CULS Working Paper Series

Transcription

1 CULS Working Paper Series No. 01 (2015) Green Luxury Goods? The Economics of Eco- Labels in he Japanese Housing Marke BY FRANZ FUERST, CULS FELLOW, UNIVERSITY OF CAMBRIDGE CHIHIRO SHIMIZU, NATIONAL UNIVERSITY OF SINGAPORE 1

2 Green Luxury Goods? The Economics of Eco-Labels in he Japanese Housing Marke BY FRANZ FUERST*, CHIHIRO SHIMIZU** Absrac Using a unique ransacion daabase of condominiums in he Tokyo meropolian area and a hedonic analyical framework, we find ha eco-labelled buildings command a small bu significan premium on boh he asking and ransacion prices. This finding is consisen wih resuls from oher counries bu in conras o hese sudies, he presen analysis also incorporaes buyer characerisics which provide furher informaion on he sources of demand for eco-labelled real esae. A separae esimaion by subgroups reveals ha he price premium is primarily driven by wealhier households ha exhibi a higher willingnesso-pay for eco-labelled condominiums, boh as a oal amoun and as a fracion of he oal sales price. Less affluen households are also shown o pay higher prices for he eco label bu he effec is less pronounced. The resuls indicae ha capialised uiliy bill savings are likely o accoun for a large proporion of he observed premium bu he higher premium paid by affluen households suggess ha more inangible benefis of living in a green building may also play a role. * Universiy of Cambridge, Deparmen of Land Economy, Cambridge CB3 9EP, UK, ff274@cam.ac.uk. ** Naional Universiy of Singapore, Insiue of Real Esae Sudies, 21 Heng Mui Keng Terrace, Singapore , cshimizu@nus.edu.sg. Research ino he profiabiliy of environmenally friendly buildings has reached a criical juncure. The seminal sudies (Miller, Spivey, Florance, 2008, Fuers and McAlliser, 2011, Reichard, Fuers and Ziez, 2012, Eichholz, Kok and Quigley, 2010, 2011 and Eichholz and Quigley 2012 o name jus a few) provided firs valuable insighs ino he pricing of susainable real esae. However, hese sudies are also characerised by imporan limiaions. Firsly, hey ypically focus on specific propery markes in specific counries and over specific ime frames which means ha heir resuls may no be readily generalisable o oher secors, places and ime periods. This is paricularly relevan as he majoriy of sudies were conduced using daa from he US office marke, possibly because of daa availabiliy. Secondly, hese sudies rely on a relaively small number of daa sources (noably from he 2

3 CoSar Group) which provide a grea wealh of informaion on propery characerisics bu are raher limied regarding he environmenal performance and general susainabiliy indicaors. The residenial secor has araced a much smaller number of academic sudies in his opic area, despie is large size and obvious relevance for boh he general economy and susainable developmen. The reasons for his lack of empirical evidence are no clear. Larger fragmenaion of invesors and a lower fracion of professional or insiuional invesmen in he marke driving he discourse around 'green value' of real esae invesmens may be a conribuing facor. Also, housing markes are highly regulaed and prone o inefficiencies in many counries which makes i more difficul o measure he conribuion of susainabiliy and energy efficiency o prices and rens. Despie he widely acceped proposiion ha moneary incenives are more effecive in reducing environmenal harm han command and conrol policies, (Requae and Unold (2003)), he housing marke seems o be lagging behind oher secors in offering an aracive business case for invesmens in susainabiliy and energy efficiency. 'Green' financial insrumens are sill no used widely in he residenial secor which makes capialisaion ino he lump-sum house price he only channel for economic rewards of susainabiliy. As his poses a significan risk for any upfron invesmen in energy efficiency, 'green value' migh no be readily observable in housing markes According o Kochen (2006), green markes can principally be undersood as a form of a privae provision of a public good and as such can have eiher beneficial or derimenal aggregae effecs depending on echnology, individual wealh levels and he iniial level of he public good. This proposiion has been evaluaed empirically, for example by Jacobsen e al (2012).in he conex of residenial elecriciy demand. However, i may be argued ha he privaised public good is a a bes a secondary consideraion of green consumers whose decision may be guided by purely privae cos savings in he form of lower uiliy bills and, equally privae, green signalling benefis o heir social peers, cosumers ec. I seems likely ha any observed green premium primarily reflecs hese privae benefis. Despie hese apparen obsacles, he exising evidence on housing markes poins o a significan green premium. An early sudy by Dian and Miranowski (1989) showed ha invesmens in energy efficiency increase house prices. Banfi e al. (2005) published research findings indicaing ha residenial enans are prepared o pay up o 13% higher ren for buildings ha have adoped energy-saving measures. Similarly, Fuers e al (2015) found a price effec of higher energy performance in he Briish housing marke for a large sample of sales ransacions in he ime period, indicaing a 14% premium of he 3

4 highes band of he Energy Performance Cerificae (EPC) over he lowes band. They also find ha his effec ends o be larger for erraced dwellings and flas compared o deached and semi-deached houses. Earlier, Brounen and Kok (2011) had examined he relaionship beween EPC raings and sale price for 31,993 residenial sale prices in in he Neherlands and repor significan premiums for more energy-efficien buildings. Alhough heir daase conains a large number of conrol variables, he adopion rae of EPCs in he Duch housing marke was relaively low a he ime (7-25% depending on he year) which may limi heir findings. Similarly, Zheng and Kahn (2008) and Zheng, Kahn and Deng (2012) find significan price premia for 'green' properies in he Chinese housing marke and a sudy by Deng, Li and Quigley (2012) finds subsanial economic reurns o green buildings in Singapore. Kok and Kahn (2014) as well as Hyland e al (2013) arrive a similar conclusions for he Californian and he Irish housing marke respecively. This paper examines he 'green value' proposiion in he conex of he Japanese housing marke. Using a unique ransacion daabase of conains roughly 25,000 housing ransacions in he Tokyo condominium marke, we seek o esablish wheher an eco-label carries a significan premium in asking and/or ransacion prices. There are a number of exising sudies on he Japanese marke for green buildings, for example Shimizu (2010) who have conduced an analysis focusing on he new condominium marke using asking prices and ransacion prices. However, due o he small size of heir sales ransacions sample, he resuls did no reach a saisfying level of saisical reliabiliy. A larger sudy was conduced by Yoshida and Sugiura (2015). Using a sample of roughly 35,000 condominiums he auhors find ha eco-labelled condominiums were sold a a discoun, raher han a premium and offer a number of empirical and mehodological explanaions for his resul. A possible reason is ha Japan is already characerised by a high level of energy efficiency in he building and appliances secor as well as he larger economy which may be why a green label carries less currency compared o a marke wih lower average levels of energy efficiency where considerable savings can be expeced from a green produc. Ineresingly, however, hey do find a premium for long-life design, suggesing a willingness o pay for superior durabiliy and slower depreciaion of a building or condominium. 1 The presen sudy seeks o clarify he conflicing findings regarding he Tokyo marke and conribues o he body of evidence by applying he larges and mos comprehensive daase 1This inerpreaion of he privae provision of public goods hypohesis in he ligh of he curren sudy of he Tokyo green housing label was provided by an anonymous reviewer of his paper whom we hank. 4

5 o dae in his invesigaion of 'green value' in he Japanese conex. Crucially, i conains informaion on he propery developmen company as a proxy for qualiy as well as buyer characerisics which were possible omied variables in he earlier sudies cied above. Daa The principal source for he sales ransacions daabase applied in his case sudy is he Tokyo Associaion of Real Esae Appraisers (2010) which collecs ransacion prices for new condominiums and used condominiums. Green labels are currenly only awarded in he Tokyo marke for new builds and no for used condominiums. While our use of new builds only may limi he applicaion of our findings o he general housing marke, i avoids he repored problems arising from any disconinuiies ha may exis in how propery characerisics are priced in he new-build and re-sale markes. The daase was colleced using a survey of house price and aribue informaion. The mos imporan piece of informaion concerns prices per uni, boh he asking price (which is he producer's offer price) and he recorded ransacion price. Furher, in order o ensure consisency wih hedonic heory, daa relaing o buyer characerisics such as income, household size, ec., was gahered. The quesionnaire survey was conduced by he Recrui Housing Insiue, saring in November Surveys were conduced in wriing, via submissions from a large number of home buyers. Conrac daa were also used o collec accurae ransacion prices. In addiion, informaion on freehold/leasehold and he form of managemen were recorded by he quesionnaire survey, i.e. is he building managed hrough visis (called parols ), hrough day shifs (a manager works in he adminisraive office during he day ime only), or by having a permanen presence on sie (a manager works in he adminisraive office and is presen on a 24-hour basis). The inuiion behind gahering his informaion was ha he qualiy and availabiliy of managemen services is said o be refleced in condominium prices. More imporanly, i can also be viewed as a proxy for oher unobservable qualiy characerisics ha migh oherwise be capured by he green label which, in he wors case, could lead o omied variable bias and oversaed green premiums. Sandard hedonic characerisics such as he oal number of condominium unis, lo area, and overall building area were also included. Moreover, we assume ha price differenials may also arise based on he developer's and (main) consrucion company's repuaion and brand power.. 2 2A variable was creaed o disinguish: leading consrucion companies (1) Takenaka Corporaion, (2) Obayashi Corporaion, (2) Kajima Corporaion, (4) Shimizu Corporaion, and (5) Taisei Corporaion; second-ier consrucion companies (6) Kumagai Gumi, (7) Toda Corporaion, (8) Pena-Ocean Consrucion, (9) Konoike Consrucion, (10) Sao Kogyo, (11) Misui Consrucion, (12) Misubishi Consrucion, (13) Sumiomo 5

6 Marke condiions and dynamics are an imporan conrol variable in any hedonic analysis. Therefore, we included he firs-monh conrac rae as a proxy. The firs-monh conrac rae reflecs he percenage of unis sold wihin he firs monh of markeing a paricular propery.. I is hough ha a higher firs-monh conrac rae reflecs beer general marke condiions bu a higher relaive rae (i.e. relaive o he marke average a he ime) also indicaes how affordable he housing uni prices are in relaion o he condominium's feaures. In addiion o he Recrui Housing Insiue's survey daa, he Japanese Real Esae Economic Insiue's daabase was used. Along wih he developer's asking price, he following key variables were drawn from he Real Esae Economic Insiue's daabase: name of he developmen company, developmen overview (developmen scale), locaion characerisics (geo-coordinaes, address, neares saion, disance o neares saion), and building characerisics (building area, land area, building srucure). This informaion was mached o he daa gahered by he Recrui Housing Insiue. Appendix 1 conains a complee overview of he variables used in he analysis. Using hese sources, a large daabase was assembled for he 10-year period from 2001 o 2011 bu he period is disregarded in our analysis as no ransacions of eco-labelled buildings were recorded during ha ime. We idenify green buildings as hose ha are labelled under he Tokyo Meropolian Governmen's Green Labelling Sysem for Condominiums. This scheme is based on he Green Building Program which was inroduced in June 2002 and mandaes ha all largescale consrucion or major refurbishmen projecs exceeding 10,000m 2 submi an environmenal plan a he ime of planning as well as a compleion noice. Addiionally, in Ocober 2005, he Green Labelling Sysem for Condominiums was sared which required he gahering and publishing of informaion based on four environmenal evaluaion iems. The four evaluaion iems are: a) qualiy of building insulaion which addresses reducion in he building's hea load; b) faciliy energy-saving performance, which addresses energysaving sysems; and c) lifespan exension and d) greening of he building, which address lifespan exension, ec., and greening. The evaluaion resuls for he respecive iems are expressed as a number of sar symbols ranging from one o hree sars. In addiion, in order o increase recogniion among consumers, condominium buildings under he obligaion o submi an environmenal plan documen had o indicae he scores of he evaluaion iems. Moreover, from January 2010 onward, he sysem was changed o cover no only owner- Consrucion, (14) Nishimasu Consrucion, and (15) Haseko Corporaion; and (16) oher. 6

7 occupied condominium buildings bu also renal condominium buildings and he floor space for which noificaion is required was lowered o 5,000 m 2 in oal. This change also sipulaed ha owners of smaller buildings were also permied o apply for his label a heir own discreion. The hedonic model used for his analysis includes a variable for buildings wih wo or more sars for eiher a) building insulaion (covering reducion in he building's hea load) or faciliy energy-saving performance (covering energy-saving sysems) and 0 oherwise. The variable is no applied o buildings which have only one sar under he Green Labelling Sysem for Condominiums as his was deemed oo low o qualify as a credible 'green' produc. Moreover, wih regard o building performance evaluaion, he exisence of a Housing Design Performance Evaluaion Documen and Housing Consrucion Performance Evaluaion Documen based on he Housing Qualiy Assurance Ac is also considered in our analysis. This is done o ensure ha he measured price conribuion of a green label is separaed from he effec of convenional Housing Performance Evaluaion and qualiy assurance documens. Adequae locaion conrols are essenial in any aemp o disenangle he facors conribuing o propery prices. In he presen analysis we use a fine-grained 500m x 500m mesh block in which he condominium is locaed as a uni. Specifically, he characerisics are buil-up area, average floor space, sandard deviaion of floor space, number of floors for each building and he sandard deviaion for he number of floors. Nex, area-based informaion on he proporion of he populaion aged 65 and over and he proporion of office workers in he peraining census mesh block were added. To accoun for unobserved spaial characerisics, we also generae a local adminisraive disric. A furher se of variables indicaes proximiy o a railway line and he ime required o Tokyo Saion from he neares saion was also included as a regressor. Buyer characerisics are an imporan feaure of our analysis as oulined above. The following variables were considered: home buyers' annual income, age, occupaion, household size, number of children and idenifier for firs ime buyers. Wih regard o occupaion, differences by employmen saus, 3 work caegory, 4 and indusry caegory 5 were 3 Regarding employmen saus, he survey uses he following classificaion: 01. permanen employee, 02. conrac employee, 03. civil servan/public organisaion employee, 04. self-employed, 05. physician/lawyer/ax accounan/accounan/ec., 06. par-ime/casual, 07. homemaker, 08. suden, and 09. unemployed. There were no samples corresponding o conrac worker, par-ime worker, homemaker, or suden. 4 The survey was conduced using he following classificaion for employmen caegory: 01. clerical job, 02. sales job, 03. echnical job, 04. service/reail job, 05. consrucion/manufacuring job, 06. specialized job, 07. 7

8 examined. In a preliminary sep, he disribuions and descripive saisics of he underlying daase are invesigaed (Table 1). The average asking price of a condominium in Tokyo in he period was million yen while he average value for he ransacion price was approximaely 1.5 million yen lower, a million. Large variaions in size are oberved, from 10m 2 sudio condominiums o large-scale condominiums exceeding 200m 2. The walking ime o he neares public ransi saion is 7 minues on average, while he average ime o Tokyo Cenral Saion is 23 minues which shows ha he properies in our sample are generally well served by public ransporaion and are hence comparable wih regard o locaion value. Looking a buyer characerisics, he average age of buyers was 37 and he average number of people in he household was 2.3, demonsraing ha hese buyers are ypical Japanese households and could hence be considered as represenaive of Japanese home buyers in general. However, an imporan cavea is ha he household head's average income was 8.51 million, a level ha is abou wice he Japanese average income. Table 1: Summary saisics of he esimaion sample average sandard deviaion minimum maximum asking price ( 10,000) 4, , , , ransacion price ( 10,000) 4, , , floor space (sq.m.) ime o neares saion (min) ime o Tokyo Cenral (min) Age of household head (yr) Household size (persons) Household income ( 10,000) , Toal unis of projec , Sie area (sq.m.) 5, , , Toal building area (sq.m.) 22, , , managemen job, and 08. company execuive. 5 The following iems were surveyed as indusry caegories: 01. agriculure, foresry, and fishing, 02. consrucion, 03. manufacuring, 04. ransporaion/warehousing, 05. finance/securiies/insurance, 06. adverising/publishing/broadcasing, 07. prining/ypeseing, 08. fashion-relaed, 09. ravel/hoel/leisure, 10. resauran/bar, 11. housing/real esae, 12. rading/wholesaling, 13. reail, 14. sofware/informaion services, 15. beauy, 16. medical/welfare, 17. educaion, 18. creaive professions, and 19. oher. 8

9 Number of observaions=23,920 Model specificaion Following he overall research sraegy oulined in he previous secion and aking ino accoun buyer characerisics which are limiing condiions for he bid price funcion, we specify he following model: P j, ) f ( Gi, X( i,, NEk, HH( i, ) P : New condominium price of condominium i and dwelling j a ime (1: asking j, ) price, 2: ransacion price) G i : Green label of condominium i X : Building characerisics of condominium i & dwelling j NE k : Locaion characerisics of region k HH : Buyer characerisics of condominium i and dwelling j This specificaion has a number of desirable properies as compared o previously esimaed hedonic funcions. Firsly, regarding he definiion of price ( P j, ) ), boh he asking and he recorded ransacion prices are known in each individual case, allowing us o invesigae wheher sellers have unrealisically high expecaions of he marke value of a 'green label'. Secondly, a price differenial is generaed ( X ) based on differences in condominium (i) feaures such as building srucure and he size of he lo area, as well as feaures relaed o he dwelling (, such as he floor space, he uni's posiion (wheher or no i is a corner uni), ec. In erms of he condominium building's feaures (i), i has increasingly been poined ou ha a price differenial is generaed by he condominium developer or he developer's brand (he developer's reliabiliy and qualiy assurance, which is difficul o observe visually) and by he consrucion company. Therefore, developer and consrucion company informaion was also gahered and incorporaed ino our analysis. Furher o hese kinds of building and dwelling characerisics, he characerisics of he surrounding environmen, such as he sreescape of he area (k), he commercial densiy, 9

10 ec., have a major effec on house prices, proxied by a neighbourhood effec ( NE ) in our esimaion. Accessibiliy o jobs, schools, shopping and oher rouine desinaions are capured by variables measuring he disance o he neares saion, ravel ime o cenral business disric, ec. k Moreover, hedonic pricing heory predics ha he addiion of buyer characerisics will yield differen bid curves reflecing differen uiliy funcions and budge consrains. For example, he marginal uiliy of an exra uni of space will be higher for larger households, all else equal. Model 3 relaxes he assumpion of a homogenous uiliy funcion of all buyers by conrolling for a series of buyer characerisics in line wih suggesions made in he exan lieraure (see Shimizu, Nishimura, and Karao, 2007). Based on hese consideraions, he hedonic price funcion is esimaed focusing on he condominium price ( P j, ) ) a ime. Firs, as a sandard model, he following model is aken as a saring poin (Model 1). logp j, ) a n 0 a T a n 5 1 NE n k a G a GT 2 6 i 3 a D i m a m 4 X m where T is a ransacion (1 in he case of he ransacion price; 0 in he case of he asking price), while D ( = 2004 o 2011) is a ime. The green label effec is capured by a binary variable ( prices can hen be derived from a cross-erm ( G i ). The difference beween green asking and ransacion G T ). Nex, i was expanded ino a hedonic funcion facoring in buyer characerisics, which in heoreical erms should normally be considered, bu which were difficul o incorporae ino he model due o daa limiaions (Model 2). i j logp j, ) a 0 n a T a n 5 1 NE n k a G a G T s 2 a s 6 i HH 3 s i a 7 m D a m 4 X m Moreover, how he green label effec ( G i ) changed in accordance wih he passage of ime 10

11 was analyzed (Model 3). logp j, ) a 0 n a T a n 5 1 NE n k a G a G T s 2 a s 6 i HH 3 s i a 7 D m a m 4 X m a G D 8 i Regarding appropriae esimaion echniques, our baseline model is an Ordinary Leas Squares (OLS) model in line wih he Gauss-Markov Theorem which saes ha OLS will yield he bes linear unbiased esimaor provided ha he usual assumpions are me. Efficiency as well as ease of compuaion and inerpreaion make OLS he mehod of choice for he majoriy of hedonic pricing sudies. However, since OLS is known o be sensiive o ouliers, we also use robus regression as an alernaive esimaion echnique o see wheher his changes he resuls for he variables of ineres. In paricular, Cook's disance measures are used o idenify bad leverage ouliers and Huber and Tukey biweighs are applied o offse he disoring influence of hese ouliers on he regression parameers as suggesed by Huber, 1964; Rousseeuw and Leroy, 1987 and Verardi and Croux, Using OLS and robus regression in andem should provide us wih a sable esimae of any pricing differenials relaed o he Tokyo Green Label. Esimaion Resuls and Discussion The esimaion resuls for he hree models are oulined in Table 2 (full esimaion resuls in Appendix 2). The baseline esimaion (Model 1) reveals ha he average asking price for a condominium wih a green label is roughly 5% higher compared o a similar condominium wihou a label. In oher words, he developers of condominiums wih superior environmenal performance offered hem a a marginally bu significanly higher price. However, he acual achieved ransacion prices are more relevan o our main research hypohesis abou he exisence of a green premium in he Tokyo housing marke. The general ransacion price variable indicaes ha ransacion prices were on average approximaely 4% lower han asking prices in he observed period When we es specifically how ransacion prices of green labelled buildings differ from heir average asking prices, he coefficiens are no saisically significan a he 5 percen level. This finding suggess ha he value of he green label was no significanly overesimaed by sellers relaive o he implici valuaion of he buyers. A more deailed lis of coefficien esimaes is conained in Appendix 2 which reveals a number of addiional insighs ino he pricing of green and non-green 11

12 condominiums. Boh he Housing Design Performance Evaluaion Documen (Par A) 6 and he Housing Performance Consrucion Evaluaion Documen (Par B) arac marginal bu significan price premiums. These evaluaion documen variables are imporan for isolaing he 'pure' effec of he green label effec from oher ypes of qualiy evaluaion of newly buil properies. Nex, similarly disinguishing buildings based on managemen coss, mainenance/renovaion invesmens, ec., shows ha such coss and ransacion prices are posiive relaed. Furhermore, he price was lower when he ype of land ownership was 'general leasehold' or considerably lower for 'fixed-erm leasehold'. The price was 2% higher when he dwelling is a corner uni. The coefficiens of all oher conrol variables including he ime-period variables exhibied he expeced signs. Model 2 in Table 2 repors he esimaion resuls using robus regression, an esimaion echnique which gives proporionally less weigh o influenial ouliers as described in he previous secion, hereby reducing he poenial bias ha a small group of properies wih exreme or unusual prices and oher aribues may inroduce o he resuls. However, he resuls are only slighly differen o he baseline model wih a 4.3% asking price premium and a negligible and insignifican ransacion price premium for green-labelled properies. Nex, we examine he impac of buyer characerisics on he model (Models 3-7 of Table 2). To his aim, we divide buyers' incomes ino four groups and esimae he impac of all price deerminans separaely for each income group 7. As expeced, green asking price premia (as a fracion of he oal price) are found o progress wih increasing incomes of buyers (from 4% o nearly 8%). Similarly,we find ha he average price premium observed in recorded ransacion prices (as opposed o asking prices) is mainly driven by households wih above-average incomes.paid for green-labelled properies. Given ha hese are percenages on he oal price, a base which is higher for more affluen buyers buying more expensive properies, he spread in erms of absolue moneary values of hese price premia is even more pronounced. This finding is significan in ha i demonsraes for he firs ime ha 'green' feaures are more likely o arac higher-income buyers despie argumens o he conrary ha claim energy efficiency and he resuling lower uiliy bills are a larger concern for more income-consrained households. 6 The Housing Performance Indicaion Sysem is based on he Housing Qualiy Assurance Ac enaced on April 1, I evaluaes housing performance based on fixed sandards, such as complying wih he obligaory 10- year defecs liabiliy period for basic srucural areas of new housing. Under his sysem, Housing Performance Evaluaion Documens are issued, which are divided ino Housing Design Performance Evaluaion Documens and Housing Consrucion Performance Evaluaion Documens. 7 Income groups are defined on a per capia basis of household members as follows: lower income: up o 2.5million p.a. (n=6,038), medium low income: million (n=6,982), medium high income: million (n=6,012), high income: 5.50 million (n=4,568) 12

13 Furher findings relaing o buyer characerisics are ha firs-ime buyers exhibi generally a lower willingness or abiliy o pay, paricularly in he lower income segmens (Appendix 2). This may be aken as an indicaion of firs-ime buyers acing more cauiously on he housing marke regardless of curren income, possibly because of heir relaively lower asse possessions compared o buyers who already own a propery and seek o 'rade up'. A price differenial also occurs based on occupaion. Independen of curren income and age, i is possible ha his variable acs as a proxy for fuure income or he cerainy of ha income. The fac ha annual income and employmen generae differences in house prices suppors our earlier proposiion ha prices of boh green and non-green properies canno solely explained by propery characerisics bu are also a funcion of socio-economic buyer characerisics. An imporan cavea o hese findings is ha he coefficien esimaes of he green asking and ransacion price premiums as repored here may overesimae he 'rue' magniude of he premium for several reasons. Firsly, he green label may be correlaed wih unobserved qualiy characerisics of he condominiums which may hen ge lumped ino he green premium esimae. While we canno rule ou compleely ha such an esimaion problem exiss, reassurance is provided by he large numbers of hedonic building, spaial and socioeconomic characerisics used in he esimaion which go far beyond he sandard se of variables used in he hedonic esimaion. The high goodness of fi of he models around 80% of explained variance is esimony o he qualiy of he daase. A second concern is ha he posiive relaionship beween household income and observed green price premiums may simply reflec beer access o capial for more affluen households. In oher words, i is possible ha he underlying willingness o pay is independen of income and wealh bu only high-income households have access o he addiional deb or equiy capial required o buy a green-labelled condominium. This poin is imporan as i may undermine he concep of he green label as an upmarke luxury good. More specifically, if he premium were merely a capialisaion of he discouned fuure uiliy cos reducion and possibly favourable governmen reamen of green-labelled condominium schemes, he differen premia obained for he household income groups may simply show he progression of incomplee capialisaion for low incomes o complee capialisaion of hese cos savings for high incomes. To es his concern empirically we would require daa on acual energy consumpion along wih a vecor of conrol variables o measure he rue cos savings associaed wih Tokyo Green Label condominiums in operaion. In he absence of such daa, we can only provide rough esimaes derived from he exising lieraure. Adan 13

14 and Fuers (2016) repor a reducion of 3 o 13 % in energy consumpion of housing in he UK following a range of heaing sysem and insulaion relaed green rerofi measures. An earlier sudy by Hong e al (2006) repors a reducion in heaing demand by 10% in cenrally heaed and 17% in non-cenrally heaed properies. Considering ha uiliy coss are only a relaively small fracion of he capial value of a house or condominium, i hus seems likely ha he average premium of approximaely 5% on average for he Tokyo marke reflecs more han mere uiliy cos savings. An example using saisics on average energy consumpion levels can illusrae his poin. According o he Family Income and Expendiure Survey compiled by he Japanese Minisry of Inernal Affairs and Communicaion, he average household gas bill (Tokyo Gas) amouns o 4,972 and he elecriciy bill o 9,472. Assuming ha households consume 20% less gas and elecriciy in green-labelled buildings, he equivalen figures for households in hese buildings would be 7,577 and elecriciy o 3,977. The nominal undiscouned combined savings would hen amoun o 866,640 (USD 7,100) over a 25 year period. Given an average ransacion price of million per condominium in our daabase, his would amoun o only 2 percen of he propery value and is hence considerably lower han he esimaed oal premium of approximaely 5 percen. Conversely, households would have o achieve reducions in heir energy consumpion by more han 50% compared o households in he non-green building peer group o jusify he measured green price premium. If he nominal undiscouned savings in our approximaion were o be calculaed o he ne presen value, he required energy savings of he green label would need o be even higher. I hence seems likely ha he green premium reflecs more han jus he pure capialisaion of cos savings. Table 2: Hedonic regression resuls (dependen variable: log price) (1) Baseline OLS (2) Robus reg (3) OLS wih HH char (4) OLS: Income low (5) OLS: Income medium low (6) OLS: Income medium high (7) OLS: Income high Green asking *** *** *** *** *** *** *** price premium 1 (14.43) (13.00) (14.38) (6.00) (6.82) (7.82) (7.13) Green ransacion price discoun 2 Transacion price discoun Propery & condo aribues (-1.44) (-1.58) (-1.46) (-0.17) (-0.14) (-1.21) (-1.36) *** *** *** *** *** *** *** (-19.93) (-20.20) (-20.13) (-12.88) (-12.09) (-9.66) (-7.31) Yes Yes Yes Yes Yes Yes Yes 14

15 Developer fixed Yes Yes Yes Yes Yes Yes Yes effecs Locaion conrols Yes Yes Yes Yes Yes Yes Yes Managemen fixed Yes Yes Yes Yes Yes Yes Yes effecs Buyer No No Yes Yes Yes Yes Yes characerisics Time fixed effecs Yes Yes Yes Yes Yes Yes Yes N 23,922 23,922 23,922 6,038 6,982 6,012 4,568 R adj. R AIC BIC saisics in parenheses * p < 0.05, ** p < 0.01, *** p < : Green asking price premium on equivalen unlabelled condominiums. 2 : The variable 'ransacion price discoun' indicaes he average discoun observed relaive o he equivalen asking price. The variable 'green ransacion price discoun' indicaes he addiional discoun applied o he ransacion of a green-labelled condominium. Boh he general ransacion price discoun and he green ransacion price discoun have o be subraced from he green asking price premium o arrive a he green ransacion price premium. For example, in Model 1 he oal green premium paid in ransacions is 1.6% (6% - 3.5% - 0.9%). Nex, we examine how he green price premium for Tokyo condominiums has changed over ime for differen income groups using he baseline hedonic model wih he full se of conrol variables (including buyer characerisics) as described above. Following he inroducion of he Tokyo green labelling sysem, he firs year wih a sufficienly large number of ransacions was Figure 1 shows how green labelled and non-labelled condominiums developed relaive o he 2004 average. For simpliciy, only wo income groups (above and below median) are considered here. Afer conrolling for a large number of variables, he esimaion resuls indicae green labelled condominiums bough by more affluen buyers commanded he highes premiums over he enire sudy period. Perhaps even more ineresingly. he premium paid by below-average green buyers is by and large higher han he premium observed for above-average non-green buyers and subsanially higher han for non-green ransacions in he below average income peer group. The green premiums have generally declined in he more recen years of he analysis from 2007/8 onwards bu he price differeniaion beween he green and he non-green producs is sill very clear. I is imporan o bear in mind, however, ha he developmen of he green premium is no merely a reflecion of he demand and supply balance of labelled properies bu was probably also influenced by he implemenaion of environmenal policies by he Japanese governmen, ypically aking he form of subsidies and favourable ax reamens of green invesmens. 15

16 Prices by income group (in % compared o 2004) Green low income Green high income Non-green high income 0.05 Non-green low income Figure 1: Esimaion resuls of condominium prices relaive o 2004 levels in % (0.25=25%) by income group and labelling saus of he purchased uni. Lower incomes < 7.5 million, higher incomes 7.5 million, R-sq=0.80, N=23,922 Conclusions This paper se ou o es wheher obaining a green label adds value o residenial properies in he Japanese housing marke using a unique daase of new condominium ransacions in he Tokyo marke. Based on our analysis, his quesion can be answered in much he same way as previous research has done across he world. The hedonic analysis shows a clear price premium for green-labelled condominiums boh for asking and ransacion prices. Taking ino accoun buyer characerisics, we find ha wealhier buyers are willing o pay a higher premium for green-labelled properies, boh in absolue and in relaive erms. I appears ha eco-labelled condominiums in Tokyo are acquired primarily by higher income households who are prepared o pay a premium for owning and occupying a green-labelled propery. Conversely, he lower price premiums found for below-median income groups may be reflecive of he igher budge consrains hey face and hey may discoun possible fuure cash savings via lower uiliy bills more heavily due o hese consrains on heir abiliy o acquire addiional up-fron capial for buying he bundle of goods proxied by he Tokyo green label. In addiion, if one looks a emporal changes in he premium, we find ha he effec of green labels became larger over ime before declining again in he final wo years of he sudy 16

17 period. In erms of he possible reasons for his, i may be ha he awareness of green buildings increased in he Japanese residenial marke in he years following inroducion of he Tokyo Green Label and ha he buyer segmen acively seeking o inves in heir value is expanding. A furher possibiliy is an increase in he supply of green buildings which may have led o a levelling off of he observed price premia. However, furher analysis is required o ascerain wheher he absence of he premium in he mos recen is a coninued rend ha marks he end of a 'green premium era' or is simply a one-off occurrence. A number of caveas remain for his analysis. Firs, i is imporan o bear in mind ha he Tokyo labelling sysem is based on applicaions from developers, i.e. i only indicaes a building's hypoheical environmenal performance a he ime of developmen. I is possible ha buyers may be relucan o pay higher premia for energy efficiency and cos savings unless hey are proven in operaion. Similarly, unless he added economic value of green buildings offses or exceeds he added developmen coss, developers are unlikely o develop many green buildings unless hey receive subsidies o make up for he shorfall. Hence, we canno rule ou ha he premium we measured may sill be oo low in comparison o he added developmen expenses which may be an obsacle o more widespread adapaion of green buildings in he Japanese marke. More deailed informaion on he cos premium for consrucing green condominium buildings would be required for his profiabiliy calculaion. Furhermore, i is also uncerain how he emerging green building segmen will be embedded in he broader housing marke ha comprises mainly of exising sock. Under he curren sysem, green labels only cover newly developed buildings, bu for green building policies o be more effecive he exension of labels o exising sock will have o be considered. Noably, when i comes o a buyer's choice of home, he decision is ypically made under considerable budge resricions. Wih he rapid changes in Japanese demographic srucure, he populaion of people in heir 30s and 40s -- which is he home-buyer segmen ha generaes he greaes demand for housing -- is decreasing significanly. This may shif he aggregae demand curve downwards, possibly in a more pronounced manner for new consrucion and newly consruced green buildings han for he general exising housing sock. In his conex, i will be necessary o keep monioring wheher here coninues o be an added value and price premiums for green buildings. Finally, he economic value of green buildings will also be affeced by more sringen environmenal regulaions o be implemened in fuure (Takagi and Shimizu, 2010) bu i is difficul o foresee how Japanese eco labels for buildings will adap o hese long-erm changes in marke condiions. 17

18 Acknowledgmen: The coninuous suppor of he Cambridge Universiy Land Sociey in enabling he research aciviies of he CULS Fellow is graefully acknowledged. References Adan, H.; Fuers, F.; (2016): Do Energy Efficiency Measures Really Reduce Household Energy Consumpion? A Difference-in-Differences Analysis, Energy Efficiency, forhcoming. Banfi, S., Farsi, M., Filippini, M., and Jakob, M. (2005), Willingness o Pay for Energy-Saving Measures in Residenial Buildings," CEPE Working Paper, No. 41. Brounen, D. and Kok, N. (2011) On he Economics of Energy Labels in he Housing Marke, Journal of Environmenal Economics and Managemen, 62, pp Deng, Y., Li, Z., & Quigley, J. M. (2012). Economic reurns o energy-efficien invesmens in he housing marke: evidence from Singapore. Regional Science and Urban Economics, 42(3), Dian, T.M., and Miranowski, J. (1989), Esimaing he Implici Price of Energy Efficiency Improvemens in he Residenial Housing Marke---A Hedonic Approach, Journal of Urban Economics, 25, pp Eichholz, P, Kok, N and Quigley, J Doing Well By Doing Good? Green Office Buildings, American Economic Review Eichholz, P. Kok, N. and Quigley, J The economics of green building. Review of Economics and Saisics, in press. Eichholz, P; Quigley, J. (2012): Green building finance and invesmens: Pracice, policy and research. European Economic Review 56.5 (2012): Ekeland, I., J. J. Heckman and L. Nesheim, (2004), Idenificaion and Esimaion of Hedonic Models, Journal of Poliical Economy, 112, pp Fuers, F., McAlliser, P., Nanda, A., & Wya, P. (2015). Does energy efficiency maer o homebuyers? An invesigaion of EPC raings and ransacion prices in England. Energy Economics. 48, Fuers, F. and McAlliser, P. (2011) Eco-Labelling in Commercial Office Markes: Do LEED and Energy Sar Offices Obain Muliple Premiums? Ecological Economics, 70, pp Fuers,F.,and P. McAlliser(2010), Green Noise or Green Value? Measuring he Effecs of Environmenal Cerificaion on Office Values, Real Esae Economics, 39, pp Hong, S. H., Oreszczyn, T., & Ridley, I. (2006). The impac of energy efficien refurbishmen on he space heaing fuel consumpion in English dwellings.energy and Buildings, 38(10), Huber, P., Robus esimaion of a locaion parameer. Ann. Mah. Sa Hyland, M., Lyons, R. C., & Lyons, S. (2013). The value of domesic building energy efficiency evidence from Ireland. Energy Economics, 40, Jacobsen, G. D., Kochen, M. J., & Vandenbergh, M. P. (2012). The behavioral response o volunary provision of an environmenal public good: Evidence from residenial elecriciy demand. European Economic Review, 56(5), Japanese Minisry of Inernal Affairs and Communicaions (2014): Family Income and Expendiure Survey, 2014 Ediion. Kahn, M. E., & Kok, N. (2014). The capialisaion of green labels in he California housing marke. Regional Science and Urban Economics, 47, Kochen, M. J. (2006). Green markes and privae provision of public goods. Journal of Poliical Economy, 114(4), Miller, N, Spivey, J. & Florance, A. (2008). Does Green Pay Off? Journal of Real Esae Porfolio Managemen, 14, Minisry of Land, Infrasrucure, Transpor, and Tourism Land and Waer Bureau (2010), Conference on he Green Value of Buildings Research Repor, Minisry of Land, Infrasrucure, Transpor, and Tourism. Reichard, A. Fuers, F. Roke, N. and Ziez, J. (2012) The business case of susainable building cerificaion: A panel daa approach. Journal of Real Esae Research, 34, Requae, T., Unold, W. (2003) Environmenal policy incenives o adop advanced abaemen echnology: will he rue ranking please sand up? European Economic Review 47, Rosen, S., (1974), Hedonic Prices and Implici Markes, Produc Differeniaion in Pure Compeiion, 18

19 Journal of Poliical Economy, 82, pp Rousseeuw, P.J., Leroy, A. (1987) Robus Regression and Oulier Deecion. John Wiley and Sons, New York. Shimizu, C. (2010), Will green buildings be appropriaely valued by he marke?,ripess (Reiaku Insiue of Poliical Economics and Social Sudies) Working Paper,No.40. Shimizu, C., K. G. Nishimura and K. Karao (2014), Nonlineariy of Housing Price Srucure - Secondhand Condominium Marke in Tokyo Meropolian Area-, Inernaional Journal of Housing Markes and Analysis, 7(3), Takagi,Y and C.Shimizu.(2010), The Environmen and Real- Esae Invesmen:Responsible propery invesing,business Ehics and Compliance Research Cener Working Paper(Reiaku Universiy),No.6. Tokyo Associaion of Real Esae Appraisers (2010), Propery environmenal friendliness and asse prices: Empirical evidence based on Tokyo condominiums. Verardi, V., Croux, C. (2009): Robus regression in Saa. Saa Journal. SaaCorp LP 9 (3), Yoshida, J., and Sugiura, A. (2015). The Effecs of Muliple Green Facors on Condominium Prices. The Journal of Real Esae Finance and Economics, 50(3), Zheng, S., Wu, J., Kahn, M (2008):. Land and residenial propery markes in a booming economy: new evidence from Beijing. Journal of Urban Economics 63: Zheng, S., Wu, J., Kahn, M. E., & Deng, Y. (2012). The nascen marke for green real esae in Beijing. European Economic Review, 56(5), Appendix 1: Variable names and daa sources for Tokyo case sudy Symbol Variable Conen Uni Source green Green label Green building = 1 Oher building = 0 (0,1) Tokyo Meropolian Governmen rans S TS Bus TT Transacion price Floor space Disance o he neares saion Bus Time o CBD(Tokyo saion) Transacion price=1 Asking price=0 Floor space of building/square meers Disance o he neares saion. bus-ransporaion area=1 walk-ransporaion area=2 Average ravel ime from he neares rail ransi saion o Tokyo Cenral Saion during dayime hours. (0,1) RECRUIT m 2 meers Real Esae Economic Insiue Real Esae Economic Insiue (0,1) Real Esae Economic Insiue minues VAL Insiue TU Land TA Cos ISP1 ISP2 Toal uni Sie area Toal building area Managemen Cos Wih Housing performance evaluaion repor A Wih Housing performance evaluaion Toal unis of condominium Sie area of condominium Toal building area of condominium Propery Managemen Cos Wih Housing performance evaluaion repor A=1 Wihou Housing performance evaluaion repor A=2 Wih Housing performance evaluaion repor B=1 m2 uni m 2 Real Esae Economic Insiue Real Esae Economic Insiue yen/monh RECRUIT (0,1) RECRUIT (0,1) RECRUIT 19

20 MG1 MG2 Corner STD RL1 RL2 TR LU g (g=0,,g) Income HD h (h=0,,h) WD i (i=0,,i) YD j (j=0,,j) LD k (k=0,,k) RD l (l=0,,l) D m (m=0,,m) repor B Managemen ype(a) Managemen ype(b) corner Sudio ype Leasehold(A) Leasehold(B) Rae of Sales Land Use regulaion Household income Employmen saus Job ype Business ype Locaion (ward) Railway line Time (yearly) Wihou Housing performance evaluaion repor B=2 Managemen ype is A=1 (0,1) RECRUIT oher =0 Managemen ype is B=1 (0,1) RECRUIT oher =0 The locaion of uni is corner =1 (0,1) Real Esae Economic Insiue Oher locaion =0 Floor space 30 square meers or less =1 Floor space over 30 (0,1) Real Esae Economic Insiue square meers = 0 Land righ is leasehold(type A)=1 (0,1) RECRUIT oher =0 Land righ is leasehold(type B)=1 (0,1) RECRUIT oher =0 Rae of sales in firs monh g-h Land use regulaion area=1 (residenial, office, indusrial) oher =0 (%) Real Esae Economic Insiue (0,1) Real Esae Economic Insiue Annual income of household yen RECRUIT h-hemploymen saus of Head of household =1, (0,1) RECRUIT oher=0 i-h job ype =1, oher=0 j-h business ype =1, oher=0 (0,1) RECRUIT (0,1) RECRUIT k-h adminisraive disric =1, (0,1) Real Esae Economic Insiue Oher disric =0. l-h railway line =1 Oher railway line = 0. m-h year =1 Oher year =0. (0,1) Real Esae Economic Insiue (0,1) RECRUIT Appendix 2: Full regression resuls of Tokyo case sudy (1) Baseline (2) Robus reg (3) OLS wih (4) Income low (5) Income (6) Income (7) Income high 20

21 OLS HH char medium low medium high log price log price log price log price log price log price log price green *** *** *** *** *** *** *** (14.43) (13.00) (14.38) (6.00) (6.82) (7.82) (7.13) rgreen (-1.44) (-1.58) (-1.46) (-0.17) (-0.14) (-1.21) (-1.36) rans *** *** *** *** *** *** *** (-19.93) (-20.20) (-20.13) (-12.88) (-12.09) (-9.66) (-7.31) sqf *** *** *** *** *** *** *** (48.74) (88.08) (43.78) (16.18) (33.19) (31.69) (31.91) sqf_squared *** *** *** *** *** *** *** (-16.85) (-34.66) (-15.79) (-4.81) (-16.70) (-11.59) (-11.41) s *** *** *** *** *** *** *** (-38.81) (-41.91) (-37.39) (-18.55) (-16.12) (-18.20) (-16.05) bus *** *** *** ** *** (-3.81) (-4.05) (-3.89) (-1.24) (-3.12) (-4.45) (-1.19) busts *** *** *** *** *** ** * (-10.15) (-14.38) (-9.94) (-7.78) (-5.77) (-2.82) (-2.33) TA -3.57e e e e e e *** (-1.20) (0.75) (-1.20) (1.46) (1.44) (-0.82) (-3.40) ISP * * * (2.48) (2.12) (2.50) (1.01) (0.50) (1.25) (0.66) ISP *** *** ** ** (3.44) (1.81) (3.34) (2.61) (0.42) (0.53) (3.13) MG (1.75) (1.58) (0.72) (0.07) (0.87) (0.39) (1.59) MG *** *** *** ** * ** (6.53) (7.03) (5.44) (2.69) (2.14) (3.04) (1.96) RL *** *** * ** (-3.98) (-3.48) (-2.11) (-2.80) (0.02) (1.17) (-1.85) RL *** *** *** *** (-3.68) (-4.60) (-3.52) (-3.37) (-0.85) (-1.15) (-0.86) corner *** *** *** *** *** *** ** (9.59) (9.71) (9.09) (6.55) (5.21) (5.13) (2.77) srucure * (1.62) (1.58) (1.45) (2.47) (0.67) (-0.13) (-0.04) TR *** *** *** *** *** *** ** (-9.70) (-9.03) (-9.64) (-3.50) (-5.54) (-4.91) (-3.16) *** *** *** *** *** (-8.65) (-12.55) (-8.70) (-6.01) (-9.53) (-1.79) (-0.51) FAR *** *** *** * *** (4.74) (7.72) (4.37) (-0.10) (1.97) (1.24) (3.83) comm *** *** *** *** * * (3.94) (5.51) (4.09) (6.29) (2.29) (2.22) (-1.43) indus *** *** *** ** ** *** ** (-10.13) (-10.41) (-9.87) (-2.65) (-3.11) (-4.76) (-2.65) renal *** *** *** * *** *** *** (-13.79) (-13.34) (-13.26) (-2.07) (-5.36) (-8.45) (-8.40) old *** *** *** ** ** (-3.83) (-5.25) (-3.66) (-3.22) (0.19) (-2.74) (-0.35) src *** *** *** *** *** *** (6.93) (7.78) (6.80) (6.53) (3.50) (4.54) (1.40) officew *** *** *** *** *** *** *** (20.92) (22.98) (20.69) (9.15) (8.93) (11.90) (6.39) cos *** *** *** * *** *** *** (10.46) (10.87) (10.23) (2.51) (6.62) (4.00) (4.18) age *** *** *** (7.93) (5.22) (1.37) (4.69) (1.79) 21