The main purpose of the paper is to present a novel approach to formal business cycle estimation. The authors propose a non-standard subsampling procedure, in order to make formal statistical inference about the properties of the business cycle. The paper shows, that business cycle can be modeled by parameters of discrete spectra of the Almost Periodically Correlated (APC) stochastic process.
In the empirical part of the paper the authors analyze the cyclical behavior of production sector in Poland with the use of the model with APC stochastic component. The model characterizes business cycle on the basis of industrial production index and also on some subsector indices.
The main conclusion presented in empirical illustration is that, during period 1995-2009, the model conﬁrms (using statistical tools) the presence of 3-4 years length of business cycle in industrial production index in Poland. This result was obtained either on the basis of the total index and also analyzing sub indices.
All indices and sub indices supported signiﬁcance of short term and middle term ﬂuctuations, attaching relatively small amplitudes for periodicity with length less than 2 years. Additionally, in all time series it has been detected existence of longer term cycle (7-8 years), interpreted in the paper as a trend or a mixture of both trend and business cycle ﬂuctuations.