This paper analyses the frequency domain properties of two well-known measures of core inflation: the trimmed mean estimator and the SVAR estimator. It also investigates whether a small modification of the trimmed mean estimator enhances its capacity of filtering high‑frequency noise. We find that the two versions of the trimmed estimator are rather similar. They work as imperfect approximations for low pass filters. Therefore, they are capturing very well trend inflation. The SVAR estimator, however, is quite different from both of them. It emphasizes intermediate frequencies rather than low frequencies, indicating that cyclical movements associated with excess demand pressures are very important in the medium run.
core inflation; frequency domain; trimmed estimator; SVAR