Brazil's receivables funds

FIDCs are Brazilian funds that buy receivables — trade credit, payroll-deduction loans, judicial claims, consumer paper — and fund themselves by issuing quotas. They file monthly to the CVM at fund level, and this project has landed those filings from 2013. It is the deepest data in this section by a long way: the cross-country panel above is one number per country per quarter, and this is four thousand funds a month.

Every figure here is RECOMPUTED from the landing tables this project ingested, not read from the research database the market was first measured in — that database is evidence and never a source. The recomputation reproduces it: at June 2026, net assets of R$ 1,001.4 bn, credit rights at risk of R$ 448.8 bn and 4,343 funds filing, against 1,001.4, 449 and 4,343 in the report. The segment mix agrees to a few hundred million on a R$ 775 bn book.

THE DUPLICATE COLLAPSE IS NOT OPTIONAL. The net-assets table carries 238 month-and-fund groups with more than one row. Of those, 201 are exact duplicates of the whole filing — the same return landed twice — and 37 are real subclasses of one fund that must both be counted. So every total sums distinct WHOLE ROWS. Collapsing on a narrower key would be worse than not collapsing at all: on the delinquency table it would delete 117 rows that differ elsewhere and merely happen to report the same figure.

A blank is absent, never zero. The filings land as text and a value is used only where it parses as a number, because a fund that filed nothing for a segment is not a fund that filed zero. Two segments look empty for that reason rather than because the market left them: factoring and brand-licensing receivables together are under R$ 0.2 bn of a R$ 775 bn book.

The currency switch selects between figures published in each currency; nothing is converted in your browser. Dollars use the monthly average of PTAX venda, the same rate every other dollar figure on this site uses. A dollar series of a Brazilian market moves with the exchange rate as well as with the market, which is worth remembering before reading a fall in the dollar line as a fall in lending. Fund counts are the same under either setting.

Registrations and cancellations come from the CVM cadastro as this project ingested it, one row per fund, and are close to but not identical with the market report’s — 1,553 registrations in 2025 against its 1,543 — because that report reads CVM’s file directly and this reads the landed snapshot, which deduplicates differently. Cancellations are drawn negative so the pair reads as a balance, and the current year is partial.

Resolution 175 took effect in 2024 and moved reporting to class level, which lifts the fund count without changing the underlying credit. A jump in the number of filers around then is a change in how the market reports itself, not in how much of it there is.

THAT CHANGE HAS A DATE, AND IT IS A BREAK IN EVERY PER-FILING DISTRIBUTION. Class-level filings are 0.0% of the panel through September 2024, 60.5% in October, 93.3% by December and 99.4% by August 2026. Before it a filing is a whole fund; after it a filing is one class of one. So the size histogram’s left tail moves down, the concentration curve ranks the pieces rather than the wholes, and subordination falls hardest of all — a fund’s subordinated class now files separately from the senior class it protects. Financial-receivable filings read a median 30.6% subordination in August 2024 and 11.1% in August 2026, and most of that is the unit rather than the market. The distributions are therefore published as monthly panels read ONE MONTH AT A TIME, with the month on the frame.

SUBORDINATION IS MEASURED ON RECONCILING CLASSES. A filing is counted only where it carries a senior subclass, reports positive net assets, and its quota rows add up to its own reported net assets within 1% — the same reconciliation the class series uses, because individual filings carry decimal-scale errors and one of them reports R$11 quadrillion. At June 2026 that is 3,100 filings, of which 53.7% carry no subordinated subclass at all, and the median is 0.0%. The market report reads 16.6% over 2,490 filings; the difference is a fund-level join that misses most class-level filers, no reconciliation filter, and a clip at 60%. The class split itself agrees to 1.3%.

THE BOX PLOT PUBLISHES QUANTILES, NEVER A FUND. Six numbers per segment reach the browser — the 10th, 25th, 50th, 75th and 90th percentiles and the count — so the box is the interquartile range, the whiskers are the 10th and 90th, and there are no outlier marks. An outlier is one fund, and no fund is in the artifact. Segments with fewer than 20 classes are not drawn rather than drawn from four.

RETURNS ARE MEDIANS OF THE CLASSES THAT PAID. A class filing a return of exactly 0.00 did not pay that month and is excluded — 51.8% of senior rows at June 2026 — because reading those as a realised zero would halve every median in the series. The trailing-twelve-month spread compounds twelve monthly returns for classes with twelve PAYING months and subtracts the CDI compounded over the same twelve. At June 2026 that is senior +3.5 points, mezzanine +5.6 and subordinated +6.6 over a CDI of 14.8%, against 3.1 / 5.8 / 6.2 in the market report. The classes themselves come from 1,825 filed labels normalised by this project’s registry, which covers 98.68% of the labels and 99.90% of the value; the class panel is unpopulated before November 2020 and every by-class series starts there.