infonet-economy

This thesis compiles three papers on different topics in computational economics: First, we present a new method to recursively integrate expectations over serially correlated latent variables with continuous support in maximum likelihood estimations, using highly efficient quadrature rules and interpolation; we apply the method to the dynamic discrete choice model of Rust (1987). Second, we present a method to use the constrained optimization approach to the estimation of dynamic models (MPEC; Su and Judd, 2012) in conjunction with grid adaption for state variables with continuous support; we use grid adaption by node movement, and derive sufficient optimality conditions for the underlying function approximation from the equioscillation theorem, which enables us to solve the estimation and the approximation problem simultaneously. Third, we present a simulation study using agent-based modelling to investigate whether a model of monopolistic competition can reach its equilibrium without common knowledge of aggregate demand and sophisticated utility optimization capabilities of the agents.