The Bayesian Model Averaging Method And Its Application In Macroeconomic Forecasting

We consider Bayesian Model Average(BMA) in order to overcome the model uncertainty that traditional macroeconomic forecasting methods ignore. In BMA, the posterior probability of each alternative model is set as a weight and we use the weighted average to forecast or estimate; because BMA considers all possible single models, and set the posterior probability that the model weights as standard to judge the pros and cons of the model, BMA deals effectively with the uncertainty of the model.This paper summarizes the two directions of the development of Bayesian statistics(model selection and combination forecasting) and describes the basic principles of the BMA and guidelines to estimate the individual model weights. Subsequently, we conclude the two main aspects of the BMA applications –the estimate of the importance of the variables based on the posterior probability as well as the forecast using BMA. To explain that, we come up with an example, that is the forecast of CPI. Empirical analysis shows that: BMA is more accurate than the stepwise regression method not only in the sense of the posterior probability, but also in the effect of prediction.