Subdivided Research on the Inflation-hedging Ability of Residential Property: A Case of Hong Kong

Transcription

1 Subdivided Research on he -hedging Abiliy of Residenial Propery: A Case of Hong Kong Guohua Huang 1, Haili Tu 2, Boyu Liu 3,* 1 Economics and Managemen School of Wuhan Universiy,Economics and Managemen School of Eas China universiy of Technology,China 2 Economics and Managemen School of Eas China Universiy of Technology, China 3 Counry Garden Holdings Company Limied Hubei Region, China *Corresponding auhor(boyu Liu, @qq.com) Absrac Wha kind of residenial real esae can effecively hedge he inflaion risk is an imporan issue ha households concern. In his paper, he long run model and he shor run error correcion model are used o analyze he hedging abiliy of residenial real esae in he Hong Kong o he expeced and unexpeced inflaion. The inflaion rae is divided ino 4 groups according o differen calculaion mehods, he residence is divided ino 5 groups according o he area size. Empirical resuls show ha, in he long run, all ypes of residenial real esae in Hong Kong have he abiliy o hedge agains inflaion. In he shor run, all ypes of residenial real esae have he sable abiliy o hedge agains he unexpeced inflaion, bu differen residenial real esae has differen abiliy o hedge agains he expeced inflaion. This implies ha households can inves in residenial real esae o hedge inflaion risk, bu hey should also ake he differen hedging abiliies of differen residenial real esae ino consideraion o choose he suiable one. On he oher hand, he governmen s real esae marke regulaion policy should also ake ino accoun his difference o opimize he srucure of regulaion o improve regulaory efficiency. Key words: Residenial real esae, risk, Abiliy o hedge, Long run effecs, Shor run effec 1. INTRODUCTION refers o he economic phenomenon in which general price level rises persisenly and hereby currency depreciaes, when money supply subsanially exceeds is demand. For households, due o heir weak invesmen abiliy and limied invesmen channels, when facing wih inflaion, he real value of heir asses and income will drop and heir living sandard will decline, even heir daily life will also be affeced. Thus how o effecively hedge inflaion risk has become an imporan bu difficul issue. This paper aemps o answer wheher he residenial real esae can hedge inflaion and wheher residences wih differen area size vary in heir abiliy o hedge agains inflaion. There is exensive lieraure examining he hedging abiliy of real esae agains inflaion, bu hese sudies provide mixed resuls and reach no consensus. Some evidence suggess ha real esae can serve as a sable inflaion hedge. Based on coinegraion es and impulse response analysis, Li and Guan (2011) conclude ha here is a wo-way causal relaionship beween China's real esae prices and inflaion. Glascock e al (2002) use VEC model and find ha US REITs have a negaive relaionship wih real inflaion, expeced inflaion and unexpeced inflaion. Based on boh auoregressive disribued lag (ARDL) models and recursive regressions, Anari and Kolari (2002) find ha house prices are a sable inflaion hedge agains inflaion in long run. However, here some empirical evidence shows ha he real esae has no hedging abiliy agains inflaion. Glasock e al (2010) analyze he hedging abiliy of real esae in Hong Kong boh in long run and he shor run. They find ha differen ypes of real esae have differen characerisics during differen periods, and ha he real esae is no a good hedge agains inflaion in Hong Kong. In addiion, some sudies indicae ha he hedging abiliy of real esae agains inflaion is no sable. In some special cases or during a specific period, he hedging abiliy may change. Di (2012) argues ha in he long run real esae is an effecive hedge agains inflaion, bu under exreme condiions, invesmen in real esae will resul in a large loss raher han hedge agains inflaion. Hardin e al (2012) discover ha alhough he REITs have he hedging abiliy, he hedging abiliy can be dominaed by he inflaion illusion effec over cerain periods. In he meanime, more aenion is paid o sudy he difference of hedging abiliy of real esae agains inflaion beween he long run and he shor run. Some sudies argue ha he hedging abiliy is week in he shor run, bu srong in he long run. Duan (2007) analyzes he mechanism of house prices affecing inflaion and oupu, and find ha he effecs of house prices on inflaion is very limied in he shor run, bu significan in he long run. Hoesli e al (2008) argues ha in he long run, REITs are a beer hedge agains expeced inflaion han he privae real esae, while boh of hem canno hedge agains inflaion in he shor run. However, some 399

2 scholars believe he hedging abiliy of real esae is sronger on shor erm han on long run. Chen and Sing (2006) find ha London residenial real esae has cerain hedging abiliy agains shor erm inflaion. Ganesan and Chiang (1998) use he model proposed by Fama and Schwer (1977) and find ha four ypes of real esae in Hong Kong can hedge agains inflaion in he shor run, bu are no able o hedge agains inflaion in he long run. Qiu (2011) uses similar model o sudy China real esae and find similar paerns. The mixed resuls may be caused by differen mehods and daa, bu we aim o es if he mixed resuls is caused by he differen characerisics of he real esae, such as he ype and he area. In his paper, he model proposed by Fama and Schwer (1977) is used o sudy he hedging abiliy of differen residenial real esae. The residenial real esae is divided ino 5 groups according o is area size, he inflaion rae is divided ino 4 groups. For each group of inflaion rae and residenial real esae, he hedging abiliy of residenial real esae agains expeced and unexpeced inflaion is analyzed. The remainder of he paper is srucured as follows. The nex secion briefly presens he mehod. Secion 3 describes he daa. The empirical resuls along wih discussions are presened in Secion 4. The las secion is he conclusion. 2. METHOD The mehod proposed by Fama and Schwer (1977) is used in he paper. Firs, he inflaion is decomposed ino expeced and unexpeced inflaion, hen he hedging abiliy of residenial real esae agains his wo componens is examined. Following he sudy of Hoesli e al (2008), hree ARIMA models are used, he ARIMA (0,1,1), ARIMA (1,0,3) and ARIMA (1,3,3). In addiion, we will also use wo moving average models MA (3) and MA (4) o esimae he expeced inflaion rae. The following model is used o es he feasibiliy of hese mehods o esimae he expeced inflaion rae: I E I e where I is he acual inflaion rae, E I is he expeced inflaion rae. If we chose an appropriae mehod, he esimaed resul should have a relaively high R 2, and should no be significan differen from 0 and should no be differen from 1. Once he expeced inflaion rae is calculaed, he unexpeced inflaion rae can I E I be derived from. Firsly, he order of inegraion is esed for each variable. For coinegraed variables, hey should have he same order of inegraion. In his paper, he DF es, ADF es and PP es are used o examine he saionariy of ime series. If all variables are inegraed of order 1, he following equaion can be used o examine he long run or equilibrium relaionship: R E I I E I where R is he asse reurn a ime, E I (1) (2) is he expeced inflaion rae a ime. 1 represens he co- represens he co-movemen feaure of he asse movemen feaure of he asse and expeced inflaion while 2 and unexpeced inflaion, which can reflec he hedging abiliy of he asse o he inflaion in he long run. Based on he coinegraion and he long run relaionship, he error correcion model (ECM) is used o esablish a shor run relaionship of variables. The shor-erm changes in asse reurns are driven by changes in he inflaion in he long run relaionship and by adjusmens o previous disequilibrium in he long run relaionship. In he shor run model, he firs differences of variables and adjusmens o previous disequilibrium in he long run model are used as explanaory variables. R E I I E I R u i 1 i 1 where represens he firs differences, he k lags of he firs differences of asse reurn are also expansionary variables in he shor run model. Coefficien =-1 indicaes he degree of adjusmen: means full -1< <0 adjusmen, 0 means parial adjusmen and means no adjusmen. Generally, he coefficien should be negaive o ensure ha he shor run relaionship can adjus o he long run or equilibrium relaionship, while posiive coefficien indicaes he deviaion of he shor run dynamic relaionship from he long run relaionship. 3. DATA SAMPLE k (3) 400

3 A quarerly daase is assembled from Hon Kong from Q o Q The Composie CPI, CPI, CPI(B), CPI(C) are obained from he websie of Hong Kong Moneary Auhoriy o calculae he inflaion rae1. The Composie CPI is consruced on he basis of he expendiure paerns of all he households in Hong Kong, while he CPI, CPI(B) and CPI(C) are based on he expendiure paerns of households in he relaively low, moderae and high expendiure range2. Privae Residenial Renal Index - (E), as well as Privae Residenial Price Index - (E) are obained from he Raing and Valuaion Deparmen in Hong Kong3. The capial leers from A o E represen five differen residenial areas4. In addiion, he hree-monh bond ineres raes are obained from Hong Kong Moneary Auhoriy. The quarerly CPI on a year-on-year basis is used o calculae he acual inflaion rae. The oal reurn for residenial real esae is he renal yield and he sale reurn. In his paper, he renal and price index are used o calculae he residenial real esae reurn. Table 1 inflaion esimaors of differen mehods Mehod Correlaion Adjused R 2 Coefficien T-Bill real rae MA(3) model of T-Bill real rae MA(4) model of T-Bill real rae ARIMA(0,1,1) model of pas inflaion ARIMA(1,0,3) model of pas inflaion * ARIMA(1,1,3) model of pas inflaion Previous value of inflaion MA(3) model of pas inflaion MA(4) model of pas inflaion * Significan a he 10 percen level. Afer calculaing he acual inflaion rae, an appropriae mehod should be seleced o esimae he expeced inflaion and unexpeced inflaion. In his paper, we esed nine mehods and he bes one is seleced by using equaion (1). These were he Treasury Bill (T-Bill) real rae, wo moving average represenaions of T-Bill real rae, hree ARIMA model of pas inflaion, he previous value of inflaion and wo moving average represenaions of pas inflaion. The resuls are shown in Table 1. Three represenaions of Treasury Bill (T-Bill) real rae will lead o a negaive correlaion beween expeced and unexpeced inflaion, which means ha hese 2 mehods are no good choices. The adjused R shows ha he explanaory power of wo moving average represenaions of pas inflaion is relaive low, while he ree ARIMA models of pas inflaion are he bes. In erms of significance, only he esimaed coefficien in ARIMA(1,0,3) model of he pas inflaion is significanly differen from 1, which suggess ha his mehod is he bes one. In addiion, we also esed he ARIMA(1,0,1), ARIMA(1,0,2) and ARIMA(1,0,4) model of pas inflaion. The shows ha he ARIMA(1,0,3) model has he lowes AIC value. Therefore, in his paper, he ARIMA(1,0,3) model of pas inflaion is seleced o esimae he expeced inflaion. To ensure he comparabiliy, he same mehod is used o esimae he expeced composiion of CPI, CPI(B) and CPI(C). Table 2 shows he descripive saisics of each series. and (E) have higher reurn rae in average, while here is lile difference in he sandard deviaions of he 5 residences reurn rae, indicaing he similar volailiy. Among he 4 ypes of inflaion rae and he respecive expeced and unexpeced composiions, here is lile difference in he means and sandard deviaions, which means he overall similar movemen of differen inflaion rae. 1 hp:// Consumer Price Indices published by Hong Kong Moneary Auhoriy in Differen series of CPIs are compiled by he Census and Saisics Deparmen (C&SD) o reflec he impac of consumer price changes on households in differen expendiure ranges. The CPI, CPI(B) and CPI(C) are compiled based on he expendiure paerns of households in he relaively low, medium and relaively high expendiure ranges respecively. A Composie CPI is compiled based on he overall expendiure paern of all he above households aken ogeher o reflec he impac of consumer price changes on he household secor as a whole. The CPI, CPI(B) and CPI(C) covers approximae 50%, 30% and 10% of he households, respecively. 3 hp:// 4 A, B, C,D, E represen residence area under 39.9 m 2, m 2, m 2, m 2, and above 160 m 2, respecively. 401

4 Table 2 Descripive saisics Mean Maximum Minimum Sandard Deviaion (B) (C) (D) (E) Composie Composie Composie (B) (B) (B) (C) (C) (C) Table 3 Uni roo es for variables ADF es PP es Series Firs Differences Series Firs Differences (B) (C) (D) (E) Composie Composie Composie (B) (B) (B) (C) (C) (C) Noe: Only he p-values are presened in able. ADF es and PP es are used o es he saionariy of all he series and he resul show ha (B), (C), (D), (E) are saionary using he ADF es, bu non-saionary using he PP es (Table 3). The firs differences of hese series are saionary using boh es, which means he four series are inegraed of order EMPIRICAL RESULT 4.1.Long Run Model Using he 5 groups of residenial real esae and 4 groups of inflaion rae, we go 20 regressions. The resuls are shown in Table 4. The adjused R 2 is quie low for all he regressions wih he larges one For he 4 groups of inflaion rae, he explanaory power is he sronges for residenial, and his explanaory power declines as he area size increases. From anoher aspec, he inflaion rae (C) has he lowes explanaory power for all groups of residenial real esae, wih he lowes adjused R 2 for residence (E). 402

5 Composie (B) (C) Table 4. Resuls for long run model e e (B) e (C) (0.004) (0.949) (0.832) (0.829) e (D) (0.010) (0.937) e (E) (0.017) (0.614) (0.907) (0.706) (0.004) - (0.714) (0.006) (0.006) (0.958) (0.011) (0.004) (0.706) (0.922) (0.840) (0.004) (0.826) (0.008) (0.942) () (0.620) (0.847) (0.944) (0.016) (0.937) (0.029) (0.003) (0.842) (0.047) (0.538) Noe: p-values are given in he parenheses In erms of expeced and unexpeced inflaion rae, we firs analyze he long run hedging abiliy of he residenial real esae o he expeced inflaion. As shown in Table 4, for each group of inflaion rae, he esimaed coefficien of expeced inflaion rae ends o decline from o (E). In erms of he significance of variables, he significance of expeced inflaion rae ends o decline from o (E) as he p-value increases. Lower p-value indicaes a sronger relaionship beween residenial real esae reurns and expeced inflaion, so ha residenial real esae is able o hedge he expeced inflaion o a larger exen. This resul indicaes ha in long run, he hedging abiliy is sronger for smaller residence. For each group of residence, he significance is sronger for and (B), while he significance is wicker for (C), which means ha for he 10% households who are a he high consumpion level, he hedging abiliy of residence agains expeced inflaion is relaive poor in he long run. Taking all he 20 regressions ino consideraion, mos of he expeced inflaion raes are significan a 1% level, and all of hem are significan a 5% level, which indicaes ha, all ypes of residence have hedging abiliy agains expeced inflaion. The esimaion resul for unexpeced inflaion in he long run model is lile differen. In erms of he magniude of he coefficien, here is no significan difference among differen ypes of residence for he each inflaion rae group. However he coefficien of he unexpeced inflaion rae for is much smaller han oher inflaion rae groups. The coefficien for is beween 7 and 8, while he coefficiens for (B) and (C) are around 16. In erms of significance of variables, all he 20 regressions show he same paern, and he unexpeced inflaion raes are significan a 1% level. The larges p- 403

6 value is in he (D)- model. This resul indicaes ha residenial real esae has more sable hedging abiliy agains unexpeced inflaion rae han agains expeced inflaion. In general, he resul of he 20 long run models indicaes ha residenial real esae is a good hedge agains boh expeced and unexpeced inflaion. As menioned in he Mehod secion, we can es he saionariy of he error based on he long run model. And if he error is saionary, he variables should be coinegraed. Afer deriving 20 error series from he 20 regressions, he DF es is used o examine he saionariy of hese error series. The resul of significan level is showed in able 5. I is obvious ha all error series are saionary. Therefore, hese error series can be used in he shor run model as an error correcion erm. Table 5 Saionariy es for residuals in long run model (B) (D) (C) (E) Composie (B) (C) Noe: Only he p-values of DF es are presened in his able. 4.2.Shor Run Model Afer obaining a long run relaionship, we can use he ECM o model he shor run relaionship among he variables. In he shor run model, we assume ha changes of housing reurn are caused by changes in he inflaion and adjusmens o previous disequilibrium in he long run relaionship. In he ECM, dependen variable is he firs difference of housing reurn, independen variables are he firs difference of expeced and unexpeced inflaion, he firs and second lag of dependen variable and he error correcion erm. In addiion, an inercep is added o he model. Table 6 shows he esimaion resuls of he 20 ECMs. Since he second lag of dependen variable is no significan in all models, i is no shown in he able. As shown in Table 6, he ECM has improved he explanaory power compared o he long run model. The adjused R 2 is generally beween 0.5 and 0.7, which indicaes ha he real esae reurn can be beer explained by he shor run model han long run model. However, differen form he long run, he explanaory power of he shor run model does no change obviously in he 20 regressions and here is no significan difference in he adjused R 2 among hese regressions. Firs, we focus on he hedging abiliy of he residenial real esae agains he expeced inflaion in he shor run, which is indicaed by he coefficien of he firs difference of he expeced inflaion rae. For he Composie, he explanaory power of he expeced inflaion rae varies for differen kinds of residence, in which i is significan a 5% level for (C), and a 10% level for and (E). For he households, residence wih area 70 m m2 has beer hedging abiliy agains shor run expeced inflaion, and residence wih area m2 and m2 has no significan hedging abiliy agains shor run expeced inflaion. Composie Table 6 Resuls for shor run model (0.079) (0.105) (0.034) Firs Difference Term Error Correcion Term (0.190) (0.190) (0.218) (0.111) (0.221) (0.071) (0.205) (0.071) (0.039) (0.019) (0.064) (0.072) 404

7 (B) (C) Firs Difference Term Error Correcion Term (0.180) (0.162) (0.188) (0.005) (0.203) (0.185) Firs Difference Term Error Correcion Term (0.080) (0.216) (0.139) (0.229) (0.045) (0.262) (0.130) (0.256) (0.090) (0.241) Firs Difference Term Error Correcion Term (0.182) () (0.233) (0.190) (0.004) (0.240) (0.060) (0.255) (0.070) (0.230) (0.078) (0.226) For, he hedging abiliy of residence agains shor run expeced inflaion increases. The esimaed coefficien of he firs difference of he expeced inflaion for (C) has he highes significance among he five groups of residence and he p value for his coefficien is The coefficien is significan a 5% level for (B) and significan a 10% level for he res groups of residence. This indicaes ha for he households a lower consumpion level, every kind of residence has cerain hedging abiliy agains shor run expeced inflaion. However, aking ino he limied purchasing power of households a lower consumpion level, i is difficul for hem o inves in residence wih large area size o hedge agains inflaion. For (B), he siuaion is similar o he Composie, where he resul is significan a 5% level for (C) and significan a 10% level for and (E). For (C), he esimaed resuls are no significan for and (C) a 10% level, and he p values of he esimaed coefficien are and 0.190, respecively, which means ha he residence wih smaller area is no a suiable hedge agains shor run expeced inflaion for households wih higher income. The resuls are significan a 10% level for (C), (D) and (E), which means ha residence wih larger area is a beer hedge agains shor run expeced inflaion for households wih higher income. The edging abiliy of residenial real esae agains shor run unexpeced inflaion is hen analyzed. From he resuls of he esimaion, we can find ha he firs difference of unexpeced inflaion rae is significan a 5% level in all models. The larges p value is for he - (C) model, while for oher models, he p value is much smaller han This resul indicaes ha all kinds of residenial real esae have he abiliy o hedge agains shor run unexpeced inflaion, which is similar as he long run model. 405

8 In he shor run model, we use he firs and second lag of he firs difference of he residenial reurn as independen variables, ha is, we resric he impac of lagged erm o half a year. In fac, only he firs lag is significan, while he second lag is no significan in all models, which means he impac of lagged erm resrains in one quarer. The error correcion erm is significan in all models and he magniude of he coefficien ranges from in he (B)- (B) model, o in he - model. The significan and negaive error correcion erm indicaes ha he error correcion is a reverse driver, he residence real esae marke will reac and correc o he dispariy beween and he equilibrium level of inflaion and residence real esae marke and his adjusmen is hrough negaive feedback from he error correcion. The large magniudes of he error correcion erms sugges ha he residence real esae reacs o innovaions in expeced and unexpeced inflaion very quickly. 5. CONCLUSION In his paper, we used long run or equilibrium model and shor run error correcion model o sudy he residenial real esae s hedging abiliy agains expeced and unexpeced inflaion in Hong Kong. The findings can be concluded as follows: Firs, in he long run, residenial real esae has cerain abiliy o hedge agains inflaion, and his abiliy does no disappear wih unexpeced change o he inflaion, i has a fairly sable characerisics. In he long run model, all ypes of residenial real esae have cerain abiliy o hedge agains all ypes of expeced inflaion. The residence wih smaller area has sronger abiliy o hedge agains expeced inflaion. In conras o he expeced inflaion, he hedging abiliy of he residenial real esae agains expeced inflaion is more sable. This sabiliy does no significanly change wih he variaion of residence area size and inflaion ypes. In fac, he long run model proves ha Hong Kong residenial real esae is a good hedge agains boh expeced and unexpeced inflaion. Second, in he shor run, he hedging abiliy of residenial real esae agains expeced inflaion differs in he residence area size. Only he residence wih moderae area is obviously able o hedge agains expeced inflaion. Based on he resuls, he residence wih area of m2 has sronger abiliy o hedge agains expeced inflaion, while residence wih area of m2 and m2 has no significan hedging abiliy. The resuls mean ha households a differen income levels should ake differen hedging sraegy when facing wih inflaion. The difference in hedging abiliy of residence wih differen area may be caused by differen price elasiciy. Since he difference in price elasiciy is mainly derived from he difference in demand for residence wih differen area, wih he developmen of economy, he hedging abiliy of differen residence agains inflaion will vary. Third, in he shor run, regardless of he residence area, all ypes of residence have srong abiliy o hedge agains unexpeced inflaion. The negaive coefficien of error correcion erm indicaes he residence real esae marke will adjus hrough a negaive feedback from he error correcion o he dispariy beween and he equilibrium level of inflaion and residence real. The large magniudes of he error correcion erms sugges ha he speed of adjusmen is relaively fas. For households a differen consumpion level, all ypes of residence can hedge agains boh expeced and unexpeced inflaion in shor run and his hedging abiliy keeps sable. Alhough his paper only uses he daa of Hong Kong, bu we believe ha residenial real esae has he abiliy o hedge agains inflaion and such abiliy differs in he residence area size. This implies ha households can inves in residenial real esae o hedge inflaion risk, bu hey should ake he difference in hedging abiliy of differen residence agains inflaion ino consideraion o choose suiable residence. On he oher hand, he governmen s real esae marke regulaion policy should also ake ino accoun his difference o opimize he srucure of regulaion o improve regulaory efficiency. References Anari, A., & Kolari, J. (2002). House prices and inflaion. Real Esae Economics, 30(1), Chen, M. C., & Foo, T. (2006). Common srucural ime series componens in inflaion and residenial propery prices. Journal of Real Esae Porfolio Managemen, 12(1), Di, J. (2012). Can real esae provide a hedge agains inflaion evidence from mainland China?. Chinese Real Esae, 04, [Chinese Journal]. Duan, Z. D. (2007).The relaionship beween house price and inflaion, oupu. The Journal of Quaniaive &Technical Economics, 12, [Chinese Journal]. Fama, E. F., & Schwer, G. W. (1977). Asse reurns and inflaion. Journal of financial economics, 5(2), Ganesan, S., & Chiang, Y. (1998). The inflaion-hedging characerisics of real and financial asses in Hong Kong. Journal of Real Esae Porfolio Managemen, 4(1), Glascock, J. L., Fen, L., Fan, L., & Bao, X. (2010). The inflaion hedging characerisic of real esae: evidence from Hongkong, China. Saisic and Decision, 1, [Chinese Journal]. 406

9 Glascock, J. L., Lu, C., & So, R. W. (2002). REIT reurns and inflaion: perverse or reverse causaliy effecs?. The Journal of Real Esae Finance and Economics, 24(3), Hardin III, W. G., Jiang, X., & Wu, Z. (2012). REIT sock prices wih inflaion hedging and illusion. The Journal of Real Esae Finance and Economics, 45(1), Hoesli, M., Lizieri, C., & MacGregor, B. (2008). The inflaion hedging characerisics of US and UK invesmens: a muli-facor error correcion approach. The Journal of Real Esae Finance and Economics, 36(2), Li, H., & Guan, T. (2011). Research on he correlaion beween inflaion and house price in China. Chinese Real Esae, 22, [Chinese Journal]. Qiu, L. (2011). The abiliy of hedging inflaion in Chinese real esae marke. Times Finance, 33, [Chinese Journal]. 407