The October Data Blackout: What Restaurant Investors Can't See This Month

It’s Friday morning, day three of the shutdown, and I am looking at the place on my second screen where the September Employment Situation release should be sitting and there is nothing there. The BLS landing page has the calendar entry crossed through. No payrolls, no hourly earnings, no leisure-and-hospitality print, no diffusion index, no revisions to July and August. The release was scheduled for this morning. It is not coming this morning. It is not coming next week either. And if the shutdown stretches the way every Beltway reporter I trust is now suggesting it will — three weeks, four, possibly into November — it is not coming this month.

Here is the trade, and here is why I am writing about it on a Friday in early October instead of waiting for the dust to settle.

The thesis, plainly. Restaurant investors entered this week without the single most important data series for the sector, and they are about to enter Q3 earnings season the same way. Leisure-and-hospitality payrolls, average hourly earnings in food service, the quits rate from JOLTS, the labor-force participation drag from the Current Population Survey — every one of those series is paused, and the only major print with a statutory reason to ship in October is CPI, because the Social Security cost-of-living adjustment is a hard legal deadline. The market is going to spend the next four weeks pricing restaurant labor off vendor data, off Toast’s monthly operator survey, off ADP, off whatever first-party tape the operators themselves choose to release on their Q3 calls — and that data is not neutral. The companies releasing it are interested parties. The vendors releasing it are selling something. And the investor who treats it like a BLS substitute is going to misprice the sector.

I want to walk through what is actually missing, what the legitimate substitutes are, where the substitutes break, and how I am personally adjusting underwrite on the deals I am looking at this week. This is INTERPRETATION where I say so, and the rest is the calendar.

What is actually paused, and for how long

The BLS Current Population Survey methods note on the shutdown impact — which went up before the lights went off and is one of the few primary-source documents I am willing to cite right now — is more candid than most agency communications. The September reference week was September 7–13. Field collection for the September CPS happened. The data is on BLS servers. What is not happening is processing, review, and publication. Estimation runs require staff. Staff are furloughed. The release does not ship.

The note is explicit on a second point that matters: when staff return, the September data does not jump to the front of the queue. It enters a backlog with the October data, which has its own reference week (October 12–18) and its own collection window. If the shutdown ends in late October, the September Employment Situation will likely be merged into a combined release with October, and the typical four-to-six-week publication lag becomes seven, eight, possibly nine. The September payrolls print I was expecting today is not coming back as the September payrolls print. It is coming back, if it comes back, as a footnote in a joint release sometime in November or December.

The same is true of the JOLTS series for August, which would have shipped this week. It is also true of the September CPI release scheduled for October 15 — with one critical exception. SSA’s COLA calculation is a statutory deadline. CPI for the third quarter of 2025 is the input. BLS does not have discretion on whether to ship that one. My read — and this is INTERPRETATION — is that CPI will ship late, but it will ship. The expected slip is roughly nine days. Call it October 24 as the working anchor. Treat that date as the only piece of federal economic data restaurant investors should plan around this month.

Everything else is dark.

The four series restaurant investors actually use

Strip away the noise and there are four BLS series that drive the labor-cost line in every restaurant model I have ever read. Knowing which ones are paused is the entire game right now.

Series one: leisure-and-hospitality payrolls. This is the headline. Month-over-month change, year-over-year change, segment share of total nonfarm. It tells you whether the industry is hiring back into recovery or whether the operators are finally taking the cost program seriously. The August print had L&H at +38,000 against a three-month average of roughly +24,000. We were expecting September to clarify whether the August number was real or noise. We will not get that clarification. It is paused.

Series two: average hourly earnings for food service and drinking places. This is the wage line. It is the input to the labor-cost-per-cover model and the variable every operator on the Q3 call is going to be asked about. The August print had food-service AHE at $19.42, year-over-year growth of roughly 4.1% — a deceleration off the 4.6% running through the spring but still above the 3.2% non-supervisory aggregate. The September number was supposed to tell us whether the deceleration was holding. It is paused.

Series three: JOLTS — job openings, hires, quits — at the food-services-and-accommodation level. The August release would have shipped this week. The quits rate is the cleanest leading indicator I know for whether the operator-side labor program is working. When quits compress, the operator is getting retention right. When quits expand, the operator is paying for turnover whether the P&L shows it or not. We expected an August quits-rate print. It is paused.

Series four: CPS labor-force participation in the 16-to-24 cohort. This is the long-cycle structural read on whether the high-school-and-college labor pool that historically staffed front-of-house is showing up. It does not move month to month. But the seasonal back-to-school step — the September print — is the read on whether the fall labor pool is normalizing or compressing. That is the input that determines whether Q4 staffing is going to require a wage step. It is paused.

Four series, four blackouts. The aggregate August data — already on the books — is the latest read we are going to have on any of these for the duration of the shutdown.

The vendor substitutes, and where they break

Here is where it gets interesting, and here is where the discipline matters.

The substitutes exist. ADP’s monthly National Employment Report ships independently of BLS and includes a small-business cut and a leisure-and-hospitality slice. The Toast operator survey runs every month and includes wage tape from roughly 140,000 restaurant locations on the platform. Square publishes a quarterly Restaurant Industry Report. Homebase publishes weekly hours-worked tape. The Atlanta Fed Wage Growth Tracker draws on CPS microdata and will go dark with CPS, but it has a backfilled August read that is still usable. NFIB publishes a small-business labor sentiment survey. Indeed has a hiring-lab series. The data is there.

The data is also not BLS data, and the difference matters in three ways the buy side does not always price.

One: selection bias. Toast’s tape is the tape of operators who pay for Toast. That is a self-selecting cohort — higher-tech-adoption, generally newer concepts, generally above-average AOV. The wage tape is not representative of the universe BLS samples. It is representative of the universe Toast sells to. The same is true of Square. The same is true of Homebase. The vendor is showing you a slice that looks like its customer base, not a slice that looks like the industry.

Two: incentive bias. Toast wants you to believe restaurants are healthy and growing — that is the story Toast’s own stock is built on. Square wants the same thing for the small-business segment. ADP is the cleanest of the three because ADP’s business is payroll, not platform optionality, and ADP has run alongside BLS for two decades with documented tracking error in the 30-to-80-thousand-job range on the headline number. ADP is the best of the substitutes. It is also the one that diverges most loudly from BLS exactly when the data matters most.

Three: methodology drift. BLS uses an establishment survey for payrolls and a household survey for unemployment. The vendor data is neither. It is administrative data — payroll runs, hours-tracking logins, transaction records — and it answers a different question. Toast’s wage tape tells you what Toast operators paid their staff. BLS’s wage tape tells you what the industry paid. Those are different numbers. They are not interchangeable. The investor who substitutes one for the other and doesn’t reprice the confidence interval is going to underwrite a tighter band than the data supports.

How I am adjusting underwrite this week

I have two live deals on my desk. Both are casual-dining roll-ups with labor running 30%-to-32% of revenue. Both have models that assume hourly wages grow at 3.5% in 2026 against a sector running roughly 4%. Neither operator has the scale to pull retention numbers that beat the segment by enough to fund the gap. The labor line is the swing variable on returns.

In a normal month, I would be reading the September Employment Situation this morning, marking the wage-growth read, and deciding whether the 3.5% assumption is defensible or aggressive. I cannot do that this morning. I will not be able to do that this month. And the call I am making — INTERPRETATION — is that I am pricing in a 50-basis-point premium to the labor line on both deals until the BLS calendar is back. That is not a small adjustment. On a $40 million revenue operator running 31% labor, 50 bps of additional wage growth assumption costs roughly $200,000 of EBITDA in year one and compounds. At a 9× exit, that is $1.8 million of equity value. Per deal. The shutdown is not free.

The wider point is one I have been writing around for two years and that the forthcoming May piece on the case against the AI premium takes directly on: the asset class has spent the last three years pricing restaurant operators on a federal-data backbone that everyone treated as free infrastructure. That backbone is not free. It is a public good provided by an agency that can go dark for political reasons unrelated to the asset class itself. Buy-side discipline in October 2025 means recognizing that the model inputs you treated as zero-cost have a non-zero failure rate, and pricing the failure mode.

What to watch between now and the CPI print

Three things, in order.

First, the CPI release. If BLS hits the late-October ship — and I expect they will, because SSA forces their hand — the headline CPI and the food-away-from-home component will be the only federal economic data point of the month. Food-away-from-home has been running roughly 350 basis points above food-at-home for the trailing twelve. That spread is the single most important macro variable for the casual-dining segment because it determines whether the trade-down narrative is real or theater. Watch that spread on the print. If it compresses, the trade-down thesis the bears have been running on Darden, Brinker, and Cracker Barrel since spring is wrong. If it widens, they are right. There is no other federal data point this month that will resolve that argument.

Second, the operator Q3 calls themselves. Toast reports the first week of November, Sweetgreen the second, the casual-dining book through the middle of the month. On every call, the operator is going to be asked the same question — what is your hourly wage growth running at, year-to-date and quarter-to-date — and the answer is going to be the proxy the sell side uses for the BLS wage line. Listen for the methodology. Listen for the cohort definition. Listen for whether the operator volunteers a comparable BLS series or whether they cite their own tape. The operators who reference a BLS comp are being transparent. The operators who only reference their own tape are not. Mark the difference.

Third, the shutdown calendar itself. If the lights come back on inside three weeks, the September Employment Situation may still ship in late October or early November as a standalone release. If the shutdown stretches into mid-October, the September and October Employment Situation prints are going to be merged, and the BLS calendar for the rest of the year is going to be sequentially compressed against the holiday schedule. The probability that we get a clean read on the third-quarter labor market before Toast’s Q3 call is, in my view, low and falling. Plan accordingly.

The bottom line on the bottom line is this. The buy side spent the last decade building models on a data backbone that was reliable enough to be invisible. The backbone is visible this month. It is also dark. Mark your model uncertainty up, mark your vendor substitutes down, and do not let the Q3 earnings cycle pull you into pricing the sector off operator-volunteered tape that has not been within a hundred miles of a BLS sample frame. The premium for waiting for real data is not zero. The premium for not waiting, in a month where the data is gone and the substitutes are conflicted, is larger than the market is currently pricing.

That is the trade. Wait for CPI. Read food-away-from-home. Underwrite the labor line wider than your model wants you to. And if a banker calls you next week with a deal that needs to clear before November, ask what they know about the September wage tape that you do not.

Then call me back when they answer.

— Marcus edits The Bottom Line for TableTransfers. Tips: ma@tabletransfers.com.