1. Vijay Krishna and John Morgan Overcoming Ideological Bias in Elections
Imagine that voters care about the quality of the candidate but also lean towards one of the two candidates. With sincere voting the larger party wins - which will be inefficient if they don't care very much and the minority party does. Now take an off-the-rack voter participation model with stochastic participation costs and lots of strategic voters. Because the party that cares more is more willing to incur participation costs they have a better chance of winning. So much so that - amazingly - majority voting gets the outcome exactly right.
2. Geoffroy de Clippel and Kareen Rozen Bounded Rationality and Limited Datasets
Imagine testing a theory of decision-making on a subset of all conceivable data. If the theory is about rationalizing choices by a preference ordering, consistency of that theory with a subset of data means that the theory hasn't been rejected "so far". That isn't true of several theories of bounded rationality: theories consistent with limited data can nevertheless be proved to be inconsistent with the larger data set --- even if we haven't observed the larger data yet!
3. Sandeep Baliga and Jeffrey Ely Torture
In the eyes of many, torture is problematic on ethical grounds, but effective and sometimes necessary. This paper develops a simple dynamic model to question this view: when the torturer does not have commitment power, torture may indeed provide stronger incentives to stay silent than to speak. If the suspect speaks and reveals to be informed, the torturer does not find it optimal to reward collaboration: he/she persists until all information is extracted. If the suspect keeps silent, on the contrary, the torturer may stop after persuading himself that the suspect is uninformed.