• time series
• filters
• Exponential Power distribution
• Business Cycle

The work assesses the impact of alternative filtering techniques on
the distributional properties of US time series. We consider filtered
time series of US output levels and growth rates and of other relevant
macroeconomic variables, and we fit the distributions. Moreover, as
regards US output growth rates, we analyse the distribution of isolate
different frequency bands. The main findings are that the majority of
US levels and growth rate distributions show tails much fatter than
those of a normal distribution, and that this result is robust to a series
of alternative filtering methods. Finally, we offers some theoretical
interpretations regarding the statistical evidences, in order to explain
why the majority of the US time series displays fat tails.