Czy dezagregacja indeksu cen poprawia prognozy polskiej inflacji?
This paper examines whether forecasting CPI components improves CPI forecast. We exploit quarterly data for Poland, disaggregated into 12 components. We follow methodology used in previous studies for Euro Area (Hubrich, 2005; Reijer and Vlaar, 2006). AR, MA, TAR and unrestricted VAR models are estimated using recursive sample and aggregated into CPI. Using out-of-sample forecasts, these models are evaluated and compared to the benchmark -- equivalents for aggregate CPI. The evidence is mixed. VAR component-forecast outperform benchmark. Contrary to VAR, for AR and TAR models we do not find substantial gain from using disaggregated data. Results for MA models are not robust. Moreover, it seems that results for AR- and VAR-based forecasts are comparable to consensus forecast.
More...