Disclosure policies in research contests with stochastic entry

This paper models research contests as an all-pay auction with stochastic entry and analyzes the effect of disclosing the actual number of participants on the expected winning bid. I find that the classic disclosure irrelevance principle on expected revenue does not extend to the ranking on expected winning bid. The ranking on expected winning bid crucially depends on the number of potential bidders, participation probabilities, and the distribution of bidders' valuations. When there are more than two potential bidders, fully concealing the actual number of participants elicits a higher expected winning bid if potential bidders have low participation probabilities, whereas the rank reverses if bidders have high participation probabilities and high chances of being high types. (C) 2020 Elsevier B.V. All rights reserved.