The Big Idea

A primer on the benchmark payroll revision

Stephen Stanley | August 16, 2024

This document is intended for institutional investors and is not subject to all of the independence and disclosure standards applicable to debt research reports prepared for retail investors. This material does not constitute research.

Once a year, the Bureau of Labor Statistics revises the payroll employment figures to incorporate the more complete benchmark data. A preliminary estimate for the next annual revision, which will cover the 12 months ended in March, is due to be published on August 21. A number of economists and Fed policymakers have suggested that the payroll gains over that 12-month period will be slashed dramatically. This piece walks through the process and offers some thoughts on the forthcoming revision.

Payroll survey data

The payroll employment figures are derived from a monthly establishment survey conducted by the BLS. According to the BLS, the survey canvasses about 120,000 business and government agencies, and the total job count from the sample represents about 27% of total nonfarm employment. As surveys go, this is a pretty big one. By comparison, the household survey captures results from about 60,000 US households out of around 130 million, or 0.05%.

Aside from garden-variety survey bias—the companies that the BLS happens to survey may be adding jobs faster or slower than the true underlying universe of firms—a key source of uncertainty for scaling the establishment survey results up to an estimate of total employment is the shift in the number of existing businesses. If new firms are being created faster or slower or if existing firms are closing faster or slower than the BLS assumes, then the true employment picture can differ from the monthly estimates. The BLS runs a birth-and-death model based on historical data to estimate these dynamics, but if there is a sharp turn in the economy, then the model’s projections may be off substantially.

All firms that participate in unemployment insurance are required to report to the federal government their employment levels once a quarter so that their taxes into the UI system can be ascertained. This amounts to a quarterly census of employment, covering almost the entire universe from which the establishment survey sample is selected. There is a small fraction of firms not included in the unemployment insurance system, which is why the level of payroll employment is somewhat higher than the raw benchmark numbers. These benchmark figures are reported on a quarterly basis with a lag of about five months. The fourth quarter 2023 report was issued on May 22, and the first quarter 2024 data are due out on August 21.

Once a year, the BLS replaces the survey estimates for payrolls with the full census results, eliminating the various errors in the preliminary data such as sampling error, birth-and-death and so on. The BLS chooses to consider the 12 months ending in March as its benchmark period. Therefore, once the first quarter numbers are available, BLS is in a position to estimate the magnitude of the benchmark revision. That announcement is what we will get in the upcoming week. Then, the actual revision to the data is implemented with the January release of the following year, the report that is published in early February. Thus, the benchmark data for the period from April 2023 through March 2024 will become official in early 2025.

To the extent that the employment census results are published quarterly, one can follow along with the discrepancy as the benchmark year proceeds. At the moment, we actually have three of the four quarterly reports that will correspond to the annual benchmark. On the basis of what we have so far, many economists and Fed policymakers have concluded that the benchmark revision for the March 2024 benchmark period will be sharply negative.

For example, Fed Governor Michelle Bowman in a speech on August 10 argued that the QCEW (quarterly benchmark) data “implies that job gains have been consistently overstated in the establishment survey since March of last year.” A footnote in her speech offers further detail: “The Q4 Quarterly Census of Employment and Wages (QCEW) administrative data show employment gains that are about 110,000 per month lower than what the Current Employment Statistics (CES) survey reported from March 2023 to December 2023.”

I was struck by the level of confidence in her statements. As I will lay out in detail, there is a significant degree of uncertainty associated with translating the quarterly QCEW data into a payroll benchmark revision. Thus, I thought it might be helpful to walk through the recent history of the relationship between establishment survey estimates of payrolls and the QCEW counts.

Crunching the numbers

Exhibit 1 shows the changes in employment levels in the survey-based payroll estimates and the QCEW census figures over the first nine months of the most recent benchmark period, the 12 months ended March 2024. The QCEW employment figures are reported only on a not seasonally adjusted basis, so I compare them to the not seasonally adjusted payroll numbers.

Just to note one other key technical point: for the payroll figures, these are the historical readings prior to the benchmark revisions. I went back and grabbed the numbers in real time for all payroll readings for which the current figures already reflect benchmark revisions. I always use the payroll estimate that incorporates the two months of regular revisions to the data but not subsequent benchmark adjustments. So, all historical comparisons compare the payroll figures in real-time before benchmark revisions with the QCEW data.

Exhibit 1: Payroll and QCEW Employment Changes, April 2023 through December 2023

Source: BLS.

As Exhibit 1 shows, the QCEW counts for three-quarters of the benchmark period suggest that the payroll gains over that period are cumulatively about 700,000 too high. Dividing that total by nine months would suggest that the payroll advances over that none-month period will be revised down by about 78,000 per month.

I would make two quick points before moving on. First, this calculation is still markedly smaller than Governor Bowman’s (which is to say, the Fed staff’s) estimate of 110,000. I do not know how that projection was derived. Second, it is important to point out that the currently estimated payroll gains for that nine-month period average a robust 233,000. So, if the QCEW figures translate in a straightforward way to the benchmark revision, then the average payroll gains from April through December of last year would be 155,000 per month, much lower but far from alarmingly weak.

We can make a prediction of the QCEW results for the first quarter to approximate the raw data that will dictate the benchmark revision. The estimates in Exhibit 2 assume that the relationship between the two series in the first quarter will be exactly the same as for the Q1 2023 period. The QCEW cell for March 2024 and the change are in red to signify that they are estimates.

Exhibit 2: Payroll and QCEW Employment Changes, April 2023 through March 2024

Source: BLS.

This calculation yields a shortfall in the QCEW of 817,000. If the benchmark revision is exactly that amount, it would work out to a downward revision on average of 68K per month. This sounds massive, but, again, note that the average payroll advance from the preliminary data for that 12-month period is 242K. So, the post-benchmark average increase would be a still-solid 174,000.

So far, all of this seems pretty straightforward, right? It is time to introduce the complications. The benchmark figures for the last benchmark period, covering April 2022 through March 2023, are shown in Exhibit 3.

Exhibit 3: Payroll and QCEW Employment Changes, April 2022 through March 2023

Source: BLS.

As you can see, the shortfall between the two series in the prior benchmark period was very similar to my estimate for the current one. For the 12 months ending March 2023, the QCEW data showed a 775,000 smaller increase than the payroll estimates, remarkably similar to my estimate for the current benchmark period of 817,000.

Here is where things get tricky. What do you suppose was the size of the benchmark revision for the 12 months to March 2023? If you said -775,000, that would be understandable. However, the correct answer is -266,000, barely more than one-third of the QCEW-payroll gap. If the March 2024 benchmark revision follows last year’s pattern, then it would be less than 300,000, and the monthly average job gain would only be shaved by about 23,000, leaving the 12-month average still well above 200,000.

Clearly, this is not as straightforward a process as it may have seemed. Exhibit 4 shows the historical relationship between the QCEW and payroll figures and the actual seasonally adjusted benchmark revisions over the past several years. As you can see, there is no easily predictable pattern.

Exhibit 4: Payroll and QCEW Employment Changes vs. Actual Benchmark Revisions

Source: BLS.

Over the six years prior to the last benchmark revision, the results were all over the map. In three of them (2022, 2019, and 2017), the benchmark revision was quite close to the gap between the two series. In the other three years, the actual benchmark revisions were quite different, including two years (2020 and 2021) when even the direction of the revision was the opposite of what would have been expected.

Conclusion

If I have left you feeling less confident about what will happen on August 21 than before reading the piece, then I have been successful. The translation from the QCEW data to the benchmark revisions has been far from simple over the years. Governor Bowman and others may end up being right that the payroll figures are going to undergo a sharp downward revision, but the degree of certitude with which she and many other economists have approached this question strikes me as misplaced. As I see it, a noticeable downward revision seems likely, but its magnitude could be moderate (as was the case a year ago) or substantial.

Moreover, while it is simple enough to extrapolate forward, a negative revision that extends from April 2023 through March 2024 does not necessarily tell us anything about the current state of the labor market. Those who are eager to call a recession would likely presume that whatever downward revisions apply to the 12 months ending in March would be as large or larger for the months since then, another assumption that may or may not be true but would certainly not be guaranteed.

Stephen Stanley
stephen.stanley@santander.us
1 (203) 428-2556

This material is intended only for institutional investors and does not carry all of the independence and disclosure standards of retail debt research reports. In the preparation of this material, the author may have consulted or otherwise discussed the matters referenced herein with one or more of SCM’s trading desks, any of which may have accumulated or otherwise taken a position, long or short, in any of the financial instruments discussed in or related to this material. Further, SCM may act as a market maker or principal dealer and may have proprietary interests that differ or conflict with the recipient hereof, in connection with any financial instrument discussed in or related to this material.

This message, including any attachments or links contained herein, is subject to important disclaimers, conditions, and disclosures regarding Electronic Communications, which you can find at https://portfolio-strategy.apsec.com/sancap-disclaimers-and-disclosures.

Important Disclaimers

Copyright © 2024 Santander US Capital Markets LLC and its affiliates (“SCM”). All rights reserved. SCM is a member of FINRA and SIPC. This material is intended for limited distribution to institutions only and is not publicly available. Any unauthorized use or disclosure is prohibited.

In making this material available, SCM (i) is not providing any advice to the recipient, including, without limitation, any advice as to investment, legal, accounting, tax and financial matters, (ii) is not acting as an advisor or fiduciary in respect of the recipient, (iii) is not making any predictions or projections and (iv) intends that any recipient to which SCM has provided this material is an “institutional investor” (as defined under applicable law and regulation, including FINRA Rule 4512 and that this material will not be disseminated, in whole or part, to any third party by the recipient.

The author of this material is an economist, desk strategist or trader. In the preparation of this material, the author may have consulted or otherwise discussed the matters referenced herein with one or more of SCM’s trading desks, any of which may have accumulated or otherwise taken a position, long or short, in any of the financial instruments discussed in or related to this material. Further, SCM or any of its affiliates may act as a market maker or principal dealer and may have proprietary interests that differ or conflict with the recipient hereof, in connection with any financial instrument discussed in or related to this material.

This material (i) has been prepared for information purposes only and does not constitute a solicitation or an offer to buy or sell any securities, related investments or other financial instruments, (ii) is neither research, a “research report” as commonly understood under the securities laws and regulations promulgated thereunder nor the product of a research department, (iii) or parts thereof may have been obtained from various sources, the reliability of which has not been verified and cannot be guaranteed by SCM, (iv) should not be reproduced or disclosed to any other person, without SCM’s prior consent and (v) is not intended for distribution in any jurisdiction in which its distribution would be prohibited.

In connection with this material, SCM (i) makes no representation or warranties as to the appropriateness or reliance for use in any transaction or as to the permissibility or legality of any financial instrument in any jurisdiction, (ii) believes the information in this material to be reliable, has not independently verified such information and makes no representation, express or implied, with regard to the accuracy or completeness of such information, (iii) accepts no responsibility or liability as to any reliance placed, or investment decision made, on the basis of such information by the recipient and (iv) does not undertake, and disclaims any duty to undertake, to update or to revise the information contained in this material.

Unless otherwise stated, the views, opinions, forecasts, valuations, or estimates contained in this material are those solely of the author, as of the date of publication of this material, and are subject to change without notice. The recipient of this material should make an independent evaluation of this information and make such other investigations as the recipient considers necessary (including obtaining independent financial advice), before transacting in any financial market or instrument discussed in or related to this material.

Important disclaimers for clients in the EU and UK

This publication has been prepared by Trading Desk Strategists within the Sales and Trading functions of Santander US Capital Markets LLC (“SanCap”), the US registered broker-dealer of Santander Corporate & Investment Banking. This communication is distributed in the EEA by Banco Santander S.A., a credit institution registered in Spain and authorised and regulated by the Bank of Spain and the CNMV. Any EEA recipient of this communication that would like to affect any transaction in any security or issuer discussed herein should do so with Banco Santander S.A. or any of its affiliates (together “Santander”). This communication has been distributed in the UK by Banco Santander, S.A.’s London branch, authorised by the Bank of Spain and subject to regulatory oversight on certain matters by the Financial Conduct Authority (FCA) and the Prudential Regulation Authority (PRA).

The publication is intended for exclusive use for Professional Clients and Eligible Counterparties as defined by MiFID II and is not intended for use by retail customers or for any persons or entities in any jurisdictions or country where such distribution or use would be contrary to local law or regulation.

This material is not a product of Santander´s Research Team and does not constitute independent investment research. This is a marketing communication and may contain ¨investment recommendations¨ as defined by the Market Abuse Regulation 596/2014 ("MAR"). This publication has not been prepared in accordance with legal requirements designed to promote the independence of research and is not subject to any prohibition on dealing ahead of the dissemination of investment research. The author, date and time of the production of this publication are as indicated herein.

This publication does not constitute investment advice and may not be relied upon to form an investment decision, nor should it be construed as any offer to sell or issue or invitation to purchase, acquire or subscribe for any instruments referred herein. The publication has been prepared in good faith and based on information Santander considers reliable as of the date of publication, but Santander does not guarantee or represent, express or implied, that such information is accurate or complete. All estimates, forecasts and opinions are current as at the date of this publication and are subject to change without notice. Unless otherwise indicated, Santander does not intend to update this publication. The views and commentary in this publication may not be objective or independent of the interests of the Trading and Sales functions of Santander, who may be active participants in the markets, investments or strategies referred to herein and/or may receive compensation from investment banking and non-investment banking services from entities mentioned herein. Santander may trade as principal, make a market or hold positions in instruments (or related derivatives) and/or hold financial interest in entities discussed herein. Santander may provide market commentary or trading strategies to other clients or engage in transactions which may differ from views expressed herein. Santander may have acted upon the contents of this publication prior to you having received it.

This publication is intended for the exclusive use of the recipient and must not be reproduced, redistributed or transmitted, in whole or in part, without Santander’s consent. The recipient agrees to keep confidential at all times information contained herein.