Review of Available Programs for Seasonal Adjustment of Weekly Data

• Activity Number: 160 - Time Series Methodology: Modern Practices in Seasonal Adjustment and Software
• Type: Topic-Contributed
• Date/Time: Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Section for Statistical Programmers and Analysts
• Author(s): Thomas Evans* and Brian C Monsell and Michael Sverchkov
• Companies: Bureau of Labor Statistics and U.S. Bureau of Labor Statistics and Bureau of Labor Statistics
• Keywords: signal extraction; high-frequency time series; calendar effects; unobserved component models

Abstract

There has been a rise in interest for the seasonal adjustment of weekly data over the last few years. However, standard seasonal adjustment programs, such as X-13ARIMA-SEATS, assume constant periodicity, but weekly data can have either 52 or 53 weeks in a year. Weekly data are also difficult to seasonally adjust for multiple reasons. The week in which official holidays occur varies from year to year; some holiday effects do not occur every year; and even if the number of weeks in a year were a constant integer, the seasonal patterns would still change from year to year since the structure of the days in a month shift. The Bureau of Labor Statistics currently adjusts two weekly unemployment insurance claims series using a regression approach developed in the early 1990s, but other innovative programs are now in various stages of development. This paper will discuss various attributes of the programs, including ease of use, available diagnostics, and any technical support.