Threshold models have been pivotal in understanding interdependent behavior, yet empirical research on individual thresholds is nonexistent. We introduce an incentive-compatible method for eliciting individual thresholds. We use it to elicit thresholds for supporting affirmative action in a large, stratified sample of the U.S. population. Consistent with threshold models, most Asian, Black, Hispanic, and White men and women condition their support on the number of others who do so. In line with our hypotheses, thresholds are influenced by personal benefit, fairness perceptions, and conformity. Our method can be used to predict societal outcomes and inform policy design across various situations.
