Suspicion over the credibility of private ratings is nothing new. Back in 2024 the National Association of Insurance Commissioners published a report casting aspersions on “private letter ratings” which, as well as causing a huge stir in the industry, gave us something to giggle about.
The NAIC report was withdrawn following accusations of selection bias that may have skewed the maths. But at the end of May a new paper was published, this time by a team of finance academics, that appears to make similar claims.
Actually, two papers came out on the exact same day — one by a group of academics based in New York and one based in London.
On Friday, Gillian Tett referenced a paper by Xuelin Li, Sangmin Oh, and Giacomo Ricciardi of Columbia Business School. Looking at insurers’ holdings of both public and privately rated bonds, they reckoned privately rated ones are associated with the kinds of impairments typically suffered by publicly rated bonds two-to-three notches lower. And this, they think, translates into maybe a $4.5bn industry-wide capital shortfall.
But we want to focus on the paper from Isabella Gschossmann, Ziang Li, Ali Uppal (all of Imperial College London) and Derek Wenning (of the Kelley School of Business, Indiana University).
This one sifts through a security-by-insurer panel organised at the insurer-security-quarter level, with around 16mn observations on 284,676 bonds held by 772 US life insurers for the 2020-24 period. Bottom line: they find that private ratings are systematically understating investment risk in the private credit market. And this, they reckon, means insurers are holding about $8bn less capital than they really should.
Even Alphaville can occasionally struggle to get excited when insurance-types talk giddily about capital charges. But they’re important, so let’s have a go at worth unpicking why.
Capital 101
Insurance capital is the money insurers bring to the game. If I’m a brand new insurer ploughing money into a marketing machine that will lead ordinary Americans to buy a gazzilion dollars of new fixed-rate annuities, capital is the deciding factor that limits how much business I can sell.
Insurers incur so-called capital charges based on a bunch of things. C-0 charges guard against the risk that their subsidiaries blow up. C-2 charges provide some margin of safety that assumptions used to write insurance turn out dud. C-3 is a capital buffer against market moves. And C-4 is a sort of cushion required for ‘everything else’ — operational risks, legal risks, etc.
But the biggest source of capital requirements comes from C-1 risk charges. These exist to make sure an insurer is good for the cover they’ve written even if bad things happen to their assets. And C-1 charges are based on the riskiness of assets that insurers buy with their capital and the premiums that they collect.
Some content could not load. Check your internet connection or browser settings.
In a 2024 paper Florian Weinlich estimated that over half of US life insurers’ risk-based capital in 2019 was held to guard against bad things happening to their assets, and almost three-fifths of this was attached to their fixed-income holdings.
As you’d maybe expect, an insurer holding bonds of shakier companies will need more capital than one holding bonds only of super-safe companies. And lower capital charges = better for a return-maximising insurer, all else equal.
But all else is hardly ever equal. Bonds issued by shakier companies tend to have higher yields — so the trick to maxxing out bang for capital buck is to hold bonds (that don’t default) combining the highest yields and the lowest capital charges. In the case of insurers, these tend to be investment grade.
Who gets to decide how shaky (or otherwise) a company is, and how large the C-1 capital charge is? Regulators. And who do regulators trust? Rating agencies.
Some content could not load. Check your internet connection or browser settings.
In ye olden days, the NAIC operated a six-category C-1 capital charge schedule, running from NAIC 1 to 6. Bonds rated AAA would fall into NAIC 1 and attract a 39bp capital charge. But so would bonds rated A-. Bonds rated BBB+ — just one rung below — fell into NAIC 2 and attracted a 126bp capital charge, and so on.
Regulators therefore incentivised insurers to load their portfolios with NAIC 1/ single-A minus bonds, which packed the highest yields for the lowest capital charge. But if these got downgraded, any insurer trying to maximise its balance sheets would be forced to pony up fresh capital or ditch the holding.
By 2021 this was generally understood to be nuts. And so the six NAIC categories were replaced by a more granular 20-category schedule in a bid to reduce risk. Under the new regime, a AAA-rated bond attracted a 16bp capital charge, and a A- bond attracted a 102bp charge. This looked far more sensible, seemingly reducing the incentive for returnmaxxing insurers to stop dangling their balance sheets on the edge of capital/rating cliffs.
But there was nothing in the regs to specify whether a rating would be public or private. And, as the academics argue, the notion that the reform reduces risk “depends on taking private ratings at face value.”
Back to private ratings
If private ratings were more generous than public ones you’d expect to see them proliferate post-2021. This way, insurers could smuggle riskier bonds into their balance sheets without having to find additional capital to grow their business.
And lo:
Some content could not load. Check your internet connection or browser settings.
Since 2021 private ratings have proliferated, more than doubling as a share of US life insurers’ holdings.
Some insurers have used them more than others. We don’t know why this might be the case, though it’s worth noting that it’s not just — or even mostly — private equity-backed insurers that have made heavy use of them:
Some content could not load. Check your internet connection or browser settings.
The researchers found that bonds with private ratings yield around 180 bps more than public ones. Fair enough — privately-rated debt tends to come with more complex structures, callability rinkydinks, etc, and might also be issued by shakier companies.
Indeed, the researchers also found that privately rated bonds on the whole attracted higher capital charges than publicly rated bonds. And this means that private ratings were, on average, lower than public ratings.
Some content could not load. Check your internet connection or browser settings.
However, once the academics start trying to compare like-with-like, the picture changed.
Looking only at bonds issued by the same issuers, they found that the ones bearing private ratings carried far lower average risk charges — associated with higher average ratings. And controlling for rinkydinks like structure, complexity, security etc made almost not a jot of difference.
Most strikingly, they found instances of individual bonds that were differently rated. For these ca 1,000 bonds, issued by around 500 companies — which showed up in year-end filings as having flipped between being privately and non-privately rated — the private ratings were, on average, higher. And this meant the risk charges on the privately rated bonds were lower, by about 1.76 per cent on average.
There can be all sorts of reasons why bonds from the same issuer can have different ratings. And the researchers don’t go into the weeds to try to tease out whether individual securities are correctly or incorrectly rated or even reconcile the differences. They just look at the overall pattern, and point to it as looking a bit sus.
This is how the authors build to their conclusion that “bonds with private ratings are inflated”, and that the industry holds around $8bn less capital at the end of 2024 than they would if everything was publicly rated.
By their calculation, $8bn amounts to around 4 per cent of industry capital — around twice the estimate arrived at by the Columbia academics. Is 4 per cent of industry capital really worth us throwing our hands in the air and proclaiming end of days?
KBRA — a rating agency that specialises in issuing credit ratings for complex structured transactions — appears to think not. In a note its analysts took issue with the Columbia academics’ reliance on broad extrapolation, with their choice of rating performance, and a number of other things. They noted that:
Based on NAIC’s 2024 Life RBC Statistics, that amount represents roughly 0.5% of industry total adjusted capital (TAC), approximately 0.6% of industry surplus, and less than 0.1% of invested assets. In other words, a small pro forma impact on the industry’s risk-based capital ratio (RBC) of approximately 38 RBC points (830% versus 868%).
In other words, no systemic risk here.
Still, this is real money. And if regulatory arbitrage is occurring systematically on a scale that increases year on year, it’s probably worth regulators taking a closer look. After all, as team-Columbia put it “the potential [capital] shortfall is ultimately borne by those the insurer owes rather than by the insurer itself.”
Unfortunately, insurance in the US is regulated individually by each state, and the quality, staffing and assiduousness of the 50 Departments of Insurance can be . . . variable. So the chances of anything happening from that front are negligible.
Further reading:
— What’s up with private credit ratings? (FTAV)
— Why are people being mean about Egan-Jones? (FTAV)
— Rating the rating data of credit raters (FTAV)
