The Atlanta Fed's nowcast printed the same 1.3% it reset to in April, and Thursday's update, not Tuesday's CPI, is the number that settles the growth question.
The Atlanta Fed's GDPNow model printed 1.3% on July 8, down from the 3.0% vintage it carried on June 17. A growth nowcast shedding more than half its estimate in three weeks is normally the start of a repricing.
This one has a problem. The model also printed exactly 1.3% on April 9, the value it reset to at the top of the previous quarterly estimation window. Same figure, to the decimal.
That coincidence proves nothing on its own. GDPNow is thin at the start of a quarter by construction: it opens with a model-based prior and then absorbs monthly source data as it lands, so its earliest prints carry almost no information and swing violently once real numbers arrive. What the coincidence does is force a question with money attached, namely whether the growth leg of this expansion is genuinely breaking or a model cleared its memory and got quoted as news.
Every other growth-sensitive series on the screen argues for the second.
The hard data refuses to cooperate
Initial claims came in at 215,000 for the week ending July 4, down from 230,000 a month earlier, a 6.5% decline over 30 days. Unemployment held at 4.2% in June. Firms staring at a demand collapse do not sit on their hands: payroll is the fastest cost to cut and the first thing to surface in the weekly claims data. A month of falling claims is not what the front edge of a 1.3% economy looks like.
Credit agrees, and credit is the market that gets paid to be paranoid about growth. High yield spreads sat at 2.7 on July 9, in from 2.78 a month earlier, a tightening of 8 basis points. The Chicago Fed's financial conditions index read -0.515 on July 3, looser than the -0.495 of a month before. Lenders own the downside and have no upside to compensate them, which is why spreads usually widen before equity indices notice anything. Spreads that tightened straight through the nowcast's collapse are spreads that never saw the collapse.
Get analysis like this delivered daily. No account required.
This analysis was produced by the Convex Research Desk from live economic data and is for informational purposes only. It does not constitute financial, investment, or legal advice. See our editorial standards and terms of service.
Convex combines institutional macro research with AI-powered signal generation. Access live trading signals, portfolio analytics, and strategy backtests.
Copper, the most reflexively growth-sensitive thing on the board, is at 6.282, effectively unchanged over the last 24 hours. The dollar index at 100.965 sits mid-range in its 99-103 band, doing nothing that resembles a flight to safety.
The front end is the tell
The cleanest evidence is in Treasuries, and it runs directly against the nowcast.
Two-year yields are at 4.16%, up 11 basis points over 30 days. Ten-year notes closed at 4.54% on July 9, up 9bp on the month, and have been quoted as high as 4.569% since. That leaves the 2s10sspread at 0.35, flattened from 0.40.
Read the sequence carefully, because the shape matters more than the levels. In a real growth scare the front end moves first and moves down: the market pulls Fed cuts forward, the 2Y rallies hard, and the curve bull-steepens. None of that has happened. Instead the front end sold off, cut expectations got pushed further out, and the effective fed funds rate has been flat at 3.62% all month. Rates traders spent the last 30 days getting more hawkish, and they kept doing it after July 8.
Real yields say the same thing from the other direction. Ten-year TIPS at 2.31% are up 15bp in a month, which more than accounts for the entire nominal selloff, since the 10-year breakeven actually fell 5bp to 2.24% over the same window. Rising real yields alongside a collapsing growth outlook is an unstable pairing. One of the two is wrong, and it is rarely the one with a fresh weekly print behind it.
Why the print still does damage
An artifact is not the same thing as harmless.
The July 8 read is the first genuinely bond-bullish data point to appear in weeks, and it arrives with scenario weights already split evenly: a reflation soft landing at 35%, stagflation emergence at 35%, a Hormuz energy shock at 15%, a growth scare at 15%. A validated nowcast break is the single event that lights up the stagflation and growth-scare branches at once, and it would be doing so against realized CPI of 4.25% year on year as of May, with the June print due Tuesday.
That is the branch with no good policy answer. Cutting into 4.25% realized inflation risks un-anchoring a breakeven already sitting at 2.24%, far below what is actually being printed. Holding tightens real policy into a decelerating economy. Neither lever works, which is why the tail worth respecting here is stagflation rather than recession: a recession the Fed can cut into, and a supply-constrained slowdown it cannot.
WTI at 71.41 and Brent at 76.01 keep that constraint alive. FRED's WTI series fell 24.0% over the 30 days to July 6, from 91.58 on June 11 to 69.60, and that decline was the crutch under the market's disinflation case. Crude now trades roughly 2.6% above that July 6 close rather than below it. Prices that merely hold these levels remove the disinflationary impulse instead of extending it, and a growth slowdown arriving with the energy premium rebuilding is not the kind that rescues bondholders.
The strongest case against this reading
The counter is straightforward, and it is why this stays a live question rather than a settled one.
Claims measure layoffs. A demand-led slowdown does not begin with layoffs. It begins with households buying less, and firms respond in a well-worn order: hours first, then hiring, then heads. By the time claims move, the spending data has been deteriorating for a month or two. GDPNow, meanwhile, is assembled from precisely the monthly series that capture spending before the labor market notices, including retail sales, trade, inventories and construction. Claims cannot see what the nowcast sees. That is the whole point of a nowcast.
Which makes Thursday, July 16, the date that actually matters this week. Advance retail sales for June and the next GDPNow update land the same day, with the weekly claims report alongside them. A retail sales miss with the nowcast confirmed near 1.3% is not two independent data points, it is one story told twice, and the labor market's good month becomes the lagging indicator it always was. Claims back above 240,000 would end the argument on the spot.
So the position is conditional, not triumphant. The weight of evidence, weekly claims, tightening credit and a bear-flattening curve, says the July 8 print was a model resetting rather than an economy cracking. But the evidence that would overturn it is scheduled, dated, and four days out.
The mispricing is in the calendar, not the level
What is striking about this setup is not that the market disbelieves the nowcast. It should disbelieve the nowcast, on the evidence available. The oddity is that the market has priced neither outcome.
The S&P sits at 7,549.5, grinding higher on tightening credit. The VIX is near 15 going into a week that contains June CPI on Tuesday, PCE on Wednesday, retail sales, claims and the GDPNow arbiter on Thursday, and consumer confidence and ISM on Friday. Gold at 4,113.7 fell 0.4% over the last session and sits far below the 4,500 to 5,000 zone the stagflation scenario maps for it. Half the scenario weight in this book, stagflation at 35% and an energy shock at 15%, lives in paths that a 15 handle on volatility does not pay for.
Tuesday's inflation print will get the attention, because inflation is the argument everyone has already agreed to have. The number that determines whether this is a reflation with an inflation problem or a stagflation with two of them lands on Thursday, and it comes from a model that may well have spent the past week saying nothing at all.