While stock market expectations are among the most important primitives of portfolio choice models, their measurement has proved challenging for some respondents. We argue that the magnitude of measurement error in subjective expectations can be used as an indicator of the degree to which economic models of portfolio choice provide an adequate representation of individual decision processes. In order to explore this conjecture empirically, we estimate a semiparametric double index model on a dataset specifically collected for this purpose. Stock market participation reacts strongly to changes in model parameters for respondents at the lower end of the measurement error distribution; these effects are much less pronounced for individuals at the upper end. Our findings indicate that measurement error in subjective expectations provides useful information to uncover heterogeneity in choice behavior.