Residential Heating Method and Housing Prices: Results of an Empirical Analysis in South Korea

Abstract

This study empirically delves into whether residential heating methods significantly affect apartment prices in Uiwang City, a suburban city near the Seoul Metropolitan area, South Korea. Using data from 1256 apartment sales, where both district heating systems (DHSs) and individual heating systems (IHSs) coexist, a hedonic price equation was estimated to analyze the impact of the heating method choices on housing values. Various housing attributes, including physical, locational, and environmental factors, were controlled, and multiple regression models were compared to identify the best-performing specification. The results show that apartments equipped with a DHS are priced, on average, KRW 92 million (USD 72 thousand) higher than those with an IHS. The price difference corresponds to KRW 849 thousand (USD 665) per m² and possesses the statistical significance at the 5% level. Moreover, it is quite meaningful, representing roughly 11.2% of the price of an average apartment. These findings suggest that the use of DHS has a positive effect on apartment prices that reflect consumers’ preferences, beyond its advantages in stable heat supply and energy cost savings. This article provides empirical evidence that DHS can serve as an important urban infrastructure contributing to asset value enhancement. Although this study is based on a specific geographic area and caution must be exercised in generalizing its findings, it reports the interesting finding that residential heating method significantly affects housing prices.

1. Introduction

Analyzing the factors that influence housing prices is critical for understanding housing market trends and providing valuable information to governments in formulating housing supply policies and to consumers considering purchasing apartments. Housing prices are determined by a variety of factors, including locational attributes, environmental conditions, transportation accessibility, and energy supply infrastructure. In particular, district heating systems (DHSs) have emerged an important determinant of real estate value by enhancing energy efficiency, reducing heating costs, and contributing to the reduction in greenhouse gas (GHG) emissions in South Korea [1].

Combined heat and power (CHP) plants produce electricity and heat simultaneously. DHSs based on CHP plants offer higher energy efficiency and more environmentally friendly characteristics compared to individual heating systems (IHSs) [2]. The transition from a traditional centralized heating system (CHS) to either an IHS or a DHS has contributed to reductions in energy consumption and GHG emissions. Notably, DHSs minimize energy loss by utilizing the surplus heat from power generation and enables large-scale, economically efficient heat production and distribution [3]. Moreover, the New European Bauhaus initiative under the European Green Deal emphasizes the role of DHSs in creating sustainable and inclusive urban environments [4].

Recognizing these advantages, the South Korean government has actively promoted the expansion of DHSs. DHSs were first introduced experimentally in Mok-dong, Seoul, the capital of the country, in 1985. Its expansion accelerated in 1987 when Korea Electric Power Corporation (Naju-si, Republic of Korea) repurposed the Seoul Thermal Power Plant (currently operated by Korea Midland Power Co., Ltd., Boryeong-si, Republic of Korea) to supply district heating using CHP-based surplus heat [5]. Since then, DHSs have been extended to first-generation new towns near Seoul (Bundang, Ilsan, Pyeongchon, Sanbon, and Jungdong), as well as major cities including Daegu, Busan, and Cheongju. However, DHS adoption remains limited outside designated district energy supply areas, with many apartment complexes continuing to rely on IHSs. Even in apartment complexes undergoing redevelopment, the adoption of DHSs faces challenges despite the proximity of heat transmission pipelines, mainly due to the high initial installation costs, the burden of maintaining existing IHS infrastructure, and a lack of consumer awareness. Consequently, DHSs and IHSs are currently in competitive coexistence [6].

This phenomenon is not unique to South Korea but has also been identified globally as a structural constraint on DHS expansion. Nielsen et al. [7] emphasized that existing investments in IHS infrastructure and the autonomy it offers consumers are major obstacles to the adoption of DHSs. They pointed out that the high initial investment cost and technical barriers associated with transitioning to DHSs limit consumer demand. Ahvenniemi et al. [8] similarly highlighted the need for shifts in consumer perception, noting that the long-term economic and environmental benefits of DHS are often inadequately communicated. They argued that trust-building and strategic marketing are essential for expanding DHS adoption, especially given that consumers accustomed to IHS value convenience and autonomy, which act as major barriers to transition. Meanwhile, some studies have reported that the proximity to energy supply facilities such as CHP plants may negatively impact real estate values due to concerns such as noise, air pollution, and visual impacts [9,10,11].

However, empirical analyses demonstrating the positive effects of the economic and environmental benefits of DHSs on property values remain relatively scarce. In light of these circumstances, this study empirically investigates the impact of heating systems—specifically, DHSs versus IHSs—on apartment prices in Uiwang City, a new town located near the Seoul Metropolitan area (SMA) in South Korea. Uiwang presents a suitable case for analysis as it features a mix of apartments that have adopted DHSs or IHSs, allowing for meaningful comparisons of price differences based on residential heating methods.

The prime purpose of this research is, therefore, to quantitatively verify whether heating methods affect housing prices and, if so, how much; this is achieved through the specific case of Uiwang City. The research attempts to estimate the housing price equation using hedonic price model. In the model, residential heating method is treated as one factor affecting housing prices. In other words, the dependent variable of the hedonic price model is housing price, and the independent variables include neighborhood variables and housing characteristic variables other than heating method. If the estimated coefficient for the heating method variable is statistically significant, the heating method has a significant effect on housing prices, all other things being equal.

Given the paucity of prior research examining the impact of heating methods on housing prices, this paper represents an important academic contribution. This is the interesting point of the paper. There are three sections in the remainder of this article. The methodology adopted for this study is described in the next section. The methodology will be covered in the following order: method, literature review, and model. The Section 3 presents the data adopted, the empirical results, and some discussions. The Section 4 provides some concluding remarks.

2. Method, Literature Review, and Model

2.1. Method: Hedonic Price Model

The hedonic hypothesis states that although certain attributes of a product such as a house are not traded on the market and it is therefore difficult to know its price, the implicit price of a specific attribute can be estimated from the housing price equation. This implicit price is also called by other names, such as the shadow price, marginal value, etc. When the product whose value is to be estimated is an attribute of another product, the hedonic price model can be usefully applied.

The price of housing is determined by a complex interplay of various physical, locational, and environmental attributes. To quantitatively isolate and estimate the values of these attributes, the hedonic price model (HPM) has been widely applied [12,13]. The HPM is regarded as an effective tool for systematically identifying the implicit prices of environmental attributes and their effects on housing values and related market dynamics. In particular, it has been extensively used in environmental economics and real estate economics to analyze how various environmental factors—such as air quality, noise, water quality, and the presence of hazardous facilities—are capitalized into housing prices.

2.2. Literature Review

The authors intend to conduct a literature review from two perspectives: prior studies applying HPM and prior studies dealing with DHS and housing prices.

Table 1. Summary of several previous studies dealing with the hedonic price model (HPM) or the effects of heating method on housing prices.

Categories	Sources	Methods	Countries and Regions	Main Results
Related to HPM	McCord et al. [14]	Space-based HPM	Denver, CO, USA	Environmental degradation negatively affects housing prices; environmental factors are capitalized.
	Carruthers et al. [15]	Space-based HPM	Puget Sound, WA, USA	Distance from pollution sources affects housing prices.
	Tsao et al. [16]	HPM	Vicinity of Taoyuan Airport, Taiwan	Aircraft noise significantly reduces housing prices; differentiated impact depending on noise levels.
	Cohen et al. [17]	Spatial HPM	Vicinity of Hartsfield-Jackson Atlanta Airport, USA	Aircraft noise lowers prices; proximity to airport raises prices.
	Sharaan et al. [18]	HPM	Red Sea Resort Area, Egypt	Coastal erosion reduces resort revenue by 3–33% by 2100 year.
	Nguyen et al. [19]	HPM	Hoi An World Heritage Beach Area, Vietnam	Coastal erosion significantly impacts tourism revenue.
	Overbeek et al. [20]	HPM	Amsterdam and major cities, The Netherlands	Environmental certification adds about 10.3% rent premium.
	Vimpari [21]	HPM	Helsinki and other cities, Finland	Houses using ground source heat pumps sell at a 5.33% premium.
Related to district heating system (DHS) and housing Prices	Millar et al. [22]	Case study analysis	Glasgow, Scotland, UK	Described barriers to DHS expansion; emphasized energy efficiency benefits.
	Kuru et al. [23]	Levene’s test	İzmir, Turkey	Central cooling system significantly impacts housing prices.
	Odgaard et al. [24]	Case study analysis	Denmark, Sweden, Estonia	DHS could influence housing prices through regulation.
	Lee [25]	HPM	Seoul Metropolitan Area, Republic of Korea	Analyzed density impacts; did not address heating method effects.
	Kim et al. [26]	Multiple regression analysis	Seoul and New Towns, Republic of Korea	Analyzed determinants without including heating system variables.
	Kim et al. [27]	Contingent valuation	Nationwide, Republic of Korea	Measured willingness-to-pay (WTP) for DHS; not direct housing price analysis.
	Yoon et al. [28]	Contingent valuation	Seoul and Incheon, Republic of Korea	Measured preference and WTP for DHS; no direct housing price analysis.
	Kim et al. [29]	HPM	Seoul, Republic of Korea	DHS apartments show a 21.7% price premium over HIS or CHS.

shows a summary of several previous studies dealing with the HPM or the effects of heating method on the housing prices. McCord et al. [14] examined how environmental health factors such as air quality, water quality, toxic emissions, and traffic volume influenced housing prices in the Denver metropolitan area in the United States (US), confirming that various environmental indicators had statistically significant effects. Carruthers et al. [15] applied an advanced HPM to analyze the impact of proximity to hazardous facilities on housing prices. Tsao et al. [16] analyzed the impact of aviation noise around Taiwan Taoyuan International Airport on housing prices. Their study quantified the negative externalities of noise pollution on housing values using HPM, demonstrating a significant negative correlation between high noise levels and housing prices.

Cohen et al. [17] employed a spatial HPM to look into housing prices around Atlanta Airport. They found that properties exposed to noise levels above 70 dB sold for 20.8% less, while proximity to the airport itself had a positive effect on prices, suggesting that while environmental factors may exert negative influences, locational advantages such as transport accessibility can offset such effects.

In a broader context, Sharaan et al. [18] investigated the effects of sea level rise on tourism revenues and real estate values along Egypt’s Red Sea coast, showing that environmental changes can significantly impact property markets. Similarly, Nguyen et al. [19] examined how coastal erosion affected tourism and real estate markets in Hoi An, a UNESCO World Heritage Site in Vietnam, demonstrating that environmental degradation causes substantial economic losses in both sectors. Overbeek et al. [20] evaluated the value-added of environmental certification in the Dutch office market using a HPM, revealing that certified buildings commanded higher rents than uncertified ones. These findings imply that sustainable energy infrastructures, such as DHSs, which are recognized as part of efforts to address climate change under the United Nations Framework Convention on Climate Change framework, can positively influence property values through energy efficiency and environmental benefits.

Vimpari [21] conducted an empirical analysis in Finland, finding that detached houses equipped with ground source heat pumps enjoyed an average price premium of 5.33%. This result illustrates that energy efficiency is reflected in consumer housing preferences and is capitalized into market prices, providing strong evidence that efficient heating systems can enhance real estate value.

Particularly relevant to DHS, Millar et al. [22] further emphasized that advancements in DHS technologies across different generations contributed to improving energy efficiency and reducing GHG emissions, thereby delivering both cost savings and environmental benefits to residential properties. Kuru et al. [23] analyzed apartment sales cases in İzmir, Turkey, and found that central cooling systems had a significant influence on housing price formation. Odgaard et al. [24] compared and analyzed DHS pricing regulation systems in Europe and explained that DHS could indirectly influence housing markets beyond their role as mere heat supply infrastructures, although no specific numerical estimates were proposed.

In the context of South Korea, Lee [25] examined determinants of housing prices focusing on apartment density and consumer preferences in the metropolitan area but did not explicitly investigate price differences based on heating system types. Similarly, Kim et al. [26] analyzed segmentation and determinants in the housing markets of Seoul and neighboring new towns yet did not explicitly address the relationship between heating system technology and housing prices.

Kim et al. [27] estimated the willingness-to-pay (WTP) for DHSs compared to IHSs in South Korea, finding that consumers were willing to pay an average premium of 8.5% relative to current heating costs for adopting the former over the latter. This result demonstrated that consumer preferences for heating systems could be expressed as economic value; however, the study did not directly examine the impact on housing market prices. Yoon et al. [28] also estimated WTP regarding the convenience of DHSs among residents in Seoul and Incheon, quantitatively illustrating consumer preferences for different heating methods. In contrast, the study by Kim et al. [29] stands out as a rare empirical analysis in South Korea that directly examines price differences between DHS- and IHS-equipped apartments in Seocho District, revealing that apartments with DHSs command a significant price premium of approximately 21.7%, a result found to be statistically significant.

While the overall share of DHSs in Seoul and Incheon remains relatively low, approximately 19.1% and 32.7% as of 2023 [30], prior studies have addressed this limitation by focusing on specific regions such as Seocho and Bupyeong, where the proportions of DHSs and IHSs are more balanced [28,29]. Following this approach, the present study selects Uiwang City, where the DHS penetration rate is around 57%, thereby providing a suitable case for comparative analysis between heating methods without introducing bias due to disproportionate system distribution. In addition, this study considers the policy context of South Korea’s public housing supply initiatives. The Korea Ministry of Land, Infrastructure, and Transport announced the “Public-Led 3080+” plan to expand housing supply in the SMA, including the development of new public housing districts in Uiwang-Gunpo-Ansan, scheduled for completion by the second half of 2031 [31]. This study seeks to provide insights into whether the adoption of DHSs should be considered in such developments, offering useful suggestions for policymakers.

After reviewing the existing literature, two important limitations and research gaps were identified. First, studies analyzing various physical and environmental factors affecting housing prices using the hedonic pricing model are relatively numerous. However, research that utilizes heating method as a key independent variable is relatively rare. In particular, there are almost no empirical studies that analyze the impact of DHSs on housing prices. There are only a few studies that compare the differences in housing prices between DHSs and IHSs in areas where one or the other is adopted. Furthermore, they are primarily focused on large cities, and there is a lack of empirical research targeting suburban cities where new town developments are planned. Second, many existing studies estimate the economic value of heating methods using data on WTP gathered from surveys, which are useful for understanding consumer perceptions but have limitations regarding capturing the implicit values or shadow prices reflected in actual market transaction prices. Therefore, this study aims to fill the research gaps in the existing literature by selecting Uiwang City, a suburban area in the metropolitan region of South Korea where DHSs and IHSs are suitably mixed, as the case study. By applying various statistical analysis techniques, the study seeks to quantitatively analyze the impact of heating methods on apartment sales prices.

2.3. Model

The HPM is an economic method used to analyze the contributions of individual attributes to the determination of the price of a product or service that is composed of multiple attributes. It has been widely applied in studies investigating the determinants of housing prices. The HPM is based on the consumer theory proposed by Lancaster [32], who argued that consumers derive utility not from a product itself but from the attributes inherent in the product. Building on this idea, Rosen [12] explained that the price of a product is determined by the value of the individual characteristics that compose it. In the context of the housing market, the price of an apartment is influenced by various factors, including its physical characteristics (such as location, floor area, floor level, and availability of amenities) and surrounding environmental factors (such as transportation accessibility, quality of school districts, and proximity to public facilities). The HPM enables the empirical estimation of the contribution of each factor to the overall price formation. The hedonic price equation, which models housing prices, is generally expressed as

$$P_i=f\left(X_{i1}, X_{i2},\dots ,X_{ik}\right)+{\epsilon }_i$$

(1)

where $P_i$ denotes the price of house $i$; $X_{ik}$ represents the explanatory variables corresponding to the individual characteristics of house $i$; and ${\epsilon }_i$ is the error term, capturing factors that influence housing prices but are not explicitly explained by the model.

Unfortunately, economic theory does not provide any information about the form of $f(·)$. Thus, researchers must make arbitrary assumptions about its form. In this study, three commonly used functional forms were adopted: the linear function, the semi-logarithmic function, and the double-logarithmic function. The linear function has the following form:

$$P_i=\alpha +\sum {\beta }_k{X}_{ik}+{\gamma DHS}_i+{\epsilon }_i$$

(2)

where $\alpha$ is a constant term, ${\beta }_k$ represents the coefficient associated with the explanatory variable $X_{ik}$, ${DHS}_i$ is a dummy variable indicating whether house $i$ adopts a DHS, and $\gamma$ represents a corresponding coefficient.

Equation (2) describes the relationship between housing price and related attributes in the simplest form, offering ease of interpretation. However, it may exhibit lower explanatory power when the relationship between variables is inherently nonlinear. A logarithmic form reflecting nonlinearity can be considered. The semi-logarithmic function applies a logarithmic transformation to either the dependent variable or the continuous independent variables, allowing nonlinear relationships to be converted into a linear form. Since we cannot take logs of dummy variables, we can only take logs of continuous variables. This transformation helps reduce data skewness and enables more intuitive interpretation. The semi-log function is generally expressed as

$$P_i=\alpha +\sum {\beta }_k{\mathrm{l}\mathrm{n}X}_{ik}+{\gamma DHS}_i+{\epsilon }_i$$

(3)

$$\ln P_i=\alpha +\sum {\beta }_k{X}_{ik}+\gamma {DHS}_i+{\epsilon }_i$$

(4)

Moreover, the double-logarithmic function applies logarithmic transformations to both the dependent and continuous independent variables, making it particularly useful for analyzing the elasticity between variables. This functional form clearly illustrates the proportional relationships between price and attributes and is generally expressed as:

$$\ln P_i=\alpha +\sum {\beta }_k{\mathrm{l}\mathrm{n}X}_{ik}+\gamma DH{S}_i+{\epsilon }_i$$

(5)

In estimating the HPMs, the least squares (LS) estimation method is usually used. The LS estimation method is the most widely applied in the literature. It obtains coefficient estimates by minimizing the sum of the squared residuals between observed data and predicted values. The LS estimates for Equation (2) are derived by solving the following equation:

$$\mathrm{m}\mathrm{i}\mathrm{n}\sum _{i=z}^n{\epsilon }_i^2=\mathrm{m}\mathrm{i}\mathrm{n}\sum _{i=1}^n{\left(P_i-\left(\alpha +\sum {\beta }_k{X}_{ik}+\gamma {DHS}_i\right)\right)}^2$$

(6)

While the LS estimation is computationally simple and offers clear interpretability, it is sensitive to outliers. The presence of extreme values in the dataset can lead to inconsistent coefficient estimates. In this case, the least absolute deviation (LAD) estimation method can be employed as an alternative. The LAD estimates are obtained by minimizing the absolute values of the residuals, providing robust results that are less sensitive to outliers [33]. While LS estimates can be significantly influenced by outliers, the LAD estimation method offers more robust coefficient estimates even when outliers are present in the dataset. Moreover, the LAD estimation method does not require the assumption of normality, making it applicable even when the data distribution deviates from normality [34]. The LAD estimates for Equation (2) are derived by solving the following equation:

$$\mathrm{m}\mathrm{i}\mathrm{n}\sum _{i=z}^n\left|{\epsilon }_i\right|=\mathrm{m}\mathrm{i}\mathrm{n}\sum _{i=1}^n\left|P_i-\left(\alpha +\sum {\beta }_k{X}_{ik}+\gamma {DHS}_i\right)\right|$$

(7)

In fact, since both linear and logarithmic forms can constrain the form of a function, a more flexible form of the function needs to be assumed. In such cases, the Box–Cox transformation can be utilized [35]. The Box–Cox transformation takes the following forms:

$$P_i^{\left(\lambda \right)}=\left\{\begin{array}{cc}{\displaystyle \frac{P^{\lambda }-1}{\lambda }}, & \mathrm{i}\mathrm{f} \lambda \ne 0\\ \ln P_i , & \mathrm{i}\mathrm{f} \lambda =0\end{array}\right.$$

(8)

Although the Box–Cox transformation can be applied to independent variables as well, this study only considers the Box–Cox transformation for the dependent variable for the convenience of estimation. Through the Box–Cox transformation, the flexibility of the functional form can be improved, thereby enhancing the predictive performance of the model. In particular, when $\lambda =0$, $P_i^{\left(\lambda \right)}$ becomes $\mathit{ln}{P}_i$ [36]. Accordingly, Equations (9) and (10) are also estimated.

$$P_i^{\left(\lambda \right)}=\alpha +\sum {\beta }_k{X}_{ik}+\gamma {DHS}_i+{\epsilon }_i$$

(9)

$$P_i^{\left(\lambda \right)}=\alpha +\sum {\beta }_k\mathrm{l}\mathrm{n}{X}_{ik}+\gamma DH{S}_i+{\epsilon }_i$$

(10)

In summary, Equations (2)–(5) are estimated using the LS and LAD estimation methods. Furthermore, Equations (9) and (10) are estimated using the maximum likelihood estimation method. Therefore, a total of 10 equations were estimated. The purpose of estimating these 10 equations is to select the HPM that best explains the data. This is because the value of $\gamma$, that is, the effect of residential heating method on housing prices, can vary depending on which functional form or estimation method is used.

3. Data, Results, and Discussion

3.1. Data

This study collected sales price data for apartment complexes located in Uiwang City, Gyeonggi Province, to empirically analyze the impact of residential heating methods (DHS vs. IHS) on apartment price. In South Korea, there are other types of housing, such as townhouses and detached houses, in addition to apartments. However, since the housing type that can adopt DHSs is limited to apartments, this study only considers apartments and no other types of housing. Although there are some apartments that have adopted central heating systems, they are only a very small number. Thus, this study only considers apartments with DHSs or IHSs.

The core hypothesis of this research is that the adoption of DHSs can have a positive effect on housing value in Uiwang City. As mentioned above, the study aims to quantitatively verify the extent to which differences in heating methods influence housing price. To this end, a variety of independent variables were established, encompassing not only the physical characteristics of the apartments but also locational and environmental factors, and the relationship with housing prices was examined through multiple regression analysis. In particular, since apartments equipped with DHSs have differentiated advantages in terms of energy efficiency and management convenience, this study focuses on identifying how such differences in heating methods affect housing prices.

To differentiate this study from previous studies, the study specifically emphasizes the selection of the study area and the effect of heating methods. While prior studies have primarily focused on major cities such as Seoul and Incheon, this study examines Uiwang City, a metropolitan area where related research has been relatively limited. With a DHS rate of approximately 57%, Uiwang provides a balanced distribution of heating systems, making it a suitable case for evaluating the impact of DHSs compared to IHSs.

Data collection was conducted in July 2023, resulting in a dataset of 1256 apartment sales listings obtained from the real estate information service system operated by Kookmin Bank. The price and various characteristics of the apartments used in this study were collected from the real estate information service provided by Kookmin Bank [37]. Additionally, supplementary data related to locational and environmental characteristics were collected from the “33rd 2022 Uiwang City Statistical Yearbook” [38]. Apartment brand reputation data were obtained from the Korea Corporate Reputation Research Institute [39].

The dependent variable of this study was set as the apartment sales price registered with Kookmin Bank, while the independent variables are classified into physical characteristics and locational/environmental factors that are expected to influence housing price. Based on the collected data, the definitions of each variable and their descriptive statistics are summarized in

Table 2. Description of variables adopted in this study.

Variables	Descriptions	Means	Standard Deviations	Expected Signs
PRICE a	Sale price for apartment (unit: KRW million)	820	330	(−)
SA	Supply area (unit: m2)	108.3	24.9	(+)
TH	Total number of households	994.6	825.9	(+)
ROOM	Number of rooms per unit	3.1	0.4	(+)
BATH	Number of bathrooms per unit	1.9	0.4	(+)
BUILT	Age of apartment (years)	15.8	9.4	(+)
HFLR	High floor dummy (1 = yes; 0 = No)	0.6	0.5	(+)
SOUTH	Main orientation to the south (1 = south-facing; 0 = other)	0.9	0.3	(+)
DHS	District heating system dummy (1 = district heating system; 0 = individual or central heating system)	0.6	0.5	(+)
NFOR	Number of foreigners per 10,000 residents	51.6	28.0	(−)
NCOM	Number of businesses per 10,000 residents	784.3	193.7	(−)
NPOOR	Number of recipients of basic living security or minimum pension per 10,000 residents	955.7	262.0	(−)
WDS	Walking distance to nearest subway station (unit: km)	2.4	1.3	(−)
PREM	Apartment premium brand dummy (1 = top 10 brand, 0 = other)	0.8	0.5	(+)

Note: ^a At the time of data acquisition, KRW 1284 was approximately equal to USD 1.

. The physical characteristic variables included supply area (SA), total number of households (TH), number of rooms (ROOM), number of bathrooms (BATH), building age (BUILT), high-floor status (HFLR), and presence of DHS. The locational and environmental characteristic variables included the number of foreign residents per 10,000 residents (FOR), number of businesses within the area per 10,000 residents (NCOM), the number of low-income households per 10,000 residents (NPOOR), walking distance to the nearest subway station (WDS), and apartment brand reputation (PREM).

The authors considered three criteria in determining the list of independent variables. First, the independent variables used in previous studies that applied the HPM presented in Table 1 were utilized. The variables were largely divided into characteristic variables directly related to housing and neighborhood variables surrounding housing. Second, the availability of data was considered. Even if a factor was certain to affect housing prices, if information on the factor by apartment unit was not available, the factor was not reflected as an independent variable. For example, the level of energy saving (in particular, thermal modernization) in residential premises may affect housing prices, but the authors cannot obtain data on this for each apartment unit. Third, the South Korean characteristics were reflected in identifying factors affecting housing prices. For example, in the country, in cases of two apartments with similar conditions, the price differs depending on which construction company built the apartment. Therefore, the brand reputation of the construction company that built the apartment was also reflected as an independent variable. The corresponding variable was PREM. This was not reflected in most previous studies.

Descriptive statistics analysis revealed that the average supply area was 108.3 m², and the average building age was 14.8 years. Among the apartments analyzed, approximately 60% were equipped with DHS. According to the correlation analysis summarized in

Table 3. Correlation matrix between sale price for apartment and variables.

Variables a	PRICE	SA	TH	ROOM	BATH	BUILT	HFLR	SOUTH	NFOR	NCOM	NPOOR	WDS	PREM	DHS
PRICE	1.00
SA	0.71	1.00
TH	0.48	0.33	1.00
ROOM	0.49	0.76	0.30	1.00
BATH	0.51	0.57	0.26	0.44	1.00
BUILT	−0.58	−0.27	−0.27	−0.15	−0.48	1.00
HFLR	0.12	0.02	0.08	0.03	0.02	−0.06	1.00
SOUTH	0.28	0.14	0.17	0.08	0.17	−0.14	0.04	1.00
NFOR	−0.44	−0.29	−0.29	−0.14	−0.30	0.29	−0.04	−0.23	1.00
NCOM	0.05	−0.02	−0.29	−0.05	−0.04	−0.07	−0.01	−0.02	0.57	1.00
NPOOR	−0.38	−0.29	−0.34	−0.19	−0.28	0.17	−0.03	−0.25	0.95	0.64	1.00
WDS	0.19	0.15	−0.19	0.04	0.17	−0.29	0.00	0.11	−0.22	0.28	−0.16	1.00
PREM	0.38	0.16	0.72	0.21	0.24	−0.39	0.07	0.19	−0.17	−0.38	−0.25	−0.41	1.00
DHS	0.61	0.35	0.42	0.22	0.33	−0.38	0.08	0.27	−0.62	−0.06	−0.59	0.24	0.22	1.00

Note: ^a These are described in Table 2.

, there was a strong positive correlation between PRICE and SA and DHS. In contrast, BUILT, FOR, and NPOOR showed negative correlations with PRICE. Notably, DHS emerged as a major factor influencing apartment prices, similar to the PREM.

These results suggest that, in addition to brand value, both the physical characteristics and locational factors of a property play important roles in housing price formation. This finding highlights the significance of analyzing the impact of DHS on housing prices, which aligns with previous studies suggesting that heating systems can act as a qualitative factor influencing residential environments.

3.2. Estimation Results of the HPMs

As addressed above, a total of ten equations were established to analyze the factors influencing apartment prices and to empirically examine the impact of DHS on housing prices. According to the Wald statistics reported in

Table 4. Results from estimating housing price equations.

Variables a	Least Squares Estimation b				Least Absolute Deviations Estimation b				Box–Cox Transformations b
	Linear-Linear	Linear-Log	Log- Linear	Log-Log	Linear-Linear	Linear-Log	Log- Linear	Log-Log	Linear	Log
Constant	−2.0893	−55.5018	0.6976	−3.7878	−1.0256	−53.1502	0.8511	−3.3305	0.5248	−12.7574
	(−2.78) **	(−10.60) **	(9.33) **	(−7.71) **	(−2.08) **	(−15.80) **	(15.84) **	(−9.29) **	(6.32) **	(−7.99) **
SA	0.0732	7.7607	0.0070	0.7669	0.0664	7.2897	0.0070	0.7920	0.0087	1.7064
	(25.08) **	(25.04) **	(23.96) **	(26.39) **	(34.66) **	(36.63) **	(33.69) **	(37.35) **	(13.81) **	(12.64) **
TH	0.0002	0.2803	0.0000	0.0460	0.0003	0.3601	0.0000	0.0464	0.0000	0.0451
	(2.27) **	(3.66) **	(1.51)	(6.41) **	(4.71) **	(7.33) **	(2.59) **	(8.86) **	(0.34)	(3.32) **
ROOM	−0.3987	0.0839	0.0098	0.1597	−0.2871	0.1084	0.0007	0.1454	0.0040	0.1862
	(−2.62) **	(0.19)	(0.65)	(3.90) **	(−2.88) **	(0.39)	(0.06) **	(4.87) **	(0.28)	(2.60) **
BATH	−0.6377	−1.7853	0.0565	−0.0070	−0.6132	−1.6087	0.0471	−0.0156	0.0503	−0.1791
	(−4.15) **	(−7.63) **	(3.69) **	(−0.32)	(−6.07) **	(−10.70) **	(4.28) **	(−0.97)	(2.77) **	(−3.27) **
BUILT	−0.0662	−0.9318	−0.0092	−0.1278	−0.0567	−0.6818	−0.0093	−0.1123	−0.0104	−0.2059
	(−9.17) **	(−9.81) **	(−12.78) **	(−14.36) **	(−11.96) **	(−11.18) **	(−17.91) **	(−17.29) **	(−9.53) **	(−7.96) **
NFOR	−0.0535	−3.6131	−0.0056	−0.3317	−0.0563	−3.6441	−0.0054	−0.3016	−0.0071	−1.0339
	(−8.29) **	(−6.86) **	(−8.65) **	(−6.71) **	(−13.27) **	(−10.77) **	(−11.60) **	(−8.37) **	(−7.04) **	(−6.54) **
NCOM	0.0062	4.7162	0.0006	0.4667	0.0064	4.8657	0.0007	0.4629	0.0007	1.2384
	(15.84) **	(14.21) **	(16.16) **	(15.00) **	(25.15) **	(22.84) **	(24.87) **	(20.39) **	(9.42) **	(9.87) **
NPOOR	0.0024	1.5551	0.0002	0.0125	0.0018	1.2273	0.0000	−0.0902	0.0004	0.5001
	(2.70) **	(1.65) *	(1.86) *	(0.14)	(3.10) **	(2.02) **	(−0.26)	(−1.40)	(3.56) **	(2.00) **
PREM	1.6988	1.4492	0.2029	0.1459	1.6926	1.6414	0.2093	0.1565	0.2549	0.4590
	(9.34) **	(7.69) **	(11.20) **	(8.25) **	(14.17) **	(13.56) **	(16.07) **	(12.13) **	(9.63) **	(8.46) **
HFLR	0.3765	0.3860	0.0379	0.0389	0.2820	0.3012	0.0424	0.0409	0.0444	0.0815
	(4.42) **	(4.52) **	(4.47) **	(4.86) **	(5.04) **	(5.49) **	(6.96) **	(7.00) **	(3.96) **	(4.25) **
SOUTH	0.4570	0.4049	0.0852	0.0758	0.3319	0.2772	0.0593	0.0596	0.0997	0.1226
	(2.89) **	(2.56) **	(5.40) **	(5.11) **	(3.20) **	(2.73) **	(5.24) **	(5.50) **	(5.06) **	(3.05) **
WDS	−0.0853	−0.6025	−0.0107	−0.0662	−0.3128	−0.8514	−0.0224	−0.0794	−0.0114	−0.2012
	(−1.95) *	(−6.08) **	(−2.45) **	(−7.13) **	(−10.90) **	(−13.39) **	(−7.17) **	(−11.72) **	(−1.92) *	(−7.64) **
DHS	0.9967	0.9876	0.1382	0.1259	0.7115	0.6749	0.1004	0.1062	0.2431	0.2402
	(6.63) **	(6.25) **	(9.24) **	(8.49) **	(7.21) **	(6.65) **	(9.34) **	(9.81) **	(12.61) **	(5.76) **
$\lambda$									0.0922	0.3378
									(3.90) **	(12.14) **
Adjusted-$R^2$ c	0.807	0.806	0.871	0.886	0.795	0.798	0.869	0.883	0.536	0.530
Wald Statistics (p-values)	44.01	39.01	74.33	63.58	52.03	44.18	78.87	86.59	154.68	30.80
	(0.000) **	(0.000) **	(0.000) **	(0.000) **	(0.000) **	(0.000) **	(0.000) **	(0.000) **	(0.000) **	(0.000) **
RMSPE d	19.24	17.76	9.82	7.91	17.99	16.30	10.03	7.79	10.53	11.11

Notes: ^a These are described in Table 2. ^b The t-values are presented in the parentheses beneath the estimates; ** and * denote statistical significance at the 5% and 10% levels, respectively. ^c The values for Box–Cox transformation indicate McFadden’s pseudo-$R^2$. ^d RMSPE denotes the root mean squared percentage error.

, the null hypothesis (that the model is meaningless or not statistically significant) was rejected at the 5% level across all models, indicating that all models secure statistical significance. Any equation that does not have statistical significance should not be used and should be discarded. Thus, all ten equations were able to be used for further investigation.

Notably, the DHS variable, along with the PREM variable, consistently exhibited a statistically significant and positive effect on apartment prices across all models. This finding clearly suggests that, under the condition that the levels of other characteristics are constant, apartments equipped with a DHS tend to be more expensive than those with an IHS. It also implies that benefits associated with DHSs, such as heating cost savings, stable heat supply, and environmental friendliness, are likely to be positively evaluated in the housing market. This is an interesting finding of this study.

As a next task, the most appropriate one among the ten equations needs to be selected. One complication involved in this selection is that the form of the dependent variable varies across models, making it difficult to compare the explanatory or predictive power of the dependent variable across models. A criterion that can be usefully employed in such a case is the root mean squared percentage error (RMSPE). The RMSPE is defined as

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{P}\mathrm{E}=100\sqrt{\sum _{i=1}^N{{\displaystyle \frac{1}{N}}\left({\displaystyle \frac{P_i-P_i^p}{P_i}}\right)}^2}$$

(11)

where $N$ is the total number observations and $P_i^p$ is a predicted value of housing price.

The RMSPE is calculated using three steps. First, the average of the squared percentage differences between actual housing price and predicted housing price is computed. Second, the square root is obtained. Third, it is multiplied by 100. Thus, RMSPE is expressed in a percentage. The smaller RMSPE is, the higher the predictive power. The smallest RMSPE was 7.79. Interestingly, this was observed in the log–log equation estimated using the LAD estimation method, which is known to be robust to the effects of outliers.

The number of bathrooms and the proportion of low-income households were found to be statistically insignificant. This suggests that these variables may not have a direct impact on apartment price determination. In contrast, apartments with larger supply areas, a higher number of households, premium brand status, proximity to subway stations, high-floor units, and south-facing orientations tended to exhibit higher prices. This finding aligns with previous studies indicating that apartments possessing highly preferred attributes in the housing market tend to command a price premium.

Notably, the coefficient for the DHS variable recorded quite a high t-value of 9.81 and was statistically significant at the 1% level. This result empirically confirms that DHSs play an important role in housing price formation, with apartments equipped with DHSs tending to achieve higher prices compared to those with IHSs. In other words, consumers do not merely perceive DHSs as a functional difference in heating methods, but rather as an important attribute that enhances the quality of the residential environment.

One point to pay attention to when interpreting the estimation results of the housing price equation is the presence of multicollinearity. If the correlation between one independent variable and another independent variable is too large, avoiding multicollinearity is difficult. The existence of multicollinearity tends to weaken the statistical significance of the estimated results by increasing the standard error of the estimated coefficients and producing low t-values. Therefore, the authors would like to examine the possibility of presence of multicollinearity.

When looking at the correlation coefficients presented in Table 3, the correlation coefficients between independent variables are large for ROOM and SA (0.76) and NPOOR and NFOR (0.95). The large correlation between ROOM and SA is intuitively natural. The number of rooms tends to be large when the area of an apartment is large. In South Korea, apartments with many rooms are preferred even if they have the same area. To reflect this preference in dealing with the HPM, both variables need to be included in the housing price equation. Since most of the foreigners living in Uiwang City are low-income workers from Southeast Asia, they tend to live in underdeveloped areas, which is why a large correlation between NPOOR and NFOR is observed. The authors considered excluding one of the two variables from the housing price equation.

This is because the problem of multicollinearity can occur when two variables with a high correlation are used as independent variables. When multicollinearity occurs, the variance of the estimated coefficients increases, which reduces their t-values, making it difficult for the estimated coefficients to secure statistical significance. However, as can be confirmed in Table 4, the estimated coefficients for both variables are generally statistically significant. The estimated coefficients for NFOR are statistically significant in all ten housing price equations, and those for NPOOR are statistically significant in seven out of ten housing price equations. It is difficult to see a problem of multicollinearity. Therefore, the authors believe that it is difficult to see a problem of multicollinearity in the housing price equations estimated in this study. Both variables were included in the housing price equation. The not high correlation coefficients among the remaining independent variables do not raise concerns about multicollinearity.

3.3. Discussion of the Results

Because a log–log housing price equation is used, the price difference between apartments with a DHS and those with an IHS must be carefully estimated. The quantitative difference in apartment sales prices was calculated based on the method proposed by Halvorsen et al. [40]. The price effect of a DHS can be directly interpreted using the estimated regression coefficient of the DHS variable (${\beta }_{DHS}$), denoted as ${\widehat{\beta }}_{DHS}$, and the price difference between apartments with DHS ($P_1$) and those with IHS ($P_0$) is expressed as follows:

$$P_1-P_0=P_0\left\{\mathrm{e}\mathrm{x}\mathrm{p}\left({\widehat{\beta }}_{DHS}\right)-1\right\}$$

(12)

However, Kennedy [41] pointed out that when an independent variable is a dummy one, biased estimates may arise when calculating marginal effects. To address this issue, he proposed a correction formula. This correction incorporates the estimated variance of the dummy variable coefficient ($\widehat{V}\left({\widehat{\beta }}_{DHS}\right)$) to reduce the bias of the original estimate. In this study, Kennedy’s [41] correction was applied to quantitatively analyze the difference in apartment sales prices based on the adoption of DHS. The difference takes the following form:

$$P_1-P_0=P_0\left\{\mathrm{e}\mathrm{x}\mathrm{p}\left({\widehat{\beta }}_{DHS}-0.5\widehat{V}\left({\widehat{\beta }}_{DHS}\right)\right)-1\right\}$$

(13)

The results from estimating the difference using this correction are summarized in

Table 5. The gap between the price of apartments with district heating system and that with other heating methods.

Models	The Gap a	t-Values
Least Squares
Linear-Linear	KRW 100 million (USD 78 thousand)	6.63
Linear-Log	KRW 99 million (USD 77 thousand)	6.25
Log-Linear	KRW 121 million (USD 94 thousand)	8.62
Log-Log	KRW 110 million (USD 86 thousand)	7.97
Least Absolute Deviations
Linear-Linear	KRW 71 million (USD 55 thousand)	7.21
Linear-Log	KRW 67 million (USD 52 thousand)	6.65
Log-Linear	KRW 86 million (USD 67 thousand)	8.88
Log-Log	KRW 92 million (USD 72 thousand)	9.31
Box-Cox Transformation
Linear	KRW 80 million (USD 140 thousand)	12.44
Log	KRW 100 million (USD 78 thousand)	5.55

Note: ^a At that time of obtaining data, KRW 1284 was approximately equal to USD 1.

. Based on the LAD (log–log) model, the analysis revealed that in Uiwang City, apartments equipped with a DHS were approximately KRW 92 million (USD 72 thousand) more expensive than those with an LIHS. This price difference accounts for about 11.2% of the average apartment sales price in the area, which is approximately KRW 820 million (USD 638 thousand). These findings suggest that DHS has a substantial impact on housing price formation. As can be seen in Table 2, the average area of the apartments used in the HPM estimation was 108.3 m². In other words, apartments with an area of 108.3 m² can be considered representative of the sample. Therefore, if the difference in apartment prices is converted into area units, it is equivalent to KRW 849 thousand (USD 665) per m².

These results imply that a DHS should be actively considered in future large-scale public housing developments. The initial installation cost of a DHS is estimated to be approximately KRW 1.5 million, calculated by multiplying the unit cost (KRW 14,040 per m²) by the average supply area (108.3 m²). This unit cost is based on the internal regulations of the Korea District Heating Corporation, a state-owned enterprise that holds about 50% of the domestic DHS market and is officially declared to the government [42]. Since most other providers adhere to similar pricing standards, the estimate is considered representative. In comparison to this estimated initial cost, the observed price premium for an apartment equipped with a DHS (KRW 92 million) significantly exceeds the initial construction cost of a DHS.

This suggests that the adoption of a DHS can make a substantial contribution to maintaining asset value. While a DHS typically requires a separate infrastructure contribution at the time of construction, an IHS does not impose such explicit charges on residents. Nevertheless, an IHS still entails certain initial costs such as the installation of condensing boilers and the expansion of city gas pipelines for heating purposes, which are generally embedded in the apartment sales price and thus indirectly passed on to residents. In light of these considerations, the price premium observed for a DHS-equipped apartment may be interpreted as a net benefit to consumers when selling the apartment, exceeding the difference in upfront investment costs.

One of the key findings of this study is that the residential heating method, particularly the application of DHSs, positively influences apartment prices. This outcome can be interpreted as consumers valuing the benefits of DHSs [24]. Importantly, this study empirically confirms that DHS infrastructure not only serves as an energy supply facility but also contributes to the formation of real estate value. This suggests that the combination of high-quality residential environments provided by premium brands and the economic advantages of a DHS can further enhance consumer preferences. Therefore, incorporating DHS into the residential development process should not merely be viewed from the perspective of energy supply but as a strategic factor in enhancing market competitiveness and increasing consumer satisfaction.

The results of this study imply that the active adoption of DHS should be considered in future public housing projects in suburban areas of the SMA. Uiwang City, one of the major hubs in the metropolitan area, has already established DHS infrastructure. Applying DHS to new public housing developments in such areas is expected to contribute positively to housing price stabilization. Given that this study confirms higher price premiums of apartments with DHS compared to those with IHS, the adoption of DHS in public housing projects is likely to contribute to the long-term maintenance of housing value.

Last but not least, the authors summarize four scientific novelties of this study. First, the study is the first to analyze the impact of residential heating methods on housing prices in a small city. In particular, it used a city in South Korean, where new cities are emerging in the SMA, as a case study. Since DHS is a heating method mainly considered in cities, estimating the price premium of apartments with DHS compared to apartments with IHS in cities provides suggestions for which heating method to select in advance when developing small cities in the future. This is because housing prices are ultimately a result that reflects consumers’ preferences.

Second, the study successfully applied HPM, which is well-established in the literature, to delve into the impact of residential heating methods on housing prices. The estimated housing price equation was statistically significant, and the estimated coefficients for major variables had the expected signs a priori and secured statistical significance. In the conventional literature dealing with HPM, heating methods are not widely discussed. However, this study explicitly used heating method as a factor affecting the apartment price, and the result was successful. Heating method is an important determinant of apartment price in the specific city in South Korea.

Third, this study attempted to ensure the rigor of the results by applying various functional forms and estimation methods and then selecting the most appropriate one and applying the HPM. In fact, economic theory does not provide much information about the functional form of the HPM. Ultimately, researchers must decide on an appropriate functional form. In this regard, this study considered not only four typical functional forms but also a more flexible Box–Cox transformation model. In addition, the LAD estimation method, which is known to be more robust than the LS estimation method, was also attempted. Efforts to adopt an appropriate functional form and estimation method can enhance the usability of the results of this study.

Fourth, the study did not only present the results but also provided suggestions for heating methods preferred by consumers for future urban reconstruction and redevelopment. Although the apartment price equation was estimated for one city, this study revealed consumers’ preferences for heating methods as reflected in apartment prices. The debate surrounding DHSs versus IHSs in the choice of heating method is ongoing in South Korea. This study does not claim the superiority of DHSs over IHSs, nor does it primarily aim to demonstrate this. Nevertheless, if consumers’ preferences are taken into account, the implications for the choice of heating method in the reconstruction of old apartments or the construction of new apartments are clear.

The implications of this study are useful to individuals who can be largely divided into three groups. The first group, central and local governments, can utilize the results of the study when deciding on the heating method of apartment complexes in advance, reflecting consumer preferences before initiating full-scale new city development. If a DHS is adopted, permits should be issued to build related facilities such as combined heat and power plants. If IHSs are chosen, city gas operators for individual boilers should be selected. The second group, the new city developers, should establish a development plan that ensures the cost of building a DHS is borne by the initial apartment buyers at a level that does not exceed the price premium of apartments with a DHS. The third group is the general public who will live in the apartments. They can decide which neighborhood and heating method to choose by considering the price premium of apartments with a DHS, as observed in the study.

4. Concluding Remarks

This study empirically delved into the impact of residential heating method, particularly the application of DHSs, on apartment prices in Uiwang City, located near the SMA of South Korea. The main findings revealed that apartments with DHSs were priced, on average, approximately KRW 92 million (USD 72 thousand) higher than those with IHSs, and this difference was statistically meaningful. It seems that the features provided by DHSs contribute to a market price premium.

This study contributes both academically and in terms of policy by empirically validating that a number of factors, including the heating method, can significantly influence apartment prices, aligning with findings from some previous studies. From an academic perspective, the study has three implications. First, the study extended HPM by investigating the effect of heating methods on apartment prices. The implications of the study are particularly valuable because this issue has not been well addressed in the literature. In fact, although many applied HPM studies have been conducted, there are few that dealt with energy issues, such as heating methods. In this respect, the study contributes to the energy-related literature. Second, this study comprehensively employed various functional forms and two estimation methods when applying the HPM and then selected the most appropriate results to ensure rigor. Moreover, it provides a methodological guide for conducting follow-up studies. Third, in estimating the difference in housing prices, the estimated coefficients of the dummy variables of interest were not simply interpreted, but the formula proposed by Kenney [41] was utilized to reduce bias and to ensure the accuracy of the estimation of the difference.

In terms of policy, three useful implications were derived. First, after controlling for the influence of other factors, DHSs had a statistically significant effect on apartment prices in Uiwang City. The finding that DHSs is one of the factors determining apartment prices is an interesting aspect of the study. Since apartment prices ultimately reflect consumers’ preferences, it can be concluded that people prefer DHSs over other heating methods. Second, the price premium of apartments with a DHS was found to be 11.2% of the apartment price, which is by no means insignificant. When developing new cities in the future, the central government, local governments, and developers can use this finding as important information when asking consumers to shoulder the costs associated with DHS supply. Third, the findings of the study provide consumers who are planning to purchase apartments with background information that can serve as a reference when making their purchase decision. This is important because consumers want to know that heating methods can be added to the list of factors that generally cause differences in apartment prices, as well as the quantitative level of the difference.

Of course, this research has further room for improvement through follow-up studies. As further research can utilize broader regional scopes and longitudinal data, it is expected that a more precise and multidimensional understanding of the relationship between energy infrastructure and the housing market will be achieved. First, our analysis was restricted to a specific region, making it difficult to generalize the findings to a nationwide context. In addition, the framework adopted in this study could be applied to other countries where various heating methods are used. Comparative studies across countries could provide interesting and new implications.

Second, while this study examined the impact of residential heating methods, it did not clearly distinguish whether the observed price differences were driven by the residential heating method itself or by locational factors, such as proximity to DHS infrastructure. Future research should consider supply distance and infrastructure accessibility to more precisely isolate the pure effect of heating systems.

Third, the analysis was based on cross-sectional data at a single point in time and thus cannot capture price dynamics over time following the adoption of DHSs. Future studies should employ long-term panel data to track these changes. Fourth, while it focused on residential heating methods, this study did not control for other energy efficiency-related factors, such as the use of renewable energy, building insulation performance, or energy ratings. Future research should incorporate these factors to provide a more comprehensive analysis of the influence of energy infrastructure on housing prices.

Fourth, given that the housing market is heavily influenced by external factors, such as government real estate regulations, interest rate fluctuations, and broader economic conditions, future research should control for these macroeconomic variables to produce more refined analyses.