Autore
Kostial, Kristina

Titolo
The fully modified OLS estimator as a system estimator: a Monte-Carlo analysis
Periodico
European University Institute of Badia Fiesolana (Fi). Department of Economics - Working papers
Anno: 1995 - Fascicolo: 8 - Pagina iniziale: 1 - Pagina finale: 50

This paper provides an extensive Monte-Carlo study on the FM - OLS estimator. The first part analyzes the FM OLS estimator under alternatives for constructing the long-run variance-covariance matrix. Taking into account six different measures, the prewhitened FM-OLS estimator employing the Bartlett kernel with an automatic choice of the bandwidth parameter proves to be the best. Section two focuses on the sensitivity of the estimator to non-Gaussian error processes; especially autoregressive e error processes with roots close to unity can constitute a serious problem. The following sections are of particular relevance for applied work. Section three illustrates that the higher the rank of the cointegrating space, the more precise will be the estimates with the FM -OLS estimator. The next section highlights the possibility of exploiting permutations of the data set to determine the rank of the cointegrating space. Finally, the results of the Monte-Carlo study are applied to estimating a data set with the FM-OLS estimator.



Testo completo: http://hdl.handle.net/1814/533

Esportazione dati in Refworks (solo per utenti abilitati)