Measurement error induced by locational uncertainty when estimating discrete choice models with a distance as a regressor

TL;DR

This paper investigates how locational uncertainty from geo-masking causes measurement errors in spatial microeconometric models, leading to biased estimates of distance effects in logistic regressions.

Contribution

It extends classical measurement error analysis to logistic models with spatial data, quantifying bias introduced by geo-masking in hospital choice studies.

Findings

Higher geo-masking distortion increases downward bias of distance coefficients.

Measurement error from geo-masking affects the accuracy of spatial microeconometric estimates.

The bias magnitude correlates with the level of locational distortion.

Abstract

Spatial microeconometric studies typically suffer from various forms of inaccuracies that are not present when dealing with the classical regional spatial econometrics models. Among those, missing data, locational errors, sampling without a formal sample design, measurement errors and misalignment are the typical sources of inaccuracy that can affects the results in a spatial microeconometric analysis. In this paper, we have examined the effects of measurement error introduced in a logistic model by random geo-masking, when distances are used as predictors. Extending the classical results on the measurement error in a linear regression model, our MC experiment on hospital choices showed that the higher the distortion produced by the geo-masking, the higher is the downward bias in absolute value towards zero of the coefficient associated to the distance in a regression model.