Misconceptions in Application of Time series Models: The Abuse of ARDL Model

Studies in applied econometrics and related disciplines are widely using time series techniques. Sound application of these techniques requires the satisfaction of their underlying assumptions. One of these techniques is ARDL Model. There are a number of examples in the published works in Iranian economic studies which suffers from misconceptions leading to biased and inconsistent estimates of parameters. For example, without assuring the exogeniety of the other RHS variables of the model, estimates are biased and inconsistent. Although fitting an ARDL model to exogenous and/or predetermined RHS I(0) and/or I(1) variables may seem appropriate, however, to continue the path leading to the extraction cointegration vectors, would be a misleading strategy. The aim of this paper is to highlight these misconceptions in the application of ARDL technique. To this end, after a theoretical survey of the concept, a dynamic simultaneous equation model (DSEM) of macroeconomic equations is introduced and simulated. The simulated series are used to estimate various versions of the model including VAR, ARDL, DSEM, VECM. Then the correct and incorrect application of ARDL model is specified by using two scenarios.