segmentation of australian housing markets: 1989–98

19
This article was downloaded by: [University of Newcastle (Australia)] On: 06 October 2014, At: 11:24 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Journal of Property Research Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/rjpr20 Segmentation of Australian housing markets: 1989–98 Yong Tu Published online: 07 Feb 2011. To cite this article: Yong Tu (2000) Segmentation of Australian housing markets: 1989–98, Journal of Property Research, 17:4, 311-327 To link to this article: http://dx.doi.org/10.1080/09599910010001420 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is

Upload: yong

Post on 17-Feb-2017

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Segmentation of Australian housing markets: 1989–98

This article was downloaded by: [University of Newcastle (Australia)]On: 06 October 2014, At: 11:24Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK

Journal of Property ResearchPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/rjpr20

Segmentation of Australianhousing markets: 1989–98Yong TuPublished online: 07 Feb 2011.

To cite this article: Yong Tu (2000) Segmentation of Australian housing markets:1989–98, Journal of Property Research, 17:4, 311-327

To link to this article: http://dx.doi.org/10.1080/09599910010001420

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all theinformation (the “Content”) contained in the publications on our platform.However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness,or suitability for any purpose of the Content. Any opinions and viewsexpressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of theContent should not be relied upon and should be independently verified withprimary sources of information. Taylor and Francis shall not be liable for anylosses, actions, claims, proceedings, demands, costs, expenses, damages,and other liabilities whatsoever or howsoever caused arising directly orindirectly in connection with, in relation to or arising out of the use of theContent.

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan,sub-licensing, systematic supply, or distribution in any form to anyone is

Page 2: Segmentation of Australian housing markets: 1989–98

expressly forbidden. Terms & Conditions of access and use can be found athttp://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 3: Segmentation of Australian housing markets: 1989–98

Segmentation of Australian housing markets:1989–98YONG TUDepartment of Real Estate, National University of Singapore, 10 Kent Ridge Crescent, Singapore 119260E-mail: [email protected]

Received 3 April 1999Revised 19 November 1999Accepted 17 December 1999

Summary

The Australian national housing market has gone through a long recovery to achieve its currentlevel. Using econometric modelling techniques, this paper has found that the real weekly earningsper employee, the nominal mortgage rates, the unemployment rates and the housing constructionactivities are the key driving forces behind to lead the national housing market out of its recession.The Australian national housing market comprises a series of segmented subnational housingmarkets. It implies the disparities of economic performance at subnational level.

Keywords: housing, prices, cointegration, causality, Australia

1. Introduction

Modelling the dynamics of housing prices has been an active research area for manyyears, for example, Whitehead (1974), Hendry (1984), Meen (1996), Bourassa andHendershott (1995), and Munro and Tu (1996). This is because the housing transactionprices have a major impact on the individual wealth holdings, the interpersonal distribu-tion of wealth, the cost of obtaining housing services as well as the level of the residen-tial housing construction activities (Hendry, 1984). The instability in the housing priceshas become an important policy issue in many countries, particularly in the countrieswith a high homeownership rate. Therefore, it has become important for policy analystsboth to understand the causes of housing price changes and also to predict when themarket might turn. The dynamic patterns of the housing prices vary from place to placeand from time to time. The housing market in a country is rooted in the environment ofthe country. Therefore, the housing models and the conclusions derived from thathousing market might not be applicable to the other countries (Chen and Chen,1998).

Journal of Property ResearchISSN 0959-9916 print/ISSN 1466-4453 online © 2000 Taylor & Francis Ltd

http://www.tandf.co.uk/journalsDOI: 10.1080/0959991001000142 0

Journal of Property Research, 2000, 17(4) 311–327

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 4: Segmentation of Australian housing markets: 1989–98

This paper is aimed at modelling the Australian national and the Australian capital cityhousing price dynamics between 1989 and 1998.1 The Australian population are verysparsely distributed with most of the residents residing in the capital cities. The nationalaverage housing prices used in this study are actually the average housing prices of sevenAustralian capital city housing prices. The Australian national real housing prices haveincreased at an average annual rate of 2.8%2 from 1989 to 1998. The increases have notbeen at a constant rate, instead the real housing prices show a strong cyclical trend as wellas a clear regional pattern. The national and Sydney real housing prices experienced amore similar dynamic pattern than that of the rest of the capital city real housing prices(Fig. A1). Melbourne housing market experienced the worst and the most prolongedrecession comparing with the rest of the capital city housing markets since the late 1980s(Fig. A1).

The complicated changing patterns have raised a number of questions. First of all, whatare the leading factors driving the Australian national housing market out of its 1989recession? Bourassa and Hendershott (1995) undertook a pooled study of the changingrates of the real housing prices across the Australian six capital cities between 1979 and1993. They predicted that, given the economic recovery taking place in the mid 1990s inAustralia, the increases in the employment, income as well as the net immigration wouldbring the Australian housing market out of its 1989 recession, while the increases in thereal mortgage rates would have the opposite effects. The disadvantage of the study is thatthis is a pooled study and each capital city is treated equally in terms of the determinantsof the real housing price changes. AETM3 (1991) investigated the real housing prices ofSydney, Melbourne and Adelaide. It is concluded that there are signi� cant differences ofthe real housing price dynamics across the three cities investigated, and the differenceswere increasing during the 1980s as the housing price appreciations were at different rates.Therefore, the national housing price model should be different from each capital city’shousing price model.

Second, do the changes of Sydney real housing prices dominate the Australian nationalhousing prices? Sydney accommodates 25% of the Australian population. It is one of themajor destinations of immigrations, and it has the highest housing price level. As a part ofthe national housing market, Sydney real housing prices signi� cantly weight up the levelof national real housing prices. However, AETM (1991) shows that the real housing priceswere only slightly more volatile in Sydney than in the other cities. Therefore, it is still aquestion if the dynamics of Sydney real housing prices has any signi� cant in� uences onthe national real housing prices or any other city’s housing prices.

Third, are the Australian housing markets at subnational level segmented?4 Thesegmentation here is de� ned as that the dynamics of the housing prices is determined byits internal factors and will not be in� uenced by the changes of other city housing prices

1 Australia is divided into six states plus one Australian Capital Territory. Canberra, the Australian capital city,is located within the Australian Capital Territory. For the rest of the states, Sydney is the capital city of the Stateof New South Wales. Melbourne is the capital city of the State of Victoria. Brisbane is the capital city of theState of Queensland. Adelaide is the capital city of the State of Southern Australia. Perth is the capital city ofthe State of Western Australia. Hobart is the capital city of the State of Tasmania.2 The � gure is calculated based on the medium established housing price data provided by CommonwealthBank of Australia and Australian Housing Industry Association .3 AETM stands for Applied Economics PTY LTD and Travers Morgan PTY LTD, Australia.4 The Australian capital cities are used here to represent subnational housing markets.

312 Yong

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 5: Segmentation of Australian housing markets: 1989–98

both in the long run and in the short run. This issue has never been formally mentioned orempirically testi� ed in the previous Australian housing research literature. According tothe estimation given by AETM (1991) in 1989, the estimated elasticity of housing pricesto household income range from 1.3 in Melbourne to 1.7 in Sydney. From 1977 to 1989the estimated elasticity of housing prices to housing supply was ranged from 2 0.39 inAdelaide, 2 0.55 in Melbourne and 2 0.65 in Sydney. These evidences imply that thecapital city housing markets perform differently although some common factors, like theaccessibity to the CBD and the environmental attributes, have been found to be stronglyrelated to an individual city’s average housing price dynamics.

This paper is organized as follows. Section 2 presents the methodology used in thispaper. Section 3 will identify both the short run and the long run national real housingprice determinants. Section 4 will testify the segmentation of the housing markets acrossseven capital cities. Some policy implications and conclusions derived from the researchwill be explored in Section 5.

2. Methodology

This section will introduce the variable selection, data collection, econometric model usedfor national housing price modelling and the statistical tests used for testing thesegmentation of the housing markets at sub-national level.

2.1. Variable selection

To model the real housing price dynamics, it is important to select a set of explanatoryvariables that are the potential variables in determining the real housing price dynamics.AETM (1991) reviews the previous Australian national and international literature ofhousing price models and summarizes 17 potential determinants of the real housing prices.Meen and Andrew (1998) reviews 25 years of housing price modelling work in the UK.They point out that few housing price models include all the potential determinants and notall variables are equally important empirically. Besides, the selection of the explanatoryvariables is also limited by the data availability. Based on the previous literature and theavailability of time series data in Australia, � ve types of the explanatory variables areconsidered in this paper.

2.1.1. Demographic variables. There are continuing changes in the Australian populationand the population structure over the years. The changes are then transformed from thepopulation totals into the households requiring housing. And hence, the changes may havelong run effects on the real housing prices. The changes mainly result from naturalpopulation changes and the international migration. The natural population changes are aslow process. Over a 10-year period, the effects on real housing prices are likely to beinsigni� cant. However, net immigrations can be a signi� cant variable in� uencing realhousing prices. Bourassa and Hendershott (1995) provides empirical evidence to show thatnet migration is a signi� cant factor in determining the capital city’s housing prices. In thisstudy, the number of the net immigrants is selected as a potential determinant of theAustralian national real housing prices.

2.1.2. Affordability. It is indicated by income and unemployment rate. Income isconsidered as the most important variable in� uencing housing prices. The evidences have

Australian housing market 313

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 6: Segmentation of Australian housing markets: 1989–98

been found both in the Australian context (AETM, 1991; Bourassa and Hendershott, 1995)as well as in the international context (Whitehead, 1974; Buckley and Ermisch, 1982;Stern, 1992 and Munro and Tu, 1996). The work of Meen and Andrew (1998) concludesthat the elasticity of real housing prices to real income is greater than 1, although the actualestimated scales vary from model to model. However, since buying a house is a long runpurchasing process through mortgage � nancing system, the demand for housing maydepend upon the expected permanent income and may not be signi� cantly affected by thetemporary changes in income (AETM, 1991; Pain and Westaway, 1997). This implies that,income may have a more signi� cant effect on housing prices in the long run than that ofin the short run. The unemployment rates are regarded as a proxy for the income growthand the income stability. When the wage rate is regulated, a constant low level ofunemployment rate indicates the prosperity of permanent income. In this study, the weeklyearnings per employee and the unemployment rates are selected as the potentialdeterminants of the Australian national real housing prices.

2.1.3. Housing � nance. Housing mortgage � nance directly in� uences a would behomebuyer’s affordability. A higher mortgage rate will increase the user cost of buying ahome, therefore, may decrease the demand, and then the real housing prices may dropdown. An argument is if the real mortgage rates or the nominal mortgage rates will havesigni� cant in� uences on the real housing price changes. The mortgage repayment iscalculated by the nominal mortgage rates that will in� uence the household’s affordabilityand their perception for the user cost of housing. This implies that the nominal mortgagerates rather than the real mortgage rates may have a signi� cant in� uence on the realhousing prices.

2.1.4. In� ationary rates. For most of the households, buying a home is an investment tohedge the in� ation. The constant high in� ation rates may make housing as a desirable assetfor the households. During the investigated period, Australia experienced a low level ofin� ation. Therefore, the housing demand during this period was unlikely to be in� uencedby the in� ation rate. The signi� cance of the in� uence will be testi� ed in this study.

2.1.5. Housing supply. The housing commences, completions, building costs and landcosts, all may have in� uences on the real housing prices. Because of the data availability,only the housing commences and the housing completions are considered in this study.

2.2. Data collection

Based on the above variables, the following time series are selected for the empiricalmodelling. The selection is limited by the data availability. The Australian national realhousing prices (RHP) are the average housing prices of seven capital city’s housing prices.There are four major housing price data sources available in Australia: ABS (AustralianBureau of Statistics), BIS (BIS Shrapnel), REIA (Real Estate Institute of Australia) andHIA/CBA (Australian Housing Industry Association and the Commonwealth Bank ofAustralia). The HIA/CBA housing price time series adjusted by the ABS housing priceindex are selected for this research. The reasons are that � rst, statistically, they provide 40quarterly data (the housing prices of the � rst and the second quarters in 1989 are estimatedusing both the HIA/CBA and the ABS data), which makes the modelling work feasible.Second, they cover both the national and seven capital city housing prices, and hence they

314 Yong

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 7: Segmentation of Australian housing markets: 1989–98

are consistent. Third, the time series cover a 10-year period and understanding thisparticular 10 year housing price dynamics is crucial. During the 10 years, the nationalhousing market went from severe recession, recovery to another price boom. The realhousing prices at national level and capital city level are obtained by de� ating the nominalhousing prices using the CPI (consumer price index) excluding housing with 1989–90 asthe base year.

For the explanatory variables used in Section 3 of the paper, the following time seriesare selected. Weekly earnings per employee (RINC), in� ation rates derived from the CPI(INFLA), unemployment rates (UEMP), housing completions (COMPL) and commences(START), and the net immigration (NIMMG) to Australia are selected from the ABSpublications. Mortgage lending rates (MORT stands for the nominal mortgage rates,RMORT stands for the real mortgage rates) are selected from the Reserve Bank ofAustralia Bulletins. Log transformation (10 as the base) is applied to all the variablesexcept mortgage rates, in� ation rates and unemployment rates. The consumer price indexis used to de� ate the nominal variables with 1989–90 as the base year.

2.3. Econometric models

The econometric analysis in this paper takes the following two steps. The � rst stepincludes unit root tests to analyse the times series stability, a cointegration analysis as wellas an error correction model to � nd both the long run and short run determinants of realnational housing prices. The empirical results are presented in Section 3 of this paper.Similar empirical analyses based on the UK housing markets can be found in Drake (1993)and Munro and Tu (1996).

The second step is aimed at testing the segmentation of Australian housing markets.Unit root tests are applied to all real housing price time series to investigate the stability.A cointegration analysis of real housing prices across all seven capital cities is undertakento investigate if there are any long run relationships among the real housing prices of sevencapital cities. Then a pairwise cointegration analysis is conducted for every two capital cityreal housing prices to identify their long run relationships. Finally, Granger causalitymethod is applied to investigate the short run relationships. The tests will show if thenational real housing price dynamics leads the changes of the capital city real housingprices, and vice versa; and if Sydney real housing price dynamics leads the changes of thenational as well as the rest of the capital city real housing prices, and vice versa. Theempirical results are presented in Section 4 of the paper. Similar empirical analysis basedon the UK housing market can be found in Alexander and Barrow (1994) and MacDonaldand Taylor (1993).

3. The determinants of the Australian national real housing prices

This section presents the results of modelling the Australian national real housing prices.Before cointegration analysis is undertaken, unit root tests are applied to all the time seriesfor integration analysis that will explore the time series properties of the selected variablesand will indicate if the time series are suitable for cointegration analysis.

The unit root tests (Tables A1 and A2) show that real weekly earning per employee(RINC) is integrated of order zero with deterministic trend (the time trend). After the time

Australian housing market 315

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 8: Segmentation of Australian housing markets: 1989–98

trend is removed from the time series (or one difference is taken), the series becomesstationary. This means that the Australian national average weekly earning increment fromone quarter to another moved around a time trend during the investigated period (Fig. A3).The country also experienced a stable net immigration (NIMMG) in� ow and the stablein� ation rates (INFLA) as both time series are stationary.

The nominal mortgage rates (MORT) are weakly integrated of order one. This impliesthat, during the investigated period, the nominal mortgage rates were decreasing at anaccelerating rate, which could be faster than the income growth or the changes of the othereconomic variables investigated in this study. It also implies that, the nominal mortgagerates may take a critical role in the Australian housing market recovery from its 1989recession. Being different, there is strong evidence to show that the real mortgage rates(RMORT) are integrated of order one. Figure A2 illustrates the trends of mortgagerates.

The rest of the variables including the Australian national real housing prices (RHP) aresigni� cantly integrated of order one. The results of unit root tests show that cointegrationmethodology can be applied to the time series selected.

Both the cointegration and the restricted cointegration analysis show that the netimmigration, the in� ation rates, the real mortgage rates and the housing commences arenot in the cointegration vector. Therefore in the long run, the real weekly earnings peremployee, the unemployment rates, the housing completions, and the nominal mortgagerates determine the national real housing prices. Full statistical diagnoses are given inTable A5. Figure A4 gives the cointegration graphic tests illustrating that only onecointegration vector should be selected (Equation 1); that constructs the long run nationalreal housing prices model. Error correction model is then applied to the time series toobtain short run national real housing prices model. The results are given below (Equations2 and 3). Full statistical diagnoses are given in Table A5. The � tness of the model isillustrated by Fig. A5.Long run real housing prices model: cointegration vector

RHPt 5 2.3280 3 RINCt 2 0.0124 3 UEMPt 2 0.6942 3 COMPLt 2 0.0039 3MORTt (1)

Short run real housing prices model: error correction model

D RHPt 5 0.2921 1 0.3920 3 D RHPt2 2 2 0.0075 3 D UEMPt 2 1 2 0.0115 3D UEMPt 2 2 2 0.0119 3 D MORTt 2 1 2 0.0103 3 D MORTt2 2 2 0.1088 3D STARTt 2 2 0.1328 3 ECRt 2 1 (2)

where,

ECRt 2 1 5 RHPt2 1 2 (2.3280 3 RINCt 2 1 2 0.0124 3 UEMPt 2 1 2 0.6942 3COMPLt 2 1 2 0.0039 3 MORTt2 1 (3)

From the long run real housing prices model (Equation 1), the following conclusions canbe derived. The long run elasticity of real housing prices to real income (weekly earningper employee) is estimated as 2.3435. The elasticity of real housing prices to the numberof housing completions is estimated as 2 0.6931. One percentage point increase in theunemployment rates may decrease the real housing prices by 2.8145%, and one percentagepoint increases in the nominal mortgage rates may decrease the real housing prices by0.8940%, holding other factors unchanged.

316 Yong

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 9: Segmentation of Australian housing markets: 1989–98

In the short run real housing price model, the ECR (Equations 2 and 3) stands for theequilibrium correction representation, which is derived from the cointegration vector(Equation 1). The modelling results show that the coef� cient of the ECR is very stable andsigni� cant. The value is very close to the one derived from the cointegration matrix. Itmeans that, if the national real housing prices (in Log value) leave its long run equilibriumpath by one unit at time t 2 1, it will be corrected back to its equilibrium path by 13.28%at time t automatically, holding other factors unchanged. The automatic adjustment can beinterpreted as the spontaneous housing demand and supply changes in response to the pricechanges (with one-quarter lag).

As expected, the income and the housing completions are not signi� cant in the short runhousing price model. However, the changes of the real housing prices are one-quarter lagof the changes of housing commences. The modelling results also show that the short rundynamics of the real housing prices do not spontaneously respond to the changes of theexplanatory variables. The in� uences of the explanatory variable on the real housing pricesare at least one-quarter lag. In the short run model, all the coef� cients are signi� cant at 5%signi� cant level, except the variable of D UEMPt 2 1. It is kept in the model, as it issigni� cant at 6% signi� cant level. The coef� cients in the short run model (Equation 2)show that mortgage rates (MORT), unemployment rates (UEMP) and housing commences(START) are the signi� cant explanatory variables to cause the short run national realhousing price � uctuations.

4. The relationships between the Australian housing markets at subnational level

As it has been argued in Section 1, that the dynamic patterns of real housing prices acrossthe Australian capital cities vary dramatically. This section is aimed at testing if thesehousing markets are segmented or inter-related. The unit root tests (Tables A3 and A4)show that all the real housing prices are integrated of order 1, except that the real housingprices in Adelaide are weakly integrated of order 0. In other words, Adelaide real housingprices are more stable comparing with the others during the investigated period.

The cointegration tests for seven capital city real housing prices (Table A6) show thatthere are three potential cointegration vectors. However, the cointegration graphic testsreject all the cointegration vectors. In other words, during the investigated period, the realhousing prices across the capital cities may follow different long run paths.

The pair wise cointegration tests show that there are 13 potential pairs, which arepotentially cointegrated (Table 1). But the cointegration graphic tests demonstrate that onlyHobart is weakly cointegrated with Canberra, Adelaide and Brisbane separately. For therest of the pairs, they are not cointegrated. For example, the tests show that the realhousing prices in Sydney and Melbourne were actually drifting apart during most of thetime between 1989–90 although the gap became narrow in the late of 1990s (see also Fig.A1).

The testing results in Table 1 prove that in the long run, the real housing prices acrossthe Australian capital cities were not moving at the same pace or moving along the samelong run path between 1989 and 1990. The next statistical tests are if there are any rippleeffects between the Australian capital city real housing prices. Ripple effects have beenfound among the British regional housing markets (MacDonald and Taylor, 1993;Alexander and Barrow, 1994; Munro and Tu, 1996).

Australian housing market 317

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 10: Segmentation of Australian housing markets: 1989–98

Table 2 summarizes the results of Granger causality tests between the real housingprices in Sydney and the real housing prices of the other capital cities, as well as betweenthe national real housing prices and the real housing prices of the capital cities. Fullstatistical diagnoses are given by Tables A7 and A8.

Table 1. The pair wise cointegration between the Australian capital citiesa

Melbourne Brisbane Adelaide Perth Hobart Canberra

Sydney 17.25 16.11 13.69 19.35b 22.45* 11.2921.38 22.97 18.77 28.32* 33.21* 13.59

Melbourne 23.55* 14.99 19.9* 14.59 15.6828.21* 21.6 29.16* 27.6* 18.57

Brisbane 19.56* 21.46* 18.81 12.221.54 26.18* 25.63* 15.48

Adelaide 21.99* 28.33** 8.14625.25 31.79** 10.73

Perth 17.95 19.36*31.15** 22.33

Hobart 20.06*23.37

Notes: a H0: Rank = 0; H1: Rank > 0.Critical value (eigenvalue test in the � rst line of each cell) = 19.0 at 5%.Critical value (trace test in the second line of each cell) = 25.3 at 5%.All tests are with a trend.b * and ** denote that it is signi� cant at 5% or 1% separately. It means that the pairs are potentiallycointegrated.

Table 2. Grange causality testing results

Causality directionbCausality testresult Causality direction

Causality testresult

Sydney ® Melbourne Yes [2]a Australia® Melbourne Yes [2]Melbourne ® Sydney No Melbourne ® Australia NoSydney ® Brisbane No Australia® Brisbane NoBrisbane ® Sydney Yes [1] Brisbane ® Australia Yes [1]Sydney ® Adelaide No Australia® Adelaide NoAdelaide ® Sydney No Adelaide ® Australia NoSydney ® Perth No Australia® Perth NoPerth ® Sydney No Perth ® Australia NoSydney ® Hobart No Australia® Hobart NoHobart ® Sydney No Hobart ® Australia NoSydney ® Canberra No Australia® Canberra NoCanberra® Sydney No Canberra® Australia NoSydney ® Australia NoAustralia® Sydney No

Notes: a The number in the [ ] denotes the lag.b A® B denotes if the changes in A will result in any changes in B.

318 Yong

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 11: Segmentation of Australian housing markets: 1989–98

A few causal relationships are found from Granger causality tests. The real housingprices in Sydney lead the real housing prices in Melbourne by two quarters, and the realhousing prices in Brisbane lead the real housing prices in Sydney by one quarter during theinvestigation period. The national real housing prices lead Melbourne real housing pricesby two quarters, and Brisbane real housing prices lead the national real housing prices byone quarter during the investigation period. Therefore, among the 13 pairs with 26potential causal relationships, only four causal relationships are proved to be signi� cant. Aunidirectional causal relationship is derived and illustrated by Fig. 1. The ripple effectshappened from Brisbane, Sydney to Melbourne during the investigated period. In otherwords, the short run real housing price � uctuations in Melbourne can be explained partlybut signi� cantly by the real housing price � uctuations in Sydney or by the � uctuations ofthe national average real housing prices. The � uctuations of the real housing prices inBrisbane have the same effects on that of the real housing prices in Sydney and thenational average real housing prices.

The above Grange causality tests explore that the short run real housing price� uctuations have a clear geographic pattern. The real housing price � uctuations in thecapital cities located along the east coast area experienced ripple effects from north tosouth (from Brisbane, Sydney to Melbourne), while there were no similar ripple effectsamong the housing markets in the rest of the cities.

5. Conclusions and implications

This paper has two main conclusions. First, the � ndings have demonstrated a number ofrobust relationships between the national real housing prices and some related economicfactors and have illustrated that the long run and short run real housing price determinantscan be different. For example, real income is the most critical factor in� uencing realhousing price dynamics in the long run, but it does not have signi� cant in� uence on theshort run housing price � uctuations, while unemployment rates and nominal mortgagerates have both the long run and short run effects on the national real housing prices.

Fig. 1. Causal relationships

Australian housing market 319

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 12: Segmentation of Australian housing markets: 1989–98

Housing completions are an important factor in determining long run real housing pricemovements, while housing commences have signi� cant in� uences on short run realhousing price � uctuations.

Second, the Australian housing markets at subnational level are highly segmented,particularly in the long run, although a few causal relationships have been found in theshort run. Sydney real housing prices neither dominate the national housing pricedynamics nor the housing price dynamics in the rest of the cities in the long run. In theshort run, it has in� uences on Melbourne housing prices, but it does not have any effectson any other cities.

Three important implications can be derived from the above conclusions. First of all, thesegmentation of Australian housing markets at subnational level implies the regionaleconomic disparities. This is because the housing market performance and housing pricedynamics are determined by the economic performance. If the regional (states inAustralian circumstance) disparities are widening in terms of economic development, thehousing markets and housing prices in each state will tend to move to the differentdirections or to the same direction but at different paces. In fact, a good deal of researchhas shown that income inequalities have been common in Australia. Cashin andStrappazzon (1998) discovered that the cross-state dispersion of per capita incomes hadwidened between 1976–1991 indicating that cross-state incomes had become less equal.

The second implication is that national housing policy may not have the same impactson the different capital cities. In other words, national housing policy should consider theregional disparities because of the different dynamic patterns of the regional housingmarkets.

Finally, national housing price models cannot represent the regional housing pricemodels. In the long run, the real housing price movements in each capital city are mainlydriven by their internal economic factors, for example, the average income, housingcompletions, mortgage rates in the city. Those economic factors and their impacts on thecity’s real housing prices can be different from one city to another city. In the short run,the real housing price � uctuations in Brisbane and Sydney should be considered as theexplanatory variables in explaining Melbourne real housing price dynamics.

Future work could be directed usefully at examining how national economic structure,institutional structure and the macroeconomic policies in� uence the housing marketperformance at subnational level. Other important research questions shall includemodelling housing price dynamics at subnational level in order to further understand thedisparities and similarities of regional housing market dynamics.

References

AETM (1991) Determinants of the Prices of Established Housing. A report published by AppliedEconomics PTY LTD and Travers Morgan PTY LTD, Australia.

Alexander, C. and Barrow, M. (1994) Seasonality and cointegration of regional house prices in theUK, Urban Studies 31(10), 1667–89.

Bourassa, S.C. and Hendershott, P.H. (1995) Australian capital city real housing prices, 1979–1993,The Australian Economic Review 3, 16–26.

Buckley, R. and Ermisch, J.F. (1982) Theory and empiricism in the econometric modelling of houseprices, Urban Studies 20(1), 83–90.

320 Yong

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 13: Segmentation of Australian housing markets: 1989–98

Cashin, P. and Strappazzon, L. (1998) Disparities in Australian regional incomes: are they wideningor narrowing? The Australian Economic Review 31(1), 3–26.

Chen, M.C. and Chen, C.M. (1998) Review of house price studies: development and comparison, apaper presented at the First Conference of Chinese Real Estate Community, and the Third AsianReal Estate Society Annual Conference, Taipei, Taiwan, August, 1998.

Drake, L. (1993) Modelling UK house prices using cointegration: an Application of the JohansenTechnique, Applied Economics 25, 1225–28.

Hendry, D.F. (1984) Econometric modelling of house prices in the United Kingdom, inEconometrics and Quantitative Economics (edited by D.F. Hendry and K.F. Wallis), BasilBlackwell, Oxford.

MacDonald, R. and Taylor, M.P. (1993) Regional house prices in Britain: long-run relationships andshort-run Dynamics, Scottish Journal of Political Economy 40(1), 43–55.

Meen, G.P. (1996) Spatial aggregation, spatial dependence and predictability in the UK housingmarket, Housing Studies 11(3), 305–72.

Meen, G.P. and Andrew, M. (1998) Modelling regional house prices: a review of the literature, areport prepared for the Department of the Environment, Transportation and the Regions, Centrefor Spatial and Real Estate Economics, Department of Economics, the University of Reading,UK.

Munro, M. and Tu, Y. (1996) The dynamics of UK national and regional house prices, Review ofUrban and Regional Development Studies 8, 186–201.

Pain, N. and Westaway, P.(1997) Modelling structure change in the United Kingdom housingmarket: a comparison of alternative house price models, Economic Modelling 14(4),587–610.

Stern, D. (1992) Explaining UK house price in� ation 1971–89, Applied Economics 24, 1327–33.Whitehead, C. (1974) The UK Housing Market: an Econometric Model. Saxon House, London.

Australian housing market 321

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 14: Segmentation of Australian housing markets: 1989–98

Appendix

Table A1. Unit root tests of time series Yt

Yt

t-DF/ADF(1) withouta deterministic trend LAG(2)

t-DF/ADF with adeterministic trend LAG

RHP 2.4255 0 0.0271 0RINC –2.6218 5 –4.1207* 5UEMP –0.8303 2 –2.0805 2MORT –1.5876 1 –2.4892 1RMORT –1.5248 0 –2.2037 0NIMMG –4.6749** (3) 0 –4.6440** 0INFLA –3.7800** 0 –3.6963* 0COMPL –1.4442 0 –2.1826 0START –2.5652 3 –2.5953 3

H0: Non-stationary H1: Stationary H0: Non-stationary H1: Stationary(4)

Unit root Critical value at 5%: –2.975 (*) Critical value at 5%: –3.587 (*)tests Critical value at 1%: –3.696 (**) Critical value at 1% –4.338 (**)

Diagnosis: Reject H0 if t-DF/ADF< Critical value

Diagnosis: Reject H0 if t-DF/ADF< Critical value

Notes:(1) Constant is included in the unit root testing model. Seasonality is also considered .(2) The maximum lag considered is 10.(3) * denotes the signi� cance at 5%, ** denotes the signi� cance at 1%.(4) When a deterministic trend is included, the null hypothesis H0 means that, yt is not a stationary time series.The alternative hypothesis H1 means that it is a stationary time series around the time trend or it is integratedof I(0) with a trend.

Table A2. Unit root tests of time series D Yt

D Yt

t-DF/ADF without adeterministic trend LAG

t-DF/ADF with adeterministic trend LAG

D RHP –4.662** 0 –3.6127* 0D RINC –5.8437** 0 –5.6704** 0D UEMP –5.8596** 0 –5.7507** 0D MORT –3.0078* 0 3.2529 0D RMORT –4.6207** 0 –4.3789** 0D NIMMG N/A N/AD INFLA N/A N/AD COMPL –5.8021** 0 –6.1449** 0D START –3.3043* 0 –3.1558* 0

H0: Non-stationary H1: Stationary H0: Non-stationary H1: Stationary

Unit root Critical value at 5%: –2.975 (*) Critical value at 5%: –3.587 (*)tests Critical value at 1%: –3.696 (**) Critical value at 1% –4.338 (**)

Diagnosis: Reject H0 if t-DF/ADF< Critical value

Diagnosis: Reject H0 if t-DF/ADF< Critical value

322 Yong

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 15: Segmentation of Australian housing markets: 1989–98

Table A3. Unit root tests of the real housing prices

Real housingprices

t-DF/ADF without adeterministic trend LAG

t-DF/ADF with adeterministic trend LAG

Australia 2.4255 0 0.0271 0Sydney 1.4158 0 –0.63183 8Melbourne –2.0409 2 0.43846 0Brisbane –0.0096 8 –1.7406 6Adelaide –3.2916* 4 –3.1411 4Perth 0.6920 1 –2.8746 0Hobart –0.3904 2 –4.0838* 0Canberra –1.8802 0 –1.7464 0

H0: Non-stationary H1: Stationary H0: Non-stationary H1: Stationary

Unit root Critical value at 5%: –2.975 (*) Critical value at 5%: –3.587 (*)tests Critical value at 1%: –3.696 Critical value at 1% –4.338

Diagnosis: Reject H0 if t-DF/ADF< Critical value

Diagnosis: Reject H0 if t-DF/ADF< Critical value

Table A4. Unit root tests of the real housing prices

D Real housingprices

t-DF/ADF without adeterministic trend LAG

t-DF/ADF with adeterministic trend LAG

D Australia –4.662** 0 –3.6127* 0D Sydney –5.0576** 0 –6.3205** 0D Melbourne –3.4165* 0 –4.1924* 0D Brisbane –5.5539** 0 –5.6582** 0D Adelaide –5.7043** 0 –5.6107** 0D Perth –7.9351** 0 –8.0486** 0D Hobart –7.7489** 0 –7.5382** 0D Canberra –6.9283** 0 –6.8918** 0

H0: Non-stationary H1: Stationary H0: Non-stationary H1: Stationary

Unit root Critical value at 5%: –2.975 (*) Critical value at 5%: –3.587 (*)tests Critical value at 1%: –3.696 (**) Critical value at 1% –4.338 (**)

Diagnosis: Reject H0 if t-DF/ADF< Critical value

Diagnosis: Reject H0 if t-DF/ADF< Critical value

Australian housing market 323

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 16: Segmentation of Australian housing markets: 1989–98

Table A5. The long run and the short run determinants of the Australian real housingprices 1989–1998

Long run model (RHPt): Co-integration vectorRHPt = 2.3280*RINCt–0.0124*UEMPt–0.6942*COMPLt–0.0039*MORTt

Co-integration testsH0: Rank = p Eigenvalue test Trace test

P = 0 38.73* 104.5**P £ 1 35.42* 65.81**P £ 2 21.63 30.39

Short run model ( D RHPt)

Variables Coef� cients Std error t-value

D RHPt–2 0.3920 0.1121 3.498**D UEMPt–1 –0.0075 0.0037 –1.995D UEMPt–2 –0.0115 0.0042 –2.753**D MORTt–1 –0.0119 0.0053 –2.231*D MORTt–2 –0.0103 0.0040 –2.585*ECRt–1 –0.1328 0.0438 –3.032**D STARTt–1 –0.1088 0.0494 –2.202*Constant 0.2921 0.0987 2.957**ECRt–1 = RHPt–1–(2.3280*RINCt–1–0.0124*UEMPt–1–0.6942*COMPLt–1–0.0039*MORTt–1s –0.0086 AR 1–3 F(3,23) = 0.0862 [0.9669]RSS = 0.0019 ARCH 3 F(3,20) = 0.4298 [0.7339]R2 = 0.74 Normality c 2 = 0.9632 [0.6178]F(7,26) = 10.638** Heterosedesticity c 2 = 0.42997[0.9303 ]DW = 1.90

Table A6. The co-integration of the Australian housing prices across the capital cities

Co-integration testsH0: Rank £ p H1: Rank > p Eigenvalue test Trace test

P = 0 64.66** 188.3**P = 1 40.96* 123.7**P = 2 37.01* 82.73**P = 3 25.29 45.72

Co-integration vectors

Sydney Melbourne Brisbane Adelaide Perth Hobart Canberra1.0000 –0.9498 –0.5205 1.9721 0.1379 1.1957 0.15192.0221 1.0000 –11.835 -2.8425 0.1705 2.0434 5.1069

–11.581 5.1790 1.0000 10.087 10.658 1.6671 –23.403

Note: * denotes the signi� cance at 5% level, ** denotes the signi� cance at 1% level.

324 Yong

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 17: Segmentation of Australian housing markets: 1989–98

Table A7. Grange causality test between the Sydney housing prices and the rest of the capital cities’housing prices

F(m n–k) m n–k R2 Causality

Sydney ® Mel 3.48** 2[2] 31 0.40 YesMel ® Syd 0.63 2 30 0.40 NoSyd ® Bris 1.25 2 31 0.14 NoBris ® Syd 2.50** 2[1] 31 0.46 YesSyd ® Adel 0.73 2 30 0.24 NoAdel ® Syd 0.90 2 29 0.41 NoSyd ® Perth 0.91 2 31 0.16 NoPerth ® Syd 1.035 2 31 0.29 NoSyd ® Hobart 0.034 2 29 0.38 NoHobart ® Syd 2.025 2 32 0.38 NoSyd ® Canb 0.002 2 29 0.12 NoCamb® Syd 0.3303 2 29 0.393 NoH0: No causalityH1: CausalityF10%(2,29) = 2.50*F5%(2,29) = 3.33**F1%(2,29) = 5.42***

F10%(2,30) = 2.49*F5%(2,30) = 3.32**F1%(2,30) = 5.39***

F10%(2,31) = 2.48*F5%(2,31) = 3.30**F1%(2,31) = 5.36***

Note: The value in the [ ] denotes the lags. ® denotes the causal direction .

Table A8. Grange causality test between the Australian national housing prices and the capitalcities’ housing prices

F(m, n–k) m n–k R2 Causality

Syd ® Aus 0.51 2 29 0.45 NoAus ® Syd 0.81 2 31 0.41 NoAus ® Mel 5.37*** 2[2] 31 0.46 YesMel ® Aus 0.093 2 29 0.46 NoAus ® Bris 1.79 2 31 0.17 NoBris ® Aus 2.89** 2[1] 31 0.52 YesAus ® Adel 1.36 2 31 0.27 NoAdel ® Aus 2.02 2 31 0.52 NoAus ® Perth 0.91 2 31 0.46 NoPerth ® Aus 1.23 2 31 0.47 NoAus ® Hobart 0.61 2 30 0.40 NoHobart ® Aus 0.06 2 30 0.44 NoAus ® Canb 1.29 2 30 0.19 NoCanb ® Aus 0.098 2 30 0.40 NoH0: No causalityH1: CausalityF10%(2,29) = 2.50*F5%(2,29) = 3.33**F1%(2,29) = 5.42***

F10%(2,30) = 2.49*F5%(2,30) = 3.32**F1%(2,30) = 5.39***

F10%(2,31) = 2.48*F5%(2,31) = 3.30**F1%(2,31) = 5.36***

Note: The value in the [ ] denotes the lags. ® denotes the causal direction .

Australian housing market 325

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 18: Segmentation of Australian housing markets: 1989–98

Source : Commonwealth Bank of Australia and Australian Housing Industry Association

Fig. A1. The Australian, Sydney and Melbourne real housing prices: 1989Q1–1998Q2

Source : Australian Bureau of Statistics

Fig. A2. The Australian nominal and real mortgage rates: 1989Q1–1998Q2

Source : Reserve Bank of Australia Bulletin

Fig. A3. The Australian national weekly real earnings per employee: 1989Q–1998Q2

326 Yong

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014

Page 19: Segmentation of Australian housing markets: 1989–98

Source : Generated by PC-FIML

Fig. A4. Cointegration relations

Source : Generated by PC-Give

Fig. A5. The � tness of the Australian national housing price model

Australian housing market 327

Dow

nloa

ded

by [

Uni

vers

ity o

f N

ewca

stle

(A

ustr

alia

)] a

t 11:

24 0

6 O

ctob

er 2

014