Abstract: I develop a framework to analyze student decision-making in multiple-choice tests. I use it to show that a broad class of benchmark assumptions imply a common behavioral prediction: when a student omits more questions, their relative share of correct answers should not decrease. Using data from a large field experiment, I document systematic violations of this principle. To rationalize the puzzle, I propose a model of imperfect metacognition where students differ in the ability to monitor their own knowledge, compatible with a Bayesian foundation. Calibrating the model parameters, I uncover meaningful gender heterogeneity: although equally metacognitively accurate, females are relatively better at recognizing low-success questions, while males high-success ones. This asymmetry implies that the gender gap in test scores should widen with knowledge -- a prediction confirmed in the data and not replicated by alternative explanations.

Abstract: We study the supply and effects of causal rhetoric in U.S. politics. We conceptualize causal rhetoric through blame and merit -- claims that assign negative or positive outcomes to political action -- and train a supervised classifier to detect these forms in over a decade of congressional tweets. Causal rhetoric has risen rapidly and pervasively, displacing affective messaging. Its production trades off revenues and costs. Quasi-random variation in Twitter adoption shows that blame increases small-donor revenues by expanding donor count, while merit raises average donation size. Fine-grained legislative data suggest that policy ownership determines relative costs: blame is cheaper for opponents, merit for proposers. Finally, causal rhetoric has downstream effects on societal outcomes, influencing protest activity, polarization, and institutional trust.

Abstract: In a political economy framework, we study how false narratives emerge in response to limitation in recipients' memory, proposing an approach based on partial identification. Coarse memory allows voters to retrieve marginal frequencies of past policies and outcomes, but prevents them from understanding their correlations. Politicians exploit such vagueness by designing narratives inflating the effectiveness of their commitment policy, while satisfying the constraint imposed by marginals (plausibility). We find that plausible narratives can be less optimistic about a policy the more it has been implemented. In a model of narrative-based competition, we show that opposing narratives are polarized and that, in the long run, political cycles arise. In particular, politicians hold office with the same frequency, irrespective of the objective quality of their commitment policy. Our mechanisms are consistent with an analysis of U.S. Congress members' rhetorical strategies.