Contract breach with overconfident expectations: Experimental evidence on reference-dependent preferences

This study examines the effect of agents' overconfident expectations in their production on their contract breach. Drawing on a reference-dependent framework, we theoretically deduce propositions for compliance to agreements where an agent exhibits overconfidence and loss aversion. We further conduct a lab experiment with a multiple-stage design and find that overconfident agents are more likely to breach the contract than non-overconfident agents. Moreover, overconfident agents breach more often and to a greater extent with increasing loss aversion. We also test the impact of a non-deterministic environment (“shock condition”) where payoff misestimation can be masked compared to a deterministic environment (“no-shock condition”). Agents breach more often in the shock condition, but breach extent remains unaffected. Results are mostly in line with the theoretical framework. In a treatment, we manipulate agents' overconfidence exogenously and use it as an instrument to establish causality. • Our reference payoff framework predicts that unmet expectations from overconfidence bias and individual loss aversion increase contract breaches. • Experimental results confirm that overconfident agents who are loss averse breach more than non-overconfident agents. • We establish causality by manipulating overconfidence with a treatment making use of the anchoring bias. • Evidence that reference points can emerge endogenously from overconfidence and lead to unethical behavior. • The study is relevant for non-enforceable agreements, promises, and contracts.