Economics Stack Exchange is a question and answer site for those who study, teach, research and apply economics and econometrics. It only takes a minute to sign up.
Sign up to join this community
I've studied a number of cointegration methods in the past few months. It seems, Johansen's multivariate VAR approach to cointegration is the most commonly used multivariate cointegration technique. Most packages appear to have in-built functions to run "Johansen cointegration analysis".
I was wondering if there are other multivariate cointegration methods in the literature though less known and are not available in let's say Eviews, Stata, RATs etc. but can be programmed.
I think the main alternative to the Johansen (statistical) approach is the methodology propounded by Pesaran and Shin in their so-called long-run structural modelling (economic) approach. The main formal reference is Pesaran and Shin (2002). The methodology is also presented to a wider audience in Garratt et al. (2012).
Although referred to as the long-run structural modelling approach, you will also read about the type of models associated with the methodology, which are called VARX* models. The distinguishing feature is that the cointegrated models are estimated using reduced-rank regression and (over-)identifying restrictions are derived from economic a priori then tested. The sequencing is slightly different to Johansen since economic theory is given priority.
In the same school of thought is the cointegrated VAR approach associated with Juselius (2006). Again, Juselius propounds an economic approach to cointegration as opposed to the statistical approach by Johansen (and others like Phillips).
Pesaran's approach has been programmed and is available in the Microfit software.
Juselius' approach has also been programmed and is available in the RATS (CATS) software.
To my understanding, you can find MATLAB code in the GVAR toolbox, and this should give an indication of what's required for the Pesaran approach. In terms of programming, I haven't seen much difference when compared to Johansen (although I just quickly investigated this). Note that the GVAR approach can be thought of as an extension to the VARX approach (the difference lies in stacking individual country models, but estimation of the individual models is the same).
References
