In order to avoid an overfitting of the model at the identification stage, which will yield a strong
in-sample explanatory performance at the same time taking the risk of a poor out-of sample
performance, the most parsimonious model will be sought after.