Reading the Reg SHO Threshold List Without Assuming a Short Squeeze Is Coming

A Regulation SHO threshold flag certifies one thing: a settlement condition at a clearing agency persisted for five consecutive settlement days. It does not certify that anyone is short the security, it does not measure how large a short position is, and it does not say whether the fails came from an abusive practice or from a routine one. The Securities and Exchange Commission states the point directly on its own investor page: "The appearance of a security on a threshold list does not necessarily mean that there has been abusive 'naked' short selling or any impermissible trading in the stock."

That sentence is easy to quote and easy to ignore. So this piece does something narrower and more checkable. It counts what was actually on one SRO's threshold file across eleven consecutive sessions, and it works out which of the rule's two numeric tests is the binding one for a given issue. Both results below were computed from the primary files for this article. Neither is a figure the SEC or any exchange publishes.

What the flag on the file certifies

Nasdaq publishes a plain-text Regulation SHO threshold file each session at a fixed path, one row per security, pipe delimited. The field definitions page describes the threshold flag as applying "Where there is an aggregate fail to deliver position for five consecutive settlement days at a registered clearing agency totaling 10,000 shares or more and equal to at least 0.5% of the issuer's total shares outstanding."

The SEC's own description on its Key Points About Regulation SHO page reads the same way: "Threshold securities are equity securities that have an aggregate fail to deliver position for five consecutive settlement days at a registered clearing agency (e.g., National Securities Clearing Corporation (NSCC)); totaling 10,000 shares or more; and equal to at least 0.5% of the issuer's total shares outstanding."

Three features of that definition matter for anyone reading the list as a signal. The test is about fails at a clearing agency, not about short interest. It is a five-session persistence test, so a single day of settlement trouble does not produce a listing. And it is a two-pronged test in which one prong is an absolute share count and the other is a fraction of shares outstanding. The file itself publishes neither the fail quantity nor the share count. It publishes a symbol, a name, a market tier and a yes-or-no flag.

The two prongs cross at 2,000,000 shares outstanding

Because one prong is absolute and the other is relative, which prong is the higher hurdle depends entirely on how many shares the issue has outstanding. Setting the two prongs equal gives the crossover: 10,000 divided by 0.005 is 2,000,000. For an issue with exactly 2,000,000 shares outstanding, both prongs are satisfied at the same 10,000 shares of fails.

Below that share count the absolute prong is the constraint. An issue with 500,000 shares outstanding clears the percentage prong at 2,500 shares, but it still needs 10,000 shares of fails, which is 2 percent of the entire issue. Above the crossover the percentage prong takes over and grows without limit. An issue with 200,000,000 shares outstanding needs 1,000,000 shares of fails, sustained for five consecutive settlement days, before the flag can be set at all.

That is a hundredfold difference in the absolute quantity of failed shares required, driven by nothing but the size of the issue. It is arithmetic taken from the rule's own two figures, and it is the reason a threshold list is not a league table of settlement stress. A large issue can carry far more failed shares than a small one and still fall short of the second prong.

Eleven sessions of one SRO's file, counted here

To see what the list is made of rather than what it is assumed to be made of, the Nasdaq threshold file was retrieved for the eleven consecutive sessions from August 17 through August 31, 2026, and every row was tallied. Securities were classified by the security-name field into three groups. A row was counted as an exchange-traded product if its name contains any of ETF, TIDAL TR, GRANITESH, DIREXION, ROUNDHILL, ISHARES, GRAYSCALE, EXCH-TRD, T ROWE PRICE, FIRST TRUST or EA SER TR; as a warrant if the name contains WT as a standalone token; and as ordinary shares otherwise. That rule is stated here so the percentages below can be re-derived from the same files.

• August 17 — 82 names, of which 37 exchange-traded products, 45.1 percent
• August 18 — 85 names, of which 38 exchange-traded products, 44.7 percent
• August 19 — 82 names, of which 37 exchange-traded products, 45.1 percent
• August 20 — 81 names, of which 36 exchange-traded products, 44.4 percent
• August 21 — 87 names, of which 40 exchange-traded products, 46.0 percent
• August 24 — 87 names, of which 42 exchange-traded products, 48.3 percent
• August 25 — 81 names, of which 40 exchange-traded products, 49.4 percent
• August 26 — 83 names, of which 41 exchange-traded products, 49.4 percent
• August 27 — 80 names, of which 41 exchange-traded products, 51.3 percent
• August 28 — 76 names, of which 40 exchange-traded products, 52.6 percent
• August 31 — 72 names, of which 38 exchange-traded products, 52.8 percent

The eleven files contain 896 symbol-days and 126 distinct symbols. The daily count averaged 81.5 and stayed inside a narrow band, 72 at the low end and 87 at the high end. Exchange-traded products accounted for 53 of the 126 distinct names, 42.1 percent, and 430 of the 896 symbol-days, 48.0 percent. Warrants accounted for 5 distinct names and 35 symbol-days. That leaves 68 distinct names of ordinary shares.

Persistence is where the file stops looking like a stress alarm. Of the 126 distinct symbols, 44 appeared on all eleven sessions, 34.9 percent of the roster. The median symbol appeared on 7 of the 11 sessions. Only 7 symbols appeared exactly once. Day-over-day turnover was small in both directions: additions ranged from 0 to 11 names and removals from 3 to 8, with a carried-over core that stayed at 69 names or more.

Split by type, the gap widens. The median exchange-traded product on the roster appeared on all 11 sessions. The median name that is not an exchange-traded product appeared on 6. Thirty of the 44 permanent residents were exchange-traded products. And the composition drifted over the window, from 45.1 percent exchange-traded products on the first session to 52.8 percent on the last, while the overall count fell from 82 to 72.

What the composition does not prove

The tempting inference from a roster that is close to half exchange-traded products is that those products are being naked shorted. The file cannot support that inference, and neither can this count. Three things are missing.

First, the file publishes no quantities. It gives a symbol, a name, a market tier and a flag. Whether a given listing reflects 10,001 failed shares or ten million is not in the data. Second, the file publishes no shares outstanding, which is the denominator of the second prong. For funds that create and redeem shares, that denominator moves. Without it, no reader can tell which prong actually bound for any specific name. Third, the file publishes no cause. A fail is a delivery that did not happen by the settlement deadline. The rule does not ask why.

One structural asymmetry does follow from the rule's own arithmetic. The percentage prong scales with shares outstanding, so the absolute number of failed shares needed to reach 0.5 percent is smaller for an issue with fewer shares outstanding than for one with more. That is a property of the test, not a finding about any issuer. This count did not measure shares outstanding for any name on the file, so it cannot say which prong bound for any listing, and it is not evidence about intent.

The one distributional fact that survives the caveats is the tier split. Of the 126 distinct symbols, 4 were in the Nasdaq Global Select tier, and those 4 accounted for 22 of the 896 symbol-days, 2.5 percent. The rest sat in the Global Market and Capital Market tiers. The market-category field records which Nasdaq tier a security is listed on. This count did not measure market capitalisation, so the split describes listing tier and nothing else.

The thirteen-day clock is not the clock that binds

The most repeated fact about threshold securities is a thirteen-day mandatory close-out. It exists, and SEC staff say it is generally not reached. The SEC's Division of Trading and Markets addresses this directly in its responses to frequently asked questions: "Although as a result of compliance with Rule 204, generally fail to deliver positions will not remain for 13 consecutive settlement days, if, for whatever reason, a participant of a registered clearing agency has a fail to deliver position at a registered clearing agency in a threshold security for 13 consecutive settlement days, the requirement to close-out such position under Rule 203(b)(3) remains in effect."

The clause that gets quoted is the tail. The clause that governs is the head. Rule 204 sets a much shorter default, and the SEC's small entity compliance guide for the 2009 amendments describes the obligation as being to "immediately purchase or borrow securities to close out the fail to deliver position by no later than the beginning of regular trading hours on the settlement date following the day the participant incurred the fail to deliver position, unless an exception applies." The same guide notes that "Rule 204 makes permanent the amendments contained in temporary Rule 204T, with limited modifications."

The exceptions are where the reading has to be careful, because they are not one deadline but three, and each is narrower than the shorthand it usually travels under. The first applies where "the participant can demonstrate on its books and records that such fail to deliver position resulted from a long sale", which is a documentary condition and not merely a characterisation of the trade; that fail runs to the beginning of regular trading hours on the third consecutive settlement day following the settlement date. The second applies to a fail "that is attributable to bona fide market making activities by a registered market maker, options market maker, or other market maker obligated to quote in the over-the-counter market", three categories rather than one, and carries the same third-day deadline. The third applies to a fail from the sale of a security the seller is deemed to own under Rule 200, where the seller intends to deliver once restrictions are removed, and runs to the thirty-fifth consecutive calendar day following the trade date.

A reader who takes only the thirteen-day figure away from the rule is holding the deadline SEC staff describe as generally not reached. Which of the Rule 204 deadlines governs a particular fail depends on which exception, if any, that fail falls under, and the threshold file reports none of that.

What Would Invalidate This

The count above covers one self-regulatory organisation's file over eleven sessions in one month. It is not the whole market, and several conditions would change the reading.

• A different SRO's file would show a different roster. The SEC staff guidance states that "Staff understands that the SROs disseminate threshold lists that contain securities that are listed on their market systems and that exceed the specified fail level for at least five consecutive settlement days." The same page adds that "A threshold security is expected to appear on one list." A screenshot from one exchange is therefore a slice, not a census.
• Eleven sessions is a short window. Composition drifted noticeably inside it. A quarter-long sample could show a different product mix, and the ordinary-share share of the roster could be higher in a month with more corporate actions.
• The classification is by security name, not by a regulatory taxonomy. Names in the file are truncated to a fixed width. A fund whose truncated name lacks an identifiable sponsor or structure token would be classified as ordinary shares here.
• The persistence figures depend on the window boundaries. A name that was flagged before August 17 or stayed flagged after August 31 is counted only for the sessions inside the window, so the eleven-session residents are a lower bound on true persistence.
• If fail quantities and shares outstanding were added, the composition argument could reverse. The fails-to-deliver data the SEC publishes separately carries quantities; this file does not, and no conclusion here should be extended to quantities.

Concrete Framework

A workable way to use the file without over-reading it, in order.

1. Read the flag as a persistence statement, not a size statement. The flag means a settlement condition held for five consecutive settlement days. It carries no quantity and no direction.
2. Check which prong plausibly binds before drawing any inference. Compare the issue's shares outstanding against the 2,000,000 crossover. Under it, the absolute prong is the hurdle. Over it, the percentage prong is, and the required fail quantity rises with the share count.
3. Pull the file yourself for the sessions in question, not a screenshot. The daily file is a fixed path with the date in the filename, and it carries the security-name and market-category fields alongside the flag.
4. Check the market-category field before treating a listing as a broad-market event. In this window the Global Select tier supplied 4 of the 126 distinct names.
5. Count the sessions the name has been present, not just today's presence. A name that has been on the file every session for weeks is a different object from one that appeared yesterday, and the file gives that history for free once it is archived daily.
6. Match the deadline to the exception, not to the headline. Ask which close-out clock applies before assuming pressure is building, and remember that SEC staff say fails generally do not remain for thirteen consecutive settlement days.
7. Treat the roster as a starting point for questions about settlement, not as evidence about positioning. Short interest, borrow rates and options positioning are separate datasets with separate publication schedules, and the threshold file does not stand in for any of them.

Two numbers in this article were produced here rather than published by any institution: the 2,000,000-share crossover between the rule's two prongs, and the eleven-session composition count of 126 distinct symbols across 896 symbol-days. The crossover can be re-derived from the two figures in the rule, and the census from the eleven daily files using the name-field rule set out above. Both should be re-derived rather than trusted if the reading matters to a decision.

This article is analysis of public regulatory data and rule text. It is not investment advice and not a recommendation to buy or sell any security.