Abstract
Low participation in public procurement auctions may reflect either weak procurement practices or a genuinely limited pool of potential suppliers. Distinguishing between these explanations requires credible estimates of potential competition. This paper develops and compares three approaches to identifying potential bidders: firms that register in tenders, machine-learning predictions of bidding probabilities, and a novel algorithm based on similarities in previous procurement activity of firms. Using comprehensive data from Finnish procurement auctions, we quantify the level of potential competition for each procurement and develop a framework to examine how well the market potential is attained. In general, more than 40% of auctions attract too few bidders relative to their market potential, indicating that a large share of potential competition remains unrealized. Moreover, there is considerable heterogeneity in performance across procurers. We show that longer bid preparation periods and the use of lot division are associated with significantly higher realized competition, consistent with procurement design playing an important role in mobilizing potential bidders.