Stochastic Optimization in Econometric Models – A Comparison of GA, SA and RSG

This paper shows that, in case of an econometric model with a high sensitivity to data, using stochastic optimization algorithms is better than using classical gradient techniques. In addition, we showed that the Repetitive Stochastic Guesstimation (RSG) algorithm –invented by Charemza-is closer to Simulated Annealing (SA) than to Genetic Algorithms (GAs), so we produced hybrids between RSG and SA to study their joint behavior. The evaluation of all algorithms involved was performed on a short form of the Romanian macro model, derived from Dobrescu (1996). The subject of optimization was the model’s solution, as function of the initial values (in the first stage) and of the objective functions (in the second stage). We proved that a priori information help “elitist “ algorithms (like RSG and SA) to obtain best results; on the other hand, when one has equal believe concerning the choice among different objective functions, GA gives a straight answer. Analyzing the average related bias of the model’s solution proved the efficiency of the stochastic optimization methods presented.