Parametric and nonparametric Monte Carlo estimates of standard errors of forecasts in econometric models

In the econometric literature simulation techniques are suggested for estimating standard errors of forecasts, especially in case of nonlinear models, where explicit analytic formulae are not available. For this purpose analytic simulation on coefficients, Monte Carlo on coefficients, Monte Carlo simulation based on parametric estimate of the underlying error distribution have been proposed, and more recently a nonparametric procedure which uses the bootstrap technique is also suggested. Main purpose of this paper is to compare, in empirical applications for real world models, parametric and nonparametrlc estimates. Furthermore, in case of linear models, the same comparisons are performed with respect to the results obtained via analytic formulae. Additional results are obtained from an error-in-variables approach.