Professional Documents
Culture Documents
2007
(hedonic equation)
(mass appraisal)
ABSTRACT
Estimating a housing price precisely is always an important issue both in real estate research
and in practice by financial institutions. According to the Basel Accord, financial institutions have
to review and to reassess their mortgage assets for a certain period of time to maintain asset quality
and control risk. However, since it is costly to reassess a real estate property in the traditional way, it
will be very troublesome and costly to reassess thousands of real estate mortgages for each institution
from time to time. Therefore, constructing a mass appraisal model to reassess a large quantity of real
estate mortgages has become a necessity for most financial institutions.
In this paper we first build a mass appraisal model using the hedonic equation model,
including housing characteristics, neighborhood attributes, and macroeconomic variables, and
then apply a dataset with more than 50,000 real estate mortgages to construct a housing price
equation. According to several statistics, including R-squared, the mean absolute percentage
error (MAPE), and the hit rate, we find that the estimation results of our pricing model are quite
satisfactory. This implies that mass appraisal could be a good model to apply in Taiwan.
(2007422007522200712)
*
64
Professor, Department of Economics, National Chengchi University, Taipei, Taiwan, Republic of China.
E-mail: nccut001@nccu.edu.tw
** 64
Ph.D. Student, Department of Land Economics National Chengchi University, Taipei, Taiwan, Republic of China.
()(1)
()
()
(2)
(single house)
()
(parametric method)
(3)
Lockword
& Ruthenford(1996)(instrumental method)Pace et al.(1998)
(grid adjustment approach)Schuiz & Werwatz(2004)
(semi-parametric model)Fotheringham et al. (1998)Leung et al.(2000a,
2000b)(geographically weishted regression)Sing et al.(2001)
(fuzzy discount cash ow analysis)Francke & Vos(2004)Pace et al.(2001a,
2001b)Schuiz & Werwatz(2004)(Kalman filter method)
McCluskey(1997)(neural network model)
James(1997)(artifical intelligence model)
Rosen(1974)(characteristics)
(1992)(1995)
(1989)Goodman(1978)Goodamn & Kawai(1982)
Rosen
Max
s.t.
YPXXPH(Z1,Z2,...Zk)H........................................................................................... (2)
UHXPXPH
Zi(i1,...,k)KY
LagrangeLagrange
LU(X,H,Z1,Z2,...Zk)[YPXXPH(Z1,Z2,...Zk)H] ............................................................ (3)
(shadow price)
......................................................................................................................... (5)
1/
(reduced form)
PHf(H,Z1,Z2,...Zk) ................................................................................................................ (6)
(6)
..................................................... (7)
Pxidjlkdjlk
1.
(1986)(1992)
2.
(1999)(1989)(1988)
(1993)(1993)
(1992)(1994)
3. (1986)(1988)
4.
(1994)
(1996)(1995)
1.
(outlier)
2. (heterogeneity)
3.
(mean absolute percentage error,
MAPE)(hit rate)(MSE)(AS ratio)
(4)
2002220045
(5)
1()
()
()
(1997)(2007)
(6)
(7)
(8)
(
9)116(10)5
()
5%95%(11)
(12)
(13)
4015
(14)
()
10
27%(23%)
1662%(57%)
()
()(a)
409.374
196.362
135.265
1058.120
(a)(b)
()
458.823
216.699
150
1170
()
39.535
15.569
7.01
175.180
()
9.231
6.766
1.500
85
()
11.722
9.012
0.010
65.310
1(c)
0.092
0.289
0
1
0.109
0.311
0
1
0.0002
0.016
0
1
0.228
0.420
0
1
()
5.082
4.343
-1
39
0.205
0.404
0
1
(d)
0.088
0.302
0
3
()
48971.210
48663.730
0
871451
()
11.771
11.309
0.050
309.480
(%)
0.147
0.127
0
0.738
0.709
0.454
0
1
0.008
0.091
0
1
0.075
0.263
0
1
0.181
0.385
0
1
0.885
0.319
0
1
0.105
0.306
0
1
0.002
0.047
0
1
0.008
0.090
0
1
0.833
0.373
0
1
0.161
0.368
0
1
0.833
0.373
0
1
0.011
0.106
0
1
0.088
0.283
0
1
0.025
0.155
0
1
()
706452.900
358812.500
107959
1224540
()
11870.150
6206.510
1232
24900
()
12283.410
6378.190
1250
25351
()
127.803
139.712
2
954
()
92.336
68.272
3
358
(%)
-0.343
2.830
-6.865
11.441
(%)
3.754
29.838
-61.941
96.651
(%)
3.749
0.796
2.470
4.963
()(%)
-0.406
6.871
-12.028
12.638
(%)
0.496
1.035
-1.476
2.337
(%)
3.964
1.991
-0.540
8.100
(a) 4081518637
(b) 0
(c)
(d) 30
10
3785
9.27
2405
12.90
3785
6578
16.12
3082
16.54
185
0.45
51
1790
11254
181
16
451
106
243
123
4.39
27.57
0.44
0.04
720
4256
43
4
1.10
350
0.60
75
0.26
0.30
77
139
3.86
22.84
0.27
0.23
0.02
1.88
0.41
12.9
6578
16.12
3082
16.54
366
0.90
94
0.50
1790
11254
467
106
243
123
1735
40815
4.25
100
371
903
18637
139
0.40
0.75
5.00
370
0.91
0.30
0.41
932
2.12
370
75
77
5.54
4850
1735
395
0.60
0.26
1.90
2261
0.75
5.00
1.01
354
12.88
932
411
1.14
22.84
2400
5.54
376
4256
3.86
14.23
2261
1.64
27.57
720
5806
12.88
670
4.39
0.40
2400
2058
2405
14.23
11.88
9.27
5806
4850
11.04
2.02
1.99
4.85
100
670
411
40815
11.88
1.64
4.25
1.01
0.91
100
2058
376
903
395
371
18637
11.04
2.02
4.85
2.12
1.99
100
()
10%
Belsley et al. (1980)DFFITS
(15)
i
i
s
i (i)
ih(i)hat matrixxi(XX)-1xipn
DFFITS2
11
DFFITS
DFFITS
10%out-sample 1
200222004590%
in-sample
out-sample 2(16)
yy()yy()
1
Y
5%10%20%
Hit Rate
100%
nN
Hit Rate
10%20%
MAPE
100%(yt0)
etytt
MAPE
outsample 1out-sample 2MAPEin-sample
MAPEout-sample 1(2)
in-sampleout-sample 1(2)MAPE(17)
()
out-sample 1out-sample 2
0.772
0.719
12
1()
(a)
(b)
(c)
(d)
t
1.19061
1.59
t
7.477**
4.49
0.772**
0.157**
-0.083**
-0.027**
-0.013**
0.040**
-0.011**
-0.249**
0.197**
-0.162**
0.057**
0.023**
-0.044**
0.263**
0.102**
-0.084**
-0.024*
0.025
0.077**
-0.080**
-0.027
-0.004
-1.033**
-0.184
0.132*
0.116
-0.081**
-0.232**
-0.124**
-0.273**
0.719**
0.176**
-0.068**
-0.032**
-0.018**
0.036**
0.014**
-0.183**
0.134**
-0.170**
0.055**
0.007
-0.056**
0.232**
0.091**
-0.207**
-0.034
-0.015
0.099**
-0.055
-0.283
-0.210
-0.438*
-0.390
-0.011
-0.008
-0.148**
-0.273
-0.183**
-0.405**
191.63
16.11
-73.37
-12.8
-23.42
5.58
-3.53
-4.24
24.54
-20.52
25.55
8.09
-14.62
119.49
37.27
-6.61
-1.74
0.24
3.77
-2.72
-0.26
-0.04
-3.83
-1.62
1.75
1.54
-3.18
-2.43
-4.69
-9.49
111.95
10.19
-49.83
-8.74
-23.38
3.03
2.59
-2.01
9.86
-13.21
15.52
1.46
-10.2
66.26
21.72
-10.17
-1.58
-0.06
2.68
-1.13
-0.91
-0.67
-1.81
-1.24
-0.05
-0.03
-3.39
-1.21
-4.06
-8.41
13
()
(e)
()
F-value
-0.332**
-0.335
-0.220
-0.296
-0.954**
0.346
-0.133
-0.566*
-0.366
-0.280
-0.194
-0.630**
-0.906**
-0.301
-0.318
-1.38
-1.46
-0.92
-1.54
-2.83
1.46
-0.45
-1.90
-1.30
-0.88
-0.65
-6.12
-4.12
-1.06
-1.02
-0.045
0.041**
-0.024
0.005
-0.009**
-0.027**
0.001**
0.009**
-2.2E-05
0.004**
0.0003
-0.83
2.50
-1.45
1.58
-3.18
-14.81
3.50
8.99
-0.65
7.20
0.77
40815
1125.70
0.8521
**5%*10%
(a) ()
(b)
(c)
(d)
(e)
t
-1.153**
-0.287
-1.363**
-1.105**
-1.141**
-1.231**
-0.952**
-0.435
-0.669*
-0.801**
-0.879**
-0.823**
-1.051**
-0.765**
-0.805**
-3.75
-1.01
-4.83
-3.33
-2.37
-2.60
-2.45
-1.16
-1.78
-2.18
-2.98
-6.51
-3.40
-2.54
-2.18
-0.436**
-3.71
-0.035
-1.27
0.073**
2.62
-0.001
-0.27
-0.002
-0.40
-0.032**
-10.22
0.0004
1.46
0.006**
3.68
1.67E-05
0.32
0.006**
7.16
-0.0007
-1.35
18637
453.40
0.8226
14
0.0230.007-0.083-0.068
-0.013-0.018
10.0400.036
0.1970.134-0.0110.014
(18)-0.162-0.170
(18)
(-0.044-0.056)(0.2630.232)
(-0.084-0.207)
(0.1020.091)
()
0.1570.176(
19)
(-0.027-0.032)(0.009
0.006)
(0.0040.006)
0.85210.8226
(20)
MAPEhit rateMAPE
MAPE15.09(out-sample 1)14.45
(out-sample 2)19.56
15
15.09
18.06
43.82
40.28
out-sample 1(%)
Hit Rate(20)
74.56
71.33
out-sample 2 (%)
Hit Rate(10)
14.45
15.64
33.73
36.09
62.28
64.39
19.56
25.49
MAPE(%)
out-sample 1 (%)
Hit Rate(10)
out-sample 1MAPE
out-sample 2(%)
Hit Rate(20)
out-sample 2MAPE
Rosen(1974)
MAPEhit rate
MAPEhit rate
0.8
16
(cross-sectional data)
17
1 2001(Basel Accord)
()
2 Shenkel(1991)
3 Geho(2003)250
1516.67
4 International Association of Assessing Officers(2003)
6
(2007)
7
8 228
10 ()()
(20032)6
11 Sample Seletion
Wooldridge(2002)
12
13
(variation)
14
15 Belsley et al.(1980)(1996)
DFFITS
16 2005678
17 MAPEDFFITS
18
10%MAPE
MAPE
2
MAPE
18 6
19 (2007)
20
0.7
0.8
21 Matysiak & Wang(1995)1030%20
70%
Ibrahim et al.(2005)
MAPE
19
1994 249-65
1988
1992 43(1)347-371
1993 121-45
1996 203-219
1995 (VAR)
71143-160
1996 1(1)29-49
1999 847-67
1993 71143-160
2007 2007
443-466
1989
1992
1989
1995
1994
20
1997 517-35
1986
1986
Belsley, D. A., E. Kuh & R. E. Welsch
1980 Regression Diagnostics. New York: John Wiley & Sons.
Fotheringham, A. S., C. Brunsdon & M. E. Charlton
1998 Geographically Weighted Regression: A Natural Evolution of the Expansion Method
for Spatial Data Analysis, Environment and Planning A. 30(11): 1905-1927.
21
22
()