Four Brazilian indices, and why no link is claimed
The four Brazilian indices do not end on the same month, and the difference is load-bearing. IPCA, the 12-month IPCA and IGP-DI run to July 2026; only IGP-M reaches August. 'Brazilian inflation to August 2026' is therefore false of three of the four, and the dataset's own coverage date is the earliest of them so the sentence cannot be written from the data at all.
All four arrive through the central bank's SGS service under ODbL 1.0, not from IBGE directly. IBGE's own statistical service is marked unpublishable in this project's registry pending a terms verdict, which is why the licence names the central bank rather than the statistics institute.
The two Brazilian panels are shown together and NO LINK BETWEEN THEM IS CLAIMED. They correlate at −0.53 over 56 months, and that is two secular trends running in opposite directions rather than one causing the other: Brazil disinflated over the period, and on-chain volume grew over the period. The frame carries no argument colour for exactly this reason — there is no series carrying an argument, because the argument is that there is no link.
The on-chain panel is settled value that MOVED, so it excludes mints and burns, nine issuer treasury and reserve wallets for one token and two for another. The XRP Ledger is included as a tenth chain with issuer-touching transfers excluded the way mints and burns are, which means the combined series is complete only through its slowest chain.
TESOURO is excluded from that panel although it is denominated in reais, because it is suspected not to be a one-to-one claim and folding it in would merge a yield signal into a payments one. It is R$799,225 across 109 transfers in January 2026, about 0.037% of that month, and it is published in full as its own dataset.
Dollar values on that panel are the Banco Central do Brasil's own PTAX closing rate. That changed on 1 September 2026: until then they were converted at a market series whose provenance could not be established, which sat more than 5% away from the official rate on 13.1% of Brazilian days. The two months this note used to quote moved by −1.58% and +0.52% in the restatement, and the exchange rate is the whole of the difference. The older series has not been withdrawn — it is still published as its own dataset, so the change stays checkable.
The regional median is a DERIVED value, not a reading of anything. It is the median across the countries that published in the newest year, it carries that base and that year, and it is not the median of a fixed set of countries over time.
The regional matrix spans deflation to hyperinflation in one grid: 30 of 331 observations since 2015 are negative, the lowest −3.08%, and the highest is +254.95%. So its colour ramp diverges around zero rather than running from low to high, and its scale is a symmetric logarithm rather than linear. Mapped linearly, every country but Argentina and Venezuela would fall in a single bucket and the matrix would encode nothing. Every cell also carries its number as text, so the colour is never the only channel.