On Hedonic Price of Second-Hand Houses in Beijing Based on Multi-Scale Geographically Weighted Regression:Scale Law of Spatial Heterogeneity
Received date: 2019-05-15
Revised date: 2020-02-03
Online published: 2025-04-11
A large number of empirical researches show that there is obvious spatial heterogeneity in the housing price influence mechanism. Although classic geographically weighted regression (GWR) can solve part of the spatial heterogeneity problem which cannot be handled by traditional linear regression models, it ignores the scale problem of spatial heterogeneity of different influencing factors and causes large estimation errors. Multi-scale geographically weighted regression (MGWR) improves classical GWR by allowing the bandwidths of each variable to be different, thereby obtaining more credible estimation results and giving the scale of influence of different variables. Based on MGWR, this paper studies the price characteristics of second-hand residential transactions in Beijing from 2011 to 2017. The results show that: 1) Previous researches based on classic GWR may not be robust. MGWR can separate different influence scales of the independent variables, and MGWR result is more reliable. 2)Beijing second hand house prices are very sensitive to location factors, and there is a high degree of spatial heterogeneity. The scale of location impact is the smallest of all variables and is close to the street scale. The number of bedrooms and the distance to the nearest subway are variables on a global scale, and the influence on space is relatively stable. The distance to the bus station, the distance to the elementary school, the structure of the building and the condition of the decoration have no significant effect on house prices. Other significant variables have certain spatial heterogeneity, and their spatial scales from small to large are transaction time, area, building age, floor, and orientation. 3) Location, orientation, number of bedrooms, and transaction time all positively affect house prices, while area, age, floors and distance to the subway station negatively affect house prices. In all the influencing factors, location is the most important factor affecting house prices, followed by the direction of transaction time. Area, transaction time, direction, and distance to the nearest subway have key impact. Floor and the number of bedrooms have a smaller impact on house prices, while the area and building age have the weakest impact.
SHEN Tiyan , YU Hanchen , ZHOU Lin , GU Hengyu , HE Honghao . On Hedonic Price of Second-Hand Houses in Beijing Based on Multi-Scale Geographically Weighted Regression:Scale Law of Spatial Heterogeneity[J]. Economic geography, 2020 , 40(3) : 75 -83 . DOI: 10.15957/j.cnki.jjdl.2020.03.009
表1 变量描述Tab.1 Descriptions of major explanatory |
变量名称 | 英文简写 | 单位 | 变量描述 |
---|---|---|---|
常数项 | Intercept | 万元 | 模型的截距项,反映了区位的影响 |
成交月 | month | 月 | 成交时间距2011年1月的月份数 |
面积 | area | m2 | 住宅面积 |
卧室数量 | bed | 间 | 住宅卧室数量 |
楼龄 | age | 年 | 房屋建成年龄,即2011年与建成年差值 |
楼层 | floor | 分值 | 房屋所在楼层(0为低楼层,1为中楼层,2为高楼层) |
朝向 | direction | 虚拟变量 | 房屋是否朝向东、南(朝向包含东或者南为1,否则为0) |
装修状况 | decorate | 虚拟变量 | 是否精装修(精装修为1,否则为0) |
建筑结构 | stucture | 虚拟变量 | 房屋是否是板楼(板楼为1,否则为0) |
最近地铁站距离 | subway | km | 到最近地铁站的距离 |
最近公交站距离 | bus | km | 到最近公交站的距离 |
最近小学距离 | school | km | 到最近小学的距离 |
表3 经典地理加权回归与多尺度地理加权回归模型带宽Tab.3 Bandwidth of GWR and MGWR |
变量 | MGWR带宽 | 经典GWR |
---|---|---|
常数项 | 43 | 142 |
成交月 | 67 | 142 |
面积 | 116 | 142 |
卧室数量 | 2 945 | 142 |
楼龄 | 481 | 142 |
楼层 | 1 114 | 142 |
朝向 | 1 890 | 142 |
装修状况 | 43 | 142 |
建筑结构 | 69 | 142 |
最近地铁站距离 | 2 970 | 142 |
最近公交站距离 | 2 970 | 142 |
最近小学距离 | 306 | 142 |
表4 多尺度地理加权回归系数统计描述Tab.4 Statistical description of MGWR coefficient |
变量 | 均值 | 标准差 | 最小值 | 中位数 | 最大值 |
---|---|---|---|---|---|
Intercept | 4.442 | 1.406 | 0.662 | 4.442 | 9.155 |
area | -0.011 | 0.005 | -0.024 | -0.011 | 0.001 |
bed | 0.437 | 0.008 | 0.416 | 0.440 | 0.448 |
direction | 0.639 | 0.113 | 0.420 | 0.634 | 0.844 |
decorate | 1.301 | 1.257 | -2.443 | 1.128 | 6.118 |
month | 0.050 | 0.024 | -0.038 | 0.048 | 0.140 |
subway | -0.138 | 0.002 | -0.143 | -0.138 | -0.134 |
floor | -0.278 | 0.123 | -0.472 | -0.309 | 0.072 |
age | -0.024 | 0.009 | -0.048 | -0.025 | -0.005 |
stucture | 0.042 | 0.366 | -0.749 | -0.015 | 1.752 |
bus | 0.056 | 0.003 | 0.050 | 0.055 | 0.067 |
school | 57.154 | 22.831 | -9.748 | 64.674 | 87.260 |
[1] |
|
[2] |
|
[3] |
|
[4] |
|
[5] |
|
[6] |
|
[7] |
|
[8] |
|
[9] |
|
[10] |
|
[11] |
|
[12] |
|
[13] |
|
[14] |
|
[15] |
|
[16] |
|
[17] |
|
[18] |
董冠鹏, 张文忠, 武文杰, 等. 北京城市住宅土地市场空间异质性模拟与预测[J]. 地理学报, 2011, 66(6):750-760.
|
[19] |
冯长春, 李维瑄, 赵蕃蕃. 轨道交通对其沿线商品住宅价格的影响分析:以北京地铁5号线为例[J]. 地理学报, 2011, 66(8):1055-1 062.
|
[20] |
宋伟轩, 毛宁, 陈培阳, 等. 基于住宅价格视角的居住分异耦合机制与时空特征研究:以南京为例[J]. 地理学报, 2017, 72(4):589-602.
|
[21] |
古恒宇, 沈体雁, 周麟, 等. 基于 GWR 和 sDNA 模型的广州市路网形态对住宅价格影响的时空分析[J]. 经济地理, 2018, 38(3):82-91.
|
[22] |
谷一桢, 郑思齐. 轨道交通对住宅价格和土地开发强度的影响:以北京市13号线为例[J]. 地理学报, 2010, 65(2):213-223.
|
[23] |
王洋, 方创琳, 盛长元. 扬州市住宅价格的空间分异与模式演变[J]. 地理学报, 2013, 68(8):1082-1 096.
|
[24] |
|
[25] |
|
[26] |
|
[27] |
古恒宇, 孟鑫, 沈体雁, 等. 中国城市流动人口居留意愿影响因素的空间分异特征[J]. 地理学报, 2020, 75(2):240-254.
|
[28] |
|
[29] |
|
[30] |
|
[31] |
|
[32] |
|
[33] |
|
/
〈 |
|
〉 |