Many important decisions within public and private organizations are based on recommendations from expert committees and advisory boards. A notable example is the U.S. Food and Drug Administration's advisory committees, which make recommendations on new drug applications. Previously the voting procedure for these committees was sequential, however, due to concerns of herding and momentum effects the procedure was changed to simultaneous voting. Exploiting a novel dataset of more than ten thousand votes cast by experts in the FDA committees under both sequential and simultaneous voting, we estimate a structural model that allows us to measure the magnitude and importance of informational herding. We show that experts, voting on important scientific questions, are susceptible to herd behavior; on average 46% of the members take into consideration the sequence of previous votes when casting their vote, 17% of these voters actually herd i.e. change their vote from what they would have voted if ignoring the preceding votes.
Keywords: Herd behavior, expert committees, structural estimation, FDA, pharmaceuticals
Frei zugängliche Version: (econstor)