Disentangling Trend and Seasonality in Panel Data: An Empirical Analysis of Food Product Sales

Abstract

This paper provides a novel approach to extract the trend and seasonal components from panel data consisting of individual entries showing both strong trend and seasonality. For such a data set, the usual principal component analysis generally fails to disentangle them. In the paper, we suggest a methodology to separately identify them using the Hodrick-Prescott filter that is commonly and widely used to remove trends in various economic data. We apply our methodology to a food product sales panel data and show that it effectively disentangles the trend and seasonal components in the data set.