Using Hedonic Regression Coefficient Estimates to Predict House Prices in India

Real estate markets in emerging economies are beset with price opacity, low transaction volumes and perverse incentives on tax evasion, leading to acute shortage of reliable price information. In this paper, we overcome the paucity of reliable price information in emerging markets through a novel approach. The aim of this paper is to use the limited available housing data from one city, to predict house prices in another city. We achieve this, by using a methodology to predict residential real estate prices from socio-economic, climatic and demographic features of a city. We use more than a million residential housing units from listed properties, in 6 metropolitan cities in India, over a period of 6 years, to construct 36 hedonic price models, and obtain a panel of coefficient estimates. Then, we regress these hedonic coefficient estimates of housing attributes, on a collection of socioeconomic, climatic and demographic features of a city. The resulting two-stage model is able to predict house prices, both within and outside our sample of analysis. For out of sample cities, the explained variation in house prices is 84%