## Highlights 2024

**WP 24/04**

In “When is Trust Robust?” Daniele Terlizzese, together with Luca Anderlini and Larry Samuelson, argue that an equilibrium with a high level of trust is fragile, as it can be disrupted by the infusion in the society of a small mass of agents that behave as if the prevailing equilibrium were one with low trust. Conversely, an equilibrium with low trust can withstand a much larger infusion of agents believing to be in the high trust equilibrium. The result is a consequence of the convexity of the cost that society imposes on cheaters. The convexity, in turn, follows naturally from the assumption that society punishes more heavily cheaters who are believed not to have a good reason for cheating (called “scoundrels” in the paper). Surprisingly, the fewer are the scoundrels, the more fragile is the high trust equilibrium.

**WP 24/03**

In "Bayesian estimation of the normal location model: A non-standard approach", Franco Peracchi, together with Giuseppe De Luca and Jan R. Magnus, consider estimating the location parameter in the normal location model and study the sampling properties of estimators derived from a Bayesian approach that places a prior on a scaled version of the location parameter, interpreted as the `population t-ratio'. This non-standard approach is motivated by the fact that, in model selection and model averaging, it is the t-ratio rather than the parameter estimate that plays an important role. They show that the finite-sample distribution of the proposed estimators is not centered at the value of the target parameter, and is generally non-normal. They also show that the speed at which the estimation bias vanishes as the sample size increases critically depends on the choice of prior. In the asymptotic theory, they prove uniform √n-consistency of the proposed estimators and obtain their asymptotic distribution under a general moving-parameter setup that includes both the fixed-parameter and the local-parameter settings as special cases. Their results have direct implications for the WALS estimator of Magnus, Powell and Prüfer (2010) and may also be helpful for other model selection or model averaging procedures.

**WP 24/02**

In “Tournament Auctions” Luca Anderlini and GaOn Kim examine auctions in which all but one of the participants bid first to gain the opportunity to compete with the last bidder. Only the winner of the first round competes with the last bidder in a second-price round. This second round finally assigns the object to be sold to the highest of the two bidders. At first sight this may look like a twist on a standard auction format, but instead it has unexpected features. Bidders in the first round consistently bid above their values. If the last bidder is sufficiently stronger than the others and there are sufficiently many bidders, the tournament auction increases the expected revenue for the auctioneer relative to the standard auction format.

**WP 24/01**

In "Fear to Vote: Explosions, Salience, and Elections" Mounu Prem, together with Juan F. Vargas, Miguel E. Purroy, Felipe Coy, and Sergio Perilla study how antipersonnel landmines thwart democratic accountability and the consolidation of post-conflict democratic institutions. They do so by exploiting the randomness in the timing of landmine explosions relative to election days, comparing the electoral outcomes of voting polls located close to a pre-election explosion with those of polls near a post-election blast. They show that landmine explosions are salient stimuli that pro- duce fear, reducing political participation. While the turnout reduction takes place across the ideological spectrum, they document that the explosions induce shifts in the political preferences of individuals who do vote, which are inconsistent with retrospective voting.