Accuracy Is (Generically) Bad For Compliance

TL;DR

This paper shows that optimizing for compliance and accuracy generally require different incentive schemes, as the set of distributions where they align is very small, challenging the common assumption that accuracy is a good proxy for compliance.

Contribution

It proves that the conditions under which maximizing compliance and accuracy coincide are extremely rare, highlighting the need for separate incentive design.

Findings

Maximizing compliance and accuracy usually require different incentives.

The set of distributions where they align is finitely shy.

Accuracy is not a reliable proxy for compliance in most cases.

Abstract

We demonstrate that the set of cost distributions under which the optimal strategy for maximizing compliance (or more generally, effort) in a binary choice environment is identical to the optimal strategy for maximizing the accuracy of the reward (minimizing Type-I and Type-II errors) is finitely shy (Anderson and Zame (2001) in the space of all smooth parameterized real-valued distributions possessing full support on the real line. In words, this implies that maximizing compliance and maximizing accuracy "almost always" call for different incentive schemes.