The D.Currie's test on the long run dynamic properties of the autoregressive-distributed lag models

Abstract

A major criticism about the autoregressive-distributed lag (AD) models is that the long run dynamic properties of such models are often not sensible and well-determined. These dynamic properties pose particular problems both for practical forecasting and simulation and because economic theory not always leads us to expect the presence of long run dynamic effects. In a recent paper D. Currie (1981) suggested a particular test which allows to examine the significance of these long run dynamic effects and to constrains such effects to zero where appropriate. The main purpose of this paper is to examine this test procedure and to show that it can easily produce dynamic misspecification. Analysing a simple first degree dynamic model, we are able to prove that the suggested restrictions to test are the same restrictions we test in the COMFAC analysis. We also show the relationship between these two tests in the 'general-to-specific' dynamic specification strategy.