Survey Non-Response in COVID-19 Times: The Case of the Labour Force Survey

Abstract

During the Covid-19 pandemic, labour-force survey non-response rates have surged in many countries. We show that in the case of the Canadian Labour Force Survey (LFS), the bulk of this increase is due to the suspension of in-person interviews following the adoption of telework in Federal agencies, including Statistics Canada. Individuals with vulnerabilities to the Covid-19 economic shock have been harder to reach and have been gradually less and less represented in the LFS during the pandemic. We present evidence suggesting that the decline in employment and labour-force participation have been underestimated over the March-July 2020 period. We argue that these non-response issues are moderate when analyzing aggregate outcomes, but that researchers should exert caution when gauging the robustness of estimates for subgroups. We discuss practical implications for research based on the LFS, such as the consequences for panels and the choice of public-use versus master files of the LFS.

Food Insecurity amid the COVID-19 Pandemic: Food Charity, Government Assistance, and Employment

“…In Quebec, where the pandemic hit the hardest during the early stage, more than one-fifth of the workers lost jobs, and nearly one-third experienced a change in employment status between March and May 2020; low-income households and workers in construction and accommodation bore the brunt of the burden ( Achou et al 2020 ). Although many workers laid off between February and April 2020 regained employment by July of the same year, recently unemployed people and those who were out of the labour force before the pandemic had difficulty finding employment; this was especially true for women and young people ( Brochu, Créchet, and Deng 2020 ). The gender employment gap widened between February and May 2020 among Canadian parents of young children, most visibly among those who were less educated ( Qian and Fuller 2020 ).…”

The Great Canadian Recovery: The Impact of COVID-19 on Canada's Labour Market

“…The bottom line is that the chief problem in the COVID-19 era has been birth nonresponse. Brochu and Créchet (2021) convincingly relate this pattern to pandemicinduced changes in interview modalities, particularly the suspension of in-person interviews, together with changes in telework and call centre arrangements. Overall, they find that the birth non-response problem was more severe for demographic groups with particular vulnerabilities to the COVID-19 shock.…”

“…5,6 At the outset, we note that an important issue in studying labour market developments during COVID-19 is data quality. At the start of the pandemic, the overall LFS non-response rate rose from 11.9% to 22.1% (Feb-March 2020) peaking at over 30% by late summer 2020 (Brochu and Créchet, 2021). This high rate of non-response has since persisted in the Canadian LFS, while the analogous non-response rate for the U.S. Current Population Survey, which peaked at about 35% in mid-2020, has since fallen to the 20-25% range but still lies above the pre-pandemic level.…”

“…In a key contribution, Brochu and Créchet (2021) decompose this non-response into "birth" non-response, a failure to initiate an interview at all, and "subsequent" non-response, a failure to achieve a follow-on interview (i.e., attrition from the birth sample). Since neither birth non-response nor subsequent non-response is directly observed, Brochu and Créchet use the incidence of Whole Record Imputation (WRI), only used for non-initial interviews (and for some specific age ranges), and the relative fall-o↵ of response (i) overall and (ii) excluding records with WRI, to assess the importance of "birth" non-response and "subsequent" non-response.…”

“…An implication of the work ofBrochu and Créchet (2021) on the problems with initial LFS interviews in the pandemic period is that there is merit in using a panel where the initial interview was conducted pre-pandemic. Also, we note that this use of the full length of the panel is rare in empirical work with the LFS and illustrates a benefit of the survey's six-month long rotation group structure.…”