Periodic correlation vs. integration and cointegration (Okresowa korelacja a integracja i kointegracja)

In this paper we present a new approach to integration and cointegration. We show that a periodically correlated time series can be divided in a natural way into subseries that are integrated. Moreover, with high probability they are cointegrated. Therefore it is enough to show periodic correlation of the original series to conclude that the subseries are integrated. In the first part of the paper we present the main features of periodically correlated processes and a method of detecting periodic correlation. We illustrate it using a data set of spot electricity prices from the Nord Pool Power Exchange. In the next section we show that the subseries (one for each day of the week) exhibit integration as well as cointegration.