An International Publisher for Academic and Scientific Journals
Author Login
Scholars Journal of Physics, Mathematics and Statistics | Volume-1 | Issue-01
A Simulation Study of Effects of Collinearity on Forecasting of Bivariate Time Series Data
Adenomon M. O, Oyejola B. A
Published: June 29, 2014 |
122
67
DOI: 10.36347/sjpms
Pages: 4-21
Downloads
Abstract
Vector Autoregression (VAR) and Bayesian VAR (BVAR) were first introduced in the early eighties of the
last century and have since proven to be practical and effective economic forecasting methodologies. This paper tends to
study the forecasting performances of the unrestricted VAR model and four versions of the Sims-Zha BVAR models in
the presence of collinearity using Monte Carlo simulation technique. We considered ten (10) collinearity levels (0.8, -0.8,
0.85, -0.85, 0.9, -0.9, 0.95, -0.95, 0.99 and -0.99), the results from our simulation study revealed that the forecasting
performances of the models vary as the collinearity levels varied. Furthermore, the values of the criteria increases as the
time series length and the collinearity levels increased. We therefore recommend that if VAR modelers know that
collinearity is acting upon the model, one can choose the forecasting model that is preferred for the criteria and the time
series length selected.