Bad loans, on one definition

The share of loans that are not being repaid, for seventeen countries, quarterly, using the International Monetary Fund's Financial Soundness Indicator for non-performing loans to total gross loans at deposit takers. One indicator code, one definition, every country.

That choice costs something and it is worth saying what. This project has landed far richer national data — seventy-five Brazilian ratios broken down by credit modality, seven Argentine ones by credit line, Peruvian overdue stocks by borrower type — and none of it is on this page. Each country's supervisor defines default for itself: Brazil counts from ninety days and redefined its problem-asset category in January 2025, Argentina's irregular portfolio is a supervisory classification, Peru's arrears are another thing again. Those series are the right answer inside a country and cannot be compared across one. The Fund's indicator exists to be harmonised, so the cross-country view uses it and the national detail stays where it belongs.

The second cost is timeliness. These are quarterly where several national series are monthly, and they lag by design. Uruguay's ends in the second quarter of 2025 and is the most overdue series anywhere in this project's registry; Bolivia's and Peru's end a quarter later; Ecuador, Nicaragua and Paraguay reach the second quarter of 2026. Those are the source's extents, not ours. A quarter a country did not publish is absent rather than zero, and any cross-section should name its own date rather than borrowing the newest date in the dataset.

Do not plot these against a Brazilian series taken from the central bank's credit register. The register reports operation by operation and the accounting aggregates do not; their ratio steps in 2019, the register redefined its problem-asset measure in January 2025, and the Fund's figure is a third measurement again. Splicing any two of them injects a level shift that reads as an event and is not one.