Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Beliefs, alongside preferences, are an important driver of behaviour. While preferences are often inferred by the choices made, measuring beliefs is not straightforward. We design a giving experiment to compare different methods of measuring beliefs, with and without monetary incentives. Consistent with a simple theoretical framework, we find that elicited beliefs about the giving decisions of others are biased and self-serving when no incentive is offered, with non-donors reporting that giving is rare. Offering a simple incentive does not reduce the bias in beliefs; however, this bias is not observed when using an incentivised method which makes the monetary outcome associated with accurately predicting beliefs more prominent. Our findings suggest that when self-interested motivations compete with accuracy incentives, beliefs are sensitive to how they are measured.