Persistence and seasonality in the US Industrial Production Index

Abstract

This paper uses a seasonal long-memory model to capture the behaviour of the US industrial production index (IPI) over the period 1919Q1-2022Q4. This series is found to display a large value of the periodogram at the zero, long-run frequency, and to exhibit an order of integration around 1. When first differences (of either the original data or their logged values) are taken, evidence of seasonality is obtained; more specifically, deterministic seasonality is rejected in favour of a seasonal fractional integration model with an order of integration equal to 0.14 for the original data and 0.29 for their logged values, which implies the presence of a seasonal long-memory mean reverting pattern.