2025

Dynamic Persuasion and Polarization  - Lara Deniz ALPER (Universitat Pompeu Fabra) 

Abstract: Why does polarization persist even when individuals are exposed to identical information? I study sender-receiver interactions, where a sender proposes a set of models to multiple receivers, each of whom may entertain a different default model. These receivers observe a common exogenous signal and they each adopt the model that best fits their prior beliefs and the observed signal. Subsequently, based on their selected models, each receiver chooses an action to maximize their expected utility. I characterize the sender’s optimal strategy and show that it always involves an extreme parameter configuration—that is, a model in which at least one signal probability is set to zero or one. Under any such strategy, receivers’ beliefs almost surely converge to the boundary of the belief space, resulting in long-run polarization. These results help explain how enduring ideological divergence can arise even when agents are exposed to the same informational environment. I apply these results to understand media and political polarization.

Strategic Disclosure of Evidence Produced On Market - Michal TULWIN (European University Institute)

Abstract: This paper studies a two-period market interplay between a monopolistic seller and two cohorts of short-lived consumers varying in private valuation for the product. Initial buyers generate honest reviews which inform next generation of customers, but are subject to strategic manipulation by a firm. This paper introduces a model of price and evidence signaling, where the informativeness of reviews and the scope for manipulation is determined by the uncertainty regarding the number of early buyers. Downward pressure on the initial price increases with uncertainty regarding the share of high valuation consumers, causing high type firm chooses to lower its price more frequently. Consumers may benefit from a decrease in quality of a low type firm, when it intensifies signaling via the initial price setting, and are weakly better-off when manipulation is possible compared to full disclosure model. Then, the effectiveness of two modes of manipulation - removing vs adding fake reviews - and their respective impact on market efficiency is discussed.

Information Provision with Endogenous States - Sanjari KALANTRI (University of Oxford) 

Abstract: In this paper, I study a dynamic model of information provision. A long-lived sender with commitment power provides information about a state of the world to a receiver, who then chooses an action to maximise his short-term payoff. The players have misaligned preferences that depend both on the state and the action. Importantly, the state evolves over time based on the action taken by the receiver. Thus, the sender faces a trade-off: the short-term gain from disclosing information that maximises her current payoff, versus the future benefit from influencing the evolution of the state, via the receiver’s action. I characterise the sender’s optimal information provision rule when the state evolution process is positively associated, negatively associated and independent of the action taken by the receiver.

Can hate speech be banned online? The effects of shutting down toxic forums on Reddit - Adam DI LIZIA (University of Warwick)

Abstract: Is deplatforming effective in reducing toxicity on social media? To answer this question we study a change in content policy on a large social network Reddit in June 2020. The change led to a simultaneous ban of approximately 2000 forums containing hateful content, but not the users of these forums. We use the complete history of comments on Reddit to examine the impact of the ban on the behaviour of affected users in a differences-in-differences design. We find that the most active users of banned subreddits increase their use of the platform after the policy change. The effect is heterogeneous: “core users” of banned left-wing subreddits leave up to 20% more comments, while users of banned right-wing forums comment up to 10% more often. This increase in activity does not appear to associated with higher toxicity: instead, the comments become more positive, contain fewer slurs, and are 10% less likely to be rated controversial by other users. Remarkably, the users not only change the way they comment in the non-banned forums they used before the policy change, but also substitute away from communities with high concentration of other users affected by the ban. Overall, the results suggest that, instead of seeking new spaces to express toxic opinions, users prefer to substitute to less extreme forums and change the way they engage with others, leading to a reduction in levels of toxicity on the platform.

Optimal Policy Design for Raising Teacher Quality - Juan PAL (Toulouse School of Economics)

Abstract: This paper studies the optimal design of higher education policies targeted at improving teacher recruitment. I leverage the introduction of a policy in Chile that aimed to raise teacher quality by crowding-in higher performing students into Education programs, while crowding-out the lower-performing ones. Exploiting the sharp assignment rule, I estimate that, at the threshold, enrollment of high performing students at teacher colleges increased by 42%, and there is a shift in the distribution of test scores. Students then proceed to perform better at the labor market, with an increase in 0.11SD in Teacher Value Added and 0.12SD in a Teacher Evaluation exam. I develop a general equilibrium model of the higher education market to study the optimal design of the policy. In doing so, I present a novel method for solving discrete-continuous games in sparse markets. Counterfactual simulations show that alternative policies could have outperformed the observed one in terms of student quality, but at the cost of either (i) a higher fiscal burden or (ii) reduced market share of teacher colleges. The distribution of college quality together with capacity constraints limits the impact on teacher value added. Several mechanisms make a policy focused on low-income students to yield further gains at no additional cost.

The Criminal Trap: Low School Attendance and Organized Crime - Ignacio LEON NINA (Université Paris 1 Panthéon-Sorbonne)

Abstract: This paper investigates the causal impact of reduced school attendance on youth criminal behavior in Ecuador. Leveraging administrative data on arrests and incarceration, I exploit variation in high school dropout rates across cohorts and cantons during the COVID-19 pandemic, implementing a difference-in-differences framework with a continuous treatment. The results indicate that pandemic-induced declines in school attendance led to a significant increase in youth crime: incarceration rates rose by 20.8 percent among high school cohorts in cantons with elevated dropout rates, relative to those with stable enrollment. The effect is more pronounced in areas with greater exposure to organized crime, driven in part by the post-2017 surge in cocaine production in Colombia. Importantly, the relationship emerges only after 2020, suggesting that the pandemic-induced labor supply shock was a necessary condition for the demand shock from organized crime to translate into higher youth criminality.

A European Safe Asset? Not Without the Investors - Giovanni BONFANTI (Columbia University)

Abstract:  I study the attempt by the European Union to issue bonds that are a common liability of the member countries through various supranational institutions. I show that this debt requires an interest rate that is higher than for comparably safe, large, and liquid sovereign issuers and more exposed to expectations about monetary policy. I develop a model that focuses on the role of investment mandates that arbitrarily reduce the size of potential investors in supranationals relative to equally safe governments. The smaller size of potential buyers translates into lower expected liquidity during crises such that supranationals must pay a premium even during normal times. This mechanism is driven by investors with potential liquidity needs during crises such as mutual funds. Expectations about asset purchases during crises, which I call conditional QE, can significantly compress this premium even if they are not targeted. 

International Risk-Sharing and Inequality Transmission: An Application to the Euro Area - Naomi COHEN (Sciences Po)

Abstract: Bilateral holdings of financial assets serve as a smoothing mechanism, aiming to alleviate the correlation between fluctuations in consumption and national output for both countries in the case of country-specific shocks. However, within countries, households exhibit unequal access to diversification opportunities, with financial assets, and particularly foreign assets, concentrated among the wealthiest households. Furthermore, asymmetric portfolios and cross-border capital holdings result in different intensity of diversification and, consequently, differing spillover effects from country-specific shocks. These two channels highlight the heterogeneous impact of portfolio diversification within and across countries, with important implications for inequality. This paper examines the interaction between inequality and portfolio diversification in the context of how country-specific shocks affect inequality in the euro area. While theoretical models assume full home bias in equity by restricting investment to domestic capital, I provide empirical evidence of portfolio heterogeneity both within and between countries. This evidence demonstrates differences in access to and intensity of diversification, using household-level and country-level data. To analyze these dynamics, I develop a two-country Two-Agent New Keynesian (TANK) model within a monetary union, calibrated to represent the Core and Periphery of the euro area. The model incorporates the ability to invest in both domestic and foreign capital through a risk-neutral mutual fund, which allocates investments based on empirically observed portfolios. I focus on the consumption responses of households within a country and among similar agents across countries when subjected to a country-specific shock. The results reveal that incorporating realistic portfolio structures significantly alters aggregate and individual consumption responses. Notably, the costs and benefits of such shocks are unequally distributed across households and countries.

The Credit Channel of Inflation - Lorenzo RANALDI (University of Bonn)

Abstract: This paper documents a novel channel through which inflation affects the real economy via bank balance-sheets: unexpected inflation erodes banks' net worth by devaluing long-term loans relative to short-term liabilities, restricting lending capacity and depressing real economic activity. Using 150 years of cross-country data, I show that inflation surprises reduce aggregate loans and banks' equity returns, even absent monetary tightening. Exploiting bank-firm micro data, I find that firms reliant on inflation-exposed banks experience declines in activity, demonstrating that banks’ balance-sheet pressures can offset traditional debt-inflation benefits.

Technological Changes and Worker Skills - Allen SHEN (University College London)

Abstract: This paper studies worker human capital responses to technological changes. Centered on a multidimensional view of human capital, we examine how exogenous skill-specific technological shocks influence worker decisions to engage in training, their occupational choices, and the resulting implications for aggregate output and efficiency. Empirically, we quantify the causal effect of technological changes on the supply of skills. We find that a one-standard-deviation increase in the rate of technological arrival at the occupation level is associated with a 9pp increase in the probability of training. We uncover the heterogeneous effect of technological changes on differently skilled workers. Furthermore, we establish a precautionary motive, embodied in skill portfolio rebalancing, in response to stochastic technological changes. Our results are robust to an instrument of technological arrival. We rationalise our empirical results with a model of multidimensional human capital investment under risk aversion and uncertain future skill productivity. This model captures important empirical features including skill heterogeneity, sluggish skill adjustment and the precautionary motive. We qualitatively demonstrate under-specialisation compared to a frictionless baseline. We are currently assessing the macroeconomic efficiency consequences of the under-specialisation by embedding the worker problem in a search-and-match framework. This will allow quantification of the skill-occupation misallocation and experiments with policy counterfactuals, such as training subsidies.

The Dual Role of Inequality for Structural Change: Cause and Consequence - Lucas WIEDEMANN (London School of Economics)

Abstract: There is a two-way relationship between inequality and structural change. To assess the different mechanisms in this intersection, I build a model with heterogeneous production intensities, non-homothetic CES preferences over sectors, and consumption as a composite of market and home origin within a sector. Calibrating the model to the U.S. over the time period 1980-2016, I find three results indicative of the role of inequality for structural change: First, exogenous increases in inequality contributed to 20% (3pp) of the rise in the high-skill services value added share. Amplification accounts for a further 5%. Second, a counterfactual tax exercise shows that if income taxes would have stayed as progressive in 2016 as they were in 1980, the high-skill services share would have increased up to 2.5pp less. Third, while the marketization of home production drives the expansion of low-skill services in the first half of the period, in the post-2000 era of high inequality, the rich contribute nearly twice as much as the poor to overall marketization.

Public Debt and the Climate Transition in Emerging Markets - Ornella TORRES (Paris School of Economics)

Abstract: Institutional investors’ preferences are quickly shifting towards green investments. Emerging markets are the first exposed to this shift: these are carbon-intensive economies, still very dependent on official funds, and with high debt burdens. Opting for a rapid green transition carries risks, as it requires significant borrowing with uncertain future benefits. However, staying on a carbon-intensive path increases the average cost of funds, as countries have to rely on expensive market debt. This paper first builds a small open economy (SOE) model with heterogeneous investors and collateral constraints to study the SOE’s investment and borrowing dynamics over the business cycle. The paper then looks at institutional investors’ optimal lending policy in a multi-country setting.

Imperfect Metacognition in Multiple-Choice Tests - Francesco BILOTTA (Università Bocconi)

Abstract: I develop a framework to analyze student decision-making in multiple-choice tests. Leveraging it, I 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 Bayesian cognitive noise. 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.

Deep Learning Across Games - Massimiliano FURLAN (University of Warwick)

Abstract: We train two neural networks adversarially to play static games. At each iteration, a row and column network observe a new random bimatrix game and output individual mixed strategies. The parameters of each network are independently updated via stochastic gradient descent on a loss defined by the individual squared regret experienced in the game. Simulations show the joint behavior of the trained networks approximates a Nash equilibrium in all games. In 2x2 games with multiple equilibria, the networks select the risk dominant equilibrium. These findings, which are robust and generalize out of distribution, illustrate how equilibrium emerges from learning across heterogeneous games.

A Behavioural Theory of Persuasion - Giuseppe PULEIO (Paris School of Economics)

Abstract: I work on a principal-agent model where a long lived- perfectly informed- sender interacts with two short lived- partially informed- cohorts of receivers. The sender offers sequentially a unique contract inducing a more favourable lottery (FOSD) to the receivers, but the exact stochastic dominance magnitude is unknown to them. Second period receivers employ the observations generated by the first period to estimate the sender type- by an MLE procedure. I show that, when receivers rely on a representative large dataset, the optimal sender strategy need not to converge to the optimal one under perfect estimation. In case of perfect estimation, I give a sufficient condition for the uniquely optimal strategy to be pooling. I discuss an extension with Bayesian receivers.

Screening Women Out? Experimenting with Salary Disclosure in Job Ads - Amen JALAL (London School of Economics)

Abstract: Firms vary in pay for otherwise similar jobs. Barriers that limit women's access to high-paying firms, or their ability to negotiate the available premiums, widen gender gaps in the labor market. This paper examines whether the absence of salary information in job ads restricts women's access to high paying firms. I find that firms are more likely to hide salaries when they are higher, for otherwise comparable jobs. In particular, large, well-paying firms use salary non-disclosure to reduce costly screening by encouraging applicants to self-screen. This practice has disproportionate costs for women, who, like men, apply more to higher paying jobs when salaries are visible, but, unlike men, do not direct search when information is missing. To investigate, I conduct an experiment on Pakistan's largest job search platform with around 20,000 jobs and 8,900 firms. Treated jobs are required to disclose salary ranges, while control jobs retain the option to hide. The intervention has limited effects on salaries, but increases applications by 49%, especially to large firms whose higher salaries are newly revealed. The reallocation of search from small to large firms is 43 percentage-points higher for women, eliminating gender gaps in sorting that were induced by lack of salaries. While the applicant pool slightly declines in average quality, top applicants to treated jobs are higher ability and 4.2% more likely to be female. Finally, salary non-disclosure disproportionately deters high ability women from applying to high paying firms, resulting in potential underutilization of talent.

Familiar strangers: Evidence from referral-based hiring experiments in India - Kartik SRIVASTAVA (Harvard University)

Abstract: Across developing countries, search and matching frictions in labor markets are associated with high turnover rates in entry-level jobs. At the same time, large fractions of the workforce remain without access to certain jobs. Firms often rely on referral-based hiring in these settings to reduce search costs and improve match quality. However, these practices can systematically exclude underrepresented groups and exacerbate inequalities. This paper asks whether reallocating referral opportunities toward historically disadvantaged groups can improve firm performance and worker welfare, focusing on the role of caste in India. In a manufacturing firm in India, I experimentally increase the probability of referral opportunities being shared with lower caste incumbents. I find that this increases their representation by 15 percentage points (62%), and reduces turnover by 4 percentage points (41%) owing to worse outside options for these entrants. Output improves by 0.09 sd, and in contrast to the literature on the adverse e!ects of mixed teams on cohesion, I find no significant e!ects on this in my setting. A lab-in-field experiment establishes a mechanism driving the results: cohesion is lower by 9% when lower caste workers are hired as outsiders without links to incumbent workers, which reduces total team output by 22% relative to when these workers are hired through referrals.

Work from Home, Quit Behavior and Monopsony Power: Evidence from the Italian Labor Market - Nicol BARBIERI (Università di Napoli Federico II)

Abstract: This paper studies the impact of Working from Home (WFH) on voluntary quits in the context of the Italian labor market. It specifically investigates how and to what extent the adoption of remote work during the COVID-19 pandemic shaped workers’ quitting behavior and how this translated into changes in employers’ market power. Using very rich administrative employer-employee data, I implement a Difference-in-Differences (DiD) event-study analysis that exploits pre-pandemic variations in remote work exposure across occupations to identify the causal effect of WFH on the likelihood of voluntarily quitting. Then, I investigate whether this shift translates into changes in the voluntary quit elasticity, particularly among occupations better suited for remote work. My results show that workers in highly remote-exposed occupations experienced a significant reduction in voluntary quits compared to those in occupations with lower remote feasibility. This drop is related to lower wages and workers’ lower sensitivity to wage changes, potentially indicating that WFH acts as a substitute for monetary compensation. Overall, these results suggest that employers in remote occupations can retain workers while offering relatively lower wages, possibly supporting monopsonistic tendencies.