Why Was a Stock Removed From an Index? The Rule Checks If It Is a Business

An index screen can delete a stock without ever forming an opinion about it.

Ask why a stock was removed from an index and the intuitive answer is performance: it shrank, it stopped trading, something went wrong inside it. On 14 August MSCI opened a consultation that answers the question a different way. The proposal adds a non-operating company screen to its Global Investable Market Indexes, the family sitting underneath an enormous amount of the world's passive money. Run as a simulation against May 2026 data, it deletes Strategy from MSCI ACWI IMI. It deletes Metaplanet too, and a uranium holding company called Yellow Cake. Strategy's free-float market capitalisation in that simulation was $23.9 billion, the largest name the screen catches.

Nothing in the rule mentions bitcoin. Nothing in it mentions uranium, or any asset at all.

This walkthrough follows Eunha, the Kodex interpreter who lives between structure and emotion and gets at things by asking the question underneath the one being asked. She works through the eligibility test itself: what it measures, why a good quarter trips it, who is obliged to sell afterwards, and how large that selling actually turns out to be.

"The names are the output," Eunha says, skipping past the list to the methodology. "The test is the part you can use again next year, on something that has not happened yet."

The screen never asks what the company owns

Eunha runs a finger down the methodology and stops on the shape of it. The proposal is not one test. It is a gate, then a scorecard.

Step one looks at the balance sheet and asks whether the asset structure holds enough operating assets: plant, equipment, inventory, receivables, the things a business uses to do the work it says it does. Several outlets report that threshold as more than half of total assets. The Block describes step one qualitatively instead, as an assessment of whether the structure contains sufficient operating assets. Until MSCI publishes the final wording, sufficiency is the safer way to hold it. Clear the gate and the company is finished with the screen entirely. Fail it and the accounts go to step two.

Step two is five flags, each carrying a number:

• operating assets below 20% of total assets
• operating expenses below 5% of total assets
• negative operating cash flow
• non-operating fair-value changes above 5% of total assets
• capital dependence above 20%

Fail four of those five and the company is ineligible for the index.

Eunha reads the list twice and then points at what is not in it. There is no profitability hurdle here. No valuation test, no governance score, no sector judgment, no view whatsoever on whether the asset sitting on the balance sheet is a sensible thing to own. Each of the five asks a version of one question, and it is not about quality. Does this entity do work, or does it hold things while the work happens somewhere else?

"It is not a verdict on the company," she says. "It is a definition of what counts as one."

That difference decides who the rule can catch, and it is why the screen behaves so strangely against intuition. The phrase removed from an index calls to mind a struggling business, because that is what the older eligibility rules were built to find: market cap that fell below a floor, liquidity that dried up, a listing in trouble. This screen is looking for something else entirely, and a company can be superbly run, deeply liquid and up several hundred percent while failing it.

The two-step design is doing something specific as well. A manufacturer or a retailer clears step one on inventory and equipment without ever noticing the test exists, so the flags only get read on entities whose balance sheets already look unusual. By the time a company is being measured against them, the question has narrowed from is this a good business to is there a business here.

Eunha thinks the sequencing is the part people will skip. "Step one is not a warning shot," she says. "It is the whole population filter. Everything after it is a conversation about companies that already failed to look like companies."

Four of the five flags fire when the quarter goes well

Take the flags in order and imagine a treasury company having an excellent year.

The asset it holds appreciates sharply. That appreciation lands in the accounts as a non-operating fair-value change, and once the gain passes 5% of total assets, one flag is up. To buy more of the asset the company issues equity, because operations are not throwing off the cash to fund purchases at that scale. Financing that way is capital dependence, and above 20% a second flag is up. The company also runs lean by design: a small team, few fixed costs. That is the efficiency the model was sold on. Operating expenses stay below 5% of total assets. Third flag. And every unit of the asset added pushes operating assets further below the 20% line, because the numerator never grows. Fourth.

Four of five. Ineligible.

Not one of those four requires anything to have gone wrong. The appreciation is the thesis working. The equity raise is the strategy executing on schedule. The thin cost base is the discipline that was promised to shareholders. Only one flag in the set, negative operating cash flow, describes something a holder would recognise as bad news, and a company can fail the screen without ever tripping it.

Eunha slows down here, because this is the part she thinks is new. "The rule reads success and failure with the same instrument," she says. "The better the quarter, the more the balance sheet looks like a holding vehicle. That is what a good quarter does to a holding vehicle."

The treasury model deletes itself by working.

There is a version of this that reaches well beyond treasury companies. A rule written about structure does not measure what you intended. It measures what your structure does when the thesis works. Optimise hard enough for one outcome and the accounts start to resemble the exact thing the rule was written to exclude.

A uranium holder fails the same arithmetic

Yellow Cake buys uranium and stores it. No mines, no processing, no crypto anywhere near the balance sheet, and the same simulation deletes it on the arithmetic that catches Strategy.

The watchlist sharpens the point further. Alongside SharpLink, the consultation flags Center Laboratories, a Taiwanese pharmaceutical company, and Lydia Holding. A screen catching a bitcoin treasury, a uranium store and a pharmaceutical firm in the same pass is not describing an asset class. It is describing a shape: assets that sit rather than work, cash that arrives from financing rather than from operations, and a cost base too small for much to be happening inside the company.

Eunha is unsentimental about what this means for someone holding one of these. "If you want to know whether your position is exposed to this, do not start with what the company owns," she says. "Start with whether the accounts look like a business doing something. The rule cannot see the asset. It can only see the shape."

Nothing has been decided yet, and the timeline matters for anyone reading this before the outcome. Feedback closes on 30 September, results are expected on 16 October, and any change takes effect no earlier than the November 2026 index review. MSCI considered a near-identical proposal earlier this year and chose not to act on it, which is the detail that keeps this a live consultation rather than an announcement. The outcome is still undecided.

Who actually has to sell when an index deletes a stock?

A fund that tracks an index does not get a vote.

When the index drops a constituent, every fund whose mandate is to replicate that index has to drop it too at the rebalance, whatever the manager privately thinks about the company's prospects. That obligation is the entire mechanism of forced selling, and it is the same one that makes a fund trade against its own view whenever the rulebook and the analysis disagree. The rulebook wins, every time, by design. That is what the fund was bought for.

JPMorgan put the passive outflow from MSCI indexes alone at roughly $2.8 billion for Strategy. Hold that number as what it is: an estimate of flow, from one bank, covering one company, under one index family, conditional on a consultation that has not concluded.

There is a second forced seller in this story, and the two get confused constantly. The first sells the stock, because an index rulebook obliges it to. The second is the treasury company itself selling its own bitcoin under balance-sheet pressure, which is when the company itself becomes the seller rather than its shareholders' index funds. Different seller, different asset, different trigger, different timetable. An index deletion acts on the first and says nothing at all about the second.

Eunha keeps them apart deliberately, because conflating them is how a flow estimate turns into a rumour. "Ask which asset is moving, and who is obliged to move it," she says. "Those two questions answer almost everything anyone gets wrong about forced selling."

How big is the shock when an index deletes a stock?

Eunha sets two findings side by side. Both are correct. They disagree by a factor of a hundred.

Robin Greenwood and Marco Sammon measured what happens to a share price when the S&P 500 adds or drops a name. The effect is shrinking toward nothing. Their paper, The Disappearing Index Effect, ran in the Journal of Finance in April 2025 and is free to read as an NBER working paper. It documents the abnormal return on addition falling from an average of 7.4% in the 1990s to under one percent over the past decade. Deletions, once a large negative, ran around 0.1% between 2010 and 2020. All of that happened while the money tied to index products grew enormously, which is the opposite of what the mechanical story predicts. Their explanation is that index changes became anticipated, and that other institutions now stand ready to take the other side of the indexers' trade.

Small caps behave nothing like that. In one recent small-cap rebalance, Park Aerospace fell about 10% on deletion and Motorcar Parts about 14%, while Kennedy-Wilson rose about 10% on inclusion. One number from that episode reconciles the two findings. A single index fund held roughly 1.5 million shares of Park Aerospace. That was about ten days of the stock's average trading volume.

So what decides which of those two worlds a given deletion lands in?

The deletion is not what sets the size of the move. The ratio is. It is how much stock the obliged sellers must shift against how much of it trades on a normal day, discounted by how much of that flow the rest of the market has already positioned for. Ten days of volume in a company worth a few hundred million dollars is a wall with a date on it. The same fund weight in a mega-cap with a deep order book is a rounding error. Arbitrage desks have had that date in the diary for weeks.

Applied to this consultation, that ratio is the only calculation worth doing, and it has to be done name by name. Strategy's free float against its daily volume gives one answer. A smaller listing with a thinner book gives a very different one, on the same $2.8 billion headline and the same rulebook.

That is the same discipline as reading the pipe rather than the print on ETF flow days. A lockup expiry becomes a supply problem on this arithmetic too, where percentage of float against daily volume decides whether new supply gets absorbed quietly or painfully.

Eunha refuses to let the flow estimate become a forecast, and she is firm about it. "$2.8 billion is a quantity of stock, not a percentage move," she says. "To turn one into the other you need the float and the volume. Then you still need to know who is waiting on the other side, and that part is not in any document."

What to read before you buy a wrapper

Eunha closes the document. What survives it is not the list of names. It is a habit, and the habit takes ten minutes.

Find the index the wrapper belongs to. Its eligibility rules are published and free, and the section that matters is the one defining what kind of entity is allowed in rather than the one listing market cap and liquidity minimums. Then ask the uncomfortable question: which of these rules does my thesis trip when it works?

For a treasury company the answer is four of five. For an ETF, the equivalent question is what the fund is obliged to hold and on what date it is obliged to sell. For a tokenized wrapper, it is who owes you the underlying and under what conditions they can stop owing it. In every case the rule is upstream of the asset, and it was written by someone who was not thinking about your position.

Eunha's version is blunter than the general form. "You are buying two things," she says. "The asset, and the membership. The asset gets priced carefully. The membership gets inherited by accident, and then it decides who sells your stock and when."

Whatever MSCI announces on 16 October, the definition is now written down and public. Definitions get borrowed.

Put the Nasdaq and Nvidia side by side in the same $5,000 paper account and hold both through a week. One is a membership list, the other is a business, and watching them come apart in real time teaches the difference faster than any rulebook does. Simulated capital, real divergence.

Start free →