Lucas GIRARD

Ph.D. Candidate in Economics

  • Address: 5 avenue Henry Le Chatelier, Palaiseau, 91120, France
  • Office: 4097
  • Phone: +33 (0)1 70 26 69 43
  • E-mail: lucas.girard[at]ensae.fr
  • Personal Website: https://faculty.crest.fr/lgirard/

Biography

I am a Ph.D. candidate at CREST (Center for Research in Economics and Statistics) under the supervision of Xavier D’Haultfoeuille.

I am interested in econometric theory and its application. My current research is devoted to two main themes.

First, I wonder how to construct reliable measures of polarization, namely indices, that quantify to which extent two exogenous groups tend to make different choices among a set of options. This abstract problem encompasses various practical settings: residential or school segregation between, for instance, natives and immigrants, speech polarization between Democrats and Republicans, buyers’ choices for different goods, etc. I investigate situations of “high dimensional choices”: when the number of observed choices in the data is low relative to the total number of distinct options. In such cases, natural measures of polarization suffer from a “small-unit bias.” Consequently, other methods need to be developed to yield measures that can reliably be compared over time or across different settings. I currently have two projects in this direction, joint works with Xavier D’Haultfoeuille and Roland Rathelot: (i) the Stata package “segregsmall” implements three methods to estimate segregation indices in the context of small units; (ii) we propose an alternative approach to Gentzkow, Shapiro, and Taddy (Econometrica 2019) to study speech polarization in the US Congress between Republican and Democrat speakers (1873-2016).

My second research theme deals with constructing nonasymptotic confidence intervals (CIs), i.e., whose probability of containing the target parameter is larger or equal to their nominal level for any sample size. It happens that the prevalent way to conduct inference in applied economics stems from asymptotic results. As a case in point, the CIs for individual coefficients in linear regressions rely on the asymptotic normality of the t-statistic. Their properties are, therefore, only asymptotic. The validity of inference for any real finite sample involves an asymptotic approximation, which might be questioned in some settings. Hence the potential interest of nonasymptotic CIs. With Alexis Derumigny and Yannick Guyonvarch, we first studied such CIs for ratios of expectations. Retrospectively, that project appears as a training exercise for a more ambitious and ongoing work about nonasymptotic CIs for individual coefficients of linear regression models.

Finally, I like teaching and think it is a vital complement to research and how to communicate and convey it. I have taught as a teacher assistant for econometrics and statistics courses at ENSAE.

Research Interests

  • Econometrics Econometric Theory

  • Economics Applied Microeconometrics

  • Statistics Natural Language Processing