Some statistical aspects on methods for detection of turning points in business cycles

Abstract

Statistical and practical aspects on methods for on-line turning point detection in business cycles are discussed. When a method is used on a real data set, there are a number of special data problems to be considered. Among these are: the effect of smoothing, seasonal variation, autoregression, the presence of a trend and problems with multivariate data. Different approaches to these data problems are reviewed and discussed. In a practical situation, another important aspect is the estimation procedure for the parameters of the monitoring system. Three likelihood based methods for turning point detection are compared, one based on a Hidden Markov Model and another including a non-parametric estimation procedure. The three methods are used to analyze an actual data set of a period of the Swedish industrial production. The relative merits of comparing methods by one real data set or by simulations are discussed.