Measuring the Cyclical Component of a Time Series: a New Proposed Methodology

Abstract

The objective of this paper is to present a simple technique that estimates the true cycle of a time series. The robustness of the technique was tested under Monte Carlo simulations using several specifications, including deterministic and stochastic switching trends. The results showed that the proposed technique is able to explain the true cycles under several statistical tests. As an alternative, we used the Hodrick-Prescott (HP) filter to obtain the cycle component of the simulated series. This technique produced cycles opposite to the true ones in some cases. Several tests indicate that this technique is greatly influenced by the erratic component on the data generated processes. Further, we applied our technique to the real Brazilian GDP – Gross Domestic Product – series. According to our methodology, the Brazilian economy showed business cycles generated by short run economic policies up to 1994. After this period business cycle was no longer the influential factor and the long run growth component was the dominant one.