Seasonal Time Series Analysis: Definitions, Modeling, Changepoints, and Adjustment

High frequency data have been widely used in official statistics for several years, especially after the Covid-19 pandemic, which increased the demand for infra-monthly indicators. Since then, they have moved from an experimental status to standard production. Many infra-monthly indicators show (multiple) seasonal patterns and need to be seasonally adjusted. Therefore, well-established seasonal adjustment methods for monthly and quarterly data have been extended to meet the methodological requirements of high frequency data and additional ad hoc methods have been developed. With a wide range of algorithms and tools available today, the user is faced with a selection problem.
We provide an empirical review of the available tools and underlying algorithms for seasonal adjustment of high-frequency data, comparing them from different perspectives: quality of the seasonal adjustment process (residual seasonality, revisions, forecasting), possibility to implement methodological refinements: automatic outlier detection and parameter selection, time-varying calendar correction. We also examine the accessibility, usability and performance of the tools in a mass production approach.
 

Forecasting in the presence of changepoints is a challenging task. If changepoints occur regularly within seasonal data, we should utilize that information in our predictions for the next season. In this paper we develop a framework for forecasting seasonal data that contains within-season changepoints. In our forecasts, we account for uncertainty in the changepoint locations both within and across seasons via a weighting approach across changepoint uncertainties. The framework is very flexible across different changepoint model assumptions and approaches to identifying confidence sets. We demonstrate the improvement in forecasting performance of our framework for two business data applications.