# RECORD 122e07c191e40cdc3321c3817fb731f2d55466e1
# TITLE: PATH SELECTION OF SPATIAL ECONOMETRIC MODEL FOR MASS APPRAISAL OF REAL ESTATE: EVIDENCE FROM YINCHUAN, CHINA
# YEAR: 2023.0  VENUE: International Journal of Strategic Property Management  DOI: 10.3846/ijspm.2023.20376
# AUTHORS: Yu Zhao; Xuejia Shen; Jian Ma; Miao Yu
# PAGES: 13  CANDIDATE TABLE PAGES (1-based): [7, 8, 9, 10, 11]
# ABSTRACT: Urbanization, national economic growth, and China’s changing population structure have elevated the importance of real estate assessment in various contexts, including mortgage financing, secondary housing market transactions, and real estate tax reform. To address this need, this study employs a time-spatial double-fixed spatial cross-section data model as a mass appraisal tool to analyze the transaction price data of 429 ordinary residential houses in Xixia District, Yinchuan, China on April 1, 2022. Specifically, this study analyzes 7 spatial cross-section data models, discerning their interconnections. It devises an assignment technique that merges distance and characteristic variable rank into a unified indicator. The results explore spatial lag effects in real estate transaction price generation and assess the descriptive capabilities of different spatial cross-section data models.


===== PAGE 1 (score 8, numeric-paren 1, table-words 0, var-words 7) =====

Copyright © 2023 The Author(s). Published by Vilnius Gediminas Technical University
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unre-
stricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Introduction
The COVID-19 outbreak has affected China’s real estate 
market, leading to reduced new construction and a rise 
in second-hand housing transactions (Song, 2021). This 
has posed a challenge to the implementation of the local 
government’s “land finance” model. On October 23, 2021, 
the “Decision to carry out pilot work on real estate tax 
reform in some areas” was ratified during the 31st meeting 
of the Standing Committee of the National People’s Con-
gress (NPC). Subsequently, since 2022, numerous cities, 
Yinchuan City included, have introduced policies aimed 
at easing property purchase restrictions. In light of the 
demand for property tax collection and the continuous 
growth of the real estate market, the accurate assessment 
of real estate values has gained heightened significance. 
As a result, there has been a growing need for efficient 
and scientific real estate valuation in the financial market, 
tax reform, and the real estate trading market. The mass 
appraisal of real estate is a key area of current real estate 
research, with a focus on providing a value reference and 
pricing basis for buyers and sellers in the secondary hous-
ing market. It also serves as a framework for appropriate 
investment, mortgage, and pricing for developers and fi-
nancial institutions, and a basis for rational and scientific 
International Journal of Strategic Property Management
ISSN: 1648-715X / eISSN: 1648-9179
2023 Volume 27 Issue 5: 304–316
https://doi.org/10.3846/ijspm.2023.20376
*Corresponding author. E-mail: shenxuejia6@163.com
PATH SELECTION OF SPATIAL ECONOMETRIC MODEL FOR MASS 
APPRAISAL OF REAL ESTATE: EVIDENCE FROM YINCHUAN, CHINA
Yu ZHAO, Xuejia SHEN*, Jian MA, Miao YU
School of Management, Shenyang Jianzhu University, Shenyang, Liaoning, China
Received 19 May 2023; accepted 1 November 2023
Abstract. Urbanization, national economic growth, and China’s changing population structure have elevated the impor-
tance of real estate assessment in various contexts, including mortgage financing, secondary housing market transactions, 
and real estate tax reform. To address this need, this study employs a time-spatial double-fixed spatial cross-section data 
model as a mass appraisal tool to analyze the transaction price data of 429 ordinary residential houses in Xixia District, 
Yinchuan, China on April 1, 2022. Specifically, this study analyzes 7 spatial cross-section data models, discerning their 
interconnections. It devises an assignment technique that merges distance and characteristic variable rank into a unified 
indicator. The results explore spatial lag effects in real estate transaction price generation and assess the descriptive capa-
bilities of different spatial cross-section data models.
Keywords: mass appraisal of real estate, spatial section data model, SDM, empirical analysis.
Online supplementary material: Supporting information for this paper is available as online supplementary material at 
https://doi.org/10.3846/ijspm.2023.20376
urban planning by local governments. Additionally, it pro-
vides a source for evaluating the real estate tax base. Mass 
appraisal, as defined by the International Association of 
Appraisal Officers (IAAO, 2012), has been widely used to 
evaluate a collection of properties at a certain date using 
shared data, standardized methodologies, and statistical 
testing. However, the mass appraisal of real estate faces 
challenges due to inconsistent criteria for data use and the 
inadaptability of the setting and assignment of character-
istic variables. The absence of a unified database and sys-
tem of characteristic variables in the hedonic price model 
and other approaches exacerbates these issues (Chen et al., 
2020). Additionally, the geographical characteristics of real 
estate as a commodity are often ignored by the hedonic 
price model, which emphasizes the variety of commodi-
ties (Feng et al., 2019). Therefore, there are concerns about 
the accuracy and flexibility of the estimation of house 
costs using mass appraisal methods. Several studies have 
utilized the average transaction price of residential units 
or the transaction price of a single unit as a sample in 
the empirical examination of real estate mass appraisal. 
However, the data distribution is often unequal due to the 
varied transaction activity of different residential areas 
at a given time, which directly affects the accuracy of re-
gression analysis. In addition, the impact of high-quality 


===== PAGE 7 (score 18, numeric-paren 0, table-words 2, var-words 12) =====

310
Y. Zhao et al. Path selection of spatial econometric model for mass appraisal of real estate: evidence...
Table 2. Selection and quantification standard of characteristic variables
Feature item
Symbol
Quantization standard
Assignment method
Expected 
impact
Area
Ar
Building area
Actual value
+
Decoration
D
Blank 2, simple 4, medium 6, hardcover 8, luxury 10
Virtual assignment
+
Floor
F
High-rise residence: low 1, middle 3, high 2
Virtual assignment
unknown
Low rise house: low 3, medium 2, high 1
Elevator/househols
EH
Number of elevators on the floor/number of single properties on 
the floor
Actual value
+
Age
Ag
Year the house was built – 2022.4.1
Actual value
–
Greening rate
GR
Green area/residential area
Actual value
+
Parking space ratio
PSR
Total number of households/parking spaces
Actual value
+
Public activity area 
for residents
PAA
Number of areas
Actual value
+
Property fee
PF
Monthly charge per square meter
Actual value
unknown
Orientation
O
Reasonable = 1, unreasonable = 0
Actual situation
+
CBD
CBD
Geographical distance
Anti-geographical 
distance
+
Shopping mall
SM
Service capability: municipal = 3, district 
level = 2, street level = 1
Service capacity/
geographical distance
+
Geographical distance
Main road
MR
Geographical distance
Anti-geographical 
distance
unknown
Bus
B
Bus route number
Quantity/distance
+
The geographical distance to the bus stop
Medical treatment
MT
Qualifications: Grade III Level Special = 10, Grade III Level A = 
9, Grade III Level B = 8, Grade III Level C = 7, Grade II Level 
A = 6, Grade II Level B = 5, Grade II Level C = 4
Qualifications/
geographical distance
+
Geographical distance
Education
E
Qualifications: high quality = 3, good = 2, ordinary = 1
Qualifications/
geographical distance
+
Geographical distance
Developer brand
DB
High quality = 3, good = 2, average = 1
Virtual assignment
+
Development zone
DZ
Rating: national = 3, provincial = 2, municipal = 1
Rating/geographical 
distance
+
Geographical distance
Park
P
Geographical distance
Anti-geographical 
distance
+
Table 3. The description of variable values
Feature item
Value range
Maximum value
Minimum value
Mean value
Area
40–200
163
44
95
Decoration
2–10
8
2
5
Floor
1–4
4
1
2
Elevator/households
0–1
1
0
0.25
Age
1–35
32
2
15.5
Greening rate
15–50%
45%
15%
28%
Parking space ratio
0–1
0.8
0.1
0.4
Property fee
0–1
1
0.25
0.65
Orientation
0–1
1
0
0.8
CBD
–
0.204
0.0833
1.01
Shopping mall
–
18.75
0.53
2.34
Main road
–
14.29
0.76
3.40
Bus
–
173.91
2.08
30.31
Medical treatment
–
90
2.73
12.05
Education
–
1320
4.57
169
Developer brand
1–6
6
1
3.5
Park
–
20
0.53
2.34


===== PAGE 8 (score 16, numeric-paren 0, table-words 6, var-words 2) =====

International Journal of Strategic Property Management, 2023, 27(5): 304–316
311
4. Results and discussion
4.1. The spatial dependency diagnosis of the 
evaluation area
The spatial weight matrix using the K-adjacency matrix is 
constructed using GeoDa1.20. The number of neighbors 
is set to 16 based on the sample’s actual situation, location 
similarity, neighborhood characteristics, and practical ex-
perience of the market method in single-case assessment.
4.1.1. Global spatial autocorrelation test
The global autocorrelation test was conducted using the 
single variable global Moran’s I test tool in GeoDa1.20, 
and the results of the significance test obtained through 
999 permutations are shown in Figure 5. It can be ob-
served that the global Moran’s I index is approximately 
0.4858 under the 0.001 significance level test, which pro-
vides strong evidence for the need of spatial econometric 
analysis.
Figure 5. Global autocorrelation test results
4.1.2. Local spatial autocorrelation
The local autocorrelation test was conducted using the 
univariate local Moran’s I test tool in GeoDa1.20, and the 
resulting Moran’s I scatter plot is displayed in Figure 6. 
The scatter plot reveals the existence of both high-high 
and low-low spatial adjacency effects within the sample 
group.
Figure 6. Local Moran’s I scatter diagram
LISA cluster diagram is shown in Figure 7, and the 
distribution of four spatial clustering phenomena can be 
intuitively observed. 
4.2. Setting and testing of spatial section data model
Before constructing the spatial cross-section data mod-
el, it’s imperative to subject the characteristic variables 
and variable function form to relevant statistical tests 
for validation. These tests encompass evaluating the 
goodness-of-fit for linear, semi-logarithmic, and full-
logarithmic models, examining multicollinearity and 
heteroscedasticity, assessing the significance of variables, 
and computing AIC and log likelihood. Table 4 shows 
the results of these tests reveal: (1) all three forms pass 
the F-statistical test, signifying substantial explanatory 
variable significance; (2) only the full logarithmic model 
exhibits significant multicollinearity and heteroscedastic-
ity; and (3) the semi-logarithmic form outperforms the 
linear model in terms of AIC, log likelihood, and adjust-
ed R-squared. Consequently, the optimal selection is the 
semi-logarithmic function form. Specific variables and 
their associated assignment methods under this form are 
meticulously outlined in Table 5.
Table 4. Inspection and selection results
Test
Semi log
Full log
Linear
Log likelihood
375.449
362.16
–3359.46
AIC
–728.898
–702.32
6740.92
Multicollinearity condition 
number
28.48
96.72
28.48
Adjusted R-squared
0.8263
0.8152
0.8113
Breusch-Pagan test
0.05727
0.00647
0.06454
F-statistic
0.00000
0.00012
0.00000
Koenker-Bassett test
0.09693
0.00207
0.05673
Number of significant 
variables
10
11
10
Note: The output results of the Breusch-Pagan test, Koenker-Bassett test 
and F-statistic in the table are the adjoint probabilities of the test results.
Figure 7. LISA cluster diagram


===== PAGE 9 (score 83, numeric-paren 48, table-words 8, var-words 1) =====

312
Y. Zhao et al. Path selection of spatial econometric model for mass appraisal of real estate: evidence...
Table 5. Significant variables and assignment methods
Variable
t-Statistic
Probability
Assignment method
Ar
11.82180 
0.00000 
Actual value
D
8.00813 
0.00000 
Actual value
F
4.94734 
0.00000 
Virtual assignment
EH
4.64670 
0.00000 
Actual value
Ag
–15.09490 
0.00000 
Actual value
CBD
10.10060 
0.00000 
Anti-geographical distance
SM
5.46333 
0.00000 
Service capacity/
geographical distance
MR
–4.24358 
0.00003 
Anti-geographical distance
P
3.17022 
0.00164 
Anti-geographical distance
B
4.59475 
0.00001 
Quantity/distance
4.3. Comparative analysis of model estimates
4.3.1. Summary of model estimation and test results
The results of LM test and Robust LM test of residuals 
based on the regression results of classical linear models 
are shown in Table 6. 
Table 6. LM inspection and robust LM inspection results
Test
Value
Probability
LM (lag)
81.7809
0.0000
LM (error)
138.9328
0.0000
Robust LM (lag)
30.0195
0.0000
Robust LM (error)
87.1714
0.0000
The results of R-squared, Log-likelihood, AIC and 
LR test of each model are shown in Table  7. In the 
first test of LR test (LR1), the constraint models cor-
responding to SLX, SLM, and SEM is CLR, the con-
straint models corresponding to SDM, SDEM, SAC, and 
GNS are SLX, SLX, SLM, and SEM respectively, and the 
constraint models corresponding to SDM, SDEM, and 
SAC in the second test (LR2) are SLM, SEM, and SEM 
respectively. 
The regression results of each model are shown in Ta-
ble 8.
Table 7. Summary of model test results
Variable
CLR
SLX
SLM
SEM
SDM
SDEM
SAC
GNS
R2
0.8263
0.8389
0.8728
0.9033
0.9037
0.9041
0.9069
0.8909
Log-likelihood
375.449
386.575
431,617
467.324
473.693
469.429
467.5302
469.869
AIC
–728.898
–745.149
–839.235
–912.648
–913.386
–906.858
–910.164
–902.126
LR1
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.1032
LR2
0.0002
0.0932
0.0740
Table 8. Summary of model estimates
Variable
CLR
SLX
SLM
SEM
SDM
SDEM
SAC
GNS
t(P)
t(P)
t(P)
t(P)
t(P)
t(P)
t(P)
t(P)
Ar
11.82
10.21
12.5
13.06
13.05
12.87
12.91
12.84
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
D
8
7.73
8.63
9.19
8.88
6.76
9.11
5.93
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
F
4.94
5.32
5.35
5.93
5.93
5.42
5.91
5.38
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
EH
4.64
4
4.29
5.45
5.22
5.43
5.42
5.46
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
Ag
–15.09
–11.75
–12.79
–16.25
–15.74
–15.85
–15.97
–15.87
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
B
4.59
4.87
3.78
2.6
2.99
2.51
2.66
2.29
(0.000)
(0.000)
(0.000)
(–0.009)
(–0.003)
(–0.012)
(–0.008)
(–0.022)
MR
–4.24
–3.88
–2.97
–2.88
–2.68
–2.97
–2.86
–2.88
(0.000)
(0.000)
(–0.003)
(–0.004)
(–0.007)
(–0.003)
(–0.004)
(–0.004)
SM
5.46
5
3.58
0.7
1.94
0.56
0.81
0.17
(0.000)
(0.000)
(0.000)
(–0.485)
(–0.052)
(–0.576)
(–0.417)
(–0.864)


===== PAGE 10 (score 47, numeric-paren 17, table-words 11, var-words 0) =====

International Journal of Strategic Property Management, 2023, 27(5): 304–316
313
Variable
CLR
SLX
SLM
SEM
SDM
SDEM
SAC
GNS
t(P)
t(P)
t(P)
t(P)
t(P)
t(P)
t(P)
t(P)
CBD
10.1
9.43
5.35
3.08
3.06
3.04
3.15
2.35
(0.000)
(0.000)
(0.000)
(–0.002)
(–0.002)
(–0.002)
(–0.002)
(–0.019)
P
3.17
3.12
3.38
2.9
2.61
2.79
2.93
2.73
(0.002)
(0.002)
(–0.001)
(–0.004)
(–0.009)
(–0.005)
(–0.003)
(–0.006)
W_Ar
1.57
–3.32
1.48
1.83
(0.116)
(–0.001)
(–0.14)
(–0.067)
W_EH
–0.18
–2.32
0.47
0.75
(0.861)
(–0.021)
(–0.639)
(–0.451)
W_Ag
–0.09
6.07
0.94
–0.24
(0.930)
(0.000)
(–0.350)
(–0.813)
W_F
2.42
–0.52
1.2
1.43
(0.016)
(–0.606)
(–0.231)
(–0.153)
W_D
3.83
–1.41
–0.66
–0.6
(0.000)
(–0.159)
(–0.507)
(–0.546)
ρ
10.96
16.22
0.65
–1.13
(0.000)
(0.000)
(–0.513)
(–0.259)
λ
18.81
17.64
11.56
12.04
(0.000)
(0.000)
(0.000)
(0.000)
Note: where: W – spatial lag term of the corresponding variable; ρ – dependent variable spatial regression coefficient; λ – error term spatial regression 
coefficient.
End of Table 8
4.3.2. Comparative analysis of appraisal models
The construction path selected based on the optimal mod-
el of ordinary residential real estate (Figure 2) and the test 
results (Tables 6 and 7) is shown in Figure 8. 
Step 1: Based on Tables 7 and 8, the significance of 
the spatial lag variable in SLX is observed, indicating that 
SLX outperforms CLR and cannot be downgraded. Thus, 
including the spatial lag effect of the dependent variable 
in the model setting is necessary.
Step 2: Table 6 shows that the dependent variable and 
the error term have passed the LM test and the Robust 
LM test, and the spatial lag effect of the error term is more 
robust than that of the dependent variable (SEM is better 
than SLM).
Step 3: Table 8 reveals that SDM is superior in terms 
of R-squared, log-likelihood, and AIC test results, and it 
cannot be reduced to a single variable model (SLM, SLX). 
Thus, the model can effectively identify the spatial lag 
effect of independent and dependent variables. Table 9 
Figure 8. Model building path (slt = Spatial lag term; iv = independent variable)


===== PAGE 11 (score 40, numeric-paren 0, table-words 15, var-words 3) =====

314
Y. Zhao et al. Path selection of spatial econometric model for mass appraisal of real estate: evidence...
shows that the two spatial lag variables of SDM are signifi-
cant, while only the spatial lag of the error term is signifi-
cant in SAC, SDEM, and GNS. Therefore, if insignificant 
variables are removed from the model, the three models 
will degenerate into SEM models. SDM is capable of ac-
curately describing the impact of neighboring real estate 
prices and characteristics on real estate prices during data 
generation as there is no significant variable competition 
between the dependent variable spatial lag term and the 
independent variable spatial lag term. However, variable 
competition exists between the error term spatial lag term 
and the dependent variable and independent variable spa-
tial lag term, resulting in the regression analysis of SAC, 
SDEM, and GNS lacking spatial lag essence. Additionally, 
the over-parameterization of GNS significantly reduces its 
overall analytical ability.
4.3.3. Optimization of model based on mass appraisal 
of real estate 
It is not difficult to see from the comprehensive compara-
tive analysis that the comparison between SDM model 
and SAC model, SDEM model, and GNS model can be 
essentially understood as a comparative analysis between 
SDM model and SEM model.
From the perspective of mass real estate appraisal, the 
functional model should exhibit the following character-
istics: (1) The estimated results of the model should be 
close to the actual market value; (2) The variable setting 
should have good interpretability; (3) The model should 
have practical utility.
While the SEM model’s spatial lag term of the error 
can enhance the model estimation’s accuracy, this variable 
is essentially a “black box.” The insignificant influence fac-
tors, hidden influence factors, and exogenous errors in the 
“black box” are difficult to observe, and researchers and 
evaluators cannot explain the economic significance of 
this variable. Consequently, the evaluation results of this 
model will be questioned by both parties, and its guiding 
significance for practical application will be lost.
In contrast, the estimation results of SDM demon-
strate that the regression coefficients of the independent 
variables, the independent variable spatial regression co-
efficients, and the dependent variable spatial regression 
coefficients are significant, and the estimated coefficients 
of each variable’s influence direction are consistent with 
expectations. Hence, it can better elucidate the econom-
ic significance of each variable than SEM. Additionally, 
even without considering the theoretical significance of 
model setting, the SDM test results for Log Likelihood, 
AIC, and R-squared are marginally better than those of 
SEM.
To conclude, SDM has the best evaluation ability. The 
optimal functional form for mass real estate appraisal 
based on ordinary residential real estate data in the study 
area is Equation (8), and the model estimation coefficients 
are shown in Table 9.
= ρ
+β ⋅
⋅θ +β ⋅
+
⋅
⋅θ +
β ⋅
+
⋅
⋅θ +β ⋅
+
⋅
⋅θ +
β ⋅
+
⋅
⋅θ +β ⋅
+
⋅
⋅θ +
β ⋅
+
⋅
⋅θ +β ⋅
+
⋅
⋅θ +
β ⋅
+
⋅
⋅θ +β ⋅
+
⋅
⋅θ
+ ε
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
9
9
10
10
( )
,
Ln Y
WY
Ar
D
W D
F
W F
EH
W EH
Ag
W Ag
B
W B
MR
W MR
SM
W SM
CBD
W CBD
P
W P
where: Y – the vector formed by the single trading price 
of each sample, Yuan/m2; βi – linear regression coefficient; 
θi – spatial autoregressive coefficient of independent vari-
able; Ar, D, Ag, F, EH, B, MR, SM, CBD, P – vectors com-
posed of corresponding eigenvalues of each sample; W – 
K adjacency matrix; ε – error term; ρ – spatial regression 
coefficient of dependent variable.
Table 9. SDM coefficient estimation
Variable
Coefficient Std. Error
t-Statistic
Probability
CONSTANT
2.4459
0.3477
7.04
0.000
Ar
0.0029
0.0002
13.05
0.000
D
0.0219
0.0025
8.88
0.000
F
0.0285
0.0048
5.93
0.000
EH
0.0990
0.0190
5.22
0.000
Ag
–0.0157
0.0010
–15.74
0.000
B
0.0005
0.0002
2.99
0.003
MR
–0.0037
0.0014
–2.68
0.007
SM
0.0035
0.0018
1.98
0.048
CBD
0.4269
0.1393
3.06
0.002
P
0.0048
0.0018
2.61
0.009
W_Ar
–0.0016
0.0005
–3.32
0.001
W_EH
–0.0758
0.0327
–2.32
0.021
W_Ag
0.0109
0.0018
6.07
0.000
W_F
–0.0058
0.0113
–0.52
0.606
W_D
–0.0080
0.0057
–1.41
0.159
ρ
0.6902
0.0425
16.22
0.000
Table 9 shows that for every 1 m2 increase in the area 
of the surrounding properties compared to the current 
property (W_Ar), the house price of the current property 
decreases by 0.0016 Yuan/m2. Similarly, for every unit in-
crease in the elevator-to-household ratio of the surround-
ing properties compared to the current property, the house 
price decreases by 0.0758 Yuan/m2. Conversely, for each ad-
ditional year in the age of the surrounding properties com-
pared to this property, the price of the property increases 
by 0.0109 Yuan/m2. Additionally, when the decoration level 
of the surrounding properties increases by 1 unit compared 
to this property, the house price of this property decreases 
by 0.0080 Yuan/m2. The coefficient of Ar is 0.0029, indicat-
ing that, in addition to the influence of spatial relations, the 
housing price increases by 0.0029 Yuan/m2 for every square 
meter increase in property area.
The coefficient ρ is 0.6902, meaning a 1  Yuan/m2 
change in surrounding residential housing price corre-
sponds to a 0.6902 Yuan/m2 change in real estate price.


===== PAGE 12 (score 9, numeric-paren 0, table-words 2, var-words 5) =====

International Journal of Strategic Property Management, 2023, 27(5): 304–316
315
Furthermore, while the CBD coefficients appear sub-
stantial, their comparison with coefficients from other spa-
tial econometric models highlights the SDM’s reasonable 
estimations. This variable has also successfully cleared the 
multicollinearity test, and the resulting model outcomes 
align with the IAAO standard ratio. As such, the model’s 
applicability is well-supported and can be confidently ac-
knowledged.
4.4. Appraisal ability measure
To further assess the practical usability of the optimal 
model, a single real estate transaction price was randomly 
selected from the 115 residential districts not included in 
the training data for testing. The model’s appraisal per-
formance was then measured against the mass appraisal 
standard outlined in section 3.2. As depicted in Table 10, 
the SDM produced three measurement outcomes that align 
with international standards and exhibit minimal Coeffi-
cient of Dispersion (COD) values. Moreover, the Mean Ra-
tio (MR) and Price-Related Differential (PRD) values are 
closely approximating 1. These outcomes underscore the 
excellent appraisal capabilities of the SDM within the study 
area, thus fulfilling international standards.
Table 10. Summary of measurement results
Indicator
MR
COD
PRD
International standard
0.9-1.1
5-15
0.98-1.03
Test result
1.0293
8.4840
1.0097
Conclusions
The study aims to build and compare the explanatory and 
appraisal abilities of different spatial cross-section data 
models using a set of data. Based on the empirical analysis 
and relevant tests, the following conclusions can be drawn:
Firstly, the significant high-high and low-low aggrega-
tion phenomena in the flat transaction price of ordinary 
residential real estate prove the necessity of spatial econo-
metric analysis and highlight the bias in the traditional he-
donic price model’s assumption of sample independence.
Secondly, the significant spatial regression coefficients 
of independent and dependent variables in the SLX and 
SDM regression results explain the influence of neighbor-
ing real estate characteristics and prices in generating real 
estate price data.
Thirdly, the comparative analysis and optimization of 
the models show that SDM has excellent appraisal ability 
to describe the spatial spillover effect in the data genera-
tion process. The measurement results conform to inter-
national standards, indicating that SDM is superior to oth-
er spatial cross-section data models in terms of appraisal 
accuracy and applicability to explain the data generation 
process.
In conclusion, the spatial Durbin model based on dou-
ble fixed effect has good applicability in the field of mass 
appraisal of real estate, and it can effectively capture the 
spatial dependence in real estate price data. These find-
ings can provide a reference for the practical application 
of mass appraisal of real estate.
Currently, the model’s scope is restricted to general 
residential real estate, excluding apartments and com-
mercial properties. Nevertheless, its usability is con-
strained by existing limitations such as non-transparent 
property transactions and the absence of standardized 
criteria for property characteristics. The following con-
tents can be further explored in the follow-up research. 
First of all, with regard to characteristic variables, it is nec-
essary to select characteristic variables according to the 
actual situation of each kind of city and to standardize 
the way of assigning values, that is, to establish a stand-
ard assignment system for mass appraisal; the second is, 
in terms of the spatial weighting matrix, when selecting 
“neighboring” properties, in addition to determining the 
“neighborhood range” based on the empirical data of 
house buyers and sellers, a comprehensive survey of the 
entire study area should be conducted in order to deter-
mine a universal “neighborhood range” and to verify the 
impact of different “neighborhood ranges” on the accu-
racy of the assessment based on actual data.
Funding
This study was supported by the Humanities and Social 
Sciences of the Ministry of Education (22YJA630121), 
Liaoning Revitalization Talents Program (XLYC2203004), 
Liaoning Federation of Social Sciences (2024lslyb-
kt-043), Social Science Foundation of Liaoning Province 
(L20BJY010), and Basic Scientific Research Project of Col-
leges and Universities in Liaoning Province (lnms202140).
Ethical statement
I testify on behalf of all co-authors that our article submit-
ted to the International Journal of Strategic Property Man-
agement: This article does not contain any studies with 
human participants or animals performed by any of the 
authors; This manuscript is original and has not been pub-
lished and will not be submitted elsewhere for publication; 
No data have been fabricated or manipulated (including 
images) to support the conclusions; All the authors listed 
have approved the manuscript that is enclosed and they 
have no conflict of interest.
References
Anderson, S. T., & West, S. E. (2006). Open space, residential 
property values, and spatial context. Regional Science and Ur-
ban Economics, 36(6), 773–789. 
https://doi.org/10.1016/j.regsciurbeco.2006.03.007 
Antipov, E. A., & Pokryshevskaya, E. B. (2012). Mass appraisal 
of residential apartments: an application of random forest 
for valuation and a CART-based approach for model diag-
nostics. Expert Systems with Applications, 39(2), 1772–1778. 
https://doi.org/10.1016/j.eswa.2011.08.077