ABSTRACT
This study examines the long – run stability of an Autoregressive Distributed Lag (ARDL) model with intercept and no constant trend, using quarterly data from CBN annual statistical bulletin that span from 1999Q1 to 2011Q2 and 2000Q1 to 2012Q4. The parameters of the long – run stability of ARDL model are not all statistically significant in their respective p – values and as such cannot be totally relied upon for statistical estimation. Then, we proposed a matrix elimination model which is motivated by the p – values of the respective parameters to systematically eliminate at a time an exogenous variable and its corresponding parameter whose p – value is less significant from the system. Using econometric package (EView), it was investigated that there is collaboration in the long – run stability between the ARDL model and the Binomial coefficient model for different order of n(K). The findings show
that the system (model) now accommodates parameters whose p – values are all statistically significant at 5% level of significance. This now paves way for an extended understanding for decision and policymakers to formulate a mechanism to maintain long – run stability in Foreign Reserves.
KEYWORDS: Binomial coefficients, Econometric View, Diagnostic Stability, Minimum information Criteria, Parsimonious Variables, Over-Parameterization