There are a number of econometrics tools to deal with the different type of situations in which cointegration can appear: I(1), I(2), seasonal, polynomial, etc. There are also different kinds of Vector Error Correction models related to these situations. We propose a unified theoretical and practical framework to deal with many of these situations. To this aim: (i) a general class of models is introduced in this paper and (ii) an automatic method to identify models, based on estimating the Smith form of an autoregressive model, is provided. Our simulations suggest the power of the new proposed methodology. An empirical example illustrates the methodology.
time series unit root cointegration error correction model identification Smith form