Calendrier du 11 octobre 2021
Roy Seminar (ADRES)
Du 11/10/2021 de 17:00 à 18:15
Salle R2-21 - Campus Jourdan - 75014 PARIS
ELY Jeff (Northwestern)
Ruth, Anthony, and Clarence
GSIELM (Graduate Students International Economics and Labor Market) Lunch Seminar
Du 11/10/2021 de 13:00 à 14:00
MSE(106, Blv de l'Hôpital 75013 Paris), salle 116
MAZET Clément (Banque de France, Sciences Po and Collège de France)
Search Frictions in Credit Markets
Motivated by empirical evidence I document on local credit markets in France using data on more than 3.5 million bank-firm relationships, I propose a theory of bank-firm matching subject to search frictions. Firms undergo a costly search process to locate and match with the right banking partner. Upon matching, agency frictions hinder banks ability to optimally screen and monitor projects. I structurally estimate my model on French data using the staggered roll-out of Broadband Internet, from 1997 to 2007, as a shock to search frictions. I confirm the model predictions that a reduction in search frictions affects the allocation of credit and the dynamics of firm-bank matching. Finally, I use the structure of my model to quantify the impact of this technology-induced reduction in search frictions on loan prices. I find that broadband internet access reduced the cost of debt for small businesses by 4.9%.
Régulation et Environnement
Du 11/10/2021 de 12:00 à 13:15
Salle R2-21 - Campus Jourdan - 75014 PARIS
ZABROCKI Leo (PSE)
Inference Design In Studies on Acute Health Effects of Air Pollution
écrit avec BAGILET Vincent
We explore statistical power characteristics of various empirical strategies implemented to estimate the short-term health effect of air pollution. Through an extensive literature review, we retrieve the estimates and standard errors of a large number of studies published on this topic. We find that a non-negligible share of studies may suffer from low power issues and could thereby exaggerate effect sizes. The analysis of published results highlights potential shortcomings of the literature but does not enable to precisely identify drivers of theses is sues. We therefore run realistic simulations to investigate how statistical power varies with the treatment effect size, the number of observations, the proportion of treated units as well as the distribution of the outcome. Usual causal identi fication methods implemented in this literature, such as instrumental variable (IV), regression discontinuity design (RD) or difference-in-differences (DiD), may yield overestimated effect sizes. This issue is driven by the imprecision of the IV estimator and the small number of exogenous shocks usually exploited in DiD and RD designs. When focusing on particular groups such as the elderly or children, researchers should be aware that statistical power is lowered by the limited average count of health outcomes