Methods to examine omitted variable bias in hedonic price studies

Abstract

Many studies use the hedonic pricing method to uncover consumers’ willingness to pay for accommodation characteristics in the hospitality industry. In most empirical applications using cross-sectional data, implicit prices might be biased if omitted time-invariant variables correlate with observed attributes. This paper proposes a set of diagnostic checks to inspect the potential bias in the estimates through exploiting repeated information for the same accommodations. Due to the limitations imposed by standard fixed and random effects panel regressions, we advocate for the use of Mundlak and Hausman–Taylor estimators. The proposed methods are applied to a study of Airbnb hedonic prices in Ibiza using a 14-month time window. In doing so, we document a novel finding: Superhosts set lower prices conditional on time-invariant quality.