Daily Points

ON DECK FOR FRIDAY, OCTOBER 4

KEY POINTS:

• Markets await nonfarm payrolls as they lose confidence in big Fed cuts
• Oil continues to surge on a Biden slip
• Nonfarm payrolls preview amid the hubris and statistical noise…
• ...including the potentially key role of seasonal adjustment factors…
• ...and why advance job market readings are of little use to nonfarm forecasts
• Markets are better balanced on 25 or 50bps Fed cut pricing into payrolls

Nonfarm payrolls (8:30amET) will dominate attention with the only other possible candidate for disruptive effects being tensions in the Middle East. There were no material overnight calendar-based developments. The suspension of the US port strike to January is welcomed, but at a steep price given a 62% wage gain over six years that will be passed onto consumers while these Luddites block automation at US ports relative to elsewhere.

Oil spiked by 5%+ yesterday and is up another 1% this morning because President Biden forgot his words. Not kidding. When asked if the US would stand with Israel’s expected retaliation against Iran if oil facilities were attacked, he said, and I quote, “We’re discussing that. I think that would be a little—anyways.” Markets took that as a sign that such an attack is even more likely and perhaps that the US would play a role.

FOMC PRICING IS MORE BALANCED GOING INTO PAYROLLS

Markets are now more balanced around payroll ties to FOMC expectations than they were just a few days ago which could make us more neutral toward what happens post-release going into it. Pricing for cumulative rate cuts over the remaining two meetings this year has been cut from 80bps to 66bps since September 25th, thereby reining in a significant amount of the back-to-back 50bps moves. Data and Fed-speak have played a role in this repricing. It seems unlikely that the Committee will pivot toward another 50bps move.

NONFARM PAYROLLS PREVIEW

Here are expectations for nonfarm payrolls:

• Consensus median: +150k
• Consensus average: 144k (only slight skewness)
• Scotia: +165k (I’m ranked 8th of 71 forecasters for payrolls fwiw).
• Range: +70k to +220k. Most within 100–175k.
• Whisper #: 151k (enter yours at WHIS <go>!)
• Revisions: no strong guidance, we’re in between benchmarking exercises
• 90% confidence interval: +/-130k (‘uuuuge…)
• Std dev: 28k
• UR: 4.2% consensus and Scotia
• Wages: 0.3% m/m SA consensus and Scotia. Wages were up 4.9% m/m SAAR in August which remains hot.
• Aggregate hours worked: Downside risk? They jumped by 0.4% m/m SA in August for the strongest gain since March and may mean revert lower. Still, US GDP growth in Q3 needs productivity since aggregate hours worked are only tracking a rise of 0.5% q/q SAAR.
• Obviously no port strike effect should be present across any of the variables since the reference period for NFP is the pay period including September 15th and for the household survey it’s the week of September 15th to the 21st.
• On wages, I would not be surprised to see cooler figures closer to 3% m/m SAAR from about 5% the prior month.

SEASONAL FACTORS COULD BE KEY

September is the first month in a while in which the seasonal adjustment factor turns more favourable. That means less of a haircut imposed on the seasonally unadjusted numbers than has been the case in several months. SA factors for like months of September in recent years have been more favourable than prior years which is the reverse of what had been the case when SA factors for several months before now were among the weakest compared to like months in history.

Chart 1 shows how SA factors for September payrolls moved higher post-pandemic.

Chart 2 makes a similar point by illustrating how all of the highest SA factors for September have been for more recent years. This is sharply different than recent months when SA factors haircut jobs more than for like months in history. For instance, this August had one of the least favourable SA factors for like months of August on record.

Chart 3 illustrates why this matters by plotting various possible outcomes for the seasonally adjustment change in nonfarm payrolls using different SA factors applied to an assumed half million seasonally unadjusted September gain which wouldn’t be unusual for the month. If this September’s SA factor is like recent years’ estimates for that month, then payrolls could come on fairly strong.

Chart 4 is another way of looking at it. If the data is seasonally adjusted, how come it is recently showing a steady pattern of bigger m/m SA gains earlier in the year, lower SA gains over summer, and then picking up somewhat again thereafter? I think issues with SA factors are the culprit.

Nonfarm isn't the only indicator having this issue. The Conference Board's 'jobs plentiful' gauge is another example among others (chart 5). It's supposed to be SA, yet in recent years it is showing a definite seasonal pattern of higher readings earlier in the year that subsequently wane.

DON’T TRUST OTHER JOB MARKET READINGS AS PREDICTORS

Here’s a rundown of some of the other advance job market signals.

• Consumer confidence jobs plentiful weakened. It’s not a great gauge (chart 6). 

• ISM gauges pointed to an outright loss of jobs in both the manufacturing and non-manufacturing gauges but have weak track records. Chart 7 shows that you wouldn’t take the correlations to the bank by any means.

• NFIB hiring plans weakened.
• NFIB ‘hard to fill’ jobs increased.
• ADP was up 143k in September. It’s not a great guide at all but based on past relationships this would suggest <10% chance of a meaningfully negative change in nonfarm payrolls. Chart 8 shows the historical spread between initial changes in ADP private payrolls and initial estimates for changes in nonfarm payrolls. One way to use the chart is to point to the far right’s columns depicting a less than 10% chance out of all of the frequency of observations shown in the chart of getting a materially negative nonfarm payrolls print.

• Challenger job cuts were 73k NSA which is similar to last September but among the higher readings compared to historical like months of September yet still compatible with job growth.
• Initial jobless claims were little changed between NFP reference periods in August and September.

Why do these other measures serve little purpose in estimating changes to nonfarm payrolls?

• Sampling error. Again, the 90% confidence interval for nonfarm payrolls is +/-130k. It's a monthly sample of about 120k businesses and government agencies. Spin the wheel!
• Measurement differences. Nonfarm counts jobs whereas the household survey counts employed bodies. Therefore, nonfarm can be heavily influenced by multiple job holders whereas the household survey only counts an employed person once.
• nonfarm by definition is a payrolls report. That means it excludes off payroll positions like many small businesses. That's another reason for not lining up well with other advance indicators.
• nonfarm is a snapshot of employment in a particular pay period that includes the 12th day of the month. If paid hourly or daily, it's measuring jobs that day, if weekly, that week, if bi-weekly, then the two weeks around that period.

And if you think nonfarm payrolls are noisy, then consider the household survey from which the other labour force metrics are derived like the change in the size of the labour force, the UR, hours worked etc. The household survey's 90% confidence interval around m/m changes in reported employment is +/-600,000. Yes, 600k.

All of which, imo, should never make one overly confident toward estimating US payrolls and related measures. The statistical noise factor alone is massive. The birth-death model adjustments to payrolls are often controversial. It's more about the clean-up in the aftermath of whatever the BLS spits out with its estimates than the hubris that characterizes the various street shops' estimates.