Confidence intervals of forecasts from nonlinear econometric models

Several methods have been proposed in the last few years for evaluating uncertainty in forecasts produced by nonlinear econometric models. Some methods resort to Monte Carlo, while others resort to different simulation techniques. This work aims at comparing these methods by means of experiments on some econometric models of small, medium and large size, used in practice for forecasting purposes. In most cases of practical interest, direct simulation of confidence intervals allows to overcome the difficulties connected with the nonexistence of finite second order moments, often encountered by the authors when applying Monte Carlo methods to real world models.