Measuring the probability of a business cycle turning point by using a multivariate qualitative hidden Markov model

A two-step procedure to produce a statistical measure of the probability of being in an accelerating or decelerating phase of economic activity is proposed. It consists of, first, an extraction of the individual linear innovations of a set of relevant macroeconomic variables whose signs are accumulated into a qualitative vector process and, second, of a factor analysis applied to this vector. The factor process is a two-state Markov process of order one whose states are described as favourable and unfavourable. Estimated on French business surveys, this measure appears to be a competitive coincident indicator.