We investigate the effect of ethnic bias on procurement auctions based on a universe of procurements in Russia. We develop a simple machine learning algorithm using Memorial Datasets on 1930s repressions and Forebears.io genealogy websites as training sets. The algorithm yields an accurate proxy for the ethnicities of public procurement officers and firm managers. We document that procurement officers with Slavic—“majority”— last names tend to match with “Slavic” firm managers, while “minority” procurement officers —with “minority” firms. A cross-sectional relationship, even conditional on granular goods, time, and region-fixed effects, could be non-causal, so we make a further step to achieve identification. Our data’s rich panel nature and quasi-exogenous changes in procurement committees’ composition give us such an opportunity. In this event-study design, we document that switching a bureaucrat from minority to Slavic increases the probability of ‘Slavic firm’ winning by 1.5%. The effect is symmetric for switches from Slavic to a minority. Note that the contract award’s baseline probabilities are 89.5% and 10.5%, so our effect is statistically significant and economically sizeable. Next, we find, somewhat counterintuitively, that this bias leads to lower procurement prices. One explanation is that bias is not taste-based, but rather co-ethnic bureaucrats have more information on co-ethnic firm managers and can screen them better. Such screening could indeed lead to lower contract prices. Naturally, if true, such a situation is not the first best, but potentially a second-best.
Рабочий язык семинара: английский.
Это заседание пройдет в режиме онлайн.
Для получения ссылки на Zoom-конференцию надо написать запрос на почту email@example.com с указанием имени и фамилии с рабочего адреса электронной почты.