Bad Loans and Bank Stress 2026: Nonperforming Loans Across 144 Economies

A loan is an asset only as long as it is being repaid. When borrowers stop servicing their debts, those loans turn from an income stream into a hole on the bank’s balance sheet — and when enough of them go bad at once, the banking system itself comes under stress. The single cleanest gauge of that strain is the nonperforming-loan ratio: bad loans expressed as a share of total gross loans. This report maps that ratio across 144 economies, from the soundest loan books in the world to the handful where roughly a third of all lending is impaired.

What the nonperforming-loan ratio measures, and why direction matters

The nonperforming-loan ratio counts loans where repayment has effectively broken down — typically loans more than 90 days past due — and divides them by the total gross value of the loan book. The unit is a share of total gross loans, and unlike a price or a growth rate, it has an unambiguous good direction: lower is better. A bank with bad loans worth 0.5% of its book is in robust health; one where the figure runs into the double digits is carrying losses it may never recover. So throughout this report a rank of 1 marks the healthiest reading — the lowest ratio — and the bottom of the ranking marks the most stressed systems.

Every figure below comes from the World Bank’s World Development Indicators (indicator FB.AST.NPER.ZS), as stored in FinStatGlobe’s derived per-country datasets. Each value carries the actual year of its latest available observation — most are recent, several from 2023, some older — and this series carries no 2026 projections, so none appear here. The “2026” in the title is editorial framing for the vintage of the survey, not a forecast of any number.

The global picture: a 3.6% median and a long, heavy tail

Across the latest observations for all 144 economies, the median nonperforming-loan ratio is 3.6%. The unweighted mean is markedly higher at 5.6% — and that gap between median and mean is the signature of a sharply right-skewed distribution. Most banking systems sit in a fairly tight band of low single digits, but a long tail of stressed economies pulls the average well above the typical reading.

That shape matters for interpretation. The typical loan book in the dataset is sound: half of all economies report a ratio at or below 3.6%. The damage is concentrated. A small group of economies with very high ratios does most of the work of lifting the global mean to 5.6%, and it is that group — not the median system — that signals where bank stress is acute.

The healthiest banking systems: the top of the ranking

The economies with the lowest nonperforming-loan ratios cluster among high-income systems with deep, well-supervised banks, though not exclusively. The single healthiest loan book belongs to South Korea at 0.2% in 2022, followed by Micronesia at 0.3% in 2021 and Norway at 0.4% in 2023.

RankEconomyNPL ratioYear
1South Korea0.2%2022
2Micronesia0.3%2021
3Norway0.4%2023
4Sweden0.4%2023
5Lithuania0.4%2022
6Canada0.5%2023
7Australia0.7%2022
8Switzerland0.8%2023
9United States0.8%2023
10Dominican Republic0.9%2023

Source: World Bank, World Development Indicators (FB.AST.NPER.ZS).

All ten of the top ten sit below 1% — the highest, the Dominican Republic at 0.9% (2023), is the one upper-middle-income entrant in the leading group — a reminder that a clean loan book is not the exclusive preserve of rich economies. Sweden, Lithuania, Canada, Australia, Switzerland and the United States fill out a top ten where bad debt barely registers.

The stressed tail: the bottom of the ranking

At the other extreme sit the economies where bad loans dominate the book. The most distressed system in the dataset is Ukraine, where nonperforming loans reached 37.4% of total gross loans in 2023 — more than ten times the global median. Chad (31.5%, 2023) and Equatorial Guinea (31.1%, 2023) follow close behind, each with roughly a third of the loan book impaired.

RankEconomyNPL ratioYear
135Angola15%2021
136Republic of the Congo15.2%2023
137Central African Republic16.2%2023
138Saint Kitts and Nevis19.4%2023
139Algeria20.3%2022
140Ghana20.6%2023
141San Marino21%2023
142Equatorial Guinea31.1%2023
143Chad31.5%2023
144Ukraine37.4%2023

Source: World Bank, World Development Indicators (FB.AST.NPER.ZS).

The bottom of the table is dominated by Sub-Saharan Africa — Angola (15%, 2021), the Republic of the Congo (15.2%, 2023), the Central African Republic (16.2%, 2023) and Ghana (20.6%, 2023) — but it is not exclusively so. The high-income microstates Saint Kitts and Nevis (19.4%, 2023) and San Marino (21%, 2023) appear here too, alongside Algeria (20.3%, 2022). Stressed loan books are a function of credit conditions and shocks, not income alone. Our companion piece on where bad loans pile up digs further into the geography of this tail.

Distribution bands: most systems are sound, a few are severely stressed

Grouping all 144 economies into bands makes the skew concrete. Forty-one economies carry bad loans below 2% of the book, and 49 fall in the 2–5% band — so 90 of 144 sit under 5%. Above that, the numbers thin out quickly: 34 economies land in the 5–10% band, 14 in 10–20%, and just 6 carry nonperforming-loan ratios of 20% or above.

NPL bandEconomies
Under 2%41
2–5%49
5–10%34
10–20%14
20% and above6

Source: World Bank, World Development Indicators (FB.AST.NPER.ZS).

The picture is one of broad health with a concentrated tail of distress. The six economies at 20% or above — Algeria (20.3%), Ghana (20.6%), San Marino (21%), Equatorial Guinea (31.1%), Chad (31.5%) and Ukraine (37.4%) — are where the systemic risk lives, and they are the reason the mean sits a full two points above the median.

Regional medians: a clear North-to-South gradient

Median ratios by region trace a familiar gradient. North America has the lowest regional median at 0.7% (across just 2 economies), followed by Europe & Central Asia at 2.4% across 48 economies and East Asia & Pacific at 2.8% across 19. Latin America & the Caribbean sits close to the global median at 3.2% across 28. The highest regional medians are in South Asia (6.1% across 6) and Sub-Saharan Africa (6.6% across 29), with the Middle East, North Africa, Afghanistan & Pakistan region between the two clusters at 4.8% across 12.

RegionMedian NPL ratioEconomies
North America0.7%2
Europe & Central Asia2.4%48
East Asia & Pacific2.8%19
Latin America & Caribbean3.2%28
Middle East, North Africa, Afghanistan & Pakistan4.8%12
South Asia6.1%6
Sub-Saharan Africa6.6%29

Source: World Bank, World Development Indicators (FB.AST.NPER.ZS).

The income gradient runs the same way. The median high-income economy reports a 2% ratio (across 56 economies), upper-middle-income economies sit at 3.8% (43), lower-middle-income at 5.8% (35), and low-income economies at 6.6% (9). A weaker loan book tends to accompany thinner credit markets — context we develop in how the world finances business and in our look at capital and business finance.

Cleanups and deteriorations over the long run

Because the series stretches back across multiple decades — the longest country tracks here run to roughly two dozen readings — the long-run movers are revealing. A falling ratio is a cleanup; a rising one is a deterioration.

The most dramatic deteriorations are clustered in the same economies that anchor the stressed tail today. In Equatorial Guinea the ratio climbed from 4.7% in 2010 to 31.1% in 2023, a rise of +26.4 pts; in Chad from 9.9% in 2010 to 31.5% in 2023, +21.6 pts; and in the Republic of the Congo from 1% in 2010 to 15.2% in 2023, +14.3 pts. San Marino added +14.1 pts (from 6.9% in 2009 to 21% in 2023), Ghana +12.9 pts (from 7.7% in 2008 to 20.6% in 2023), and Angola +11.8 pts (from 3.2% in 2010 to 15% in 2021).

EconomyEarliestLatestChange
Equatorial Guinea4.7% (2010)31.1% (2023)+26.4 pts
Chad9.9% (2010)31.5% (2023)+21.6 pts
Republic of the Congo1% (2010)15.2% (2023)+14.3 pts
San Marino6.9% (2009)21% (2023)+14.1 pts
Ghana7.7% (2008)20.6% (2023)+12.9 pts
Angola3.2% (2010)15% (2021)+11.8 pts

Source: World Bank, World Development Indicators (FB.AST.NPER.ZS).

The cleanups are just as striking, and several come out of post-crisis recoveries. Ukraine shed −20.6 pts, falling from 58% in 2005 to 37.4% in 2023 — and yet still ranks as the most stressed system in the dataset, a measure of how deep the starting hole was. Bosnia and Herzegovina cleaned up −17.4 pts (from 21.2% in 2000 to 3.8% in 2023), Belize −15.5 pts (20.4% in 2017 to 4.9% in 2022), and the Maldives −12.6 pts (20.9% in 2012 to 8.3% in 2023).

EconomyEarliestLatestChange
Ukraine58% (2005)37.4% (2023)-20.6 pts
Bosnia and Herzegovina21.2% (2000)3.8% (2023)-17.4 pts
Belize20.4% (2017)4.9% (2022)-15.5 pts
Maldives20.9% (2012)8.3% (2023)-12.6 pts
North Macedonia14.9% (2005)2.7% (2023)-12.2 pts
Moldova16.4% (2009)5.6% (2023)-10.8 pts

Source: World Bank, World Development Indicators (FB.AST.NPER.ZS).

North Macedonia (−12.2 pts, from 14.9% in 2005 to 2.7% in 2023) and Moldova (−10.8 pts, from 16.4% in 2009 to 5.6% in 2023) round out the cleanup leaders — two more Eastern European systems that worked through legacy bad debt over a decade and a half. The pattern is symmetric: where loan books deteriorated most, credit stress followed; where they were cleaned up, financial systems steadied. The same forces shape the rise of newer credit channels, a theme we track under digital lending.

Sources & method

All statistics in this report come from the World Bank’s World Development Indicators — specifically bank nonperforming loans to total gross loans (FB.AST.NPER.ZS), measured as a share of total gross loans — as stored in FinStatGlobe’s derived per-country datasets. The ranking universe is the 144 economies that have a latest available value for the indicator; ranks run from 1 (lowest ratio, healthiest loan book) to 144 (highest ratio, most stressed), because for this measure lower is better. The median (3.6%) and mean (5.6%) were computed across every economy in that universe using each economy’s latest available observation, the median following the standard convention (for an even count, the mean of the two middle values). Regional and income-group medians were computed the same way over the economies in each group, with the group size shown in every table. Each figure is labelled with its actual observation year; no 2026 projections are shown, because none exist for this series. For the full pipeline — how a World Bank snapshot becomes a country page — see the methodology page.