# RECORD 757bd683e867096bd3de6214d8e05c02d711dbcc
# TITLE: Effects of Educational and Cultural Facilities on Housing Prices in Seoul from an Accessibility Perspective
# YEAR: 2021.0  VENUE: Journal of people, plants, and environment  DOI: 10.11628/ksppe.2021.24.5.529
# AUTHORS: Minki Sung; Junghoon Ki
# PAGES: 16  CANDIDATE TABLE PAGES (1-based): [6, 9, 10, 11]
# ABSTRACT: Background and objective: A great deal of previous research has highlighted the value of educational and cultural facilities embedded in housing prices, by taking a large spatial area as the focus, such as the city or district level. However, few studies have investigated the extent to which educational and cultural facilities influence the formation of housing prices from an accessibility perspective. This study aims to identify the value of educational and cultural facilities embedded in the housing prices in Seoul Metropolitan City with a focus on the concept of the residents’ neighbourhood and accessibility. Methods: To this end, this research used a spatial regression model with educational and cultural facilities as the independent variables and housing prices as the dependent variable. The model assessed the accessibility of cultural and educational facilities by considering geographic effects. Results: The findings are as follows. First, the spatial error model was found to be the best fit for multi-unit housing, while the spatial lag model was more appropriate for single-unit housing and apartments. Second, private educational facilities and art museums had positive effects on single- and multi-unit housing prices, while historical sites had a negative effect. Finally, private educational facilities positively influenced apartment prices, whereas public libraries and urban park areas had a negative effect. Conclusion: These findings indicate that the accessibility of educational and cultural facilities reflects residents’ preferences and needs, which will ultimately influence housing prices.


===== PAGE 1 (score 12, numeric-paren 0, table-words 1, var-words 10) =====

Minki Sung and Junghoon Ki
Journal of People, Plants, and Environment Vol. 24, No. 5, 2021∙529
J. People Plants Environ. Vol. 24 No. 5: 529-544, October 2021
https://doi.org/10.11628/ksppe.2021.24.5.529
JPPE
Effects of Educational and Cultural Facilities on Housing Prices in 
Seoul from an Accessibility Perspective
Minki Sung1 and Junghoon Ki2*
1Postgraduate student, Department of Archaeology, Durham University, Durham DH1 3LE, United Kingdom
2Professor, Department of Public Administration, School of Social Science, Myongji University, Seoul 03674, South Korea
ABSTRACT
Background and objective: A great deal of previous research has highlighted the value of educational and cultural facilities
embedded in housing prices, by taking a large spatial area as the focus, such as the city or district level. However, few 
studies have investigated the extent to which educational and cultural facilities influence the formation of housing prices 
from an accessibility perspective. This study aims to identify the value of educational and cultural facilities embedded in
the housing prices in Seoul Metropolitan City with a focus on the concept of the residents’ neighbourhood and accessibility.
Methods: To this end, this research used a spatial regression model with educational and cultural facilities as the 
independent variables and housing prices as the dependent variable. The model assessed the accessibility of cultural and 
educational facilities by considering geographic effects. 
Results: The findings are as follows. First, the spatial error model was found to be the best fit for multi-unit housing, while
the spatial lag model was more appropriate for single-unit housing and apartments. Second, private educational facilities
and art museums had positive effects on single- and multi-unit housing prices, while historical sites had a negative effect.
Finally, private educational facilities positively influenced apartment prices, whereas public libraries and urban park areas
had a negative effect. 
Conclusion: These findings indicate that the accessibility of educational and cultural facilities reflects residents’ 
preferences and needs, which will ultimately influence housing prices.
Keywords: neighbourhood; public facilities; residents’ preference; spatial regression model
This work was supported by the National Research Foundation of Korea [grant number: NRF-2017R1D1A1B03032415]
Received: July 6, 2021, Revised: August 10, 2021, Accepted: September 13, 2021
First author: Minki Sung, mingi5678@nate.com, 
 https://orcid.org/0000-0001-6665-9383
*Corresponding author: Junghoon Ki, johnki70@gmail.com, 
 https://orcid.org/0000-0001-7554-9573
Ⓒ 2021 by the Society for People, Plants, and Environment. This is an Open Access article distributed under the terms of the Creative Commons 
Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, 
and reproduction in any medium, provided the original work is properly cited.
Introduction
How is housing appraised? Are the physical character-
istics of housing the only indicators of its value? While these 
have been ongoing questions in housing research, many re-
searchers have also discussed housing prices per se or in 
relation to locational characteristics (Dong et al., 2019; Li 
et al., 2019; Wittowsky et al., 2020). In reality, housing prices 
reflect not only the physical characteristics of a house itself 
but also the various elements of its location and surrounding 
environment (Adair et al., 2000; Dong et al., 2019; Lan et al., 
2018; Li et al., 2019; Zahirovic-Herbert and Gibler, 2020). 
In particular, the neighbourhood plays an important role 
(Bramley et al. 2008; Lan et al., 2018; Wittowsky et al., 2020), 
in that housing prices vary significantly among different sub-
markets segmented according to accessibility (Adair et al., 2000; 
Feng et al., 2020; Ki and Sung, 2020; Sung and Ki, 2021).
Previous Seoul-based case studies, which mainly focused 
on how urban public facilities affect housing prices over a 
large area, such as a residential district (Jin and Sohn, 2005; 
Kim, 2012; Kim and Lee 2007; Lee, 2010), have primarily 
addressed the extent to which access to educational facilities 
influences housing prices. Further, recent studies on cultural 
facilities and green spaces have explicitly demonstrated that 


===== PAGE 6 (score 14, numeric-paren 3, table-words 3, var-words 5) =====

Effects of Educational and Cultural Facilities on Housing Prices in Seoul from an Accessibility Perspective
534∙Journal of People, Plants, and Environment Vol. 24, No. 5, 2021
where Y denotes the housing price, x is a matrix of observations 
on the explanatory variables, and ρwy is the spatial lag (or 
error) variable. ρ assesses the degree of spatial correlation and 
w is the spatial weights matrix. The value of ρ ranges be-
tween -1 and 1, with ρ = 0 signifying no spatial correlation. 
Under a positive ρ, housing prices are expected to be higher if 
spatially adjacent housing prices are higher, while ε is a vector 
of the independent and equivalently distributed error terms 
(Ward and Gleditsch, 2008).
Spatial regression models include SLM and SEM, which 
use spatial regression dependencies and spatial regression 
errors, respectively. We determined which model was more 
suitable by comparing them with the results of the OLS model. 
The OLS model was tested for suitability under the assumptions 
of normality, equal variability, and spatial independence of 
errors (Park and Ham, 2018). Similarly, the spatial depend-
encies of dependent variables and regression model error 
terms were examined using the Lagrange multiplier (LM) lag 
and error. If the LM lag and error are meaningful at the 5% sig-
nificance level, the null hypothesis of the spatial independence 
of dependent variables and error terms is rejected. If only one 
of either the LM lag or error is statistically significant, selecting 
a more adequate model should be considered. If both values 
are significant, the robust LM lag and robust LM error statistics 
test functions are used for reselection, as a last resort (Park 
and Ham 2018). Park and Ham (2018) relied on the log-like-
lihood, Akaike information criterion (AIC), and Schwarz 
criterion (SC) results to determine the suitability of spatial 
regression models. Compared to the results of the OLS mod-
el, the model with a higher log-likelihood and lower AIC 
and SC was considered appropriate.
To determine the effects of educational and cultural factors, 
we used the number of public libraries and private educational 
facilities as a proxy for the educational factor (independent 
variable). Regarding the cultural factor, art museums, museums, 
historical sites, and urban parks are used as independent 
variables. Table 3 and Table 4 present detailed statistical 
information for the independent and dependent variables.
Results and Discussion
As shown in Fig. 3, we illustrated housing prices and 
their independent variables using ArcGIS 10.8. On average, 
the number of private educational facilities was large (31.65) 
and showed a high standard deviation (17.74) compared 
to other facilities. Fig. 3 also shows that private educational 
facilities tend to be clustered in the south-eastern part of 
Seoul, where housing prices are generally higher regardless of 
the housing type. However, public libraries had a relatively 
low standard deviation (1.26), implying that a similar number 
of public libraries are distributed across districts, as can 
also be seen in Fig. 3.
As for cultural facilities, art museums, museums, and 
Fig. 2. Global and local Moran’s I analysis results; MULTI: multi-unit housing, SINGLE: single-unit housing, APT: apartment.


===== PAGE 7 (score 13, numeric-paren 0, table-words 2, var-words 8) =====

Minki Sung and Junghoon Ki
Journal of People, Plants, and Environment Vol. 24, No. 5, 2021∙535
Variable
Dong (sub-district)
Average
Standard deviation
Max
Min
Dependent 
variable
Price by housing type
Multi-unit housing
249688516.3
155834809.2
1100000000
104363359.5
Single-unit housing
831456068.7
617399014.6
4288716364
190000000
Apartment
517997051
290439973.6
1956893793
183663265.3
Independent 
variables
Educational factor
Public library
1.43
1.26
20
0
Private educational 
facility
31.65
17.74
195
0
Cultural factor
Art museum
0.11
0.57
7
0
Museum
0.26
1.14
11
0
Historical site
0.35
3.16
49
0
Number of urban parks
6.05
2.53
27
0
Urban park area
0.06 km2 (4%)
0.97
14.55 km2 (80%)
0.00003 km2 (0.01%)
Note: All independent variables are expressed per 10,000 population. Urban park areas are expressed as a proportion of the total sub-district (dong) area. The 
unit of housing price is Korean won.
Table 4. Descriptive statistics of independent and dependent variables
Variables
Definition
Dependent 
variable
Price by 
housing type
Multi-unit housing
Real estate transaction 
price
Housing types are classified as follows: single-unit 
housing, multi-unit housing, or apartment. Single- and 
multi-unit housing tends to be occupied by less affluent 
families, whereas relatively affluent families reside in 
apartments (Kang et al., 2019). More than half of the 
population (50.7%) in Seoul resides in apartments (Kang 
et al., 2019). Furthermore, Jeon (2015) suggested that 
most residents in Korea prefer apartments because an 
apartment complex that has access to educational and 
cultural facilities is more attractive than single- or 
multi-unit dwellings when raising children.
Single-unit housing
Apartment
Independent 
variables
Educational 
factors
Public library
Number of public libraries 
in Seoul
Public library means a library established and operated by 
local government for information, reading activities, and 
cultural activities by the public. The libraries we 
employed as an independent variable are local public 
libraries, not university or special libraries, specially built 
for residents.
Private educational facility
Number of registered 
private academies and 
study centres in Seoul
Private educational facilities in this study are categorized 
into two types: those with curricula that facilitate entrance 
into prestigious universities and those providing lifelong 
or vocational education.
Cultural factors
Art museum
Number of art museums in 
Seoul
The reason for distinguishing museums from art museums 
is that the Museum and Art Gallery Support Act of South 
Korea (Government of the Republic of Korea, 2018b) 
considers them to be different.
Museum
Number of museums in 
Seoul
Historical site
Number of historical sites 
in Seoul
In this study, historical sites cover every historical 
property registered by central and local governments. In 
South Korea, historical sites can be categorized into three 
types: nationally assigned cultural properties, city 
assigned properties, and nationally registered properties.
Number of urban parks
Number of urban parks in 
Seoul
The park concept we employed in this study includes 
parks established and managed within the urban residents’ 
immediate neighbourhood structure; it includes 
small-sized, children’s, and neighbourhood parks.
Urban park area
Urban park areas in Seoul
Table 3. Definitions of independent and dependent variables


===== PAGE 9 (score 34, numeric-paren 4, table-words 10, var-words 6) =====

Minki Sung and Junghoon Ki
Journal of People, Plants, and Environment Vol. 24, No. 5, 2021∙537
Table 7 presents the analysis results of the OLS model, 
SLM, and SEM using single-unit housing price as the de-
pendent variable. The Jarque-Bera test statistic of 1286.0811 
(p < .05) signifies an abnormality in the spatial autocorrelation 
test of error terms. Therefore, this test leads to spatial re-
gression models instead of the OLS model. Using the results of 
the spatial regression model analysis, we selected the more suit-
able model between SLM and SEM. The R2 value (0.303476) 
of the SLM is higher than that of the SEM, and is also more 
robust. Additionally, the log-likelihood value of the SLM 
(-9,084.95) is higher and their AIC and SC are lower than 
those of the SEM. Furthermore, the likelihood ratio test of 
the SLM generates statistically significant results, indicating 
it is more suitable than the SEM.
For this reason, the SLM was considered the most suitable 
model based on the analysis results. The SLM results reveal 
that educational (private educational facilities) and cultural 
(art museums) factors exert positive influences on single- 
unit housing prices. However, historical sites adversely affect 
single-unit housing prices.
Table 8 shows the analysis results of the OLS model, SLM, 
and SEM using apartment price as the dependent variable. 
The Jarque-Bera test statistic of 1,008.7673 (p < .05) signifies 
an abnormality in the spatial autocorrelation test of the error 
term. Moreover, the Breusch-Pagan test statistic (26.1987, 
p < .05) indicates heteroscedasticity in the error term. Therefore, 
spatial regression models are preferrable to the OLS model.
The spatial regression model analysis allowed the most 
suitable model between the SLM and SEM to be selected. 
The R2 value (0.171330) of the SLM was more robust and 
higher than that of the SEM. Moreover, the log-likelihood 
value of the SLM (-8,803.23) was higher and the SC lower 
Dependent variable (multi-unit housing)
OLS
SLM
SEM
Constant
2.53016e + 008***
2.30529e + 008***
2.52693e + 008***
Educational factor
Public library
-1.14233e + 007
-1.18353e + 007
-1.21706e + 007
Private educational facility
548281**
546127**
544701**
Cultural factor
Art museum
8.81226e + 007***
8.72462e + 007***
8.69073e + 007***
Museum
4.83881e + 006
6.14577e + 006
6.9372e + 006
Historical site
-1.37727e + 007***
-1.39005e + 007***
-1.39798e + 007***
Number of urban parks
-1.28692e + 006
-1.21817e + 006
-1.16495e + 006
Urban park area
-0.996065
-0.957873
-0.914984
Rho
0.0890597
Lambda
0.092414
R2
0.081162
0.085508
0.085678
Adj. R2
0.065664
Log-likelihood
-8561.41
-8560.71
-8560.70
Akaike information criterion
17138.8
17139.4
17137.4
Schwarz criterion
17171.2
17175.9
17169.8
Koenker-Bassett
3.5660
Breusch-Pagan
19.1758**
18.8382
18.8002
White test
60.6184**
Moran’s I
1.3146
Lagrange multiplier (lag)
1.4785
Robust LM (lag)
0.0085
Lagrange multiplier (error)
1.4964
Robust LM (error)
0.0265
Likelihood ratio test
0.23762
1.4261
Note:* p < .05, ** p < .01, *** p < .001. OLS: ordinary least squares; SLM: spatial lag model; SEM: spatial error model; LM: Lagrange multiplier.
Table 6. Analysis of multi-unit housing


===== PAGE 10 (score 34, numeric-paren 7, table-words 7, var-words 7) =====

Effects of Educational and Cultural Facilities on Housing Prices in Seoul from an Accessibility Perspective
538∙Journal of People, Plants, and Environment Vol. 24, No. 5, 2021
than that of the SEM. For this reason, the SLM was re-
garded to be the most suitable based on the analysis results. 
The results reveal that the educational factor exerted both 
positive (private educational facilities) and negative (public 
libraries) influences on apartment prices. This finding is 
important, because while public libraries were statistically 
significant in the spatial regression model, they were insig-
nificant in the OLS one. This indicates that spatial re-
gression analysis considers spatial effects and must be ap-
plied to measure the spillover effects of sites with a specific 
space.
Dependent variable (single-unit housing)
OLS
SLM
SEM
Constant
6.52325e + 008***
5.36521e + 008***
6.5393e + 008
Educational factors
Public library
-2.79845e + 007
-2.76873e + 007
-2.86462e + 007
Private educational facility
2.79395e + 008***
7.4858e + 006***
7.5038e + 006***
Cultural factors
Art museum
258232550.280***
2.68802e + 008***
2.68357e + 008***
Museum
6.32358e + 007
6.68108e + 007
6.65248e + 007
Historical site
-3.71002e + 007***
-3.78481e + 007**
-3.84802e + 007**
Number of urban parks
-6.80577e + 006
-6.92999e + 006
-6.8878e + 006
Urban park area
-6.46759
-5.90286
-5.8235
Rho
0.0890597
Lambda
0.10986
R2
0.293818
0.303476
0.298736
Adj. R2
0.281906
Log-likelihood
-9087.08
-9084.95
-9086.07
Akaike information criterion
18190.2
18187.9
18188.1
Schwarz criterion
18222.5
18224.3
18220.5
Koenker-Bassett
2.5375
Breusch-Pagan
12.4651
11.8637
2.0222
White test
41.6741
Moran’s I
1.5495
Lagrange multiplier (lag)
4.6786*
Robust LM (lag)
3.9209*
Lagrange multiplier (error)
2.1215
Robust LM (error)
1.3638
Likelihood ratio test
4.2727*
2.0222
Note: * p < .05, ** p < .01, *** p < .001. OLS: ordinary least squares; SLM: spatial lag model; SEM: spatial error model; LM: Lagrange multiplier.
Table 7. Analysis of single-unit housing
Dependent variable (apartment)
OLS
SLM
SEM
Constant
4.76082e + 008***
4.84887e + 008***
4.76125e + 008***
Educational factors
Public library
-2.39781e + 007
-2.39659e + 007*
-2.39323e + 007*
Private educational facility
2.5649e + 006***
2.56447e + 006***
2.5631e + 006***
Cultural factors
Art museum
4.78958e + 007
4.83697e + 007
4.81688e + 007
Museum
3.89872e + 007
3.87268e + 007
3.88224e + 007
Historical site
-5.82057e + 006
-5.82256e + 006
-5.81122e + 006
Number of urban parks
-678318
-668196
-678375
Urban park area
-6.0852**
-6.10283**
-6.09184**
Table 8. Analysis of apartments


===== PAGE 11 (score 23, numeric-paren 3, table-words 6, var-words 6) =====

Minki Sung and Junghoon Ki
Journal of People, Plants, and Environment Vol. 24, No. 5, 2021∙539
Conclusion
We analysed Seoul Metropolitan City’s dong-unit data-
base and obtained the following results. First, while the 
SEM is suitable for analysing multi-unit housing prices, 
the SLM is suitable for single-unit housing and apartment 
prices. Second, private educational facilities and art museums 
have positive influences on most housing prices, while historical 
sites have a negative influence. Finally, public libraries and urban 
park areas have negative effects on apartment prices. Thus, this 
study can imply that historical sites, public libraries, and urban 
park areas can be interpreted as less influential factors in 
housing prices compared to private educational facilities 
and art museums.
Table 9 presents a comparison of our research results with 
those of Ki and Sung (2020). Both studies adopted the same 
independent variables but at different spatial scales: the res-
idents’ immediate neighbourhood (dong) and the residential 
district (gu), respectively. The test of the SLM is more robust 
in terms of residents’ immediate neighbourhood (dong) analysis, 
especially between single-unit housing and apartments, sug-
gesting that the most proximate public facilities can be con-
sidered a relevant factor in explaining residential housing 
prices, which is also consistent with the results of Wittowsky 
et al. (2020).
Based on this comparison, we suggest the following policy 
implications. First, private educational facilities positively 
influence all types of housing at the gu and dong levels, 
which is a finding consistent with most of the previous studies 
(Cho et al., 2020; Ha and Lee, 2013; Kim and Kim, 2019). 
The residents of Seoul closely associate private educational 
facilities with students’ academic performance and consider 
them as essential educational facilities (Ha and Lee, 2013). 
Consequently, parents of students tend to relocate close to 
private educational facility clusters (Park and Lee, 2011) to 
allow easy access to education for their children, thereby 
positively influencing housing prices in Seoul at both the 
gu and dong levels.
Second, art and general museums have positive effects on 
the prices of all housing types in a residential district (gu). 
Art museums have a positive effect on single- and multi-unit 
housing prices in residents’ immediate neighbourhood (dong). 
In particular, art museums positively affect social and edu-
cational environmental satisfaction, and in turn, residential 
environmental satisfaction (Ki and Sung, 2020). For example, 
a recent study (Han, 2020) suggested that art museums have 
Dependent variable (apartment)
OLS
SLM
SEM
Rho
-0.0169392
Lambda
-0.006901
R2
0.171184
0.171330
0.171206
Adj. R2
0.157204
Log-likelihood
-8803.26
-8803.23
-8803.253223
Akaike information criterion
17622.5
17624.5
17622.5
Schwarz criterion
17654.9
17660.9
17654.9
Koenker-Bassett
6.2613
Breusch-Pagan
26.1987***
26.0083***
26.1088***
White test
36.1292
Moran’s I
-0.0090
Lagrange multiplier (lag)
0.0560
Robust LM (lag)
0.2264
Lagrange multiplier (error)
0.0083
Robust LM (error)
0.1787
Likelihood ratio test
0.0528
0.0079
Note: * p < .05, ** p < .01, *** p < .001. OLS: ordinary least squares; SLM: spatial lag model; SEM: spatial error model; LM: Lagrange multiplier.
Table 8. Continued


===== PAGE 12 (score 15, numeric-paren 0, table-words 1, var-words 13) =====

Effects of Educational and Cultural Facilities on Housing Prices in Seoul from an Accessibility Perspective
540∙Journal of People, Plants, and Environment Vol. 24, No. 5, 2021
gained in popularity as art education program venues for 
nearby elementary schools. Therefore, the art museums in Seoul 
serve not only as exhibition spaces but also as educational 
facilities. Additionally, another study (Seresinhe et al., 2016) 
showed that the image of art represented by art museums 
can affect housing prices. These findings indicate that the 
accessibility of art and general museums reflects residents’ 
needs and preferences, and in turn, influences housing prices.
Third, public libraries negatively influence apartment prices 
at both the gu and dong levels. This implies that the percep-
tion of public libraries needs to change from ‘book lending 
library’ to ‘community cultural complex’ (Noh and Kim, 2019). 
Notably, residents regard public libraries as community spaces 
for cultural activities (Seoul Metropolitan City, 2019). If public 
libraries do not meet their expectations, they lose their appeal 
and fail to motivate residents to work within the community 
for cultural reinforcement (Adkins and Hussey, 2006), which 
would have an adverse effect on apartment prices. Although 
apartment residents accept public libraries as a type of edu-
cational facility, they consider them less important than pri-
vate educational facilities. Therefore, public libraries are re-
quired to serve as both educational and cultural facilities for 
residents. Enduring efforts and prompt support from both local 
and central governments are necessary to achieve this con-
version for public libraries. Beyond this, policy-makers could 
adopt measures to transform public libraries from basic fa-
cilities to complex systems with various educational and 
cultural programs (Gu, 2019; Seoul Metropolitan City, 2019).
Fourth, historical sites negatively influence the prices of 
most housing types, at both the gu and dong levels. Historical 
sites are often perceived as a hindrance to local physical 
development (Been et al., 2014) because the disadvantages 
of construction restrictions appear to outweigh their cultural 
benefits, thus adversely affecting social environment sat-
isfaction (Ki and Sung, 2020). In other words, historical 
sites fail to attract public interest as cultural heritage sites 
(Hyundai Research Institute, 2018), due to the lack of under-
standing of the inherent value and information they provide 
(Korea Cultural Heritage Administration, 2018). To address 
this, public efforts must be made to improve the quality of the 
interpretive content of signs, interpretation of cultural heritage, 
and relevant cultural programs to help residents appreciate 
the value of historical sites.
Finally, urban park areas negatively affect the prices of all 
housing types at the gu level, and of apartment prices at 
the dong level. Although urban parks were initially created to 
meet residents’ needs (Kim et al., 2006), negative external-
ities, such as noise or disturbances from frequently used com-
munity sports facilities in the park, can cause inconvenience 
to residents, adversely affecting cultural environmental sat-
isfaction (Ki and Sung, 2020) and in turn, the housing pri-
ces in the area (Lin, 2016). For instance, apartment resi-
dents often register civil complaints against park users due 
to noise disturbances (Sung and Ki, 2021). In particular, 
considering that most apartments in Seoul Metropolitan 
City have their own green spaces, urban park areas - which 
negatively affect apartment prices at the dong level - seem 
unnecessary. Therefore, urban park planners must find al-
ternative means to minimize negative externalities, such as 
employing landscape treatments to screen out disturbances 
(More et al., 1988).
Despite the significance of this study, we acknowledge its 
limitations. First, we assumed that the ‘other location factors’ 
in the spatial dataset are similar, because Seoul’s dong-level 
Dependent Variables
Spillover Effects of Independent Variables
Best Model
Positive Effects
Negative Effects
Residents’ immediate 
neighbourhood 
(dong)
Multi-unit housing
SEM
Private educational facility, art museum
Historical site
Single-unit housing
SLM
Private educational facility, art museum
Historical site
Apartment
SLM
Private educational facility
Public library, urban park area
Residential district 
(gu)
Multi-unit housing
SEM
Private educational facility, art museum, museum
Historical site, urban park area
Single-unit housing
SEM
Private educational facility, art museum, museum
Historical site, urban park area
Apartment
SEM
Private educational facility, art museum, museum Public library, historical site, urban park area
Note: SLM: spatial lag model, SEM: spatial error model.
Table 9. Comparison of research results