13F holdings

Definition

Form 13F requires institutional managers above $100M to file their US equity long positions quarterly, within 45 days of quarter-end — the public census of professional ownership. Its uses are structural (who owns what, how crowded, how concentrated) far more than imitative: the data is stale by construction (45 + up to 90 days), longs-only (shorts and derivatives are invisible), and clone strategies inherit both limits.

How it works / structure

• The data (engine-executable): quarterly snapshots per manager (position, shares, value); derived measures — ownership breadth (number of holders), concentration (top-holder share), crowding (overlap across momentum- correlated managers — “hedge-fund hotels”), and quarter- over-quarter NEW/EXIT position flags (the trades, inferred from snapshots).
• The evidence: institutional holdings and trades carry modest predictive information in academic samples (Chen- Jegadeesh-Wermers: trades more than holdings); breadth changes correlate with subsequent returns in some samples; cloning high-conviction concentrated managers has practitioner-documented but decay-suspect results.
• The blind spots (structural): no shorts (a long-short manager’s 13F shows half the position — the visible long may hedge an invisible short), no international, no futures/swap exposure, 45-day lag, and quarter-end window dressing (documented).
• Crowding as risk data: hedge-fund-hotel names carry documented deleveraging correlation — when one crowded manager degrosses, the shared names fall together (port-correlation-budgets hidden-factor concentration; bias-herding at the professional tier).

When it applies

Crowding measurement on single names (the risk use — the platform’s primary use); ownership-structure context for liquidity events (who must sell in a downgrade/deletion); idea-flow observation (new concentrated positions by documented long-horizon managers as research prompts, never as signals); float analysis (institutional lock-up of effective float).

Risk profile & failure modes

• Staleness imitation: buying a position filed 45 days ago that the manager may have exited — clone decay’s first cause.
• Half-position illusion: cloning the visible long of a hedged structure (merger arb longs are the classic — event-mergers-acquisitions positions read as conviction longs).
• Hotel fire risk: crowded-name concentration reads as validation until the correlated unwind — the risk read and the imitation read are opposites.
• Window-dressing snapshots: quarter-end holdings are managed for optics (documented); mid-quarter reality differs.

Evidence & limits

Filing rules are SEC-documented. The academic record supports modest information in institutional trades with substantial decay and the structural blind spots above. Clone-fund marketing overstates replicability — labeled folklore- adjacent where unreplayed. The crowding-risk use rests on documented deleveraging episodes (2007 quant unwind, 2021 degrossing waves).

Falsifiable-thesis examples

Illustrations only, not signals:

• “Top-decile hedge-fund-crowding names will underperform the index in the next drawdown month by at least 2% (unwind-correlation thesis)” — falsified by the episode’s cross-section.
• “New concentrated positions (>5% of a documented long-horizon manager’s book) will outperform their sector over 12 months (idea-flow thesis)” — falsified by the cohort return.

Cross-references

• The frame: lens-sentiment; individual twin: sent-insider-transactions
• The risk use: port-correlation-budgets (crowding as hidden factor), bias-herding
• The blind-spot interactions: event-mergers-acquisitions (hedged longs), sent-short-interest (the invisible side)