How can decisions taken by the researcher during the estimation process affect the bias of the synthetic control estimator? : an investigation by Monte Carlo simulations

• Grant number: 14/23731-3
• Support Opportunities: Scholarships in Brazil - Master
• Start date: February 01, 2015
• End date: January 31, 2016
• Field of knowledge: Applied Social Sciences - Economics - Quantitative Methods Applied to Economics
• Agreement: Coordination of Improvement of Higher Education Personnel (CAPES)
• Principal Investigator: Cristine Campos de Xavier Pinto
• Grantee: Vítor Augusto Possebom
• Host Institution: Escola de Economia de São Paulo (EESP). Fundação Getúlio Vargas (FGV). São Paulo , SP, Brazil

Abstract

The master's thesis proposed in this research project aims to investigate how decisions taken by the researcher during the estimation process can affect the bias of the synthetic control estimator. In order to achieve this goal, we will run Monte Carlo simulations to evaluate, for instance, the impact of adopting the cross-validation method or including more predictors or comparison units on the results estimated by the synthetic control technique. Using the knowledge discovered with these simulations, we aim to elaborate a recommendation list that will guide two empirical exercises: (i) reestimate the effect of terrorism on the Basque Country's economy and compare it with the results found in \citeonline{Abadie2003} in order to evaluate how sensitive the synthetic control estimator is to decision taken by the researcher in an applied context and (ii) estimate the effect of the creation of the Free Zone of Manaus on the economy of this state capital, since the existing data about Brazilian municipalities is rich in its cross-sectional and temporal dimensions, allowing us to test the sensitivity of the synthetic control estimator to several different specifications. Furthermore, we aim to give a theoretical contribution to the synthetic control literature: we propose a small modification in one of the minimization problems solved by the synthetic control estimator, giving larger weights to prediction errors made in periods closer to the beginning of the intervention whose impact the researcher wants to evaluate. (AU)

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

FIRPO, SERGIO; POSSEBOM, VITOR. Synthetic Control Method: Inference, Sensitivity Analysis and Confidence Sets. JOURNAL OF CAUSAL INFERENCE, v. 6, n. 2, SEP 2018. (14/23731-3)