Identification of Influencing Factors of Housing Price Divergence in Different Cities
Since 2013,the growth rate of housing prices in China's Tier 1,Tier 2,and Tier 3 cities has exhibited a clear trend of differentiation,and"city-specific policies"have gradually become the mainstay of housing price regulation.To implement"city-specific policies",it is necessary to accurately identify the primary factors that affect the differentiation of housing prices in different cities.This paper comprehensively employs machine learning methods such as XGBoost and SHAP value interpretability methods,based on panel data of 70 large and medium-sized cities nationwide from 2009 to 2019,to calculate and analyze the primary driving factors behind the multi-round differentiation of housing prices in Tier 1,Tier 2,and Tier 3 cities.The research results indicate that:First,expected factors played a crucial driving role during the multi-round differentiation of housing prices;Second,monetary policy itself is not the primary driving factor of housing price differentiation,but it can cause housing price differentiation by affecting expectations;Third,demand factors and supply factors themselves are not the primary driving factors of housing price differentiation,but they can also cause housing price differentiation by affecting expectations.In view of this,the key to"city-specific policies"in different cities is to stabilize the public's expectations of housing price trends based on their own situation,starting from the demand side or the supply side.For Tier 1 cities,due to the continuous influx of population outside the city and relatively scarce land supply,they are in a state of relatively tight housing supply,which also makes residents'expected housing prices show an upward trend.To stabilize housing price expectations,it is necessary to consider appropriately increasing housing supply under the premise of controlling land costs,so as to weaken the expected housing price increase caused by the imbalance between housing supply and demand.In addition,government should continue to improve the housing rental market to alleviate the pressure on the housing buying and selling market,and further weaken the expected housing price increase in Tier 1 cities.For Tier 2 and Tier 3 cities,since these cities often face the situation of population outflow,residents'expected housing prices will decline.In order to stabilize housing price expectations,it is necessary to further upgrade urban infrastructure,improve education,medical and environmental quality,and improve residents'housing purchase demand.At the same time,Tier 2 and Tier 3 cities with relatively abundant housing supply should also formulate more scientific and reasonable talent introduction plans to attract population inflows,so as to further stabilize housing price expectations.The marginal contribution of this paper is reflected in two aspects.First,although some literature attempts to explore the primary reasons for the differentiation of housing prices in different cities,traditional methods such as panel regression models are mainly used.Also,the number of factors that can be examined and the identified nonlinear relationships are relatively limited.Machine learning methods can better overcome the shortcomings of traditional methods and more accurately identify the primary factors affecting housing price differentiation.Second,although the central government has been calling for"city-specific policies"in recent years,the academic community's research on the reasons for the differentiation of housing prices in different cities is not yet clear.Existing literature mainly studies whether a certain factor has a significant impact on housing price differentiation,and cannot clearly compare the relative importance of different factors.This paper uses SHAP value interpretability method to directly rank the relative importance of various influencing factors,to identify the primary factors affecting housing price differentiation and provide more clear decision-making references for real estate regulation and control.
万方数据
维普
