城市结构对轨道交通系统与房地产价格关系的影响——以北京、杭州、南京、成都4个城市为例

Impact of Urban Structure on Subway System and Real Estate Price:Evidence from Beijing,Hangzhou,Nanjing and Chengdu

文章构建衡量地铁通勤时间节省度的指标--地铁指数。基于该指标的回归分析显示,多数城市房价对交通便捷度的溢价为负。这一反直觉结果反映了城市多中心结构对地铁房价关系的影响。为识别不同的中心区块,提出一种带约束的聚类算法。该算法的优势在于其分区结果完全是数据导向的,相对于传统的行政分区,更能反映各区块的内在联结程度。同时,相较于传统的聚类算法,该算法在最小化组内距离的同时,加入一个在地铁指数的基础上构建的约束条件,该约束条件使得聚类分区结果能够同时反映城市经济学中关于交通便捷度溢价的基本假说和房地产市场的空间连续性特征。在分区的基础上重估模型,研究发现,在多中心城市的不同区块,地铁所带来的交通便捷度增益与其带来的居住环境恶化两种力量的对抗导致十分复杂的溢价效果。同时,诸如城市是否存在多个中心、城市核心区的大小、多中心城市的副中心与核心区距离的远近、不同中心区块经济发展程度的差异等城市结构因素也都会对地铁溢价的方向与大小产生影响。而这些复杂关系是在使用常系数特征价格模型分析单一城市的地铁房价关系时无法得到体现的。最后,基于约束聚类算法的分区结果还反映城市内不同区域间的经济地理一体化趋势与隔离趋势,因而可以为城市规划领域的有关问题提供有价值的建议。

This paper constructs an index system of time saving level of metro commuting to measure the improvement of commuting efficiency induced by living close to a metro station. Based on the regression analysis, the results find that except for Beijing, subway systems in Hangzhou, Chengdu and Nanjing all have negative impact on real estate price,which reflects the impact of the poly-centric structure of the three cities on their housing markets. To identify the correct housing submarket associated with every subcenter, this paper develops a constrained clustering algorithm which can partition the urban area into a set of subregions, this algorithm is fully data-oriented and can reflect deep-level features of urban structure. Compared with the standard K-means clustering, this clustering algorithm adds a constraint based on metro index. It is this constraint that can simultaneously reflect the space continuity of real estate market and the basic hypothesis which is the premium of transportation convenience. From the estimation, we find a complicated relationship between metro stations and real estate prices, the complexity is arising from the interaction of positive effect of subway system in terms of its commuting efficiency and its negative impact on living environment nearby. In addition, many structural features of cities can also significantly affect both the direction and scale of premiums of metro stations, such as the size of urban core, the distance between subcenters and urban core and so on. The complexity found in this paper is not detectable when it is used to analyze the real estate prize of single city based on the hedonic price model with constant coefficient. Finally, apart from pinpointing housing submarkets, constrained clustering can also shed light on the economic integration and segregation trend among different subregions of a city, which are useful information for urban planners.