Key takeaways
- Across S&P's global corporate pools from 1981 to 2024, 0.14% of BBB rated issuers defaulted within a year, against 26.12% of those rated CCC/C.
- The step from BBB to BB quadruples the one-year rate, 0.14% to 0.56%. Over 10 years the gap widens from 2.86% to 10.44%.
- In S&P's March 2026 SEC filing, 16.77% of issuers rated B on 31 December 2015 had defaulted 10 years later and 68.60% had the rating withdrawn.
- S&P had 4 AAA rated corporate issuers outstanding at the end of 2024, Moody's 13 and Fitch 2. The top row of every table is that thin.
- The ICE BofA BBB index paid 1.00 percentage points over Treasuries on 18 August 2026, while the CCC and lower index paid 10.27.
Default rates by credit rating: what 44 years of pools actually did
You want to know how often a bond rated BBB defaults, and how much worse BB is. Here's the longest published answer, with one caveat to carry through it: default rates by credit rating are counts of what happened, not forecasts.
S&P Global Ratings groups every corporate issuer it rates into a pool on 1 January each year, then follows that pool forward. Its 2024 Annual Global Corporate Default And Rating Transition Study runs those pools from 1981 to 2024, which is 44 years. Within one year, 0.14% of the issuers rated BBB had defaulted. At BB it was 0.56%, at B 2.93%, and at CCC/C 26.12%. Not one issuer rated AAA defaulted inside a year across the whole period.
Stretch the horizon and the ranking holds. The cumulative default rate over 10 years was 0.67% for AAA, 0.65% for AA, 1.03% for A, 2.86% for BBB, 10.44% for BB, 22.02% for B and 50.43% for CCC/C. Investment grade as a block came to 1.69%. Speculative grade came to 19.15%.
| Rating | 1 year | 3 years | 5 years | 10 years | 15 years |
|---|---|---|---|---|---|
| AAA | 0.00 | 0.13 | 0.34 | 0.67 | 0.86 |
| AA | 0.02 | 0.11 | 0.28 | 0.65 | 0.90 |
| A | 0.05 | 0.19 | 0.39 | 1.03 | 1.56 |
| BBB | 0.14 | 0.67 | 1.36 | 2.86 | 4.01 |
| BB | 0.56 | 3.12 | 5.75 | 10.44 | 13.05 |
| B | 2.93 | 10.46 | 15.60 | 22.02 | 25.11 |
| CCC/C | 26.12 | 41.32 | 46.53 | 50.43 | 52.30 |
| Investment grade | 0.08 | 0.37 | 0.77 | 1.69 | 2.38 |
| Speculative grade | 3.54 | 9.55 | 13.64 | 19.15 | 21.93 |
The chart above plots the five categories that carry the shape; AAA and AA are left out because at 0.67% and 0.65% they don't render against a 50.43% bar. AA looks marginally safer than AAA at the 10-year mark, 0.65% against 0.67%. S&P deals with that directly. Small samples, it says, "do not imply that, for example, 'AAA' rated companies are riskier than 'AA+' rated companies, but rather that both are highly unlikely to default".
The interesting step is BBB to BB, not AAA to AA
Most of the table is flat. From AAA to A, the 10-year cumulative default rate moves from 0.67% to 1.03%. That's a difference of 0.36 percentage points over a decade, spread across three whole rating categories.
From BBB to BB it moves from 2.86% to 10.44%. One notch boundary, the line between investment grade and speculative grade, carries more risk than the entire span above it. In the first year the same shape shows up: 0.14% against 0.56%, four times as often, off a base so small that both look like rounding. By year five they've separated properly, at 1.36% and 5.75%.
That's the arithmetic under the phrase "fallen angel". A downgrade from BBB- to BB+ doesn't change what the company does the next morning. It changes which row of the table the bond belongs to, and the row below behaves differently. Whether that reaches you depends on whether you own the bond or a fund, which the piece on bond funds vs individual bonds works through.
A rating transition matrix tells you more than a default rate does
A default rate reports one outcome. A rating transition matrix reports the whole distribution: where the issuers holding a letter ended up a year later.
On S&P's 1981 to 2024 averages, an issuer rated AAA on 1 January was still AAA a year later 87.28% of the time, and rated AA 8.92% of the time. An issuer rated BBB stayed BBB 87.33% of the time, rose to A 3.05% of the time, fell to BB 3.21% and defaulted 0.14%. An issuer in the CCC/C category stayed there 45.07% of the time and defaulted 26.12%.
Read down the diagonal and stability falls with the rating. That's the part a default rate hides. A BBB issuer downgraded to BB hasn't defaulted, so it never appears in the default column, but the price has already moved and next year's odds now come from a different row.
Withdrawn ratings, not defaults, are the main exit from the pool
Here's the largest hole in every default table, and it sits in the method rather than the data.
S&P's pools contain only actively rated issuers. When a rating is withdrawn, the issuer drops out of later pools. Between 1 January 1981 and 31 December 2024, S&P added 23,831 first-time-rated organisations, excluded 3,556 that defaulted, and excluded 13,432 that were no longer rated. Withdrawals outnumber defaults by close to four to one. S&P does keep watching withdrawn issuers and books a later default back to the pools that issuer belonged to, so the leakage isn't total. It's still the dominant exit.
The March 2026 filings with the US Securities and Exchange Commission make the scale visible, because every agency has to report the same table. Of the 847 corporate issuers S&P rated B on 31 December 2015, 16.77% had defaulted by 31 December 2025 and 68.60% had the rating withdrawn for reasons other than default. Only 3.54% were still rated B. At CCC+ the split was 51.28% defaulted and 43.59% withdrawn, with nobody left at the starting notch.
What happened to the withdrawn majority isn't in the filing. S&P's form has a separate "Paid Off" column and it is empty in every corporate row, so acquisitions, redemptions and issuers who simply stopped buying a rating all land in one bucket.
The same letter at two agencies is not the same number
For the year to 31 December 2025, S&P reported 4,720 corporate issuer ratings outstanding, Moody's 3,170 and Fitch 3,151. The SEC warns against reading much into the totals: "The classification of ratings into the five rating categories is not necessarily consistent across NRSROs."
The letters are more interesting. Over the decade to 31 December 2025, 18% of the 142 issuers Moody's rated B2 defaulted, 50% were recorded as paid off and 4% as withdrawn for other reasons. S&P's comparable B row defaulted at 16.77%, within about a point, but recorded nothing as paid off and 68.60% as withdrawn. The default columns nearly agree. The bookkeeping around them does not, and the bookkeeping is what decides how big the unknown is.
None of that is an accusation. It's the reason default rates by credit rating from one agency can't be pasted onto another agency's letters without a footnote.
At the top of the scale the sample is four companies
The AAA row is the one people quote and the one with almost nothing in it. On 31 December 2024, S&P had 4 corporate issuers rated AAA. Fitch had 2. Moody's had 13 at Aaa. Ten years earlier, at the end of 2015, S&P's AAA cohort was 12 issuers, of which a third were still AAA a decade later, half had moved to AA+ and the rest to AA-.
A 0.00% one-year default rate computed from four companies is a true statement about four companies. It is not a probability estimate, and S&P doesn't present it as one. Its own filing defines an issuer credit rating as "a forward-looking opinion about an obligor's overall creditworthiness", and defines the BBB category as an obligor with "adequate capacity to meet its financial commitments". No number appears anywhere in the definitions. The default rates in this piece are what happened afterwards, counted by the firm that issued the opinions.
Nor does a regulator check the mapping. The SEC's 2025 staff report notes that Congress "did not authorize the Commission to assess whether an applicant's procedures and methodologies for determining credit ratings are well designed or whether the applicant's credit ratings are accurate measures of creditworthiness". That test was handed to the market, via a rule that institutional buyers certify they've used an applicant's ratings for at least three years. The large agencies held 93.38% of all outstanding ratings at the end of 2024.
The strongest objection: the agency also defines the default
The serious criticism of a table like this isn't that the numbers are wrong. It's that both the numerator and the denominator belong to the same firm.
S&P decides which issuers are in the pool, what letter each one carries, and what counts as a default. That last definition has been doing unusual work lately. The number of distressed exchanges in 2024 "rose to its highest level since 2008, accounting for 59.3% of defaults", and S&P treats an exchange as a default when the issuer is distressed and the deal "offers investors materially less than the original promise of the debt". That is a judgement, made by the same firm that set the rating.
The defence is that the ordering keeps working anyway. S&P's own summary measure of rank ordering, the one-year global Gini ratio, was 89.4% in 2024, and there were no defaults that year of issuers rated A or above, "the 15th consecutive year with no defaults from these higher rating levels". Where the ordering does break, it breaks in small subsets. In 2024 the BBB category default rate among emerging and frontier markets issuers, at 0.2%, ran above the BB rate, while globally BB at 0.2% still exceeded BBB at 0.1%.
An average across 44 years hides the cycle that made it
These are averages of 44 separate annual pools, and the annual numbers look nothing like the average. The global corporate default rate was 0.37% in 2007 and 4.15% in 2009. The speculative grade default rate ran from 0.91% to 9.89% across those same two years. In 2024 it was 3.94%, and in 2001 the all-issuer rate was 3.70%.
So a BB bond bought in 2007 and a BB bond bought in 2008 faced very different odds, and neither faced the long-run average. S&P's own filing explains why the two views can't be blended. The annual study uses an average cohort method and the SEC filing a single cohort, and "both approaches are valid, depending on the needs of the user, but they do not yield comparable information".
Timing shows up a second way. Among global corporate defaulters from 1981 to 2024, the average gap between carrying a AAA rating and defaulting was 27.4 years. From BBB it was 8.8 years. From CCC/C it was 0.9 years. A high rating that eventually fails does so slowly, through a long chain of downgrades, so the letter on the bond at default is rarely the letter you bought.
What the market pays for the same letters right now
A default rate is half the calculation. The other half is what you're paid to carry it.
On 18 August 2026 the ICE BofA BBB US Corporate Index traded at an option-adjusted spread of 1.00 percentage points over Treasuries. The CCC and lower US high yield index traded at 10.27. Set those against the table: BBB has defaulted at 0.14% a year on average and pays 1.00, while CCC/C has defaulted at 26.12% a year and pays 10.27. Per £100 of exposure the CCC index pays £9.27 a year more than the BBB index, against 25.98 percentage points more annual default frequency on the long-run record.
The second comparison isn't as lopsided as it looks, because a default is not a total loss and the spread pays for more than expected default. It also pays for liquidity, for uncertainty about recovery, and for the chance that realised losses run above the long-run average. The piece on credit spreads takes that decomposition apart, and the one on high-yield spreads follows what the CCC tail does when dispersion widens.
What this table cannot tell you
Start with the unit. Default rates by credit rating are issuer counts, not money. S&P is explicit: "All default rates that appear in this study are based on the number of issuers rather than the dollar amounts affected by defaults or rating changes." A pool in which the single largest issuer defaults and every small one survives reads as a low default rate whatever the money says.
The sample is not the bond market either. Only rated issuers appear, so private credit, unrated companies and small borrowers are absent by construction. The rated population itself turned over heavily across the 44 years, with 23,831 issuers joining the pools and 13,432 leaving them unrated.
Geography is uneven too. On S&P's regional tables the US BBB 10-year cumulative default rate is 3.80% and Europe's is 1.43%. Same letter, same window, different populations, and the global figure of 2.86% sits between the two while describing neither.
And the whole thing looks backwards. Forty-four years is a long sample and it still isn't a forecast. It contains two severe credit cycles and long quiet stretches like 2007, when the all-issuer rate was 0.37%, and the next 44 years don't have to rhyme with it.
What would change the conclusion
If withdrawals stopped being a black box. More than two-thirds of B rated issuers left S&P's 10-year window without defaulting and without a stated reason. If the agencies split "repaid" from "stopped paying for a rating", the long-horizon cells would either firm up or move, and nobody outside the agencies can say which.
If the top of the scale keeps emptying. S&P's AAA corporate cohort went from 12 issuers at the end of 2015 to 4 at the end of 2024. Carry that on and the AAA and AA rows become historical artefacts rather than live estimates, and the practical top of the scale becomes A.
If a cycle arrives that the sample doesn't contain. The two worst years on record ran at 3.70% in 2001 and 4.15% in 2009, and both sat almost entirely in the speculative grade tail: investment grade defaulted at 0.23% and 0.33% in those years, against 9.70% and 9.89% for speculative grade. A default wave that started in investment grade instead would break the rank ordering the table rests on, and no year in the 44 looks like that: the highest investment-grade default rate in the whole sample is 0.42%, in 2008.
The figure worth tracking isn't the headline default rate. It's the share of rated issuers sitting at CCC+ and below, because that's the row where 26.12% default inside a year, and where a cycle turns up first.