C O V I D - 1 9 R E P O RTS
Continuity of household surveys after the coronavirus disease (COVID-19) pandemic December 2020
Summary In the context of the global coronavirus disease (COVID-19) pandemic, the Economic Commission for Latin America and the Caribbean (ECLAC) is preparing a series of short publications with key policy recommendations for this period. This note, the third on household surveys, offers some considerations regarding the continuity of this type of statistical instrument in a context of initial lifting of the restrictions on movement in the region that were put in place to contain the spread of the pandemic.
Introduction1 In the COVID-19 pandemic, most countries of Latin America and the Caribbean took measures to restrict citizens’ movement. These restrictions prevented household surveys from being taken face-to-face. As a result, some surveys were suspended, and others were conducted by telephone.
Summary Introduction A. Suggestions for mixed-mode household surveys in the return to a new normal B. Comparability and the impact of COVID-19 C. Conclusions Bibliography
The change in the data collection mode from face-to-face to telephone —which was necessary to continue producing employment and income statistics, that are particularly important in the pandemic— has posed new challenges for national statistical offices and other public agencies that take surveys. These challenges were shared and analysed in various virtual seminars as part of the work of the Knowledge Transmission Network of the Statistical Conference of the Americas of ECLAC, as well as during the nineteenth meeting of the Executive Committee of the Conference (see ECLAC, 2020a and 2020b). The ECLAC Statistics Division also made some recommendations to guarantee the quality of statistics with the changes in collection methodology and has provided technical assistance to several countries for their implementation (see ECLAC, 2020c and 2020d). The recommendations focused on the inherent properties of sampling strategies, namely the survey sampling designs and estimation procedures used to draw inferences from collected data. In short, it was recommended to: ● Base publication of official statistics from household surveys on probabilistic selection of samples and not on predictive models. ● Define probabilistic monitoring panels, based on the most recent months of face-toface collection, to conduct surveys by telephone, maintaining the population contained in the national statistical offices’ master sampling frame as the reference population. ● If response rates were low, publish statistics at the national level, without the usual disaggregation, explicitly mentioning the period over which the information was gathered. ● Use models to correct selection and coverage bias, which would allow for adjustment of weights based on the auxiliary information available in the monitoring panel. 1
The cut-off date for the information used to prepare this report is 31 October 2020, unless otherwise indicated.
1