About
I am a Research Fellow at the University of Nottingham and a PhD Candidate in Economics at Trinity College Dublin. I am currently visiting Harvard Government hosted by Professor James Snyder.
My research interests lie in empirical Political Economy with a specific focus on Media Economics. I employ advanced text analysis techniques, including BERT-based models and large language models (LLMs).
You can find my CV here. Email: matteopograxha@gmail.com
Research
Working Papers
-
Online Virality, Broadcast Visibility, and the Incentives for Extreme Rhetoric
Show abstract Hide abstract
I study whether social media virality increases politicians' access to television. Using 700,000 tweets by Italian Members of Parliament (2018-2022) linked to daily television appearance records, I show that MPs who go viral receive substantially more television invitations the following day. An event study design reveals that the effect is immediate, short-lived, and concentrated on private channels. I then show that extreme rhetoric, measured as institutional delegitimization, group-based discrimination, or zero-sum framing, boosts online engagement and increases the likelihood of virality. Extreme rhetoric is asymmetrically distributed: extreme right MPs are far more likely to use these strategies. A supply-side model shows that politicians rationally exaggerate rhetoric to generate visibility signals that influence television booking decisions, trading reputational costs for media access even when citizens discount their credibility.
-
Man Bites Dog: Editorial Choices and Biases in the Reporting of Weather Events
— with
Nicola Mastrorocco,
Arianna Ornaghi, and
Stephane Wolton (R&R, Journal of Public Economics)
Show abstract Hide abstract
Every day, editors of media outlets decide what is news and what is not. We unpack the process of news production by looking at the share of newscasts devoted to weather events by local TV stations in the United States. Coverage increases with the severity of the weather event that day. Stations in Democratic leaning markets devote more time to extreme weather and mention climate change more than outlets in Republican leaning markets. We interpret these publication and presentation patterns with a stylised model of news production and consumption.
Work in Progress
- (Do not) Keep Calm and Carry On: TV and Heated Debates
— with
Nicola Fontana
Show abstract Hide abstract
Does confrontation on television attract more viewers? This paper studies whether heated debates draw larger audiences than respectful discussions, and whether spectacle driven formats impair citizen learning. We develop a framework using machine learning to extract debate intensity, namely interruptions, overlapping speech, and volume, from audio of Italian political talk shows, linking these features to minute by minute audience data. We implement a dynamic difference in differences design that exploits within episode variation in debate intensity to measure viewership responses. We plan to run a randomized experiment to test whether spectacle captures attention while reducing information retention.
- Uncovering Disagreement in Central Bank Communication: Social Media and Monetary Policy Surprises
— with
Donato Masciandaro,
Conor Parle, and
Davide Romelli
Show abstract Hide abstract
Central bank announcements generate vast social media discussions that reveal how markets interpret monetary policy. We assemble a dataset of over 125 million tweets surrounding ECB, Fed, and BoE policy decisions. Using AI based classification, we distinguish relevant tweets and categorize them as hawkish or dovish, constructing a real time measure of disagreement about future policy. Higher social media disagreement is strongly associated with larger financial market reactions following announcements, highlighting social media's value for understanding how central bank communication shapes expectations.
- Scoring Goals, Spreading Hate: Drivers of Hate Crime in Germany - with Francesco Barilari, and Enrico Cavallotti
Research & Training
- Oct 2024 – Sept 2025 UN Research Consultant — Text Analytics & Legal Digitization
Led large-scale digitization and variable extraction of South African Government Gazettes using advanced layout detection and Named Entity Recognition (NER).
- Sept 2025 Guest Instructor — National Treasury (Pretoria), South Africa
Delivered a two-day seminar on text analysis for policy and research teams, covering NLP pipelines, document layout parsing, and practical use cases with legal and administrative texts.
Teaching
- Spring 2022 Introduction to Macroeconomics - Undergraduate
- Fall 2022 Introduction to Microeconomics - Undergraduate
- Spring 2024 Statistics - Undergraduate