Parameter estimation of nonlinear econometric models using particle swarm optimization

摘要

Global optimization is an essential component of econometric modeling. Optimization in econometrics is oftendifficult due to irregular cost functions characterized by multiple local optima. The goal of this paper is to applya relatively new stochastic global technique, particle swarm optimization, to the well-known but difficult disequilibriumproblem. Because of its co-operative nature and balance of local and global search, particle swarm issuccessful in optimizing the disequilibrium maximum likelihood function, providing better values than thosereported in the literature obtained using other stochastic techniques. These encouraging results suggest thatparticle swarm optimization may be successfully applied to difficult econometrics problems, possibly in conjunctionwith existing methods.