Rating Agenices: Can they keep growing?

John Moody—a former bank clerk with a taste for statistics—nearly went broke in the Panic of 1907 when printing costs for his encyclopedic railroad manuals outran subscriptions. Two years later he stripped the tome down to a single table of alphanumeric grades—an embryonic version of today’s “Aaa” through “C”—and investors snapped up the $50 booklet faster than he could ink them.

Not to be outdone, Henry Varnum Poor’s heirs launched their own ratings guide in 1916 from a Pearl Street brownstone. Legend has it that salesmen toted the cloth‑bound volumes on the Boston & Albany line, pitching trust‑company officers between Worcester and Back Bay.

Across the Hudson, John Knowles Fitch debuted an even plainer “AAA”‑to‑“D” ladder in 1924 and stapled sample pages to the cover of his Fitch Stock and Bond Manual, betting that a free taste would hook paying readers.

Fast‑forward to 2024 and S&P Global’s ratings arm rang up roughly $4.2 billion in revenue—up 31% from 2023 and the fastest growth inside the company. Rival Moody’s Investors Service wasn’t far behind, posting $3.79 billion. Fitch Ratings, now wholly owned by Hearst, generated an estimated $1.7 billion. Together, the “Big Three” still control about 95% of the global ratings pie, leaving only crumbs for privately held challengers such as DBRS Morningstar and Kroll Bond.

Where is the Moat?

Regulatory cement: Since 1975 the Securities and Exchange Commission has blessed only a handful of firms as Nationally Recognized Statistical Rating Organizations (NRSROs). Just 10 hold the license today, and winning one can take years of petitions, onsite inspections and capital‑adequacy tests. Europe’s ESMA and Japan’s FSA run similarly tight gates. Banks, insurers and many mutual funds must use those ratings to satisfy capital rules, so the NRSRO badge itself is a fortification.

Network effects: Roughly $44 trillion of global bonds carry covenants, margin agreements or index rules that hinge on an S&P or Moody’s level—down one notch and you may face forced selling. Bloomberg tracks more than 50,000 such “ratings triggers.” That embedded code locks in demand for incumbents’ alphabets and makes any newcomer’s scale feel like Esperanto.

Data flywheels: When Moody’s publishes its annual default study, it draws from 119 years of outcomes; S&P’s Cumulative Average Default Rates reach back to 1921. Fitch’s loan‑recovery tables cover every leveraged borrower since the 1980s. Those deep archives feed predictive models and stress tests that new entrants simply cannot replicate without decades of seasoning—especially through crises.

Brand and legal safe harbor: In the U.S. an issuer can file a “well‑known seasoned issuer” shelf if at least one of the Big Three assigns investment‑grade status. Indenture definitions of “investment grade” typically cite Moody’s or S&P by name, conferring a quasi‑statutory imprimatur. Courts have also treated ratings as a form of opinion journalism, giving agencies First‑Amendment breathing room that would be harder to win anew.

Switching costs for issuers. A corporate treasurer eyeing a better price can shop banks overnight; swapping rater is messier. Investors discount an unrated or little‑rated bond, so an issuer must often pay for a Big‑Three grade anyway. CLO indentures, for instance, demand at least one major rating on every underlying tranche. In effect, the customer is captive even when the service disappoints.

Put simply, ratings are the lingua franca of global debt; nobody wants to publish a prospectus in Esperanto.

Threats to the Moat: Regulators

Regulators with long memories still haunt the agencies; Washington and Brussels, chastened by the mortgage‑ratings debacle of 2008, are exploring ways to curb the issuer‑pays model and have even floated a publicly backed competitor in Europe. Private credit’s rapid ascent is another headache: direct‑lending deals—on course to top $1.8 trillion by 2027—rarely bother with public ratings, narrowing a lucrative fee stream.

Why Have Breakup Efforts Stalled?

Politicians on both sides of the Atlantic have pledged for decades to pry open the ratings market, yet genuine rivals remain bit players. The core problem is network credibility: asset‑management mandates, swap‑margin tables, index‑eligibility rules and central‑bank collateral frameworks all cite Moody’s and S&P by name. Any newcomer must persuade thousands of counterparties—and the lawyers who draft their boilerplate—to rewrite centuries of fine print, a herculean coordination chore no regulator has forced.

Capital intensifies the barrier. Moody’s says it has spent more than $2 billion compiling 119 years of default and recovery data; S&P’s archive runs even deeper. Re‑creating that statistical bedrock from scratch is beyond most venture budgets. Even firms that did clear the NRSRO gauntlet—Egan‑Jones, DBRS Morningstar, Kroll—together command barely 5% share after decades in the field, largely because investors still demand a Big‑Three seal before they will quote a price.

Europe’s experience illustrates the pitfalls. In 2012 Brussels flirted with a taxpayer‑backed “European Rating Agency” to counter perceived U.S. bias during the sovereign‑debt crisis. Member states balked at footing the bill, and the idea collapsed under quarrels over governance and legal liability. Scope Ratings of Berlin, launched in 2002, enjoys rhetorical support from Berlin and Paris but covers only about 800 issuers; most European borrowers still hire an American agency so their paper qualifies for dollar‑denominated indices. Ironically, ESMA’s post‑crisis rulebook—designed to discipline the Big Three—raised compliance costs for all, burdening would‑be challengers with the same heavy reporting load.

As a result, global share has barely budged since Lehman fell. Dismantling the moat would likely require regulators to rewrite capital rules to accept new alphabets—or the bond market to stop caring about ratings altogether, a shift that even the booming private‑credit sector has yet to prove.

Threats to the Moat: AI

Technology pressures are the newest battering ram. London‑based Credit Benchmark uses machine learning to fuse the internal risk views of more than 40 global banks into consensus grades on 115,000 borrowers—roughly 90 percent of them unrated by the majors. Toronto’s Overbond ingests live bond prices, liquidity ticks and macro data, then lets neural nets spit out implied ratings and downgrade alerts across a universe of 100,000 securities. And bondIT’s Berlin‑Israeli “Scorable” engine employs explainable‑AI models to flag rating changes months before the agencies move; Liquidnet just embedded the feed in its trading terminals. Large asset managers are also tinkering with proprietary risk scores that shave fees and sidestep legacy scales.

Meanwhile, the scramble to bolt environmental grades onto legacy ratings has drawn political fire, threatening fresh regulation just as sovereign issuers in China and India nurture local champions to guard home turf. None of these threats will topple the wall tomorrow, but each chisels a brick from an edifice that once looked impregnable.

The big agencies insist they will meet algorithms with algorithms. Moody’s has assembled a 300‑person data‑science team and spent more than $500 million on tech acquisitions from Bureau van Dijk to PassFort. S&P paid nearly $550 million for Kensho’s natural‑language engines and now feeds them transcripts from 20,000 earnings calls a year. Fitch is rolling out a large‑language‑model companion called “Fedora” that drafts surveillance notes on 10,000 issuers overnight. The incumbents argue that AI without proprietary data is like a Ferrari without gasoline, and they own a century’s worth of fuel.

Yet speed cuts both ways. The same models that help agencies flag anomalies faster also let challengers reverse‑engineer scores at marginal cost—turning what was once a $300,000 rating fee into a $30‑per‑month API call. If that price compression spreads, the agencies’ 60‑plus‑percent operating margins could look as vulnerable as newspaper classifieds did in 2005.

Financial Scorecard

Raters still print money when bond printers run. Global issuance rebounded almost 40% in 2024 as refinancing roared back, lifting S&P’s transaction fees 54% and Moody’s 42%. Margins swelled—to roughly 63% at S&P’s ratings unit and 60% at Moody’s—because payroll and cloud costs rise far slower than deal volume. Even Fitch, smaller and more exposed to sovereign work, benefited from a rush of CLO paper.

But cyclicality cuts both ways. If the Federal Reserve keeps rates “higher for longer,” underwriting desks slow, and ratings revenue can dry up in weeks. In 2022—a lean year for issuance—S&P’s ratings sales fell 18%.

Balance‑sheet muscle helps. S&P threw off $5.7 billion in free cash last year and returned more than $4.4 billion to holders via dividends and buybacks. Moody’s free cash flow topped $2 billion, enough to keep its 14‑year streak of double‑digit dividend hikes alive.

Conclusion

Investors betting on the raters are buying a paradox: a near‑monopoly that is also a cyclical play on capital markets. The moat—regulatory, reputational, historical—remains formidable, but the mortar now faces its sternest test since the photocopier era: an AI arms race that could democratize credit scoring the way Google democratized information.

Regulators will shape the outcome. Brussels is mulling disclosure rules for algorithmic ratings that could punish black‑box providers but also strip incumbents of their opacity premium. In Washington, the SEC’s climate‑risk proposal contemplates machine‑readable emissions data—an invitation for software to rate what analysts once did by hand. Legal safe harbors may erode if courts decide that a fully automated grade is more model than opinion.

For now, fat cash flows, contractual lock‑ins and that priceless data trove give the Big Three time to adapt. Their most likely path is co‑option: buy the best AI shops, pour their datasets into the engines and sell the output back to the market as premium insight. Profit margins would compress, but the franchises could endure—much as stock exchanges survived electronic trading by becoming technology vendors.

The real crack in the wall would come only if the bond market itself—index rules, collateral schedules, covenant triggers—accepted alternative alphabets at scale. That remains a distant prospect. As one veteran bond banker put it, “You only need a rating until you can teach the market a new language.” The agencies still control the dictionary, but every new algorithm scribbles in the margins. Whether the moat survives will hinge less on code than on whether investors ever decide those scribbles read like gospel. Find out if one of the rating agenices makes it to our Global Quality portfolio.