How Credit Rating Systems Emerged

The fundamental problem that credit rating systems exist to solve is ancient but the institutional solution is surprisingly recent. When a lender considers extending credit to a borrower, the lender faces an information problem: the borrower knows far more about their own financial condition, intentions, and ability to repay than the lender can easily discover. This information asymmetry creates adverse selection — if interest rates are set at a level that reflects average borrower quality, good borrowers find the rate unattractive and drop out, leaving the pool of borrowers increasingly concentrated in poor credit risks. The market for credit, like the market for used cars that George Akerlof analyzed in his famous “lemons” paper, can fail precisely because the party with better information can exploit that advantage at the expense of the counterparty. Solving this problem — or at least reducing it to manageable scale — is the economic function of credit rating.

Modern economies generate credit on a scale that makes direct bilateral assessment by each lender of each borrower practically impossible. A bank making one mortgage loan can investigate the borrower thoroughly. A mutual fund investing in a portfolio of thousands of corporate bonds cannot independently assess each issuer’s creditworthiness at acceptable cost. Capital markets depend on mechanisms that allow investors to delegate the credit assessment function to specialists who can develop information, analytical capacity, and reputational accountability that individual investors cannot maintain. This is the market niche that credit rating agencies occupy — they are information intermediaries whose product is a standardized assessment of credit quality that allows investors to screen large numbers of issuers efficiently.

The origins of formal credit reporting in the United States are traceable to Lewis Tappan, a New York abolitionist and silk merchant who founded the Mercantile Agency in 1841. Tappan’s insight was that merchants extending trade credit to customers in distant cities faced exactly the information asymmetry problem: they needed to know whether a retail merchant in Cincinnati or New Orleans was creditworthy before shipping goods on 60- or 90-day credit terms. Reputation information existed in local commercial communities where merchants knew each other, but it did not travel well across the distances of a continental economy. Tappan’s solution was a national network of correspondents — lawyers, local merchants, and other well-connected individuals — who reported on the credit standing and business reputation of merchants in their localities. Subscribers to the Mercantile Agency paid for access to this accumulated intelligence, receiving assessments of specific merchants’ creditworthiness on request.

The business model rested on a fundamental asymmetry: Tappan’s correspondents collected information from many sources about many merchants, aggregated it, and sold access to this aggregated intelligence at a price below what individual merchants would have paid to acquire comparable information independently. This is the classic economics of information goods: the cost of producing the information is fixed (roughly), while the cost of distributing it to an additional subscriber is close to zero, so spreading fixed collection costs across many subscribers generates large per-subscriber savings. The Mercantile Agency grew rapidly, and its competitors — most importantly Robert Dun, who eventually acquired the Mercantile Agency’s successor, and John Bradstreet, who founded his own rival firm in 1849 — confirmed that the market was real. Dun and Bradstreet merged in 1933 to form the company that still bears their names.

The rating of financial securities — bonds issued by governments and corporations — developed through a parallel but distinct institutional path. The railroad bond market of the mid-nineteenth century created the immediate practical demand. American railroad construction was enormously capital-intensive, and individual railroads could not finance it from retained earnings or bank loans alone. They issued bonds to the public, and those bonds were traded on secondary markets where investors who had not participated in the original issuance bought and sold them. Investors needed some basis for evaluating railroad bond quality beyond simply trusting the issuing company’s representations, and the financial press — particularly Henry Varnum Poor’s Manual of the Railroads of the United States, first published in 1868 — developed detailed standardized financial statistics on individual railroads that gave investors comparable data.

John Moody made the decisive innovation in 1909, when he published Moody’s Analyses of Railroad Investments and introduced letter-grade ratings — Aaa, Aa, A, Baa, and so on down — to summarize his assessment of bond credit quality in a single symbol. The rating was an opinion, not a guarantee, but it compressed complex financial analysis into a form that investors could use without themselves performing the underlying work. Poor’s Publishing Company (the ancestor of Standard & Poor’s) and Fitch adopted similar rating scales. For the next several decades, ratings agencies grew steadily as the bond market expanded and as institutional investors — insurance companies, pension funds, savings institutions — increasingly faced regulatory requirements to hold only “investment grade” bonds, a category defined by reference to rating agency assessments. The agencies had moved from being informational conveniences to being gatekeepers whose assessments had legal and regulatory force.

The economic logic of the early rating system was sound in structure if imperfect in execution. Agencies charged investors — the subscribers who received rating books — for access to their assessments. This investor-pays model aligned agency incentives with investor interests: agencies that produced consistently accurate ratings would retain and grow their subscriber bases, while agencies that failed to identify credit deterioration or that awarded undeservedly high ratings would lose credibility and business. The reputational mechanism was imperfect — rating errors often became apparent only slowly, and agencies could maintain subscriber loyalty through switching costs and incumbent advantage even with mediocre accuracy — but the basic incentive structure pointed in a useful direction.

The transformation of this incentive structure occurred gradually through the 1970s as the rating market consolidated and as agencies shifted from the investor-pays to the issuer-pays business model. The Securities and Exchange Commission’s designation of Moody’s, S&P, and Fitch as Nationally Recognized Statistical Rating Organizations (NRSROs) in 1975 created an oligopoly with regulatory backing, reducing competitive pressure on the established agencies. Simultaneously, as bond markets grew and issuers became more sophisticated about the rating process, agencies found that charging issuers for ratings was more lucrative than charging investors — issuers had strong incentives to pay for ratings that would allow them to access capital markets, while investors could often free-ride on publicly available rating information. By the 1980s, the issuer-pays model had become standard.

The conflict of interest embedded in the issuer-pays model is not subtle. An agency paid by the entity it is rating has a financial incentive to assign favorable ratings, because a dissatisfied issuer can take its business to a competing agency. This “ratings shopping” dynamic — where issuers approach multiple agencies and reveal their planned structures only to those likely to assign favorable ratings — pushes ratings quality systematically downward. Each agency faces the choice between maintaining standards (and losing business to less scrupulous competitors) or relaxing standards (and retaining the fee revenue). In a competitive market with concentrated players and reputational capital to protect, this might be self-limiting; but as structured finance grew in the 1990s and 2000s, the complexity of the products being rated made it increasingly difficult for any outsider to assess whether the ratings were accurate, reducing the reputational consequences of error.

The structured finance crisis of 2007-08 revealed in dramatic fashion how thoroughly the incentive problem had corrupted the rating system. Mortgage-backed securities and collateralized debt obligations — financial instruments that pooled large numbers of individual mortgages and sliced the resulting cash flows into tranches of varying seniority and credit exposure — required rating agency opinions to be sold to the institutional investors who dominated the buyer market. Rating AAA tranches of CDO structures as equivalent in credit quality to AAA corporate bonds or Treasury securities was the necessary predicate for the instruments’ marketability; without the AAA rating, the instruments could not be held by pension funds, insurance companies, or money market funds whose mandates restricted them to investment-grade assets.

The ratings on structured mortgage products turned out to be wildly incorrect. Securities rated AAA suffered catastrophic losses — not the near-zero loss rates implied by the rating, but losses of 50% or more on nominally senior tranches. The agencies’ models for assessing structured product credit quality had relied on historical mortgage loss data from periods before the 2000s housing bubble, had assumed geographic diversification would prevent correlated losses (an assumption violated when housing prices fell simultaneously across the country), and had used correlation assumptions provided by the issuers themselves rather than independently derived. The result was a systematic overrating of structured products on a scale that channeled trillions of dollars of capital into instruments whose actual risk was not remotely captured by their ratings.

The agencies did not discover in 2007 that their models were wrong; they had received warnings from analysts who questioned the assumptions, and internal emails published in subsequent congressional investigations revealed awareness at various levels that competitive pressure was driving rating decisions that methodological rigor could not support. The famous formulation from one S&P analyst — “let’s hope we are all wealthy and retired by the time this house of cards falters” — captured the institutional knowledge that standards had been compromised and that consequences would eventually follow. The issuer-pays model had done exactly what its critics predicted: it created a race to the bottom in which the agencies most willing to accommodate issuer pressure grew their structured finance revenues most rapidly, while those that maintained higher standards lost market share.

What the history of credit rating systems reveals is a case study in how institutional solutions to information asymmetry problems can themselves become information asymmetry problems. Lewis Tappan’s Mercantile Agency existed to give merchants accurate information about counterparty creditworthiness; its value depended entirely on the accuracy of the information it provided. When the agencies that descended from that tradition acquired regulatory gatekeeping status and shifted to the issuer-pays model, the incentive to provide accurate information was replaced by the incentive to provide favorable information — and in a market where product complexity made independent verification nearly impossible, favorable information was what the market received.

The deeper lesson is about the conditions under which information intermediaries can maintain their integrity. Intermediaries whose reputation is their primary asset and who face informed, monitoring customers have strong incentives for accuracy. Intermediaries whose market position is protected by regulatory designation, whose products are too complex for customers to verify, and whose revenues depend on the satisfaction of the entities they assess are systematically compromised. The credit rating agencies in 2007 had all the characteristics of the second type: regulatory protection that insulated them from competitive accountability, products complex enough to resist independent verification, and business models that made them financially dependent on the issuers they were rating. The system had evolved from an information solution to an information problem — and the financial system paid the price.

The institutional lesson has not been fully absorbed. Post-crisis reforms attempted to reduce mandatory regulatory reliance on ratings, require greater disclosure about rating methodologies, and create liability for materially inaccurate ratings. But the fundamental issuer-pays structure remains in place, and the oligopolistic market structure has if anything become more concentrated. The same agencies that catastrophically misrated structured products in the pre-crisis period remain the regulatory gatekeepers for capital market access today. The history of credit rating demonstrates that information institutions, once captured by the interests they are supposed to assess, are extraordinarily difficult to reform without dismantling and rebuilding from different first principles.