Credit ratings look like measurements and are formally opinions, and the gap between those two things has caused real damage.
What is being rated
The likelihood that a borrower will fail to meet its obligations, and in some scales the expected loss if it does.
Ratings apply to sovereigns, corporations, financial institutions and structured securities, using scales that are broadly comparable and not identical between agencies.
The distinction between investment grade and below is the most consequential single boundary, because of what depends on it.
Why the boundary matters so much
Regulation and investment mandates reference ratings directly.
Capital requirements for banks and insurers depend on ratings of what they hold.
Many funds are prohibited from holding below-investment-grade securities by their own mandates.
Which means a downgrade across that boundary triggers forced selling regardless of any investor's own view, and the price effect can be severe.
Regulators have worked to reduce mechanical reliance on ratings since the financial crisis, with partial success.
The issuer-pays problem
The structural conflict at the centre of the business.
The entity being rated generally pays for the rating.
Which creates an obvious incentive problem, since the customer prefers a higher rating and can take its business elsewhere.
Agencies argue that reputation is their entire asset and that inflating ratings would destroy it, which is a genuine counterweight and was insufficient in the structured finance episode.
What went wrong with structured products
Complex securities built from pools of mortgages received top ratings and subsequently defaulted at rates the ratings implied were near-impossible.
The failures included models assuming house price declines could not occur simultaneously across regions, insufficient data on the underlying loans, and competitive pressure between agencies for issuer business.
Investigations found evidence of pressure on analysts and of concerns raised internally and not acted on.
Settlements followed and the business model did not change.
The alternatives
Investor-pays models exist and struggle commercially, because ratings are difficult to keep exclusive once published.
Proposals for assignment of raters by a public body, removing issuer choice, have been discussed and not implemented at scale.
Reducing regulatory reliance so that ratings matter less is the approach regulators have actually pursued.
Sovereign ratings
A particular source of controversy.
Downgrades of sovereigns during crises have been criticised as procyclical, worsening the conditions being assessed by raising borrowing costs.
Agencies respond that they report deterioration rather than causing it, which is true and does not resolve the feedback effect.
The methodology involves substantial qualitative judgement about political stability and institutional quality, which is why sovereign ratings attract accusations of bias more than corporate ones.
How to read a rating
As one input reflecting a specific assessment of default risk, produced by an interested party, using a published methodology.
The methodologies are available and are more informative than the letter.
Outlook and watch designations indicate direction and are frequently more informative than the current rating, since they signal expected change.
And market pricing frequently moves before ratings do, which means spreads carry information that ratings have not yet incorporated.
Unsolicited ratings
Ratings issued without the issuer requesting or paying for them.
Which are less common and have been criticised from both directions — as a pressure tactic to win business, and as a valuable independent check.
Disclosure requirements now generally require unsolicited status to be stated.
Smaller agencies
The market is dominated by a small number of firms, which regulators have identified as a concentration concern.
Registration regimes were intended partly to lower barriers to entry, and market share has shifted little.
The obstacle is that regulatory and mandate references frequently name specific agencies or require recognition that smaller firms find difficult to obtain, which entrenches incumbents.
Structured finance ratings now
The methodologies have been substantially revised, with more conservative assumptions and more disclosure of underlying loan data.
Regulators require issuers to publish loan-level information in several jurisdictions, which allows independent analysis rather than reliance on the rating alone.
Whether the incentive problem has been addressed is a different question, and the answer is that it has been mitigated rather than removed.
Internal ratings
Large institutions maintain their own credit assessments alongside agency ratings.
Regulatory frameworks permit banks to use internal models for capital purposes subject to approval, which reduces the mechanical reliance on external ratings.
The criticism of internal models is that firms have an incentive toward assumptions producing lower capital requirements, which regulators address through review and through floors on the results.
Which relocates the incentive problem rather than removing it, and both approaches remain in use for that reason.
Environmental and governance ratings
A separate and growing category, assessing non-financial characteristics.
Correlation between providers' scores for the same company is notably low, which indicates they are measuring different things under the same label.
Regulation of these providers has been introduced in some jurisdictions, largely focused on transparency of methodology rather than on standardising it.