Challenges Related to Regression Composite Estimation in Statistics Canada's Labour Force Survey

• Activity Number: 334 - New Methods for Survey Weighting and Variance Estimation
• Type: Contributed
• Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Survey Research Methods Section
• Author(s): Martin Pantel* and Jack Gambino and Wisner Jocelyn and Emmanuel Benhin and Dany Faucher
• Companies: Statistics Canada and Statistics Canada and Statistics Canada and Statistics Canada and Statistics Canada
• Keywords: composite estimation; rotating panels; diagnostics; labour

Abstract

Like many labour force surveys around the world, the Canadian Labour Force Survey (LFS) uses a rotating panel design. The estimation method used by the LFS, called regression composite estimation (RCE), calibrates the unit weights to leverage the month-to-month correlation of labour force status due to the overlapping sample. This can lead to substantial improvement in the efficiency of the estimates. The version of RCE used for the LFS includes a combination of change-driven and level-driven calibrations, allowing it to compromise between improving estimates of change and level. In contrast to these benefits, the application of RCE in a practical setting such as the LFS brings some challenges. Some of the underlying assumptions are affected by non-response, we cannot exploit month-to-month correlation for new (and outgoing) rotation groups and, as noted by Fuller and Rao (2001), the change-driven form of RCE may lead to a “drift” in estimates of level. This presentation will focus on the identification and measurement of the “drift”, and on the effect of incoming rotation groups on key LFS results. The diagnostics, and some potential corrective measures, will be explored.