The Disclosure Lag Problem: Reading Stale Positioning Data
Every positioning disclosure has two clocks — when the position changed and when you learned of it — and the damage the gap does depends on who filed, not just how late.
Sophisticated investors know positioning data arrives late — everyone can recite that a quarterly holdings filing may be weeks old on publication. The error is what they do with that knowledge: they treat the lag as a uniform tax, a fixed discount applied evenly to whatever the filing says. It is not uniform. The lag destroys some kinds of information completely, leaves other kinds nearly intact, and how much it destroys depends less on the filing than on the filer. Reading stale positioning data well is not about knowing it is stale; it is about knowing which parts of it are still true.
Start with the frame. Every disclosure runs on two clocks: the position clock, which records when the holder actually transacted, and the publication clock, which records when you were allowed to know. The regime sets the maximum gap between them, and the gaps vary enormously by filing type: corporate insider transactions must generally be reported within two business days, while quarterly institutional holdings can arrive up to forty-five days after the quarter ends — meaning a position established early in that quarter can be over four months old when you first see it. Congressional disclosures allow up to forty-five days from the trade. Exchange-published short interest is compiled on a twice-monthly settlement cycle and released with a further processing delay. Fund flow data aggregates on weekly and monthly cycles. The naive fix is to rank sources by lag and prefer the fast ones. The better fix is to recognize that lag alone does not determine staleness.
Effective staleness: lag times turnover
Here is the framework phrase we use on the desk: effective staleness equals disclosure lag times holder turnover. It is a relationship, not an equation you compute to decimals — but it reorders the entire dataset. A concentrated owner who holds positions for years and files quarterly is, in practical terms, barely stale at all: the odds that the position you are reading still exists, at roughly the size disclosed, are high, because the filer's behavior changes on a clock far slower than the disclosure cycle. A high-turnover trading firm filing the same form on the same date is publishing an archaeology report: the portfolio it describes was dead before the ink dried, and treating its line items as positions is a category error. The same forty-five days; opposite information content.
What the archive never contained
Staleness is the visible flaw; incompleteness is the invisible one, and it is more dangerous precisely because a stale filing at least announces its date while an incomplete one announces nothing. The standard quarterly holdings regime captures long positions in covered securities — and omits short positions, most derivative exposure, and anything held through instruments outside the reporting requirement. A manager can be economically short a name that their filing shows them long, once options and swaps are netted. A manager can carry enormous exposure that appears in no filing at all, because swap-based exposure has historically sat outside the holdings-disclosure perimeter altogether.
The illustration: when the biggest position was the missing one
The 2021 collapse of a large family office, Archegos, is the cleanest modern teaching case, stated here as history. The fund had accumulated enormous, leveraged, concentrated exposure to a handful of names — and because the exposure was held through total return swaps with its prime brokers rather than through the underlying shares, it appeared in no public holdings filing in proportion to its true size. Market participants reading the disclosure record saw the visible shareholder register; the position that would determine those stocks' path — a single holder whose forced unwind would cascade through every name at once — was structurally invisible until it was collapsing. The episode generalizes beyond its specifics: the disclosure regime is a perimeter, and the most consequential positioning is often built deliberately at or beyond its edge, because size and invisibility are jointly valuable to the holder. A positioning picture assembled only from filings is therefore biased in a known direction — it understates leverage, understates shorts, and understates exactly the concentrated exposures most capable of producing discontinuous price action.
The strongest case against this framework
The efficient-markets objection runs: filings are public the moment they publish, armies of quantitative readers parse them within seconds, and therefore whatever value stale positioning data contains is in the price before a human finishes downloading the file. On this view the entire category is a solved problem — the lag does not merely degrade the data, the market's processing of it eliminates the residue, and time spent on positioning analysis is time spent competing with machines on their home field.
The objection is correct about one layer and wrong about the others, and the layer structure is again the answer. The event content of a filing — the surprise, the new stake, the exited position — is indeed arbitraged on publication, and we would not encourage anyone to trade the announcement. But the structural content is not an event and is not arbitraged, because it is not a mispricing: knowing that a holding's register is concentrated among leveraged fast-money holders rather than patient institutions does not tell you the price is wrong, it tells you how the name will behave under stress. That knowledge pays at a different time than it is acquired — in drawdowns, in liquidity events, in deciding position size — and payoffs displaced in time from their information source are precisely the ones fast money does not compete for. The machines read the filing faster; they are not using it for the same purpose. A second objection — that turnover-adjusting is unreliable because filers change behavior — earns a narrower answer: regime changes in a filer's turnover are themselves visible in the filing history, with lag, and a framework built on calibrated discounts fails gracefully where a framework built on taking filings at face value fails catastrophically.
How to apply this framework
- Stamp both clocks on every disclosure before reading it. Establish the maximum age of the position, not just the age of the filing — a quarterly holding can be over four months old on publication day — and let that bound which claims the document can still support.
- Discount by the filer, not the form. Estimate each filer's turnover from their own filing history and apply the product rule: lag times turnover. A slow-moving concentrated owner's filing is read as a position; a fast trader's filing is read as a historical document.
- Consume the structure layer, discard the timing layer. Ownership concentration, holder-type mix, and crowding survive the lag; entry prices and this-moment intent do not. If a use of positioning data requires knowing what a holder is doing now, the data cannot support the use.
- Correct for the perimeter. Filings understate shorts, derivatives, and swap-held exposure by construction — so treat any filings-only positioning picture as a floor on leverage and concentration, never a ceiling, especially in names showing stress that visible ownership cannot explain.