Zusammenfassung:
This paper introduces two methodological improvements to the Hodrick- Prescott (HP) filter for decomposing GDP into trend and cycle components. First, we propose a robust univariate filter that accounts for extreme observations - such as the COVID-19 pandemic - by treating them as additive outliers. Second, we develop a multivariate HP filter that incorporates time-varying, import- adjusted budget shares of GDP sub-components. This adaptive weighting minimizes cyclical variance and yields a more stable trend estimate. Applying the framework to U.S. data, we find that private investment is the dominant source of cyclical fluctuations, while government expenditure exhibits a persistent counter-cyclical pattern. The proposed approach enhances real-time policy analysis by reducing endpoint bias and improving the identification of cyclical dynamics.