Dynamic and Geographic Paerns of Home Ownership by Jørgen Lauridsen, Niels Nannerup and Moren Skak Discussion Papers on Business and Economics No. 9/2006 FURTHER INFORMATION Deparmen of Business and Economics Faculy of Social Sciences Universiy of Souhern Denmark Campusvej 55 DK-5230 Odense M Denmark ISBN 87-91657-08-3 Tel.: +45 6550 3271 Fax: +45 6615 8790 E-mail: lho@sam.sdu.dk hp://www.sam.sdu.dk/deps/virkl/abou.shml
Dynamic and Geographic Paerns of Home Ownership (**) Jørgen Lauridsen(*), Niels Nannerup, Moren Skak Deparmen of Business and Economics, Universiy of Souhern Denmark (*) Corresponding auhor: Campusvej 55, DK-5230 Odense M, Denmark, Phone +45 6550 2142 E-mail jl@sam.sdu.dk (**) Aknowledgemen: The paper is wrien as a par of he Cener for Housing and Welfare - RealDania Research Projec and he CARINAS projec. Economic suppor from RealDania is acknowledged Preliminary work please do no quoe wihou permission 1
Absrac. Deerminaion of he demand for home ownership is analysed. Deerminans include prices and shor- and medium-erm price changes, public regulaion (regulaion of house ren, housing subsidies, axaion), compeiion from alernaive residence forms (measured by supply of subsidized housing), social composiion of populaion (age, social benefi receivers, household composiion, civil saus, educaion, naionaliy), economic abiliy (income), and congesion (measured by populaion densiy and degree of urbanisaion). Danish aggregae daa for 270 Danish municipaliies, available annually for he period 1999-2004. The sudy applies a spaially adjused SUR approach, so ha dynamic as well as spaial paerns are conrolled for simulaneously. I is revealed ha ignorance of conrolling for spaial spillover srongly skews conclusions regarding effecs of deerminans, as deerminaion of housing marke behaviour is no resriced wihin municipaliies, bu raher spills over across municipaliies. JEL Classificaion: C21; C33; P25; R21; R31. Keywords: Housing marke; Demand for home ownership; SUR; Spaial spillover; spaial auoregression; spaial disribued lag Word coun: 6165. 2
1. Inroducion Over he las decades here has been growing aenion abou he aggregae home ownership rae in indusrialized counries. Recen economic sudies of deerminans for home ownership emphasize he complexiy of he basic housing enure choice beween owning and rening whereas earlier ime series sudies such as Rosen e al (1984) and Henderson and Ioannides (1983) mainly focused on user cos of home ownership and rens, and he flucuaions over ime in hese variables as decisive facors for variaions in home ownership. More recen empirical work poins o a range of economic as well as socio-economic and demographic facors as imporan deerminans for he aggregae home ownership rae. Raher han consiuing an alernaive explanaion in undersanding enure choice, however, such facors can be seen as complemenary o he radiional user-cos facors. One reason for he increased aenion in home owning is he generally higher focus of porfolio analysis of privae households in hese years, and hus a broader porfolio choice explanaion has emerged o enure choice ha basically considers real esae asses as an inegraed par of porfolio invesmen in households. This issue is sudied in he conex of price hedging, ren risk and income risk under various individual and srucural condiions. See for example Brueckner (1997), Goezmann (1993), Oralo-Magne and Rady (2002), and Sinai and Souleles (2003). Sinai and Souleles (2003) provide a summary of he lieraure on he relaion beween porfolio choice and enure choice. In addiion, wihin social capial research here is ineres in idenifying he linkages beween he physical environmen and social ineracions of individuals, and among oher issues he impacs of home ownership on social connecion is examined. For more reasons home owners appear o have a high sake in proecing he local communiy and hey herefore pu more effor in he upkeep and appearance of a neighbourhood. For insance, home owners, unlike reners, have made a financial 3
invesmen in he dwelling and hey also appear more saionary han reners. This again may lead o aciviies and behaviour ha serves o reduce vandalism, hef and oher crimes in he area and generally increases social ineracion and responsibiliy beween residens. Several sudies documen a range of posiive effecs from ownership. They include, among ohers, Glaeser and Sacerdoe (1999), Perkins e al. (1996), Rohe and Basolo (1997) and Whie (2001). The presen paper can be seen as exending he lieraure on demand for home ownership by accouning for an undersudied elemen in empirical analysis of home ownership. All empirical sudies of home ownership raes (a leas o he knowledge of he auhors) ignore imporan issues relaed o spaial variaion of he daa applied. I is well known from sudies concerning small area variaion ha i is necessary o conrol for spaial spillover in order o obain proper conclusions regarding effecs of deerminans (Anselin, 1998; Anselin and Bera, 1998; Anselin, 2000). This evidence ypically perain o housing sudies insofar as housing markes are no resriced o ac wihin he borderlines of single small areas. Raher, marke condiions and marke behaviour of coniguous areas may be expeced o spill over beween he jurisdicions. Thus, solely assuming he housing marke behaviour of a (small area) o be condiioned on he deermining facors of his small area alone may well lead o skewed conclusions. Specifically, i is he purpose of he presen sudy o relae our resuls o he findings in Lauridsen e al. (2006). This sudy esablished an economeric model for he fracion of homes ha are owner occupied in Denmark for he period 1999 o 2004. Theoreical deerminans included prices and shor- and medium-erm price changes, public regulaion (regulaion of house ren, housing subsidies, and axaion) along he lines of he user cos approach. Moreover he sudy esed for facors such as compeiion from alernaive residence forms (measured by supply of subsidized housing), social composiion of populaion (age, social benefi receivers, household composiion, 4
civil saus, educaion, and naionaliy), economic abiliy (income), and congesion (measured by populaion densiy and degree of urbanisaion). Issues relaed o he applicaion of pooled cross secional daa were furher discussed. Lauridsen e al. (2006) included parameric insabiliy over ime, adjusmen for dependency caused by repeaed observaion, and idenificaion of he effec of prices on home ownership raes. I was found ha parameric insabiliy over ime could be mosly ascribed o ime rends in he parameers so ha a simplified specificaion wih common parameers across ime, combined wih ime ineracions, could be esablished. By analysing he deerminans of home ownership while conrolling simulaneously, no only for dynamic paerns as done by Lauridsen e al (2006), bu also for spaial spillover effecs, he presen analysis provides an opporuniy o formalise and analyse geographical aspecs of home ownership in a small area seup. One such aspec is endogenous spillover, which implies ha high ownership raes in one area induce high ownership raes in neighbourhood areas. Anoher aspec is exogenous spillover, which implies ha facors deermining he home ownership rae in one area also affec home ownership in surrounding areas. Par 2 of he presen sudy briefly summarises he heoreical foundaions from Lauridsen e al. (2006) regarding deerminans of home ownership raes. The applied daa are briefly described in Par 3, upon which Par 4 ouline mehodological aspecs. As pooled daa are applied, a Seemingly Unrelaed Regression (SUR) framework is advocaed in order o capure dynamic paerns efficienly. Furher, poenial spaial spillover is conrolled for by exending he SUR wih spaial auoregression (SAR) and spaial disribued lag (SDL) specificaions. Nex, Par 5 oulines he esimaion resuls and hroughou demonsraes he fallacies of no simulaneously conrolling for dynamic paerns and spaial spillover, as conclusions regarding effecs on home ownership raes of 5
deerminans varies heavily across he adjused and non-adjused specificaions. Finally, Par 5 rounds off wih a few commens and suggesions. 2. Demand for owner occupied homes in Denmark We address he basic household choice of owning or rening a residence, focusing on home ownership by considering he demand for owner occupied residenial unis relaive o oal demand for hese unis. Based on Danish daa, empirically significan deerminans for his fracion are idenified. Theoreically speaking, a household choose o own a dwelling if owning is he oucome of is uiliy maximizaion given specific economic condiions for his household. The following discussion on deerminans for home ownership draws on heoreical findings by among ohers Linneman (1986), Rohemberg e al (1991), Hansen and Skak (2005), and Elsinga and Hoeksra (2004). 1 In general, house prices and propery values impac on ownership raes as mainly low income groups may be expeced o reduce or delay demand for ownership occupaion when rising prices occur. As price changes over longer periods also lead o changing price expecaions his may again affec demand of dwellings, he consequence being ha here is no unique relaionship beween owner occupaions and house prices, bu raher disincive shor erm and a medium o long erm relaionships. Various forms of governmen inervenion in housing markes via axaion and subsidizaion end o capialize in marke prices and may as well affec relaive price expecaions for owned and rened dwellings. These inervenions may herefore play a significan role for he choice of enure ype. The mos imporan ax and subsidy measures in he Danish housing marke will be esed for 1 For a broader discussion of he heoreical basics see Lauridsen e al (2006). In addiion Aerhög (2005) surveys recen empirical sudies on home ownership deerminans from various counries. 6
direcly or indirecly as explanaory facors in he analysis. We hus include he real propery ax rae. Furher, as ren subsidies are commonly offered o reners, we es for he influence of his policy by considering he share of households and he share of he populaion receiving ren subsidies. In indusrialised counries morgage loans ypically consiue he major share of real esae finance. Morgage erms and credi raing of households by lender insiuions are likely o depend on a variey of individual characerisics. Chiuri and Japelli (2004) provide empirical evidence from 14 counries ha he morgage availabiliy affecs home ownership disribuion across age groups primarily due o income differences beween he groups. Furher, Canner and Smih (1991) find ha ehniciy maers for morgage availabiliy. Oher facors ha may affec credi raing are educaional level and job perspecives. In he analysis we es for such characerisics. Comparing advanages and disadvanages of ownership relaive o rened dwellings may help idenifying furher poenial deerminans for he analysis. In more respecs here are addiional coss of owning raher han rening. The heoreical lieraure poins o disadvanages of owners as o swiching coss of moving (salaries o real esae agencies and lawyers, uncerainy abou sales prices ec.) which hereby cause relaively low geographical mobiliy of owners. This indicaes ha individuals being more inclined o move (such as younger people, unmarried people, younger couples wihou children) may choose rened dwellings. In addiion Linneman (1986) poins ou ha high producion efficiency by landlords (i.e. as o mainenance coss) in high densiy areas is an imporan reason why ownership raes end o fall from counry side o urban areas. We seek evidence for his hypohesis by esing he significance of populaion densiy. One advanage of ownership is he wide scope for individual adapaion of he residence, and households clearly pu differen value on such an opion. Preferences for housing auonomy may differ wih age and career 7
posiion as younger couples plan o have (more) children and educaed people expec increasing fuure income. Furher, one can argue ha self employed people may be more individualisic oriened han employees and for ha reason prefer home ownership. I appears from hese reasonings, however, ha incenives for choosing enure ype are mixed for some of hese groups. For insance, while younger couples may evade swiching coss of moving by being reners hey may on he oher hand prefer ownerships for reasons of housing auonomy (which in some sense can provide he same services as obained from moving o a new residence). The same argumens in principle also apply o divorced people (anicipaing fuure marriage). In all hese consideraions lead us o es he variables shown in Table 1 for empirical significance in explaining home ownership raes in Denmark. [Table 1 around here] 3. Daa The daa o be applied are aggregae cross secion daa observed for 270 Danish municipaliies (5 municipaliies on he island of Bornholm were omied due o daa problems) annually from 1997 o 2004. Daa were colleced from five sources: The Saisical Bank a Saisics Denmark, he Key Figure Base [Nøglealsbasen] a he Minisry of he Inerior, he Minisry of Urban and Housing Affairs (2000) repor on regulaion of housing rens, and he Danish Tax Auhoriy s [Told & Ska] (2004) repor on propery sales prices. Table 2 presens an overview of he daa applied, including variable shor-hands, definiions and a few descripive saisics. [Table 2 around here] 8
Figure 1 shows he disribuion of he variables (averaged over years) across municipaliies. Several indicaions of spaial clusering are observed for he home ownership raes as well as for he explanaory variables. Thus, spaial spillover may poenially be presen and needs o be adjused for. [Figure 1 around here] 4. Mehodology The poin of deparure is he linear regression model defined for he N=270 municipaliies by (1) y = β + υ, υ ~ N(0, σ 2 I) X where X is an N by K dimensional marix of K explanaory variables, y an N dimensional vecor of endogenous observaions, and β a K dimensional coefficien vecor. While pooled daa for T years are applied, he residuals beween years are correlaed, and he variances wihin each year will vary across years, i.e., beween any wo years, he residual covariance reads as (2) E 2 ( ' υ s ) = σs υ, s = 1,.., T. Thus, o obain efficien esimaes of β, Lauridsen e al. (2006) applied Feasible Generalised Leas Squares (F-GLS) esimaion o obain he Zellner (1962) Seemingly Unrelaed Regression (SUR) esimaes forβ. Furher, o allow for variaion of β across years, hey furher added ineracion erms beween some of he X variables and a ime rend T. As he model is esimaed wih regional daa, dependencies beween he cross-secions have o be aken ino accoun. I is inuiively clear ha he housing marke is no resriced o realise iself 9
wihin a single municipaliy, bu raher flows over he municipaliy borderlines. Operaionally, he home ownership rae ( y ) may no only be deermined by he explanaory variables in he municipaliy iself ( X ), bu also by values of X in he surrounding municipaliies, i.e., exogenous spaial spillover may occur. Furher, if he demand for home ownership in he surrounding municipaliies is high, his demand may spill over and induce demand in he municipaliy in quesion, i.e. endogenous spaial spillover may occur. Alike any oher omission of relevan variables, ignorance of spaial spillover may bias he resuls obained (Anselin, 1988). Tradiionally, conrol for spaial spill-over is obained by adding spaial parameers o he model in quesion (Paelinck and Klaassen, 1979; Cliff and Ord, 1981; Anselin, 1988; Florax, 1992; Anselin, 2000). Operaionally, endogenous spaial spillover may be conrolled for by adding he average of y in he neighbourhood municipaliies (denoed by spaially auoregressive (SAR) specificaion (Anselin, 1988) W y ) as an explanaory variable in (1) o obain he W (3) y = y λ + X β + υ, where λ is a parameer specifying he magniude of spill-over, formally resriced o he inerval beween (-1) and (+1), bu for mos pracical purposes resriced o be non-negaive. Likewise, exogenous spaial spillover may be conrolled for by adding he averages of neighbourhood municipaliies (denoed W X spaially disribued lag (SDL) specificaion (Florax, 1992) X in he ) as explanaory variables in (1) o obained he W (4) y = X β + X δ + υ, while boh ypes of spillover may be conrolled for simulaneously by simply adding boh W y and W X o obain a combined SAR-SDL specificaion. One furher approach commonly applied is o 10
defer he enire maer o be a residual spaial spillover by respecifying (1) as he spaially auocorrelaed (SAC) specificaion (Anselin, 1988) W (5) y = X β + ε, ε = λε + υ, bu his specificaion is merely a special case of he SAR-SDL obained by imposing resricions on he λ and δ parameers, and will herefore no be applied here. All hese specificaions are easily inegraed wih he SUR framework, so ha spaial spill-over and dynamic paerns are conrolled for simulaneously (Anselin, 1988; Florax, 1992) 2. Finally, an issue relaed o idenificaion was considered by Lauridsen e al. (2006): The prices of home ownership depress he demand for owner-occupied housing. A he same ime a shif in he demand funcion will affec he equilibrium prices in he same direcion as he shif. Hence, prices and home ownership are simulaneously deermined, so ha any of he above esimaion will yield biased resuls (Greene, 2003). A proper soluion is o use insrumenal esimaion (Greene, 2003) where a supply-side variable is applied as an insrumen for prices. As such an insrumen, Lauridsen e al. (2006) applied he amoun of finished buildings per capia. The presen sudy will follow his approach hroughou and hus apply insrumenalised prices. To provide devices for comparison of alernaive models, some quaniies are applied. One is a 2 pseudo-r-square ( R ), calculaed as he square of he correlaion beween y and is prediced values. This measure is readily calculaed for he SUR and he SUR-SDL models, bu i is no 2 One difficuly of hese spaial SUR specificaions is relaed o esimaion. While he SDL adjused SUR specificaion W can be consisenly esimaed using he F-GLS procedure by simply adding X o he explanaory variables, he SAR and he SAR-SDL adjused SUR specificaions canno be esimaed consisenly by he F-GLS due o he conemporaneous correlaion among he observaions in y (Anselin, 1988). Consisen esimaes are obained using he following Maximum Likelihood approach: We did a grid search of he relevan values of λ from -1 o +1. W Condiioned on each λ, he F-GLS procedure was performed using ( y λy ) insead of y. Finally, he se of resuls which maximized he log likelihood funcion (Anselin, 1988) was seleced. 11
defined for he SUR-SAR specificaion. A second device applied is he familiar Akaike Informaion Crierion (AIC) calculaed as (-2LogL + 2K). Finally, nesed models are esed agains each oher using Likelihood Raio (LR) es, calculaed as wice he difference beween he values of he log likelihoods of he wo models. 5. Resuls Table 3 repeas in he second column he SUR model esablished by Lauridsen e al. (2006). I is especially noiced ha he effecs of prices as well as of shor- and medium-erm price changes have he expeced signs, alhough he price is insignifican. As demonsraed by Lauridsen e al. (2006), some of he deerminans exered an effec on home ownership rae which decreased or increased hrough years. For subsidized housing, he impac is negaive bu gradually reduced hroughou he period from 1999 o 2004 as indicaed by he ineracion erm wih ime. The impac of housing subsidies is posiive in he beginning of he period bu gradually moves oward significanly negaive hroughou he period as shown by he ime ineracion. Ren subsidies has a negaive impac as expeced, and he ineracion wih ime show ha his effec is gradually srenghened in he period from 1999 and onwards. Ren regulaion has he expeced negaive impac, bu his impac is gradually reduced hrough he period as illusraed by he ineracion erm. For urbanisaion and proporion of divorced, he effecs are negaive, bu heir ineracion erms wih ime illusrae ha hey significanly reduce oward zero over ime. Considering proporion of 7-16 year olds, he effec is gradually moving from insignificanly posiive/negaive o significanly negaive during he period as shown by he ineracion wih ime. Thus, hese demographic variables share a common feaure of having an effec on home ownership rae, which is significan, bu mosly reduced in magniude during he period. The remaining deerminans were no found o exer ime-varying effecs on home ownership. Wih a few excepions, he signs of hese effecs 12
corresponded o prior expecaion. As excepions, propery axes and proporion of unemployed exered significanly posiive effecs, while proporion wih furher educaion exered a negaive effec. [Table 3 around here] To benchmark he SUR resuls, a simple OLS, which is no adjused for dynamic paerns alike he former, is provided in he firs column of Table 3. Wih he excepion of proporion of social disabiliy pensioned, for which he effec changes from significanly negaive o significanly posiive, none of he effecs exer a significan shif in sign, bu i is eviden ha several deerminans loose explanaory significance. Furher, he log likelihood as well as he AIC and he LR es srongly favour he SUR for he OLS specificaion. Thus, he inefficiency occurring if ignoring dynamic paerns is clearly demonsraed. Nex, aenion is urned o he spaially adjused models, which are repored in he remainder of Table 3. For he SAR-SUR (3), which conrols for endogenous spaial spillover by adding a spaial lag of home ownership raes o he SUR specificaion, i is firs of all noiced ha he LR es for he SAR-SUR versus he unadjused SUR (2) does no rejec he laer in favour of he former. This conclusion is suppored by he insignificance of he spaial lag of home ownership raes and by he log likelihood and AIC values, which are pracically equal for he wo specificaions. Furher, a quick look hrough he esimaed effecs of deerminans reveals ha neiher he signs nor he magniude or he significance levels of hese deviae across he wo specificaions. A differen picure is obained when comparing he unadjused SUR (2) o he SDL-SUR (4) which adjuss for spaial spillover of he deerminans (i.e. he erms W X of equaion (4)). The LR es of (4) versus (2) srongly rejecs he unadjused SUR in favour of he SDL-SUR. This conclusion is 13
suppored by he log likelihood and AIC values, which are larger and smaller, respecively, for he SDL-SUR han for he unadjused SUR. A closer look a he effecs of deerminans, repored in he firs column of (4), reveals subsanial differences as compared o he unadjused effecs of (2). I is especially noiced ha he effecs of shor- and medium-erm price changes are no longer significan. Thus, he effecs of local price dynamics seem o be overesimaed when ignoring conrol for spaial spillover occurring from he supra-municipal naure of he housing marke. On he oher hand, several effecs of deerminans capuring populaion srucure seem o be overesimaed by he unadjused SUR. This is especially he case for populaion proporions of divorced, educaed, early reired, social benefi receivers, and inhabians from hird counries. Thus, i is clearly demonsraed ha ignorance of conrolling for spaial spillover (occurring from a discrepancy beween he small area or inra-municipal naure of he daa and he large area or supramunicipal naure of he housing marke) leads o seriously skewed conclusions regarding effecs of imporan deerminans of home ownership raes. The second column of he SDL-SUR (4) repors he effecs of spaial lags of deerminans. Though i should be kep in mind ha hese effecs are no of explici ineres, bu raher added as conrols o ensure proper conclusions regarding effecs of he deerminans, ineresing informaion regarding he funcioning of he housing marke may be obained. Thus, a weakly posiive effec of spaially lagged prices is found. This indicaes ha high prices in he surrounding municipaliies may ceeris paribus lead o increased homeownership raes in he municipaliy. An alike posiive spillover effec is exered by high propery axes in surrounding municipaliies. As he direc propery ax effec is posiive, he posiive spill-over reflecs some clusering of municipaliies wih relaively high home ownership. Applying municipal cross secion daa, his clusering phenomena is no surprising given ha planned localisaion of owned dwellings in he pas (mainly one-family house areas) as well as he localisaion of aparmen complex areas has no followed municipal borders of 14
oday. The posiive spill-over effec could also be he resul of a endency a local policy-makers (which in Denmark decide on real propery axes) o follow he levels of axes imposed in he region he purpose being o pre-emp ax evasion from own locals ino surrounding locals. The significanly posiive spillover exered by he populaion proporions of divorced and inhabians from hird counries appear less inuiive, given ha he direc effec on homeownership of hese variables are boh negaive. As o he divorce facor, a possible explanaion appears from a combinaion of he facs ha home owners geing divorced on he one hand end o give up ownership (parly for direc economic reasons as wo formerly muual incomes need o finance wo fuure households, and parly for lowering expeced fuure residenial swiching coss following from a higher likelihood of a fuure change from single saus o new marial saus). On he oher hand, people s preferences for saying in a region afer divorce may remain and hey herefore prefer o sele down in less expensive rened dwellings in nearby areas. This would explain he counervailing spaial effec on he direc divorce effec. A similar movemen paern for immigrans may in fac be behind he posiive spaial spillover for 3 rd counries. The Danish Inegraion Ac give municipaliies responsibiliy for inegraion of refugees, and for he purpose of achieving a more equal disribuion of refugees among municipaliies, newly arrived refugees are generally required o remain in heir allocaion municipaliy in a hree-year inroducion period. Presumably o follow ehnic nework, here is however a clear endency for refugees o move o rened dwellings in nearby larger ciies wih larger populaions of immigrans 3. To round off he analysis of spaial spillover, he SAR-SDL-SUR (5) conrols for endogenous as well as exogenous spillover simulaneously. Looking hrough he effecs of he deerminans as well as heir spaial lags reveals no pracical differences as compared o he SDL-SUR. Thus, he 3 See Nielsen and Blume (2006) for a horough saisical descripion of selemen paerns and mobiliy of refugees in Denmark. 15
conclusions from he reamen of he SAR-SUR and he SAR-SDL separaely are confirmed, i.e. i is spaial spillover of deerminans raher han endogenous spillover which maers when focus is on proper evaluaion of deerminans of home ownership raes. On he oher hand, he effec of endogenous spillover is significanly posiive in he SAR-SDL-SUR, which i was no in he SAR- SUR, and he LR es of (5) versus (4) as well as he log likelihood and AIC values slighly prefer he SAR-SDL-SUR for he SDL-SUR. Thus jus as a bypass i may be seen ha in order o obain a proper picure of spaial spillover effecs, one has o conrol for spaial spillover. However, wheher he full SAR-SDL-SUR or he resriced SAR-SDL specificaion is preferred, he cenral message is unchanged: The necessiy of conrolling for spaial spillover, in order o obain proper conclusions regarding he effecs of deerminans on home ownership raes, is urgen. 6. Conclusions The presen invesigaion of deerminans of home ownership raes adds o previous knowledge and suggess revision of radiional modelling pracice (like applicaion of OLS esimaion). I is confirmed ha adjusmen for iner-emporal residual correlaion and heerogeneiy is essenial in order o obain efficien esimaion of he effecs of explanaory characerisics on home ownership raes, when applying pooled small area cross secions. Furher, he necessiy of conrolling for spaial spill-over effecs is demonsraed. Endogenous spaial spill-over is found o be of some bu less relevance, while exogenous spaial spill-over maers seriously. Especially, he effecs of price dynamics seem o be heavily overesimaed if conrol for spaial spill-over is ignored, while he effec of several populaion characerisics seem o be srongly underesimaed. Finally, i is demonsraed ha he spaial spillover effecs in hemselves may conain valuable informaion regarding he funcioning of he housing marke. In paricular, he spaial lags of prices were found o induce demand for home ownership. Thus, he urgency of simulaneously adjusing for he cross 16
secional naure as well as he dynamic properies of he daa when analysing (large area) housing marke behaviour using pooled (small-area) cross secion daa is clearly illusraed. 17
References Anselin L 1988: Spaial economerics: Mehods and models. Norh-Holland: Kluwer Academic Publishers Anselin L 2000: Spaial Economerics. In Balagi B (ed) A Companion o Theoreical Economerics. Oxford: Blackwell Publishers Anselin L and Bera A 1998: Spaial dependence in linear regression models wih an inroducion o spaial economerics. In Ullah A, Giles D (eds) Handbook of applied economic saisics. New York: Marcel Dekker Aerhög, Mikael (2005). Imporance of Governmen Policies for Home Ownership An Inernaional Survey and Analysis, Working paper, Building and Real Esae Economics, Royal Insiue of Technology, Sockholm. Brueckner, Jan 1997: Consumpion and Invesmen Moives and he Porfolio Choices of Homeowners, Journal of Real Esae Finance and Economics Vol 15, pp.159-180. Canner, G. and D. Smih 1991: Home Morgage Disclosure Ac: Expanded Daa on Residenial Lending. In: The color of credi. Morgage discriminaion, research mehodology, and fair-lending enforcemen. (Eds, Ross, S. and Yinger, J.), Cambridge: The MIT Press. Chiuri, M. C. and T. Jappelli 2003: Financial marke imperfecions and home ownership: A comparaive sudy. European Economic Review, 47:857-875. Cliff A, Ord J 1981: Spaial Processes, Models an Applicaions. London: Pion 18
Danish Tax Auhoriy 2004: Ejendomssalg 1. halvår 2004 [Propery sales 1. half-year 2004]. Copenhagen: Danish Tax Auhoriy Elsinga M. and Hoeksra, J. 2004: Homeownership and housing saisfacion: a sudy of he lieraure and an analysis of he European Communiy Household Panel, ENHR Conference Paper, Cambridge, 2-6 July. Florax RJGM 1992: The Universiy: A Regional Booser? Economic Impacs of Academic Knowledge Infrasrucure. Avebury, Aldersho Glaeser, E: L. and B. Sacerdoe (2000). The Social Consequences of Housing, Journal of Housing Economics 9, pp. 1-23. Goezmann, William 1993: The Single Family Home in he Invesmen Porfolio. Journal of Real Esae Finance and Economics, vol. 6 (3), pp. 201-22. Greene WH 2003: Economeric Analysis. Fifh Ediion. NJ: Prenice-Hall Gyourko J 2003: Access o Home Ownership in he Unied Saes: he Impac of Changing Perspecives on Consrains o Tenure Choice. In O Sullivan, Tony and Kenneh Gibb. Housing Economics and Public Policy. Oxford: Blackwell Science Ld Hansen JD and Skak M 2005: Economics of Housing Tenure Choice. Working paper. Odense: Universiy of Souhern Denmark, Deparmen of Business and Economics Henderson, J.V. and Ioannides, Y.M. 1983: A Model of Housing Tenure Choice, American Economic Review, Vol. 73 (1), pp. 98-113. 19
Lauridsen J, Nannerup N, Skak M 2006: Explaining Home Ownership Raes in Danish Municipaliies. In Doling J, Elsinga M (eds) Home Ownership: Geing In, Geing From, Geing Ou, Par II. Delf: Delf Universiy Press (forhcoming) Lindberg G and Lindén A-L 1989: Social segmenaion på den svenska bosadsmarknaden [Social segmenaion on he Swedish housing marke]. Lund: Universiy of Lund, Deparmen of Sociology Linneman P 1986: A New Look a he Homeownership Decision. Housing Finance Review. 5: 159 87 Minisry of Social Affairs 2004: Redegørelse fra Ekspergruppen vedr. Lejelovskommisionens modererede lejelovsmodel [Repor from he Task Group conc. The Ren Ac Commission s Modified Ren Ac Model]. Copenhagen: The Minisry of Social Affairs Minisry of Urban and Housing Affairs 2000: Huslejen 1999 [The house ren 1999]. Copenhagen: Minisry of Urban and Housing Affairs Munch JR, Rosholm M and Svarer M 2003: Are Home Owners Really More Unemployed? Working Paper 2003:3. Copenhagen: The Economic Council Nielsen C. P. and Jensen, K. B. 2006: The Danish Inegraion Ac s Significance for he Selemen Paerns of Refugees, Repor from Insiue of Local Governmen Sudies (AKF), Copenhagen. Oralo-Magné F and S. Rady 2002: Tenure choice and he riskiness of non-housing consumpion. Journal of Housing Economics, 11, 266-79. Oswald A 1997: Thoughs on NAIRU. Journal of Economic Perspecives, 11: 227-28. Paelinck J, Klaassen L 1979: Spaial Economerics. Farnborough: Saxon House 20
Perkins, D.D., B.B. Brown, and R.B. Taylor 1996. The Ecology of Empowermen: Predicing Paricipaion in Communiy Organisaions, Journal of Social Issues, 52, pp. 85-110. Ren Ac Commission [Lejelovskommissionen] 1997: Lejeforhold [Tenancy]. Beænkning nr 1331 [Repor no. 1331]. Copenhagen: The Minisry of Social Affairs Rohe, W.M. and V. Basolo 1997. Long-erm effecs of homeownership on he self-percepions and social ineracion of low-income persons. Environmen and Behaviour, 29, 793-819. Rosen, Harvey, Kenneh Rosen and Douglas Holz-Eakin 1984: Housing Tenure, Uncerainy, and Taxaion, Review of Economics and Saisics, vol. 66, no 3, pp. 405-416. Rohemberg J, Galser GC, Buler RW and Pikin J 1991: The Maze of Urban Housing Markes. Theory, Evidence, and Policy. Chicago: The Universiy of Chicago Press Sinai, T. and N.S. Souleles 2003: Owner-Occupied Housing as a Hedge Agains Ren Risk, Working paper, The Wharon School, Universiy of Pennsylvania. The Economic Council 2001: Dansk Økonomi, forår 2001 [Danish Economy, Spring 2001]. Copenhagen: The Economic Council Whie, Garland F. 2001: Home Ownership Crime and he Tipping and Trapping Processes, Environmen and Behavior, Vol. 33, No. 3, May 2001, pp. 325-342. Zellner A 1962: An Efficien Mehod of Esimaing Seemingly Unrelaed Regressions and Tess of Aggregaion Bias. Journal of he American Saisical Associaion, 58: 977-92 Ærø T 2002: Boligpræferencer, boligvalg og livssil [Housing preferences, housing choice and life syle]. Hørsholm: Danish Building Research Insiue 21
Table 1. Variables affecing home ownership raes Variable Prices Acual price (-) One-year price change (-) Three-year average price change (+) Favourable ax reamen of home owners Tax bracke (+) Ren subsidy (-) Ren conrol (-) Urban resricion on ownership (-) Financial capaciy Income (+) Naionaliy (?) Educaional level (+) Oher personal characerisics Special life evens (e.g. divorce, beques, loery) Expeced occupaion ime Age (-) Rae of under educaion (-) Job ype Producion efficiency for landlords vs. owner-occupiers Congesion (-) Households differs in benefi from adaping heir home Self employed (+) More han one child (+) High ren area (-) Social heriage Parens enure choice Lifesyle Rae of single households (-) Explanaion High prices and shor-run price increases make i difficul o buy home. Medium-erm price changes simulae he expecaion of fuure increasing prices and hus he propensiy o buy. A favourable ax reamen riggered by ownership ends o raise ownership raes; such reamen, e.g. a low impued ren, is ypically more valuable for higher income ax brackes. Home ownership raes are reduces if an income subsidy is riggered by rening vs. owning. If ren conrol arificially keeps he ren on rened homes below he marke equilibrium his also reduces demand for owned housing. If, e.g. for social reasons, only a fracion of homes can be owned, his poenially reduce home ownership raes. Wih asymmeric informaion on financial markes, various indicaors of borrowers (home owners) repaymen abiliy will influence home ownership raes. Ownership sars wih closing or conracing coss ha have o be balanced agains benefis in each occupaion year. If he expeced number of occupaion years is low, ownership raes end o fall. Expeced occupaion years may also fall wih some job ypes. Where many live ogeher landlord scale economies for producion of housing services may be pronounced. Idiosyncraic variaions in he benefi households or individuals ge from individual adapaion of homes leads o a marke screening where owners benefi mos. High rens reduce ne benefi mos for owners and squeeze some owners ino reners. People end o demand he ype of dwelling hey used o live in as child. Modes of living, e.g. free single life vs. ied family life influence ownership raes. Noe: A (+) indicaes a posiive correlaion beween he variable and he home ownership rae. 22
Table 2. Daa applied Variable Definiion 25% quarile Median 75% quarile Home ownership % of housing unis occupied by owner (cooperaive housing and suden hosels 62.00 71.00 76.00 (dependen variable omied) (1) Price Average sales price (real DKK) per square meer of one-family houses (4) 51.86 55.73 73.48 Shor erm price change Defined as (Price i, Price i,-1 ) / Price i,-1 0.034 0.055 0.079 Medium erm price change Defined as (Price i, Price i,-3 ) / Price i,-3 0.095 0.225 0.285 Subsidized housing of populaion living in subsidized housing [almennyige boliger] (2) 5.00 9.00 17.00 Housing subsidy % of households receiving housing subsidies [boligydelse] (2) 8.90 10.90 13.25 Ren subsidy % of 15-66 year old receiving ren subsidies [boligsikring] (2) 2.90 4.00 5.90 Regulaed Ren Regulaion Ac assumed by 2000 (1=yes, 0=no) (3) Proporion yes =0.556 Propery ax Real Propery Tax (in 0/00) [Grundskyldspromille] (2) 8.00 12.00 15.00 Tax rae Municipal + couny ax rae (in %) [Udskrivningsprocen] (2) 20.20 20.80 21.30 Tax base Tax base [beskaningsgrundlag] per inhabian (100.000 DKK) (2) 9.94 10.97 12.10 Populaion densiy Inhabians per square kilomere (10000 (2) 48 69 147 Urbanisaion % of populaion living in urban areas (2) 61.00 71.00 86.00 7-16 year % of populaion aged 7-16 (1) 11.90 12.90 13.90 17-25 year % of populaion aged 17-25 (1) 8.07 9.09 10.21 26-35 year % of populaion aged 26-35 (1) 11.74 12.82 13.89 36-66 year % of populaion aged 36-66 (1) 40.55 42.33 44.27 67+ year % of populaion aged 67 and over (1) 12.00 13.50 15.00 Widowed % of populaion widowed (1) 5.91 6.61 7.37 Divorced % of populaion divorced (1) 4.86 5.82 7.40 Unmarried % of populaion unmarried (1) 41.91 43.54 44.80 Adul children % of households wih children over 18 (1) 7.68 8.76 9.88 No children % of households wihou children under 18 (1) 0.00 3.06 5.62 Educaed % of populaion wih higher educaion (2) 11.50 13.60 16.45 Social Disabiliy Pension % of populaion on social disabiliy pension [føridspension] (2) 6.25 7.40 8.80 Social Benefi receivers % of populaion receiving social benefis [konanhjælp] (2) 6.70 8.00 9.60 Unemployed % of populaion (17-66 year) unemployed (2) 3.60 4.40 5.40 3rd counries Number of ciizens from counries ouside EU, Scandinavia and Norh America per 10.60 15.70 23.60 10,000 inh. (2) Finished new buildings Finished new buildings (m 2 per capia) (1) 0.84 1.37 2.06 Sources: (1) Saisics Denmark; (2) The Key Figure Base; (3) The Minisry of Urban and Housing Affairs; (4) The Danish Tax Auhoriy. 23
Table 3. OLS and SUR models Variable (1) OLS (2) SUR (3) SAR-SUR (4) SDL-SUR (5) SAR-SDL-SUR Consan 183.72***(14.36) 149.81***(10.44) 150.04***(10.41) 125.88***(22.75) 107.99***(22.97) Price 0.061***(0.014) -0.005 (0.004) -0.005 (0.004) -0.005 (0.004) 0.014* (0.007) -0.005 (0.003) 0.014** (0.007) Shor erm price change -5.220* (2.840) -2.222***(0.589) -2.249***(0.591) -1.521 (0.976) -0.819 (1.105) -1.536 (0.973) -0.476 (1.103) Medium erm price change 5.196** (2.201) 1.678** (0.714) 1.679** (0.715) 1.495 (1.131) 0.848 (1.360) 1.527 (1.128) 0.577 (1.355) Subsidized housing -0.504***(0.038) -0.567***(0.031) -0.569***(0.030) -0.571***(0.031) -0.017 (0.043) -0.562***(0.030) 0.043 (0.045) Housing subsidies -0.732***(0.104) -0.127***(0.047) 0.128***(0.047) 0.127***(0.048) 0.041 (0.058) 0.105** (0.047) 0.043 (0.057) Ren subsidies -0.577***(0.183) -0.115 (0.098) -0.114 (0.097) -0.085 (0.099) -0.092 (0.107) -0.102 (0.098) -0.049 (0.107) Regulaed -1.052** (0.493) -2.725***(0.568) -2.722***(0.565) -3.193***(0.535) 0.227 (0.725) -3.181***(0.530) 0.490 (0.728) Propery ax 0.141***(0.015) 0.036** (0.018) 0.036** (0.017) 0.028 (0.018) 0.098***(0.032) 0.026 (0.017) 0.093***(0.032) Tax rae -0.166** (0.077) -0.036 (0.071) -0.034 (0.070) -0.005 (0.070) -0.017 (0.133) -0.008 (0.069) -0.016 (0.132) Tax base -0.696***(0.065) 0.008 (0.047) 0.007 (0.046) -0.015 (0.050) 0.101 (0.090) -0.014 (0.049) 0.096 (0.089) Populaion densiy -18.25***(1.469) -25.03***(3.066) -25.77***(3.294) -26.78***(5.318) -22.90***(8.775) -26.05***(5.333) -19.92** (8.814) Urbanisaion -0.037** (0.018) -0.099***(0.018) -0.099***(0.018) -0.089***(0.018) 0.035 (0.029) -0.088***(0.017) 0.043 (0.028) 7-16 year -0.190 (0.218) -0.049 (0.117) -0.047 (0.116) 0.011 (0.119) -0.338* (0.205) 0.004 (0.18) -0.313 (0.203) 17-25 year -0.894***(0.147) -0.561***(0.106) -0.558***(0.106) -0.553***(0.107) -0.060 (0.205) -0.551***(0.106) 0.028 (0.204) 26-35 year -0.306 (0.201) -0.391***(0.125) -0.387***(0.125) -0.406***(0.125) 0.294 (0.245) -0.413***(0.124) 0.345 (0.243) 36-66 year -0.485***(0.150) -0.223* (0.115) -0.217* (0.114) -0.261** (0.114) 0.053 (0.224) -0.263** (0.113) 0.084 (0.222) 67+ year -0.654***(0.162) -0.829***(0.126) -0.822***(0.124) -0.907***(0.127) 0.332 (0.250) -0.908***(0.126) 0.437* (0.250) Widowed -0.892***(0.175) -0.353** (0.152) -0.348** (0.151) -0.276* (0.151) 0.034 (0.301) -0.267* (0.149) 0.093 (0.299) Divorced -0.817***(0.195) -0.965***(0.142) -0.969***(0.142) -1.194***(0.146) 0.687***(0.222) -1.162***(0.144) 0.737***(0.220) Unmarried -0.953***(0.091) -0.525***(0.081) -0.520***(0.080) -0.467***(0.082) 0.157 (0.162) -0.474***(0.081) 0.228 (0.162) Adul children 0.483***(0.093) 0.195***(0.049) 0.194***(0.049) 0.178***(0.049) 0.024 (0.098) 0.177***(0.049) -0.014 (0.097) No children 2.507** (1.109) 0.531 (0.410) 0.533 (0.410) 0.278 (0.411) -1.545* (0.874) 0.313 (0.408) -1.561* (0.868) Educaed 0.042* (0.023) -0.035 (0.038) -0.036 (0.037) -0.104** (0.042) 0.098 (0.067) -0.104** (0.042) 0.099 (0.067) Social Disabiliy Pensioned 0.117* (0.061) -0.184** (0.076) -0.179** (0.076) -0.209***(0.079) 0.005 (0.139) -0.212***(0.078) 0.020 (0.138) Social benefi receivers 0.026 (0.054) -0.116***(0.042) -0.114***(0.041) -0.147***(0.042) 0.034 (0.080) -0.151***(0.041) 0.042 (0.079) Unemployed 0.014 (0.073) 0.118***(0.046) 0.119***(0.046) 0.087 (0.054) 0.095 (0.080) 0.082 (0.053) 0.079 (0.079) 3rd counries 0.004 (0.020) -0.020* (0.012) -0.020* (0.012) -0.027** (0.012) 0.054***(0.016) -0.028** (0.011) 0.057***(0.015) Time*Subsidized housing 0.012** (0.006) 0.015***(0.003) 0.015***(0.003) 0.016***(0.003) 0.014***(0.003) Time*Housing subsidies -0.029 (0.017) -0.043***(0.008) -0.043***(0.008) -0.044***(0.009) -0.040***(0.008) Time*Ren subsidies -0.052* (0.029) -0.046***(0.013) -0.046***(0.013) -0.053***(0.014) -0.047***(0.013) Time*Regulaed 0.131 (0.084) 0.160***(0.042) 0.161***(0.042) 0.174***(0.041) 0.164***(0.040) Time*Urbanisaion 0.009***(0.003) 0.003* (0.002) 0.003* (0.001) 0.005***(0.002) 0.004***(0.001) Time*7-16 year -0.001 (0.013) -0.020***(0.007) -0.020***(0.007) -0.024***(0.008) -0.021***(0.008) Time*Divorced 0.056* (0.029) 0.063***(0.014) 0.064***(0.013) 0.063***(0.015) 0.055***(0.014) Time*3rd counries -0.005 (0.004) -0.001 (0.002) -0.001 (0.002) -0.001 (0.002) -0.001 (0.001) Home ownership (spaial lag) -0.012 (0.018) 0.122***(0.026) R-Square 0.961 0.932-0.944 - LogL -2361.73-655.18-656.36-611.97-605.68 AIC 4797.46 1426.35 1428.74 1391.94 1381.37 LR (2)-(5) versus (1) 3413.10*** 3410.74*** 3499.52*** 3512.1*** LR (3)-(5) versus (2) 2.36 86.42*** 99.00*** LR (5) versus (3)-(4) 101.36*** 12.58*** Noe. Significance indicaed by ***(1%), **(5%), *(10%). Sandard errors in parenheses. Second columns of (4) and (5) are coefficiens of exogenous spaial lags. 24
Figure 1. Geographical disribuion of variables (average 1999-2004). 25
Figure 1 (coninued) 26
Figure 1 (coninued) 27