Bayesianisches vektorautoregressives Verfahren zur Prognose der Schweizer Wirtschaft

This book employs a method for forecasting real GDP and inflation growth in Switzerland. In this study introduced by Litterman, forecasting models for the Swiss economy are developed. Initially, autoregressive distributed lag models (ARDL) are calculated, followed by the framework of Bayesian models. Bayesian vector autoregressive models (BVARs) heavily rely on the VAR framework but allow for better utilization of all available information. Using data from 1980, out-of-sample forecasts were calculated from 2000 to 2014. This study proposes four categories into which variables can be classified and notes that Bayesian VAR models improve forecast errors, particularly for inflation. An extension of the model is conducted using foreign data, further reducing forecast errors. It was found that asset prices contain valuable information for forecasting real GDP and especially for predicting inflation growth. However, BVARs cannot replace a complete structural method for analyzing economic policy.