Готовьтесь к еще одному масштабному пересмотру негативных платежных ведомостей, что даст зеленый свет снижению ставки на 50 б.. в сентябре

It was almost exactly one year ago that the constant stream of BS coming out of the BLS (Bureau of Labor Statistics) finally snapped.

For years, ZeroHedge had been writing that Biden’s Bureau of Labor Statistics had been massaging the jobs data to make the economy appear stronger than it was – if only to give the braindead puppet in the White House talking points with which to fill countless awkward silences – chronically revising every prior month’s data lower while overestimating the current month (until next month’s downward revision) while using every trick imaginable to cover up the deteriorating reality: fewer full-time jobs, more part-time jobs, a staggering gap between the Establishment and Household surveys, chronic overestimates of birth-death adjustments, generous boosts from seasonal adjustments, record multiple job holders as well as non-native (read illegal immigrant) workers. It all came to a head in March 2024 when we reported that „Philadelphia Fed Admits US Payrolls Overstated By At Least 800,000.”

As we wrote at the time, according to estimates from the far more accurate Quarterly Census of Employment and Wages (QCEW) report „the BLS had overstated payrolls by 800,000 through Dec 2023 (and more if one were to extend the data series into 2024). It’s truly statistically remarkable how every time the data error is in favor of a stronger, if fake, economy.„

The analysis also found that far from the stellar 230K average monthly increase in payrolls in 2023, which the White House would spin time and again as direct evidence of the benefits of Bidenomics, the true average monthly payroll increase in 2023 was only 130K! The full monthly change in payrolls as originally reported by the BLS (in green) and the actual monthly number, as per the QCEW (in red) is shown below.

And while there was a small if vocal response to the report, most promptly went on about with their lives, moving on from the latest accusation that the BLS had generously misreported the labor market data.

But five months later it all came crashing down when on August 21, the BLS confirmed everything we had said, when it unveiled in its annual CES Preliminary Benchmark Announcement, that „the preliminary estimate of the benchmark revision indicates an adjustment to March 2024 total nonfarm employment of -818,000 (-0.5 percent)” or just above the 800,000 was said to expect back in March.

How big is the 818,000 revision in context? As the chart below shows, the 2024 revision was the biggest in the past decade, and the second biggest on record, with just the 824K downward revision in 2009 just (barely) greater.

More importantly, it helped seal the jumbo rate cut that followed just 3 weeks later when on Sept 17, Powell shocked markets by cutting not 25bps as everyone had expected, but a whopping 50bps. There were various explanations for the shockingly big cut, and the best one was delivered by Rabobank’s Michael Every who correctly called out the fact that Powell was desperately pushing the strings to get Kamala elected…

Powell had a clear incentive to deliver a 50 bps cut before Election Day, because Trump has already made clear that he would not reappoint him as Fed Chair. In fact, he may decide to remove him prematurely. So Powell’s only chance of another term is by pleasing Kamala Harris and her fellow Democrats in the Senate

… but since Powell would never admit that the Fed was a political organization, he had to come up with an economic justification for the move, and that was precisely the near record revision to catch down to the QCEW that came into play. Note this exchange between Economist reporter Simon Rabinovitch and Jerome Powell during the Fed’s post jumbo cut press conference:

SIMON RABINOVITCH. Thank you, Chair Powell. Simon Rabinovitch with the Economist. You’ve mentioned how closely you’re watching the labor market, but you also noted that payroll numbers have been a little bit less reliable lately because of the big downward revisions. Does that put your focus overwhelmingly on the unemployment rate? And given the SEP projection of 4.4 [percent] basically being the peak in the cycle, would going above that be the kind of thing that would trigger another 50 basis point cut?

CHAIR POWELL. So we will continue to look at that broad array of labor market data, including the payroll numbers. We’re not discarding those. I mean, we’ll certainly look at those, but we will mentally tend to adjust them based on the QCEW adjustment, which you referred to.

Of course, Powell was looking at the near record QCEW adjustment just three weeks earlier, if for no other reason then to have a pretext for the first jumbo rate cut since the covid crash: a jumbo rate cut which would take place just two months before the election and in a time when the liberal media propaganda apparatus was praising the state of the US economy and labor market to anyone who bothered to listen.

Well, Powell is about to have to listen once again, because in just ten days, the US economy is about to be hit with another monster negative jobs revision, one which will likely tip the scales for another jumbo 50bps rate cut on September 17, when consensus for a 25bps rate cut is already in the bag.

Many may have missed it due to the excitement over the presidential transition, and the looming holidays, but on December 17, the Philly Fed once again stunned the world’s economists when it revealed the latest data fabrication out of the BLS, when it reported that in the second quarter of 2024 instead of a 1.1% increase in US jobs, the true change in US payrolls was actually down 0.1%. We detailed this finding in „Biden Lied About Everything: Philly Fed Finds All Jobs „Created” In Q2 Were Fake.”

So with that in mind it isn’t at all difficult to anticipate that the latest data from the Philadelphia Fed would expose even more data rigging and manipulation by the BLS, and even lower real job numbers. Which of course, we now know was the case… but there is much more to come.

First, recall that last month there was a shocking development when not only did the July payrolls come in far below estimates, at just 73K, the lowest since last year, but the real shocker were the massive monthly revisions which cut the previous two months by a near record amount:

• May was revised down by 125,000, from +144,000 to +19,000,
• June was revised down by 133,000, from +147,000 to +14,000.

With these revisions, employment in May and June combined was 258,000 lower than previously reported, which according to Goldman was one of the worst two-month revisions in history.

It culminated with a furious Donald Trump firing the BLS commissioner – Biden appointee Erika McEntarfer – and appointing BLS skeptic, E.J. Antoni, chief economist at the Heritage Foundation and someone who has frequently digested and amplified our labor market observations and analyses. But that is just the start.

On September 9, 2025, the Bureau of Labor Statistics (BLS) will release its preliminary annual benchmark revision for nonfarm payrolls. This will be a significant revision to previous monthly jobs reports, as it will adjust nonfarm payroll data from April 2024 through March 2025. That will be this year’s version of the shocker from August 2024, which as noted above was the trigger for Powell to do a jumbo rate cut.

The problem, as we have extensively discussed in recent years, is that there is ample justification for another massive negative revision if for no other reason than the BLS’s chronic overestimation of Birth Death adjustments, not to mention everything else that the Bureau assumes in order to put lipstick on the pig of a job market, only to quietly revise it all away. But don’t take our word for it.

In a report published on August 17 (available here to pro subscribers), Goldman economist David Mericle wrote that after the shockingly bad July jobs report, the bank’s estimate of trend job growth „is now clearly below even that low bar at 30k per month” but more importantly, „future revisions to job growth are more likely to be negative because the birth-death model is likely a bit too generous, changes in trend payroll growth can initially be partially misattributed to changes in seasonal factors, revisions to the raw payrolls data tended to be negative in past slowdowns, data from ADP raise doubts about officially reported payroll growth in healthcare, and the household survey is now overstating immigration and employment gains.”

They sound almost like us. Of the above we will drill down only on two of the points brought up by Goldman, starting with the Birth Death adjustment, a fudge factor which we have repeatedly disparaged in the past decade as a chronic source of overestimation of true labor strength.

Why 1 million jobs will be quietly removed from the payrolls? Because as noted below, 57% of all YTD jobs are statistical fakes from Birth/Death adjustment which assume the same new business vibrancy as just after covid (which was mostly to facilitate PPP fraud) https://t.co/YHvKRyYFds

— zerohedge (@zerohedge) August 20, 2024

In a separate report discussing how and why the BLS is overestimating payroll growth (available here to pro subs), Goldman explained that „job gains in the QCEW and establishment survey can differ because the QCEW offers a more comprehensive picture of net job gains from opening and closing businesses. The establishment survey does not capture employment at opening firms. As a result, the BLS excludes job losses at closing firms and imputes net job creation from business openings and closings—the birth-death model.”

To check whether the birth-death model may be leading payroll job gains to be overstated, in the next chart Goldman „update its model of net job gains based on higher-frequency business formation and bankruptcies data as well as detailed data from the business employment dynamics (BED) program.” Goldman’s model suggests that the birth-death model probably overstated payroll gains in 2024H2 — and would justify about 45k/month of the large downward revision from the QCEW — but that that overstatement has shrunk more recently, as the birth-death estimates have come down somewhat and net business formation has stabilized.

An almost identical analysis comes from Steven Englander, head of FX research at Standard Chartered, who agrees that Birth/Death is the consummate fudge factor, and recently wrote that his best estimate is that published NFP data overstates true NFP growth by about 70k per month: „Investors and policy makers who take the published numbers literally probably contribute a hawkish bias to monetary policy expectations. We think the NFP overstatement arises from the BLS birth-death adjustment, which is based on an autoregressive model, is pre-determined early in the year (ignoring subsequent labour-market conditions). It has also increasingly diverged from an accurate count of jobs created by net firm openings.”

More shocking is Englander’s discovery that if one takes published NFP data literally, continuing firms in the private sector have added only 25k jobs per month since the start of 2024, while net firm openings have added more than 100k jobs per month. This contradicts Business Employment Dynamics (BDM) data – regarded as the gold standard – suggesting that net firm openings accounted for about 20% of total job creation in 2024. Moreover, the BDM data have shown a far more dramatic drop than the NFP data in jobs created by net firm openings.

The BDM data provide an accurate calculation of jobs creation from net firm openings, and they show a drop in 2024 to less than 20% of 2022 levels! If the birth death-adjustment reflected the same drop, published NFP would be running 70k lower on a monthly basis (chart below). Englander estimates that the NFP number published by BLS would need to be around 170k to keep the underlying labor market in equilibrium. This 170k comprises 100k ‘real’ jobs (the estimate of underlying equilibrium growth driven by demographics and immigration) and 70k jobs that reflect the bias in the BLS data.

As an aside, the BDM data shown above are universally regarded as more accurate, albeit lagging (the latest available data is for 2024). They are based on the same administrative data used to produce the Quarterly Census of Employment and Wage (QCEW) data, which are used to benchmark NFP (and is what tipped us off last year about the looming 800K downward revision). BDM data are not sample-based; rather, they represent a count based on all payments into the unemployment insurance system. If BDM is saying that job creation from net firm openings has dropped, this is based on a count of employment at firms that have just started to pay into the unemployment insurance system minus those that have just stopped.

In any case, it is extremely likely that this trend has continued in 2025, as both BDM and NFP data showed a low rate of net job creation by continuing firms in 2024, and NFP data show this continuing well into 2025 (Figure 2). NFP is benchmarked to QCEW/BDM though March 2024, but data since then show the same weakness. As Goldman also notes above, it is very unlikely that net firm openings will be a major source of net new jobs if continuing firms have created only 25k jobs per month on average since the start of 2024. Job creation from net firm openings tends to correlate closely with job creation from continuing firms (Figure 3).

Of course, it’s not just the birth death model. As Goldman also highlighted, seemingly echoing everything we have said in recent years, there are several other reasons why the jobs data is likely far worse than originally reported (it is curious why the bank didn’t report these negative factors at the time but only after the dramatic negative BLS revisions to the jobs data, almost as if to goalseek their conclusion). Among these are:

• Collapse in illegal immigration: Based on Goldman’s estimates of monthly illegal immigration. the bank finds that the breakeven rate of payroll growth is 80k per month today and will continue to fall in the years ahead as the effect of lagged entry by those who came during the 2022-2024 surge fades. Furthermore, the legacy complexities of the immigration chaos left by the Biden administration make this estimate much more uncertain than usual. What share of immigrants awaiting an asylum hearing and a work permit work anyway, whether those who do are captured in the official employment data, and to what extent worksite raids, the loss of Temporary Protected Status, and other aspects of the tightening of immigration policy this year make immigrant workers less likely to work and to be captured in the official data are all very hard to know.
• Seasonal adjustments: the seasonal adjustment algorithm tends to initially misattribute part of a change in the trend pace of payroll growth to a change in the seasonal factors. When data for subsequent months confirm that the trend has indeed slowed, the algorithm reverses that change to the seasonal factors and correctly revises seasonally-adjusted payroll growth down.
• Chronic negative adjustments: There is a tendency for the raw (not seasonally adjusted) payroll data to be revised down as more information came in late when the economy slowed in the past, as Governor Cook recently noted. This occurred, for example, in every recession but one since 1979, when data on the composition of revisions begin.
• ADP: The divergent data reported by ADP raises doubts about the official BLS data on payroll growth in the healthcare industry. The healthcare industry is particularly important at the moment because it has accounted for more than all total payroll growth in each of the last three months. The BLS-ADP gap has widened to over 100k jobs per month, whereas the gap in the real-time data had been close to 50k in previous years. While the BLS numbers are more consistent with trends in healthcare spending, employment counts from large healthcare companies and views from our healthcare sector analysts suggest that the truth might be somewhere in the middle
• Household Survey: The household survey is now likely overstating population growth and consequently household employment growth. The reason is that it is using an estimate of immigration that was reasonable at the start of the year but is now likely much too high, the reverse of the problem in past years when it underestimated immigration. The assumed rate of population growth might be as much as a million per year too high, which would imply that household employment growth—which has averaged -30k per month this year on a payrolls-equivalent basis—could be overstated by around 50k per month.

Which brings us to the latest QCEW estimate. Looking at the latest Philly Fed Early benchmark revision data and applying the same analysis to the underlying excel data we did back in March 2024 when we correctly predicted the 800K drop, we find that another major negative revision is on deck.

The chart below shows three data series. The first one (in green) is the originally reported payrolls data, which however was dramatically revised back in February for the series ending December 2024 and which noted at the time, resulted in a loss of 610K jobs from the pre-revised data to the revised payrolls. The blue line is the most recent payrolls data, while the dotted blue line shows the recent negative revisions on the monthlies. This however does not represent the comprehensive revision which will be published on Sept 9 and which we are trying to estimate. Finally, the redline is the latest preliminary QCEW data reported by the Philly Fed. Like the last time we looked at it, it shows that the BLS is once again dramatically overestimating jobs data, and as the two series converge, we expect the BLS to unveil another dramatic negative revision in ten days, when on Sept 9 the Dept of Labor will announce that another 550,000 jobs never existed!

Putting it all together, we can conclude that the labor market is far weaker than it has been represented and despite recent revisions – either as part of annual wholesale or recurring month revisions – there is still substantial catch down to the true underlying data, largely as a result of flawed Birth-Death assumptions and far fewer illegal immigrants entering the labor market. Attempting to quantify the data shows that the looming negative revision is anywhere between 40K per month (as part of the QCEW catch up) or as much as 70K (as per Standard Chartered calculations), suggesting that on Sept 9 the BLS will revise the „actual” payroll number lower by 550K-800K.

The implications are huge: as one of Goldman’s top traders correctly said, all that matters now – after Powell’s dovish pivot on is jobs – is not inflation but „payrolls that will dictate Powell’s pace”, and another massive downward revision to jobs at exactly the same time as we got it last year when the same QCEW haircut was cited by Powell as a pretext for his jumbo cut, will force the Fed chair to pursue a similar jumbo cut as he did last September, if only to confirm that the 50bps rate cut last September was not political but drive by economic reasons.

And just in case the Fed Chair forgets what’s at stake, his replacement Chris Waller will be sure to remind him. Just last night Waller said that „based on the data in hand”, he did not “believe that a cut of larger than [0.25 percentage points] is needed in September” but he was quick to add that his view „could change if the employment report for August, due out a week from tomorrow, points to a substantially weakening economy and inflation remains well contained”.

Needless to say, another huge negative jobs revision – which is certainly not in the data at hand – will be just the indicator that the economy is weakening enough to warrant not a 25bps but a jumbo 50 basis point rate cut… just like it did one year ago according to Waller’s soon-to-be-former boss.

Much more in the full notes from Goldman (here and here) and Standard Chartered (here) available to pro subs.