Research & Old Projects
Research & Old Projects
Work In Progress (comments more than welcome!):
Do School Bureaucrats Discriminate against Foreigners? Experimental Evidence from ItalyÂ
I study whether public Italian schools discriminate against foreigners. I conduct a randomized correspondence experiment in which each school is contacted with an identical email signed by an Italian, French, or Arabic alias. Schools are 5.3% less likely to reply to foreigners, and when they do respond, answers are 17.6% less likely to be informative and 4.8% less likely to be polite. Discrimination is stronger in lower-income municipalities, with nearly 10% of schools exhibiting biases above 11%, and varies substantially across regions. These findings are inconsistent with statistical discrimination and contact theory. I contribute by (i) documenting discrimination within the school system itself, (ii) showing that widely cited mechanisms fail to explain it, and (iii) highlighting the local nature of bureaucratic bias. These results suggest that anti-discrimination efforts may be most effective when targeted at specific regions, as nationwide interventions risk being diluted. Slides. Code.
Old Projects:
Germany Unplugged: Forecasting Grid Congestion to Reduce Welfare Loss (with Neri F., Opocher G., Tozzi S.)
The paper is the result of our work at the 2025 Econometric Game, as members of the University of Bologna team.
Due to the complex dynamics of the electricity market, grid congestion is hard to prevent and therefore produces large economic costs and welfare losses. In this paper, we propose two methodologies to forecast network congestion and rationalize the role of the transmission system operator through the lenses of a toy model. After investigating the principal components that drive redispatches, we run a time series analysis using an INAR(1) model which accounts for the count nature of the data and provides an interpretable structure for forecasting. As an alternative specification, we consider a fully flexible Recurrent Neural Network (RNN) model that allows to account for the unbalancedness of the congestion frequency attaining high forecasting accuracy (0.94). RNN outperforms INAR(1), as it better captures the true negative congestion events in the test data and therefore potentially reduces the costs related to readjusting energy production. Finally, we run a counterfactual analysis in which we artificially introduce a 75% cap on the price of electricity and study the effects on congestion. We find that introducing this policy would significantly increase the frequency of redispatches.