Title: Bayesian vector autoregressive models for forecasting inflation rate in Nigeria
Authors: Uchechukwu George - Michael Okpara University of Agriculture (Nigeria)
Joy Nwabueze - Michael Okpara University of Agriculture, Umudike. Abia State, Nigeria (Nigeria) [presenting]
Abstract: Maintaining Price stability is a key function of central banks, there is therefore need for inflation forecasting. The focus is on using multiple time series method for forecasting of inflation rate in Nigerian. The Bayesian approach to the estimation of Vector Autoregressive (VAR) model is applied. This allows combination of prior information and data information. Forecasts of inflation rates in Nigeria are provided by using six different Bayesian VAR priors (Diffuse prior, Minnesota prior, Natural Conjugate prior, Independent Normal Wishart prior, Stochastic Search Variable selection prior-Wishart and Stochastic Search Variable selection prior-VAR). The forecast performance of various models is evaluated using root mean square error (RMSE). The Stochastic search variable selection-Wishart outperforms other methods. The forecast from this model together with the impulse response function is given. It is concluded that the stochastic search variable selection prior-Wishart gives a reliable forecast of Inflation rate in Nigeria and therefore, we recommend it for short term inflation forecasting in Nigeria.