Modelling GDP in Nigeria using Bayesian Model Averaging

• O. B. Akanbi
• J. F. Ojo
• M. O. Oluneye

Statistical modelling has played significant roles in various field of endeavors and this has helped policy makers and planners to take decisive decision. This study attempts to model Gross Domestic Product (GDP) in Nigeria using Bayesian Model Averaging (BMA). It sought to know factors that have contributed to the growth of GDP as well as best model for modelling GDP in Nigeria. Essentially, empirical growth research faces a high degree of model uncertainty. Therefore, this study adopts the use of Bayesian Model Averaging in a bid to overcome the model uncertainty in a single model selection process. We obtained estimates of the posterior probabilities via Markov chain Monte Carlo (MCMC) which was further used as weights to model averaged estimates and predictions. Each model was weighted accordingly with a model prior and a parameter prior. Model uncertainty and posterior inclusion probability of each predictor was determined. The result shows that Exchange Rate is the most important variable affecting the GDP of the Nigerian economy followed by Interest Rate. The best model consists of an average of two predictors with Exchange Rate as one of the major contributor. The top five models were checked and it explained 40.16% uncertainty, which reveals the risk of depending on single model prediction. Every predictor, irrespective of its minute contribution provides some information which may be lost if the predictor is not considered

Select Volume / Issue:

• Volume 5,Issue 3, May 2018

Year:

2018

Type of Publication:

Article

Keywords:

Posterior Inclusion Probabilities; Posterior Model Probabilities; Model Uncertainty; Gross Domestic Product

Journal:

IJASM

Volume:

5

Number:

3

Pages:

22-27

Month:

May

ISSN:

2394-2894