"Substantial doubt" is easy to grep and easy to misread: notes from checking every 10-Q for going-concern warnings

Every quarter, US companies have to decide whether there is "substantial doubt about an entity's ability to continue as a going concern" over the next year, and if there is, say so in their 10-Q. It's one of the few places where a company tells you in its own words that it might run out of money.

I recently counted how many companies say it, using the latest 10-Q filed since July 1 for 4,899 companies. The headline: one in six Nasdaq operating companies (382 of 2,281) carries the warning, against 56 of 1,670 on the NYSE and NYSE American. Getting to those numbers took a lot more than a full-text search. Here's what tripped me up.

1. The phrase shows up in sentences that say the opposite

Searching for "substantial doubt" finds the warnings, but also:

• Alleviated doubt. "These conditions raise substantial doubt... management's plans alleviate the substantial doubt." That's not a warning. 107 operating companies wrote some version of it.
• Conditional risk factors. "If we are unable to raise additional capital when needed, that could raise substantial doubt" describes a hypothetical. Damora Therapeutics, worth $1.3 billion, uses exactly that sentence while saying its cash funds operations into the second half of 2029.
• The negative. A plain sufficiency statement ("sufficient to fund operations for at least 12 months from the issuance date") or "substantial doubt does not exist". Hundreds of reports discussed going concern like this without a warning.
• Someone else's doubt. National Health Investors, a healthcare REIT, wrote that a tenant has been on a cash basis since 2022, when it "received financial information from Bickford raising substantial doubt about its ability to continue as a going concern". The REIT itself carries no warning.

Spyre Therapeutics makes a good test case: "If we are unable to raise additional capital when needed, we will not be able to continue as a going concern." Strong words, but still conditional.

So I sorted each report into five buckets: warning, alleviated, discussion without a warning, nothing found, and unclear. The 42 unclear reports were left out rather than guessed. Then I read samples in both directions: 42 of 42 random warnings stated the doubt plainly, and none of 20 discussion-only reports or 11 reports with no going-concern language turned out to hide one when I read them in full.

2. Blank-check companies swamp the count

SPACs (SIC code 6770) must complete a merger by a deadline or return the trust money, and the cash they hold outside the trust is small, so most carry the warning once the deadline is within a year: 207 of 301 in my data. That says little about shareholder risk, because the money sits in the trust. Leaving them in would add a couple of hundred warnings that mean something else entirely, so I counted them separately.

3. The SEC's exchange file merges two exchanges

The exchange for each company came from the SEC's company_tickers_exchange.json. It's the best free mapping I know of, but its "NYSE" label also covers NYSE American: Cheniere Energy Partners, Uranium Energy and United States Antimony, all NYSE American listings, show up as "NYSE". For a Nasdaq vs NYSE comparison that matters, since NYSE American is home to a lot of small companies. I labeled the group "NYSE and NYSE American" rather than pretend I could split it.

The same file is a good completeness check. 532 companies that filed a 10-Q weren't in my data, and 501 of them had no ticker in that file at all (debt-only issuers, non-traded funds). Worth checking before you call a dataset "every filer".

4. Prices for OTC names can be years stale

For the size breakdown I multiplied shares outstanding by the latest close. For one OTC company, Creatd, the latest close in my price table was $1,850, from September 2022, which made it a $6 billion company on paper. Rather than patch names one at a time, I limited the size analysis to exchange-listed companies and used the October 26 close.

Share classes need the same care as in any market-cap calculation: an Up-C company's total share count can be more than double its listed Class A count.

5. "No later report" doesn't mean "dropped the warning"

To see whether warnings persist, I followed the 780 companies that warned in a 10-Q filed from January to May. Comparing each with its next 10-Q loses everyone who doesn't have one. The EDGAR quarterly form index fills the gap:

• 46 filed a 10-K instead, because their fiscal year ended in between.
• 15 filed only an NT 10-Q or NT 10-K, the notice that a report will be late.
• 11 filed a Form 15 to stop reporting altogether.

Of the 671 with a later 10-Q I could read, 632 still carried the warning.

6. A filing is a snapshot

Summit Therapeutics, worth $11.6 billion, wrote the warning in its report for the quarter to June 30. Two months later, AstraZeneca agreed to buy $2.0 billion of its convertible preferred stock. The warning was true when it was written and may well be gone from the next report. If you build anything on going-concern status, store the filing date with it and check the 8-Ks filed since.

Where it ended up

The exchange comparison, size and industry breakdowns and the follow-up are in the study, with accession numbers for every company named:

https://equibles.com/research/one-in-six-nasdaq-companies-doubts-it-can-last-a-year

If you've worked with going-concern data: do you classify from the 10-Q text, from the auditor's opinion in the 10-K, or from a vendor flag, and how do you handle the conditional sentences?

Data: SEC 10-Q filings and EDGAR indexes, processed by Equibles.