Housing Price Growth Detection and Classification Using Dynamic Time Warping

Authors: Ziye Zhang*, , Yuanshuo Xu, Cornell University
Topics: Economic Geography, Quantitative Methods, Spatial Analysis & Modeling
Keywords: housing price growth, dynamic time warping, housing policy, Beijing
Session Type: Paper
Day: 4/4/2019
Start / End Time: 9:55 AM / 11:35 AM
Room: Roosevelt 0, Marriott, Exhibition Level
Presentation File: No File Uploaded


The past decade witnessed a housing market boom in Beijing, China. Since 2011, Beijing has issued a series of policies to regulate the market, such as housing purchase restrictions. The market price responded to these regulations drastically, but continued its growth until recent. One interesting observation is that housing prices exhibit different growth rates and response magnitudes across the city. In another word, housing units with different attributes have different patterns in housing price appreciation. For example, the increase rate of housing price could be much higher for housing units nearer to the city center. The adjustment and response of market to policies had something to do with investment and speculative incentives. Such cross-sectional heterogeneity in the time series of housing prices could provide valuable information to understand people's expectation and preference and the impact mechanism of housing policies. The foundation of such analysis is the detection and classification of housing price growth patterns. This paper collects a large housing dataset from 2012 to 2016 in Beijing, involving over 4,000 locations, based on which housing price time series are generated. To detect the similarities between them, we apply Dynamic Time Warping (DTW) technique, which is commonly applied for time-dependent series optimum alignment between two time series. The results show that the growth patterns could be classified into several distinct types, which are different in volatility and in timing to respond to policies. Housing market regulations disseminated impacts across different housing attribute spaces.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login