Radar Perene / Archive / science

BCB open data: from the SGS to a Diário reading

◦ Index methodology v2.2 (working papers with DOI). See the methodology.

Science

Every business day, a handful of numbers runs through Radar Perene's Diário: the Selic rate, the CDI, accumulated inflation, the closing exchange rate. The reader finds them ready, contextualized, compared against their own history. Almost nobody asks where they come from — and the answer is more interesting than it seems, because it involves no expensive terminal, no subscription, no privileged access. It involves a public system anyone with a browser can query right now.

The SGS — the Brazilian Central Bank's Time Series Management System — is the open repository that gathers thousands of economic and financial series, freely searchable, with no registration, through a web interface or an API. Each series has a fixed numeric code, a frequency and a history that, in many cases, spans decades.

What exactly is in there?

The catalog goes far beyond the policy rate. The SGS holds inflation series in all their breakdowns, monetary aggregates, credit by type, the expectations collected by the Focus survey, the exchange rate in its several quotes, activity and external-sector indicators. Each one answers to a code: the Selic target is series 432, the daily CDI is 12, the monthly IPCA change is 433, the PTAX selling dollar is series 1. The code is the data's true name — two researchers citing "series 433" can be certain they mean the same number, which is not always the case when one simply says "the IPCA."

Querying is direct: the web interface builds tables and charts, and the API returns any series in JSON or CSV from a simple URL. There is no access key and no usage cap that affects an individual researcher. The SGS also coexists with a younger sibling: the Central Bank's open data portal, which gathers datasets beyond time series — exchange quotes by institution, payment-system statistics, financial-system data. For most historical-series questions, though, the SGS remains the address. For research, this has a consequence worth underlining: a study built on SGS codes can be checked by any reader, at the same address, for free — citing the data already provides the path to it.

From raw series to published reading

Between an SGS number and a publishable sentence lies a journey, and that is where the work lives. The house's journey has three stages that can be described without secrecy.

First, disciplined collection: series come in by code, stamped with date and source, and the stored value is the published value — if the agency later revises the history, the revision is logged as an event, not overwritten as if there had never been another number. Second, context: today's value is set against the series' own past — the step that turns "this month's CDI was X" into information, because it says whether X is routine or rarity. Third, editorial judgment: not every statistical rarity is news, and the Diário records the difference.

The house's archive holds a published example of that third step. In May 2026, monitoring flagged a statistical anomaly in the monthly CDI — a historically tame series — and the month's memo highlighted the number and, in the same paragraph, disarmed it: more series noise than economic event. The episode is told in "The extreme number that wasn't news". In numbers: the Selic closed that month at 14.5% a year — a figure any reader can verify in SGS series 432, on the Central Bank's website, without asking anyone's permission.

That is the sense in which an SGS number "becomes" a Diário reading: the raw material is public and the same for everyone; what the house adds is organized history, a statistical ruler and the decision — signed and dated — about what deserves the record.

Why this matters for researchers

Open data with stable codes solves, at once, two classic problems of applied research: provenance (where the number came from) and reproducibility (how another researcher obtains it). A table citing SGS codes needs no data appendix — the appendix is the Central Bank itself. For those researching outside institutions, without budget for commercial vendors, this infrastructure is what makes the work possible; the house's published working papers lean on it extensively.

Frequently asked questions

Is the SGS really free?

Yes. Web queries and the API are open, with no registration, no key and no charge.

How do the SGS and the Focus survey differ?

Focus is the Central Bank's market expectations survey; its medians are published as series within the SGS. The SGS is the repository; Focus is one of the sources feeding it.

Are SGS data revised?

Some are, because the originating agencies revise their histories. Careful research records the query date and, when the revision matters, treats the original and revised values as distinct pieces of information.

Where does a first-time user start?

With the series search in the SGS web interface, on the Central Bank's website. Once the code of the series of interest is found, the same query works forever — the code does not change.

---

Continue the track: B3 open data: what is public and what requires a subscription →

House readings: these series' destination is today's reading, in the Diário; the precedents they illuminate, in the Atlas.

Selecting and treating a specific series for a concrete question is work the house does on commission, at another depth.