Methodology — 13F Common-Ownership Pair List + Crowding

Transparency is a feature. This page explains what the common-ownership pair list shows and why — in plain terms. The precise formulas and numeric conventions are documented in our internal methodology; the summary here is deliberate, not an omission.

Two stocks can be linked in a way no price chart shows: the same investors own big pieces of both. If those holders ever need to sell one, the other tends to feel it. For every pair of your book's US-listed names, this panel measures how much of the two companies is held in common by actively managed 13F filers — the classic "connected stocks" measure (Antón & Polk, Journal of Finance, 2014).

What the number is

For each pair we sum, across every screened active institution that holds both names, its combined position in the two companies, and divide by the two companies' combined market value — all measured as of the same 13F report quarter: filed share counts × that quarter-end's prices. We never mix a filed dollar value with today's market cap; mixed vintages make wrong numbers.

Share classes of one issuer are combined (GOOGL + GOOG is one Alphabet), and a name whose quarter-end market value we cannot verify shows "unmeasured" with the reason — never a partial or guessed figure.

Who counts as "active"

Index funds hold (nearly) everything by construction, so their presence in two names says nothing about shared conviction. Before measuring, we exclude:

named index complexes — a versioned list (Vanguard, BlackRock, State Street, Geode, and peers), its version shown on the tile;
quasi-indexers — filers that trade almost nothing and own almost everything, screened by portfolio turnover computed from the filings' own implied prices;
gap filers — institutions without a consecutive prior-quarter filing (their turnover is unknowable, so they are excluded, not assumed passive);
duplicate filings of the same book under two registrants.

Concentrated, low-turnover managers (the Berkshires) are exactly who this measure is for — low turnover alone never excludes anyone.

How to read the colors

Raw common-ownership levels rise mechanically with company size, so we never color the raw number. Each pair is ranked against random pairs of similarly-sized US companies measured the same quarter under the same screen; the percentile is what you see. The flag to look for is a pair with a high ownership percentile and low price correlation — a shared-holder linkage the correlation matrix alone would miss.

Crowding & exit liquidity (the same panel)

Three additional layers, all built on the same screened-ownership data:

Days of volume per holder. In each pair's drill-down, every named common holder's position is also shown in days of that name's average daily trading volume — how long a full exit would take at 100% of typical volume. Volume comes from FINRA's consolidated short-interest file (the settlement date is shown); it is the only volume source in our stores that is measured consistently across stock splits. The fire drill applies the same arithmetic to your book's largest common holder: if it sold everything, how many days of selling pressure would each of your names absorb? Clearly labeled a hypothetical — arithmetic, not a prediction.
Short-crowded pairs. A pair is flagged when both names sit in the top tenth of days-to-cover (short interest ÷ average daily volume) among served US companies worth $500M or more. Where FINRA caps the reported figure we recompute it from the underlying counts; a name whose figure cannot be measured says so — it is never treated as calm.
Who else owns your book. The screened active filers whose disclosed portfolios most resemble yours (a similarity score over the same active-filer pool, with the index complexes and hyper-broad books removed). Each row shows its percentile within that pool — a raw similarity list would just rank the broadest funds.

Honest limits

13Fs disclose long US-listed positions only — shorts, most derivatives, bonds and non-US listings are invisible. Foreign lines read "not covered", never falsely calm.
Positions are as of the stated quarter end; filings arrive up to 45 days later and our bulk ingest refreshes quarterly, so the worst case on our cadence is about 165 days stale. The tile states the actual quarter. Volume figures carry their own (bimonthly) settlement date — the two vintages are both shown.
Days-of-volume figures assume 100% of typical daily volume; real liquidations stretch longer at realistic participation rates.
This is a statement of ownership and crowding facts about your book — not a prediction of returns.