Efficient importance sampling in applied econometrics

This thesis focusses on econometric applications requiring multivariate numerical integration. Models that attempt to capture real world complexities are typically nonlinear and display many unobservable factors. These characteristics imply that the likelihood function of these models contain high-dimensional integrals that often cannot be solved analytically, and thus have to be approximated numerically. Importance Sampling is a Monte Carlo simulation method often used to solve high-dimensional integration. In the present work, the Efficient Importance Sampling method developed by Richard and Zhang (2007) was used to overcome the problem of finding good multivariate importance samplers for the integrand of likelihood functions from different models. It was shown how importance sampling can be used to efficiently solve high dimensional integration problems in econometric problems involving panel data, and time series.