The ability to produce high-quality inflation forecasts is crucial for modern central banks. Inflation forecasts are needed for understanding current and forthcoming inflation trends, evaluating the effectiveness of previous policy actions, making new policy decisions, and building the credibility of a central bank in the eyes of the public. This motivates a constant search for new approaches to producing inflation forecasts. This paper analyses the empirical performance of several alternative inflation forecasting models based on structural vs. data-driven approaches, as well as aggregated vs. disaggregated data. It demonstrates that a combined ARMA model with data-based dummies that uses the disaggregated core inflation data for Ukraine allows to considerably improve the quality of an inflation forecast as compared to the core structural model based on aggregated data.
展开▼
