R&D, political campaigns, science competitions, job promotions and lobbying are often viewed as contests – and a key question for organisers of such contests is how much information to reveal when players do not know other contestants'' ability, valuation and/or competence? The answer is fully to reveal information about contestants if they are good enough but otherwise to keep some uncertainties.
That is the main finding of research by Jun Zhang and Junjie Zhou, published in the Economic Journal. Their study has important applications in situations where contest organisers can control the amount of information released to contestants. For example:
- In the United States, the government can decide the level of transparency by requiring lobbyists to provide financial information. Such transparency requirements could potentially leak information about lobbyists'' private interests.
- In job promotions, companies can decide whether to announce the list of candidates and also whether to reveal candidates'' past experience. Such information conveys signals correlated with workers'' private competence, and, once disclosed, could lead to updates in beliefs.
- US political candidates are required by the Federal Election Campaign Act to reveal the sources of campaign contributions and report campaign expenditures, which convey information about the extent of a candidate''s financial support.
In earlier studies of these issues, organisers are assumed to make a zero-or-one choice by comparing no disclosure and full disclosure. This raises the question of whether restricting to the zero-or-one choice causes loss of generality.
Zhang and Zhou show that when a contestant''s valuation follows a binary distribution, it is without loss of generality to focus on no disclosure or full disclosure, and one of them will be optimal among all disclosure policies. But when a contestant''s valuation takes more than three different values, the researchers show that the simple zero-or-one choice generally fails to maximise the contest organiser''s objective.
Zhang and Zhou''s study focuses on a model with two contestants: contestant A, with commonly known valuation, and contestant B, with privately known valuation. The novelty of their research is to show that it is without loss of generality to focus on posteriors that have no more than two positive probabilities over contestant B''s possible valuations.
Specifically, Zhang and Zhou show that when contestant A is strong enough, full disclosure is optimal; when contestant A is a bit weaker, pooling the highest and the second-highest valuations and fully separating the others is optimal.
''Information Disclosure in Contests: A Bayesian Approach'' by Jun Zhang and Junjie Zhou is published in the November 2016 issue of the Economic Journal. Jun Zhang is at the University of Technology Sydney. Junjie Zhou is at the Shanghai University of Finance and Economics.
