We conduct a natural field experiment in fish markets where sellers frequently cheat on weight and face negligible economic penalty. Exploiting exogenous variations in fish prices, an indicator of marginal economic benefit from cheating, we examine how dishonest behavior varies with rising economic benefit from cheating. We find that most sellers cheat but that cheating almost never exceeds ten percent of purchased quantity, and that the value of cheating is small. The data reveal a non-monotonic relationship wherein cheating initially increases and thereafter decreases in the fish price.