國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
31
Chapter 5. Empirical Result
The VECM results are showed as following tables. First, I use all sample data
including presale and new house to test the price lead-lag relationship with Sinyi index.
The result represented on the Table 4. First, we can find a significant long term
conintegration relationship between TAIALL and Sinyi index t value is -12.9023. It
means the Sinyi index will increase 0.7916 when TAIALL increases 1 unit in long term
equilibrium. Also, We also find correction term is significant which t value is 2.39920 in
D(TAIALL). When Sinyi index is high and deviated from long term equilibrium, the
TAIALL will adjust its price higher to the long term equilibrium relationship. The speed
of adjustment in TAIALL will be faster than Sinyi. Therefore, we can say the TAIALL
leads the resale market because developers have more information to forecast the price
trend of real estate market compared to resident buyer. The result coincides the
hypothesis one. In the short term perspectives, the 1st difference of TAIALL have a
negative price adjustment in Sinyi which t value is -2.2494. The resale price will rise
when the real estate marketing index goes down in short adjustment. The coefficient of
RATE is significantly negative, which is -0.039735. It means that when the rate is low,
the second hand house market price will be spurred by cheaper capital. Also, resale
‧
Table 4. VECM result of TAIALL and Sinyi Cointegrating Eq Coint Eq1
Sinyi(-1) 1
TAIALL(-1) -0.791622
(0.06136) [-12.9023]
C -0.151966
Error Correction: D(Sinyi) D(TAIALL)
CountEq1 -0.062082 0.327771***
t-value [-0.55648] [2.39920]
D(Sinyi (-1)) -0.013553 0.106169
t-value [-0.07004] [0.44806]
D(Sinyi (-2)) 0.015262 0.248309
t-value [0.09506] [1.26300]
D(TAIALL(-1)) -0.276585** -0.019604
t-value [-2.24494] [-0.12994]
D(TAIALL(-2)) -0.057581 0.115486
t-value [-0.45585] [0.74660]
C 1.510210 -0.139196
t-value [1.45451] [-0.10948]
LOGGDPPER -0.233386** 0.009275
t-value [-2.42695] [0.07876]
LOGSTOCK 0.156441*** 0.002115
t-value [3.55086] [0.03920]
RATE -0.039735*** 0.013420
t-value [-3.05962] [0.84384]
R-squared 0.547204 0.536782
Adj. R-squared 0.434005 0.420977
F-statistic 4.834005 0.038226
Log likelihood 93.17542 84.86882
Akaike AIC -4.106118 -3.700918
Schwarz SC -3.729968 -3.324768
*,**, and *** Singnificant at 10%, 5% and 1% levels, respectively
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
33
house is influenced significantly by stock market. The Sinyi price index movement
aligns with stock market, it may be because people have more money to spend driving
house price increased. While, the GDPPER variable is -0.233386, it means when GDP
per capita goes down, the house price goes up. It obeys normal understanding which
price will be higher when people become rich. There may be a possible explanation that
house price deviated from fundamental demand.
Then, we separated all samples into two groups, presale and new house, to isolate
two different markets features. Table 5 showed the result of presale house. We can see
the same result that there is a significant cointegration relationship between Sinyi and
PRESALE which t value is -10.5787. According to the correction term, the presale
market correction the price discrepancies when deviating the long equilibrium. We can
find a significant lead-lag relationship that the speed of price adjustment of future
market leads spot market. It aligns with our hypothesis one which the market
participants with more information will reflect the price fast. Also, we can observe the
same result in short term perspectives. There is a significantly negative adjustment in
Sinyi index by D(PRESALE(-1)) which t value is -2.24598. And presale market is
influenced significantly by its first and second order lag term which net effect positive.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
34
It means there is a short term momentum that the high price of presale house will drive
the price higher in the next period. Also, we can explain all of developers act at the
same direction. The stock and rate variables represented the same pattern in explaining
resale market as previous result. The stock variable is significantly positive and the rate
variable significantly negative. The coefficient of gdpper is -0.229821, also significant.
In new house, we can see the table 6 that we can find that there is also a
cointegration relationship between Sinyi and NEW. As a result, the new house market
leads the resale market too in the long term perspectives. The developers take the price
advantage in pricing. If the price deviated form long term equilibrium then developer
will adjust the price of new house first. The result also met our hypothesis one. None of
short term adjusted variables are significant. There is no short term price adjustment
between Sinyi and NEW. The possible explanation is that the new house and second
house are largely homogeneous. The price movement in new house and resale market
follow the same pattern. The feature causes an insignificant short term
‧
Table 5.VECM result of Presale House and Sinyi Cointegrating Eq Coint Eq1
Sinyi(-1) 1
PRESALE (-1) -0.77938
(0.07367) [-10.5787]
C -0.15736
Error Correction: D(Sinyi) D(PRESALE)
CountEq1 -0.099314 0.308890**
t-value [-1.10391] [2.24977]
D(Sinyi(-1)) 0.090895 -0.055457
t-value [0.49863] [1.21620]
D(Sinyi(-2)) -0.069824 0.294760
t-value [-0.43967] [1.21620]
D(PRESALE (-1)) -0.198309** -0.238459
t-value [-2.24598] [-1.76966]
D(PRESALE (-2)) 0.007509 0.344375**
t-value [0.07059] [2.12139]
C 1.516119 -0.226851
t-value [1.42277] [-0.13949]
LOGGDPPER -0.229821** -0.006280
t-value [-2.37026] [-0.04244]
LOGSTOCK 0.150291*** 03031919
t-value [3.44675] [0.47966]
RATE -0.039577*** 0.018072
t-value [-3.05114] [0.91292]
R-squared 0.547617 0.581127
Adj. R-squared 0.434521 0.476408
F-statistic 4.842058 5.549425
Log likelihood 93.19410 75.86238
Akaike AIC -4.107029 -3.261580
Schwarz SC -3.730879 -2.885430
*,**, and *** Singnificant at 10%, 5% and 1% levels, respectively
‧
Table 6. VECM result of New House and Sinyi Cointegrating Eq Coint Eq1
Sinyi(-1) 1
NEW(-1) -0.923576
S.D (0.07294)
T-value [-12.6616]
C -0.041205
Error Correction: D(Sinyi) D(NEW)
CountEq1 -0.034365 0.590356**
t-value [-0.32036] [2.99454]
D(Sinyi(-1)) -0.026293 -0.125411
t-value [-0.13435] [-0.34868]
D(Sinyi(-2)) -0.033351 -0.067060
t-value [-0.19252] [-0.21063]
D(NEW(-1)) -0.161259 0.259927
t-value [-1.57245] [1.37912]
D(NEW(-2)) -0.024645 0.012249
t-value [-0.20276] [0.05484]
C 2.138945** 0.875439
t-value [2.14694] [0.47813]
LOGGDPPER -0.272209** -0.113549
t-value [-2.74442] [0.56795]
LOGSTOCK 0.138883** 0.050506
t-value [2.87022] [0.56795]
RATE -0.047793*** 0.022021
t-value [-3.13798] [0.78673]
R-squared 0.511136 0.409369
Adj. R-squared 0.388920 0.261711
F-statistic 4.182230 2.772414
Log likelihood 91.60422 66.65254
Akaike AIC -4.029474 -2.812319
Schwarz SC -3.653324 -2.436169
*,**, and *** Singnificant at 10%, 5% and 1% levels, respectively
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
37
relationship between two markets. The stock and rate variables showed consistent
significantly consequence with pervious results, which coefficient are 0.138883,
-0.047793. And the variable of GDPPER is -0.272209, also significant.
After separating all samples into two different markets, We can find the future
market still leads the spot market at each group. In the presales market, there is a short
term momentum of presale market. However, the short term relationship of price
adjustment was disappeared in new house sample because its homogenous features.
Here, I also test the price relationship between new house and presale house. In
both of markets, Developers decide the listing price. The result represented on table 7.
There is also a significant cointegration between new house and presale. However, in
the table 7, we can observe the correction term is significantly negative in the new
house market. Therefore, as developers’ view, when the price of new house is high and
away from long run equilibrium, the developers will revise the price of new house down
rather than presale because the carrying cost of new house is high and it will tight
developers’ capital. So, we can say the new house market will lead presale market when
the price away from long run equilibrium. The result aligns our hypothesis two which
participants with higher carrying cost will reflect price fast. Also, the new house price
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
38
index is influenced positive by its lag term which t value is 1.78431. There is a short
term momentum like presale market. And the stock positively explained the spot
market.
‧
Table 7. VECM result of Presale and New House Cointegrating Eq Coint Eq1
NEW(-1) 1
PRESALE(-1) -0.784778
(0.07635) [-10.2781]
C -0.208894
Error Correction: D(NEW) D(PRESALE)
CountEq1 -0.491791** 0.053432
t-value [-2.19534] [0.29572]
D(NEW(-1)) 0.408789 0.197939
t-value [1.78431]* [1.07117]
D(NEW(-2)) -0.044788 0.096877
t-value [-0.15151] [0.40631]
D(PRESALE(-1)) -0.085683 -0.300616
t-value [-0.36927] [-1.60628]
D(PRESALE(-2)) -0.041414 0.331329*
t-value [-0.18291] [1.81430]
C -2.635695 -1.339638
t-value [-1.01155] [-0.63744]
LOGGDPPER 0.11309 0.049105
t-value [0.49324] [0.26553]
LOGSTOCK 0.150215 0.085494
t-value [1.77003] [1.24898]
RATE -0.001087 0.011905
t-value [-0.03574] [0.48537]
R-squared 0.274264 0.475258
Adj. R-squared 0.092830 0.344072
F-statistic 1.511645 3.622790
Log likelihood 62.42966 71.24294
Akaike AIC -2.606325 -3.036241
Schwarz SC -2.230175 -2.66091
*,**, and *** Singnificant at 10%, 5% and 1% levels, respectively
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
40