Public credit guarantees should be provided to firms that are both creditworthy and credit constrained. We use Machine Learning (ML) predictive tools to propose a targeting rule that includes both objectives. The study elaborates on the case of Italy's Guarantee Fund and demonstrates, by means of ex-post evaluation methods, that the program effectiveness can be increased by ML targeting. We discuss some of the problems in using algorithms for the implementation of public policies, such as transparency and manipulation.

Negative Emissions Technologies (NETs) — a range of methods to remove carbon dioxide from the atmosphere — are a crucial innovation in meeting temperature targets set by international climate agreements. However, mechanisms which undo the adverse consequences of short-sighted actions (as NETs) can fuel substitution effects and crowd out virtuous behaviors (e.g., mitigation efforts). For this reason, the impact of NETs on environmental preservation is an open question among scientists and policy-makers. We model this problem through a novel restorable common-pool resource game and use a laboratory experiment to exogenously manipulate key features of NETs and assess their consequences. We show that crowding out only emerges when NETs are surely available and cheap. The availability of NETs does not allow experimental communities to either conserve the common resource for longer or accrue higher earnings and makes the earnings distribution more unequal. 

This paper investigates the political consequences of perceived health risks associated with immigration in Italy. We leverage the exogeneity of the 2014 Ebola epidemic, which resulted in almost no cases in Italy but triggered a significant public reaction, with extreme right-wing politicians claiming ongoing immigration could endanger citizens’ health. In a differences-in-differences framework, we examine the changes in the vote share of the main right-wing and anti-immigration party, Lega, across Northern Italian municipalities before and during the Ebola outbreak. Treatment is based on perceived exposure to risk-Ebola immigrants, proxied by the local historical concentration of immigrants from countries affected by Ebola in 2014. Results document a drop in political support for Lega in municipalities with a larger share of risk-Ebola migrants. Our findings, robust to falsification tests and alternative treatment definitions, suggest that strategically exploiting a health crisis to garner support for anti-immigrant policies can eventually backfire.

In times of hardship, politicians often leverage citizens' discontent and scapegoat minorities to obtain political support. This paper tests whether political campaigns scapegoating migrants for a health crisis affect social, political, and economic attitudes and behaviors. Through an online nationally representative survey experiment in Italy, we analyze the effects of such narratives through information-provision treatments, which include facts also emphasizing the alleged health consequences of ongoing immigration. Results show that narratives associating immigration with health threats do not generate sizeable add-on effects compared to those based on immigration only. If anything, they increase disappointment towards co-nationals, reduce institutional trust, and undermine partisanship among extreme-right supporters. Results are consistent with a theoretical framework where party credibility and support, and institutional trust are influenced by political discourse. Our experiment underpins the prediction that political campaigns based on extreme narratives can be ineffective or socially and politically counterproductive, providing an example of how populism can backfire.

In the absence of conclusive evidence from the literature and in the attempt to better understand some of the so-far unexplained variation of cooperation levels observed in experimental data, this paper investigates the role of individuals’ social preferences in shaping cooperation in Infinitely Repeated Prisoner Dilemmas. To this aim, we measure individuals' social preferences and analyze behavior across contexts that differ in terms of strategic incentives for cooperation, as defined by the environmental game parameters. This allows us to neatly test whether social preferences have an effect per se on cooperation, and to analyze whether the effect of social preferences is confounded in contexts where cooperation could be sustainable even in the absence of social preferences. Novel experimental evidence is complemented by the evidence from a meta-analysis run on an extended version of the dataset collected by Dal Bó and Fréchette (2018), where we rely on simple supervised-learning algorithms to test the ability of environmental game parameters to predict cooperation. Results from the meta-analysis show that strategic incentives alone, defined by environmental game parameters, predict cooperation increasingly better over supergames, as subjects gain experience, and in contexts where cooperation is not sustainable as an equilibrium, as opposed to the cases in which it is. Our experimental evidence proves that social preferences do, indeed, play a relevant role in shaping cooperative attitudes, both in one-shot and Infinitely Repeated PDs. It also highlights the key role played by beliefs, which explain a large fraction of the variation observed, serving as a transmission channel for the effect of social preferences on behavior.

Many digital platforms operate as two-sided markets, facilitating the matching between sellers and buyers. The Amazon case is particularly interesting because of the dual role the tech giant plays: Amazon both controls the rules of the game, through platform design choices, and participates as a player itself, operating as a seller. What do consumers know about Amazon’s platform functioning? Would they behave differently if they knew how the type and the amount of information they are provided with is filtered/manipulated? This paper focuses on a peculiar and distinctive algorithmic-driven Amazon’s feature: the Buy-Box. Through a framed field experiment, we collect individual-level data on real consumers’ search and purchase behavior on the marketplace in a controlled environment, where we can manipulate the amount of information disclosed on the functioning (and the hidden risks) of Amazon’s Buy-Box. Experimental data reveal that - in the absence of any information provision - more than 80% of the participants do not even inspect other than the Buy-Box (default) deal. Providing participants with (full) information on the Buy-Box functioning and the availability of alternative deals by competing sellers significantly increases the probability of inspecting alternative deals before finalizing the purchase, and the propensity to select a seller other than the one in the default Buy-Box position. At the same time, the average time spent to finalize the purchase is not significantly different.

We propose an experimental analysis to investigate the impact of social learning on individuals' acquisition of information before making a choice and how behavioral biases and the perceived reliability of the information source affect this process. We use a stylized version of the canonical sequential search model and solve it with and without social information. Agents without social information act in isolation, while those with social information observe a peer's choice but not their search process. We aim to explore the effects of social information on search behavior, the influence of the perceived quality of the information source, and the potential of alternative institutions to counteract behavioral biases and restore the benefits of social information. Our results can guide the development of policies that enhance the efficiency of decision-making processes by optimizing the use of social information.