Although early economic approaches to misbehavior merely compare the monetary utility of accessible options, self‐concept maintenance models introduce moral considerations to the equation. These assume that people trade off possible gains to be made from moral transgressions with associated decreases in self‐esteem. On the basis of the assumption that the development of moral values among children is weaker than in adults, we expected children's behavior to be close to that of the hypothetical homo economicus, with their decisions as to whether to cheat therefore being influenced by decision theory elements: normative elements such as temptation magnitude, and behavioral elements such as framing. As children should pay no heed to moral considerations in dynamic multiple task settings, behavior opposite to “moral cleansing” was expected, with a first lie predicting later lies. In addition to testing the above ideas, the present study adopted a novel methodology. Our hypotheses were tested in a lab‐in‐the‐field study using a modified “roll a die” method in a naturalistic setting for children of ages 7 to 10. We modified the method to identify both true and declared values of die rolls (a novel DICE+ electronic die was used). As expected, children were sensitive to temptation and cheated more willingly for more attractive prizes. Girls (but not boys) lied more (in terms of both frequency and magnitude) to avoid losses (with loss framing) than to make gains (with gain framing). Previous lying correlated positively with lying on subsequent tasks.