Are people conditionally honest? The effects of stakes and information about others' behavior

We study theoretically and empirically how monetary incentives and information about others' behavior affects dishonesty. We ran a laboratory experiment with 560 participants inspired by the "observed game" developed by Kajackaite and Gneezy (2017). We find that the extensive (the fraction of liars) and intensive (the size of the lie) margin of dishonesty decrease when stakes are very high. On average, information about others slightly increases the fraction of liars but has no effect on the size of the lie. Distinguishing subjects by their belief on others' behavior, we find that information decreases the fraction of liars among over-estimators and increases the fraction among under-estimators. This pattern is the same across payoff levels.