Investigating heteroscedastic disturbances in some transformed economic models while eliminating multicollinearity

Abstract

When economic variables move together over time, it may cause increase/decrease in economic development. Trend factors in time series are the strongest sources of multicollinearity. The recent recession in Nigeria suggest multicollinearity among some Nigeria economic variables. This research examined the presence of multicollinearity among some economic variables using Revised Frisch’s Confluence (or Bunch –map) Test and Variance Inflation Factor in some transformed models namely; Semi-logarithmic (model B), Inverse Semi-logarithmic (model C), Double logarithmic (model D) and basic linear model (model A). These models were compared to identify the best fitted using AIC as a comparison criterion. Furthermore re-parameterization of the models was done to eliminate multi-collinear variables and also show the coefficients effect of each model. In addition, the presence heteroscedastic disturbance was seen to be present in model A and model B using Golfeld-Quandt test. A parsimonious model Yi =f(X3, X4) of the inverse semi logarithm model C was found to be the best fitted model. We also observed that heteroscedastic models (model A and B) had very high AIC’s in comparison to the homoscedastic models (model C and D).