Does CPI disaggregation improve inflation forecast accuracy?

If the consumer price index (CPI), one of the main indicators of inflation, consists of several components, would it be more accurate to forecast them separately? International experience shows that the aggregate of individual forecasts is often more accurate than the forecast of the aggregated index. In this paper, we explore this issue for Russia and test whether the quality of inflation forecasts can be improved by the CPI individual components forecasting. Using the panel data of Russian regions for the period from 2010 to 2021 we partially confirm the usefulness of a disaggregated approach. Individual modelling of the short-term price dynamics of individual commodity groups is ahead in terms of accuracy of the overall inflation model, including standard benchmark models, but only under certain conditions. First, it is necessary to include the factors of trend inflation in the models, which helps to separate the trend inflation acceleration/deceleration from short-term idiosyncratic fluctuations. Secondly, the models should have the property of inflation convergence to its long-term level, determined by the Bank of Russia's goal. Under these conditions, the disaggregated approach gives a more accurate forecast on short horizons than the aggregated one and a forecast of comparable to non-structural models’ accuracy on longer ones. Additionally, good predictive properties of the “anchored” forecast model were established (the “anchored” forecast is equal to the target inflation rate). The accuracy of this forecast turns out to be higher than the accuracy of standard models and does not deteriorate with an increase in the forecast horizon. This allows us to recommend this model as a simple non-structural benchmark for measuring the quality of inflation forecast models in Russia.