Identification and estimation of dynamic games with switching costs

• Grant number: 15/13336-2
• Support Opportunities: Scholarships abroad - Research
• Start date: July 01, 2015
• End date: December 31, 2015
• Field of knowledge: Applied Social Sciences - Economics - Quantitative Methods Applied to Economics
• Principal Investigator: Fábio Adriano Miessi Sanches
• Grantee: Fábio Adriano Miessi Sanches
• Host Investigator: Sorawoot Srisuma
• Host Institution: Faculdade de Economia, Administração e Contabilidade (FEA). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Institution abroad: University of Surrey, England

Abstract

This project develops estimators to reduce the computational burden in the estimation of dynamic games. In these models the state space grows exponentially when players, actions and other state variables are included in the model. This implies that the estimation of a typical dynamic game involves the inversion of large matrices. This task can be either very time consuming or unfeasible. A two stage identification procedure is developed. In the first stage the dynamic components (switching costs) of the model are identified. In the second stage the static components are identified. The identification is constructive. It directly provides expressions to estimate the parameters of interest. It is shown that this procedure halves the size of the matrices one has to invert in order to estimate the parameters of the model. This can reduce severally the computational burden and make it possible the estimation of models that cannot be estimated based on traditional estimation techniques. The estimators have closed-form. No numerical method is needed. (AU)