Behavioral Foundations of the Spatial Model: the Choice of the Distance Metric
4.01.2018 13:30
YER : AB5, 5202
Uğur Özdemir
The University of Edinburgh

Spatial models are ubiquitous within political science. Whenever we confront spatial models with data, we need to measure distances on the underlying space. The most commonly used measure is the Euclidean measure. It is not clear why we take this choice for granted though. Why should we expect the individual preferences to be in accord with some geometric notion of distance? In this paper, I employ a latent utility based spatial voting framework to look into this question. Using machine learning algorithms on individual level survey data for different country-elections, I estimate the best fitting metric and test whether the choice of Euclidean measure is justified. It turns out that it is hardly the case. The paper has implications for any empirical study on decision making which is based on a notion of distance.