What Dados B3 does not fix: 401 companies left the register since 2010 and none of them is in the base

The universe comes from the regulator's register with one filter: only companies whose status is ACTIVE. That is 456 companies. Outside them the register carries 2,135 non-active records, and 401 are Category A companies — those required to file full statements — that cancelled registration from 2010 onwards. The intersection between the two sets is 0: not one of them is in the base.

The ones that leave are not a random draw. Out go those acquired, those that went private, and those that failed — and it is the third that ruins a backtest, because the bad return is exactly what disappears. The six most recent, from the register itself:

Fixing it would require the historical series of the cancelled companies, which the regulator does not publish in usable form: each year's file carries whoever was required to report that year, and rebuilding the eligible universe for every date is work we have not done. Measuring and publishing is what can be done honestly. Promising the rest would repeat the very error this page corrects.

The base holds the companies that are active today. A backtest run over 2015 with it is picking among the firms that made it to 2026 — information nobody had in 2015. That inflates returns, and inflates them more the longer the window.

This limit is no longer just a size. The page <a href="/saiu-da-bolsa?lang=en">the companies that left the exchange</a> shows which ones they are, how many years of statements each filed and — the part that matters to anyone thinking of reconstructing the universe — why only a fraction can be tied back to the share they traded.

There is no exit date, no final price, no return through the close. A company simply stops appearing. Anyone needing the return of a position up to the delisting will not find it here.

We do not know who was in the main index in 2013. Our sector index is built from the regulator's sector classification over the current universe: it compares sectors against each other, it does not reproduce an index portfolio.

That is when Brazilian accounting converged to IFRS. Before it the chart of accounts is a different one, and stitching the two eras together would produce a series continuous in appearance and incomparable in substance.

what we DO fix: the look-ahead from price and filing · transparency · comparison