Abstract
There is a growing body of research showing that people altruistically enforce cooperation norms in social dilemmas. Most of this research analyzes situations where norm violators are known and group members enforce cooperation among each other. However, in many situations norm violators are unknown and detection and punishment is enforced by third parties, such as in plagiarism, tax evasion, doping or even two-timing. Our contribution is threefold. Conceptually, we show the usefulness of inspection game experiments for studying normative behavior in these situations. Methodologically, we present a novel measurement of strategic norm adherence and enforcement, asking for continuous, "frequentistic" choice probabilities. Substantively, we demonstrate that norm adherence in these situations is best understood by coexisting distinct actor types. Self-regarding types learn the inspection rate and calibrate their norm violations to maximize own payoffs. Other-regarding types reciprocate experienced victimizations by stealing from other, unknown group members; even at additional costs. We specify both mechanisms by agent-based simulation models and compare their relative strength by behavioral and attitudinal data in inspection game experiments (N=220). Our results suggest a modern sociological perspective, which combines homo oeconomicus with homo sociologicus. Further, our findings contribute to understanding conditional norm compliance in "broken windows" dynamics, since we show under controlled conditions that it may result jointly from self- and other regarding mechanisms.