Abstract
This paper evaluates a wide range of models for forecasting euro area services inflation, with a particular emphasis on the potential benefits of disaggregated approaches compared to aggregate-none models. It assesses the forecasting performance of state-of-the-art Vector Autoregressive (VAR) models, which incorporate features such as time-varying variances and mixed-frequency data handling. A comprehensive real-time forecasting exercise is conducted, using the cut-off dates of ECB projection exercises between September 2016-September 2025. The results indicate that disaggregating services inflation into categories such as travel-related services, rents, and other services yields better forecasting performance compared to alternative disaggregation methods. Additionally, models for total services inflation prove competitive, particularly for medium-term forecasts. Overall, the findings highlight that incorporating survey data improves the accuracy of services inflation forecasts.