Productivity and Costs, First Quarter 2020, Revised: Adjusting March 2020 employment and hours to account for effects of COVID-19

BLS quarterly estimates of labor productivity growth combine output data published by the Bureau of Economic Analysis (BEA) with hours data primarily from the Current Employment Statistics (CES) survey. The reference period for the CES, the pay period that includes the 12th of the month, largely predated many of the COVID-19-related job losses and business closures that occurred in the latter part of March. (Please see the frequently asked questions for March that accompanied the March Employment Situation release and the frequently asked questions for April that accompanied the April Employment Situation release.) In order to publish an estimate of total hours that reflects the entire first quarter of 2020, BLS made adjustments to its estimates of total hours worked. This adjustment is consistent with BEA estimation methods used to capture the changes in economic activity that occurred at the end of the first quarter of 2020.

Total hours are produced by combining data on employment and average weekly hours. For the preliminary estimates of first quarter labor productivity, the adjustments to total hours were made through adjustments to employment, and were based primarily on the Department of Labor’s Employment and Training Administration weekly reports of the number of initial claims for unemployment insurance (UI) benefits. No adjustment was made to average weekly hours or to total hours worked by the self-employed.

The revised estimates of first quarter labor productivity use the same basic methodology to estimate employment, but also use data from the April CES to better account for job losses that occurred in the latter half of March. The revised estimates also adjust average weekly hours of wage and salary workers and total hours worked by the self-employed.

As was done in the preliminary estimate, BLS combined data from the CES with the initial UI claims data to estimate March employment week by week. BLS extrapolated recent trends in employment to estimate employment for the first week of March. These estimates account for employment growth that likely occurred between the end of the reference period in February and the first week of March. The revised March 2020 CES data were used directly to estimate employment for week two of March.

To account for the job losses that occurred in the latter part of March, BLS used not seasonally adjusted initial UI claims data to estimate job losses for 3-digit NAICS industries. In the preliminary estimate, BLS estimated employment for the third week of March using week two employment adjusted by the initial UI claims for the week ending on March 21. For the remainder of the month, BLS similarly estimated employment using the initial UI claims for the week ending on March 28, as well as a portion of the initial UI claims filed in the week ending April 4 to account for the last few days of March.

The revised estimate followed the same procedure for employment, except that total UI claims for the week ending on March 14 through the week ending on April 11 were controlled to the decline in CES employment between March (revised) and April.

Average weekly hours for wage and salary workers were estimated week by week by interpolating between the March and April estimates from the CES. Total hours worked by the self-employed were similarly estimated using data from the March and April CPS. Both interpolations were weighted so that about half of the declines in hours between March and April were assigned to March. In addition, outlier identification was used in the seasonal adjustment process to account for extreme movements in the April data.

These estimates are subject to revision as source data are revised and new data become available.

Last Modified Date: June 4, 2020