©CopyrightJASSS

YanMa,ZhenjiangSHENandMitsuhikoKawakami(2013)

Agent-BasedSimulationofResidentialPromotingPolicyEffectsonDowntownRevitalization

JournalofArtificialSocietiesandSocialSimulation 16(2)2<http://jasss.soc.surrey.ac.uk/16/2/2.html>

Received:12-Nov-2011Accepted:13-Oct-2012Published:31-Mar-2013

Abstract

Inrecentdecades,compactcitieshavebecomeanewconcerninurbanplanninginmostJapanesecities.ThemainreasonforthistrendamongJapanesecitiesisthephenomenonofde-urbanizationanddowntowndeclinethatgraduallyoccurredafterthe1990s.Assuch,atpresent,therearedispersed,small,built-upportionsofsuburbanareasthathaveresultedinhouseholdmobilityoutsidethedowntown.Therefore,somelocalgovernmentsinJapanareattemptingtorealizecompactcitiesthroughpolicyintervention,suchasencouraginghouseholdstorelocatefromsuburbantodowntownareasinordertoaddressthepopulationdeclineinurbanareas.Recently,onesuchresidentialpolicyhavebeenpromotedbyJapaneselocalcitygovernments.Byofferingalocalhousingallowance,thispolicyencourageshouseholdstorelocatetodowntownareas.Wedevelopedanagent-basedhouseholdresidentialrelocationmodel(HRRM)tovisualizetheeffectofthisresidentialpolicy,thatis,thelocalhousingallowance.TheHRRMisbuiltonhouseholdsÄôadaptivebehavioursandinteractionsthroughhousingrelocationchoicesandpolicyattitudes,andsoitcansimulatethediversifiedresidentialrelocationsofhouseholdsinvariouslifecyclestages.ThroughsimulationusingtheHRRM,theeffectivenessofthisresidentialpolicycanbevisualized,andtheHRRMwillhelplocalgovernmentstounderstandtheeffectsofresidentialpolicies.

Keywords:HouseholdRelocation,DowntownDecline,CompactCity,UrbanShrinkage,PolicyEffect

Introduction

1.1 Foranumberofdecades,urbanissuesrelatedtohousingmarketsandresidentialmobilitywereconcernedprimarilywithsuchtopicsasurbanizationandsprawlingsettlements(Haaseetal.2010;Antrop2004;Kazepov2005).Now,however,urbanshrinkageisahottopicamongurbanplanners(Rieniets2005;Rieniets2009).Asthepopulationdensitydecreases,householdsinalower-density,built-upcityrequiremoreprivatevehicles(Kaido2005).Thistrendappearsnottofollowthatdesiredbyurbanplanners—namely,toreducethenegativeenvironmentalimpactsassociatedwithcardependency(NewmanandKenworthy1989;Banister1997;Banisteretal.1997).Thus,methodsforrevitalizingdowntownareasarebeingconsideredbymanygovernmentsandurbanplanners.InmostWesternsocietiestoday,policyprescriptionhasincreasinglyfavouredacompactcityapproachinordertoaddresstheadverseeffectsofurbanshrinkage(Howleyetal.2009).Inthispaper,wewillintroducethehouseholdresidentialrelocationmodel(HRRM),whichisanagent-basedmodel(ABM)forsimulatingthehouseholdresidentialrelocationprocesseffectedbyaresidencepromotionpolicy.TheHRRMintegratestheadaptivebehavioursofhouseholdsintermsofresidentialrelocationswithpolicyinteractionstovisualizetheimpactofaresidencepromotionpolicyondowntownrevitalization.

1.2 Residentialrelocationactuallyisnotanewtopic;muchresearchalreadyexistsinthisfield.Fromabroaderperspective,researchersgenerallyattempttoaddresstheirconcernaboutresidentiallocationissuesbyanalyzingtherelationshipbetweenhouseholdresidentiallocationchoiceandtransportation.Statedpreferenceexperimentsinthisfieldfocusprimarilyondeterminingtherelationshipbetweentransportcharacteristicsandresidentiallocation(Kimetal.2005;MolinandTimmermans2003;RouwendalandMeijer2001).Thesestudiesarebasedprimarilyonstatisticsinsteadofonagent-basedsimulations.Inaddition,researchersalsohaveinvestigatedinteractiveprocessesbetweentransportationandlanduseinordertomodelthedistributionofpopulationsacrossspace.Asanexample,LandUseTransportInteraction(LUTI)wasfirstdevelopedasanaggregatedmodel(Timmermans2003),describingtheallocationofpopulationasanaggregatecategory.Gradually,LUTImodelswereimprovedtoagent-based(Benenson1998;Milleretal.2004;Waddelletal.2003;Ettemaetal.2006;Moeckeletal.2005),whichcandescribethelocationbehaviourofindividualhouseholds.AnotherrepresentativemodelisUrbanSim(Nothetal.2003),whichadoptsamicrosimulationapproachinwhichitrepresentsindividualagentswithinthesimulation.InUrbanSim,ahouseholdmobilitymodelispresentedtosimulatetherelocationofahouseholdclosertoemployment,whichisanevolutionaryprocessresultingfrominteractionsbetweendifferenturbanactors,landuse,andtransportation.Thus,LUTIandUrbanSimaresimilarinthattheyarebothintegratedframeworksforsimulatingtheinteractiveprocessoflanduse,transportation,andresidentialchoice.Theyshowastrongabilitytosimulatethespatialprocessbuttheirfocusontheinteractionsofadjacentagentsisrelativelyweak.

1.3 Becausethisresearchsimulateshouseholdbehaviourduringresidentialrelocationwithrespecttoaresidencepromotionpolicy,however,themodelhereneedsastrongcapacitytoreflectagents'policyinteractions.Thus,anagent-basedapproachismoresuitablethanmicrosimulations.Asreportedpreviously(Jager2007),anABMisexpectedtocontributetoexploringtheeffectivenessofpolicymeasuresincomplexenvironmentsthroughbehaviour-environmentinteractions.AnumberofstudieshaveusedanABMtoassessfuturethesocio-ecologicalconsequencesresultingfromland-usepolicies(Leeetal.2010),andotherstudieshavefocusedontheuseofmulti-agentsimulationforpolicydevelopment(Bergeretal.2006).Asanagent-basedsimulation,theresidencepromotionpolicywillbetakenasakeyfororganizingagentinteractionsintheHRRM.Unlikemicro-simulation,theHRRMemphasizesinteractionbetweenhouseholdsduringthedecisionprocessesofhouseholdagentswithrespecttoresidentiallocation.Duringsuchatime,householdagentsevolvestochasticallytoadaptthemselvestourbanspaceinresponsetothechangesoflifecyclestagesandtheresidencepromotionpolicy,whereastraditionalmicrosimulationtransitionprobabilitieslackevolutionaryandspatialdimensions.Thedecisionprocessofhouseholdresidentiallocationcanbesimplifiedintotwophases:theevaluationofthecurrentresidenceandtheselectionofanewone(Boyleetal.1998).Ahouseholdresidentiallocationchange,asdefinedbypreviousstudies,canbeclassifiedaseitheraninducedrelocationorasanadjustmentrelocation(Cadwaller1992;ClarkandOnaka1983).Aninducedrelocationislinkedtochangesinanindividual'slifecyclestage(KuluandMilewski2007;MulderandWagner1998).Thismeansthatindividualswhoenteranewlifecyclestagearethemostlikelytorelocate(Kulu2007).Ontheotherhand,adjustmentrelocationisrelatedtodissatisfactionwiththecurrentlocation(Kähriketal.2012),withthedecisiontorelocatedependingonthesatisfactionoftheresidentswiththeircurrentlocation.Whenthesatisfactionwithcurrenthousingisbelowacertainthreshold,individualswillstarttosearchforanalternativeplaceofresidence.BothinducedandadjustmentrelocationsareincludedintheHRRMasadaptivebehavioursintheresidentialrelocationprocess.

1.4 ABMsarewidelyusedforresidentialsimulationfromtheviewpointofeconomicswithrespecttorealestate,inwhichhousingpricesareanimportantfactor.Basedontheassumptionthatgentrificationoccursbecausecapitalflowsbacktotheinnercityandcreatesopportunitiesforresidentialrelocation,DiappiandBolchi(2006)presentedadynamicmodeldevelopedonaNetlogoplatformandusingamulti-agent/cellularautomatasystemapproach.Bydescribingtherelocationsofhouseholds,researchershave,tosomeextent,addressedthehousingmarketprocessesandpriceformation(Ettema2011;Dawn2008).Inparticular,Moeckelsimulatedtheprocessbywhichhouseholdsevaluateindividualdwellingsuntiltheyeventuallyfindandacceptadwellingthatoffersasignificantimprovementovertheircurrentdwelling(Moeckeletal.2003).ThisissimilartotheproposedHRRM,whereutilityisusedbyhouseholdstoevaluatedifferentlocations.However,unlikeinthestudybyMoeckel,intheproposedHRRM,utilitycomparisonisonlyonestepoftheentirerelocationprocessforadaptinghouseholdstotheurbanenvironment.Furthermore,thepresentstudydoesnotfocusontheurbansprawlprocessofresidentiallocationchoices.Rather,wefocusonhowtosimulatehouseholdresidentialrelocationchoicesasinfluencedbyalocalresidencepromotionpolicy;wefurtherconsiderhowthatpolicyinfluenceshouseholdresidentialrelocationandconsequentlydowntownrevitalizationduringurbandecline.

1.5 Withrespecttoourcontribution,thismodelcansimulatetheprocessofhouseholdresidentialrelocationinfluencedbyaresidencepromotionpolicyfordowntownrevitalization.Unlikeconventionalsimulationsofresidentialrelocation,theHRRMnotonlycansimulatehouseholds'adaptivebehavioursofresidentialrelocationsinapredefinedurbanspacebutalsocanreflecttheinfluencesofpolicyimplementationonhouseholds'relocationprocessthroughorganizingpolicyinteractionsamonghouseholdagents.Thesimulationresultcanbevisualized,andhouseholdsthathaverelocatedtodowntownareascanthenbeidentified.Inthefollowingsection,wedescribehowtheHRRMisconstructedandused.Theremainderofthispaperisorganizedasfollows:theHRRMwillbedescribedindetailinsection2.Section3illustratestheinitialconditionsandsimulationconfigurationforusingHRRM.Section4isatestofmodelsensitivityonchangesowingtobothhouseholdlifecyclestagesandpolicyeffects,andisalsoacomparisonbetweensimulatedresultsandtherealcitydataofatypicalJapanesecity.Finally,theresearchresultsarediscussedandconclusionspresentedinsection5.

ModelFormulation

StructureoftheHRRM

2.1 Regardingtheincentivesofresidentiallocations,someexistingresearchhasproposedthatthechangesinhouseholdlifecyclestagesleadtorelocationbehaviour(FontaineandRounsevell2009;Torrens2001).Thisconclusionisnotabsolutebecause'induced'and'adjustment'movesplaydifferentroles.IntheHRRM,weassumethatwhenthelifecyclestagechanges,ahouseholdagentmay

http://jasss.soc.surrey.ac.uk/16/2/2.html 1 14/10/2015

expressinterestinrelocation,althoughthefinaldecisionastowhethertorelocatedependsonthehousehold'ssatisfactionwiththecurrentlocation.Thus,inthepresentstudy,weseebothinducedrelocationandadjustmentrelocationasadaptivebehaviours.Inducedrelocationisassumedtobethebasisofadjustmentrelocationintheadaptivedecisionprocess.Inotherwords,wedevelopahouseholdlifecyclestagemoduleinwhichhouseholdagentsfirstidentifytheneedforadaptingthemselvestonewlifecyclestagesbeforetheydecidetoadaptthemselvestonewresidentiallocations.Thus,asshowninFig.1,theHRRMincludesthreemodules:ahouseholdlifecyclestagemodule,anevaluationmoduleforhouseholdrelocationdesire,andahouseholdrelocationmodule.

Figure1.UMLstatediagramoftheHRRM

Figure2.DiagramofthehouseholdlifecyclestageintheHRRM

Householdagent

2.2 TheHRRMisaspatiallyorientedagent-basedmodel,inwhichtherearehouseholdagentsandurbanspace.Householdagentsrepresentthehumanpopulation,andurbanspace(i.e.landunits)representsthespacewherepeoplelive.IntheHRRM,peoplecorrespondnottoindividualagentsinthevirtualcity,butrathertomembersofhouseholds.Ahouseholdisanagent,whichisacoherentunitofsimulationintheHRRM,thatcanmakedecisionsasasingleentity.Thissingleentityisassumedtobecomposedofafamilyconsistingofoneormorepeople.IntheHRRM,thehouseholdagenthassuchattributesasage,marriagestatus,members,deposit,income,andmeansoftransportation.

2.3 AllofthehouseholdagentsinthepresentsimulationaredesignedtofollowthelifecycleprocesspresentedinFig.2.Wedividedthetotallifecycleofahouseholdintosevenstagesinordertoclarifythepossibilitiesofrelocationineachlifecyclestage.AsshowninFig.2,inthefirststage'Independent',anindependenthouseholdiscreated.Afterseveralyearsofindependent,singlelife,theindividualofthishouseholdmeetssomeoneanddecidestogetmarried.Inthesecondstage'Married',thecouplefindsalargerhouse,andanewhouseholdisformed.Thoughsomehouseholdsdonotenterthesecondstage,anumberofhouseholdswillenterthethirdlifecyclestage.Inthethirdstage'Raisechildren',couplesfindthattheircurrenthousesaretoosmallortoofarawayfromlocalschools,sotheydecidetorelocateagain.Inthefourthstage'Childrenleavehouse',anewgenerationofhouseholdsiscreated,and,asshownbyadottedline,whenachildinahouseholdreaches18yearsofage,heorshewillfindanewresidenceandbeginthefirstlifecyclestageofanewhousehold.Duringthisprocess,theindividualswhoremainintheoldhouseholdcontinuetoageandeventuallyretire.Oneoftheindividualswilleventuallydie,andtheremainingindividualbecomessingleagain.Finally,theremainingindividualwilldieanddisappearfromthesimulationmodel.

Adaptivebehavioursofhouseholdagentsintheresidentialrelocationprocess

Decisionprocessforhouseholdrelocationdesire

2.4 IntheHRRM,weseethatwhenthecurrentlifecycleofahouseholdagentmovestothenextstage,theagentwilldecideonwhethertomoveornotinordertoadapttothenewlifestage.Thisprocesswedefinedashouseholddecision-makingonresidentialrelocationdesire.Forthisprocess,ahousehold'ssatisfactionwithitscurrentlocationanditsabilitytoaffordanewonearekey.AsshowninFig.3,adecisionprocessisdesignedsothathouseholdagentswillmakerelocationdecisionsduringeachlifecyclestagebasedontheageofthehousehold.Asshowninthisfigure,thehousehold'ssatisfactionwillfirstbeevaluatedinordertojudgewhetherhouseholdagentiissatisfiedwiththecurrentlocationorifrelocationisdesired.IftheresultsindicatethatsatisfactionwiththecurrentlocationSiisbelowasatisfactionthreshold,thenhouseholdiwillconsiderrelocation.Afterthesatisfactionestimation,householdagentipredictstheexpectedcostsofthedesiredrelocationandthencomparesthesecostswithcurrentsavings.Thehouseholdwillrelocateifthesecostscanbeborne.Otherwise,thehouseholdmustremaininitscurrentlocation.Accordingly,therelocationprocesswillbeimplementedbythehouseholdrelocationchoicemoduleintheHRRMinordertodecidewheretorelocate.

http://jasss.soc.surrey.ac.uk/16/2/2.html 2 14/10/2015

Figure3.Decisionprocessforhouseholdrelocationdesire

Decisionprocessforhouseholdrelocationchoice

2.5 Weassumethathouseholdsbelongingtothesameagegroupexhibitsimilarutilitypreferenceswithrespecttorelocation.Inthissection,weproposethathouseholdswhichwanttorelocatewillfollowthedecision-makingflowshowninFig.4.Unlikesomesimulationsthatfocusonresidentialsprawl(VegaandReynolds-Feighan2009;LiandMuller2007;BrownandRobinson2006),thepresentstudyattemptstorevealtheeffectivenessofaresidencepromotionpolicyfordowntownrevitalization.Thus,weconsidertherelocationprocessbetweendowntownandotherurbanareas.AsshowninFig.4,theurbanareasaredividedintothreedifferentregions.Foreachhouseholdagent,weassumethattheresidentialutilitiesprovidedbydifferenturbanareasaredifferent(thethreeregions),whereastheutilitiesarehomogenouswithinoneregion,witharandomrangefollowinganormaldistributionthatrepresentsindividualpreferences.Whenhouseholdsrelocate,theymustcomparethedifferentutilitiesofresidentiallocationsforadaptingthemselvestonewlocationsbasedonthenecessitiesoftheirnewlifestages.Thealternativelocationsarerandomlydistributedwithinthesethreeurbanareas—i.e.thecitycentrearea(CCA),theurbanpromotingarea(UPA),andtheurbancontrolarea(UCA).Basedonthiscomparison,householdswilleventuallychooseanareathatprovidesthegreatestutility.

http://jasss.soc.surrey.ac.uk/16/2/2.html 3 14/10/2015

Figure4.Decisionprocessofhouseholdrelocationchoices

Policyinteractionsbetweenhouseholdagentsintheresidentialrelocationprocess

2.6 Interactionsbetweenhouseholdagentsareconsideredtotakeplaceintheiradaptivebehavioursduringresidentialrelocationprocessinresponsetotheresidentialpromotionpolicy.Forrepresentingtheinteractionsinsimulation,theinteractionsbetweenhouseholdagentsaredesignedasonecomponentoftheutilitymodel.Althoughutilitymodelsarewidelyusedinconsideringresidentiallocations(Moeckeletal.2007),theutilitytheoryconventionallydoesnotreflecttheinfluencesofneighbours.Inthepresentstudy,theutilitymodeliscombinedwiththeagent-basedsimulationinordertoclarifytheinteractionsbetweenhouseholdagentsintheresidentialrelocationprocessduringdifferentlifecyclestagesofhouseholds.Inordertoclarifytheinteractiveinfluencesbetweenhouseholdagentsregardingtheresidencepromotionpolicy,interactionsbetweenhouseholdagentsaredesignedtooccurontwolevelsintheHRRM.

2.7 Weproposethathouseholdrelocationbehaviourswillbeinfluencedbythepolicyattitudesofhouseholdsfromtheentirecityandtheirneighbours,whicharedefinedashouseholdinteractionsattheglobalandneighbourhoodlevels—namely,globalinfluenceandneighbourhoodinfluence.Theglobalinfluencerepresentsthepolicyattitude,whichistheproportionofhouseholdsinacitythatacceptandplantousethelocalresidencepromotionpolicyforrelocation.Attheneighbourhoodlevel,neighbourhoodinfluencewillbeconsideredinordertorepresenttheeffectofneighboursonhouseholdresidentialrelocation.Itreflectsthedeliveringofinformationaboutthepolicybyhouseholdswithinasmallneighbourhood.Thedetailsofthesetwofactorsareexplainedbelow.

1. Neighbourhoodinfluence:theratioofneighboursthatusetheresidentialallowancepolicytorelocatetoadowntownareadividedbythenumberofneighboursthatdonotusethepolicyforrelocation.Here,neighbourhoodinfluencecanrepresentagentchoicesinfluencedbyneighbourswhoplantousethepolicy.Thisindicatorisaddedtotheutilitymodelasanextracomponentofutilityforreflectinginteractionsbetweenneighbours.TheneighbourhoodisdefinedasthenumberofhouseholdagentswithintheninecellsofMoorneighboursinthiswork.

2. Globalinfluence:ratioofthetotalnumberofhouseholdsthatusethepolicyforrelocationdividedbythetotalnumberofhouseholds.Thisfactorrepresentstheproportionofhouseholdsinthecitythatacceptthepolicyandplantousethepolicyforresidentialrelocationtoadowntownarea.Thisindicatorisalsodefinedasonecomponentofutility,whichisthepolicyimpactonindividualrelocationdecisionsfromtheentirecity.

Modelsforadaptivebehavioursofhouseholdagentsintheresidentialrelocationprocess

Householdsatisfactionwithcurrentlocation

2.8 Whenahouseholdagentconsidersrelocating,thefinaldecisiondependsonthesatisfactionlevelwiththecurrentlocation.However,anewhouseholdagentmayfindanewlocationwithoutevaluatingthesatisfactionwiththecurrentlocation.IntheHRRM,thesatisfactionofahouseholdwithitscurrentlocationcanbeevaluatedbasedontheattributesofthehouseholdagentsandspatialinformation—namely,theattributesofurbanspace.Eachcellinanurbanspacehasaseriesofpredefinedspatialattributes.Themathematicalmodelsusedintheevaluationofhouseholdsatisfactionwiththecurrentlocationareshowninthefollowingequations:

(1)

(2)

(3)

(4)

whereSiisthesatisfactionofhouseholdiwiththecurrentresidentiallocation,bugijisavectorofretrospectivecoefficientstovariablej,anduindicatesthestepofthelifecycleofhouseholdiinincomegroupg.Here,xijsisthesatisfactionofhouseholdiinlocationsproducedbyvariablejandhasfourlevels:1)extremelyunsatisfied,2)unsatisfied,3)satisfied,and4)verysatisfied.Ifonehouseholdagent'ssatisfactionwiththecurrentlocationislessthantheSthreshold,thisagentwillconsiderrelocating.However,thedecisiontorelocatewillbemadebasedontheutilityoftherelocationcandidates.

Utilityprovidedbyurbanspacetohouseholdagents

2.9 Householdsmakedecisionsonnewlocationsbasedontheutilityofthelocation.Theutilitymodelusedtodescribethesubjectivedifferencebetweentheagent'schoicesisgiveninEqs.5through7.InEq.5and6,VisistheutilityofhouseholdiprovidedbylocationswithouttheunobservedrandomcomponentandinteractionbetweenhouseholdagentsViinter.xijsisavectorofobservableexplanatoryvariablejdescribingtheattributesofhouseholdiinlocations.AsshowninEq.7,xijscanbedefinedintwoforms,oneofwhichistheevaluationofhouseholdiofthespatialattributionxjinlocations;anotheristhedistancebetweenhouseholdiinlocationsandpublicfacilityjinthenearestposition—forexample,schools,shops,andsoon.Forreflectingthedifferenceofxijsbetweenallhouseholdagentsiinlocations,arandomnumberαiisgenerated,whichfollowsnormaldistributionwithameanof0andastandarddeviationof0.1,asshowninEq.7.

2.10 Inthepresentstudy,weobservethedifferencesinresidenceutilitypreferencesofhouseholdsindifferentlifecyclestages.Forthispurpose,wesetbugijasshowninEq.8ascoefficientsoftheobservedcomponentsjtohouseholdiintheulifecyclestageofthegincomegroup.Utilitypreferences1through6areexplainedinTable1,inwhichβisarandomperturbationwithameanof0andastandarddeviationof0.1,generatedwithanormaldistributiontorepresentindividualpreferences.Here,weuseadecisionruleasshowninFig.5todeterminedifferentutilitypreferencesbetweenhouseholdagentsintheHRRM.

2.11 InEq.9,Viinterstandsfortheinteractionofhouseholdiwithotherhouseholdagentsinurbanspace,whichcanbedividedintotheneighbourhoodinfluenceNViandtheglobalinfluenceGViofhouseholdagenti.InEq.10,NViisdefinedastheratioofneighboursthatusetheresidentialallowancepolicytorelocatetoadowntownareaNmovedividedbythenumberofneighboursthatdonotusethepolicyforrelocationNnomove.GViisdefinedasratioofthetotalnumberofhouseholdsGmovethatusethepolicyforrelocationdividedbythetotalnumberofhouseholdsGTotal.AsshownbyEq.11,componentεisreflectstheunobservedrandomcontributiontoutilityVisandViinter.ThisrandomelementεisfollowsaGumbledistributionandcanbegeneratedbyEq.11,inwhichrfollowsarandomuniformdistribution,andconstantsμandβaresettobe-4.5and2,respectively,becauseitispreferabletofixtherangeofεisbetween-10and10.Inaddition,Qisistheprobabilityofhouseholdi'schoosinglocations,whichisintheformofEq.12.

(5)

http://jasss.soc.surrey.ac.uk/16/2/2.html 4 14/10/2015

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Table1.Parametersforutilitypreferencebyhouseholdsindifferentlifecyclestages

http://jasss.soc.surrey.ac.uk/16/2/2.html 5 14/10/2015

Figure5.Residenceutilitypreferencesofhouseholds

Policy,VirtualUrbanSpaceandParametersforSimulationConfiguration

PolicyapproachfordowntownrevitalizationinJapan

3.1 Generally,therearethreestagesofurbandevelopment(KlaassenandPaelinck1979):urbanization,suburbanization,andurbandecline.ThesocialbackgroundregardingurbanizationinJapanisintroducedherebrieflyasrelevanttothisresearch.TheresidentialpopulationdensityinJapaneseurbanareas—suchasHokkaido,Honshu,Shikoku,andKyushu—reachedamaximumvalueofapproximately105.6peopleperhectarein1960duringrapidurbanization(Kaido2005).Theincreasedpopulationdensityinthedowntownareas,higherincomes,andgenerallycheapertransportationthatresultedfromthisurbanizationperiodledtoincreasedhousingdemandinsuburbanareas.Althoughresearchhasindicatedthatpeopleinurbanandsuburbanareashavedifferingcircumstances,improvementsinaccessibilitythroughtheuseofpublictransportationandprivatevehiclesmeansthatrelocatingtothesuburbsisnotexpectedtoleadtoinconvenientlivingconditions—i.e.lessaccesstocomfort(Marcellini2007).Instead,homeownershipcanprovideafeelingofsecuritytopeoplewhoarenotfinanciallywelloff(Rogers1999).Thus,theresidentialpopulationdensityrateinJapanesecitycentreshasdecreasedoverallsincethe1980s.Inparticular,asithasbeenproventhatanincreaseincommutingdistancewillnotnecessarilyresultinasignificantincreaseincommutingtimebecauseofdevelopmentsintransporttechnology(MaandKang2010;Kim2008;Schafer2000),livinginthedowntownisnotasattractiveasbefore.Thissituationdiffersinsmallcitiesversusbigmetropolitanareas;forexample,theTokyoMetropolitanAreaexperiencedsuburbanizationafter1965whileitstotalpopulationcontinuedtoincreaseinthe1980s(Okamoto1997).However,smallercitiesinJapanbegantolosepopulationtowardtheendofthe1980s,enteringaneraofurbandecline.

3.2 InJapan,localgovernmentsareincreasinglyinterestedinusingpolicyapproachestorevitalizetheirdowntownareasandmaketheircitiesmorecompact.Someofthesepolicyapproacheshaveinvolveddowntownregenerationefforts,includingdevelopmentcontrolsonlarge-scaleshoppingcentres(Shenetal.2011).InsomelocalcitiesofJapan,suchasKanazawaCity,aresidencepromotionpolicyhasbeenimplementedinordertorevitalizeitsdowntownareasbyencouraginghouseholdstorelocatetodowntownareas.Themainstrategyofthispolicyistoprovideresidentswhorelocatetodowntownareaswithlocalhousingallowances.ThedetailsofthisresidencepromotionpolicyinKanazawaCity(issuedin2005)areshowninTable2.Thispolicyisthebackgroundofourworkinfocusingondesigningamodel—namely,theHRRM—tosimulatepolicyimpactsondowntownrevitalization.

Table2:ResidencepromotionpolicyofKanazawaCity

BuildingTypes Utilizationtypes AllowancesHouse Buynewhouse Singlehousehold 10%payment,2millionJPY

Twohouseholds 10%payment,lessthan3millionJPYBuyorrepairoldhouse Basicpart+supplementarypart,lessthan500,000+200,000JPY

Apartment Buynewapartment Basicpart(5%ofpayment)+supplementarypart(1%),lessthan1million+200,000JPYOldapartment Buy

Repair 50%ofdesignpayment,lessthan1millionJPY

Virtualurbanspaceforaccommodatinghouseholdagents

3.3 Inordertosimulatetheeffectsoftheresidencepromotionpolicyondowntownrevitalization,wedesignedavirtualspacethatreproducestheurbanplanningconditionsofatypicalJapanesecity.InJapan,citiesthatimplementcityplanninglawsarereferredtoasdelineationcities(DCs),whereanurbanizationcontrolline(UCL)isestablishedinordertodividetheurbanplanningareaintoUPAandUCA.TheUPAistheareainwhichthelocalgovernmentiswillingtopromoteurbanizationthroughland-usezoning.TheUCAistheareainwhichurbanizationmustbeconstrained.Land-usezonesarefurtherclassifiedintothreemajorgroups:residentialuse,commercialuse,andindustrialuse,whichcanbefurtherbrokendowninto12moredetailedcategoriesofland-usezones.Theproposedmodelassumesthatthevirtualurbanspace(patchdata)hasthetypicalcharacteristicsofdelineationcitiesinJapan:ithas1)atraditionalCCAlocatedintheheartofthecity,2)anurbanplanningareadividedintotheUPAandtheUCA,and3)definedland-usezoneswithintheUPA.Inthiswork,weembodytheconceptsofthetypicallocalcityofJapaninamono-centralvirtualspaceasshowninFig.6.Thevirtualurbanspaceconsistsof2,500cells(50×50)whereeachcellmeasures500m×500m.Inadditiontotheplanninginformation,eachcellinthevirtualurbanspacewillhavepredefinedspatialattributes,suchasshoppingarea,greenspace.

3.4 Itisimpossible,however,togenerateatypicalJapanesecityinspacewithoutincludingnumericinformation,suchasthesizesofCCA,UPA,andUCA,andhouseholddensity.Forthisreason,wereferredtoKanazawaCity,atypicalJapaneselocalcity,fororganizingthevirtualurbanspaceinthepresentwork.AsshowninTable3,theattributesofthevirtualspaceplanningareas,includingtheareasoflandusezoning,arebasedonthesituationofKanazawaCity.Theglobalparameters,includingbirthrate,deathrate,andcouplingratearebasedontheBasicCensusSurveyofKanazawaCity(2005-2007).Withrespecttotheattributesofhouseholds,thehouseholdlocationsinthisvirtualcityfollowthehouseholddensitiesintheland-usezoningandresidentialsuitabilityrestrictionsofJapan.Householdsarefurthergroupedintothreeincomelevels:rich,middleclass,andpoor.Wesetthepercentagesofpopulationbelongingtothethreeincomelevelsat20%,60%,and20%,respectively,allocatingtheminthevirtualspaceaccordingtothepercentageofincomegroupsinthedifferentlandusezoningareasofKanazawaCity.Withrespecttothenumberofhouseholds,1,500householdagentswereinitiallygeneratedinthisvirtualcityaccordingtohouseholddensitydefinedbyplanningregulation—i.e.thefloorarearatiosindifferentlandusezones.Accordingly,householddensityinthevirtualurbanspacehasbeencreatedasshowninthethirdimagefromtheleftinFig.6,andhouseholdincomeisrepresentedintheright-mostimageofFig.6.Wealsoassumethatallhouseholdshavecars.Inthepresentresearch,thedesignedvirtualurbanspaceandagentdistributionshavealreadybeenutilizedbyShenetal.(2011).

3.5 Furthermore,eachcellofthisvirtualspacehas18spatialattributes(x1throughx18).Thesefeatureswillbeusedbyhouseholdagentstoevaluatetheirsatisfactionwithcurrentlocationsortomakerelocationchoices.Adetailedintroductionofthe18spatialfeaturesisgiveninTable4.The18featuresatthecelllevelofthevirtualcityaredividedintotwogroups.Onegroupconsistsofobjectivefeatures,suchasthedistributionofpublicfacilities,andwemarkedtheseobjectivefeaturesinthevirtualcity.Theothergroupconsistsofsubjectivefeaturesthatwecannotmarkinthespace.Thesesubjectivefeaturesarerandomlyvaluedatacertainrankfollowingcertainrules,asintroducedinTable4,basedonthegeneralsituationofatypicalJapanesecity.

http://jasss.soc.surrey.ac.uk/16/2/2.html 6 14/10/2015

Figure6.Virtualurbanspaceforaccommodatinghouseholdagents

Table3:Globalparameters,Attributesofvirtualspaceandhouseholdagents

Parametername Predefineddata

Globalparameters

Birthrate 0.90%Deathrate 0.80%Couplingrate 0.59%Thresholdforsatisfaction 0.1,KidaniandKawakami(1996)Policyscenariooption UseresidencepromotionpolicyornotCommute-over Maximumcommutingdistance

AttributesofVirtualspace

Landusezoning 12typesoflandusezoningsHouseholdDensity HouseholdnumbersbasedonFloorAreaRatiosBasedonlandusezoningPlanningareas CCA(4%),UPZ(32%),UCA(64%)18spatialfeaturess 18variables(asTab.5)evaluatedbasedonlocation.Houseprice 800(CCA),400(UPA),200(UCA)thousands(JPY)/3.3m2

AttributesofHousehold

Income Low(20%),middle(60%)andhigh(20%)Carship YesHouseholdage accordingtohouseholdlifecyclestagesdescribedinFig.2Familymembers 1-6basedonlifecyclestagesSaving RandomNumberbasedonincomegroupsandlifecyclestagesGlobalinfluence Asdescribedinsection2.4Neighborhoodinfluence Asdescribedinsection2.4PersonalSatisfactionthreshold Theglobalthreshold0.1andarandomnumberinnormaldistributionwithmean

0andSD0.01.Satisfactionofcurrentlocation Asdescribedinsection2.5.1IndividualparametersofUtilitiesregardingrelocationalternatives

AsdescribedinTable5.

Location Coordinationx,y

http://jasss.soc.surrey.ac.uk/16/2/2.html 7 14/10/2015

Table4.Thespatialfeaturesofthevirtualcityforhouseholds'makingdecisiononresidentialrelocationchoices

Parametersforsimulationofhouseholds'adaptivebehavioursandpolicyinteractionsonresidentialrelocation3.6 Theparametersoftheutilitymodelinthepresentworkrepresenttheprinciplesofadaptivebehavioursandpolicyinteractionsofhouseholdagentsinaresidentialrelocationprocess.Thoseparameters

areretrievedfromtheresultsofanempiricalsurveyinKanazawaCityinordertoassociatethepreferencesofhouseholdagentsinrealsocietywiththoseinthevirtualspace.Theempiricalsurveyfocusedontwoparts:1)satisfactionwithcurrentlocationsand2)knowledgeoflocalresidencepromotionpolicy.Forthefirsttheme,thequestionnaireasksthequestion,'Wouldyouliketorelocate?'Basedontheresultsofthesurvey,weestimatedtheparametersthrougharegressionanalysisofSiandxijs,usingRstatisticalsoftwaretodeterminethecoefficientsbj(b1tob18)forhouseholdagents'evaluationofsatisfactionwiththeircurrentlocations.TheseresultsarelistedinTable5.Forthesecondfocus,thequestionnaireasksthequestion,'Doyouplantorelocatetoadowntownareaasaresultofthispolicyandindoingsoreceiveanallowance?'Residentswereaskedtochooseananswerfromamongfouroptions.Aregressionanalysiswasconductedinordertoestimatethecoefficientsrepresentingthepolicyeffectsonhouseholdrelocationbasedonthequestionnaire.TheseestimatedpartialcoefficientsarelistedinTable5asb'j.Inaddition,theHRRMwasimplementedontheNetlogoplatformformodeltesting.

Table5:Partialcorrelationcoefficientofimpactfactorsonhouseholdrelocationutilityandsatisfaction

Variablesforsatisfactionevaluation Coefficients(bj) Partialcoefficient(nopolicyinteraction)

Significant Coefficients(b'j) Partialcoefficient(withpolicyinteraction)

Significant

x1:Buildingsize b1 0.170634 **** b'1 0.069904 ·

x2:Buildingsecurityfromearthquake,typhoon b2 0.140753 **** b'2 0.153358 ***

x3:Buildingsecurityfromfire b3 0.045609 · b'3 0.085280 *

x4:Buildingimpairment b4 0.029561 · b'4 0.112268 **

x5:Barrier-freestructuresforoldpeople b5 0.067244 · b'5 0.051591 ·

x6:Surroundingsafetyequipments b6 0.125595 *** b'6 0.072290 *

x7:Safetywhilewalkingonsurroundingpavement b7 0.126829 *** b'7 0.105173 *

x8:Crimerate b8 0.109048 *** b'8 0.085835 *

x9:Airornoisepollution b9 0.162371 **** b'9 0.098698 *

x10:Accessibilitytoworkorschool b10 0.056896 · b'10 0.079705 *

x11:Shoppingconvenience b11 0.118102 *** b'11 0.075495 *

x12:Accessibilitytocommunityhospital b12 0.111277 *** b'12 0.052196 ·

x13:Culturalfacilities(e.g.Distancefromlibrary) b13 0.097474 ** b'13 0.058622 ·

x14:Parkorplayinggroundforchildren b14 0.100275 ** b'14 0.100814 *

x15:Greenspace b15 0.079093 ** b'15 0.123672 **

x16:Theareasofoutspace b16 0.172469 **** b'16 0.027558 ·

x17:Streetlandscape b17 0.087435 ** b'17 0.102831 *

x18:Communicationfeasibilitywithneighbors b18 0.061328 · b'18 0.020922 ·

Interaction1:Neighborhoodinfluence(x19) b'19 0.064556 ·

Interaction2:Globalinfluence(x20) b'20 0.174542 ****

Significantvalues:0.1"·"0.05"*"0.001"**"0.0005"***"0.00001"****"

http://jasss.soc.surrey.ac.uk/16/2/2.html 8 14/10/2015

ModelTest

Sensitivitytestforhouseholds'adaptivebehaviourinaresidentialrelocationprocess

4.1 IntheHRRM,theadaptivebehavioursofhouseholdagentsinaresidentialrelocationprocessweredevelopedusingalifecyclestagemodelandautilitymodelasintroducedinsection2;therefore,itwasnecessarytoprovethatadaptivebehavioursbasedonthesemodelsintheHRRMcouldproducereasonablesimulationoutputs.Forthispurpose,weconductedasensitivityanalysisoftheparametersrelatingtoadaptivebehaviours(e.g.satisfactionthreshold,parametersoflifecyclestages)andpolicyinteractions.TheinterfaceoftheHRRMisshowninFig.7.Onetickinthesimulationrepresentsoneyear,and30ticksaresimulatedineachexperiment.Thesimulationresultsthusrepresenttheresultsfor30years,followingfromtheinitialstage.Theentiresensitivitytestwasrepeatedin50experimentswiththeinitialvaluesoftheparameterslistedinTables3and5,buttheparametertestedchangedaccordingtotheneedsofthesensitivityanalysis.

Satisfactionthresholdanddesireofahouseholdtorelocate

4.2 InFig.8(a),weshowthesensitivitytestofthesatisfactionthresholdhouseholdsusetoevaluatetheircurrentresidentiallocation.Wetestedourmodelbyadjustingthesatisfactionthresholdfrom-0.9atthefirsttickto1.2atthe210thtickbymeansofsettingthetickintervalto0.01.AsshowninFig.8(a),thenumberofhouseholdssatisfiedwiththeircurrentlocationdecreasesasthethresholdincreases.Thenumberofhouseholdssatisfiedwiththeircurrentlocationandthenumberthataredissatisfiedintersectwhenthethresholdvalueis0.25atthe115thtick,andtheintersectioncorrespondstothethreshold0.25,atwhichtheproportionofhouseholdsthataresatisfiedwiththeircurrentlocationis50%.Previousstudieshaverevealedthat,inKanazawaCity,theproportionofhouseholdsthatwishtorelocateis30%(KikuchiandNojima2007;KawakamiandTakama1978).Thus,wesetthesatisfactionthresholdofhouseholdresidentialrelocationas0.1inorderthattheratioof'unsatisfied'householdagentswillremainaround30%insimulation,asshowninFig.8(b).

Figure7.InterfaceoftheHRRMintheNetlogoplatform

a)Sthreshold(-0.9to1.20) (b)Sthreshold=0.10Figure8.Satisfactionthresholdinfluenceonthedesireofhouseholdstorelocate

Testontheparametersofhouseholdlifecyclestage

4.3 Thenumberofhouseholdagentsintheurbanspacechangesbasedonthechangesofparametersrelatingtothelifecyclestages.Inordertoinvestigatehowtherelocationprocessisinfluencedbythelifecyclestage,asensitivitytestwasconductedofsuchparametersasthebirthrateanddeathrateasdefinedinthelifecyclestage.Weconducted50experimentsinwhichthebirthrateanddeathratewerevariedinordertodeterminethesensitivityoftheeffectsofthetwoparametersonhouseholdresidentialrelocations.InFig.9(a1)and(a2),thebirthratewassetto0.1and0.5,respectively.First,asensitivityanalysisofthebirthratewasconducted.Thedeathrateandcouplingrateweremaintainedconstantattheirinitialvalues.AsshowninFig.9(a1),therearemorehouseholdsintheUCAthanintheCCAduringtherunningofthesimulations.Asthebirthrategrewto0.5,thenumbersofhouseholdsinthethreeurbanareasincreasedwhilemaintainingthesamerelativerelationshipinFig.9(a2).

4.4 ThefiguresthroughFig.9(b1)and(b2)representhouseholdsthatrelocatetodifferenturbanareas.Asthebirthrateincreases,althoughthetotalnumberofhouseholdsthatrelocatetodifferenturbanareaseventuallyincreases,mostofthesehouseholdsrelocatetotheUCA.Inthemeantime,wefurtherusedimagestoshowtheaveragevalueofrelocatedhouseholdsatdifferentagelevels.Onethingshouldbeclarifiedhereisthatthehouseholdagewetalkedherearetheagesofhouseholders.AsshowninFig.9(c1),whenthebirthrateequatesto0.1,mostrelocationsoccuramonghouseholdsthataremorethan60yearsold,followedbyhouseholdsbetween40and50yearsold.Householdsyoungerthan20(around18)orapproximately30yearsoldshowalowpotentialforrelocation.Whenthebirthrateisincreasedto0.5,however,thetrendgraduallychanges;mostrelocationshappenamonghouseholdsyoungerthan20yearsold.Thesharpincreaseinthebirthratewillincreasethepercentageofyounghouseholdsinthepopulation,therebyresultinginanincreaseofrelocationsamonghouseholdslessthan20yearsold.Thus,ahigherbirthrateresultsinincreasingthenumberofyounghouseholds(lessthan20yearsold).Fromthisviewpoint,althoughahigherbirthratecanrelievethesituationofanagedsociety,itwillnotincreasetheresidentialrateinurbancentres,asindicatedinthesimulatedresultinFig.9(b2).Therefore,withoutspecialpolicies,downtownareaswillnotberevitalizedeventhoughthetotalnumberofhouseholdsisincreasing.

4.5 Incontrast,ifwekeepthebirthrateconstantatitsinitialvaluebutincreasethedeathratefrom0.008to0.1,asshownasFig.10(a1),boththetotalnumberofhouseholdsandthenumbersofhouseholdsindifferenturbanareasdecrease.TheincreaseddeathratedoesnotchangethetrendofhouseholdrelocationwithrespecttoagelevelsinFig.10(a2).Householdsthataremorethan60yearsoldmakeupthemajorityofhouseholdsthatrelocate.Furthermore,aminorityofhouseholdsthatareapproximately30yearsold,alongwithmostotherhouseholds,relocatetotheUCA.Thishelpstoclarifythesituationoflocalurbandeclineandtheagedsociety.Withoutincreasingthebirthrate,thedeathratehasnoobviouseffectondowntowndeclineortheagedsociety.Moreover,withoutspecialinterventionwiththegoalofdowntownrevitalization,changesinthebirthrateorthedeathratedonotinfluencedowntowndecline.Althoughincreasingthebirthratecanincreasethenumberofyounghouseholdsandimprovetherelocationfrequency,mostrelocationsstilloccurintheUCA.

4.6 Asaresult,wecanconcludethatintheHRRM,thechangesintheparametersofthelifecyclestagecanproducereasonablechangesinhouseholdadaptivebehavioursinresidentialrelocation,butitisimpossibletorevitalizethedowntown.InordertoincreasetherelocationfrequencytoCCA,specialpolicyinterventionisneeded.Inthefollowingsection,wewillinvestigatepolicyimpactsontherelocationofhouseholdagentsthroughagents'interactions.

http://jasss.soc.surrey.ac.uk/16/2/2.html 9 14/10/2015

(a1)Householdsinurbanareabybirthrate0.1 (a2)Householdsinurbanareabybirthrate0.5

(b1)Householdsrelocationtourbanareabybirthrate0.1 (b2)Householdsrelocationtourbanareabybirthrate0.5

(c1)Householdsrelocationbyhostagebybirthrate0.1 (c2)Householdsrelocationbyhostagebybirthrate0.5Figure9.Householdchangeindifferenturbanareasaccordingtobirthrate

(a1)Householdsinurbanareasbydeathrate0.008 (a2)Householdsinurbanareasbydeathrate0.1

(b1)Householdsrelocationtourbanareabydeathrate0.008 (b2)Householdsrelocationtourbanareabydeathrate0.1

http://jasss.soc.surrey.ac.uk/16/2/2.html 10 14/10/2015

(c1)Householdsrelocationbyhostagebydeathrate0.008 (c2)Householdsrelocationbyhostagebydeathrate0.1Figure10.Householdchangeindifferenturbanareasaccordingtodeathrate

Policyimpactonhouseholds'decisionmakingconcerningtheresidentialrelocationprocess

4.7 AsshowninFig.3,householdagentswhohaveadesiretorelocatewillcomparetherelocationcostswiththeirsavings.Ifhouseholdsavingsareinsufficient,thehouseholdwillconsiderapplyingfortheallowancestipulatedinthelocalresidencepromotionpolicyandwillrelocatetoadowntownarearatherthansimplytotheareawiththehighestutility.

4.8 Inordertoinvestigatethepolicyimpactonthedecisionofhouseholdagentsinthehouseholdrelocationprocess,asensitivityanalysiswasconductedonneighbourhoodinfluenceandglobalinfluence,whicharerelatedtotheratioofagentswhoagreetotakeadvantageofthepolicyandrelocatetoadowntownarea.First,wetestedthemodelwiththeneighbourhoodinfluenceas0.064556andglobalinfluenceas0.174542(estimatedresultsinTable5),whicharetheinitialvaluesandnamedinitialinteractionsinFig.11.Wethenchangedthevaluesofneighbourhoodinfluenceandglobalinfluencebothto0.5toshowthechangesinhouseholdresidentialrelocationsduringanincreaseinpolicyinteractions.Thereafter,weconductedacomparativeanalysisbetweenthesimulationresultswithinitialpolicyinteractionsandtheresultswithincreasedpolicyinteractionsbymeansofRsoftware.Duringalltheseprocesses,allotherparametersofHRRMweremaintainedconstantattheirinitialvalues.

4.9 IfFig.11(a1)iscomparedwithFig.9(a1)orFig.10(a1),itcanbeseenthatwhenthepolicyinteractionwasincorporatedinthesimulation,thenumberofhouseholdsintheCCAincreased(asshownbythegray).Meanwhile,Fig.11(b1)showsthatthenumberofhouseholdsthatrelocatedtotheCCAalsoincreasedcorrespondingly.Thisresultshowsthathouseholdswouldbeaffectedbythisresidencepromotionpolicyduringtheirdecision-makingonresidentialrelocationsifthepolicywaspublicizedwithintheneighbourhoodandentiresociety.Figs.11(c1),9(c1),andFig10(c1),however,showthatthemostrelocationsoccurredamonghouseholdsthataremorethan60yearsold,followedbyhouseholdsbetween40and50yearsold,householdsthatareyoungerthan20(around18)andhouseholdsthatareapproximately30yearsold,inthatorder.Theseresultswouldseemtoindicatethatthetrendofhouseholdsdoingrelocationsindifferentagegroupswillnotchangewhetherthereisapolicyinterventionornot.Next,wefurtherinvestigatedhowhouseholdrelocationwouldchangeifwechangedtheintensityofthepolicyimpactonhouseholdinteraction.

4.10 Thus,weincreasedtheneighbourhoodinfluenceandglobalinfluenceto0.5.ThecomparativesimulationresultsareshowninFigs.11(a2),(b2),and(c2).Thevirtualurbanspaceinthepresentworkwasdesignedasaclosedurbanspacewithoutpopulationmobilityfromtheoutside.Thissituationdecreasesthetotalnumberofhouseholdsduringsimulation.Whenneighbourhoodinfluenceandglobalinfluenceareboth0.5,thegrayboxesinFig.11(a2),whichrepresentthenumberofhouseholdsintheCCA,furtherincrease,butthehouseholddensityintheUCAandUPAdecreasessignificantly.ThedecreaseinthenumberofhouseholdsintheUCAandUPAispartiallyaresultofthedecreaseinthetotalnumberofhouseholds,andtheincreaseinthenumberofhouseholdsintheCCAisaresultoftheresidencepromotionpolicy.Inaddition,asthevaluesofneighbourhoodinfluenceandglobalinfluenceincrease,thenumberofhouseholdsintheCCAincreasesobviouslyinFig.11(b2)comparedwithFig.11(b1);specifically,thenumberofhouseholdsthatrelocatetotheCCAincreasesinbothofthesefigures.TheseresultsdoindicatethataresidencepromotionpolicycaninfluencehouseholdresidentialrelocationtoadowntownareaandthattheHRRMcanrepresentandvisualizethisprocess.

(a1)Householdnumberinurbanareaswithinitialinteractions (a2)Householdsinurbanareawithincreasedinteractions

(b1)Householdsrelocationtourbanareawithinitialinteractions (b2)Householdsrelocationtourbanareawithincreasedinteractions

http://jasss.soc.surrey.ac.uk/16/2/2.html 11 14/10/2015

(c1)Householdsrelocationbyhostagewithinitialinteractions (c2)HouseholdsrelocationbyhostagewithincreasedinteractionsFigure11.Householdchangeindifferenturbanareasaccordingtopolicyinteractions

4.11 However,asindicatedbyFigs.11(c1)and(c2),aswiththebirthrateanddeathratesensitivityanalyses,thenumberofhouseholdsthatrelocatedtotheCCAare,indecreasingorder,householdsthataremorethan60yearsold,householdsthatarebetween40and50yearsold,householdsthatareapproximately20yearsold,andhouseholdsthatareapproximately30yearsold.Thus,wecanconcludethatthelimitedallowancesforrelocationarenotsufficienttoencourageyoungerhouseholdstochooseresidencesindowntownareas.AsshownbyFig.10(c1)and(c2)andFig.11(c1)and(c2),themiddle-agedandelderlyhouseholdsarebothmorelikelytorelocatetoadowntownareausingthispolicy,ascomparedtohouseholdsthatarelessthan30yearsold.Therefore,thispolicyprobablytendstobemoreattractivetohouseholdsmorethan40yearsoldthatarewell-off.Fromthisviewpoint,goodpublicizingoftheresidencepromotionpolicyhaveaneffectonhouseholdsrelocatingtodowntown,butthiseffectislimited.

TheMooreTestforSpatialRepeatability

4.12 FromFigs.9to11,wecanconfirmthatthenumberofhouseholdagentsandrelocationsarestatisticallystableinthe50timessimulationresults.Inthissection,wewouldliketofurtherprovethespatialrepeatabilityofsimulationresultsbymeansoftheHRRM.Theexperimentsweredone50timeswiththeparametersintroducedinTables3and5.Ineachexperiment,weranthesimulationfor30ticks.Simulationresultstakenevery10tickswereusedtodeterminethespatialrepeatabilityofthenumberofhouseholdsineachcellforthesameticknumberamongthe50simulations.

4.13 First,themeansamongthe50simulationsofthenumbersofhouseholdsforacellat10,20,and30ticksarevisualized,asshowninFig.12.AWelch'sttestofthenumberofhouseholdsfordifferentsimulationswasconducted,andSig.waslessthan0.05inallcases.Thus,thesimulationresultsobtainedusingtheHRRMarestable.WealsoconductedaMooreTestofthenumberofhouseholdsineachcellusingR;theresultsrevealedastablecorrelationandshowedthatSig.waslessthan0.05inallcases.

4.14 SummarizingthesensitivityanalysisinSection4,whenweadjusttheparametersrelatedtolifecyclestageandhouseholdinteractions,theHRRMcanreflectreasonablywellthechangesinthenumberofhouseholdsindifferenturbanareas.Moreover,thespatialdistributionofhouseholdagentsinthesimulationprocessisrepeatable,too.

(a)Averageagentnumberineachcellattick10 (b)Averageagentnumberineachcellattick20 (c)Averageagentnumberineachcellattick30Figure12.Householddistributioninvirtualurbanspaceinsimulationprocess

Note:AWelch'sttestandaMooreTesthavebeenconductedfornumbersofhouseholdsbetweenticks10,20,and30.Significanceissmallerthan0.05inallcases

ComparingsimulationdataandrealdataforKawakawaCity

4.15 Followingtheexplanationinsubsection3.2,wedesignedthevirtualurbanspacebasedonthesituationofatypicalJapanesecity.Specifically,weretrievedparametersfromanempiricalsurveyofKanazawaCityinordertoassociatethebehaviourpreferencesofhouseholdagentsinsimulationwiththoseofatypicalJapanesecity.Thus,thesimulatedpopulationdistributionoftheHRRMisexpectedtobesimilartothecharacteristicsofKanazawaCity.AsmentionedinSubsection3.2,eventhoughurbanspacesandpopulationdistributionsofdifferentplanningareasinvirtualspaceandthoseofKanazawaCityaredifferent,weassumethathouseholdagentswillcomparetheutilitiesonlybetweenalternativesselectedrandomlyfromtheUPA,UCA,andCCA.Thus,inthepresentstudy,itispossibletocomparetheproportionsofhouseholdnumbersindifferentplanningareasbetweenthevirtualspaceandKanazawaCity.

4.16 Consequently,thesimulationofhouseholdresidentialrelocationusingavirtualspacewasdesignedfor15years.Inordertomakeacomparisonwith1985-2000inKanazawaCity,wecomparedthesimulationresultswithlocalstatisticaldatabyconvertingthesimulationresultsintohouseholdratiosfordifferenturbanplanningareas.TheresultsarecomparedinTable6,whichshowsthatthelocalcensussurveyin1985giveshouseholdratiosof33.9%inCCAand66.1%inUPAandUCA,whereastherespectivevaluesobtainedbythesimulationare32.7%and67.3%.Thus,thesimulationresultsareingoodagreementwiththerealdata.Upuntil1990,thehouseholdratioinCCAwas31.9%fortherealdatasetand32.1%forthesimulation,whichagainagreewell.Thisisalsothecasefor1995.However,thedifferencebecamelargerintheyear2000:thesimulationresultsforhouseholdsinCCAare10%greaterthanintherealdata.ThisresultprobablystemsfromachangeintheBuildingStandardsActin1998,inwhichthetraditionalsixtypesofland-usezoneswerechangedto12typesofland-usezones(BuildingStandardsAct1998).

Table6:Comparisonofhouseholdratiosindifferenturbanareasbetweenrealdataandthesimulationresults

Years: 1985 1990 1995 2000Realdataset CCA 33.9% 31.9% 29.0% 26.6%

UPA+UCA 66.1% 68.1% 71.0% 73.4%Simulatedresults CCA 32.7% 32.1% 30.4% 33.7%

UPA+UCA 67.3% 67.9% 69.6% 66.3%

Conclusions

5.1 TheHRRMintegratedtheadaptivebehavioursandpolicyinteractionsofhouseholdagentsinaresidentialrelocationprocess,whichcanmimictheprocessofhouseholddecision-makinginresidentialrelocationbyintegratingABMwithhouseholdlifecyclestages,satisfactionevaluationofthecurrentresidentiallocation,andselectionofanewlocation.

5.2 Regardingtheevaluationmoduleforhouseholdrelocationdesireandchoiceinthehouseholdrelocationprocess,wedevelopedautilitymodeltoreflectindividualchoicesbyintroducinganormallydistributedrandomperturbationtothepartialcorrelationcoefficientsofallvariablesintheutilitymodel,whichcanberecognizedasthepreferenceofhouseholdsforadaptingthemselvestodifferentlocationsinnewlifecyclestagesintheHRRM.Interactionsbetweenhouseholdagentsareconsideredtotakeplaceatthelevelsoftheneighbourhoodandtheentirecity,reflectingtheirattitudestotheresidencepromotionpolicyduringtheresidentialrelocationprocess,whichwereincludedasonecomponentofutilitymodel.

5.3 Inordertotesttheadaptivebehavioursandinteractionsofhouseholdagentsinsimulation,weexaminedthemodel'ssensitivitytotheparametersdefinedinthelifecyclestagemoduleandtheparametersdefinedinhouseholdpolicyinteraction.Comparingthesimulationresultswithdifferentpolicyparameters,itwasdeterminedthatthenumberofhouseholdsrelocatingtotheCCAincreases

http://jasss.soc.surrey.ac.uk/16/2/2.html 12 14/10/2015

whenthepolicyparametersareintroduced.Thismeansthatdowntowndeclinecanberelievedbyimplementingthisresidentialpolicyasmodelledhere.Althoughthelocalresidencepromotionpolicyaffectshouseholdresidentialrelocationtoadowntownarea,thepolicyisnotveryeffectiveeitherforhouseholdsthatareyoungerthan20yearsoldorforhouseholdsthatareapproximately30yearsold.Thus,weanalyzedwhethertheallowanceforrelocationwasinsufficientforencouragingyoungerhouseholdstochooseresidencesindowntownareasandfoundthatthispolicytendstobemoreattractivetowell-offhouseholds—thatis,thosethathavesufficientsavingstorelocate.Comparedtohouseholdslessthan30yearsold,middle-agedhouseholdsandelderlyhouseholdsbothexhibitedagreatertendencytorelocatetodowntownareasusingthispolicy.

5.4 TheparametersemployedforthevirtualspacearebasedonthesituationofKanazawaCity,andtheparametersoftheutilitymodelareestimatedbasedonresponsestoanempiricalsurveyconductedinKanazawaCity.Thus,theparametervaluesusedintheHRRMcanbeassociatedwiththoseinKanazawaCity.Finally,wecomparedthesimulationresultsobtainedusingtheHRRMandtherealstatisticaldataforKanazawaCity,whichrevealedthatthesimulationresultsforthehouseholdresidentialrelocationaresimilartotherealstatisticalresultsforactualhouseholdresidentiallocationsoverthepast20yearsinKanazawaCity.Thus,theHRRMcanvisualizethepossibleresultsofpolicyimplementation,therebymakingitpossibletojudgethepotentialeffectivenessofaresidencepolicyforrevitalizingthecitycentreofatypicalcityinJapan.

References

ANTROPM.(2004).LandscapechangeandtheurbanizationprocessinEurope.LandscapeandUrbanPlanning,67,9-26.[doi:10.1016/S0169-2046(03)00026-4]

BERGERT.,SchreinemachersP.,WoelckeJ.(2006).Multi-agentsimulationforthetargetingofdevelopmentpoliciesinless-favoredareas.AgriculturalSystems,88(1),28-43.[doi:10.1016/j.agsy.2005.06.002]

BANISTERD.(1997).Reducingtheneedtotravel.EnvironmentandPlanningB:PlanningandDesign,24(3),437-449.[doi:10.1068/b240437]

BANISTERD.,WatsonS.andWoodC.(1997).Sustainablecities:transport,energy,andurbanform.EnvironmentandPlanningB:PlanningandDesign,24(1),125-143.[doi:10.1068/b240125]

BROWND.G.andRobinsonD.T.(2006).EffectsofHeterogeneityinResidentialPreferencesonanAgent-BasedModelofUrbanSprawl.EcologyandSociety,11(1),46-67.

BENENSON,I.(1998).Multi-agentsimulationsofresidentialdynamicsinthecity.Computers,EnvironmentandUrbanSystems,22,25-42.[doi:10.1016/S0198-9715(98)00017-9]

BUILDINGStandardsAct.(1998).MinistryofLand,Infrastructure,TransportandTourism,Japan.JapanesewebsiteArchivedat:http://www.mlit.go.jp/

BOYLE,P.,Halfacree,K.,&Robinson,V.(1998).Exploringcontemporarymigration.Harlow,NewYork:Longman.

CADWALLER,M.(1992).Migrationandresidentialmobility.Microandmacroapproaches.Wisconsin:UniversityofWisconsinPress.

CLARK,W.A.V.,&Onaka,J.L.(1983).Lifecycleandhousingadjustmentasexplanationsofresidentialmobility.UrbanStudies,20(1),47-57.[doi:10.1080/713703176]

DAWNC.P.andFilatovaT.(2008).Aconceptualdesignforabilateralagent-basedlandmarketwithheterogeneouseconomicagents.Computers,EnvironmentandUrbanSystems,32,454-463.[doi:10.1016/j.compenvurbsys.2008.09.012]

DIAPPIL.andBolchiP.(2006).GentrificationWavesintheInner-CityofMilan-Amultiagent/cellularautomatamodelbasedonSmith'sRentGaptheory,VanLeeuwen,J.P.andH.J.P.Timmermans(eds.).InnovationsinDesign&DecisionSupportSystemsinArchitectureandUrbanPlanning(pp.187-201),Dordrecht:Springer.

ETTEMAD.(2011).Amulti-agentmodelofurbanprocesses:Modellingrelocationprocessesandpricesettinginhousingmarkets.Computers,EnvironmentandUrbanSystems,35,1-11.[doi:10.1016/j.compenvurbsys.2010.06.005]

ETTEMA,D.,deJong,K.,Timmermans,H.,&Bakema,A.(2006).PUMA:Multi-agentmodellingofurbansystems.InE.Koomen,A.Bakema,J.Stillwell,&H.Scholten(Eds.),Landusemodeling.Springer.

FONTAINEC.M.,RounsevellM.D.A.(2009).Anagent-basedapproachtomodelfutureresidentialpressureonaregionallandscape.LandscapeEcological,24,1237-1254.[doi:10.1007/s10980-009-9378-0]

HAASED.,LautenbachS.,SeppeltR.(2010).Modelingandsimulatingresidentialmobilityinashrinkingcityusinganagent-basedapproach.EnvironmentalModelling&Software,4,1225-1240.[doi:10.1016/j.envsoft.2010.04.009]

HOWLEYP.,ScottM.,RedmondD.(2009).Anexaminationofresidentialpreferencesforlesssustainablehousing:ExploringfuturemobilityamongDublincentralcityresidents.Cities,26,1-8.[doi:10.1016/j.cities.2008.10.001]

JAGERW.,Mosler.H.J.(2007).SimulatingHumanBehaviorforUnderstandingandManagingEnvironmentalResourceUse.JournalofSocialIssues,63(1),97-116.[doi:10.1111/j.1540-4560.2007.00498.x]

KAZEPOVY.(2005).CitiesofEurope:ChangingContexts,LocalArrangements,andtheChallengetoSocialCohesion.UK:BlackwellPublishing.[doi:10.1002/9780470694046]

KAWAKAMIM.,TakamaC.(1978).InvestigationofhouseholdresidentialrelocationdesireinKanazawaCity(inJapanese).ArchitectureInstituteofJapan,13,67-72

KIKUCHIY.,NojimaS.(2007).Resident'sMindaboutResidenceSelectioninSuburbanHousingEstates:Casestudyof4suburbanestatesinFukuiCity.Cityplanningreview:Specialissueoncityplanning,42(3),217-222.

KIMC.(2008).Commutingtimestability:atestofaco-locationhypothesis.TransportationResearchA,42,524-544.[doi:10.1016/j.tra.2008.01.001]

KAIDOK.(2005).UrbandensitiesandlocalfacilityaccessibilityinprincipalJapanesecities,In:M.JenksandN.Dempsey(Eds.),FutureFormsandDesignforSustainableCities(pp.311-337).ArchitecturalPress.

KLAASSEN,L.andPaelinckJ.(1979).Thefeatureofthelargetowns.EnvironmentandPanningA,11,1095-1104.[doi:10.1068/a111095]

KIM,J.H.,Pagliara,F.,Preston,J.(2005).Theintentiontomoveandresidentiallocationchoicebehaviour.UrbanStudies,42(9),1621-1636.[doi:10.1080/00420980500185611]

KIDANIH,KawakamiM,1996,"Actualconditionandplanningthemeofbuildingactivityinurbanizationcontrolarea"JournaloftheCityPlanningInstituteofJapan,31,583-588.

KULU,H.(2007).Fertilityandspatialmobility:EvidencefromAustria.EnvironmentandPlanningA,40(3),632-652.[doi:10.1068/a3914]

KULU,H.,&Milewski,N.(2007).Familychangeandmigrationinthelifecourse:Anintroduction.DemographicResearch,17(9),567-590.[doi:10.4054/DemRes.2007.17.19]

KÄHRIKA.,LeetmaaK.,TammaruT.(2012).Residentialdecision-makingandsatisfactionamongnewsuburbanitesintheTallinnurbanregion,Estonia.Cities,29,49-58.[doi:10.1016/j.cities.2011.07.005]

LEEQ.,ParkB.,VlekS.J.P.L.G.(2010).LandUseDynamicSimulator(LUDAS):Amulti-agentsystemmodelforsimulatingspatio-temporaldynamicsofcoupledhuman-landscapesystem:2.Scenario-basedapplicationforimpactassessmentofland-usepolicies.EcologicalInformatics,5(3),203-221.[doi:10.1016/j.ecoinf.2010.02.001]

LIY.andMullerB.(2007).Residentiallocationandthebiophysicalenvironment:exurbandevelopmentagentsinaheterogeneouslandscape.EnvironmentandPlanningB:PlanningandDesign,34,279-295.[doi:10.1068/b31182]

MARCELLINIF.,etal.(2007).AginginItaly:Urban-ruraldifferences.ArchivesofGerontologyandGeriatrics,44,243-260.[doi:10.1016/j.archger.2006.05.004]

MAK.R.andKangE.T.(2010).Time-spaceconvergenceandurbandecentralisation.JournalofTransportGeography,inpress.

MOECKELR.,SchwarzeB.,SpiekermannK.,WegenerM(2007)Simulatinginteractionsbetweenlanduse,transportandenvironment.Paperpresentedat11thworldconferenceontransportresearch.UniversityofCalifornia.Berkeley,24-28Jun2007

MOLIN,E.,Timmermans,H.(2003).Accessibilityconsiderationsinresidentialchoicedecisions:accumulatedevidencefromtheBenelux.TRB2003AnnualMeetingCD-ROM,Washington,DC.

MILLER,E.J.,Hunt,J.D.,Abraham,J.E.,&Salvini,P.A.(2004).Microsimulatingurbansystems.Computers,EnvironmentandUrbanSystems,28,9-44.[doi:10.1016/S0198-9715(02)00044-3]

MOECKEL,R.,Wegener,M.,&Schwarze,B.(2005).Simulatinglandusechangebymodellingpersons,householdsanddwellings.InPaperpresentedatCUPUMconference,London.

MOECKELR.,SpiekermannK.,SchürmannCarsten.,WegenerM.,(2003).MicrosimulationofLandUse.InternationalJournalofUrbanSciences,7(1),14-31.[doi:10.1080/12265934.2003.9693520]

MULDER,C.H.,&Wagner,M.(1998).First-timehome-ownershipinthefamilylifecourse:AWestGerman/Dutchcomparison.UrbanStudies,35(4),687-713.[doi:10.1080/0042098984709]

NEWMANP.andKenworthyJ.(1989).CitiesandAutomobileDependence:AnInternationalSourcebook.Aldershot:GowerPublishingCompany.

http://jasss.soc.surrey.ac.uk/16/2/2.html 13 14/10/2015

NOTH,M.,Borning,A.,Waddell,P.(2003).Anextensible,modulararchitectureforsimulatingurbandevelopment,transportation,andenvironmentalimpacts.Computers,EnvironmentandUrbanSystems,27(2),181-203.[doi:10.1016/S0198-9715(01)00030-8]

OKAMOTOK.(1997).SuburbanizationofTokyoandTheDailyLivesofSuburbanPeople.In:P.P.KaranandK.SKentucky(Eds.),TheJapaneseCity(pp.79-105).TheUniversityPressofKentucky.

RIENIETST.(2005).Shrinkingcities-GrowingDomainforUrbanPlanning?Availableat:http://aarch.dk/fileadmin/grupper/institut_ii/PDF/paper_presentation_EURA2005.pdf

RIENIETST.(2009).Shrinkingcities:causesandeffectsofurbanpopulationlossesinthetwentiethcentury.NatureandCulture,4(3),231-254.[doi:10.3167/nc.2009.040302]

ROGERSC.C.,(1999)Changesintheolderpopulationandimplicationsforruralareas.EconomicResearchService(RuralDevelopmentResearchReportNo.90),U.S.DepartmentofAgriculture,Washington,DC.Availablefrom:http://www.ers.usda.gov/publications/rdrr90/i.

ROUWENDAL,J.,Meijer,E.(2001).Preferencesforhousing,jobs,andcommuting:amixedlogitanalysis.JournalofRegionalScience,41(3),475-505.[doi:10.1111/0022-4146.00227]

SCHAFERA.(2000).Regularitiesintraveldemand:aninternationalperspective.JournalofTransportationStatistics,3(3),1-31.

SHENZ.J.,YaoX.A.,KawakamiM.,ChenP.(2011).Simulationspatialmarketsharepatternsforimpactsanalysisoflarge-scaleshoppingcentersondowntownrevitalization.EnvironmentandPlanningB:PlanningandDesign,38,142-162.[doi:10.1068/b35138]

TORRENSP.M..(2001).CanGeocomputationsaveurbansimulation?Throwsomeagentsintothemixture,simmer,andwait...UniversityCollegeofLondon,CenterforAdvancedSpatialAnalysis,WorkingPaperseries:Publication32.

TIMMERMANS,H.J.P.(2003).Thesagaofintegratedlanduse-transportmodeling:Howmanymoredreamsbeforewewakeup?InPaperpresentedatthe10thinternationalconferenceontravelbehaviourresearch,Lucerne,10-15August,2003.

VEGAA.andReynolds-FeighanA.(2009).Amethodologicalframeworkforthestudyofresidentiallocationandtravel-to-workmodechoiceundercentralandsuburbanemploymentdestinationpatterns.TransportationResearchPartA,43,401-419.[doi:10.1016/j.tra.2008.11.011]

WADDELL,P.,Borning,A.,Noth,M.,Freier,N.,Becke,M.,&Ulfarsson,G.(2003).Microsimulationofurbandevelopmentandlocationchoices:DesignandimplementationofUrbanSim.NetworksandSpatialEconomics,3,43-67.[doi:10.1023/A:1022049000877]

http://jasss.soc.surrey.ac.uk/16/2/2.html 14 14/10/2015