Macroeconomic and demographic determinants of residential property prices in Malaysia.

Transcription

1 MPRA Munich Personal RePEc Archive Macroeconomic and demographic deerminans of residenial propery prices in Malaysia. Ivan D. Trofimov and Nazaria Md. Aris and Dickson C. D. Xuan Kolej Yayasan Saad (KYS) Business School, Malaysia, Universiy of Malaysia Sarawak 6 April 2018 Online a hps://mpra.ub.uni-muenchen.de/85819/ MPRA Paper No , posed 11 April :16 UTC

2 Macroeconomic and demographic deerminans of residenial propery prices in Malaysia Ivan D. Trofimov 1 ; Nazaria Md. Aris 2 ; Dickson C. D. Xuan 2 Absrac This paper sudies he relaionship beween residenial propery prices and macroeconomic and demographic deerminans in Malaysia. In he years following he Asian financial crisis, propery prices in Malaysia rose subsanially, resuling in an affordabiliy crisis and ulimaely policy responses o he problem. Using uni roo, Johansen-Juselius coinegraion, VECM-based Granger causaliy ess and variance decomposiion, and considering quarerly daa ha covers period, we esablished ha residenial propery price growh is principally driven by srong demographic performance and populaion growh and is backed by he low ineres rae environmen and rising consumer prices. Household income and level of GDP do no appear o conribue o propery price growh. Cerain disorions and asymmeries in he Malaysian real esae markes are documened: oversupply in he higher price segmen of he marke coupled wih he lack of affordable housing in he lower price segmen; household income growh lagging behind GDP and propery price growh, hereby dampening housing demand; growing renal markes in major urban areas as a resul of he affordabiliy crisis; and a qualiy mismach beween buyers preferences and housing supply. Keywords: propery prices; housing; coinegraion JEL Codes: C22, R21, R30, R38 Inroducion Economic and sociological lieraure documens he imporance of home ownership and residenial propery for social reproducion and economic developmen and sabiliy. Residenial housing is a 'big icke' iem and a major componen of household expendiure. A major deerminan of savings a he micro and macro levels, is affordabiliy is also a major deerminan of individual and social well-being, affecing qualiy of life, life saisfacion, family economics, and durabiliy of social fabric. In addiion, he consrucion secor, real esae, and real esae banking and lending are major pars of he naional economy, whils regulaion peraining o residenial real esae is one of he mos imporan governmen policies. Malfuncioning in he propery markes has broad domesic and inernaional implicaions, compromising financial sabiliy and economic growh, as was evidenced by he global financial crisis of Consideraion of residenial propery prices (as a major componen of housing affordabiliy) and heir deerminans is herefore a salien heoreical, empirical and policy issue in any economy. Residenial housing prices are subjec o a plehora of influences relaed o macroeconomic and demographic condiions, he sae of local housing markes, sociological and socio-culural characerisics of buyers and sellers, and policy facors. In his sudy, we focus on cerain 1 Kolej Yayasan Saad (KYS) Business School, Malaysia, ivan.rofimov1@gmail.com (Corresponding auhor) 2 Universiy of Malaysia Sarawak, nazzoes@gmail.com. Auhors acknowledge he suppor by Universiy of Malaysia, Sarawak (UNIMAS Special Shor Term Gran - F01/SpSTG/1574/2017). 1

3 macroeconomic and demographic deerminans. The purpose of his sudy is o examine he effec of seleced deerminans (populaion growh, GDP level, household income, level of lending raes, and consumer prices) on he residenial housing prices in Malaysia. In conras o he empirical analysis of he residenial propery markes in he OECD and seleced developing economies (paricularly hose where housing affordabiliy is he mos acue, or where deformaions in he residenial markes are mos visible, e.g. China, Hong Kong, Singapore), he sysemaic research of housing price deerminans in Malaysia will be limied. I will focus on an analysis of supply-side facors, policy insrumens ha affec propery prices, he sae of he residenial markes in paricular localiies, and affordabiliy faced by various social groups, or will use daa from earlier periods. The novely of his paper will be is consideraion of residenial propery prices for Malaysia as a whole, and is examinaion of he influence of macroeconomic and demographic variables on prices in he mos recen period. The essay is organised as follows. Secion 2 will provide background informaion relaed o he developmens aking place in he Malaysian residenial propery marke. Secion 3 will provide an overview of heoreical views and empirical analyses peraining o residenial propery price deerminans. Secion 4 will consider daa sources, models, and economeric echniques. Secion 5 will presen empirical findings. Secion 6 will offer concluding remarks, provide inerpreaion of empirical resuls, discuss he limiaions of he research, and ouline avenues for fuure research. Background Residenial propery markes in Malaysian in he s were characerised by conflicing developmens and subjec o mulidirecional forces. Since he early 2000s, house price increases were experienced in all saes of Malaysia, and price growh paricularly acceleraed in he afermah of he global financial urmoil of (NAPIC, 2014). For all residenial housing caegories, he 2014 house price as a muliple of 2000 house price sood a 2.14 for Malaysia as a whole, rising as high as 2.99 and 2.58 in he sae of Sabah and in Kuala Lumpur respecively (i.e. prices over 15 year period increased by up o 200%). The propery marke in jus one sae (Melaka) was characerised as affordable, wih median muliple affordabiliy sanding a 3.0, whereas markes in all oher saes and federal erriories were classified as moderaely, seriously or severely unaffordable (Khazanah Research Insiue, 2015). The underpinning facor of hese developmens was srong populaion growh; however, i was also complemened by real esae speculaion and governmen policies ha encourage i. Hashim (2010) provides evidence of subsanial divergence of he mean from he median house prices, poining o disorions in housing supply, specifically o proliferaion of high-end residenial properies and an inadequae supply of affordable and budge housing for working and middle class people. The problem has been aggravaed by inconsisen housing policy, wih posiive incenives for foreign home ownership, exempions from samp duy for higher-end propery purchases, aboliion of he real propery gains ax (RPGT) and low base rae have all encouraged real esae speculaion. On he oher hand, he imporance of affordable housing for Malaysian macro-, urban and social economies has been recognised by he Malaysian governmen. The rise in propery prices ou of synchronizaion wih income growh saw policy responses o improve affordabiliy. Firsly, he 2

4 sevenh and eighh Malaysia Plans saw he consrucion of 1.6 million properies, he majoriy argeing he housing needs of he low and middle income caegories. Secondly, wihdrawals from employees Providen Fund accouns o pay off home loans were permied from January 2008, in order o reduce he morgage burden whils conaining propery speculaion. Thirdly, My Firs Home Scheme, and he '1Malaysia Housing Program' were launched in 2011 o improve affordabiliy. Fourhly, some previous quesionable policies were reversed, such as he re-imposiion of PRGT in I remains o be seen which endency will dominae. Hashim (2010) argues ha affordabiliy has been improving naionwide since he Asian financial crisis of On he oher hand, a repor by he Khazanah Research Insiue (2015) idenifies an acue affordabiliy problem in mos saes of Malaysia, driven by he demographic boom and insufficienly addressed by public policy. The laer has ended o pursue a narrow agenda of provision of public housing o low income households, wihou due consideraion of he affordabiliy problems faced by middle income households, paricularly in urban areas and large urban agglomeraions, ineligible for housing assisance and a he same ime unable o purchase propery in a free marke. This paper does no focus on he issue of home affordabiliy or on measuring he exen of unaffordabiliy. However, knowledge of he driving forces of propery prices is insrumenal in housing policy formulaion and selecion of he appropriae policy insrumens. Lieraure review A number of heoreical explanaions are advanced o explain he funcioning of he residenial propery markes. Firsly, inelasic supply and supply side consrains were idenified as a driving force. Brueckner (1991) and Dowell (1984) poin o reduced supply of land (as an objecive facor or a resul of growh conrols imposed by governmens), leading o higher developmen coss, slack in residenial developmen, and higher and more volaile propery prices (Huang & Tang, 2012). The sae of he consrucion indusry including he cos of consrucion maerials, wages, echnological improvemen and innovaion likewise affecs he qualiy and amoun of supply. Secondly, Mankiw and Weil (1989) and Poerba (1991) poin o he imporance of demographic facors, such as overall populaion growh, household formaion process and populaion ageing. Populaion growh in he curren period would have posiive effecs on he demand for housing and house prices in he subsequen periods, i.e. subsanial lags are presen. Mankiw and Weil argue ha in developed economies he demand for residenial propery will decline when he large populaion cohor (he Baby Boomer generaion) ages and he smaller populaion cohor (he Baby Bus generaion) reaches adulhood, and ages laer on, resuling in a significan decline in prices (asse-meldown hypohesis). Thirdly, Glaeser (2013) considers he effec of expecaions and collecive psychology facors, finding ha opimisic expecaions end o have greaer effec on housing prices and cycles han credi condiions, wih a failure o esimae he supply response resuling in over-opimisic views abou propery and land prices in he fuure. 3

5 Fourhly, local condiions and he qualiy of individual propery are salien, wih implicaions for prices in specific areas: locaion, aesheic and healh facors; proximiy o proeced areas, ameniies, and ranspor infrasrucure; proximiy o negaive exernaliy producers; or oher adverse facors, such as elecriciy infrasrucure and high volage ransmission lines (Gregory, Von Winerfeld, 1996; Poudyal e al., 2009; Debrezion e al., 2006). Fifhly, he general sae of he economy is considered a major deerminan of housing prices. Imporanly, causaliy beween he wo is bidirecional. Macroeconomic condiions affec real esae prices, bu real esae prices condiion he sae of economy; for example, housing booms may have posiive effecs on household consumpion (Quigley, 1999; Girouard & Blondal, 2001). As noed by Case e al. (2005), he effecs of housing prices on real aciviy end o be srong, relaive o he effecs from he sock marke, wih a sharp decline in house prices having a much bigger impac on oupu growh han equiy price buss. Sixh, he availabiliy of credi and he ineracion beween bank lending and demand for propery are preponderan. On one hand, as argued by Minsky (1982), Ausrian school economics (Thornon, 2009), and more recenly by inernaional economic regulaors (IMF, 2000; Borio & Lowe, 2002), he availabiliy of credi affecs propery demand and (given fixed supply in he shor-run) propery valuaions. The cycles in propery prices are hereby credi driven, wih credi availabiliy being he funcion of cenral banks prime raes, innovaion in lending pracices, and he sophisicaion of housing finance producs. On he oher hand, propery prices affec credi supply and banks lending capaciy and capial posiion hrough he valuaion and performance of banks real esae porfolios. Through he wealh effec, propery prices aler he value of collaeral and he borrowing capaciy of households, and hus demand for credi (Bernanke & Gerler, 1989; Kiyoaki & Moore, 1997). Empirical evidence ends o suppor boh views. Goodhar (1995) idenifies ha propery prices subsanially affeced credi growh in seleced developed economies, whils Collyns and Senhadji (2001) repored significan conemporaneous effecs of credi on residenial propery prices in Asian economies. Sevenh, he sae of oher financial markes and he level of prices in hese markes (specifically sock prices) are likely o be deerminans of residenial prices. On one hand, he rise in sock prices and he appreciaion of sock porfolios and associaed increase in financial wealh boos invesmen in real esae (wealh effec), resuling in higher demand for real esae and higher propery prices. On he oher hand, he opposie causal effecs (from propery o sock markes) are also likely (Kapopoulos & Siokis, 2005). As a per credi-price effec, he increase in propery prices and respecive increase in collaeral values for loans spurs lending o firms and households across he economy, simulaing invesmen and appreciaion of financial asses, including socks. Finally, he insiuional and policy facors affec housing marke dynamics, specifically he sae and he developmen of housing finance (marke srucure in he housing finance indusry, markeing of housing loan producs, refinancing opporuniies), and consumer proecion regulaion, bankrupcy law, and ax incenives (such as he axaion of capial gains and presence of inheriance axes). Empirical research has considered a number of developed and developing economies over various periods. The earlier economeric analyses included Nellis and Longboom (1981), which looked a residenial propery prices in he UK during he s, and Case and Shiller (1990), which invesigaed house price dynamics in four US ciies using quarerly daa covering

6 1986. More recenly, he deerminans of residenial house prices were examined in various individual economies: he UK (Xu & Tang, 2014), Spain (Eseban & Aluzarra, 2008; Gimeno & Marinez-Carrascal, 2010), Hong Kong (Chow & Shih, 1995; Tse e al., 1999; Leung e al., 2008), Taiwan (Chen e al., 2007; Tsai & Peng, 2011), Singapore (Lum, 2002), China (Liu & Shen, 2005; Wang & Zhang, 2013; Guo & Wu, 2013), and Iran (Pour e al., 2013). Major comparaive sudies included Greiber and Sezer (2007), who examined European economies and he US in he periods of and , respecively; as well as Hofmann (2003), who considered 20 developed economies in Europe, America, and Asia. Economeric approaches included ime-series, cross-secional and panel daa models, including convenional OLS, Engle Granger and Johansen-Juselius coinegraion, a variey of uni roo ess, Granger causaliy, Harris and Inder ess, and ohers. The resuls were raher conradicory, given he diverse mehodologies adoped and diverse variables examined. Growh of GDP, household income and wealh, and low levels of unemploymen were all found o have posiive effecs on demand, residenial house prices, and morgage repaymens in mos sudies (Nellis & Longboom, 1981; Case & Shiller, 1990; Meen, 2002; Abelson e al., 2005; Liu & Shen, 2005; Greiber & Sezer, 2007; Eseban & Aluzarra, 2008; Tsai & Peng, 2011; Guo & Wu, 2013). In Iran, however, he GDP-house price relaionship was negaive, semming from he massive oversupply of propery and large volume of consrucion aciviy during he periods of GDP growh (Pour e al., 2013). The negaive income-house price relaionship in he UK during (Xu & Tang, 2014) was aribued o conflicing consumer choices, wih households preferring oher ypes of expendiure and invesmen (such as child educaion) o real esae. Non-conemporaneous links beween disposable income, credi, consrucion cos, ineres raes, and house prices were also idenified, wih possible leads and lags, and shor- and long-run relaionships (Xu & Tang, 2014). Land supply consrains, residenial developmen resricions, and consrucion coss were principal drivers on he supply side. In he land-scarce markes of Singapore and Hong Kong (Lum, 2002; Leung e al., 2008), as well as in counries wih subsanial rural-urban migraion and fas growh of major urban areas (Guo & Wu, 2013; Wang & Zhang, 2013), hese facors were he mos salien facors in housing shorages and price growh, and were amplified by rapid income growh in he case of China. Wih regard o asse meldown hypohesis, he empirical sudies ha followed Mankiw and Weil (1989) conradiced he predicions: prices coninued o grow in he s in mos developed economies, including hose wih demographic and homeownership parameers similar o hose of he US (Engelhard & Poerba, 1991; Pikin & Myers, 1994; Piergallini, 2018). We noe ha a number of sudies of developed economies eiher found no evidence of any significan effec of demographic variables on prices (Peek & Wilcox, 1991; Forin & Leclerc, 2000), or idenified hese effecs as uneven (Levin e al., 2009). Oher sudies noneheless mainain ha a posiive link beween populaion growh and propery prices exiss (Tse e al., 1999; Liu & Shen, 2005; Eseban & Aluzarra, 2008). Overall, he size and direcion of demographic effecs is uncerain as demographic variables may be confounded wih oher variables and hus canno be isolaed. In addiion, consumpion-smoohing over he life cycle, beques moives, immigraion, capial invesmen abroad, and forward-looking financial markes may play role in offseing asse meldown. 5

7 Posiive effecs of inflaion on house price growh were experienced in Hong Kong (Chow & Shih, 1995), China (Liu & Shen, 2005), as well as in Iran, which experienced periods of hyperinflaion (Pour e al., 2013). A wo-way relaionship beween CPI and housing prices is noed, given ha housing is included in he CPI calculaion (Liu & Shen, 2005). The role of moneary and credi variables and financial markes in explaining house price movemen was emphasised by Greiber and Sezer (2007) and Goodhar and Hofmann (2008). An increase in money supply and low ineres raes were found o be associaed wih he formaion of real esae bubbles and a susained increase in propery prices (Tsai & Peng, 2011). Srong negaive effecs of ineres raes on prices were experienced in European economies, including he UK, and in Hong Kong, whils in he US he effecs were no significan (Nellis & Longboom, 1981; Chow & Shih, 1995; Suon, 2002; Greiber & Seser, 2007; McQuinn & O Reilly, 2008). In China, a posiive relaionship beween ineres raes and house prices was idenified, in conras o he majoriy of findings (including hose of Guo and Wu, who idenified a negaive relaionship in he Shanghai real esae marke). Liu and Shen (2005) aribue his o he regulaion of ineres raes by he Chinese governmen. For credi variables, a saisically significan co-movemen beween credi cycle and house prices was idenified in he USA (Greiber & Seser, 2007). The sudy of house prices in Spain by Gimeno and Marinez-Carrascal (2010) and Eseban and Aluzarra (2008) likewise esablished a link beween he cos of credi, demand for housing, and he level of household indebedness. Here, he cos of morgage financing was found o be negaively relaed o demand and propery prices, and hus he level of deb. Regarding he ineracion beween sock and propery markes, he majoriy of empirical sudies idenified posiive correlaion beween he wo, and in many insances long-run relaionships and he causaliy running from sock o real esae prices. The sudies include Hoesli and Hamelink, 1997 (Swizerland); Lizieri and Sachell, 1997 (UK); Bonnie, 1998 (USA); Abelson e al., 2005 (Ausralia); Oikarinen, 2010 (Finland), among ohers. Several reservaions were made. Firsly, he sock marke could affec real esae hrough naional income (Suon, 2002); secondly, he effecs could be mos pronounced in he expensive segmen of he propery marke (Kakes & Van Den End, 2004); hirdly, in some cases he effecs on he sock marke could be shor-erm in naure (Leung e al., 2008), whils in oher cases, he long-run coinegraion beween he wo markes could be presen (Takala & Pere, 1991); and fourhly, conemporaneous relaions beween he wo markes could be absen (Quan & Timan, 1999). Chen e al. (2007) argue ha he cos of consrucion, he cos of land, and seasonal facors were he principal deerminans of Malaysian residenial propery prices. Ong (2013) suggess ha he quaniy and qualiy of labour force in he consrucion secor affec he supply of housing. In addiion, he Real Propery Gains Tax (RPGT) provides incenives for lae disposal of properies and affecs he purchase decisions of households, including demand for residenial properies. The effec of RPGT was negaive, hough speculaion aciviies and purchasing decisions by high-ne-worh individuals were no affeced by RPGT. Tan (2010) considers he influence of ineres raes on residenial housing prices. Using quarerly daa and pooled random effec model, Tan shows a negaive relaionship beween he level of he base lending rae and he volume of residenial properies aciviies and purchases. 6

8 Furhermore, a non-significan relaionship was found beween he volume of residenial propery ransacions and average propery prices, specifically for cerain local markes and properies of inferior qualiy (hose wih unsaisfying locaions, for example). Zandi e al. (2015) examines he effecs of macroeconomic facors on properies locaed in urbanised and indusrialised areas (Penang sae) in he period of A significan posiive relaionship wih possible lagged effec beween he lending rae and price level was idenified, despie he majoriy of previous sudies idenifying a negaive relaionship. GDP and gross naional income had posiive bu insignifican effecs on prices, whils inflaion had no effec. In a similar analysis of macroeconomic deerminans, Hui (2013), using quarerly daa, considers he effecs of privae consumpion, gross invesmen, sock prices, money supply, ineres raes, and bilaeral exchange raes on housing prices in Malaysia in he period of Gross domesic produc, invesmen, and sock price did no subsanially affec housing prices. In conras, exchange raes were found o have negaive effec on prices. Money supply was also seen o have a relaively srong effec on housing prices, wih Granger causaliy running in boh direcions. In conras, an analysis of macroeconomic variables by Ong (2013) esablishes a posiive effec of Malaysian GDP and populaion growh on prices, and an absence of effec by ineres raes. In he broader conex of housing affordabiliy, Bujang e al. (2015) consider he dynamics of residenial markes in he sae of Johor and he associaed effecs for he naive ehnic group (Bumipuera). I was found ha despie he poenial o moderae prices, he policies and producs provided by financial insiuions did no have srong effecs on demand and prices. Wih ongoing rural-urban migraion, demand was rising subsanially, effecively precluding Bumipuera from home ownership. Despie loans wih more aracive feaures, he defaul raes and repaymen burden remained high, paricularly among lower-income Bumipuera. An analysis of housing markes and affordabiliy by Hashim (2010) delivers more opimisic findings. Indeed, low affordabiliy of housing was becoming a problem for Malaysia, wih incomes lagging behind propery prices, and an insufficien supply of low and medium cos properies. However, levels of affordabiliy were no uniform. There was a subsanial dispariy in housing prices, semming from income and regional developmen inequaliy, wih he poores saes (Kelanan) experiencing moderae levels of unaffordabiliy in some years. In addiion, a Cenral Bank inervenion resuled in lower ineres raes and loan cheapening, whils solid economic growh led o a gradual rise in household incomes. Given he conradicory findings, as well as he fac ha many of he sudies were conduced for specific regions of Malaysia and focused on he effecs of policy insrumens or supply side facors, he effec of macroeconomic and demographic variables on residenial propery markes mus be re-examined. This sudy used more recen daa, exending o Acknowledging he local naure of markes and regional dispariies, his sudy focused on he overall dynamics of residenial housing in Malaysia and used aggregae housing price index as a dependen variable. 7

9 Mehodology Daa descripion This sudy aimed o examine he relaionship beween Malaysian residenial propery prices (real residenial propery price index wih a base in 2010, calculaed using he hedonic price mehod) and heir deerminans, specifically he lending rae, inflaion rae, real household income in 2010 consan values, real gross domesic produc in 2010 consan values, and he counry s populaion. The sudy was conduced using quarerly daa from he period. Residenial propery was defined o include erraced houses, semi-deached houses, deached houses, and high-rise unis. As shown in Table 1, he secondary daa were rerieved from he CEIC Global Daabase, Deparmen of Saisics of Malaysia, Thomson Reuers DaaSream, and he daabase of he Naional Propery Informaion Cenre (NAPIC). Table 1: Variables and Corresponding Secondary Daa Sources Variables Secondary Daa Source House Price Index (HPI) CEIC Daabase Base Lending Rae (BLR) Thomson Reuers DaaSream Consumer Price Index (CPI) CEIC Daabase Household Income (HI) Malaysia Deparmen of Saisics Populaion Growh (POPGROWTH) Gross Domesic Produc (GDP) Thomson Reuers DaaSream CEIC Daabase Theoreical model The heoreical model is formulaed wih residenial house prices as he funcion of he base lending rae, inflaion rae, counry populaion, gross domesic produc (GDP), and household income: A linear funcional form is adoped: HPI BLR INF HI P GDP (1) Where HPI is house price index in year, inflaion rae in year, HDI is household income in year, BLR is he base lending rae in year, P is populaion in year, INF is GDP is gross domesic produc in year, 0 is consan, 1, 2, 3, 4, 5 are parial regression coefficiens, and is an error erm. To ensure direc inerpreaion of coefficiens as percenage changes, he model is run on he naural logarihms of he series. A posiive relaionship is hypohesised beween house prices and populaion, GDP, household income, and inflaion; whils a negaive relaionship is likely o be presen beween house prices and he base lending rae. Thus, HPI HPI HPI HPI HPI 0, 0, 0, 0, 0. BLR INF HDI P GDP 8

10 Economeric mehod The saionariy and he order of inegraion of he variables were esed by he Augmened Dickey-Fuller (ADF) and Phillips-Perron (PP) ess. The Augmened Dickey-Fuller (ADF) es is based on he auxiliary regression as follows: Y b b b Y b Y e i i i 1 k (2) where y is he logarihm of he respecive variable a ime, y 1 is he lag in he firs difference, is rend, e is he error erm adjused for serial correlaion and k is he number of lags chosen o remove serial correlaion. The null and alernaive hypoheses are represened as: H0: b2 0 (The series is non-saionary and uni roo is presen) Ha : b 1 (The series is saionary and uni roo is absen) 2 The Phillips-Perron (PP) uni roo es is used o complemen he ADF es, which suffers from several shorcomings including low power in he case of near uni roo processes. Being a nonparameric es, i deals wih serial correlaion and heeroscedasiciy of any ype by making correcions o -saisic in he es regression hereby making i unnecessary o specify lag lenghs, as in he ADF es. The es regression is given as: ' y D y 1 (3) where D represens deerminisic erms, I(0) and correcion for serial correlaion and heeroscedasiciy is performed on. The modified saisics are: Z Z 1/2 ^ T SE( ) ^ 2 ^ ^ SE 2 2 ^2 ^ 1 T ( ) T ( ) 2 (5) (4) ^2 where and ^2 are consisen esimaes of variance. The wo hypoheses se for PP es are similar o hose in he ADF es: H 0 : 0 (The series is non-saionary and uni roo is presen) 9

11 H a : 1 (The series is saionary and uni roo is absen) The Johansen-Juselius coinegraion approach is adoped o es he presence of he long run relaionship among six variables. The coinegraion long-run equaion is derived as: k 1 Z Z Z D (6) 1 i i i 1 Where Z are variables examined, is a whie noise disurbance wih zero mean and finie variance, D is a vecor of deerminisic variables, is he n n coefficiens marix, wih he rank of deermining he number of coinegraing vecors. Two likelihood raio ess (he maximum eigenvalue and he race es) are used o idenify he number of coinegraing vecors (Johansen, 1988). The race es saisic is represened as follows: p T ln(1 ) (7) race i q 1 i Where race is he likelihood raio saisic, T is he number of observaions used in he esimaion, p sands for he number of variables, and q is he rank of marix. The null hypohesis is ha he rank of marix (and hence he number of coinegraing relaionships) is q, in effec: H0 : rank( ) q. The alernaive hypohesis is ha he rank of he marix is higher han q bu lower or equal o n, which is defined as he maximum possible number of coinegraing relaions, i.e. Ha : q rank( ) n. The race es proceeds sequenially unil he firs non-rejecion of he null. The maximum eigenvalue es saisic is shown as: T ln(1 ) (8) max q 1 The null hypohesis is ha here are no coinegraing relaions and he rank of marix is zero, in effec: H0 : rank ( ) 0. The alernaive hypohesis is ha here is a single coinegraing relaion, in effec: Ha : rank( ) q 1. Thus, he es considers sequenially wheher he larges eigenvalue is zero (under null hypohesis) or he nex larges eigenvalue is zero (under he alernaive hypohesis). Given ha he resuls of he Johansen-Juselius ess may suffer from size bias (over-rejecion of he null hypohesis of no coinegraion), a correcion o race and maximum eigenvalue saisics may be needed in small samples. We adop he correcion facor derived by Reinsel and Ahn (1992) as ( T pk) / T, where T is he sample size, p is he oal number of variables, and k is he lag lengh. 10

12 Following deerminaion of he number of coinegraing vecors and of he fac ha variables comove in he long-run, Granger causaliy is examined in order o esablish he direcion of influence among he variables in he shor- and long-run. Granger causaliy is esed wihin he VECM framework if variables in levels are nonsaionary and here exiss a coinegraing relaionship among hem. In conras, in he absence of coinegraion, he VAR model is run on he differenced variables and causaliy is esablished in he VAR conex. VECM (as a resriced VAR for nonsaionary variables) resrics he long-run dynamics, whils allowing for shor-run adjusmens owards long-run equilibrium. For correcly specified VECM, he respecive value of he error correcion erm is negaive and belongs o he (0; -1) range, hereby allowing such adjusmen. In VECM, wo sources of causaliy are idenified: long-run, capured by error-correcion erm (ECT), and shor-run, capured by lagged explanaory variables. The VECM in a wo-variable case augmens bivariae VAR and is represened as follows: Y e Y Y X X (9) k k k k 1 X X X Y Y (10) k k k k 2 where e 1 and 1 are he lagged value of ECT, and 1 and 2 are whie noise disurbances. For a five-variable case, and residenial propery prices as he dependen variable, VECM is rewrien as: k k k k k k HPI HPI BLR INF HI P GDP 1 1 2i i 1 i 1 i 1 i 1 i 1 i 1 ( HPI BLR INF HI P GDP ) (11) Where is he ECT coefficien and hereby indicaes he adjusmen o long-run equilibrium, ( HPI 1 1BLR 1 2INF 1 3HI 1 4P 1 5GDP 1 ) is he error-correcion erm, is a whie noise disurbance, and are consans, and 1 6are coefficiens of he variables ha indicae a shor-run relaionship. The null hypohesis for he VECM based Granger causaliy es is ha variables do no cause one anoher; in effec: in Equaion (9) X does no Granger-cause Y and in Equaion (10) Y does no Granger-cause X. Thus, H0 : 11 1k 0 or H0 : 21 2k 0. Respecively, shor-run Granger causaliy in Equaion (9) is esablished if coefficiens 11 1k are joinly significan. Likewise, 21 2k are joinly significan in Equaion (10). The presence of significan coefficiens in boh equaions would poin o bilaeral causaliy. The saisical significance of e 1 and 1 would indicae long-run Granger causaliy. 11

13 Forecas variance decomposiion is performed o show he breakdown of he forecas error variance for a variable a differen horizons, or o idenify how shocks o specific variables pass hrough he sysem. Specifically, he mehod compares he relaive imporance of populaion growh, GDP, inflaion, ineres raes, and household income in explaining propery prices, and given ha forecas error variances evolve over ime esablishes how hese relaive conribuions change. Empirical resuls As a firs sep, uni roo ess were conduced. The ADF uni roo es resuls (Table 1) show ha all variables are non-saionary in levels (under inercep and inercep plus rend specificaions). For insance, he -saisics for he naural logarihm of he house price index is (in levels and wih inercep), which is smaller han he criical value of a a 1% level of significance. Therefore, he null hypohesis is no rejeced and i is concluded ha he naural logarihm of he house price index conains a uni roo. ADF uni roo es resuls indicae ha all of he variables in quesion are saionary in firs differences (under boh specificaions menioned above), indicaing he absence of a uni roo. For insance, he -saisics for he naural logarihm of he consumer price index is in he firs difference (specificaion wih inercep), exceeding he criical value of a a 1% level of significance. Table 1: Resul of Augmened Dickey Fuller (ADF) Uni Roo Tes Variables Level 1s Difference Inercep Trend and Inercep Inercep Trend and Inercep LHPI *** *** 0 LBLR *** *** 0 LCPI *** *** 0 POPGROWTH *** *** 0 LINCOME *** *** 0 LGDP *** *** 1 Noe: LHPI is he naural logarihm of house price index, LBLR is he naural logarihm of base lending rae, LCPI is he naural logarihm of consumer price index, POPGROWTH is populaion growh, LINCOME is naural logarihm of household income, LGDP is he naural logarihm of gross domesic produc. (*), (**) and (***) indicae 10%, 5% and 1% levels of significance, respecively. Lags are seleced based on Schwarz Informaion Crierion (SIC). Figures in parenheses are lag lenghs. The PP es yields similar resuls (Table 2). All variables are non-saionary in levels under boh inercep and rend plus inercep specificaions. The PP es also shows ha all variables are saionary in firs differences. Given ha all variables are I(1), or non-saionary in levels bu saionary in firs difference, he use of he Johansen-Juselius mehodology is jusified. 12

14 Table 2: Resul of Philips-Perron (PP) Uni Roo Tes Variables Level 1s Difference Inercep Trend and Inercep Inercep Trend and Inercep LHPI *** *** 3 LBLR *** *** 1 LCPI *** *** 8 POPGROWTH *** *** 2 LINCOME *** *** 3 LGDP ** * 3 Noe: As per Table 1. Table 3 (below) shows he resuls obained from he Johansen-Juselius coinegraion es. The number of lags of he firs differenced erms o be used in he es is se o he minimum: k = 2. Given ha series have non-zero mean and likely conain sochasic rends, he es is run wih inercep (bu no rend), which belongs only in he coinegraing relaion. The maximum number of relaionships is se o r = 5, and he es is run sequenially unil he firs non-rejecion of he null hypohesis. Table 3: Resul of Johansen-Juselius Coinegraion Tes Trace Max Eigenvalue Null Alernaive Unadjused Adjused 99% C.V. Unadjused Adjused 99% C.V. k=3, r=1 r = 0 r = * * r 1 r = r 2 r = r 3 r = r 4 r = r 5 r = Noe: The lag lengh is k and r is he number of coinegraing vecors under race and maximum eigenvalue ess. (*) represens rejecion of he hypohesis a 1% level of significance. The adjused es saisics for he race es is , exceeding he criical value of a a 1% level of significance. Thus, he null hypohesis of no coinegraion (r = 0) is rejeced for LHPI, LBLR, LCPI, POPGROWTH, LINCOME, and LGDP. The maximum eigenvalue es saisic is , exceeding he criical value of a 1% level of significance. Since he null hypoheses for he maximum eigenvalue and he race ess are similar, he null hypohesis is also rejeced. Jus one coinegraing vecor wih a 1% level of significance is idenified, suggesing ha here is a single long-run relaionship beween he variables (LHPI, LBLR, LCPI, POPGROWTH, LINCOME, and LGDP). The normalised coinegraing vecor is represened as follows (wih - saisics indicaed in parenheses): LHPI POPGROWTH LGDP LINCOME 2.153LCPI 0.160BLR (14.071) ( 5.641) ( 8.242) (7.476) ( 2.318) 13

15 The coefficien of he populaion growh variable has a posiive sign, indicaing ha when here is a 1% increase in populaion growh, here is a 0.999% increase in residenial house prices. Wih regard o he GDP variable, he respecive coefficien has a negaive sign, suggesing ha a 1% increase in GDP brings a 0.854% decrease in he level of residenial propery prices. A similar negaive relaionship is observed beween he naural logarihm of household income and level of prices (a 1% increase in he former resuling in a 0.294% decrease in he laer). Changes in he consumer price index have a posiive associaion wih changes in residenial propery price levels, wih a 1% increase in he consumer price index bringing an increase of 2.153% in he level of prices. This is in line wih he majoriy of sudies, which poin o he posiive conribuion of consumer goods inflaion o he residenial propery inflaion. Finally, here is a negaive relaionship beween he base lending rae and level of residenial propery prices, wih a 0.16% decrease in he house price index per 1% increase in lending rae. A possible explanaion for he phenomena is ha a decrease in he base lending rae conribues o a cheapening of morgages and home loans, hereby encouraging residenial propery purchases, in urn resuling in higher demand for homes and higher prices. (In he conex of he Malaysian propery marke developmens in recen years, a negaive relaionship beween base lending rae and prices would likely reflec he spur in speculaive aciviies.) A posiive relaionship beween populaion growh and propery price change is no surprising, given he saus of Malaysia as a counry wih rapid (albei deceleraing) populaion growh (Deparmen of Saisics, 2015), and he effec ha populaion growh has on he demand for residenial propery. The resuls are herefore consisen wih Liu and Shen (2005), Ong (2013), Guo and Wu (2013), and Tse e al. (1999). The negaive long-run relaionship beween GDP and residenial propery prices is unexpeced. A heoreical explanaion is proposed by Pour e al. (2013): ha consisenly high economic growh may over-simulae residenial consrucion, resuling in excessive residenial consrucion and a decrease in propery prices. Indeed, as documened by Bujang e al. (2010), he propery overhang in Peninsular Malaysia in 2005 has been subsanial, wih he number of unsold properies oalling 755,000 unis. A similar propery oversupply has been observed by Bank Negara Malaysia in recen years, paricularly in he higher price segmen (Ling e al., 2017). This developmen is coupled wih undersupply of housing in oher segmens, resuling in decreased affordabiliy and higher propery prices faced by lower- and middle-income households. The negaive relaionship beween household income and propery prices is suppored by he findings by Xu and Tang (2014) and Bujang (2010). The consumpion preference owards rening (raher han ownership of propery), or owards paricular ypes of propery and oher consumpion iems (vehicles, for example) may dampen demand for propery, despie growing income levels. As noed by Bujang (2010), housing developmen decisions by local and sae auhoriies end o be based on populaion rends, wihou due sudy of consumer preferences. For insance, in Johor Bahru, a survey of households idenified a preference for new housing schemes wih beer design and higher qualiy, implying he persisence of a propery overhang in cerain caegories. The preference for higher qualiy lifesyle housing was confirmed by Sarip and Lee (2015). 14

16 An addiional facor behind he negaive income-price relaionship, which was likely o offse higher propery prices (due o populaion growh and a lack of affordable housing consrucion), is a persisen gap beween populaion and GDP growh on one hand and income and wage growh on he oher, resuling in low labour share of GDP. In he propery marke, his endency leads o lower demand paricularly among lower-income or newly formed households and exercises a downward pressure on prices, despie moderae income growh. Empirical evidence appears o provide some suppor (albei no unequivocal) o he GDP-income gap hypohesis. On one hand, he change in household income in recen years has been posiive and he labour share of income has increased (Ng, 2017). On he oher hand, some of he developmens in he labour marke were adverse. Despie he Economic Transformaion Programme (ETP) launched in Sepember 2010, and allied iniiaives, he majoriy of jobs ha have been creaed since hen have been in he low and mid skill and salary segmens, whils he unemploymen rae for he age group has remained high (Lim, 2016). Similar endencies are idenified by Rasiah e al. (2015): since he lae 1990s, he labour marke in Malaysia has been characerised by slow manufacuring wage growh, increased presence of foreign workers, ousourcing and conracing, and a low degree of unionisaion. The Johansen-Juselius coinegraion es deermined ha here is a single coinegraing vecor, and uni roo ess deermined ha variables (when represened in levels) were non-saionary. Thus, he use of vecor error correcion model (VECM) and subsequen esing for causaliy in VECM conex are jusified. The vecor error correcion model (VECM) Granger causaliy es resuls are shown in Table 4 below. There exiss unidirecional shor-run Granger causaliy significan a he 1% level from populaion growh o household income, from GDP o consumer price index, from he base lending rae o consumer price index, and from personal income o he house price index. In addiion, a a 5% level of significance, he unidirecional shor-run causaliy runs from populaion growh o GDP, and from GDP o he base lending rae. A a 10% level of significance, house price index Granger causes household income, populaion growh Granger causes house price index, and household income Granger causes populaion growh. Overall, wo bidirecional causaion cases are idenified: house price index and income, and populaion growh and income. In he long-run, Granger causaliy runs from house price index o populaion growh. Given ha his sudy focused specifically on he deerminans of residenial propery prices, he evidence from Granger causaliy ess suggess ha shor-run causaliy runs from demographics and broader economic facors (growh of populaion and income) o residenial propery prices, bu no from moneary variables, such as ineres raes (nowihsanding he fac ha he long-run equilibrium relaionship indicaes coinegraion beween lending raes and consumer price index on one side and house prices on he oher). This resul is in line wih he findings of Tse e al. (1999), Liu and Shen (2005), and Eseban and Aluzarra (2008) in he conex of OECD and seleced developing economies, and Ong (2013) in he Malaysian conex. 15

17 Table 4: VECM Granger causaliy Variables LHPI POP GROWTH LGDP LCPI LBLR LINCOME ECT χ 2 -Saisics Coefficiens -sa LHPI (0.052) (0.981) (0.900) (0.429) (0.000)** POPGROWTH (0.491) (0.976) (0.765) (0.177) (0.077) LGDP (0.650) (0.024)** (0.487) (0.143) (0.784) LCPI (0.340) (0.196) (0.001)** (0.000)** (0.844) LBLR (0.819) (0.538) (0.037)** (0.727) (0.669) LINCOME (0.069) (0.000)** (0.996) (0.708) (0.165) Noe: The Block Exogeneiy Wald Tes is employed o es for causaliy. (*), (**) and (***) indicae rejecion of he null hypohesis and saisical significance of 1%, 5% and 10%, respecively. The lag order of VECM was se o 3. T-saisics are indicaed in parenheses. 16

18 Anoher imporan implicaion of he Granger causaliy es findings is ha residenial propery prices are sensiive o changes in household income, bu no o GDP growh, hereby poining o he imporance of disribuional and affordabiliy issues and policy incenives, raher han economic growh per se. The long-run causaliy from propery prices o populaion growh may appear spurious; however, as argued by Mulder (2006), he effec ha house prices have on household formaion and marriage, as well as on decisions o have children, is significan in many insances. Wheher hese effecs are presen in Malaysia (and wheher economic facors have primacy over culural, religious and oher non-economic facors, paricularly among Muslim Malays) is a opic ha requires separae analysis. We also noe ha he Granger causaliy resuls presened above indicae emporal ordering variables and 'firsness' of any paricular variable, raher han causaliy in a sric economic and philosophical sense. Defined in erms of predicabiliy and forecasing power, Granger causaliy may, however, provide some indicaion of he rue economic influence (Geweke, 1984; Dawson, 2003). Resuls of he forecas variance decomposiion for endogenous and exogenous variables are presened in Table 5. The respecive forecas horizon is se a en quarers (hough only he values for he second, fifh, sevenh, and enh quarers are shown), and facorisaion using Cholesky decomposiion is performed. The sandard error column represens he forecas error for respecive variables a differen horizons. I can be seen ha own shocks o propery prices are salien boh in he shor- and long-run, conribuing o 95% of propery price forecas error variance afer wo quarers and 94% afer en quarers. The relaive conribuion of he consumer price index o propery price forecas error variance rises o 4% afer en quarers, whils oher variables are insignifican in explaining changes in propery prices in he shor- and long-run. Propery price index is herefore he mos exogenous variable in he sysem. Similar paerns are observed in he case of variance decomposiion for consumer price index and he base lending rae, where own shocks conribued o 71% and 84% of variance in he respecive variable afer en quarers. In conras, he own shocks o GDP, populaion growh, and household income are of a smaller magniude, conribuing o 42%, 43%, and 4% of variance in he variables afer en quarers. Therefore, GDP, populaion growh, and household income are he mos endogenous variables. Table 5: Forecas variance decomposiion resuls Variance decomposiion of HPI Period S.E. LHPI LCPI LBLR LGDP POP GROWTH LINCOME Variance decomposiion of CPI

19 Variance decomposiion of BLR Variance decomposiion of LGDP Variance decomposiion of POPGROWTH Variance decomposiion of LINCOME Conclusion This sudy aimed o explain he macroeconomic and demographic deerminans of Malaysian residenial propery prices during he period of Quarerly daa was used and imeseries economeric mehods (uni roo ess, Johansen-Juselius coinegraion, Granger causaliy based on VECM, and variance decomposiion) were employed. I was found ha populaion had a significan and posiive effec on he demand for residenial properies and, consequenly, on he price of residenial properies in Malaysia. The resuls obained are in line wih hose of he majoriy of previous sudies (Liu & Shen, 2005; Ong, 2013; Guo & Wu, 2013; Tse, Ho & Ganesan, 1999). Gross domesic produc was negaively and significanly relaed o he prices of residenial properies in Malaysia. The resul obained is in line wih ha of Pour e al. (2013): economic growh gives rise o housing supply expansion, which depresses prices. Given ha propery prices coninued o grow over a susained period, i is unlikely ha his effec played a dominan role excep in he high-end segmen of he propery marke, where oversupply was eviden. The exisence of a significan and negaive relaionship beween household income and housing prices is suppored by he findings of Xu and Tang (2014), Bujang (2010), and Rasiah e al. (2015): more sophisicaed consumer preferences, which are no mached by he curren housing supply, coupled wih a growing rend for rening raher han owning propery (paricularly in overpriced locaions), and income growh no caching up wih propery prices, were likely o exer a downward pressure on he demand side, hus lowering prices. 18

20 The sudy also idenified a negaive and significan relaionship beween base lending rae and prices for residenial properies, in line wih economic heory (high financing coss discouraging propery purchases) and empirical research by Tan (2010), Nellis and Longboom (1981), and Guo and Wu (2013). Finally, a significan and posiive relaionship beween he consumer price index and residenial propery prices (he laer being a consiuen par of he former) was observed, in line wih previous findings by Liu and Shen (2005) and Pour e al. (2013). Several policy measures (some of which have already been acceped by regulaors) follow from his paper's findings. Firsly, given he srong dependence of prices on demographic facors, ongoing supply expansion is needed, based on he idenificaion of consumer preferences for qualiy housing and he paricular areas where he affordabiliy crisis is he mos acue. This requires furher modernisaion and growh of he consrucion indusry in order o cope wih growing demand, as well as adjusmen of consrucion sandards and proocols. Secondly, oher supply-side measures may be needed, including greaer allocaion of land for residenial developmen, in order o decrease he high cos of land for residenial housing: a deparure from previous land policies ha have envisaged subsanial se-asides of land for public areas and commons. Of paricular imporance are measures o simulae residenial consrucion in he lowprice range. Thirdly, given he negaive relaionship beween propery prices and incomes, furher social policy and economic resrucuring measures may be required o address significan income inequaliy and wages lagging behind propery price growh. On a broader level, as proposed by he Naional Economic Advisory Council (2010), a move owards a new economic model based on labour force skill upgrading, more vigorous produciviy growh, and specializaion in high value added producion may be necessary. Finally, given ha low ineres raes in recen years have fuelled propery speculaion and brough in higher prices across he board, sricer regulaion of propery financing is jusified; for example, measures o resric lending o buyers wih muliple propery porfolios, and assessmen of oal deb obligaions prior o lending. A fuure analysis of he propery marke drivers in Malaysia could consider addiional deerminans, such as he sae of he renal marke, he level of wealh (as measured by he aggregae value of he sock marke), he influence of exchange raes on propery invesmen decisions by foreign buyers, he elasiciy of he housing supply, and he sae of he consrucion indusry. In addiion, given he high level of income sraificaion in Malaysia and he segmened naure of he propery marke (wih propery oversupply in he op-end segmen being paralleled by under-supply and rising prices in he low end), fuure research could consider specific secions of he marke and he purchasing decisions of paricular groups of households, such as working aduls or newly formed households. References Abelson, P., Joyeux, R., Milunovich, G., Chung, D. (2005). Explaining House Prices in Ausralia. Economic Record, 81 (s1), pp Bernanke, B., Gerler, M. (1989). Agency Coss, Collaeral and Business Flucuaions. American Economic Review, 79 (1), pp