Handbook of Economic Forecasting

4.1.2 Do Traditional Macroeconomic Time Series Beat an Autoregressive Benchmark Model in Out-of-Sample Forecast Comparisons Tests?

We next consider the predictive ability of the same macroeconomic variables for forecasting inflation and output growth out-of-sample. The benchmark is the autoregressive model, and the forecast horizon is four quarters. Results are broadly similar for the random walk without drift benchmark and for other forecast horizons. We consider both traditional out-of-sample forecast comparison tests as well as Giacomini and Rossi’s (2010a) forecast comparisons tests robust to instabilities.⁵⁷

Tables 2 and 3 report results of traditional out-of-sample forecast comparison tests. The first line in Table 2 reports the RMSFE ...