Common Factors of CPI Sub-aggregates and Forecast of Inflation

In this paper, we investigate whether incorporating common factors of CPI sub-aggregates into forecasting models increases the accuracy of forecasts of inflation. We extract factors by both static and dynamic factor models and then embed them in ARMA and VAR models. Using quarterly data of Iran’s CPI and its sub-aggregates, the models are estimated over 1990:2 to 2008:2 and out of sample forecasts are produced for 2008:3 to 2012:1. The results show that in most cases the performance of the models containing common factors of CPI sub-aggregates is better than the Autoregressive, as one of the benchmark models. But, only for the horizon of two-step ahead, the performance of the factor models are significantly better than that of benchmark. Also, the FAVAR performs better than the other factor models in forecasting inflation.