House price-stock price relations
69
International Journal of Housing
Markets and Analysis
Vol.3No.1,2010
pp.69-82
#Emerald Group Publishing Limited
1753-8270
DOI 10.1108/17538271011027096
Received 3March 2009Revised 6June 2009
Accepted 22September 2009
House price-stock price relations in Thailand:an
empirical analysis
Mansor H.Ibrahim
Department of Economics,Faculty of Economics and Management,
Universiti Putra Malaysia,Serdang,Malaysia
Abstract
Purpose –The purpose of this paper is to empirically evaluate the wealth and credit-price effects in the relations between housing prices and stock prices for Thailand using quarterly data from 1995to 2006.Design/methodology/approach –The analysis relies on a four-variable vector autoregression (V AR)framework consisting of house prices,stock prices,real output and consumer prices.Granger causality tests,impulse-response functions and variance decompositions simulated from the estimated V AR systems are adopted as bases for inferences.
Findings –The results obtained from Granger causality tests,impulse-response functions and variance decompositions all suggest a unidirectional causality that runs from stock prices to house prices.Thus,the wealth effect is unequivocally supported for the Thai case.The paper also documents the importance of real activity in influencing both house and stock prices.Likewise,stock prices do exert significant effects on real output and to some extent the general price level.These results have an implication that stock market stability is critical for the stability of the housing market as well as the goods market.
Originality/value –The paper provides an emerging market perspective on stock price –house price relations,which seem to be lacking in the literature.
Keywords Housing,Prices,Stock prices,Autoregressive processes,Thailand Paper type Research paper
1.Introduction
The causal relations between house prices and stock prices,which may be due to the wealth effect or credit-price effect or both,have captured a great deal of academic interest.Basically,houses and stocks are considered as investment alternatives.At the same time,the former is also viewed as consumption goods.The unexpected gains in stock prices,reflecting increasing share of the stocks in the investment portfolio and wealth,motivate households to rebalance their portfolios by investing in or consuming more housing services.This so-called wealth effect,thus,posits a causal direction from stock prices to house prices.Meanwhile,the credit-price effect tends to suggest a reverse causation from house prices to stock prices and admits the possibility of persistent spiraling upturns in both prices.Under the credit-price effect,real estates act as collateral to especially credit-constrained firms.The increase in house prices,thus,would be favorable to their balance-sheet position in that they may get access to lower costs of borrowing.With the expanded investments,firms’value and,thus,firms’stock prices rise.As noted by Sim and Chang (2006),this in turn may lead to the increase in the demand for real estates,resulting in their persistent feedback effects.
Empirically,many studies have attempted to assess the relations between housing and stock prices.Early empirical studies on the advanced markets especially the
US
The current issue and full text archive of this journal is available at https://www.doczj.com/doc/177320228.html,/1753-8270.htm
The author would like to express his thanks to two anonymous referees for providing comments on the earlier draft of the paper.The remaining errors,however,remain the author’s responsibility.
IJHMA 3,1
70and UK markets focus on correlations between the two assets’returns.For the US case, such studies as Ibbotson and Siegel(1984),Hartzell(1986),Worzala and Vandell(1993), and Eichholtz and Hartzell(1996)document the correlations between the two variables to be negative ranging fromà0.06toà0.25using either annual or quarterly data over different sample periods.For the UK case,Worzala and Vandell(1993)find the correlation between housing and stock returns to be positive(0.039)while Eichholtz and Hartzell(1996)note their negative correlation(à0.08).Later,Quan and Titman (1999)investigate whether real estate prices and stock prices move together by ways of cross-sectional and panel regressions for17developed and emerging markets.The results generally indicate positive correlation between the two prices.However,while the panel results remain robust,their positive correlation in cross-sectional regression disappears once business cycle variables are added.They attribute these results to the role of economic fundamentals in affecting both prices.
It should be noted that the aforementioned studies provide no indication as to whether the wealth effect or credit-price effect or both are operative in these markets as no inference can be made on their causal direction.More recently,several works have applied time-series techniques of vector autoregression(V AR)to disentangle dynamic or causal interactions between housing and stock prices.Among recent notable studies include Chen(2001), Sutton(2002),Kakes and Van den End(2004),Kapopoulas and Siokis(2005),and Sim and Chang(2006).Chen(2001)examines causal relation between the two asset prices for Taiwan using a bivariate V AR framework.Applying the framework to quarterly data from 1973to1992,he notes significant explanatory power of past stock prices on variations in housing prices and the absence of feedback causal relation.The evidence,thus,is indicative of the wealth effect at work.Sutton(2002)attempts to explain housing prices in six advanced economies–USA,UK,Canada,Ireland,the Netherland,and Australia–in a four-variable V AR framework consisting of housing prices,stock prices,real income,and interest rate.The evidence he documents,which suggests positive responses of house prices to stock price changes,is uniform across countries.The presence of wealth effect is further ascertained by Kakes and Van den End(2004)and Kapopoulos and Siokis(2005) for respectively the Netherlands and Greece.However,supporting the credit-price effect, Sim and Chang(2006)note that real estate prices tend to lead stock prices for Korea.
The present paper seeks to contribute to this line of research by examining the issue for Thailand.Prior to the1997/1998Asian crisis,the Thai economy had grown remarkably well at an average rate of8percent per annum.In parallel,it had witnessed a run-up in both stock market and property market especially in Bangkok metropolitan area,the capital city and business center of Thailand.Indeed,anecdotal evidence tends to suggest active speculative activities in both stock and housing markets during the years (Sheng and Kirinpanu,2000).However,both asset markets succumbed to the Asian crisis with the stock market nosedived first and followed by the property market–suggesting possible contribution of the stock market to housing market decline.In the lead-up years to the crisis,the Thai stock market index was well over1,000points.During the crisis episode,it drastically dropped and reached its lowest point in the third quarter of1998. The market fluctuated around380points until the end of2003.Since then it went up and stabilized around600-750points until the end of2006.Meanwhile,the property prices have again been on the rise especially since2001(Sriphayakand and Vongsinsirikul, 2007).In light of the above,Thailand provides an interesting case study to discern causal relations between house prices and stock prices.Moreover,the focus on Thailand brings up an emerging market perspective to the issue,adding to existing studies on predominantly advanced markets.
House price-stock price relations
71
In line with recent studies,we make use of standard V AR approach,outlined in the next section,to assess the causal relations between the stock and property markets.Then,section 3presents data preliminaries.The results on dynamic interactions among the variables are detailed in section 4.The final section,section 5,summarizes the main findings and provides concluding remarks.
2.Empirical approach
The focal variables in the analysis are house prices and stock prices.However,employing bivariate analysis as in some previous studies may not be https://www.doczj.com/doc/177320228.html,ly,as Quan and Titman (1999)demonstrate,their relations may be spurious reflecting responses to common factors.This means that other control variables need to be added.Although the approach we employ permits the inclusion of any macroeconomic variables deemed relevant in influencing or being influenced by stock and house prices,its implementation is constrained by available sample size.Accordingly,following the lead of existing literature,we include real gross domestic products (GDP)into the analysis.While,Quan and Titman (1999)have incorporated real GDP as a business cycle variable,the importance of real GDP in affecting housing supply and demand as well as in influencing stock prices through asset pricing valuation is well noted.Apart from real GDP,we also include the aggregate price level in the analysis.This is in line with the inflation hedging literature and,again,can be argued through the asset valuation model[1].
Thus,our framework consists of four variables –house prices,stock prices,real GDP and aggregate price level.To begin,we specify a linear equation that relates these variables as:
HP t ? 0t 1S t t 2Y t t 3P t tu t
e1T
where HP is house prices,S is stock prices,Y is real GDP and P is consumer prices.Equation (1)can be viewed as a long-run equation that ties the variables together.Before considering the appropriate econometric model to disentangle short-run dynamic interactions among them,statistical verification for the validity of (1)as a long-run equation is needed.Accordingly,we follow the standard procedures of time-series analysis.These include unit root tests,cointegration test,causality test and finally analyses of impulse-response functions (IRF)and variance decompositions (VDC)simulated from estimated V AR models.
We apply the commonly used augmented Dickey-Fuller (ADF)and Phillips-Perron (PP)tests for unit root.Then,the V AR-based approach of Johansen (1988)and Johansen and Juselius (1990)–henceforth JJ test –is employed to test for cointegration or long-run relationship among the variables[2].It needs to be noted that implementing the JJ test requires pre-specifying the V AR lag order.Following the suggestion by Hall (1989)and Johansen (1992),we specify the lag order such that the error terms are serially uncorrelated.Based on the unit root and cointegration tests,we then proceed to Granger causality test.More specifically,with the finding of cointegration among the variables,the dynamics of housing prices is framed in an error-correction form as:áHP t ? t u t à1t
X k i ?1
i áHP t ài t
X k i ?1
i áS t ài t
X k i ?1
’i áY t ài t
X k i ?1
i áP t ài t"t
e2T
IJHMA 3,1
72whereáis the first difference operator,u is the error-correction term from(1) measuring the deviations of the HP from its long-run value,and all variables are as defined above.
Specification(2)conveniently combines the long-run information in the data as well as their short-run https://www.doczj.com/doc/177320228.html,ly,it incorporates the long-run information in the data through the error-correction term,representing deviations of the variables from their long-run values and,thus,allows the house prices to adjust to the long-run equilibrium with the speed of adjustment, .Meanwhile,the lagged first-differenced right-hand-side variables reflect their dynamic short-run influences on house prices.The former is aptly termed as the long-run causality while the latter as the short-run causality in Granger sense among the variables.The presence of these causalities is respectively based on the significance of error-correction coefficient using standard t test and coefficients of the lagged first-differenced variable of interest using standard F test.
Note that the dynamics of stock prices and other variables in the model can be tested in a similar manner.From the tests,we can discern four alternative causal patterns between pairs of variables;more specifically to our interest is the causal pattern between house prices(HP)and stock prices(S).These are:
.unidirectional causality from S to HP,
.unidirectional causality from HP to S,
.feedback effect between S and HP,and
.S and HP are independent or the two markets are segmented.
The presence of the first causal pattern provides support for the wealth effect while that of the second is indicative of the credit-price effect.The feedback effect between the two is consistent with both effects and can be potential explanation of spiraling upturns of both prices.
As a further analysis,we also estimate V AR models and simulate IRF and VDC. These steps make causality analysis more complete and add robustness to the results as the Granger causality test is simply within-sample causality test while simulated IRF and VDC can be viewed as out-of-sample causality tests.The IRF traces the temporal responses of a variable to a one standard deviation shock in other variables. From the functions,we can assess the direction,magnitude,and persistence of the responses of say house prices to stock price shocks.Meanwhile,the VDC indicates the percentage of a variable’s forecast error variance attributable to its own shocks and shocks in each variable in the model.Thus,it serves as natural measures for the relative strengths of various shocks in accounting for variation in a variable of interest. In other words,central to our theme,we can measure the relative importance of stock price shock in explaining variations in house prices and vice versa.
3.Data preliminaries
The data employed are quarterly from1995.Q1to2006.Q4,the span of which is dictated by data availability.Four different housing indexes are used,namely,semi-detached houses(SD),semi-detached houses with land(SDL),townhouses(TH),and townhouses with land(THL)[3].We employ these indexes alternatively in our analysis, forming four separate V AR systems.For stock prices,we use the Stock Exchange of Thailand Composite Index(S).Meanwhile,real gross domestic product and consumer price index respectively represent real income(Y)and the general price level(P).The GDP figures are in real terms at the source;they are computed by deflating nominal
House price-stock price
relations
73
GDP with the GDP deflator.For the price level,we settle with the consumer price index since disaggregated price indexes such as the investment price series are not available.All data except the consumer price index are sourced from Bank of Thailand website (www.bot.or.th).The consumer price index is taken from the International Financial Statistics (CD-ROM).These data are expressed in natural logarithm.Examining the data series,we note their abrupt changes during the Asian crisis.In addition,seasonal patterns in real GDP are evident.Accordingly,we incorporate the Asian crisis and seasonal dummies in the ensuing analyses.
Table I presents descriptive statistics of stock and housing returns and autocorrelations of stock prices and house prices while Table II provides their cross-correlations.Note that,while the stock market records an average negative return over the sample period,all housing returns are positive.Moreover,the stock market return is relatively more volatile.Among the housing indexes,the semi-detached houses have the highest return while the semi-detached houses with land have the highest volatility.The autocorrelations indicate that shocks to these prices tend to be persistent especially for semi-detached house prices,townhouse prices and stock prices.Moreover,the cross-correlations tend to suggest the leading role of stock prices for especially semi-detached houses with land and townhouses with land prices by roughly four quarters.However,for semi-detached and townhouse price indexes,their lead-lag relationship with stock prices is not clear-cut.While these simple statistics is suggestive of the wealth effect,to be concrete,formal analyses of the data are required.To proceed further,we first subject each time series to ADF and PP unit root tests,which are presented in Table III.The tests uniformly suggest that the variables are non-stationary in level but become stationary when first-differenced.In other words,they are integrated of order one,or I (1).Having verified their stochastic property,we apply the JJ cointegration test to validate the presence of Equation (1)as a long-run equation.As noted earlier,the V AR lag order is determined such that the error-terms are serially uncorrelated.We find lag 4to be sufficient for the semi-detached house system and lag 2for others[4].The results from the JJ test,reported in Table IV ,suggest the presence of a long-run relationship among the variables in all systems,regardless of the housing prices used.It should be noted that,this result is not overturned even if we adjust the test statistics for
Table I.
Descriptive statistics and
autocorrelations
Lag Stock prices (S )
Semi-detached house (SD)Semi-detached house-land (SDL)Townhouse (TH)Townhouse –land (THL)
(a)Descriptive statistics of logarithmic-difference in prices Mean à0.01240.00920.00490.0039
0.0031Std.Dev.0.18470.03430.05260.02980.0473Max 0.39360.07610.19940.08150.1804Min à0.5408à0.1101à0.2326à0.09140.0473(b)Autocorrelation of price series 10.8990.9110.7090.8730.65620.8020.8280.6480.7950.59630.6700.7390.5430.6920.45340.5470.6280.4580.5740.40850.4250.5300.3110.4740.25760.2860.4290.2490.3930.17770.1840.3170.1320.3140.06780.0820.2100.034
0.230
à0.016
IJHMA
3,1
74
Table II.
Cross-correlations
House types
S and H(±i)SD SDL TH THL t60.10920.43960.18590.5250t50.10000.42790.21950.5302t40.08590.48760.21580.6064t30.06310.44870.18480.5602t20.06530.37910.18170.4735t10.06280.31700.16950.3951 00.06050.23730.16970.3017à10.08300.23270.19520.2929à20.11900.16420.20470.1920à30.15310.17300.23600.1941à40.18340.16060.24620.1641à50.21720.19070.26770.1867à60.23840.17400.28000.1586
Table III. Unit root tests
Level First difference Variables ADF PP ADF PP SDà1.377à1.493à6.567*à6.564* SDLà1.446à2.571à11.07*à12.06* THà1.096à2.042à10.16*à10.41* THLà1.457à2.219à11.10*à12.63* Sà1.527à1.528à8.001*à8.001* Yà2.052à0.986à3.374***à7.754* Pà20.54à2.557à5.394*à5.422* Notes:*,**,and***indicate significance at1per cent,5per cent,and10per cent,respectively
Table IV. Johansen-Juselius cointegration tests Test Null hypothesis
Statistics None At Most1At Most2At Most3 (a)Semi-detached house(Lag?4)
Trace78.71033.6568.8500.195 Max Eigen45.05424.8068.6550.195 (b)Semi-detached house including land(Lag?2)
Trace64.25634.93714.5810.378 Max Eigen29.32020.35614.2030.378 (c)Townhouse(Lag?2)
Trace58.72827.42111.937 1.288 Max Eigen31.30615.48410.649 1.288 (d)Townhouse including land(Lag?2)
Trace64.75833.34614.4690.190 Max31.41218.87814.2790.190 Critical values(5per cent)
Trace47.85629.79715.495 3.841 Max27.58421.13214.265 3.841
House price-stock price
relations
75
small sample size bias as suggested by Reinsel and Ahn (1992).Likewise,relying on the bounds testing procedure as suggested by Pesaran et al.(2001)yields similar conclusion[5].With these findings,we proceed next to investigating dynamic interactions among the variables,particularly between stock prices and house prices.
4.Dynamic interactions
This section reports results of the variables’dynamic interactions via Granger causality tests,IRF,and VDC.The Granger causality tests are presented in Table V .Meanwhile,Table VI summarizes the VDC and Figures 1-4plot the IRF.
Several interesting results emerge from Table V .More particularly,regardless of the house price indexes,we note significant adjustments of the house prices towards the long-run equilibrium with the adjustment speed ranges from 0.26for townhouses to 0.87for semi-detached houses.Thus,given deviations from the long-run equilibrium,house prices seem to adjust pretty fast.We also document evidence that consumer prices bear the burden of making adjustments towards the long-run equilibrium.By contrast,with insignificant error-correction coefficients,stock prices and real GDP are weakly exogenous in the system.These results further reaffirm the presence of wealth effect in the Thai market since in the long-run house prices make adjustments once there are shocks in other variables in the system including stock prices.In other words,there is a unidirectional long-run causality from stock prices and other variables to house prices.
Table V .
Granger causality tests
2-statistics of lagged first-differenced terms Dep.variables áSD áS áY áP
Coeff.of ECT (t-stat)
(a)Semi-detached houses áSD –13.889[0.008]16.367[0.003]
8.267[0.082]à0.875*(5.303)áS 1.582[0.812]– 2.746[0.601]
3.332[0.504]0.359(0.284)áY 1.677[0.795]11.895[0.018]–
11.264[0.024]
0.007(0.087)áP 3.826[0.430] 1.180[0.881]11.453[0.022]–0.043(0.933)(b)Semi-detached (including land)houses
áSDL áS
áY áP
áSDL – 3.175[0.204] 2.223[0.329] 1.825[0.401]à0.619*[4.073]áS 1.826[0.401]–0.491[0.782]
5.452[0.066]0.869[1.402]áY 3.791[0.150] 1.734[0.420]–
4.192[0.123]
à0.085[1.574]áP 7.759[0.021] 1.099[0.577]0.825[0.662]–à0.083*[3.192](c)Townhouses
áTH áS
áY áP
áTH –0.332[0.847] 1.504[0.471] 1.124[0.570]à0.263*(2.475)áS 0.620[0.733]–0.955[0.620]
1.338[0.512]0.689(0.895)áY
2.199[0.333]7.169[0.028]–
3.313[0.191]
à0.049(0.049)áP 0.246[0.884] 1.727[0.422]0.944[0.624]–à0.103*(3.462)(d)Townhouses (including land)
áTHL áS
áY áP
áTHL – 3.399[0.183] 2.395[0.302] 1.132[0.568]à0.595*(3.880)áS 3.476[0.176]–0.713[.700] 5.893[0.052]0.912(1.370)áY 3.019[0.221] 1.203[0.548]–
4.531[0.104]
à0.104(1.733)áP 5.316[0.071] 1.271[0.530]
0.751[0.687]
–
à0.097*(3.398)
Notes:Numbers in squared brackets are p -values while those in parentheses are absolute values of t-ratios;*significance at 5per cent level
IJHMA 3,1
76
Note that the short-run causality results are not so revealing since they tend to depend on which house price index is https://www.doczj.com/doc/177320228.html,ly,short-run interactions are prominent only when semi-detached house price index is employed,in which case we note significant causal influences of all variables to house price variations.In other systems,there seems to be very limited cases of dynamic short-run interactions among them.Still,notable among these limited cases is the causal nexus from stock prices to real output in the systems using semi-detached house prices and townhouse prices.This means that stock prices may also have influenced house prices indirectly through aggregate output,which strengthens the case for the wealth effect.
Figures 1-4report the IRF simulated from estimated levels V AR models with the lag chosen based on the requirement that the error-terms are serially uncorrelated.Note that,in simulating IRF,the responses of a variable to innovations in other variables cannot be
Table VI.
Variance decompositions
Explained by innovations in Horizons HP
S Y
P
(a)Variance decompositions of house prices System I:SD,S.Y,P 444.23535.665
18.043 2.056830.80925.66241.215 2.3141226.20622.71446.838 4.242System II:SDL,S.Y,P 468.458 3.81813.70714.017845.47514.44329.85510.2271237.61017.32836.5768.486System III:TH,S.Y,P 474.3967.9038.8698.832852.48112.68028.688 6.1501239.57614.02541.666 4.732System IV:THL,S.Y,P 475.238 3.33510.77010.657850.59713.60227.5078.2931243.568
15.798
33.423
7.211
(b)Variance decompositions of stock prices System I:SD,S.Y,P 40.33844.35144.10811.2038 1.94642.59343.60811.85312 4.20439.94941.70514.142System II:SDL,S.Y,P 4 4.02752.55839.621 3.7948 3.32050.57542.603 3.50212 3.29149.19843.880 3.631System III:TH,S.Y,P 4 3.62764.75424.966 6.65387.53757.75527.0737.634127.34653.99530.4618.198System III:THL,S.Y,P 4 5.99151.10539.912 2.9928 5.20449.24942.849 2.69712 5.203
47.972
44.025
2.800
House price-stock price relations
77
adequately represented in the presence of contemporaneous correlations among the shocks (Lutkepohl,1991).This makes the normally-used Cholesky decomposition as suggested by Sims (1980)to identify the shocks through orthoganization process inappropriate since it requires a pre-specified causal ordering among the variables and accordingly,initial impact of shocks on the variables ordered first cannot be detected.To circumvent the ordering issue,we report the generalized IRF developed by Koop et al.(1996)and Pesaran and Shin (1998).In this way,the historical patterns of correlations among different shocks are fully incorporated making the functions to be invariant to alternative orderings.Making the case for their studies on equity markets,Ewing et al.(2003)note the useful application of the approach in allowing quick price transmissions and contemporaneous adjustments of included variables.We believe that the approach is also appropriate for looking at dynamic interactions among asset prices which,in our case between house and stock prices.
The case for the wealth effect is further reaffirmed by the responses of house prices to shocks in the stock market,as presented in Figure 1.At the same time,the responses of stock prices to house prices are basically absent (Figure 2).As may be noted from Figure 1,a one standard deviation shock in stock prices exerts positive responses from house prices.While the semi-detached houses tend to respond quite immediately,the responses of semi-detached houses with land prices and townhouses with land prices turn significant after four quarters to roughly seven quarters.However,the
responses
Figure 1.
Responses of house prices
to shocks
IJHMA 3,1
78
of stock prices to impulses in house prices are insignificant at all horizons.It is also pleased to note from Figures 1and 2that real output shocks lead to positive responses from both house prices and stock prices.In this case,the following notes are in order.First,despite the common responses of both asset prices to output shocks,the house prices still independently respond to stock market fluctuations.Note that the stock prices tend to respond immediately to innovations in output.Meanwhile,the significant responses of house prices come with lags.Accordingly,it can be said that real output shocks may be transmitted to the housing market via the stock market,in addition to its independent influences on house prices.
One notable point that needs to be raised from Figure 1is that the semi-detached houses with land prices and townhouses with land prices tend to respond in a similar manner to stock market innovations.Meanwhile,no clear-cut evidence is observed in the cases of housing prices without land,i.e.SD and TH .This is interesting since it may be argued that the inclusion of land prices in the calculation of housing price index tends to matter in the relations between housing prices and other variables.Whether the land prices account for these results,however,need to be verified.With the availability of land price index for Thailand,we estimate a similar V AR system for the land price index and simulate corresponding IRF[6].Interestingly,the results show that the responses of
land
Figure 2.
Responses of stock prices to shocks
House price-stock price relations
79
prices to stock price shocks are not significant at all horizons while their responses to real GDP and consumer price shocks are significant.Thus,the inclusion of land prices alone may not account for the results we obtain.Perhaps,the positive wealth effect of stock market increase leads to portfolio rebalancing towards houses with lands.This explanation,however,is tentative and requires further verification.
From Figure 3,we may observe that real output does respond positively and significantly to stock price shocks,suggesting feedback effect between real output and stock prices.Meanwhile,its reaction to house price innovations is virtually absent.From Figure 4,the role of consumer price shocks to house and stock prices cannot be observed.However,at long horizons,stock price shocks do lead to positive reaction from consumer prices.These results indicate potential wealth effects of stock price fluctuations by increasing possibly the economy’s aggregate demand.However,perhaps consumer prices tend to be sticky or the time period covered is mostly characterized by the availability of idle resources,the increase in aggregate demand is translated first to increasing in real output and followed by increasing in consumer prices.This story tends to be in line with the responses of consumer prices to output shocks in Figure https://www.doczj.com/doc/177320228.html,ly,in response to increase in output,consumer prices increase.The corresponding VDC yield similar results on the dynamic behavior of house prices and stock prices (Table VI)[7].As can be observed from panel (a)of the table,
at
Figure 3.
Responses of real GDP
to shocks
IJHMA 3,1
80
12-quarter horizon,the proportion of house price forecast error variances attributable to stock market shocks range from 14.0percent for townhouse prices to 22.7percent for semi-detached house prices.By contrast,from panel (b),house price variations explain no more than 8percent of stock price forecast error variances at all horizons.Likewise,we note the relative importance of real output shocks in accounting for fluctuations in both house prices and stock prices.Finally,as in the IRF,the role of consumer prices in explaining variations in both prices are rather limited.These VDC further substantiate our conclusions made earlier.
To sum up,the evidence we obtained for the wealth effect seems unequivocal.Stock price fluctuations affect house prices not only directly but also indirectly through real output.We believe that the decline in the market observed during the crisis may have contributed directly to the observed bursting bubble of Thai housing market during the time.Ensuing output contraction,then,made things worse.Stated differently,our analysis places the stock market at the central stage for stability in the housing market as well as in the goods market for the Thai economy.
5.Conclusion
The paper assesses the causal relations between house prices and stock prices for the case of Thailand.Apart from the two asset prices,it also includes real output
and
Figure 4.
Responses of consumer prices to shocks
House price-stock price relations
81
consumer prices in the analysis.Having documented the presence of their long-run relation via JJ cointegration tests,vector error-correction models (VECM)are estimated to discern dynamic causal interactions among the variables.Then,for completeness,IRF and VDC are also generated.The results we obtained unequivocally suggest the presence of the wealth effect in the relation between house prices and stock https://www.doczj.com/doc/177320228.html,ly,in a vector error-correction setting,house prices seem to bear the burden of making adjustment towards the long-run relationship that ties the variables together.Additionally,house prices also react positively to innovations in stock prices as observed from the simulated IRF.Then,the VDC indicate quite sizable fraction of the house price variations attributable to stock price shocks.Finally,stock prices may also indirectly contribute to house price fluctuations via the real output.
In a nutshell,stability in the stock market seems to be critical for stability in other markets such as housing and real markets for the Thai economy.Indeed,we believe that the stock market decline during the 1997/1998Asian crisis may have contributed to housing market decline,which was worsened by ensuing contraction in output.From a policy point of view,Thai policy makers should be cautious in implementing its policies so that not to generate instability in the stock market.At the same time,measures to further develop its stock market in a way to promote stability are much needed.Otherwise,the negative wealth effect from stock market decline may have far-reaching implications.
Notes
1.We wish to incorporate interest rates in the analysis.However,the interest rates in
Thailand have been relatively flat and exhibit little variations since the crisis,which forms substantial part of our sample.Thus,they are not included.
2.These tests are now standard in time-series literature and,accordingly,are not explained
here.Refer to Johansen (1988)and Johansen and Juselius (1990)for details.
3.We have no information as to the proportion of semi-detached houses and townhouses in
Thailand.We also do not have information on regional housing prices.The availability of regional data would surely make the analysis more complete and insightful.As normal for any developing market,data unavailability remains an important caveat.4.As appropriately note by a referee,lag 4seems to be quite large for our small sample.
We also test the cointegration of semi-detached house system using lag 2and reach similar conclusion.
5.The ARDL test results are not reported to conserve space but available upon request
from the author.
6.We do not report the results here to conserve space and due to our focus on housing
prices.However,we believe that the exercise can provide insight on the issue raised.7.To conserve space,we report only the variance decompositions of house and stock
prices,which are the main theme of the paper.Variance decompositions for other variables,however,are available upon request.References
Chen,N.-K.(2001),‘‘Asset price fluctuations in Taiwan:evidence from stock and real estate prices
1973to 1992’’,Journal of Asian Economics ,Vol.12,pp.215-35.
Eichholtz,P .and Hartzell,D.(1996),‘‘Property shares,appraisals and the stock market:an
international perspective’’,Journal of Real Estate Finance and Economics ,Vol.12,pp.163-78.Ewing,B.T.,Forbes,S.M.and Payne,J.E.(2003),‘‘The effects of macroeconomic shocks on sector-sector specific returns’’,Applied Economics ,Vol.35No.2,pp.201-7.
IJHMA 3,1
82Hall,S.G.(1989),‘‘Maximum likelihood estimation of cointegration vectors:an example of the Johansen procedure’’,Oxford Bulletin of Economics and Statistics,Vol.51,pp.213-8. Hartzell,D.(1986),‘‘Real estate in the portfolio’’,in Fabozzi,F.J.(Ed.),The Institutional Investor: Focus on Investment Management,Balliger,Cambridge,MA.
Ibbotson,R.and Siegel,L.(1984),‘‘Real estate returns:a comparison with other investments’’, AREUEA Journal,Vol.12,pp.219-41.
Johansen,S.(1988),‘‘Statistical analysis of cointegration vectors’’,Journal of Economic Dynamics and Control,Vol.12,pp.231-54.
Johansen,S.(1992),‘‘Testing weak exogeneity and the order of cointegration in UK money demand data’’,Journal of Policy Modeling,Vol.14,pp.313-34.
Johansen,S.and Juselius,K.(1990),‘‘Maximum likelihood estimation and inference on cointegration:with applications to the demand for money’’,Oxford Bulletin of Economics and Statistics,Vol.52No.2,pp.169-210.
Kakes,J.and Van Den End,J.W.(2004),‘‘Do stock prices affect house prices?Evidence for the Netherlands’’,Applied Economics Letters,Vol.11,pp.741-4.
Kapopoulos,P.and Siokis,F.(2005),‘‘Stock and real estate prices in Greece:Wealth versus‘credit-price’effect’’,Applied Economics Letters,Vol.12,pp.125-8.
Koop,G.,Pesaran,M.H.and Potter,S.M.(1996),‘‘Impulse response analysis in nonlinear multivariate models’’,Journal of Econometrics,Vol.74,pp.119-47.
Lutkepohl,H.(1991),Introduction to Multiple Time Series Analysis,Springer Verlag,Berlin. Pesaran,M.H.and Shin,Y.(1998),‘‘Generalized impulse response analysis in linear multivariate models’’,Economics Letters,Vol.58,pp.17-29.
Pesaran,M.H.,Shin,Y.and Smith,R.J.(2001),‘‘Bounds testing approaches to the analysis of level relationship’’,Journal of Applied Econometrics,Vol.16,pp.289-326.
Quan,D.C.and Titman,S.(1999),‘‘Do real estate prices and stock prices move together?An international analysis’’,Real Estate Economics,Vol.27No.2,pp.183-207.
Reinsel,G.C.and Ahn,S.K.(1992),‘‘Vector autoregressive models with unit roots and reduced rank structure:estimation,likelihood ratio test,and forecasting’’,Journal of Time Series Analysis,Vol.13,pp.353-75.
Sheng,Y.K.and Kirinpanu,S.(2000),‘‘Once only the sky was the limit:Bangkok’s housing boom and the financial crisis in Thailand’’,Housing Studies,Vol.15No.1,pp.11-27.
Sim,S.-H.and Chang,B.-K.(2006),‘‘Stock and real estate markets in Korea:wealth or credit-price effect’’,Journal of Economic Research,Vol.11,pp.99-122.
Sims,C.(1980),‘‘Comparison of interwar and postwar business cycles:monetarism reconsidered’’, American Economic Review:Papers and Proceedings,Vol.70,pp.250-7. Sriphayakand,A.and Vongsinsirikul,S.(2007),‘‘Asset prices and monetary policy transmission in Thailand’’,Bank of Thailand Discussion Paper,DP/01/2007.
Sutton,G.D.(2002),‘‘Explaining changes in house prices’’,BIS Quarterly Review,September,pp.46-55. Worzala,E.and Vandell,K.(1993),‘‘International direct real estate investments as alternative portfolio assets for institutional investors:an evaluation’’,paper presented at the1993 AREUEA Meetings,Anaheim,CA.
Corresponding author
Mansor H.Ibrahim can be contacted at:mansorhi@https://www.doczj.com/doc/177320228.html,.my
To purchase reprints of this article please e-mail:reprints@https://www.doczj.com/doc/177320228.html,
Or visit our web site for further details:https://www.doczj.com/doc/177320228.html,/reprints