Randomization inference in shift share designs with an application in banking
Título da Revista
ISSN da Revista
Título de Volume
This Master Thesis is comprised of two parts. The first one, presented in Chapter 1, is a theoretical advance in econometrics for Shift-share Designs, based on joint work with Luis Alvarez and Bruno Ferman, "Randomization Inference Tests for Shift-Share Designs" (see the reference Alvarez et al., 2022). It shows that, by choosing a properly studentized statistic for performing Randomization Inference, we are able to (i) control size in finite samples under relatively strong hypotheses, such as homogeneous treatment effects and known assignment process, and (ii) control size asymptotically under milder hypotheses, such as a "well-behaved" treatment heterogeneity and randomization distribution, even if it is different from the original assignment process. The second part, in Chapter 2, is an empirical application of this technique to the expansion of the physical bank network in Brazil during the commodity boom of the 2000's and 2010's. It seeks to measure to what extent the increase in the number of branches in the period was a response to greater economic activity. This is of interest because, at the time, the physical network was a vector of financial inclusion. Understanding the extent to which this was spurred by exogenous demand shocks may be informative for future policy.