Forecasting Some Selected Macroeconomic Variables with BVAR Models under Natural Conjugate Prior.

Abstract

Bayesian VARs are mostly used in computational analysis of macroeconomic variables of a nation. The natural conjugate prior when combined with the likelihood function gives a posterior distribution that belong to the same distributional family. The study investigates the forecasting of some selected macroeconomic variables with BVAR
Model using special types of natural conjugate prior called the symmetric and asymmetric natural conjugate prior. The data for the macroeconomic variables were obtained from the statistical bulletin of the Central Bank of Nigeria (CBN) ranging from 1986 to 2019. The forecasting assessment used is the Root Mean Square Factor Error (RMSFE).
The RMSFE with small value indicates a better forecast performance. Forecasting the Macroeconomic variables with BVAR under the natural conjugate prior (symmetric and asymmetric natural conjugate prior), it was discovered from the study that asymmetric natural conjugate Prior is the best natural conjugate prior that should be used in forecasting macroeconomic variables of developing country like Nigeria. There is an inverse relationship between unemployment rate and other selected macroeconomic variables used in this study. Therefore, the policy makers should endeavor to formulate policies that will reinvigorate the economy so that single–digit inflation rate can be achieved and unemployment rate will be reduced to the barest minimum. This will help to boost the GDP
and economy at large.