Abstract:
The global Vector autoregressive (GVAR) approach directly models the interlinkages using trade- weighted observable macroeconomic aggregates and financial variables and allows for interdependencies at a variety of levels in a transparent manner that can be empirically evaluated. However, with a modeling task of this size, it would be surprising if a single model were universally preferred over any other. Recognizing that a broader set of models might be needed to tackle the problem, we apply Bayesian Model Averaging (BMA) with a Minnesota prior to appropriate GVAR models-for five East African countries on a quarterly basis from the year 2000 to the year 2013-in order to arrive at better overall forecasts. The results obtained indicate that the GVAR models are dynamically stable. The results also indicate that the GVAR methodology is valid since the weak exogeneity is rejected for most of the foreign variables and that the trade weights are „granular‟. The study also uses the AIC, SBC and log-likelihood selection criterion in order to select the lag orders of the domestic and foreign variables respectively. Our results also show that, taking international linkages into account improves forecasts in inflation and exchange rates while for interest rates forecasts of the univariate benchmark models remain difficult to beat. Moreover, there is a notable outcome whereby the standard GVAR model employed in our literature does not perform well than the simple models.
