Controlling Unobserved Heterogeneity in Repeat Sales Models: Application to Anchoring to Purchase Price

Abstract

Stocks, art, housing and other assets are often studied in a two-stage repeat sales framework to control unobserved heterogeneity. But unobserved variables that influence both first and second sales remain a troubling source of omitted heterogeneity. We develop an iterative simulation algorithm to mitigate omitted variable bias which can be estimated with standard hedonic data in many repeat sales studies. We illustrate our model with an application to seller anchoring to the price paid for a house. We find that premiums on expected losses are reduced by about 50% compared to conventional models, as expected if our algorithm reduces unobserved heterogeneity. We also show that the probability of sale is strongly negatively associated with quality-adjusted expected losses, consistent with theory and previous empirical evidence on anchoring and sales volume. In contrast, losses measured conventionally produce the opposite results. Our study has general implications for two-stage estimation in the finance literature.

Keywords: Repeat Sales, Unobserved Heterogeneity, Simulation Algorithm, Anchoring, Loss Aversion

JEL Classification: C1, E00, R3