Modelli con correzione dell'errore e stima dei moltiplicatori di lungo periodo

Abstract

The present note considers recent developments on the use of various transformations of a general autoregressive-distributed lag (AD) model for estimation of long run multipliers. We observe that Bardsen (1989) and Banerjee, Galbraith and Dolado (1990) ECM transformations do not provide an efficient research strategy since they do not provide the near-orthogonality of regressors and a direct test of long-run proportionality or homogeneity hypothesis. Our main point here is that there is no need in searching for other ECM generalizations since the Hendry, Pagan and Sargan (1984) ECM transformation provides a real efficient research strategy; this form does not impose any restriction to the AD model, but it provides a direct test of the homogeneity hypothesis and the near-orthogonality of regressors. Moreover, we show that this form is particularly convenient for estimation of long-run multipliers and of their variances. Finally we suggest a way to overcome Wickens and Breusch (1988) criticism to the ECM transformation.