Estimating bid-auction models of residential location using census data with imputed household income

Abstract

Residential location is a key component of the land-use system and modelling residential location is essential to understand the relationship between land-use and transport. The increasing availability of censuses such as the German Zensus 2011 has enabled residential location to be modeled with a large number of observations, presenting both opportunities and challenges. Censuses are statistically highly representative; however, they often lack variables such as income or mobility-related attributes as in the case of Zensus 2011. This is particularly problematic if missing variables define the choice options in a location model. For estimating bid-auction location models for household income groups and comparing them to a similar model for household size, we impute household income in census data by ordered regression. We find that the income model performs sufficiently well as it reveals reasonable aspects of the location patterns that the size model does not cover. Imputing choice variables should thus be considered in the estimation of residential location models.