Forecasting Peruvian Macroeconomic Variables with Bayesian Vector Autorregressions with Time-Varying in the mean.

Macroeconomic Forecasting in a changing and uncertain environment over time is a great challenge today. This paper uses a Bayesian VAR with a time-varying mean and stochastic volatility, in order to elaborate forecasts for the Peruvian economy. The model is flexible enough to consider the structural changes that potentially occur in the economy. Forecasts are made mainly for variables such as inflation and GDP growth, although the model might be adapted to include other variables. The empirical model uses information from the survey of macroeconomic expectations as observables linked to the long-term means, following Banbura and van Vlodrop (2018). Results show a good fit, and reaffirm the idea associated with the use of expectations surveys to reduce long-term uncertainty, while the time varying parameters improve the predictive power of the model.