BUILDING A HOUSE PRICES FORECASTING MODEL IN HONG KONG

BUILDING A HOUSE PRICES FORECASTING MODEL IN HONG KONG BUILDING A HOUSE PRICES FORECASTING MODEL IN HONG KONG Xin Jane Ge and Associae Professor Ka-Chi Lam Ciy Universiy of Hong Kong INTRODUCTION The purpose of building a house prices forecasing model is o esimae he impac of housing demand and housing supply in Hong Kong. The propery marke plays a very imporan role in he economy of Hong Kong. The real esae secor conribued approximaely 10.2 per cen of GDP in 1996 (Hong Kong Governmen, 1998). More han 45 per cen of all bank loans, over HK$500 billion as a he end of 1997, were direcly ied o properies (Hong Kong Governmen, 1998). Income from land aucions, raes and samp duies accouned for approximaely 24 per cen of oal Governmen revenue in 1997/1998 (Chan, e al., 2001). Propery and consrucion company socks conribued 25 per cen o Hong Kong s sock marke capializaion as well as o over 60 per cen of capial invesmen expendiures (Newell and Chau, 1996). Smooh changes in house prices hus help o mainain sable economic growh in Hong Kong. To achieve a sable house price level, housing supply mus mach he demand for houses. However, house prices have a imes been very volaile as a resul of mismached housing demand and supply in Hong Kong. Figure 1 below shows he behaviour of real house prices and plos he quarerly ime series daa over he las wo decades. I is eviden from Table 1 ha here have been significan booms and buss since he lae 1980s. The real price index of privae residenial propery rose 87 per cen from he hird quarer 1984 o he second quarer of 1989, 71 per cen from he hird quarer 1989 o he hird quarer 1992, and 50 per cen from he fourh quarer of 1995 o he hird quarer of 1997. The reason for such growh during hese periods was ha he demand for houses was growing faser han supply and his generaed speculaive aciviy in he propery marke. The periods of low prices were comparaively less volaile han he boom periods. Two dramaic declines have been observed, i.e., from he second quarer of 1981 o he fourh of 1983, wih a real fall of prices of 47 per cen and a 42 per cen real fall in he hird quarer of 1997. The Asian financial crisis resrained speculaive aciviies dramaically as he suddenly decline in propery prices changed asses ino liabiliies for many households. Consequenly, households reduced heir non-housing consumpion and he lack of confidence in he economy as a whole creaed a vicious circle, furher lowering he value of propery. The house prices have dropped a furher 18 per cen since hen. Table 1: Housing Price booms and Buss (Source: Raing and Valuaion Deparmen, Hong Kong Governmen) Booms 4.79 2.81 3.84 2.89 3.89 3.92 1.93 2.94 4.95 3.97 Buss 2.81 4.83 2.94 4.95 3.97 4.98 2.99 4.00 Time Period % Change (Nominal) % Change (Real) 50.0% 21.9% 144.0% 86.6% 128.6% 71.3% 38.0% 25.7% 65.9% 50.0% -31.3% -12.4% -41.6% -24.8% -47.1% -22.3% -41.5% -18.4% THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2 57

XIN JANET GE AND KA-CHI LAM Figure 1: Real Residenial Housing Prices Index in Hong Kong (Source: Raing and Valuaion Deparmen, Hong Kong Governmen) 5 4 3 Real HP 2 1 0 year 1980.1 1981.4 1983.3 1985.2 1987.1 1988.4 1990.3 1992.2 1994.1 1995.4 1997.3 1999.2 2001.1 2002.4 I is an imporan role for he Hong Kong governmen o forecas he housing marke and o provide a maching supply of land o he marke. A house price forecasing model is one way his may be done. The objecive of his sudy is o develop a house prices forecasing model for Hong Kong. I sars from he assumpion ha housing in Hong Kong is raded in an efficien, free marke. The firs sep is o idenify, hrough a lieraure review, he variables ha conribue o changes in he demand for and supply of houses. The second is o use a muliple regression for he empirical esimaion. Quarerly ime series daa from 1980 o 2001 are used for he analysis. Some variables are ransformed ino logarihms and/or by use of moving averages o remove irregulariies and/or seasonal paerns before applicaion of he reduced form of he house prices model. The hird sep is o es he model by examining he significance of saisical indicaors. Three ypes of variables, namely macroeconomic indicaors, housing relaed variables and demographic variables are used in he analysis. From hese variables, eigh models are derived for he analysis. These models indicae ha household income, he size of he populaion, land supply, he Hang Seng Index and uni ransacion volumes are he major indicaors of changes in house prices. LITERATURE REVIEW I is widely acceped ha house prices are deermined by he demand for, and supply of, houses in he propery marke. Thus, prices will adjus o ensure ha he marke clears in he long run (Nellis and Longboom, 1981). Bajic (1983) suggess ha he marke is no generally in shor run equilibrium and ha changes in housing prices are frequen and rapid, while DiPasquale and Wheaon (1994) sugges ha here is, insead, a coninuous adjusmen as acual prices converge owards equilibrium prices. The effecive housing demand is he amoun of housing for which he populaion is willing and able o pay. Individuals view housing no merely as a consumpion good, bu also, simulaneously, as an invesmen (Dusansky and Wilson, 1993). Reicher (1990) suggess ha naional economic facors such as morgage raes, and local facors such as populaion shifs, employmen and income rends have a unique impac on housing demand, and hus on housing price. The demand for housing also depends on facors like cos of morgage finance, real incomes and he general level of consumer confidence (www.uor2u.ne, 2002). Muh (1960) concluded ha housing demand is highly responsive o changes in income and prices. The empirical resuls indicae ha he mos imporan facor in he deerminaion of house prices is real income because rising income increases he absolue value of he marginal rae of subsiuion beween goods and owner-occupaion. If household behaviour is consisen, hen he appropriae income measure should be long run permanen income (Megbolugbe, e al., 1999). Demographic variables such as family size and age composiion are major deerminans 58 THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2

BUILDING A HOUSE PRICES FORECASTING MODEL IN HONG KONG of household consumpion paerns (Pollark and Wales, 1981). Mankiw and Weil (1989) have found ha aggregae demand growh would slow markedly as he baby boom generaion grows older and reduces is consumpion of housing. However, he conclusion has been criicised by mos scholarly commenaors (DiPasquale and Wheaon, 1994; Woodward, 1991; Engelhard and Poerba, 1991). I is commonly acceped ha an increase in household formaion due o an ageing populaion, increasing numbers choosing o remain single and rising raes of divorce and separaion will lead o increased demand for housing. Wong (1993) aribues he fas growh in demand during he boom in 1991 o a number of demographic as well as economic facors. In his sudy, he populaion age group 20 o 59 is used as represening demographic and permanen income facors because hese people generally have sufficien savings and income o finance he purchase of houses. Hong Kong began o experience a surge of populaion in his home purchasing age group afer 1986 (Wong, 1993) (Figure 2). Expecaions abou he fuure direcion of he economy affec curren demand. Consumers are more likely o buy houses when hey expec an expanding economy o provide hem wih boh job securiy and rising income in he fuure. Governmen policy, inflaion, ineres rae changes and rae of reurn on propery all have a grea impac on consumer confidence and he demand for housing (Wheeler and Chowdhury 1993). The sock marke is anoher indicaor of economic performance. An empirical sudy by Fu e al. (1993) found a paern in which he sock marke leads he propery marke in price change. Thus he Hang Seng Index is used as a proxy for macroeconomic impac in his sudy. Housing price appreciaion simulaes invesmen demand for houses. House price rises may lead o speculaion, and speculaion has been considered as a possible deerminan of house price by number of auhors such as Case and Shiller, (1989, 1990) and Levin and Wrigh (1997). Speculaion opporuniies arise from he gaps beween he iming of purchase and sale conracs, circumsances where he expeced growh rae in house prices exceeds he ineres rae charged on bridging loans, and he opporuniy o rade up wihou incurring any incremenal ransacions coss exiss (Levin and Wrigh, 1997). Therefore, he ransacion volume of residenial properies in erms of he number of sales and purchase agreemens in Hong Kong is adoped for his sudy. Mos homeowners use morgages o finance heir home purchase (Chan, 1996). To qualify for a morgage, borrowers usually inves equiy in a down paymen (Harris and Ragonei, 1998). The availabiliy of morgage credis and firs down paymen have been criical for housing invesmen demand. Figure 2: Populaion age composiion in Hong Kong, 1980 2000 people POPULATION STRUCTURE IN HONG 5000000 4500000 4000000 3500000 3000000 2500000 2000000 1500000 1000000 500000 0 0-19 20-60 age groups > 60 1980 1986 1991 1996 2000 THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2 59

XIN JANET GE AND KA-CHI LAM Expecaions abou he fuure direcion of he economy affec curren demand. Consumers are more likely o buy houses when hey expec an expanding economy o provide hem wih boh job securiy and rising income in he fuure. Governmen policy, inflaion, ineres rae changes and rae of reurn on propery all have a grea impac on consumer confidence and he demand for housing (Wheeler and Chowdhury 1993). The sock marke is anoher indicaor of economic performance. An empirical sudy by Fu e al. (1993) found a paern in which he sock marke leads he propery marke in price change. Thus he Hang Seng Index is used as a proxy for macroeconomic impac in his sudy. Housing price appreciaion simulaes invesmen demand for houses. House price rises may lead o speculaion, and speculaion has been considered as a possible deerminan of house price by number of auhors such as Case and Shiller, (1989, 1990) and Levin and Wrigh (1997). Speculaion opporuniies arise from he gaps beween he iming of purchase and sale conracs, circumsances where he expeced growh rae in house prices exceeds he ineres rae charged on bridging loans, and he opporuniy o rade up wihou incurring any incremenal ransacions coss exiss (Levin and Wrigh, 1997). Therefore, he ransacion volume of residenial properies in erms of he number of sales and purchase agreemens in Hong Kong is adoped for his sudy. Mos homeowners use morgages o finance heir home purchase (Chan, 1996). To qualify for a morgage, borrowers usually inves equiy in a down paymen (Harris and Ragonei, 1998). The availabiliy of morgage credis and firs down paymen have been criical for housing invesmen demand. The effecive demand for privae housing is volaile, while he supply side is deermined no only by he producion decisions of builders of new dwellings bu also by he decisions made by owners of housing concerning conversion of he exising sock, as housing is a durable good (DiPasquale, 1999). Supply side facors including vacancies, housing sars and ineres rae all play a role in housing price movemens (Ley and Tuchener, 2001). Land supply has been addressed by Peng and Wheaon (1994) and Ho and Ganesan (1998) in he Hong Kong housing marke. They claim ha he quaniy of land supply deermines housing prices. In Hong Kong, land is a highly scarce naural resource. Governmen land policy may impose a conrived effec on he supply of land. The Sino- Briish Join Declaraion sipulaed ha 50 hecares of land was he maximum ha could be sold by he Hong Kong Governmen in a single year during he ransiion period (May 27, 1985 June 30, 1997). Land leased by he Hong Kong Housing Auhoriy for he consrucion of public renal housing was exemped from he land sales limi. In his sudy, residenial unis wih consen o commence work, i.e., housing sars in erms of gross floor area, are aken as a proxy for land supply as suggesed by Ho and Ganesan (1998). DEVELOPMENT OF A REDUCED FORM HOUSE PRICES FUNCTION A reduced form equaion for he price funcion is derived based on he supply and demand funcions for owner-occupied housing and hen invered under an equilibrium assumpion (DiPasquale and Wheaon, 1994). Table 2 shows ha reduced form equaions have been employed by many researchers in differen applicaions. An example is Reicher (1990) who has used a reduced form equaion o derive a regional housing prices model. He found ha morgage raes, populaion shifs, employmen and income rends ofen have a unique impac on housing prices. Many models of house price changes concenrae on demand facors (Muellbauer and Muphy, 1992) as supply facors are more difficul o measure. Some sudies have uilized naional aggregae ime series daa (Nellis and Longboom, 1981; Buckley and Ermisch, 1983; Mankiw and Weil, 1989). Ohers have made use of pooled ime series cross-secional daa (Case, 1986; Mancheser, 1987; Reicher, 1990; Abraham and Hendersho, 1996). To develop a house price reduced form model, he firs sep is o derive a demand equaion. In accordance wih lieraure review, he quaniy demand for houses can be denoed as follows: Qd = f(g, H, D, ) ( = 1, 2, 3, n) (1) G = g(x1, xi, xm, ) (i = 1, 2, 3, m) (2) 60 THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2

BUILDING A HOUSE PRICES FORECASTING MODEL IN HONG KONG Table 2: Applicaion of reduced form house prices model Auhor(s) Year Tile Key Findings Muh 1960 The Demand for Non-Farm Housing Aggregae daa, concluded ha here is a perfecly elasic supply curve. Follain 1979 The Price Elasiciy of Long Run Supply of New Housing Consrucion Aggregae annual daa, he assumpion of a perfecly elasic long run supply curve canno be rejeced. Nellis and Longboom Ozanne and Thibodeau Forura. and Kushner 1981 An Empirical Analysis of he Deerminaion of Housing Prices in he Unied Kingdom 1983 Explaining Meropolian Housing Price Differences 1986 Canadian Iner-Ciy House Price Differenials Mancheser 1987 Inflaion and Housing Demand: A New Perspecive Manning 1989 Explaining Inerciy Home Price Differences Reicher 1990 The Impac of Ineres Raes, Income, and Employmen upon Regional Housing Prices Muellbauer and Murphy Follain, Leavens and Velz Case and Mayer 1992 Booms and Buss in UK Housing Marke 1993 Idenifying he Effecs of Tax Reform on Mulifamily Renal Housing 1996 Housing Price Dynamics Wihin a Meropolian Area Malpezzi 1996 Housing Price Exernaliies, and Regulaion in U.S. Meropolian Areas Aggregae daa, house prices is relaively more responsive o demand facor. The change in he price of houses lagged one period and nominal morgage sock were found imporan in shor run. Ren and house price indexes used o measure he variaion among 54 meropolian areas and able o explain 88% of he variaion in renal prices and 58% of he variaion in house prices. Idenify he sources of iner-ciy house price differenials in Canada. Demand facors are imporan explanaory variables; a 1% increase in he income of households raises house prices by 1.11%. Naionwide ime-series daa cross ciies. The ineracion beween axes and inflaion as well as cash-flow consrains has srong effecs on he relaive price of houses. Aggregaed daa, explanaion for 84% of inerciy variaion in owner-occupied housing prices. Naionwide daa, various regions respond in a similar fashion o cerain naional facors and sugges moneary and ax policy should ake ino consideraion boh naional facors and regional rends. Housing demand is examined aking ino accoun expecaions, credi consrains, lumpy ransacions coss and uncerainy. Cross-secion ime series daa. Examine he empirical relaionship beween ren and user cos. Changes in user cos significanly affec consrucion, bu no he level of rens. Boson house price paern. Changes in he cross-secional paern of house prices are relaed o differences in manufacuring employmen, demographics, new consrucion, ec. Cross-secion analysis. Log ransformed daa, Increasing local marke regulaions of land increase home prices hrough increasing rens. THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2 61

XIN JANET GE AND KA-CHI LAM H = h(y1, yi, ym, ) (3) D = d(z1, zi, zm, ) (4) Therefore, Qd = f( xi, yi,, zi,, ) (5) Where Q d = aggregaed quaniy demand for new houses during period, G = macroeconomic variables, H = housing relaed variables, D = demographic variables, x i = macroeconomic variables such as GDP, ineres rae, Hang Seng Index, ec., y i = housing relaed variables such as house prices, permanen income, unemploymen rae, ec., z i = demographic variables such as populaion, number of marriages, birh raes, ec. I is assumed ha homeowners maximize uiliy and invesors maximize heir profis (Reicher, 1990). The mehod is applied for he supply equaion as he second sep. The supply of housing is a funcion of house prices, consrucion coss including ineres raes, maerial coss and labour coss, and land supply. Qs = f(s, ) (= 1, 2, 3, n) (6) S = s(v1, vi, vm, ) (i = 1,2,3, m) (7) Qs = f(vi,) (8) Where Q s = aggregaed quaniy of new supply during period, S = Supply variables, v i = variables such as house prices, consrucion coss and land supply. Under an assumpion of supply-demand equilibrium wihin he given period, i.e., Q d = Q s, he funcions (5) and (8) give a reducedform price funcion: P = f(qd, Qs, ) ( = 1,2,3, n) (9) P = f(xi, yi, zi, vi, ) (i =1,2,3, m) (10) Where P = house prices of new unis sold during period as dependen variable. x i, y i, z i, v i are he independen variables. Assuming he generalized consanelasiciy demand funcion wih a muliplicaive relaionship according o Reicher, (1990) gives: β 1 β2 β3. β4 P = β 0xi. yi. zi vi (11) The funcional form in (11) can be convered ino a linear equaion suiable for esimaion by sandard muliple regression echniques by expressing i in logarihmic form. A one period lagged auoregressive error erm P -1 is applied o he model. Thus, he muliple populaion regression equaion for houses demand becomes: ln P = β 0 β1 ln xi β 2 ln yi β 3 ln zi β 4 ln vi β 5 ln P 1 ε (12) Where β 0 β 5 represens he inercep and he regression coefficiens (or elasiciies) associaed wih heir respecive explanaory variables, ln = he naural log of he coninuous variables, ε = he populaion disurbance erm for quarer. Where ε ~ WN (0, σ 2 ). Daa Preprocessing and Esimaing Procedures Secondary daa sources were uilised in he sudy. Unless specify, quarerly ime-series economic indicaors were absraced from he Hong Kong Monhly Diges of Saisics complied by he Census and Saisics Deparmen in Hong Kong over he las wo decades. The inerpreaion of each variable employed is as follows: Table 3: The percenage disribuion of households by expendiure Index Approximae percen of households covered Monhly expendiure range in 1984/85 CPI (A) 50 $2,000 $6,499 CPI(B) 30 $6,500 $9,999 Hang Seng CPI 10 $10,000 $24,999 Source: Hong Kong Monhly Diges of Saisics 62 THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2

BUILDING A HOUSE PRICES FORECASTING MODEL IN HONG KONG General Economic Indicaors: GDP represens Gross Domesic Produc a ime which is a measure of he oal value of producs of all residen producing unis of a erriory in a specified period, before deducing allowance for consumpion of fixed capial. GDP is mos widely used measure of economic performance. The growh in he GDP also underlines he viabiliy of he housing marke and lends suppor o he rising aspiraion of home ownership. I is used as an independen variable ogeher wih Hang Seng Consumer Price Index (HCPI ) which is used o produce consan (2000 = 100) prices. GDPC (deflaed by he HCPI) represens he gross value of invesmen expendiure in land, building and consrucion, plan, machinery and equipmen by he public and privae secors in consan erms. There are hree Consumer Price Index series derived from he Household Expendiure Survey, defined in erms of he percenage disribuion of households by expendiure as shown in Table 3. The remaining 10 percen of households a he op and boom of he expendiure scale are excluded. The Hang Seng Consumer Price Index (HCPI ) is used in his sudy because i represens he expendiure group mos likely o affec privae housing prices. Hang Seng Index (HSI ) is compiled by he Hang Seng Bank Ld based on informaion on share prices supplied by The Sock Exchange of Hong Kong. HSI covers 33 blue chip socks lised on he Exchange and is weighed by marke capializaion. The las daa for each quarer are used for his sudy. Median Monhly Domesic Household Income (HHI ) is he median household income which represens purchasing power in he period. Real household incomes are consruced by dividing he household incomes by he Hang Seng Consumer Price Index (HCPI ). The ineres rae r is he bes lending rae a he period, expressed as per cen per annum. The Hong Kong dollar is linked o he U.S. dollar hence he local ineres rae is beyond he governmen s conrol. House prices will increase when morgage and ineres raes decline and he propery marke will slow down when morgage raes rise. Therefore i is expeced o have a negaive sign. Real ineres rae is he nominal rae (i ) minus inflaion rae (i f ). Tha is: r = i i i f f CPI CPI = CPI 4 4 (13) (14) The real morgage rae (rm) is derived by dividing nominal morgage rae (im) by he HCPI, i.e.: im r m = HCPI (15) Demographic Facors The demographic variables such as oal populaion (GPL ), people a age group of 20 59 (PL ), marriages (MN ) and number of birhs (BN ) a he period respecively are considered. Increasing demographic facors will increase he pressure on house prices. Only mid-year and end-year populaion figures are available. Quarerly figures are calculaed as follows: GPL GPL GPL 2 1 1 = GPL 1 (16) The figures relaing o birhs, deahs and marriages refer o such evens as were regisered wih he Direcor of Immigraion every quarer. Seasonal adjusmen is made o eliminae seasonal effec. Housing Relaed Facors Saisics on price and renal cos indices for privae domesic premises are provided by he Raing and Valuaion Deparmen, Hong Kong. There are four ypes of privae domesic premises ha are lised in Table 4. The overall price indices are used for he sudy. Real housing prices (HP ) are derived by dividing nominal prices by he Hang Seng Consumer price index (HCPI ). A sudden scarciy of land raises housing prices because of suppressed curren housing producion and higher invesmen demand (Peng and Wheaon, 1994). Thus, he availabiliy of land is an imporan facor o be considered in he model. Consen o commence work on residenial flas is used as a proxy of land supply (LS ). The measuremen of land supply is defined as he oal gross floor area of land supply acually pu o he marke (Ho and Ganesan, 1998). THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2 63

XIN JANET GE AND KA-CHI LAM Table 4: Propery price indices for privae domesic premises Privae Domesic Premises (square meer) (1989=100) Year Up o 39.9 40 69.9 70 99.9 100 & above Overall 1992 210 219 229 205 215 1993 223 244 261 250 237 1994 263 306 341 351 293 1995 252 282 306 314 272 1996 269 310 334 352 298 1997 376 435 488 514 420 1998 274 308 336 348 299 1999 231 265 287 302 257 Sources: Hong Kong Monhly Diges of Saisics, various issues Number of houses compleed (HS ) is he major measure of housing supply. I is relaively inelasic in he shor run; his is because here are ime lags beween a change in price and an increase in he supply of new properies becoming available, or homeowners deciding o pu heir properies ono he marke. The long run impac on prices depends on he supply response deermined by he price elasiciy of supply (DiPasquale, 1999). The consrucion cos index is sourced from Leve and Bailey Charered Quaniy Surveyors Ld in Hong Kong. Their ender price index is a quarerly weighed index ha measures he coss of building maerial, labour coss, plan coss, rens, overhead coss and axes. Oher Indicaors Poliical evens (PO ) conribue o house price flucuaions such as occurred when housing prices kep falling from 1981 ill he end of 1984 because of uncerainy over Hong Kong s poliical fuure afer he Sino- Briish negoiaions over Hong Kong in 1997, or he propery boom afer he Sino-Briish Join Declaraion in 1985. The Tiananmen Square evens caused an immediae bu brief fall in propery prices. Hence i is eviden ha Hong Kong s housing marke is highly responsive o changes in he poliical climae (Chou and Shih, 1995). In he analysis, 1 indicaes he occurrence of an even and 0 indicaes oherwise. Confidence is vial in he housing secor. Wha people hink will happen in he fuure influences curren purchasing decisions. I is no an inpu variable for he house prices equaion because confidence is hard o quanify. The deails variables descripion is in he Table 5. Dae Transformaion To make hem more suiable for quaniaive analysis, he daa are examined and ransformed or manipulaed as required. All daa are examined o esablish (a) wheher he daa for individual variables are normally disribued; and (b) wheher he independen variables are linearly relaed o he dependen variable. Daa are ransformed hough he following mehods: 1) Log ransformaion of variables o make relaionships linear, such as house price index and renal index; 2) Moving average o minimize he effec of seasonal and irregular variaions, hereby indicaing he daa s general rend. For example, new house compleions where a 4- year moving average eliminaes, o a grea exen, he flucuaions in he original daa. The equaion is: 1 1 p = x 1 4 i= 0 y (= p, n) (17) Where here are n values in a ime series x 1 x n. The cenred moving average is applied o mach ime series (Waxman, 1993). The moving average will be disored by any unusual evens occurring during he ime under consideraion, hus any unusual curve can be deeced if here is policy change or some special even occurred in he period. 64 THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2

BUILDING A HOUSE PRICES FORECASTING MODEL IN HONG KONG Table 5: Definiion of variables Type Name Definiion y i PO Poliical evens. 1 if occurs, 0 oherwise. HP The Privae Housing Price Index (1989=100), inflaion adjused. R The Privae Renal Index (1989=100), inflaion adjused. U Unemploymen Raes (percen) z i GPL Toal Populaion Number HN Household Number PL Populaion Number age a 20 59 who are he sources of income group. MN Number of Marriages BN Number of Birhs x i Y The Median of Household Income, Hong Kong Dollars per household HSI Hang Seng Index (1964=100), inflaion adjused. GDP Gross Domesic Produc a consan 2001=100, Hong Kong Dollars Million GDPC Gross Domesic Produc consrucion. Hong Kong Dollars Million HCPI Consumer Price Index (10.1999 09.2000=100) r The Morgage Raes percen per annum from Hang Seng Bank, inflaion adjused. v i LS Land supply for privae residenial developmen. Residenial unis/flas wih consen o commence work by floor area (square meer) as proxy. HS Residenial Unis Compleed by Privae Number of Unis C Consrucion Cos Index (1968=100), inflaion adjused. 3) Differencing echnique, i.e., by subracing a lagged version of he series from he original ime series daa. A new ime series is creaed from he firs difference (or he difference of order 1) such as changes in housing compleion. z = y y-1 (18) Similarly, a difference of order 4 can be derived by: z = y y-4 (19) The inflaion rae is deermined using his echnique. 4) Principle componen analysis is adoped for selecing bes effecive variables. I is applied o avoid he use of variables wih srong posiive relaionships as such relaionships may reduce he validiy of he model. Many indicaors, especially macro economic variables, are srongly correlaed. When his is he case, he calculaed coefficiens may no represen a rue causal relaionship beween dependen and independen variables. This ype of analysis can separae he naure of variables by making caegories and give exracion sums of squared loadings for considering variable selecion. 5) There are leading, coinciden and lagging characerisics of indicaors. To ensure ha he economic indicaors ruly reflec he growh or decline of housing prices, he degree of ime lag or lead should be esablished. Pearson correlaions are produced o es if here are significan correlaions beween he dependen and independen variables, and o find lead/lagged relaionships beween variables. The correlaion value is considered significan if he p value is less han 0.05. Esimaion Procedures The forecasing procedures are depiced in Figure 3. A reduced-form model is a specific forecasing sysem used for he sochasic simulaion for housing prices based on economic indicaors. THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2 65

XIN JANET GE AND KA-CHI LAM Figure 3: Depics procedures for he housing prices forecas using log-linear model. Procedures for Daa Preprocess Daa Transformaion Procedures for Forecasing Qualiaive Daa Quaniaive Daa Defined Dependen & Independen Variables Selec & Apply Forecasing Model Daa Collecion Daa Transformaion Variables Selecing Sepwise Variable Selecion Organizaion of Colleced Daa Correlaion Analysis Esimaion Parameers No Yes Yes No Daa Analysis Principle Componen Analysis Analysis Model Fi Inerpreaion Resuls Daa for Forecasing Forecasing & Inerpreaion Verificaion Sage One Sage Two Sage Three Figure 4: Periods of forecasing T1 Hisorical daa T3 T2 Time Backcasing Esimaion period Ex-pos forecas period Today Source: Pindyck and Rubinfeld, 1991 Ex-ane Forecas period To es he accuracy of he forecasing models, 80 per cen of he hisorical daa are used for esimaion and 20 per cen of he daa are adoped for ex-pos forecasing. Ex-ane forecas will be applied for analysis of policy implicaion as showed in Figure 4. Sepwise selecions are used. The decision o ener or remove variables in he model is based on how much hey conribue o muliple R 2 and on F and values. The mean squared error (MSE) and adjused R 2 2 ( R ) are employed as crieria in he examinaion of he model fi. MSE is used because i effecively esimaes ou-of-sample mean square predicion error he smaller he beer (Diebold, 1998). N 2 e = MSE = 1 N where N is he sample size and = HP HP e R = 1 (20) 1 2 s (21) 2 ( HP HP) /( N 1) 2 = T where s 2 is he variance. R 2 is he coefficien of deerminaion which expresses he 66 THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2

BUILDING A HOUSE PRICES FORECASTING MODEL IN HONG KONG proporion of he oal variaion in he dependen variable ha is explained. Adjused R squared is an esimae of how well he model would fi from he same populaion. Thus, minimizing he sandard error of he regression maximizes adjused R squared o esablish he model wih he bes fi. Empirical Resuls The ordinary leas-squares regression mehod is employed in his analysis. The advanage of he leas-squares mehod is ha i expresses he secular rend in a mahemaical formula which permis objecive exrapolaion ino he pas, presen and fuure. The disadvanage is ha i is based on he assumpion ha all variables have linear relaionships which is no always he case. Table 6 shows hree ses and a oal of eigh models, chosen from many derived models. The dependen variable is he real privae residenial house price index (1989=100). The models are significan a he 95 per cen confidence level. The independen variables are differen in each case for he purposes of comparisons. Case one uses he populaion of age 20 59, land supply, morgage rae, Hang Seng Index and unis of ransacions as independen variables. I is found ha boh he F es and he es for each variable, excep he morgage rae, are saisically significan and have he expeced signs. A one per cen increase in populaion during he given period is associaed wih a 2.67 per cen increase in housing prices during he same period. A one per cen decrease in land supply during he given period is associaed wih 0.104 per cen increase in housing prices during he same period. A one per cen increase in he Hang Seng Index and uni ransacions volume during he given period are associaed wih 0.36 and 0.258 per cen increase in housing prices during he same period respecively, ceeris paribus. The implicaions are (1) he higher he populaion and he higher he permanen income, he higher he housing prices; (2) here are speculaive aciviies in he housing marke; (3) macroeconomic facors impac on housing prices. The problem in his model is ha morgage rae does no have he expec sign. The Durbin-Wason es rejecs he null (ρ=0 ) hypohesis. Case wo uses he same variables as case one bu increases he ime span (sample sizes) from 63 o 75 quarers, which improves he Durbin-Wason es. To furher improve he Durbin-Wason resul, he real housing price, lagged one period, is applied. The Durbin-Wason reaches 1.68 which is in he inconclusive range. However, he negaive sign of populaion variable for models 3 8 has indicaed ha here may be problems wih mulicollineariy in he model. In cases four o eigh, differen variables are esed in he model. I is found ha household income is significan, i.e., a one per cen increase in household income a a given period is associaed wih a 1.97 per cen increase in housing prices for he same period. An ineresing finding is ha poliical evens have posiive impacs on housing prices. However, here are negaive signs on he oal size of populaion. The demographic facors indicae grea changes in Hong Kong during he pas wo decades. I is implied ha he size of he oal populaion may no be he bes proxy in building house prices model. The reasons are as follows: (a) Populaion Ageing The proporion of he populaion aged 65 and over has grown progressively (Champion, 2001) in Hong Kong, from 8 per cen o 11 per cen from 1988 o 1999 (Deparmen of Census and Saisics, April 2001). This age group is mosly saying a governmen public housing or living wih heir children, raher han adding o he demand for housing. (b) Children of School Age This age group is decreasing in relaive erms. The proporion of people aged less han 19 fell 25 per cen from 1980 o 2000 and he birh rae is declining, from 1.2 per cen in 1989 o 0.8 per cen in 1999 in Hong Kong (Deparmen of Census and Saisics, April, 2001). (c) Changes in Marriages Raes Marriage raes have decreased coninuously over he pas en years. The median age a firs marriage for men increased from 28.6 in 1988 o 29.8 years in 1998, and for women from 25.8 o 26.9 years, indicaing a rend owards laer marriages (Census and Saisics Deparmen, 1999). THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2 67

XIN JANET GE AND KA-CHI LAM Table 6: Regression Resuls wih Dependen Variable = LOGHP Variables Consan LOGPL LOGGPL LOGLS LOGR LRHP -1 LOGHHI LOGLS 2 LOGHSMA POLICY LOGHHI 2 LOGHSMA (1) Expeced 82 2001 80 97 80 2001 Sign Case 1 Case 2 Case 3 Case 4 Case 5 Case 6 Case 7 Case 8-18.55-10.92 2.04-3.269 (-4.2) (-3.3) (2.85) (-7.99) 2.674 1.46 -.482 (3.85) (2.78) (-4.3) -4.566 (-5.01) - -0.104 (-2.2) -0.155 (-3.7) - 0.875 (4.2) 0.871 (3.92) 0.15 (2.96) 0.361 (4.40) 0.835 (36.9) 0.253 (3.44) 1.98 (12.34) 0.0886 (5.16) - -0.201 (-2.69) /- 0.0081 (5.46) - -2.911 (-6.02) -4.55 (-4.17) 0.322 (3.29) 1.954 (10.16) 0.105 (5.16) -0.275 (-3.12) -3.423 (-11.4) -4.634 (-6.93) 0.375 (6.23) 0.102 (8.21) -0.162 (-2.95) -3.504 (-13.2) -3.829 (-10.3) 0.399 (7.50) 1.852 (23.2) 0.10 (8.83) -0.131 (-2.87) -3.437 (-8.4) -3.339 (-6.08) 0.403 (4.9) 1.799 (15.03) 0.0877 (4.84) 0.009 0.0087 0.0068 (10.53) (11.96) (4.94) 1.972 (16.74) -0.153 (-2.17) LHSI 0.36 0.292 0.074 (3.15) (2.78) (3.4) LSP 0.258 0.419 0.121 (3.39) (6.77) (8.81) Adjused 0854 0.81 0992 0.962 0.946 0.980 0.981 0.955 R 2 0.566 0.757 1.678 1.413 0.922 1.465 1.451 1.648 D-W Raio 74.99 65.1 1606 302.03 248.08 569.33 732.89 298.26 F-Raio 63 75 75 71 71 71 84 84 Sample size (d) Migraion Mainland China is he major source of immigrans. In 1999, 54,625 mainland residens came o sele in he Hong Kong under he one-way permi scheme (Census and Saisics Deparmen, 2001). The populaion will coninue o increase, however, new immigrans may no have much purchasing power. Though he models have indicaed significance in erms of adjused R 2 and ess, he unexpeced sign in some of he models implies ha here may be a problem of mulicollineariy in he models. Mulicollineariy is he correlaion among he independen variables. I can disor he sandard error of esimae and may, herefore, lead o incorrec conclusions as o which independen variables are saisically significan. I is 68 THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2

BUILDING A HOUSE PRICES FORECASTING MODEL IN HONG KONG suggesed ha he ordinary leas square mehod may no be suiable for a housing prices model in he Hong Kong siuaion. CONCLUSION This sudy has aemped o consruc house price forecas models for Hong Kong. All seleced variables were ransformed ino logarihms before applying he muliple loglinear funcional forms (Anas and Eum, 1984; Harringon, 1989) of regression analysis. In he sudy i is found ha macroeconomic elemens such as he Hang Seng Index and household income have impacs on housing prices. Demographic variables, such as populaion of age 20 59, are also significan, while housing relaed facors such as land supply, and uni ransacion numbers and compleion of new houses are he main variables ha influence house prices. Over he sudy period, policy facors have also been imporan for he fi of he models. The implicaion is ha he Hong Kong Governmen may formulae a sable and suiable long erm housing policy. Land policy may affec invesmen demand for housing and herefore influence house prices. Of he eigh models, five (3, 4, 6, 7 and 8) are saisically saisfacory. However, i is indicaed ha here may be a problem of mulicollineariy buil ino he models which suggess ha hey can be improved hrough co-inegraion, error correcion models. An alernaive mehod ha could be considered is an arificial neural nework model. REFERENCES Abraham, J. M. and Hendersho, P. H. (1996) Bubbles in Meropolian Housing Markes. Journal of Housing Research, 7, 191 207. Bajic, V. (1983) Urban Housing Markes Modelling: Shor-run Equilibrium Implicaions, Areuea Journal, 11, (3). Ball, M. J., (1993) Recen Empirical Work on he Deerminans of Relaive House Prices, Urban Sudies, 10, 213 233 Dusansky, R. and Wilson, P. W. (1993) The Demand for Housing: Theoreical Consideraions, Journal of Economic Theory, 61, 120 138. Engelhard, G. V. and Poerba, J. M. (1991) House Prices and Demographic Change (Canadian Evidence), Regional Science and Urban Economics, 21, 139 546, Norh- Holland. Buckley, R and Ermisch, J. (1983) Theory and Empiricism in he Economeric Modelling of House Prices, Urban Sudies, 20, 83 90. Case, K. E. (1986) The Marke for singlefamily Homes in he Boson Area. New England Economic Review. May/June, 38 48. Case, K. E., and Shiller, R. (1989) The Efficiency of he Marke for Single Family Homes. American Economic Review 79 (1), 125 37. Case, K. E. and Shiller, R. J. (1990) Forecasing Prices and Excess Reurns in he Housing Marke, Areuea Journal, 18, (3). Case, K. E. and Mayer, C. J. (1996) Housing Price Dynamics wihin a Meropolian Area, Regional Science and Urban Economics, 26, 387 407. Census and Saisics Deparmen in Hong Kong, Hong Kong Monhly Diges of Saisics, various issues. Census and Saisics Deparmen in Hong Kong, Hong Kong Propery Review, various issues. Chan, S. (1996) Residenial Mobiliy and Morgages, Regional Science and Urban Economics, 16, 287 311. Chan, Hing Lin; Lee, Shu Kam and Woo, Kai Yin (2001) Deecing raional bubbles in he residenial housing markes of Hong Kong. Economic Modelling, 18, 61 73. Chou, W.L. and Shih, Y.C. (1995) Hong Kong Housing Markes: Overview, Tenure choice, and Housing Demand, Journal of Real Esae Finance and Economics, 10, 7 21. DiPasquale, D. and Wheaon, W. C. (1994) Housing Marke Dynamics and he Fuure of Housing Prices, Journal of Urban Economics, 35, 1 27. DiPasquale, D., (1999) Why don we know more abou Housing Supply? Journal of Real Esae Finance and Economics, 18 (1), 9 23. Follain, J. R., (1979) The Price Elasiciy of Long Run Supply of New Housing Consrucion, Land Economics, 55, 190 191. Follain, J. R., Leavens, D. and Velz, O. T. (1993) Idenifying he Effecs of Tax Reform on Mulifamily Renal Housing, Journal of Urban Economics, 34, 275 298. THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2 69

XIN JANET GE AND KA-CHI LAM Forura, P. and Kushner, J. (1986) Canadian Iner-Ciy House Price Differenials, AREUEA Journal, 14 (4), 525 536. Fu, Yuming, Leung, W.K. and Lo, Wai Chung (1993) The Dynamics of Residenial Propery Markes and he Sock Marke in Hong Kong. Asia-Pacific Financial and Forecasing Research Cener. Harris, R. and Ragonei, D. (1998) Where Credi is Due: Residenial Morgage Finance in Canada, 1901 o 1954, Journal of Real Esae Finance and Economics, 16 (2), 223 238. Ho, Winky and Ganesan, Sivaguru (1998) On land supply and he price of residenial housing. Journal of Housing and he Buil Environmen, 13 (4). Levin, E. J. and Wrigh, R. E. (1997) Speculaion in he Housing Marke?, Urban Sudies, 34, (9) 1419 1437. Ley, D. and Tuchener, J. (2001) Immigraion, Globalizaion and House Price in Canada s Gaeway Ciies, Housing Sudies, 16, 199 223. Mankiw, N. G., and Weil, D. N. (1989) The Baby Boom, he Baby Bus, and he Housing Marke, Regional Science and Urban Economics, 19, 235 258, Norh-Holland. Manning C., (1989) Explaining Inerciy Home Price Differences. Journal of Real Esae Finance and Economics, 2, 131 149. Malpezzi, S. (1996) Housing Price Exernaliies, and Regulaion in U.S. Meropolian Areas, Journal of Housing Research, 7, 209 241. Mancheser, J., (1987) Inflaion and Housing Demand: A New Perspecive, Journal of Urban Economics, 21, 105 125. Megbolugbe, I., Sa-Aadu, J. and Shilling, J. D. (1999) Elderly Female-Headed Households and he Decision o Trade Down, Journal of Housing Economics 8, 285 300. Muellbauer, J. and Murphy, R. (1992) Booms and Buss in UK Housing Marke, The Economic Journal, 107, 1701 1727. Muh, R. F. (1960) The Demand for Non- Farm Housing. In: Quigley, J. M. (ed), The Economics of Housing, Volume I, Edward Elgar Publishing, Inc., 1997. Nellis, J. G. and Longboom, J. A. (1981) An Empirical Analysis of he Deerminaion of Housing Prices in he Unied Kingdom, Urban Sudies, 18, 9 21. Newell, G. and Chau, Kwong Wing (1996) Linkages beween direc and indirec propery performance in Hong Kong. Journal of Propery Finance, 7 (4), 9 29. Ozanne, L. and Thibodeau, T. (1983) Explaining Meropolian Housing Price Differences, Journal of Urban Economics, 13 (January), 51 66. Peng, R. and Wheaon, W. C. (1994) Effec of Resricive Land Supply on Housing in Hong Kong: An Economeric Analysis, Journal of Housing Research, 5 (2), 263 292. Pollak, R. A. and Wales, T. J. (1981) Demographic Variables in Demand Analysis, Economerica, 49, (6) November. Reicher, A. K. (1990) The Impac of Ineres Raes, Income, and Employmen upon Regional Housing Prices, Journal of Real Esae and Economics, 3, 373 391. Waxman, P. (1993) Business Mahemaics and Saisics, Third Ediion. Prenice Hall. Wheeler, M. and Chowdhury, A. R. (1993) The Housing Marke, Macroeconomic Aciviy and Financial Innovaion: An Empirical Analysis of US daa, Applied Economics, 385 1392. Wong, R. (1993) Propery and Housing Markes in Hong Kong: Issues and Analyses, HKCRE Leers, No. 18, January. Woodward, S. E. (1991) Economiss Prejudices: Why he Mankiw Weil sory is no credible, Regional Science and Urban Economics, 21, 531 537, Norh-Holland. www.uor2u.ne (2002). 70 THE AUSTRALIAN JOURNAL OF CONSTRUCTION ECONOMICS AND BUILDING VOL.2 NO.2