Abstract:
Since the adoption of inflation targeting, the seasonal appears to be the component that explains the major part of inflation's total variation in Mexico. In this context, we study the performance of seasonal time series models to forecast short-run inflation. Using multi-horizon evaluation techniques, we examine the real-time forecasting performance of four well-known seasonal models using data on 16 indices of the Mexican Consumer Price Index (CPI), including headline and core inflation. These models consider both, deterministic and stochastic seasonality. After selecting the best forecasting model for each index, we apply and compare two methods that aggregate hierarchical time series, the bottom-up method and an optimal combination approach. The best forecasts are able to compete with those taken from surveys of experts.