.Wenotethatthisisthefirstapplicationofthisweightedparallel

trends(WPT)DDapproach.Tosavespace,werelegateamoretechnicaldiscussion, includinga

Monte‐Carlo evaluation of the small‐sample properties of the estimator, to a companion paper

Easevs.noise 14

(Ahlfeldt2018).5Inlinewithotherweighting‐basedmatchingtechniques,weviewthe1881‐1890

trendinlandpricesasacovariatetobebalanced;however,balancingmustbeachievedwithrespect

totwocorrelatedtreatmentassignments,noiseandstationdistance.Undertheidentifyingassump‐

tionthatthecorrelationbetweentreatmentsandunobservedfactorsthat interactwithtimeare

time‐invariant,successfuleliminationoftreatment‐trendcorrelationsduringthepre‐treatmentpe‐

riodimpliesthatnon‐paralleltrendsarealsoremovedinpotentialoutcometrendsduringthepost‐

treatmentperiod.Toachievethispurpose,wedefinethefollowingparcelweights:

∑, , , , (4)

where, , … , areparameterstobeidentified. , ,isoneofmvariablescapturingobserva‐

bletime‐invariantparcelcharacteristicsthatenterstheweightsinaGaussiantransformation:

, ,1

√2exp

12

, 2

, (5)

wherethebandwidths aresetaccordingtotheSilverman(1986)ruleandtheupperbarindi‐

cates themeanofadistribution.WeusetheGaussiantransformationbecausewepresumethat

parcelsthataremore“normal”withrespecttoaplotcharacteristic , aremorelikelytobeona

similartrend.Furthermore,wepresumethatparcelsthatarerepresentativewithrespecttodiffer‐

entcharacteristics , arelikelyondifferenttrends.Thisapproachhasbeenchosensoastomix

thesedifferenttrendsinawaythatensuresthattheaveragetrendintheweightedsampleisor‐

thogonaltothetreatments.ApositivecollateraloftheGaussiantransformationisthatall ,

, , arepositiveandinthesamedimension.Inthebaseline,weusedistancefromtheCBD,

distancefromasub‐centre,and1881‐1890pricegrowthasparcelcharacteristics inthealgo‐

rithm.Insearchingforavector thatminimisestheobjectivefunction,wesearchoveraparameter

space defined by 0, 0.01, 0.02, … ,1, 0, 0.01, 0.02, … ,1, 0, 0.01, 0.02, … ,1, which

equatesto101^3=1,030,301combinations.Weselect thatminimizesthesumofsquaredpartial

correlationsbetweenour treatmentmeasures(railnoiseandstationaccess)and the landprice

growthoverthe1881to1890period.6

5 Thecompanionpapercitesanearlierworkingpaperversionofthispaper.

6 Tothisend,werunrregressionsoftheform∆ln , ,where∆ln , isthechangeinloglandpricefrom1881to1890andtildedenotesnormalizationbystandarddeviation.Ineachregression,observationsareweightedbyWi,whichdependsonthevector , … , .Weselectthecom‐

binationofparametersthatminimizes∑ , .

Easevs.noise 15

Tooveridentifyourparcelweights,weuseinformationthatdidnotentertheweightsconstruction.

Wehavetwomorepre‐openingperiodsinourdataset(1890‐1896,1896‐1900)whichweuseto

evaluatewhetherthecommontrendsassumptionholdswithintheweightedsample.Wehaveex‐

perimentedwithalternativesetsofparcelcharacteristicsandobjectivefunctionsandourchoices

arebasedontheirperformanceintheoveridentificationtestreportedinappendixsection4.There,

wealsoevaluatewhethertheweightingchangesthecompositionofthesamplewithrespecttoob‐

servableparcelcharacteristics.Theweightedsampleresemblestheunweightedsampleintermsof

observablecharacteristics(seeappendixsection4.1).Whileeveryweightedanalysisresultsina

localestimate,inourcaseitisatleastnotobviousthattheweightedDDeffectsareidentifiedfrom

parcelswithveryparticularcharacteristicsthatwouldimpedegeneralizabilitywithinoursample.

3.2 Baselineresults

InFigure2,weillustratethetime‐varyingtreatmenteffects,estimatedaccordingtotheDDmodel

(1)usingthetreatmentfunction(2)andtheweightsdefinedin(4)and(5).Wereportrailnoiseand

stationdistanceeffects,estimatedunconditional(solidlines)andconditional(dottedlines)oneach

other.Estimatedstationdistanceeffectsaremultipliedby‐1toensurethatpositivenumbersmean

normativelypositiveeffects.Ourweightedestimationapproachachievesitspurposeofeliminating

pre‐trends,i.e.,thereisnosignificantcorrelationbetweenthe1881‐1890landpricetrendonthe

onehandandproximity tostationsorexposuretorailnoiseontheother.Proximityeffectsare

insignificantin1896and1900andthenoiseeffectisinsignificantin1900(yearsthatwerenotused

intheconstructionoftheweights),indicatingthatthecommontrendsassumptionholdswithinthe

weightedsample.

Stationdistanceeffectsremaininsignificantduringallyearspriortotheopeningofthelineand

becomesignificantlypositiveafterwards,witha tendencyto increaseover time.Theabsenceof

anticipationeffectsincombinationwiththegradualadjustmentaftertheopeningofthelineare

consistentwithaninterpretationthatthelinerepresentsanovelmodeoftransportationwhose

benefitswereyettobeexperienced.Controllingforrailnoise,aone‐kilometerdecreaseindistance

fromthestationincreaseslandpricesinthelong‐runbysomenotable0.3logpoints(35%).

Easevs.noise 16

Fig.2. Difference‐in‐differences:Time‐varyingtreatmenteffects(WPTmodels)

Note: Time‐varyingtreatmenteffects and basedonbaselineDDequation(1)andtreatmentfunction(2).

WPTmodelsuseweightsconstructedtominimizetheconditionalcorrelationsbetweennoiseandthe1881‐1890landpricetrendaswellasaccess(distancefromstation)andthe1881‐1890landpricetrend.Accessparameters(effectsofdistancefromstation)multipliedby‐1sothatpositiveshiftsindicatepositiveeconomiceffects.Verticalerrorbarsindicatethe95%confidenceintervalbasedonstandarderrorsthatareclusteredonparcels.Solidverticallinesdenotetheyearofopeningofthemetroline(1902).

Theestimatedweightedrailnoiseeffectsalsodisplayanintuitivepattern.Controllingforstation

distanceeffects,a10‐decibelincreaseinrailnoiseisassociatedwithareductioninlandpricesby

slightlymorethan4%inthelong‐run.Incontrasttoourresultsforstationdistanceeffects,wefind

notableanticipationeffectsofrailnoisefor1896.Thisfindingisplausible inlightoftheintense

publicdebateabout theaestheticappealofelevatedrail lines.Theconflictwassettledafter the

announcementtoimprovethearchitecturaldesignofthestationsandtheviaductandthedecision

tobuildanundergroundlinewithintheboundariesofthecityofCharlottenburg,explainingwhy

theanticipationeffectdisappearsin1900.Inkeepingwithintuition,estimatedstationdistanceef‐

fectsincreasebyaboutonethirdifrailnoiseeffectsarecontrolledfor.Theeffectofcontrollingfor

stationdistanceeffectsonrailnoiseeffectsisevenlarger.

Easevs.noise 17

InformedbyFigure2,wenowproceedtoestimatingparametricbefore‐afterDDeffects,usingour

baselinespecification(1),thetreatmentfunction(3),and,again,theweightsdefinedin(4)and(5).

TheresultsarereportedinTable1.Forcomparison,wepresentweightedDDestimatesofstation

distanceeffectsnotcontrollingforrailnoiseeffects(columns1‐2)andrailnoiseeffectsnotcontrol‐

lingforstationdistanceeffects(columns3‐4). Incolumns(5‐6)ofthetable,wethenreportour

preferredstationdistanceandrailnoiseeffectsestimatedconditionaloneachother.Wecontrolfor

anticipationeffectsin1896and1900asindicated.

Whenwedonotcontrol forrailnoiseeffects,ourestimationresults indicate that thepriceofa

parcellocatedrightatastationincreasesby12.7%(=exp(0.120)‐1)aftertheopeningoftheline,

comparedtoaparcelonekilometerawayfromastation.Railnoiseeffectsareclosetozeroand

statisticallyinsignificantifstationaccessibilityisignored.Controllingforanticipationeffectsinei‐

thercasehasaminorimpactontheestimatedraileffects.Acomparisonoftheseresultstocolumns

(5‐6)highlightstheimportanceofjointlyidentifyingatransportationinfrastructure’samenityand

disamenityeffects.Asshownincolumn(6),thestationdistanceeffectincreasesto20.2%inour

preferredmodel.Moreover,inlinewithFigure3,the(negative)railnoiseeffectisnowstatistically

significant.Thepointestimatesindicatethata10‐decibelincreaseinrailnoisecausesarelative

declineinlandpricesby3.7%.Comparingourestimatesacrossthedifferentspecifications,thebias

thatresultsfromignoringcountervailing(dis)amenityeffectsamountstoasmuchasabout35%

0.184 0.119 /0.184 instationdistanceeffectsandtoabout85%inrailnoiseeffects.Inthis

context,itisworthnotingthatconsistentwiththeinsignificantnoiseeffectincolumns(3‐4),our

preferredestimatesincolumn(6)suggestthatpositiveaccessibilityeffectsaboutoffsetthenega‐

tivenoiseeffectfortheparcelsexposedtothehighestlevelsofnoise(seeappendixsection4.2for

details).

ThetreatmenteffectsreportedinTable1arederivedfromacomparisonofthemeanlandpriceat

theparcel levelintheperiods1881‐1890and1904‐1914.Sincethismodelignorespricetrends

aftertheopeningofthe line,theeffectsaresmallerthanthe1914treatmenteffectsreportedin

Figure2.Theseparametricestimates,however,areclosertothestandardapproachinthelitera‐

ture,thereforeprovidingamorereasonablestartingpointforacomparisonofourquantitativere‐

sultstocontemporaryestimatesintheliterature.

Easevs.noise 18

Tab.1.Noiseanddistanceeffects:Historicalweighteddifference‐in‐differencesestimates

(1) (2) (3) (4) (5) (6) Lnlandprice(1881‐1914)Distance(km)xafter

1902 ‐0.120***(0.025)

‐0.119***(0.032)

‐0.173***(0.031)

‐0.184***(0.040)

Noise(10db)xafter1902

0.001(0.006)

‐0.004(0.008)

‐0.029***(0.007)

‐0.036***(0.010)

Parceleffects Yes Yes Yes Yes Yes YesYeareffects Yes Yes Yes Yes Yes YesAnticipationeffects ‐ Yes ‐ Yes ‐ YesN 37,933 37,933 37,933 37,933 37,933 37,933r2 .93 .93 .93 .93 .93 .93

Notes: Weightedmodelsuseweightsconstructedtominimizetheconditionalcorrelationsbetweennoiseandthe1881‐1890 landpricetrendaswellasaccess(distance fromstation)and the1881‐1890 landpricetrend.Afterisadummyvariableindicatingyearsafterthelineopening(1902).Announcementeffectsaredistanceandnoisevariablesinteractedwith1896and1900effects.Balancedpanelofrepeatedparcelobservationsfor1881,1890,1896,1900,1904,1910,1914.Standarderrorsinparenthesesareclusteredonparcels.*p<0.10,**p<0.05,***p<0.01.

3.3 Robustnesschecksandcomplementaryanalyses

Wehaveperformedanumberofperturbationsof thebaselinemodelreported incolumn(6)of

Table1toaddressvariousconcerns.Forinstance,weobtainsimilarresultswhenweusedifferent

covariatesandobjectivefunctionsintheweights‐generatingalgorithm.Wealsofindthatthebase‐

lineresultsarereasonablyrobusttoallowingfortime‐varyingimplicitpricesofvariouslocation

characteristics(capturedbycontrols×yeareffectsinteractions).Allowingforinteractionsofnoise

anddistancevariableswithseparatetimetrendsbeforeandaftertheopeningofLineAresultsin

cumulatedeffectsafter10yearsthatareveryclosetothebaselineestimates.Addingadummyvar‐

iableindicatingparcelswithanunobstructedviewoftheelevatedlinedoesnotsignificantlyaffect

thenoise(orthedistance)effect.Similarly,theresultshardlychangeifallparcelswithadirectview

oftheelevatedlineareexcluded.Avieweffectisonlysignificantifthenoisemeasureisexcluded

fromthemodel.Notcontrollingfornoise,parcelswithadirectviewexperiencedarelativedecrease

inthelandpriceof4.4%,whichissubstantiallylessthanimpliedbythenoiseeffectatthesame

location(about‐9.5%;seepreviousparagraph).Itis,therefore,unlikelythatournoiseestimates

areconfoundedbyaviewdisamenityeffectoradisamenityfromsubwayvibrations(asbotheffects

shouldbehighlycorrelated).Wehavealsoevaluatedthespatialdecayinthedistanceeffectusinga

seriesofdummiesdenotingparcelsinmutuallyexclusive100‐meterstationdistancebins.Wefind

thatthedistanceeffectis largelyconfinedtothefirst400meters,withnoevidencefornegative

congestioneffectsatclosedistances.Comparingtheeffectintheinnermostringversustheouter‐

mostresidualcategoryresults inaneffect that isalmost identical to theone‐kilometerdistance

effectfromthebaselinemodel.Wehavealsoevaluatedthestabilityofthehedonicfunction(Kumi‐

Easevs.noise 19

noffandPope,2014)aroundtheopeningdatesbycomparingmarginaleffectsofotherspatialat‐

tributesover timeandexperimentedwithvarying levelsof spatial clustering.Theserobustness

tests and complementary analyses arepresentedanddiscussed indetail in appendix section4,

wherewealsopresenttheresultsofanunweightedOLSanalysesfortheinterestedreader.Asa

finalandparticularlypowerfulrobustnesscheck,wealsoevaluatethenoiseeffectexploitingadis‐

continuityinnoiseatthetunnelentranceclosetoNollendorfplatz,findingqualitativelyandquan‐

titativelysimilarresults.Thisanalysisispresentedinappendixsection5.

4 Contemporaryestimates

4.1 Empiricalstrategy

Intheabsenceofvariationovertimeinthemetrorailnetworkduringthecontemporarystudype‐

riod(1990‐2012),weestimateacross‐sectionalmodel.Toimprovetheidentificationofnoiseef‐

fects,werestricttheidentifyingvariationtothesharpchangeinnoisethatarisesatninetunnel

entranceswhereelevatedlinesturnintoundergroundlines.Thereasonsforthetransitionandthe

selectionofthelocationofthetunnelentrancesareoftenspecifictotheline(Bohle‐Heintzenberg,

1980).Inparticular,weestimatemodelsoftheform:

ln ε , (6)

where isthepropertytransactionpricenormalizedbythelotsizeofapropertyjsellingat

timetwithinthecatchmentareaofstationcandwithinanetworkcorridore.Asdiscussedinsection

2.6,thisspecificationaccountsforendogenoushousingqualityandyieldsmarginaleffectsofrail

noiseandrailaccessthataredirectlycomparabletothehistoriclandpriceeffectsestimatedinsec‐

tion3.Incontrasttoconventionalhedonicanalysesusingsalesprices(correspondingto inno‐

tationsofsection2.6),housingattributeslikethenumberofbathroomsorbedroomsmustnotbe

controlledfor. / isdirectlyobservedinthedataandtheoreticallyonlydependsonfactors

thataffect the landprice, i.e. locationalcharacteristics. Incontrast to the theoretical framework

outlinedinsection2.6,however,housingisdurablesuchthattheactualbuildingcapitaldoesnot

necessarilycorrespondtotheequilibriumvaluesincecapitaldepreciates(seeappendixsection6.1

forestimatesofthedepreciationrate).Therefore,wecontrolforageinthevector ,whichalso

containsahostoflocationalcontrolvariables.

ThevariablesSandNareourrespectivemeasuresofstationdistanceandrailnoiseasbefore, is

afixedeffectforstationcatchmentareasand isayearfixedeffect.Sincesubwayandcommuter

Easevs.noise 20

railuseasimilartechnologyinthecontemporaryperiod,wetreatbothtypesofstationsasperfect

substitutes.Stationcatchmentareasare, therefore,defined forgroupsofpropertiessharing the

sameneareststation.Inourbaselinespecification,werestrictthesampletoareaswithinonekilo‐

meter of thenearest station.As evident fromFigure3, thedensity of stations is relatively high

withinthecentralpartsofBerlin,furtherreducingthesizeofacatchmentarea.Themeancatch‐

mentareais just1.3squarekilometers(about0.8squaremiles)asopposedtomorethanthree

squarekilometersimpliedbyacirclewithaone‐kilometerradius.Withtheinteractioneffects

,we, thus, provide a strong control forunobserved location characteristics such as pollution,

changesinlocationalcharacteristicsandchangesintheimplicitpricesoflocationcharacteristics.

Fig.3. Contemporaryrailnetworkandstationcatchmentareas

Notes: Own illustration using the Urban Environmental Information System of the Berlin Senate Department

(SenatsverwaltungfürStadtentwicklungBerlin2006).

Critical for the identificationof thenoiseeffect, isasetof fixedeffect forrailcorridors.Each

corridoriscenteredontheintersectionoftherailnetworkandoneoftheninetunnelentrances

indicatedbytheorthogonalsinFigure3.Weusecorridorsdefinedbasedonatrackdistanceof100

metersandadistancefromtheorthogonalof1000meters.Theinteractionfixedeffects

capturearbitraryshockstoanyofthesecorridors.Wedefineanauxiliaryrunningvariable that

takes the distance from the nearest tunnel entrance (negative distances in the tunnel section)

withinacorridoreandavalueofzeroelsewhere.Wethenuseadummyvariableindicatingthe

Easevs.noise 21

elevatedpartsofthosecorridors 0 ( isonewithinanyofthecorridors)asanin‐

strumentfornoisetorestricttheidentificationtothedifferenceinnoiseacrosselevatedandun‐

dergroundsegmentswithincorridors.

4.2 Baselineresults

Figure4illustratesrailnoiseandcontemporarypropertypricesalongtherailcorridorsandtunnel

entrances.Wepresentmeanvaluesofoutcomeswithin100‐meterbinsandconfidenceintervals

thatsummarizewhetherthewithin‐binmeanissignificantlydifferent(atthe90%level)fromthe

meanacrossallobservationswithinacorridorontheothersideofatunnelentrance.

Fig.4. Contemporaryspatialdifferencesinnoiseandpropertyprices

Notes. Eachcircleillustratesthemeanvalueofadependentvariablewithinagridcell.Onedimensionofthegrid

cellsare200‐mbinsdefinedbasedonthedistancefromthetunnelentrance.Theotherdimensionisa100‐m‐distancebufferaroundthetrack.Negativedistancesfromthetunnelrefertotheundergroundsection.Solidhorizontallinesindicatethemeans(weightedbythenumberofobservations)withintheunderground(neg‐ativedistance)andelevated(positivedistance)segments.Errorbarsarethe90%confidenceintervalsbasedonrobuststandarderrorsfromseparateparcel‐levelregressions(withinthebuffer).Foreachoutcome,werunoneregressionoftheoutcomeagainstdummiesindicatingpositivedistance(≥0)bins,andanotherre‐gressionoftheoutcomeagainstdummiesindicatingnegativedistance(<0)bins.Foreachbin,theerrorbarrepresentsatestifthemeanwithinthebinisdifferentfromthespatialcounterfactual(thedashedline).Theboundaryeffectcorrespondstothedifferencebetweenthetwohorizontallines.Transactionpricesaretheresidualsplustheblockfixedeffectcomponentfromregressionsofthenaturallogofthetransactionpricenormalizedbylotsizeagainstahostofhedoniccontrols,yeareffects,andblockfixedeffects,severaldistancevariables,includingdistancefromthecentralbusinessdistrict,distancefromthenearestlake,riverorcanal,distancefromnearestparkorforest,distancefromnearestlandmarkbuilding,distancefromnearestplay‐ground,distancefromnearestmainstreet,streetnoise(excludingrailnoise).

Withintheserailcorridors,thelevelsofrailnoisealongtheelevatedsegmentsexceedthatofthe

undergroundsegmentsbyabout18decibels.Theadditionalnoisecomeswithadiscountonland

pricesof‐0.26logpoints.Fouroutoffivehighnoisebins(elevatedsection)havemeanpricesthat

Easevs.noise 22

aresignificantlylowerthanthemeanpricewithinthelownoise(underground)sectionandfour

outofsixlownoise(undergroundsection)binshavemeanpricesthataresignificantlyhigherthan

themeanpricewithinthehighnoise(elevated)section.Theimpliedpriceeffectofa10‐decibel

increaseinrailnoiseisabout‐0.14logpoints,morethanthreetimesthelandpricecapitalization

effectinthehistoricalperiod.

Table2reportstheestimatesforseveralvariantsofequation(6).Incolumns(1‐3),wepresent,for

comparison,theresultsofaconventionalhedonicmodel,whichexcludesallcorridor‐relatedvari‐

ablesanddoesnotusetheinstrument.OurpreferredSDspecificationsforthenoiseeffectsidenti‐

ficationaretabulatedincolumns(4‐6).Forbothvariants,wereportresultsofmodelsthatexclude

(1and4)andinclude(2and5)stationcatchment×yeareffectsaswellasmodelsthatusealltrans‐

actions(1‐2and4‐5)orsamplesrestrictedtopropertieswithinonekilometeroftheneareststation

(3and6).

Theestimatedstationdistanceeffectsarerelativelystableacrossallspecifications.Ourpreferred

estimateoftheper‐kilometerstationdistanceeffectisthe exp 0.144 1 /100 15.4%esti‐

matefromcolumn(3),forseveralreasons.Inmodel(3),stationcatchment×yeareffectscontrol

forarbitraryshocksatarelativelylocallevel.Moreover,therestrictiontoaone‐kilometerstation

radiusfurtherincreasesthestrengthofthiscontrolandmakestheresultsmorecomparabletoour

historicalanalysis.Importantly,themodelcontrolsfornoisealongallelevatedsegmentsofthenet‐

workwhereasintheSDspecificationmuchofthevariationinnoiseisintentionallywipedoutby

theinstrument.

TheSDmodelsconsistentlypointtorelativelylargeandnegativenoiseeffects.Themostconserva‐

tiveestimatesuggeststhata10‐decibelincreaseinnoisereducesthepropertypriceperlandunit

(andundertheassumptionsmadeinsection2.6alsothe landprice)byabout11.5%.Giventhe

geographyoftheBerlinrailnetwork,itisintuitivethatthehedonicmodelsincolumns(1‐3)yield

smallerestimates.Thesubwaynetworkoftenfollowsmajorboulevardsthatwerelaidoutinthe

1862Hobrecht‐Plan(Bernet2004),whichborrowedmanyfeaturesfromHaussmann’sdesignsfor

Paris(deMoncan2009).Theseboulevardsprovidethenecessaryspacefortheconstructionofvia‐

ducts forelevated linesor facilitate the cost‐effectiveopenconstructionof tunnels. Suchboule‐

vards, however, also possess desirable features such as distinctive architecture, tree coverage,

shops,boutiquesandrestaurants,whicharenotobservedinthedata.Ifthesefeaturesareempiri‐

callyconfoundedwithrailnoise,thenoisedisamenitywillbeunderestimated.

Easevs.noise 23

Tab.2.Contemporaryanalysis

(1) (2) (3) (4) (5) (6) Lnpropertytransactionprice/lotsizeDistance(km) ‐0.128*** ‐0.126*** ‐0.144*** ‐0.127*** ‐0.126*** ‐0.152*** (0.003) (0.007) (0.021) (0.003) (0.007) (0.022)Railnoise(10decibel) 0.050*** ‐0.021 ‐0.032** ‐0.166*** ‐0.143*** ‐0.122** (0.011) (0.015) (0.015) (0.032) (0.049) (0.049)Controls Yes Yes Yes Yes Yes YesYeareffects Yes ‐ ‐ Yes ‐ ‐Stationxyeareffects ‐ Yes Yes ‐ Yes YesCorridorxyeareffects Yes Yes YesNoiseinstrument Yes Yes YesSample All All Station

distance<1km

All All Stationdistance<1km

N 71,313 71,313 46,143 71,313 71,313 46,143r2 .259 .584 .608 .261 .586 .61

Notes: Unitofanalysisispropertytransaction.Controlsincludestructureage,dummiesforlocationwithinablock(cornerlot,streetfront,backyard,etc.),dummiesforbuildingcondition(poor,good),distancefromnearestlake,riverorcanal,distancefromnearestparkorforest,distancefromnearestlandmarkbuilding,distancefromnearestplayground,distancefromnearestmainstreet,streetnoise(excludingrailnoise).Stationeffectsidentifygroupsofpropertieswhichhave the samenearest rail station.Corridoreffects identifygroupsofpropertieswithin100‐meterbuffersalongarailline,spreading1,000meterinbothdirectionsfromatunnelentrance.Noiseinstrumentisadummyvariabletakingthevalueofonewiththeelevatedsegmentofanyrailcorridorandzerootherwiseinmodels(4‐6).Standarderrorsinparenthesesarerobustin(1)and(4),clus‐teredstationxyeareffectsinallothermodels.*p<0.10,**p<0.05,***p<0.01.

4.3 Robustnesschecksandcomplementaryanalyses

Wehaveexpandedtheanalysisofcontemporarypropertypriceeffectsinseveraldirections.We

haveevaluatedtheancillarypredictionfromthetheoreticalframeworkinSection2.6thatincreases

inlandvaluesduetolocationalamenitiesshouldbeaccompaniedbyinvestmentsinbuildingcapital

andalargerquantityofhousingservices.Wefindthatincreasesinstationdistancebyonekilometer

andincreasesinrailnoiseby10decibelsreducethesupplyoffloorspaceperlandunitbymore

than20%andabout10%,respectively.Thereisalsoanegativeeffectonbuildingconditionsaswell

asthepropensityofbuildingswithfeaturessuchaselevators,basements,orundergroundparking.

Toallowforamoreexplicitcomparisontothehistoricalanalysis,weestimatedistanceandnoise

effectswithintheone‐kilometerbuffersurroundingtheelevatedpartofLineAdepictedinFigure1.

Theamenityanddisamenityeffectswithinthebufferareverysimilartotherestofthecityarea.If

anything,thedistanceeffectappearstobesomewhatlarger(‐19.3%perkilometer),althoughthe

differencebetweentheeffectsinbothareasisnotsignificant.Withasimilaraim,weestimatethe

distanceeffectforthesubway(U‐Bahn)andcommuterrail(S‐Bahn)networkseparately.Thedis‐

tanceeffectforthesubwaynetworkof21.9%perkilometeris,again,somewhatlargerthaninthe

baseline.Inrobustnesschecks,weanalyzethesensitivityoftheresultstovariationsinthedefini‐

tionoftherailcorridoranddifferentattemptstoachieveamorelocalidentificationinareduced‐

Easevs.noise 24

formframework(usingthenoiseinstrumentasanexplanatoryvariable).Narrowerdefinitionsof

therailcorridor(75or50meters)resultinsimilarpointestimates,butlargerstandarderrors.Fur‐

therrestrictingtheidentificationtovariationclosertothetunnelentrancebyweightingobserva‐

tionsbydistanceoraddingdistancetrendsresultsinlargernoiseestimates.Acomplementaryanal‐

ysisofnon‐lineardistanceeffectsrevealsthatthedistanceeffectslargelycapitalizewithinthefirst

500meters,withnoevidencefornegativecongestioneffectsatclosedistances.Thepeakcapitali‐

zationeffectclosetothestationrelativetotheone‐kilometerstationdistancemargin,atabout20%,

issomewhatlargerthanimpliedbythebaselineestimate.Wealsofindthatconditionaloncontrols

thedifferenceinroadnoisewithinelevatedandundergroundsegmentsofourrailcorridorsisclose

toandnotstatisticallydistinguishablefromzero.Thus,withthechosenresearchdesign,roadnoise

isunlikelyapotentialconfounderforrailnoiseeffects,andsoareotherdisamenitiessuchaspollu‐

tionthatarelikelycorrelatedwithroadnoise.Amorecompletepresentationanddiscussionofthe

extensionsandrobustnesschecksisinappendixsection6.

5 Interpretation

5.1 Comparisonofhistoricalandcontemporaryestimates

Thusfar,wehaveprovidedcontemporaryandhistoricalestimatesofcapitalizationeffectsofnoise

andrailaccessintolandprices.UsingthetheoreticalframeworkdiscussedinSection2.6,itispos‐

sibletoretrievetheimpliedhousepricecapitalizationeffects.Toobtainestimatesofthelong‐run

incomeelasticitiesof(dis)amenityvaluesofnoiseandaccess,wemakesomefurtherassumptions.

Inparticular,weassumethat,withineachperiod(historicandcontemporary),(i)preferencesfor

allgoods(includingnoiseandaccess)arehomogeneous,andsoareexpendituresharesonhousing

andlandsharesintheproductionofhousing(thisdoesnotprecludedifferencesacrossperiods);

(ii)real incomesgrowataconstantrateforallpopulationgroups(thisdoesnotprecludelevel‐

differencesacrossgroups);and(iii)theestimatedmarginaleffectsofnoiseandaccessarecausal

andconstantacrossthedistributions(fornoisethisconcernsvaluesexceeding40decibels).Wecan

thendefinethewillingnesstopay(WTP)foraunitamenityincreaseinperiodtastheproductof

thecapitalizationeffectinhousepriceterms 1 (1 isthelandshareasdefinedinsec‐

tion2.6),income ,andtheexpenditureshareonhousing : 1 .Taking

log‐differencesandrearrangingtheWTPequationgivestheincomeelasticityoftheamenityvalue:

∆ ln∆ ln

1∆ ln∆ ln

∆ ln 1∆ ln

∆ ln∆ ln

, (7)

Easevs.noise 25

Ofcourse,theassumptionsmadearedisputableandaresubjecttoacriticalassessmentinappendix

section7,wherewealsoprovideadetaileddiscussionofthecalibratedvaluesfor∆ ln 1 , ∆ ln

and∆ ln .Acknowledgingthatconsiderableuncertaintysurroundsourestimatesofbothlong‐run

incomeelasticities,weprovideasummaryofourmaintakeawaysbelow.

5.1.1 Noise

Overaperiodofabout100years,theeffectofa10‐decibelincreaseinnoiseonlandpricesroughly

tripledfrom‐4.2%(Table2,column3)to‐13.0%(Table3,column6).Undertheassumptionsmade,

thiscorrespondstoanincreaseintheper‐decibelhousepricecapitalizationeffectfrom‐0.1%to

‐0.4%, the latterbeingwithin the rangeof contemporaryestimatesof aircraft noise (Boes and

Nüesch,2011report‐0.5%perdecibel)androadnoiseeffects(Graevenitz,2018reportsarange

of‐0.1%to‐1.4%perdecibel).Theimpliedincomeelasticityofthenoisedisamenityvalueis2.2.

This long‐run incomeelasticity estimate iswithoutprecedent, but complements cross‐sectional

stated‐preferenceestimatesthatpointtoanincomeelasticityofthemarginalcostofnoisebelow

unity(Wardmanetal.2005citeacentralestimateof0.5).

Onepossibleconcernwiththeinter‐temporalcomparisonwemakeisthatwedonotobservehis‐

toricrailnoise.Forthereasonsdiscussedinsection2.3,contemporaryrailnoiselevelslikelyun‐

derstatehistoricalnoise levels, implyingthatourhistoricalnoiseestimatesareupwardlybiased

andthelong‐runincomeelasticityofthenoisedisamenityvalueislikelylargerthanthevaluewe

infer.Anotherconcernisthat,inthepast,roadnoiselevelswerelikelylowerduetotheabsenceof

affordablemass‐producedcars.Thiswillbeapotentialproblemifwerelaxtheassumptionofa

constantmarginaleffectofnoise.Ifthedisamenityeffectsofrailandroadnoiseweremutuallyre‐

inforcing,anincreaseinroadnoiseovertimewouldleadtoahighernoisecapitalizationeffecteven

intheabsenceofachangeinnoiseaversion.However,inanancillaryanalysis,wefindthattherail

noisecapitalizationeffectdecreases in thepresenceofhigher levelsofroadnoise, i.e. railnoise

matterslessifthereisalreadyalotofroadnoise.So,withoutapresumedincreaseinroadnoise

levelsovertime,therailcapitalizationeffecttodaywouldlikelybeevengreater,implying,again,a

largerincomeelasticity.Ifwerelaxtheassumptionofhomogeneouspreferences,itseemsreason‐

abletoexpectthatafter100yearsofsortingmostnoisesensitivehouseholdswillhaveleftthenois‐

iestareas(KuminoffandPope,2014).This,again,mutesthecontemporarynoisecapitalizationef‐

fectandincreasestheimpliedincomeelasticity.However,theoverallincreaseinnoiselevelsacross

thecitycouldalsoleadtothemarginalbuyerinanoisyareabeingmorenoisesensitive,sothatthe

Easevs.noise 26

neteffectofsortingisambiguous.Importantly,rapidrailtransitinBerlinwasrelativelymorepop‐

ularamongwealthypeopleinthepastsincefareswhererelativelyhigherand,intheabsenceof

cars,rapidtransitwasthefastestmode.So,likely,averageincomeinthestudyareaincreasedata

ratelowerthancalibrated,implyingalikelydownwardbiasinourincomeelasticityestimate.Thus,

onbalance,webelievethat2.2isalower‐boundestimateoftheincomeelasticityofthenoisedis‐

amenityvalue.

5.1.2 Access

Accordingtoourestimates,thelandpricecapitalizationeffectofaone‐kilometerreductionindis‐

tancefromthenearestmetrostation(treatingsubwayandcommuterrailassubstitutes)declined

fromabout20.2%to15.5%.Becauseoftheincreaseintheshareoflandinthevalueofhousingthis

decreaseinthelandpricecapitalizationeffectcorrespondstoanincreaseinthehousepricecapi‐

talizationeffectfrom3.6%to5.0%.Thisiswithintherangeofrecentdifference‐in‐differenceesti‐

matessuchasbyGibbons&Machin(2005),whoreporta1.5%to5%range,orDubéetal.(2013),

whoseestimates implyaper‐kilometereffectof7%.The implied incomeelasticityof theaccess

amenityvalueis1.4.Becausethedistance‐from‐stationcapitalizationeffectcapturesthevalueof

theassociatedwalkingtime(Gibbons&Machin2005),theincomeelasticityofthevalueofstation

accessshouldgeneralizetothevalueoftime.Itisthereforenotablethatourestimatesaresignifi‐

cantlylargerthanthecross‐sectionalestimatesoftheincomeelasticityoftraveltimevalueinthe

literature,whichtendtobebelowunity(Börjessonetal.2012reportacentralestimateof0.6‐0.7).

Oneconcernregardingthecomparabilityofthehistoricandcontemporaryestimatesisthatrail

transitwasrelativelymorevaluableinthepastsincemass‐producedcarswerenotyetavailableas

affordablesubstitutes.Atthesametime,themetrorailnetworkhasexpandedsubstantiallyover

time,nowofferingconnectionstoagreatervarietyoflocations,whichshouldincreaseitsvalue.In

anetworkanalysis,wefindthatthetwooffsettingeffectsarelikelyofcomparablemagnitude.The

effectsofsortingwithrespecttotheaccessamenitygo,again,bothways.Preference‐basedsorting

overacenturymakesitmorelikelythathouseholdswithlargepreferencesforrailtransitlocate

closetostations.However,theexpansionofthenetworkmakesitmorelikelythatthemarginal

buyerinawell‐connectedareatodayhasarelativelylowerpreferenceforrailaccessthaninthe

past.Giventhatincomesortinglikelyleadstoususinganexaggeratedvalueforincomegrowthnear

metrostations,wetentativelyconcludethat1.4isalower‐boundestimateoftheincomeelasticity

oftherailaccessamenityvalue.

Easevs.noise 27

5.2 Fiscalcaseforundergroundmetrolines

Buildinganundergroundlineissignificantlymoreexpensivethanbuildinganelevatedline.Under‐

groundlines,conversely,avoidsizabledisamenities.Inthissection,weprovidesomesimpleback‐

of‐the‐envelopecalculationstoevaluatehowlongittakestorefinancetheextracostsviaproperty

taxrevenues.Tothisend,weestimatetheextracostofahypotheticalundergroundLineA,theextra

propertyvaluegeneratedinthiscounterfactual,andtheassociatedextrataxrevenues.

5.2.1 Extracost

Bousset(1935)reportstheper‐kilometerconstructioncostsfor31segmentsoftheBerlinmetro

railnetworkopenedby1930,includingper‐kilometercostofabouttwomillionReichsmark(RM)

forafive‐kilometerslongsubsegmentoftheelevatedpartofLineA.Multiplyingtheper‐kilometer

costbythetotallengthoftheelevatedsectionofeightkilometersyieldsconstructioncostsofabout

16millionRM.Toapproximatetheextracostassociatedwithahypotheticalundergroundsection,

we run an auxiliary regression of the natural log of per‐kilometer construction costs against a

dummyindicatingundergroundsections,controllingfortrackwidthandperiod(fiveyears)effects.

Theresults,reportedinSection8intheappendix,indicatethatbuildinganundergroundsectionin

theearly20thcenturyinBerlinwasaboutthreetimesasexpensiveasbuildinganelevatedsection.

MultiplyingtheestimatedconstructioncostofLineAbythisfactoryieldsacounterfactualconstruc‐

tioncostofabout50millionRMandanextracostfortheundergroundlineofabout34millionRM.

It isnoteworthythatthecurrentruleofthumbsuggestscostsofanundergroundlineareabout

twicethecostofanelevatedline(Flyvbjergetal.2008).So,theextracostfortheconstructionof

undergroundlineshavedeclinedovertime.

5.2.2 Extrapropertyvalue

Tocomparetheextracostofconstructiontotheaggregatedeffectonpropertyvalues,weaggregate

theplot‐levellandpriceobservationstoa50×50‐metergrid,whichallowsforrichspatialvariation

inrailnoiseand,atthesametime,ensuresthatwecovertheentirebuilt‐uparea.Undertheas‐

sumptionsmade in section 2.6, the noise‐induced change in property value in each grid cell is

ln / , where is noise level attributable to Line A and ln /

1 lnΩ / istherelativehousepricecapitalizationeffectofaone‐decibelincreaseinnoise.

SincetheCobb‐Douglashousingproductionfunctionimpliesthat 1/ 1 Ω ,wecanex‐

presstheimpactonpropertyvalueasafunctionoftheestimatedhousepricecapitalizationeffects

andtheaggregatelandvalue:

Easevs.noise 28

11

Ωln

, (8)

Intuitively,inequation(8),weholdthecapitalstockconstantsuchthatthevalueoftheproperty

increasesduetoanincreaseinthevalueoftheunderlyingland,exclusively.Thisway,weonlyac‐

countfortheincidenceontheimmobilefactor,i.e.weavoidtheproblemthatapolicy‐inducedin‐

creaseinthequantityofhousingstockatonelocationdisplacesdemandinotherareas.Theresult‐

inglandpriceeffectsbygridcellare illustratedintheappendix(section9). Inthiscontext, it is

worthemphasizingthatourplots includeall typesof landuses; theaggregate landvalueeffect,

therefore,reflectsbothchangesinutilityandproductivity.

Table3providesacomparisonoftheextracostforanundergroundvariantofLineAandtheag‐

gregatedimpactonbuildingvaluesthatwouldresultfromtheassociatednoisereduction.Wepro‐

videthecomparisonfortheactualhistoricalscenario(usingourhistoricallandpricecapitalization

estimates)andacounterfactualscenarioinwhichweapplythecontemporaryestimateoftheland

pricecapitalizationeffect .Thiscounterfactuallandpricecapitalizationeffectinflatesthees‐

timatedcontemporarylandpricecapitalizationeffect bytheratiosofthecontemporaryover

thehistoricalland(1 )andhousingexpenditure( )sharestoreflectthatthesamewillingness

topaywith lowershareparameters impliesa largerpercentage landpricecapitalizationeffect:

.

Basedonourhistoricalnoiseestimates,theaggregateincreaseinpropertyvaluesinacounterfac‐

tualscenariowithanundergroundLineAamountstoslightlymorethanonehalfoftheextracost

ofgoingunderground(18.6millionRM).Itisimportanttonotethattheseresultsdonotrejecta

welfarecaseforanundergroundLineAsincepositivehealthbenefitsarelikelyimportant,butun‐

likelytofullycapitalizeintopropertypricesduetolackofpublicawareness(Navrud,2002).Also,

anundergroundlinerelativetoanelevatedlinegenerateswiderbenefitstootherthanlocalresi‐

dentsandfirms(e.g.,tovisitorsandtourists).Yet,applyingthecounterfactualcontemporaryland

pricecapitalizationeffect,thegeneratedpropertyvaluealonealreadymorethanoffsetstheextra

costsofgoingunderground.Intheory,locallandlordswouldbeabletobeartheextracostforan

undergroundlinewithoutmakinglosses.

5.2.3 Extrataxrevenues

Whilelandvaluecaptureschemesareoftendifficulttoimplementinpractice,theincreaseinthe

property tax basemechanically generates revenues and, therefore,may be a less controversial

Easevs.noise 29

meansofrefinancinginthelong‐run.InGermany,thepropertytaxisdeterminedastheproductof

the tax base (the assessed value of the property, the so calledEinheitswert), a tax rate (Grund‐

steuermesszahl)andataxfactor(Hebesatz).SincetheEinheitswertisfixedatahistoricvalue,prop‐

ertytaxrevenuesareinsensitivetochangesinlocational(dis)amenities.However,propertytrans‐

actiontaxesrespondimmediatelyastheyareleviedonactualtransactionprices.Toapproximate

theyearlytaxrevenuesresultingfromnoise‐inducedchangesinpropertyvalue,weconsiderthe

6%propertytransactiontaxratecurrentlyapplicableinBerlinaswellasahistoric(pre‐1998)rate

of3.5%.Moreover,weconsider5%and10%probabilitiesofanypropertybeingtransactedina

givenyearsinceempiricalevidencepointstoaverageholdingperiodsbetween10and20years

(Collettetal.,2000:Fisheretal,2004).Inappendixsection11,wediscusstheGermanpropertytax

environment ingreaterdetailandshowthat inmoreconventionalproperty taxsettingssimilar

fiscalrevenueswouldbegenerated.

Inafurthersetofauxiliaryregressionsofthenaturallogoflandpriceonlocationfixedeffectsand

ayeartrend,wefindthatannuallandpriceappreciationratestendedtofluctuatearound5%in

Berlinfromthelate19thcenturytotheearly21stcentury,whichisclosetothemeaninterestrate

acrossyearsinthesameperiod.Moreover,thereisapositivecorrelationbetweenthetwovariables

(seesection10intheappendix).Thus,itseemsreasonabletomakethesimplifyingassumptionthat

inthelong‐runlandpricesgrowataratethatequatestotheopportunitycostofcapital.

Tab.3.Thefiscalcaseforanundergroundline

(1) (2) (3) (4) (5) (6) (7) (8)Noisepreference Historic ContemporaryRailnoisecapitalizationeffectonhsoueprices 0.41% 0.41% 0.41% 0.41% 3.32% 3.32% 3.32% 3.32%Estimatedtotalcost(million1900RM) 15.94Estimatedundergroundextracost(1900RM) 34.36Aggregatednoiseeffectbuildingvalue(millionRM) 18.6 18.6 18.6 18.6 151 151 151 151Transactiontaxrate 0.04 0.04 0.06 0.06 0.04 0.04 0.06 0.06Transactionprobability 0.05 0.10 0.05 0.10 0.05 0.10 0.05 0.10Yearlytaxrevenue(million1900RM) 0.03 0.07 0.06 0.11 0.26 0.53 0.45 0.91Yearstorecoverundergroundextracosts 1056 528 616 308 130 65 76 38

Notes: Contemporarylandpriceeffectadjustedforchangesinlandshareandhousingexpenditureshare(landpricecapitalizationeffectinflatedbytheratioofcontemporaryoverhistoricshares).CostestimatesbasedonBous‐set(1935).Estimatedtotalcostresultfrommultiplyingthereported1902perkmcostsofoverelevatedsec‐tionsby8km(thelengthoftheelevatedsectionsoftheLineA).Theestimatedundergroundextracostresultmultiplyingthetotalcostbythepercentageextracostsforundergroundsegmentsobtainedfromanauxiliaryregressionreported inSection5of theappendix.Yearstorecoverextracostsarecalculatedundertheas‐sumptionthatpropertyvaluesgrowataratesimilartocostofcapital(seeappendix9forajustification).

Undertheassumptionsmade,itturnsoutthatbasedonourestimatesofthehistoricallandprice

capitalizationeffects,itwouldhavetakenhundredsofyearstorecovertheextracostsviaproperty

taxes.Therefore,itisperhapsnosurprisethatLineAwasbuiltasanelevatedlineandthatittook

Easevs.noise 30

majorprotestsandpoliticalpressuretoforcethelineundergroundwithintheboundariesofChar‐

lottenburg.Incontrast,underthecounterfactualcontemporarycapitalizationeffect,tax‐revenues,

dependingontheassumedtaxrateandtransactionprobability,wouldhaverefinancedtheextra

cost foranunderground linewithin38to130yearsand, thus, likelywithin thepast lifetimeof

LineA.

6 Conclusions

Weusedifference‐in‐differencesandspatialdifferencesdesignstoestimatethelandpricecapitali‐

zation effects of the contemporarymetro rail network in Berlin andGermany’s first electrified

metrorailline,LineA,whichopenedmorethanacenturyago.Wefindthatthelandprice(implied

houseprice) capitalizationeffectof a10‐decibel reduction in railnoise increased from4.2% to

13.0%(1%to4%).Theeffectofaone‐kilometerreduction indistance fromtheneareststation

decreased(increased)from20.2%to15.5%(3.6%to5.0%.).Fromtheseestimates,weinfernovel

estimatesofthelong‐runincomeelasticitiesofthevalueofnoisereductionandtransportaccessof

2.2and1.4.Whilesignificantuncertaintysurroundstheseelasticityestimates,weviewthemas

likelylower‐boundestimates.Thus,ourtentativeconclusionisthatthelong‐runincomeelasticities

oftransport(dis)amenityvalueslikelyexceedtheirshort‐runcounterpartswhichhavebeenesti‐

matedatbelow‐unityvalues.

Thisfindinghasimportantimplicationsfortransportinfrastructureappraisalsasitsuggeststhat

timeandenvironmentalqualityareluxurygoodswhosevalueswilllikelyincreaseinabsoluteand

relativetermsasincomesrise.Whiletheexistingbelow‐unitycross‐sectionalincomeelasticityes‐

timatesarecertainlyrelevantfortheassessmentofthedistributionalconsequencesofinvestments

within generations, larger values may be required for the assessment of distributional conse‐

quencesacrossgenerations.Aswedemonstrate,usingBerlin’sLineAasacaseinpoint,thewelfare

caseforconstructingundergroundraillinesismuchstrongertodaythanacenturyagobecausethe

valueofaquietenvironmenthasincreasedmorethanproportionatelytoincome.Inanticipationof

likelyincreasesinrealincomes,infrastructureappraisalsthatseektofullycapturenet‐benefitsto

futuregenerations,shouldinflateratherthandeflatecontemporary(dis)amenityvalues.

References

Abadie, A.&Gardeazabal, J., 2003. TheEconomicCosts of Conflict: A Case Study of theBasqueCountry. American Economic Review, 93(1), pp.113–132. Available at:http://www.aeaweb.org/articles?id=10.1257/000282803321455188.

Ahlfeldt,G.M.,Redding,S.J.,Sturm,D.M.&Wolf,N.,2015.TheEconomicsofDensity :EvidencefromtheBerlinWall.Econometrica,83(4).

Easevs.noise 31

Ahlfeldt, G.M., 2018.Weights to AddressNon‐parallel Trends in Panel Difference‐in‐differencesModels. CESifo Economic Studies, 64(2), pp.216–240. Available at:http://dx.doi.org/10.1093/cesifo/ify013.

Ahlfeldt,G.M.&Maennig,W.,2015.Homevotersvs.leasevoters:Aspatialanalysisofairporteffects.JournalofUrbanEconomics,87.

Ahlfeldt,G.M.&McMillen,D.P.,2018.Tallbuildingsandlandvalues:HeightandconstructioncostelasticitiesinChicago,1870‐2010.ReviewofEconomicsandStatistics,100(5).pp.861‐875.

Ahlfeldt, G.M., Moeller, K. & Wendland, N., 2015. Chicken or egg? The PVAR econometrics oftransportation.JournalofEconomicGeography,15(6).

Ampel,F.J.&Uzzle,T.,1993.Thehistoryofaudioandsoundmanagement.AESconventionpaper,94(3598),pp.1–12.Availableat:http://www.aes.org/e‐lib/browse.cfm?elib=6566.

Anderson,M.L.,2014.Subways,Strikes,andSlowdowns:TheImpactsofPublicTransitonTrafficCongestion. American Economic Review, 104(9), pp.2763–2796. Available at:http://www.aeaweb.org/articles?id=10.1257/aer.104.9.2763.

Ashenfelter,O.&Card,D.,1985.UsingtheLongitudinalStructureofEarningstoEstimatetheEffectofTrainingPrograms.TheReviewofEconomicsandStatistics,67(4),pp.648–660.Availableat:http://www.jstor.org/stable/1924810.

Baltzer, F., 1897. Die elektrische Stadtbahn in Berlin von Siemens & Halske. Zeitschrift fürKleinbahnen.

Bernet, C., 2004. TheHobrecht Plan (1862) and Berlin’s urban structure.UrbanHistory, 31(3),pp.400–419.

Berry,B.J.L.,1976.GhettoExpansionandSingle‐FamilyHousingPrices.JournalofUrbanEconomics,3(4), pp.397–423. Available at:http://search.ebscohost.com/login.aspx?direct=true&db=buh&AN=6039560&site=ehost‐live.

Bertrand,M.,Duflo,E.&Mullainathan,S.,2004.Howmuchshouldwetrustdifference‐in‐differenceestimates?TheQuarterlyJournalofEconomics,119(1),pp.249–275.

Boes,S.&Nüesch,S.,2011.Quasi‐experimentalevidenceontheeffectofaircraftnoiseonapartmentrents.JournalofUrbanEconomics,69(2),pp.196–204.

Bohle‐Heintzenberg, S., 1980. Architektur der Berliner Hoch‐ und Untergrundbahn. Planungen,Entwürfe,Bautenbis1930.,Berlin:Arenhövel.

Bolt,J.&vanZanden,J.L.,2014.TheMaddisonProject:collaborativeresearchonhistoricalnationalaccounts. The Economic History Review, 67(3), pp.627–651. Available at:http://dx.doi.org/10.1111/1468‐0289.12032.

Börjesson,M.,Fosgerau,M.&Algers,S.,2012.Ontheincomeelasticityofthevalueoftraveltime.Transportation Research Part A: Policy and Practice, 46(2), pp.368–377. Available at:file://www.sciencedirect.com/science/article/pii/S0965856411001613.

Bousset,E.H.J.,1935.DieBerlinerU‐Bahn.,Berlin.

Brown,S.R.,2014.Loudandclear:NewYorkerssaynoiseistheirtopcomplaintaboutcitylife.NewYorkDailyNews,January19.

Brueckner, J.K., Thisse, J.‐F.& Zenou, Y., 1999.Why is central Paris rich anddowntownDetroitpoor?:Anamenity‐basedtheory.EuropeanEconomicReview,43(1),pp.91–107.Availableat:http://www.sciencedirect.com/science/article/pii/S0014292198000191.

Cellini,S.R.,Ferreira,F.&Rothstein,J.,2010.TheValueofSchoolFacilityInvestments:EvidencefromaDynamicRegressionDiscontinuityDesign.TheQuarterlyJournalofEconomics,125(1),pp.215–261.Availableat:http://qje.oxfordjournals.org/content/125/1/215.abstract.

Easevs.noise 32

Chay,K.Y.&Greenstone,M.,2005.DoesAirQualityMatter?Evidence fromtheHousingMarket.Journal of Political Economy, 113(2), pp.376–424. Available at:http://www.journals.uchicago.edu/doi/10.1086/427462.

Clausen,U.,Doll,C.,Franklin,F.J.,Franklin,G.V.,Heinrichmeyer,H.,Kochsiek,J.,Rothengatter,W.&Sieber,N.,2012.ReducingrailwaynoisepollutionP.D.B.S.andC.Policies,ed.Availableat:http://www.europarl.europa.eu/studies.

Collet,D,Lizieri,C.,Ward,C.,2000.Timingand theHoldingPeriodsof InstitutionalRealEstate.WorkingPaper.

Coffman, C. & Gregson,M., 1998. RailroadDevelopment and Land Value. JournalofRealEstateFinance&Economics,16(2),pp.191–204.Availableat:http://10.0.3.255/A:1007707801970http://search.ebscohost.com/login.aspx?direct=true&db=bth&AN=18766839&site=ehost‐live.

Currie,J.,Davis,L.,Greenstone,M.&Reed,W.,2015.Environmentalhealthrisksandhousingvalues:Evidencefrom1,600toxicplantopeningsandclosings.AmericanEconomicReview,105(2),pp.678–709.

Davis, L.W., 2004. The effect of health risk on housing values: Evidence from a cancer cluster.AmericanEconomicReview,94(5),pp.1693–1704.

Domke,P.&Hoeft,M.,1998.Tunnel,Gräben,Viadukte:100JahreBaugeschichtederBerlinerU‐Bahn,Berlin:KulturbildVerlag.

Dubé,J.,Thériault,M.&DesRosiers,F.,2013.Commuterrailaccessibilityandhousevalues:ThecaseoftheMontrealSouthShore,Canada,1992–2009.TransportationResearchPartA:Policyand Practice, 54, pp.49–66. Available at:http://www.sciencedirect.com/science/article/pii/S0965856413001377.

Environment,S.D.forU.D.andthe,2013.StrategicNoiseMapsS.D.forU.D.andtheEnvironment,ed.

Epple, D., Gordon, B. & Sieg, H., 2010. American Economic Association A New Approach toEstimatingtheProductionFunctionforHousing.AmericanEconomicReview,100(3),pp.905–924.

Fama,E.F.,1970.EfficientCapitalMarkets:AReviewofTheoryandEmpiricalWork.TheJournalofFinance,25(2),pp.383–417.

Fisher,J.,Gatzlaff,D.,Geltner,D.,Haurin,D.,2004.AnAnalysisoftheDeterminantsofTransactionFrequency of Institutional Commercial Real Estate Investment Property. Real EstateEconomics,32(2),pp.239‐264.

Flyvbjerg,B.,Bruzelius,N.&vanWee,B.,2008.ComparisonofCapitalCostsperRoute‐KilometreinUrbanRail.EuropeanJournalofTransportandInfrastructureResearch,8(1),pp.17–30.

Garrioch,D., 2003. Soundsof the city: the soundscapeof earlymodernEuropean towns.UrbanHistory,30(1),pp.5–25.

Gibbons,S.,2015.Gonewiththewind:Valuingthevisualimpactsofwindturbinesthroughhouseprices. JournalofEnvironmentalEconomicsandManagement, 72,pp.177–196.Availableat:https://www.sciencedirect.com/science/article/abs/pii/S0095069615000418 [AccessedDecember30,2018].

Gibbons,S.&Machin,S.,2005.Valuingrailaccessusingtransport innovations. JournalofUrbanEconomics,57(1),pp.148–169.

Glaeser,E.L.,Kolko,J.&Saiz,A.,2001.Consumercity.JournalofEconomicGeography,1(1),pp.27–50.Availableat:http://joeg.oxfordjournals.org/content/1/1/27.abstract.

Graevenitz, K. (2018). The amenity cost of road noise. JournalofEnvironmentalEconomicsandManagement,90.1‐22.

Easevs.noise 33

Greenstone,M.&Gallagher, J.,2008.DoesHazardousWasteMatter?EvidencefromtheHousingMarketandtheSuperfundProgram*.QuarterlyJournalofEconomics,123(3),pp.951–1003.Available at: https://academic.oup.com/qje/article‐lookup/doi/10.1162/qjec.2008.123.3.951.

Gwilliam,K.M.,1997.TheValueofTimeinEconomicEvaluationofTransportProjects,LessonsfromRecent Research, Washington D.C.: World Bank. Available at:http://www.worldbank.org/html/fpd/transport/publicat/td‐ot5.htm.

Hahn,J.,Todd,P.&VanderKlaauw,W.,2001.IdentificationandEstimationofTreatmentEffectswith a Regression‐Discontinuity Design. Econometrica, 69(1), pp.201–209. Available at:http://dx.doi.org/10.1111/1468‐0262.00183.

Hainmueller,J.,2012.EntropyBalancingforCausalEffects:AMultivariateReweightingMethodtoProduce Balanced Samples in Observational Studies. Political Analysis, 20(1), pp.25–46.Availableat:https://www.cambridge.org/core/article/entropy‐balancing‐for‐causal‐effects‐a‐multivariate‐reweighting‐method‐to‐produce‐balanced‐samples‐in‐observational‐studies/220E4FC838066552B53128E647E4FAA7.

Hämer, H.‐W., 1990. Behutsame Stadterneuerung. In Senatsverwaltung für Bau‐ undWohnungswesen, ed. Stadterneuerung Berlin. Berlin: Senatsverwaltung für Bau‐ undWohnungswesen.

Hanna, B.G., 2007. House values, incomes, and industrial pollution. Journal of EnvironmentalEconomics and Management, 54(1), pp.100–112. Available at:https://www.sciencedirect.com/science/article/abs/pii/S0095069607000204 [AccessedDecember30,2018].

Hernán,M.A.,Brumback,B.&Robins,J.M.,2001.MarginalStructuralModelstoEstimatetheJointCausalEffectofNonrandomizedTreatments.JournaloftheAmericanStatisticalAssociation,96(454),pp.440–448.Availableat:https://doi.org/10.1198/016214501753168154.

Hilber,C.A.L.&Vermeulen,W.,2015.TheImpactofSupplyConstraintsonHousePricesinEngland.TheEconomicJournal,p.n/a‐n/a.Availableat:http://dx.doi.org/10.1111/ecoj.12213.

Kabak,B.,2015.NYCCan’tAffordtoBuildtheSecondAvenueSubway,andItCan’tAffordNotTo.CITYLAB,July17,2.

Kau,J.B.&Sirmans,C.F.,1979.UrbanLandValueFunctionsandthePriceElasticityofDemandforHousing. Journal of Urban Economics, 6(1), p.112. Available at:http://search.ebscohost.com/login.aspx?direct=true&db=buh&AN=7182643&site=ehost‐live.

Knoll,K.,Schularick,M.&Steger,T.,2017.NoPriceLikeHome:GlobalHousePrices,1870‐2012.American Economic Review, 107(2), pp.331–353. Available at:http://www.aeaweb.org/articles?id=10.1257/aer.20150501.

Kuminoff,N.V.,Pope,J.C.2014.Do“capitalizationeffects”forpublicgoodsrevealthepublic'swill‐ingnesstopay?InternationalEconomicReview,55(4),p.1227‐1250.

Kurzweil,L.G.,1979.Ground‐bornenoiseandvibrationfromundergroundrailsystems.JournalofSound and Vibration, 66(3), pp.363–370. Available at:http://www.sciencedirect.com/science/article/pii/0022460X79908538.

Lamartina,S.&Zaghini,A.,2011.IncreasingPublicExpenditure:Wagner’sLawinOECDCountries.GermanEconomicReview,12(2),pp.149–164.Availableat:http://dx.doi.org/10.1111/j.1468‐0475.2010.00517.x.

Lemke,U.&Poppel,U.,1996.BerlinerU‐Bahn,Munich:AlbaPublikation.

Linden,L.&Rockoff, J.E.,2008.Estimatesof the ImpactofCrimeRiskonPropertyValues fromMegan’s Laws. American Economic Review, 98(3), pp.1103–1127. Available at:http://www.aeaweb.org/articles.php?doi=10.1257/aer.98.3.1103.

Easevs.noise 34

McDonald, J.F. & Bowman, H.W., 1979. Land value functions: A reevaluation. Journal of UrbanEconomics, 6(1), pp.25–41. Available at:http://www.sciencedirect.com/science/article/B6WMG‐4DBC5R7‐2N/2/7ac70204465541d0b553ea4c219ef8dd.

McMillen, D.P., 1996. One Hundred Fifty Years of Land Values in Chicago: A NonparametricApproach. Journal of Urban Economics, 40(1), pp.100–124. Available at:http://www.sciencedirect.com/science/article/B6WMG‐45MGSV8‐N/1/b659c6c08818d12b728f413bea6beded.

McMillen,D.P.&McDonald,J.F.,2002.LandValuesinaNewlyZonedCity.TheReviewofEconomicsandStatistics,84(1),pp.62–72.

Melke, J., 1988.Noiseandvibration fromundergroundrailway lines:Proposals forapredictionprocedure. Journal of Sound and Vibration, 120(2), pp.391–406. Available at:http://www.sciencedirect.com/science/article/pii/0022460X88904518.

Mills, E.S., 1969. The value of urban land. In H. Perloff, ed. The quality of urban environment.Baltimore,MA:ResourcesfortheFuture,Inc.

deMoncan,P.,2009.LeParisd’Haussmann,Paris:Mécène.

Murphy,E.&King,E.,2014.EnvironmentalNoisePollution:NoiseMapping,PublicHealth,andPolicy,Amsterdam:Elsevier.

Navrud,S.,2002.TheState‐Of‐The‐ArtonEconomicValuationofNoise,FinalReporttoEuropeanCommissionDGEnvironment.

Neitzel,R.,Gershon,R.R.M.,Zeltser,M.,Canton,A.&Akram,M.,2009.NoiseLevelsAssociatedWithNewYorkCity’sMassTransit Systems.American JournalofPublicHealth, 99(8), pp.1393–1399.

Nelson,A.C.,1992.EffectsofElevatedHeavy‐RailTransitStationsonHousePriceswithRespecttoNeighborhoodIncome.TransportationResearchRecord,1359,pp.127–132.

Nelson, J.P., 2004. Meta‐Analysis of Airport Noise and Hedonic Property Values: Problems andProspects. Journal of Transport Economics and Policy, 38(1), pp.1–27. Available at:http://www.jstor.org/stable/20173043.

Nelson,JonP.(2008).HedonicMethodsinHousingMarkets,ChapterHedonicPropertyValueStud‐iesofTransportationNoise:AircraftandRoadTraffic.SpringerVerlag.

Oates,W.E.,1969.TheEffectsofPropertyTaxesandLocalPublicSpendingonPropertyValues:AnEmpiricalStudyofTaxCapitalizationandtheTieboutHypothesis.JournalofPoliticalEconomy,77(6),pp.957–971.Availableat:http://www.jstor.org/stable/1837209.

Parry,I.W.H.&Small,K.A.,2009.ShouldUrbanTransitSubsidiesBeReduced?AmericanEconomicReview, 99(3), pp.700–724. Available at:http://www.aeaweb.org/articles?id=10.1257/aer.99.3.700.

Passchier‐Vermeer,W.&Passchier,W.F.,2000.NoiseExposureandPublicHealth.EnvironmentalHealthPerspectives,108,pp.123–131.Availableat:http://www.jstor.org/stable/3454637.

Ram,R.,1987.Wagner’sHypothesisinTime‐SeriesandCross‐SectionPerspectives:Evidencefrom“Real” Data for 115 Countries.TheReview ofEconomicsand Statistics, 69(2), pp.194–204.Availableat:http://www.jstor.org/stable/1927226.

Rosen,S.,1974.HedonicPricesandImplicitMarkets:ProductDifferentiationinPureCompetition.Journal of Political Economy, 82(1), pp.34–55. Available at:http://dx.doi.org/10.1086/260169.

Rosenbaum, P.R.&Rubin,D.B., 1983. The central role of the propensity score in observationalstudies for causal effects. Biometrika, 70(1), pp.41–55. Available at:http://biomet.oxfordjournals.org/content/70/1/41.abstract.

Easevs.noise 35

Senatsverwaltung für Stadtentwicklung Berlin, 2000. Digitale Schwarzpläne. Available at:http://www.stadtentwicklung.berlin.de/planen/stadtmodelle/de/innenstadtplaene/sp/index_sp‐vt3.shtml.

SenatsverwaltungfürStadtentwicklungBerlin,2006.UrbanandEnvironmentalInformationSystem,Berlin.

Silverman,B.W.,1986.DensityEstimationForStatisticsandDataAnalysis.MonographsonStatisticsandAppliedProbability.

Tanaka,S.&Zabel,J.,2018.Valuingnuclearenergyrisk:EvidencefromtheimpactoftheFukushimacrisisonU.S.houseprices.JournalofEnvironmentalEconomicsandManagement,88,pp.411–426. Available at:https://www.sciencedirect.com/science/article/abs/pii/S0095069617301626 [AccessedDecember30,2018].

U.S.DepartmentofLabor,2006.100YearsofU.S.ConsumerSpending,Report991.

Wagner,A.,1890.Finanzwissenschaft,Leipzig:Winter,C.F.

Walls,M., Gerarden, T., Palmer, K. & Bak, X.F., 2017. Is energy efficiency capitalized into homeprices?EvidencefromthreeU.S.cities.JournalofEnvironmentalEconomicsandManagement,82, pp.104–124. Available at:https://www.sciencedirect.com/science/article/abs/pii/S0095069616304508 [AccessedDecember30,2018].

Wardman,M.,Bristow,A.&Arsenio,E.,2005.ApplyingStatedPreferenceMethodstotheValuationofNoise:SomeLessonstoDate.Conferencepaper:The2005CongressandExpositiononNoiseControlEngineering.

WorldHealthOrganization,2009.Nightnoiseguidelines forEurope.Copenhagen:WHORegionalOfficeforEurope.

Yeates,M.H.,1965.SomeFactorsAffectingtheSpatialDistributionofChicagoLandValues,1910‐1960. Economic Geography, 41(1), pp.57–70. Available at:http://www.jstor.org/stable/141856.

ISSN: 1438-2733

Documents

Darmstadt Discussion Papers in ECONOMICS