Seasonal adjustment of CPIs during the COVID-19 pandemic and beyond

This paper examines the presence of seasonality in CPI in 36 OECD economies that provide monthly CPI data and reviews the properties of standard methods, namely X-13 and TRAMO-SEATS, in performing the adjustment. Evidence from statistical tests points to the presence of seasonality in headline CPI and its components, with stronger seasonality in some components. There are also indications of changes in seasonal pattern from 1980 to 2022, but it is not systematic across countries. Simulations suggest that differences between the two methods are small when applied to CPI in OECD countries in normal times. Differences between the direct (adjusting all-item CPI and components independently) and the indirect (aggregating the seasonally adjusted components) approaches are also minimal when using the 12 divisions of the 1999 COICOP classification, limiting the need for a reconciliation method. Although large shocks should in theory affect seasonal adjustment, there is no strong evidence of a change in CPI seasonal patterns following the COVID-19 pandemic. This issue needs, however, to be revisited once the effects of the shock, including its impact on inflation have fully dissipated. The extent of revisions implied by the seasonal adjustment should be among the criteria for choosing a seasonal adjustment method, as CPI is often used in indexation and legal documents. The paper provides a summary of how communication is handled by selected OECD countries and provides a list of best practices that can be drawn upon by a National Statistical Office aiming to publish seasonally adjusted CPI.