Who[m] Do You Trust, the BLS or the ISM?
Before he took over hosting The Tonight Show” in September 1962, Johnny Carson hosted an afternoon quiz television show called “Who Do You Trust” (originally titled “Do You Trust Your Wife”). I used to watch it when I came home from school before digging into my homework. Looking at the revisions to the monthly report of nonfarm payroll employment, the show’s title gained more relevance to me now. Billions of dollars, maybe trillions, of transactions in financial instruments take place in reaction to the first estimate of the change in nonfarm payrolls on the day each month that the Bureau of Labor Statistics (BLS) releases its report. That first estimate of the monthly change in nonfarm payrolls gets revised in each of the following two months. But those revisions, although sometimes mentioned by the mainstream financial media, tend to be put aside in the frenzy of trading that takes place in the nanosecond after the BLS releases its monthly report of the Employment Situation.
I decided to take a look at how much the monthly changes in total nonfarm payrolls get revised from their preliminary estimate to their “final” estimate two months later. (I put quotation marks around final because there are of course, subsequent benchmark revisions.) Plotted in Chart 1 are the BLS first estimates for monthly changes in total nonfarm payrolls (the red bars) and its third estimates (the blue bars). (Ignore the November and December 2023 data points inasmuch as the BLS has not yet reported its final estimates for the changes in nonfarm payrolls for these months.) Plotted in Chart 2 are the monthly differences in changes in total nonfarm payrolls between the third estimate by the BLS and its first estimate. Notice that in the 10 months ended October 2023, there is only one month, July 2023, in which the third estimate is greater than the first estimate. Plotted in Chart 3 are the 10-month cumulative totals in the monthly differences between the third estimates of the changes in total nonfarm payrolls and the first estimates. Attention should be focused on the last data point, October 2023, which reads minus 417 thousand. So, in the first 10 months of 2023, the cumulative total of monthly changes in nonfarm payrolls was 417 thousand less when the third estimate of the monthly change in nonfarm payrolls is compared with the first estimate. But these 417 thousand jobs that got revised away probably had no effect on the analyses of the state of the US labor market made by the talking heads on CNBC and Bloomberg when the November and December 2023 first estimates of changes in nonfarm payrolls were reported by the BLS.
If this is not enough to make you skeptical of the first estimates of monthly changes in nonfarm payrolls, consider that according to David Rosenberg, founder and president of Rosenberg Research, “[m]ore than 40% of payroll growth in 2023 didn’t even come from the survey but from the fairy-tale ‘Birth-Death’ model”. (As an aside, David and I were the proverbial skunks at the garden party in 2006 when we were grizzly bears in the midst of raging bulls.) The BLS has developed a “Birth-Death model to adjust sample- based estimates of the Payroll Employment Survey to reduce a primary source of non-sampling error which is the inability of the sample to capture, on a timely basis, employment growth generated by new business formations”. Lastly, “the 90-percent confidence interval for the monthly change in total nonfarm employment from the establishment survey is on the order of plus or minus 120,000. Suppose the estimate of nonfarm employment increases by 50,000 from one month to the next. The 90-percent confidence interval on the monthly change would range from -70,000 to +170,000”. And policy decisions and investment decisions are made on the basis of the first-reported monthly change in nonfarm payrolls?!
If there were only a measure of economic activity that never got revised and whose behavior comported with the business cycle. Oh wait, there is, the Institute of Supply Management (ISM) manufacturing survey. The not-seasonally adjusted data in this survey do not get revised. Once a year, the seasonal factors are updated. Plotted in Chart 4 are the quarterly averages of the differences between the percent of respondents reporting an increase in production minus the percent reporting a decline in production. The quarterly data are seasonally-adjusted, not by ISM but by a Haver Analytics program using all available data. The shaded areas in the chart represent periods of recession. These net differences in production generally move higher during periods of economic recovery/expansion and generally decline during periods of recession or near recession (1963, 1995 and 2019). These net production data are not leading indicators, but rather are coincident indicators. If one were to chart the data for the net differences in new orders, you would see a similar pattern as to that shown in Chart 4.
Changes in nonfarm payrolls also are coincident indicators. But as I have shown, nonfarm payrolls are subject to significant revisions. The underlying ISM data never get revised, save for seasonal adjustment revisions. We all would like good leading indicators that do not get revised. But for a policymaker and investor isn’t it useful to have an indicator that provides a good reading on the current state of the economy and is not subject to radical revisions? So, who[m] do you trust, the BLS or the ISM?
Paul L. KasrielAuthorMore in Author Profile »
Mr. Kasriel is founder of Econtrarian, LLC, an economic-analysis consulting firm. Paul’s economic commentaries can be read on his blog, The Econtrarian. After 25 years of employment at The Northern Trust Company of Chicago, Paul retired from the chief economist position at the end of April 2012. Prior to joining The Northern Trust Company in August 1986, Paul was on the official staff of the Federal Reserve Bank of Chicago in the economic research department. Paul is a recipient of the annual Lawrence R. Klein award for the most accurate economic forecast over a four-year period among the approximately 50 participants in the Blue Chip Economic Indicators forecast survey. In January 2009, both The Wall Street Journal and Forbes cited Paul as one of the few economists who identified early on the formation of the housing bubble and the economic and financial market havoc that would ensue after the bubble inevitably burst. Under Paul’s leadership, The Northern Trust’s economic website was ranked in the top ten “most interesting” by The Wall Street Journal. Paul is the co-author of a book entitled Seven Indicators That Move Markets (McGraw-Hill, 2002). Paul resides on the beautiful peninsula of Door County, Wisconsin where he sails his salty 1967 Pearson Commander 26, sings in a community choir and struggles to learn how to play the bass guitar (actually the bass ukulele). Paul can be contacted by email at firstname.lastname@example.org or by telephone at 1-920-559-0375.