Data-driven approach for measuring neighbourhood value: A study of neighbourhood value polarization in 20 US cities using housing transactions data
Abstract
This thesis explored changing patterns in neighbourhood value for 20 US cities from 2004 to 2015. Neighbourhood Value was defined and measured as a combination of distance from key locations within the city and local conditions within the neighbourhood, as reflected in housing prices. Hedonic modeling was used to determine neighbourhood value from a database of four million housing transactions, which included for each house the sale date, purchase price, geolocation, and structural characteristics. Spatial interpolation was used to visualize neighbourhood value in each city over 12 years. Moran’s I was used to analyze polarization in neighbourhood value. The findings suggest that for some cities, distance factors drove neighbourhood value while in other cities, local attributes drove neighbourhood value. Rather than applying generalized rules, each city must be considered individually to understand how neighbourhood value works within it. Which factors drive neighbourhood value likely impact the effectiveness of municipal policy.