On Specification of Almon Distribued Lag Model in the Presence of Multicollinearity

In Almon distributed lag modeling, presence of multicollinearity, choice of appropriate approximating polynomial of suitable degree and choice of optimal truncation lag-length required for the implementation of Almon distributed lag model constitute three major problems confronting Almon distributed lag modeling.Given the fact that the choice of appropriate approximating polynomial of suitable degree and optimal truncation lag-length can significantly reduce problem of multicollinearity in Almon model, we employ variance inflation factor as an alternative but credible criterion to choose an appropriate approximating polynomial of suitable degree as well as circumventing the problem of underfitting and overfitting of truncation lag-length associated with standard information criteria.For empirical illustration, several Almon models were fitted to Nigerian inflation data against Naira-based exchange rate data. From the results of data analysis,we showed that approximating polynomial of order 2 is the most appropriate and the lag-length of Almon model with the lowest variance in the inflation factor was selected as the optimal truncation lag-length that guarantee minimal multicollinearity.