Title: A new filter for long memory time series
Authors: Adriana Cornea-Madeira - University of York (United Kingdom) [presenting]
Joao Madeira - University of York (United Kingdom)
Abstract: Macroeconomic and financial time series may have long memory (that is, are integrated of order larger than zero but smaller than one). In such series deviations from the long-run mean decline slower than exponential decay. We study how the properties of time series which display long memory are affected by the application of filters (such as the Hodrick-Prescott and Butterworth) which extract cyclical and trend components. Without relying on any model assumptions, we then propose a new filter designed to take into account for the possible presence of long memory and apply it to unemployment, current account and price-dividend ratio time series.