Differential Measurement Error in House Price Indices

Abstract

This article investigates measurement errors when using indices to model house prices over time. Our analysis, comparing index prices to actual transaction values, reveals that in many cases, widely-used indices display measurement errors correlated with the index values. Measurement error correlated with predictors constitutes “differential measurement error” at the level of the data generating process (DGP). We further explore the presence of differential measurement error within the context of mortgage lending. Our findings uncover substantial measurement errors in mortgage data, which not only diminish the predictive accuracy of models but also introduce notable biases in the coefficient estimates of variables.