E12 - General Aggregative Models: Keynes; Keynesian; Post-KeynesianReturn

The paper deals with forecasting the ability of the most common macroeconomic methods. The main goal is to predict the percentage GDP growth while using many methods and at the end assess the performance of these methods. The performance is measured by the Root Mean Square Error statistics. Methods used in this paper are: naive Auto Regression with one lag, Vector Auto Regression with two lags, Bayesian Vector Auto Regression with two lags, Dynamic Stochastic General Equilibrium (DSGE) and a DSGE-VAR model which obtains priors from a DSGE model and then estimates it like Vector Auto Regression. The next contribution of this paper is specification of a New Keynesian DSGE model and coding it in Matlab. Final results summarize the performance of each method. On the other hand, structural models as DSGE and DSGE-VAR perform better than a benchmark. In case of a DSGE model it is mainly caused by its structural nature. This model also contains forward looking variables which take into account the behavior of households and firms which are basic cornerstones of a NK DSGE model. The DSGE-VAR model performs better than a benchmark due to fact that priors are taken from the DSGE model. It means that structural information can be transferred using these priors. Nevertheless, according to the RMSE statistics the best performing method is the DSGE model.