Quality information gaps in housing listings: Do words mean the same as pictures?

Abstract

The incomplete or defective information on the listed housing quality may impair the market clearing processes, may lead to the adverse selection and would have its reflection in the data needed to perform market analyses. To pinpoint the existing information gaps, this paper inspects the quality-related information disclosure strategies of agents who list apartments for sale or rent in online listing platforms. The first part of the empirical study compares the direct declarations of the listed housing quality with the quality indirectly signaled via photos attached to listings and shows that there extists a discrepancy between them. Combined with the high share of observations, for which the full information on quality is not being disclosed it contributes to the existence of informations gaps. The second part of the study answers, whether indirect textual quality signals (descriptions of apartments), processed with the use of Wordscores—a supervised machine learning algorithm are consistent with visual quality signals. It has been documented, that the sales listings’ quality signals have agreed in 63–90% of cases, depending on the model’s variant. For the rental market, the descriptive quality signals have matched the visual ones in 71–83% of cases. Moreover, it has been shown that among rental listings the consistency has been higher for listings posted by landlords, while for sales listings—by brokers. Finally, it may be argued that to fill the existing information gaps one may utilize the often-unused information conveyed via textual quality-signalling channel.