A BVAR projection model for Peruvian inflation

This article investigates the effectiveness of a Bayesian Vector Autoregressive (BVAR) forecast model for inflation in Peru, in the context of monetary policy under the inflation targeting regime implemented by the Central Reserve Bank of Peru since 2002. The study focuses on the criticism that current forecast models are not very replicable. To address this issue, BVAR models incorporating Litterman-style assumptions and different variance specifications are estimated. Using data from 1994 to 2004, the analysis shows that BVAR models—especially the simplest one, which includes six key variables—significantly outperform traditional statistical models, such as the random walk, in predicting inflation and GDP growth. The article concludes that the proposed Bayesian methodology is robust and can be applied to other economies with inflation-targeting frameworks, suggesting improvements in future forecasts through the incorporation of confidence intervals. (Abstract and audio: Department of Economic Publications)