Spurious Periodic Autoregressions

Posted: 9 Apr 1999

See all articles by Tommaso Proietti

Tommaso Proietti

University of Rome II - Department of Economics and Finance

Abstract

The class of periodic autoregressive (PAR) models, suitably extended so as to allow for 'periodic integration', has recently found widespread application to economic time series as an alternative to the time-invariant models available in the literature. An elaborate modelling strategy has been proposed, and new tests for periodic integration have been envisaged, whose empirical performance tends to support the notion that the kind of non-stationary stochastic dynamics observed in time series arises as a consequence of periodic integration. This paper aims at challenging this view by means of a Monte Carlo experiment: we generate data according to a trend with a seasonality model such that the trend is a random walk with drift and the seasonal component is generated according to a stochastic trigonometric model. It is found that all the fundamental tools of PAR modelling will tend to provide spurious evidence in favour of a periodic model, and conclude that, as long as macroeconomic time series are concerned, PAR models are an overelaborate way of capturing essential features, such as indeterministic trends and seasonals, that are more parsimoniously accommodated by a time-invariant model.

JEL Classification: C22, C32

Suggested Citation

Proietti, Tommaso, Spurious Periodic Autoregressions. Available at SSRN: https://ssrn.com/abstract=156706

Tommaso Proietti (Contact Author)

University of Rome II - Department of Economics and Finance ( email )

Via Columbia, 2
Rome, 00133
Italy

Do you have negative results from your research you’d like to share?

Paper statistics

Abstract Views
664
PlumX Metrics