Bufetico
Sort thirty thousand companies by how cheap they look and something predictable happens. The top of the list is full of things that are broken.
Mine was topped by companies sitting 94% below what the calculation said they were worth. Every one of them had stopped trading. There were 4,041.
Why a dead company still has a price
A delisted company doesn’t disappear from a data feed. Its last traded price sits in the record, and depending on who’s serving the data, it keeps coming through as if it were current. Meanwhile the accounts from the years before it delisted are still there, still complete, still perfectly good inputs.
So the calculation runs. Real cash from real filings, divided by a share count that was real, compared with a price that was real on the day the company left the exchange.
The output says this business is available at six cents on the dollar.
Every input was real, and the whole thing still fell over on an assumption nobody writes down: that a quoted price is a price you can get.
The other batch
The same sort turned up a different problem. Companies around 97% below value, none of them delisted, all apparently trading normally.
They were preferred shares and listed debt, 505 of them.
A preferred share carries its parent company’s ticker family and, in a bulk feed, often inherits the parent’s accounts. The model then divides the parent’s entire cash flow by the preferred share’s price, which is a fraction of the common stock’s. The result looks like a company trading at a rounding error of its worth.
The instrument got misidentified before the math ever started.
Where the errors go
Data has errors. The part worth knowing is where they show up, because it isn’t spread evenly. Sort a list by how good things look and the broken ones pile up at the top. Any company whose price is wrong in the downward direction rises to exactly the spot where you’re most likely to be looking.
Which is why checking a screen on a sample of normal companies proves very little. Go to its extremes, because that’s where people read it.
Where this breaks
Both categories are excluded now. That exclusion is a blunt instrument and it has a price.
Some companies delist for benign reasons, an acquisition or a move to private ownership. Removing them removes a little legitimate history.
And a rule will always catch something it shouldn’t. The alternative, deciding case by case whether a price is genuinely current, can’t be done one company at a time across thirty thousand of them. A rule that’s slightly too wide beats deciding by eye, which can’t be done at that scale.
They’re still findable by search, with their valuation squares gray. Someone who looks up a company that stopped trading should be told it stopped trading, and that’s all the square should say.
That fences the problem in. It doesn’t solve it. These are the two categories I found. A list this size, put together from bulk feeds, probably holds others I haven’t found yet, and I’ll find them the same way I found these. By looking at the top of the list and asking why it’s there.