Simulation-Based Analysis of Real-Time Reliability for Trend/Cycle Decompositions

When comparing methods for performing trend/cycle decompositions, real-time reliability properties are often highlighted as an important aspect to consider. The measures employed to assess the real-time reliability properties are frequently calculated based on real-world economic time series and give an indication of how real-time estimates of trend and cycle components are revised as more data become available and estimates are updated. This use of real-time data is absolutely essential when assessing and comparing the performance of filters. But since revision properties relying on real-world data are calculated as an average over a time period, properties that vary with time are not necessarily portrayed in enough detail when applying this approach. The current article suggests that using also a simulation-based methodology can shed light on revision properties that vary with time-series sample positions for various trend/cycle decomposition methods. Using a set of commonly applied filters, the relevance of such time variation, and hence the relevance of the simulation-based approach, is illustrated.