IS A TEMPORARY DICHOTOMY EMERGING BETWEEN ESTABLISHMENT DATA AND HOUSEHOLD DATA REGARDING U.S. EMPLOYMENT?

Yesterday’s (Friday, March 6) release by the U.S. Bureau of Labor Statistics (BLS) of the change in total nonfarm payroll employment (seasonally adjusted) showed that the U.S. economy added approximately 273,000 jobs in February according to the establishment survey of employers.  (https://www.bls.gov/news.release/empsit.nr0.htm)  By contrast, the household survey from the BLS indicated that on a seasonally adjusted basis, the U.S. economy added about 45,000 jobs – a much smaller number.  (https://www.bls.gov/news.release/empsit.a.htm)  While the February 2020 establishment survey indicates at least some strength if not considerable strength in the U.S. labor market before the potential negative impact of the COVID-19 coronavirus begins to impact the data, the household survey casts some doubt on just how strong the labor market was. 

Readers should know that with data not seasonally adjusted, my calculations show using data from the household survey an increase in employment of more than one million jobs in February 2020, but the same was also true in 2018, 2019, and perhaps other earlier years (at least based on the initial releases of those times).  (Table A-1 in each of the following:  https://www.bls.gov/news.release/archives/empsit_03062020.htm, 

https://www.bls.gov/news.release/archives/empsit_03082019.htm, and

https://www.bls.gov/news.release/archives/empsit_03092018.htm)  Thus, the February 2020 increase with data not seasonally adjusted does not necessarily reflect unexpected strength in the labor market.  In fact, my calculations find that the February 2020 increase in employment from the BLS household survey was about 130,000 jobs less than the increase in February 2019 and more than 480,000 jobs less than the increase in February 2018.

Reviewing some seasonally adjusted data from 2020 and relatively recent years, the BLS household survey data indicate that the estimates of total U.S. employment from both January and February 2020 are less than the estimate from December 2019.  (series LNS12000000 available at:  https://data.bls.gov/cgi-bin/surveymost)  This is the first time since December 2008 and January and February 2009 that this happened!  It seems that in recent months, the change in employment indicated by the BLS establishment survey has been a more favorable measure of labor market conditions than the change from the household survey.  For example, the seasonally-adjusted household survey data accessed yesterday (March 6, 2020) show that total estimated employment in the U.S. fell in January 2019, March 2019, April 2019, November 2019, and January 2020.  The January 2019 estimated decrease reached nearly 200,000 jobs, and the March 2019 estimated decrease amounted to over 100,000 jobs.   No month in 2019 or 2020 showed a decrease in jobs in the establishment survey.  However, for February of 2019, the estimate from the establishment survey shows an increase of only about 1,000 jobs. 

Based on a comparison of data from 2018 and 2019, the difference (at least in terms of whether net jobs are added or lost) between the establishment survey and the household survey may be growing.  Only one month in the year 2018 had a decrease in estimated employment in either BLS survey, and that was in the household survey only.  I calculate from the data that estimated employment in the household survey fell by nearly 400,000 jobs in August 2018.  My understanding is that the BLS household survey is more volatile that the establishment survey.  Could the August 2018 BLS household survey decrease in employment be more of an outlier, while decreases in 2019 be part of a trend?

In light of more months recently where the BLS establishment survey reports an increase in jobs but the household survey reports a decrease, both with seasonally adjusted data, has a temporary dichotomy arisen in the household survey versus the establishment survey?  Are there conclusions to draw?  To the extent that the household survey results are accurate, they help to explain what some have described and as mentioned in my earlier blog entry (https://harrisonhartman.blogspot.com/2020/02/mixed-signals-about-us-economy-part-two.html) as a somewhat weak economic start to the new year.  Given the likely negative economic impact of the coronavirus, if the U.S. economy enters a short-term contraction or a recession in 2020, will this divergence in the two BLS employment surveys last? 

(Please note that subsequent data revisions may change the analysis in this blog entry.  Thank you for your understanding.)