At the annual Economics PhD meeting of Paris-Saclay University, I gave a presentation of my research agenda on exploring Twitter data on the French 2022 legislative elections.

Abstract:

Econometrics is the interplay between economic reasoning and statistical analysis. Systematic differences in estimates given by experimental and non-experimental methods made economists increasingly rely on natural experiments. Econometricians are thus interested under what conditions a treatment effect can be estimated in such a setting. In such efforts, machine learning is used as a tool for providing good proxy variables and improving estimation methods. I will use observational data from the French 2022 legislative election and the Twitter activity of candidates to give new insights on the following question: what topics politicians bring up during elections and how they adapt their discourse based on the rival they are facing? I will attempt to provide a good estimation of a treatment effect (e.g. qualifying for second round) on an outcome (e.g. topics distance between candidate and rival), and explain how machine learning plays a role in providing the key proxy variables: political topics and measures of discourse similarity.

More details about the workshop: