Ghost cities of China: Identifying urban vacancy through social media data

Abstract

“Ghost Cities” have become a phenomenon of global interest since 2009 when popular media highlighted the existence of Ordos, a large Chinese city that was almost entirely vacant. The term is used to describe housing vacancy associated with overdevelopment and can refer to small communities, neighborhoods, or even whole cities that lie vacant. Ghost Cities are particularly prevalent in China where housing vacancy has become a serious concern for many second and third-tier cities. Measuring the extent of these vacant areas has been challenging due to Chinese data restrictions. This research tests whether it is possible to collect data, scraped from Chinese social media open access API's including Dianping (Chinese Yelp), Amap (Chinese MapQuest), Fang (Chinese Zillow), and Baidu (Chinese Google Maps) to develop a computational model to identify areas considered to be Ghost Cities. The model created for this study is based on the idea that thriving communities need access to basic amenities. Therefore Hansen's gravitational model was applied to give an “amenities score” for residential locations based on their accessibility to restaurants, banks, grocery stores, beauty salons, KTV, medical facilities, schools, and malls Moran's I spatial autocorrelation was applied to the amenity scores below the mean to determine spatially clustered residential locations with low scores. The results were considered potential Ghost Cities and were visited in Chengdu and Shenyang to confirm the accuracy of the model. These site visits showed that the model identified transitional, underperforming, or vacant housing in these cities, illustrating that it is possible to use data scraped from social media to identify underused residential developments.