What does a credit rating actually claim to measure?
A credit rating is a letter — a published opinion about a borrower's ability and willingness to meet its obligations. It is enormously influential and routinely misread, so it pays to be precise about what it claims and what it does not.
The scale, generically described
The major global agencies use scales that differ in punctuation but agree in shape. From the top:
- A triple-A category — the highest opinion of creditworthiness.
- Then double-A, single-A, and triple-B.
- Then double-B, single-B, triple-C, and below.
- Finally a default or near-default category.
Within most categories there are three notches, marked either with plus and minus signs or with the digits 1, 2 and 3. So the step from the middle of the triple-B category to its lowest notch is one notch, not one category.
The scale is ordinal. It ranks. A double-A is judged more creditworthy than a single-A; the letters do not claim that the gap between them is the same size as any other gap.
Issuer ratings versus issue ratings
Two different objects share the same alphabet:
- An issuer rating describes the borrower as a whole — how likely it is to fail to pay. The agencies differ on exactly what that letter measures: some scales are defined purely as default-probability opinions, while at least one major scale is defined as an expected-loss opinion and therefore already blends in how much would be lost if default happened.
- An issue rating describes a specific bond, and factors in seniority and expected recovery. A secured bond from an issuer may be rated a notch or two above the issuer rating; a subordinated bond a notch or two below.
On a default-probability scale, this is the ratings system encoding the same PD-and-LGD split you met in the expected loss arithmetic: one letter for how likely default is, the other adjusted for how much you would lose if it happened. On an expected-loss scale, both pieces already sit inside the issuer letter and the notching refines the loss half instrument by instrument. Knowing which philosophy a scale follows is part of reading it — and it is one reason the same borrower can carry different letters from different agencies without either of them being wrong.
What the letters imply about default rates
Agencies publish long-run default studies, and the shape is consistent even though the exact numbers vary by study, period and methodology. Illustratively and heavily rounded, five-year cumulative default rates run from a fraction of a percent at the top of the scale, to low single digits around the triple-B area, to the mid-teens or higher for single-B credits, and far higher below that.
The important feature is not any single figure but the convexity: default rates do not rise linearly as you descend. They accelerate. Two notches down near the bottom of the scale changes the loss arithmetic far more than two notches down near the top.
Ratings move slowly. Spreads move constantly.
Agencies rate through the cycle: they deliberately try to look past ordinary ups and downs and avoid whipsawing an issuer's rating with every quarter. Market spreads do the opposite — they reprice by the second on any news at all.
So ratings and spreads disagree constantly, and both are informative. A bond trading much wider than peers with the same rating tells you the market has formed a view the agencies have not acted on, or that something non-credit — liquidity, size, complexity — is in the price. That disagreement is an observation worth investigating, not a signal to act on.
Agencies also publish outlooks (a direction of travel over a year or two) and watch or review designations (a shorter-fuse flag, often tied to a specific pending event). Both are part of the opinion.
What a rating is not
It is not a probability. It is not a measure of price, volatility or liquidity. It is not a recommendation to buy or sell — the agencies say so explicitly. And it is not a guarantee: ratings are opinions, they are revised, and they are sometimes revised late.
In the data
The table holds six governments' ratings from Moody's, S&P and Fitch. They are letters, not numbers: Aa1, AA+ and AA+ for the United States.
Nothing arithmetic works on them. Ratings cannot be averaged, interpolated or differenced, and the distance between Aa1 and Aa2 has no defined size. Every row also carries the same once-a-year date: this is an opinion collected on a schedule, not a measurement taken continuously.
Try it now
Connect letters to numbers:
- In the table above, translate each letter to a notch on a common scale (Aaa = AAA, Aa1 = AA+, Baa3 = BBB−) and note which countries carry different letters from different agencies. One of them sits on opposite sides of the investment-grade line depending on whose letter you read.
- The same six countries, priced by the market: what it costs each year to insure their debt, as a credit default swap spread. Take two several notches apart and compare their spreads in basis points.
- Check whether the spread ordering matches the rating ordering. Where it does not, write down one question about why — that question is the beginning of credit analysis.