1# RECORD 1956ccee53765fb417ccdc3f95db189441def9312# TITLE: Analyzing the Impact of Accessibility on Residential Property Price: A Case study of Hanoi's Central Districts3# YEAR: 2025.0 VENUE: Tạp chí Khoa học liên ngành và Nghệ thuật DOI: 10.64711/dq21w9974# AUTHORS: Le Minh Son; L. T. K. Ly; Trang Thu Doan5# PAGES: 27 CANDIDATE TABLE PAGES (1-based): [8]6# ABSTRACT: This research focuses on real estate values in central Hanoi, with the objective of analyzing factors influencing real estate prices, with particular emphasis on locational characteristics, for the case of Hanoi’s central districts. A geographically weighted regression (GW)R model is employed to address the limitations of traditional Hedonic Pricing Models (HPM), based on data from 1,545 properties across 12 central districts of Hanoi during the period from November 2024 to March 2025. The study evaluates the impact of 7 accessibility variables: distance to hospitals, parks, schools, universities, shopping centers, administrative offices at various levels, and the city center. Results show that the relationship between distance to public amenities and property prices varies significantly according to geographic location - a characteristic not reflected in traditional HPM models. Notably, many variables that lack statistical significance in the HPM model show differentiated impacts at specific locations in the GWR model, with higher coefficients of determination, demonstrating better explanatory power for real estate price variations. The research has demonstrated spatial heterogeneity in the relationships between valuation variables, indicating that location characteristics and spatial variation have not been adequately considered in current valuation methods in Vietnam. These findings not only address the limitations of HPM models but also provide important reference perspectives for policymakers regarding the significance of spatial factors.789===== PAGE 1 (score 8, numeric-paren 0, table-words 0, var-words 8) =====1011Tạp chí Khoa học liên ngành và Nghệ thuật12Tập 1, Số 2 (2025)138114NGHIÊN CỨU – TRAO ĐỔI151 School of Interdisciplinary Sciences and Arts, Vietnam National Univeristy, Hanoi16* Email: sonminhle@vnu.edu.vn17ABSTRACT18This research focuses on real estate values in central Hanoi, aiming to analyze factors 19influencing real estate prices, with particular emphasis on locational characteristics, for the 20case of Hanoi’s central districts. A geographically weighted regression (GWR) model is 21employed to address the limitations of traditional Hedonic Pricing Models (HPM), based 22on data from 1,545 properties across 12 central districts of Hanoi during the period from 23November 2024 to March 2025. The study evaluates the impact of 7 accessibility variables: 24distance to hospitals, parks, schools, universities, shopping centers, administrative offices 25at various level, and the city center. Results show that the relationship between distance to 26public amenities and property prices varies significantly across geographic locations – a 27characteristic not reflected in traditional HPM models. Notably, many variables that lack 28statistical significance in the HPM model show differentiated impacts at specific locations in 29the GWR model, with higher coefficients of determination, demonstrating better explanatory 30power for real estate price variations. The research has demonstrated spatial heterogeneity 31in the relationships between valuation variables, indicating that location characteristics and 32spatial variation have not been adequately considered in current valuation methods in Vietnam. 33These findings not only address the limitations of HPM models but also provide important 34reference perspectives for policymakers regarding the significance of spatial factors.35Keywords: Geographically weighted regression; GWR; hedonic; HPM; real estate; 36valuation; Hanoi.37ANALYZING THE IMPACT OF ACCESSIBILITY ON RESIDENTIAL 38PROPERTY PRICE: A CASE STUDY OF HANOI’S CENTRAL DISTRICTS39Le Minh Son 1*, Le Thi Hoa Ly 1, Doan Thu Trang 140Lê Minh Sơn 1*, Lê Thị Hoa Ly 1, Đoàn Thu Trang 141Received: 15 May 202542Revised: 28 August 2025; Accepted: 10 September 202543Ngày nhận bài: 15 tháng 5 năm 202544Ngày nhận bài sửa: 28 tháng 8 năm 2025; Ngày duyệt đăng: 10 tháng 9 năm 202545PHÂN TÍCH TÁC ĐỘNG CỦA KHẢ NĂNG TIẾP CẬN 46ĐẾN GIÁ BẤT ĐỘNG SẢN NHÀ Ở: NGHIÊN CỨU TRƯỜNG HỢP TẠI 47CÁC QUẬN TRUNG TÂM HÀ NỘI48TÓM TẮT49Nghiên cứu này tập trung vào giá trị bất động sản tại khu vực trung tâm Hà Nội, nhằm 50phân tích các yếu tố ảnh hưởng đến giá bất động sản, đặc biệt chú trọng đến đặc điểm vị trí 51đối với trường hợp các quận trung tâm của Hà Nội. Mô hình hồi quy có trọng số theo địa lý 52(GWR) được sử dụng để khắc phục những hạn chế của Mô hình Định giá Hedonic (HPM) 53truyền thống, dựa trên dữ liệu từ 1.545 bất động sản tại 12 quận trung tâm của Hà Nội trong 54giai đoạn từ tháng 11 năm 2024 đến tháng 3 năm 2025. Nghiên cứu đánh giá tác động của 7 55biến số về khả năng tiếp cận: khoảng cách đến bệnh viện, công viên, trường học, trường đại 56học, trung tâm mua sắm, cơ quan hành chính các cấp và trung tâm thành phố. Kết quả cho 57thấy mối quan hệ giữa khoảng cách đến các tiện ích công cộng và giá bất động sản thay đổi 585960===== PAGE 8 (score 10, numeric-paren 6, table-words 1, var-words 2) =====6162VNU Journal of Interdisciplinary Sciences and Arts 63Vol.1, No.2 (2025)648865STUDIES – DISCUSSION66In which, is the dependent variable – 67usually the monetary value of the property 68 are dependent variables and 69 are coefficients to be estimated 70in the model.71Model 3.1 can be rewritten in the more 72general matrix form as followed:73(3.2)74In which, is the matrix of size nx1 75containing the values of observed dependent 76variables. X is the matric of size n x (k+1) 77containing observed independent variables. 78 is the matrix of size (k+1) x 1 containing 79variable coefficients and is matrix of size 80n x 1 with residual terms. In mathematical 81form, these matrices are written as:82The estimation process of the impact of 83each independent variable in X on Y, i.e. 84estimation of values, is shown as:85(3.3)86And since these matrices are not of the 87same size, we multiply both sides with 88transposed matrix XT to achieve:89(3.4)90 in formular 3.4 is the vector of 91coefficients 92 in the model. In 93the conventional HPM approach, the value 94of these coefficients is global in the whole 95study area. 96The GWR extends this model by 97incorporating spatial factors into the model, 98to take into account local factors within the 99study area. First, for each observation point, 100the GWR model assigns spatial coordinates 101to the model:102(3.5)103Similarly, we can rewrite the general 104form of GWR model as:105(3.6)106where geographical coordinates 107 are 108the assigned to each observation point . This 109allows the relationship between the dependent 110variable and the independent variables to vary 111from location to location. In matrix form, the 112GWR model is shown as:113(3.7)114Where is a matrix of coefficients that 115vary with geographic location, and ⊗ is 116a logical matrix multiplication (between 117two matrices of the same size): each value 118of 119 is multiplied by the corresponding 120value of X. Matrix 121 now take a new 122updated form:123 124125126===== PAGE 9 (score 18, numeric-paren 2, table-words 0, var-words 16) =====127128Tạp chí Khoa học liên ngành và Nghệ thuật129Tập 1, Số 2 (2025)13089131NGHIÊN CỨU – TRAO ĐỔI1326 For calculation of weightings applied, we used the default settings in ArcGIS to select bandwidth and applied the Akaike Information 133Criterion (AIC) for optimal calculation based on works by Fotheringham et al. (1998) and Bellefon & Floch (2018) which argued for 134AIC as a reliable and most efficient settings. 1357 “SD” is short for “shortest distance”, with HOS, PRK, SCH, UNI, SHP, CENL, CEN correspondingly stand for Hospital, Park, School, 136University, shopping malls (shopping centers), Center (local) – the position of district level people’s committee buildings, and Center 137(Hanoi) – the position of Hanoi people’s committee building. 138The values in each row of 139 is now 140estimated as:141(3.8)142Where the matrix 143 is a diagonal matrix 144of size n x n, where 145 is the weight assigned 146to the observation data point at location , 147based on coordinates 148 This weight6149 reflects the influence of observation point 150 on point . The weight of a point close 151to will be higher than a point far from . 152The weight of a location decreases with 153increasing distance from .154The 155fundamental 156difference 157between 158the HPM and the GWR model is shown in 159equations 3.4 and 3.8. In the conventional 160HPM, 161the 162estimated 163coefficients 164are 165spatially 166constant 167while in the GWR model, the estimated 168coefficients can vary spatially – expressed 169numerically as taking values within a range 170 171This spatial variation is adjusted through the 172weight matrix 173. The main results of the 174GWR analysis are usually presented in the form 175of multiple maps. Each independent variable 176requires at least one map to represent the 177variation of the estimated coefficients in space.1783.3. Model specification179From the conventional HPM approach 180(Model 3.1) we first define the standard 181HPM with the following form:182(3.9) 183Where PRICE is the advertised selling 184price of residential property (measured 185in million VND/m²), SDHOS, SDPRK, 186SDSCH, SDUNI, SDSHP are independent 187variables respectively represent the shortest 188distance as-the-crow-flies from real estate 189 to the nearest urban services which are 190hospitals, parks and green spaces, schools, 191shopping centers, universities. SDCENL, 192SDCEN 193are 194independent 195variable 196representing shortest distance of real estate 197i to the nearest local people’s committee 198location and to the Hanoi People’s 199committee location7. is the error term and 200 are coefficients to 201be estimated in the model.202Access to these urban services – or 203amenities – (Figure 2), we argue, are basic 204and essential matter-of-fact need that 205everyday life requires to happen: medical 206facilities in case of sickness; education for 207themselves and family members; green 208spaces and shopping as utility enhancers; 209engagement with public institution for 210essential daily services and information. 211The selection of these variables align with 212other empirical works which include several 213variables to measure accessibility based on 214distance to different amenities such as by 215Wu et al. (2019) with variables representing 216distance to central business district, district 217center, high-speed rail, metro station, 218density, park, hospital, school. Wang et al. 219(2020) distinguished between distance to 220shopping mall, food market, supermarket 221