Individual choices are often inconsistent with economic theories, which has motivated a variety of ways to measure how far choices are from a given theory. Recent work has investigated the correlation between “measures of rationality” and observable information such as education or income. This paper investigates the sensitivity of this analysis to the units used to measure rationality, in particular we examine measures in percentage expenditure vs dollars expenditure. We find that correlations can change sign when we change the units, (...) but this sensitivity might be mitigated with experimental methods. (shrink)