CPS and COVID-19
As the COVID-19 pandemic continues to up-end many aspects of daily life, the Census Bureau continues to collect monthly Current Population Survey (CPS) data. At time of writing, the Bureau of Labor Statistics (BLS) has not indicated that data releases will be delayed or interrupted as a result of the current public health crisis. IPUMS CPS will continue to process data as quickly as possible. Updates will be provided to registered IPUMS CPS users via email; follow @ipums on Twitter for the latest data release information.
Beginning with March of 2020, the BLS is releasing supplementary information about the impact of the COVID-19 pandemic on collection and processing of the data underlying the monthly Employment Situation news release (referred to hereafter as the impact statement). Information relevant to CPS data will be summarized here with links to the original documentation. As the situation evolves, this page will be updated with pandemic-related news about CPS data collection and processing.
General Data Collection
During March of 2020, CPS data continued to be collected with some modifications in procedure for the safety of respondents and Census Bureau employees.
Enumeration for March began on the 15th. Households that normally receive in-person interviews are those first entering the survey rotation (month-in-sample one) and those that were re-entering the survey after the eight-month hiatus (month-in-sample five) (MISH). Households at other points in the CPS rotation are typically contacted via phone. The Census Bureau suspended in-person data collection on the 20th and closed two of the call centers that assisted in collecting CPS data. However, data collection continued by phone and efforts were made to reach households via phone that would normally have received an in-person visit. Response rates were lower than average by about 10 percentage points in March 2020 (see table below). Response rates were even lower for those households in rotation groups that would normally have received an in-person visit (over 20 percentage points lower for month-in-sample one and over 10 percentage points for month-in-sample five).
For more information see item 5 in the BLS March 2020 impact statement.
Interviews for April were conducted exclusively by phone. The two Census Bureau call centers that usually assist with the collection of CPS data remained closed. Response rates continued to be low in April, over 10 percentage points below the average (see below). Response rates continued to be particularly low among rotation groups one and five who would have normally received a visit from the enumerator. Month-in-sample one and five response rates were similar to those same groups in March. Additionally, those households that entered the survey for the first time in March had a similarly low response rate for their second interview in April.
For more information see item 5 in the BLS April 2020 impact statement.
Interviews were conducted exclusivley by phone again in May. The two Census Bureau call centers that usually assist with the collection of CPS data remained closed. Response rates continued to be lower in May both relative to April and to response rates for this month in 2019. Response rates for those groups entering and re-entering the survey rotation remained espeically low. Low response rates persisted among those groups who entered the survey in March and April. See the CPS Response Rates, 2019-2020 table below for details.
For more information see item 5 in the BLS May 2020 impact statement.
Interviews were conducted exclusivley by phone again in June. The two Census Bureau call centers that usually assist with the collection of CPS data remained closed. Response rates continued to be lower in June both relative to May and to response rates for this month in 2019. Response rates for those groups entering and re-entering the survey rotation remained espeically low. See the CPS Response Rates, 2019-2020 table below for details.
For more information see item 4 in the BLS June 2020 impact statement.
In-person interviews began again in some areas of the country in July, and the Census Bureau call centers that usually assist with CPS data collection also returned to interviewing in a limited capacity in this month. Response rates rose slightly compared to May and June, but remain lower than average. This is particularly true among incoming rotation groups. Low response rates persist for those rotation groups that entered the CPS in May and June.
For more information see item 3 in BLS July 2020 impact statement
In-person interviews continued in some areas of the country in August, and the Census Bureau call centers that usually assist with CPS data collection also continued to conduct interviews in a limited capacity in this month. Response rates rose slightly compared to May, June, and July, but remain lower than average. This is particularly true among incoming rotation groups. Low response rates persist for those rotation groups that entered the CPS in May, June, and July.
For more information see item 3 in BLS August 2020 impact statement
In-person interviews were expanded to all areas of the country in September. In-person interviews were conducted only after attempting to reach all households by phone. The two Census Bureau call centers that normally assist with CPS interviewing also continued to conduct phone interviews on a limited basis. Response rates remained low in September, but represent an increase over previous months.
For more information see item 3 in BLS September 2020 impact statement
|CPS Response Rates, 2019-2020|
|Year||Month||MIS 1||MIS 2||MIS 3||MIS 4||MIS 5||MIS 6||MIS 7||MIS 8||Total||2019||January||80.44||83.81||84.17||84.92||81.10||81.66||83.54||85.49||83.14|
AttritionThere is evidence of non-random attrition in the CPS since the beginning of the COVID-19 pandemic. In the appendix of their working paper, "Determinants of Disparities in COVID-19 Job Losses", Montenovo et al. demonstrate that those who respond to the CPS in February 2020, but are not present in the March sample are, on average, younger, more diverse, and less educated than those who also respond in both February and March of 2020.
The ASEC is a combination of the March Basic Monthly sample and hispanic oversample collected in November of the previous year and SCHIP oversamples collected in August-November of the previous year and February and April of the survey year. As described above, the coronavirus pandemic affected CPS data collection in March and April. As the ASEC interviews collected in April are only from those in the first and fifth rotation groups, where response rates were particularly low, there were fewer respondents available for the ASEC supplement in this month.
The many months from which the ASEC is drawn and the possibility that a respondent answers the basic monthly portion of the survey but not the supplement make calculating a response rate for the ASEC complicated. The Census Bureau estimates that the unweighted combined supplement response rate was 61.1% in 2020, compared to 67.6 percent in 2019. Median income and educational attainment of 2020 ASEC respondents are highter than those in 2019, suggesting that non-random attrition also affects the ASEC in 2020.
For more information on the impact of the COVID-19 pandemic on ASEC data, see Appendix G of the 2020 ASEC documentation.
Before March interviews began, enumerators were given specific instructions on how to categorize COVID-19-related deviations from normal work/employment during the reference week. These instructions remained in effect during April through August interviews.
Those who reported usually working full time (WKSTAT) but reported working less than 35 hours during the reference week (AHRSWORKT), are asked for the reason they worked less than their usual hours during the reference week (UH_HRRSN3_B1). COVID-19-related responses were categorized as follows:
- "Own illness, injury, or medical problem" includes those who were under quarantine or self-isolating due to health concerns during the reference week.
- "Slack work or business conditions" includes those who had hours reduced as a result of the coronavirus outbreak but were not themselves ill or quarantined.
- "On layoff (temporary or indefinite)" includes those who did not work at all during the reference week but were not themselves ill or quarantined. For example, those working at entertainment venues at which all shows had been canceled or restaurant workers whose places of business are temporarily closed are included in this category.
By the Census Bureau's definition, respondents who reported doing no work during the reference week must expect to be called back to work within six months in order to be classified as "unemployed on temporary layoff". If a respondent was uncertain whether they would be able to return to work within six months due specifically to the COVID-19 pandemic, their response to this question was recorded as "yes" (UH_LAY6M_B1) so that they would be included in the unemployed on temporary layoff category.
For more information on pandemic-related enumerator instructions in
- March 2020: see items 7 and 10 in the BLS March 2020 impact statement
- April 2020: see items 8 and 11 in the BLS April 2020 impact statement
- May 2020: see items 8 and 11 in the BLS May 2020 impact statement
- June 2020: see items 7 and 11 in the BLS June 2020 impact statement
- July 2020: see item 6 in the BLS July 2020 impact statement
- August 2020: see item 6 in the BLS August 2020 impact statement
- September 2020: see item 6 in the BLS September 2020 impact statement
Despite these instructions, there is evidence that some persons who should have been classified as "unemployed on temporary layoff" due to COVID-19-related closures or reductions in business were miscategorized as "employed but not at work." In accordance with standard procedure, BLS did not attempt to recategorize survey responses. However, as this misclassification error appears to be non-trivial and has persisted since March of 2020, Before June collection began, Census conducted additional traning and review of the guidance discussed above, and included these updated instructions in the survey to assist enumerators. These measures seem to have been somewhat effective, reducing the share of those misclassified as "employed but not at work" in June compared to preceeding months. Further training will was administered in order to further reduce this error before July collection began. Misclassification rates were much lower in July, August, and September than in preceeding months.
For more information on misclassified responses to employment questions in
- March 2020: see items 11 and 12 in the BLS March 2020 impact statement
- April 2020: see items 12 and 13 in the BLS April 2020 impact statement
- May 2020: see items 11-14 in the BLS May 2020 impact statement
- June 2020: see items 11-14 in the BLS June 2020 impact statement
- July 2020: see items 5-8 in the BLS July 2020 impact statement
- August 2020: see items 5-8 in the BLS August 2020 impact statement
- September 2020: see items 5-8 in the BLS September 2020 impact statement
The value for "unemployed on temporary layoff" is recoded to "unemployed, experienced worker" in the IPUMS CPS variable EMPSTAT. Original codes and labels are available to researchers as part of the unharmonized variable UH_MLR_B3. For more information on IPUMS CPS unharmonized varibles, see our documentation here.
To attempt to shed light on the effects of the COVID-19 outbreak on the American labor market, the Bureau of Labor statistics will add five new questions about work in the time of COVID-19 to the basic monthly CPS questionnaire.
These questions were first asked in May of 2020, however it is unclear when data from these questions will become publicly available. These questions can be previewed here. Tables made using these data from May and June have been made available here. There is currently no timeline for the release of the underlying microdata. IPUMS CPS will process these data as soon as they become available and make them available through our extract system. Follow @ipums on twitter for the latest data release information.