Radar Perene / Archive / science

Ipeadata: more than 11,000 series that are not just textbook macro

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

Science

The earlier articles on this trail visited four houses — Central Bank, B3, IBGE, Treasury, CVM — each the owner of its slice of Brazilian data. One house remains, one that produces almost nothing and for that very reason solves a problem none of the others solves: that of the researcher who needs series from five different sources, on the same table, in the same format. That house has existed since the 1990s, has an interface that betrays its age, and remains irreplaceable.

Ipeadata is the public database of IPEA — Brazil's Institute for Applied Economic Research — which aggregates, on a single portal and in a uniform format, thousands of economic, financial, regional and social series produced by dozens of institutions: from nineteenth-century exchange rates to municipal indicators of the latest census. By the portal's own count, more than eleven thousand series.

The key word is aggregates. Ipeadata is a curator, not a producer: the Selic found there comes from the Central Bank; the IPCA, from the IBGE; the fiscal series, from the Treasury and the BCB. The value lies in the gathering — and in three services the gathering provides.

Three services only the aggregator provides

The first is historical depth. Ipeadata keeps long series that the primary sources do not always expose with the same reach — exchange rates, prices and aggregates going back decades, in some cases centuries, stitched from documented historical sources. For this house's questions, which live on comparing the present with the largest available past, that reach is direct raw material.

The second is the frontier nobody else covers at the same counter: the regional and social series. Municipal GDP, state-level indicators, demographic and income data — this article's subtitle exists because the portal's reputation ("textbook macro") hides precisely that half. Applied research that crosses markets with territory — credit by region, employment and the local cycle — finds there what would otherwise demand digging through half a dozen sources.

The third is standardization: series from different origins come out in the same structure, with name, source, periodicity and unit declared. Whoever has manually harmonized three file formats from three agencies understands the size of the service.

From catalog to hypothesis: the pipeline

This section's agenda promises to describe how a series "becomes research", and the house's path can be told in four steps, with no method secrets.

Step one: the question before the catalog — browsing eleven thousand series without a hypothesis is tourism, pleasant and infertile. Step two: locating the candidate series and, immediately, reading its record — original source, periodicity, unit, end date. Step three, the most important and the least practiced: verification at the primary source. The aggregator is a mirror; mirrors lag and, rarely, distort — the house treats Ipeadata as a discovery index and confirms the decisive datum at the origin before publishing it. Step four: the series enters this trail's common protocol — consultation date stamped, contrast against its own history, hypothesis declared before the test.

An example from the archive closes the circuit. The essay on August 2015 — the month three series entered rare territory at the same time — has as its protagonist the public debt as a share of GDP, which reached 62.97% ("Public debt in rare anomaly"). The series is compiled by the Central Bank, as the previous article on this trail detailed — and it is also in Ipeadata, under the public finance theme, next to the exchange rate and the interest series the same essay mentions. For the reader who wants to redo that entire August, the aggregator is the only place where all the episode's pieces sit on the same shelf. That is exactly its role in research: not to replace the sources, but to make them comparable.

The usual warning, in the right dose

Aggregated data inherits the revisions, discontinuities and methodology changes of its sources — and Ipeadata's records document part of that, not all. This trail's ruler applies in full: citation with the exact series name and consultation date, verification at the primary source when the number carries the argument, and methodical suspicion of series spliced from different methodologies, which sometimes hide a seam where the researcher sees a trend.

Frequently asked questions

Is Ipeadata the primary source of anything?

Of almost nothing — and that is its function. The series have their origin declared at other institutions; IPEA gathers, standardizes and documents them. Rigorous citation names both: the original source and the aggregator consulted.

How does it differ from the Central Bank's SGS?

The SGS is the primary source of monetary and financial series and mirrors some aggregates; Ipeadata covers a larger territory — regional, social, historical — aggregating dozens of sources. For a pure financial series, the SGS suffices; for cross-source work, Ipeadata saves weeks.

How often are the series updated?

Each series inherits the original source's calendar, with an aggregator's own lag. Each series record states the last update — a datum that belongs in any citation.

Is there API access?

Yes, with public documentation, in addition to consultation and export through the web interface. The choice between the two is one of scale, not of access.

---

The open data trail ends where the whole section began. With the five houses on the table — series, tables, stocks, filings — the question stops being "where do I find data?" and returns to the first of all: what exactly does one want to test? It is the rereading, now with data in hand, of How research is born.

House reading: these series flow every day into the reading of the Diário; the episodes they reconstruct, into the Atlas.

Choosing and treating a specific series for a reader's question is the kind of work the house does on commission.