# RECORD f8260edb97a0dbb7c5a255fca59c10da9cde8ca8
# TITLE: Real estate listings and their usefulness for hedonic regressions
# YEAR: 2020.0  VENUE: Empirical Economics  DOI: 10.1007/s00181-020-01992-3
# AUTHORS: Jens Kolbe; Rainer Schulz; Martin Wersing; A. Werwatz
# PAGES: 31  CANDIDATE TABLE PAGES (1-based): [10, 21]
# ABSTRACT: Real estate platforms provide a new source of data which has already been used as a substitute for transaction data in hedonic regression applications. This paper asks whether it is valid to do so in the established research areas of (1) willingness to pay estimation, (2) automated valuations, and (3) price index construction. It therefore compares listings and transaction data and regression results derived from them. We find that ask prices stochastically dominate sale prices, mainly because the composition of characteristics differs between the two data sets. But estimates of implicit prices also differ. As a result, willingness to pay estimates from listings data can be widely off when compared with estimates from transaction data. Listings data are not very useful to predict market values of individual houses either, as these predictions suffer from upward bias and large error variance. We find, however, that an ask price index complements a sale price index, as it is useful for nowcasting.


===== PAGE 1 (score 4, numeric-paren 0, table-words 0, var-words 4) =====

Empirical Economics (2021) 61:3239–3269
https://doi.org/10.1007/s00181-020-01992-3
Real estate listings and their usefulness for hedonic
regressions
Jens Kolbe1 · Rainer Schulz2 · Martin Wersing2 · Axel Werwatz1
Received: 14 January 2020 / Accepted: 15 November 2020 / Published online: 13 January 2021
© The Author(s) 2020
Abstract
Real estate platforms provide a new source of data which has already been used as
a substitute for transaction data in hedonic regression applications. This paper asks
whether it is valid to do so in the established research areas of (1) willingness to pay
estimation, (2) automated valuations, and (3) price index construction. It therefore
compares listings and transaction data and regression results derived from them. We
findthataskpricesstochasticallydominatesaleprices,mainlybecausethecomposition
of characteristics differs between the two data sets. But estimates of implicit prices
also differ. As a result, willingness to pay estimates from listings data can be widely
off when compared with estimates from transaction data. Listings data are not very
useful to predict market values of individual houses either, as these predictions suffer
from upward bias and large error variance. We find, however, that an ask price index
complements a sale price index, as it is useful for nowcasting.
Keywords Hedonic modelling · Nowcasting · Price prediction · Stochastic
dominance
Supplementary Information The online version contains supplementary material available at https://doi.
org/10.1007/s00181-020-01992-3.
B Axel Werwatz
axel.werwatz@tu-berlin.de
Jens Kolbe
j.kolbe@tu-berlin.de
Rainer Schulz
r.schulz@abdn.ac.uk
Martin Wersing
martin.wersing@abdn.ac.uk
1
Chair for Econometrics and Business Statistics, Institute for Economics and Business Law,
Technische Universität Berlin, Straße des 17. Juni 135, 10623 Berlin, Germany
2
University of Aberdeen Business School, Edward Wright Building, Dunbar Street, Aberdeen AB24
3QY, UK
123


===== PAGE 10 (score 15, numeric-paren 0, table-words 3, var-words 7) =====

3248
J. Kolbe et al.
Table 3 Stochastic dominance tests. The statistic d j,k ( j ̸= k) tests the null hypothesis that distribution j
dominates distribution k weakly. In Panel A, the variable tested for is −AGE, d j,k is the signed two sample
KS test statistic, defined in Eq. (1). The p values for the null are calculated as exp{−2(d j,k)2}, see Barrett
and Donald (2003, p. 78). In Panel B, d j,k is the KS maximal t-statistic as defined in Chernozhukov et al.
(2013, p. 2222). The p-values for the null are calculated as R−1 
r 1(d∗
j,k,r > d j,k), where d∗
j,k,r is the
r′th bootstrap test statistic, see Barrett and Donald (2003, p. 8). The number of bootstrap replications is
200
da,s
p value
ds,a
p-value
Panel A. Marginal distributions
Price
0.000
1.000
20.577
0.000
Age
0.286
0.849
24.689
0.000
Floor area
0.000
1.000
34.102
0.000
Plot area
0.027
0.999
9.007
0.000
Panel B. Price decomposition
Price
0.000
0.915
48.777
0.000
Characteristics
0.000
0.860
53.132
0.000
Implicit prices
1.076
0.705
10.978
0.000
Ask data
0
.02
.04
.06
.08
Density
0
20
40
60
80
100
Age
Sale data
0
.02
.04
.06
.08
Density
0
20
40
60
80
100
Age
0
.005
.01
Density
0
200
400
600
Floor area
Ask data
Sale data
0
.0005
.001
.0015
.002
Density
0
500
1000
1500
Plot area
Ask data
Sale data
Fig. 2 Distributions of house characteristics. Top panel shows histograms of building age for the observa-
tions in the ask and the sale data, respectively. Lower panel shows kernel density estimates of floor area
(left) and plot area (right) for the observations in the data. Bandwidths are chosen with Silverman’s (1986)
rule-of-thumb
123


===== PAGE 11 (score 20, numeric-paren 0, table-words 6, var-words 8) =====

Real estate listings and their usefulness...
3249
5800000
5810000
5820000
5830000
370000
380000
390000
400000
410000
Longitude
Latitude
0
5
10
15
20
Relative frequencies
 of listings [%]
Ask data
5800000
5810000
5820000
5830000
370000
380000
390000
400000
410000
Longitude
Latitude
0
5
10
15
20
Relative frequencies
 of transactions [%]
Sale data
Fig. 3 Spatial distribution of ask and sale observations. Shows the relative frequency of observations in the
ask and sales data across Berlin’s 12 administrative districts. Crosses give locations of the 59,502 (12,218)
individual observations in the ask (sale) data for which we have coordinates
Table 2 shows that the proportion of detached houses is higher in the ask than in the
sale data and the proportion of terraced houses is lower.12 This could explain why the
floor and plot area distributions of the ask data dominate those of the sale data. Figure 3
shows that the observations of the two data sets follow a similar spatial cluster and
are, at the level of Berlin’s administrative districts, nearly identical with a correlation
of ρ = 0.97. The observed characteristics can therefore not explain why the houses
in the ask data are dominantly younger.
In summary, the analysis of the core characteristics age, floor and plot area shows
strong dominance of observations in the ask over their counterparts in the sale data.
We explore next whether this is the sole reason for the strong dominance of ask over
sale prices.
2.4 Decomposition analysis
We use the inference procedure of Chernozhukov et al. (2013) to assess the importance
of house characteristics for the differential of ask and sale price distributions. The
procedure is similar in spirit to the decomposition at the means of Blinder (1973) and
Oaxaca (1973). The starting point is the following formula which—akin the law of
iterated expectations—relates unconditional and conditional distribution functions
Fj|k(p) ≡

Xk
FPj|X j (p|x)dFXk(x)
(2)
FPi|Xi (p|x) is the price distribution conditional on the vector x of characteristics and
FXi (x) is the distribution of these characteristics. If k = j, both the conditional price
12 t-tests (not reported) show that the proportions of the three house types are different between the data at
the usual significance levels.
123


===== PAGE 21 (score 28, numeric-paren 0, table-words 2, var-words 17) =====

Real estate listings and their usefulness...
3259
Table 5 Willingness to pay for house characteristics. Reports WTP estimates and regression diagnostics for penalised least squares estimates of Eq. (6). WTPs for continuous
(discrete) house characteristics are computed with Eq. 9 (Eq. 10). Standard errors are computed using the pairs bootstrap with 200 replications. ¯R2 is the adjusted coefficient
of determination. DCV is the double cross-validation score. Significant at ∗∗∗0.001 level, ∗∗0.01 level, ∗0.05 level
Ask data
Sale data
(1)
(2)
(3)
(4)
(5)
WTP
SE
WTP
SE
WTP
SE
WTP
SE
WTP
SE
Age
243.99∗∗∗
47.25
−78.96
43.88
−1173.73∗∗∗
103.73
−1461.39∗∗∗
106.05
−1562.89 ∗∗∗
110.14
Floor area
1419.07∗∗∗
10.39
1293.92∗∗∗
7.41
1047.99∗∗∗
17.46
906.73∗∗∗
14.82
830.24∗∗∗
16.87
Plot area
1465.18∗∗∗
7.95
1380.25∗∗∗
7.46
1152.82∗∗∗
18.14
1011.70∗∗∗
14.60
940.84∗∗∗
16.79
Detached
6881.76∗∗∗
1424.74
17771.60∗∗∗
1263.12
−7918.00∗∗
2468.32
4130.49
2198.43
6546.97∗∗
2999.16
Semi-detached
−1042.73
1205.12
5074.97∗∗∗
1011.21
−8160.82∗∗
2630.25
3741.82
2315.45
3705.94
2158.45
Listed
12599.23∗∗
3702.38
Prefabricated
−6544.52∗∗∗
1686.19
Converted attic
3426.77∗∗
1141.96
Swimming pool
14744.26∗∗
5509.92
Flat roof
−5158.22∗∗∗
1426.43
No basement
−21888.12∗∗∗
1724.46
Backland develop.
396.67
1363.94
Waterfront
59947.35 ∗∗∗
6417.56
Poor condition
−53313.91∗∗∗
2341.98
Good condition
29031.33∗∗∗
1657.96
Poor amenities
−6263.71∗∗∗
1720.23
Good amenities
16298.74∗∗∗
2243.77
Excel. amenities
37654.98∗∗∗
9613.84
f4(L AT , LON)
No
Yes
No
Yes
Yes
123


===== PAGE 22 (score 5, numeric-paren 0, table-words 2, var-words 0) =====

3260
J. Kolbe et al.
Table 5 continued
Ask data
Sale data
(1)
(2)
(3)
(4)
(5)
WTP
SE
WTP
SE
WTP
SE
WTP
SE
WTP
SE
District dummies
Yes
No
Yes
No
No
Year dummies
Yes
Yes
Yes
Yes
Yes
Buyer/Seller dummies
No
No
No
No
Yes
DCV
0.048
0.039
0.068
0.056
0.050
¯R2
0.775
0.816
0.665
0.731
0.761
N
68,070
59,502
12,524
12,218
12,218
123