Working papers

Raise your voice! Activism and peer effects in online social networks (PDF)

Abstract: Do peers influence individuals’ involvement in political activism? To provide a quantitative answer, I study Argentina’s abortion rights debate through Twitter - the social media platform. Pro-choice and pro-life activists coexisted online, and the evidence suggests peer groups were not too polarized. I develop a model of strategic interactions in a network - allowing for heterogeneous peer effects. Next, I estimate peer effects and test whether online activism exhibits strategic substitutability or complementarity. I create a novel panel dataset - where links and actions are observable - by combining tweets’ and users’ information. I provide a reduced-form analysis by proposing a network-based instrumental variable. The results indicate strategic complementarity in online activism, both from aligned and opposing peers. Notably, the evidence suggests homophily in the formation of Twitter’s network, but it does not support the hypothesis of an echo-chamber effect.

Media: uc3nomics (English); Nada es Gratis (Spanish).

Work in progress

Hate speech in social media: evidence from Bolsonaro’s election in Brazil

with D. Marino Fages.
Status: draft in progress.

Abstract: How does newly available information affect individuals’ perception of social norms and, consequently, behavior? We examine the impact of Bolsonaro’s victory in the 2018 Brazilian presidential election on the prevalence of online hate speech. This project relies on Twitter data from 2017 to 2019 and employs text analysis techniques to detect hate speech in tweets’ content. To causally identify the impact of Bolsonaro’s election on hate speech through Twitter, we follow a difference-in-differences approach, using the election result as an information shock. We estimate two difference-in-differences models, the traditional and another with a continuous treatment variable. In the latter, the election result at each municipality measures the local incidence of this information shock. Our findings reveal that online hate speech experienced a surge following the elections, particularly in municipalities where Bolsonaro’s popularity was relatively low. These results are further supported by individual-level regressions, which show that both extensive and intensive margins of individual hate speech contributed to the overall increase. We interpret these findings within the framework of belief updating, specifically emphasizing the process of revising social norms that govern what is acceptable to say (or not) in public.

Silence in social networks

Status: draft in progress.

In short: How do social interactions affect what we publicly say and what we do not? I study a model of social norms, assuming interactions are structured through a network. Individuals choose whether to conform to a social norm – have a speech – or not – stay silent. Social norms may be controversial, and their compliance is observable. Then, individuals may remain silent (i) if no norm aligns with their preferences or (ii) if they do not want to conform to a social norm different from their friends, i.e., by social pressure. I investigate how the network structure and distribution of preferences regarding social norms affect the equilibrium outcome and under which circumstances the social pressure mechanism arises.

The long memory of poverty: the Historical Unsatisfied Basic Needs in Argentina

with E. Nicolini.
Pre-doctoral work.