Seasonal adjustment of weekly data

This paper summarizes and assesses several of the most popular methods to seasonally adjust weekly data. The industry standard approach, known as X-13ARIMA-SEATS, is suitable only for monthly or quarterly data. Given the increased availability and promise of non-traditional data at higher frequencies, alternative approaches are required to extract relevant signals for monitoring and analysis. This paper reviews four such methods for high-frequency seasonal adjustment. We find that tuning the parameters of each method helps deliver a properly adjusted series. We optimize using a grid search and test for residual seasonality in each series. While no method works perfectly for every series, some methods are generally effective at removing seasonality in weekly data, despite the increased difficulty of accounting for the shock of the COVID-19 pandemic. Because seasonally adjusting high-frequency data is typically a difficult task, we recommend closely inspecting each series and comparing results from multiple methods whenever possible