Seasonality, Cycles and Unit Roots

Inference on ordinary unit roots, seasonal unit roots, seasonality and business cycles are fundamental issues in time series econometrics. This paper proposes a novel approach to inference on these features by focusing directly on the roots of the autoregressive polynomial rather than taking the standard route via the autoregressive coefficients. Allowing for unknown lag lengths and adopting a Bayesian approach we obtain posterior probabilities for the presence of these features in the data as well as the usual posteriors for the parameters of the model