Theories may be informative to agents. When agents' actions are determined by the predictions of a theory, the predictions endogenize the data that they are tested against. In this case, it is unclear in what manner these theories can be regarded as scientific. I attempt to characterize this problem in a general framework, and state an idealized criterion for theory selection. I then informally suggest the application of this framework in the analysis of various settings.

The paucity of numerical data need not result in scholars implementing ad hoc empirical methods in their study of ancient economies. Using the method of partial identification, I give an example of how formal econometric analysis may be implemented in the study of ancient economies, and show how this method allows scholars to explicitly demonstrate the effect on analyses of their beliefs regarding the representativeness of archeological evidence.