The (Un)Compromise Effect

Abstract

The compromise effect—i.e., the preference for the middle option—is an established bias in behavioral economics, but has not been experimentally validated in the field. In the current study I test the compromise effect in a natural context, and whether this bias can be used to stimulate active choice—the (un)compromise effect. In a mail fund raiser for a large US hospital, I evaluate their baseline ask string [$10, $50, $100, $ ] to an extended version [$10, $50, $100, $250, $500, $ ]. In line with the compromise effect, the extended ask string increases the average amount given and the share of donors giving $100, which is now the middle option. Importantly, however, and in line with a model of contextual inference, revealing the middle option is not necessary for the effect to arise. The (un)compromise ask string [$10, $500, $ ] generates the same average amount given and same share giving $100, as the extended ask string—the only difference being that 90 percent of donors, instead of 30 percent, use the open-ask alternative. Hence, by only providing informative end points of a distribution, organizations can benefit from the compromise effect and at the same time promote individuality by stimulating active choice. I discuss theoretical and practical implications of the results.