Brazil Opens Climate-Risk Data, but Smaller Institutions Remain Harder to Compare

Brazil's central bank will put institutions in prudential segments S1 through S4 on a new climate-risk reporting architecture from January 1, 2027. The reports must remain public for five years and also be available as open data.

The files will be easier to find and process. They will not make every institution equally comparable.

Under Resolution BCB 586, S1 and S2 institutions must provide all 12 mandatory GRSAC tables for the December 31, 2027 data date. S3 institutions can defer nine of those tables until the 2028 data date. S4 institutions can initially publish governance alone and will never face the full quantitative set under the current rule.

That distinction defines what investors and supervisors can compare. Open data solves an access problem. It does not fill a disclosure gap.

Problem: one framework, three effective scopes

The rule applies to S1, S2, S3 and S4 institutions. Members of the same prudential conglomerate must report on a consolidated basis. The BCB's segmentation framework deliberately applies simpler prudential requirements as institutions become smaller and less complex.

Proportionality makes regulatory sense. It also means the first expanded GRSAC cycle will not support a single league table across the regulated universe.

The coverage matrix is stark:

The optional opportunities table sits outside these counts. Individual fields identified as optional in the templates remain optional.

Publication timing adds another layer. Each report covers a December 31 data date. The normal deadline is 90 days later, while the first report may be published within 180 days. The data date is therefore not the date on which the market receives the file. Comparing a December snapshot with a June publication would confuse measurement time with delivery time.

The rule also requires institutions to correct and republish inconsistent information immediately. A corrected version restarts the five-year availability period and must identify that an update occurred. Users will need to track versions, not merely download the latest file.

Analysis: standard fields cannot repair missing tables

Normative Instruction BCB 772 points institutions to a 41-page set of standard GRSAC layouts. The templates define common headings, fields and units. Several quantitative tables use fixed formats. Sector classifications must follow a BCB reference mapped to Brazil's CNAE activity codes.

That is real progress. The expanded set covers financed emissions and credit exposures by sector, agricultural exposures by region, drought and heavy-rain risk, transition plans, environmental sanctions, tailings dams and rural credit linked to native-vegetation removal. Brazil's Ministry of Finance describes the change as an increase in consistency and comparability.

The improvement is strongest within the same reporting scope. An S1 lender's sector exposure can be compared with another S1 lender's corresponding field. For the 2027 data date, the same field may be absent from an S3 report because the institution used a permitted deferral. It will be absent from an S4 report because the rule does not require that table.

Those blanks have different meanings. One is temporary. One is structural. Neither is zero.

The S3 gap is especially important in the first cycle. The segment can defer the strategy table and tables V through XII. What remains is governance, risk management and the institution's methodology for identifying climate risk. All three are flexible-format, qualitative disclosures. S3 readers may therefore see how an institution says it governs climate risk without receiving the fixed exposure tables needed to test the scale or concentration of that risk.

S4 coverage is narrower by design. After phase-in, those institutions must provide governance and two tables describing voluntary social, environmental or climate commitments. They need not publish the expanded tables on sector exposure, financed emissions, physical hazards or social and environmental exposures.

A parser can ingest what exists. It cannot infer an institution's drought exposure from a table that the rule did not require.

The new phase-in should also be kept separate from the previous delay. Resolution BCB 445 had moved the former regime's open-data requirement to the December 2026 data date. Resolution 586 revoked that transitional clause and created a new schedule. The S3 and S4 relief concerns disclosure content, not simply the arrival of machine-readable files.

Implications: compare like with like, then mark the gaps

Supervisors can make the framework more useful by publishing a coverage map alongside any aggregate analysis. Each observation should carry the institution's segment, data date, publication date, table version and whether a permitted deferral applies. Structural omissions should be labelled separately from missing or corrected submissions.

Bank risk officers face a choice beyond minimum compliance. An S3 institution that publishes deferred tables early can make its exposures easier to compare with larger peers. An S4 institution can provide voluntary detail, but investors should not assume a common schema where the rule supplies none.

Institutional investors should keep comparisons inside matched cohorts. Governance language can be compared across S1 through S4. Fixed exposure metrics can be compared across institutions that actually publish the same tables for the same data date. Moving from one cohort to another changes the denominator and the available evidence.

The reporting gaps say nothing by themselves about the quality of an institution's climate-risk management. A smaller institution may manage risk well and disclose less because proportional rules permit it. A larger institution may disclose every required field and still carry concentrated exposure.

Brazil is making climate-risk information more accessible and more standardized. That is not the same as making every regulated institution comparable. The new files will show more. Their most important metadata may be what each institution was never required to show.