A Comparative Prediction Study of Housing Price Index based on Web Search Data - Evidence from Beijing and Lanzhou in China

Abstract

Recently, using web search data to predict public health trends and economical indicators, like consumption, unemployment, tourism is arising more and more researchers interest. Based on the thought that web search data could reflect searchers' attention and concern, we study the correlation between housing price index and web search data, then make a comparative analysis of two regions with different economic levels. One the one hand, a theoretical framework illustrating the relationship between web search data and housing price index has been established. One the other hand, empirical studies of Beijing and Lanzhou has been conducted to verify the predict ability of web search data. Comparing with models not including web search data, the mean absolute percentage error decreased and the goodness-of-fit improved in models with web search data. In addition, based on the comparison of Beijing and Lanzhou, we find the predict ability of web search data has a certain relationship with economic development level of the region, and web search data has a better explanation ability for the fluctuation of housing price in developed region, like Beijing. Importantly, using web data to predict is able to achieve true real-time monitoring, and provide as references for government sector to make macro-control policy.