Frédéric Koessler*, Vasiliki Skreta
This article was originally published in the March 2021 edition of 5 papers… in 5 minutes!
Any economic decision is determined by the available information, and most times this information is the deliberate outcome of some entity attempting to influence behavior. Politicians control information affecting voters’ sentiments, entrepreneurs filter information revealed to potential investors, and pharmaceuticals run public experiments to prove the efficacy of their treatments.
The vast literature on information design studies how to design experiments in order to achieve a specific goal, when one is bound to reveal any outcome truthfully and cannot alter the design once any counterparty has observed it. For example, a pharmaceutical must register a drug trial in advance, and is committed to accurately publish the results. A range of questions addressed in this literature includes the informativeness of optimal experiments, as well as the limits of what outcomes the designer can achieve. Importantly, a party in a position to design informative experiments producing new data, is often privileged with superior information: prior to registering a specific trial, a pharmaceutical might have privately collected data on the efficacy of the drug in question.
It is these situations where the designer is privately informed that are addressed by Frederic Koessler and Vasiliki Skreta in this article. The authors offer a broad analysis of the problem and various insights on what outcomes might be obtained in these scenarios. When the designer is privately informed, the nature of the problem changes in two fundamental ways. Firstly, the incentives of a designer prior to obtaining any private information (ex-ante) and its incentives once some data has been collected (interim) could differ substantially. Secondly, rational actors understanding the potential presence of private information, will attempt to infer something from the designer’s choice of experiment. For example, if a pharmaceutical knowing of some side-effects chooses a design which focuses away from them, then consumers could infer their presence. The authors provide sufficient conditions for when an ex-ante optimal design arises as an outcome when the designer is informed. These include situations when an uninformed designer would find a fully revealing experiment optimal, as well as situations where decision makers take a simple ‘approval’ action, and the designer has ‘transparent motives,’ in that he is disinterested in the underlying state, caring only about inducing approval. In these situations, the ability of the designer to choose an experiment before becoming informed is irrelevant. Finally, on a technical level the authors compare and contrast various solution concepts for the problem at hand, importing and adapting ideas from the literature on mechanism design by an informed designer.
Original title of the article: Information Design by an Informed Designer
Published in: PSE Working Paper n°2021-03
Available at: https://halshs.archives-ouvertes.fr/halshs-03107866/
Credits (picture) : Shutterstock SFIO CRACHO
* PSE member