What do 13F filings tell you about institutional buying?

There is a public dataset that shows what the largest investors in the world own, updated four times a year, free to read. It is one of the few genuinely informative disclosures a retail investor has, and it is underused because the raw form is tedious.

What is a 13F?

Any institution managing more than 100 million dollars in US listed securities has to file a form 13F with the Securities and Exchange Commission every quarter, listing its positions. The filings are public and free.

One filing on its own is a snapshot: here is what this manager held on that date. The useful part comes from comparing two consecutive filings, which shows whether a position grew, shrank or stayed flat. That change is the signal.

Why BlackRock, Vanguard and State Street?

Because they are the three largest asset managers in the world, and because they hold positions in nearly every large listed company. That ubiquity is the point. A test built on managers who appear in some filings and not others would mean different things for different stocks, which makes it useless as a consistent filter.

It is worth being clear about what their buying does and does not mean. A large part of what these three manage is index funds, which buy because an index says to, not because an analyst formed a view. So this is not a read on conviction. It is a read on flows and inclusion: money is moving toward the stock rather than away from it, and it is doing so at the largest scale that exists.

What exactly does Signal desk require?

The rule is deliberately not "all three increased", because that turns out to be almost never true and would reject nearly everything.

Instead: at least two of the three must have increased their position, and the third must not have reduced it. Flat counts as acceptable for the third. Any of the three missing from the data counts as a failure, not as a neutral.

Why does this filter reject so many stocks?

Two reasons, and both are about missing data rather than about bad businesses.

First, ownership data is commonly published only for the top ten holders. If BlackRock or State Street sits at position eleven for a given stock, their change is simply not in the data, and the check cannot be completed. Signal desk counts that as a failure. Real stocks have been rejected on this alone.

Second, and more fundamentally, 13F is a US disclosure regime. Brazilian stocks listed on B3 have no equivalent filing, so a B3 stock cannot produce the data this filter needs and therefore cannot pass it. That is an honest limit of the system rather than a judgement about Brazilian companies, and it is stated plainly on the site rather than hidden.

How current is the data?

Not very, and this is the filter's real weakness. Filings are quarterly and are due 45 days after the quarter closes, so the freshest available picture can be several months old by the time anyone reads it.

That is why this is used as a slow moving check on the quality of a business and never as a timing signal. Timing comes from the daily chart. This answers a different question: is serious money accumulating this company, or leaving it?

Where to read the filings yourself

The primary source is the SEC's EDGAR full text search, which is free and requires no account. Searching for an institution's name and filtering to form type 13F-HR returns their filings directly. Comparing the current one to the previous quarter's is exactly the calculation described above, done by hand.