In our recent ECGI working paper, A Strict Liability Regime for Rating Agencies, we study how to induce Credit Rating Agencies (CRAs) to produce ratings as accurate as the available forecasting technology allows.
Referring to CRAs, Paul Krugman wrote that: “It was a system that looked dignified and respectable on the surface. Yet it produced huge conflicts of interest. Issuers of debt […] could choose among several rating agencies. So they could direct their business to whichever agency was most likely to give a favorable verdict, and threaten to pull business from an agency that tried too hard to do its job.”
However, the conflicts of interest stemming from the issuer-pays model and rating shopping by issuers are not sufficient to explain rating inflation. Because ratings are valuable only as far as they are considered informative by investors, in a well-functioning market, reputational sanctions should prevent rating inflation.
The economic literature shows that the problem is more complex. If a large proportion of investors are naïve, the market will fail to punish rating inaccuracy. The same result occurs if there are enough investors that reap sizable regulatory benefits from investing in highly rated assets, because those investors always prefer rating inflation. Moreover, ratings may turn out to be inaccurate ex-post because the forecasting technology available to CRAs is imperfect and hence they make mistakes. Because the relative importance of these problems is not known and is likely to vary across different contexts, we argue that accuracy of ratings cannot be fostered by detailed regulations of the markets in which they are supplied. On the contrary, we argue that an appropriately crafted strict liability rule supports rating accuracy simply by recreating opposing interests between supply and demand for ratings.
The imposition of liability on CRAs has been part of the regulatory debate following the global financial crisis, but so far, only negligence rules have been considered. We overcome the ambiguities of setting and enforcing a diligence standard by designing a simple and legally workable strict liability rule: CRAs should be liable to pay damages whenever a bond or a company they rate defaults. This leaves us with three problems. Firstly, the default of a large issuer could be enough to bankrupt any CRA irrespective of rating accuracy. Secondly, because we do not live in a world of perfect foresight, CRAs should not be liable if they make mistakes due to Knightian uncertainty. Thirdly, under a strict liability rule CRAs would be acting as de facto insurers against the risk of defaults. As the defaults of firms and financial assets can be significantly correlated, CRAs need to be shielded from systemic risk to avoid crushing liability.
We introduce three corrections to address each of these issues. To begin with, we introduce a damage cap to limit CRAs’ exposure to liability. Liability should be calculated multiplying the income received from the issuer by the inverse of the highest probability of default associated with the rating class in which the issuer (or its asset) is included. This liability regime disallows profits from rating inflation without discouraging ratings altogether. The CRAs will prefer to supply lower ratings in order to reduce their expected liability, whereas issuers and regulated investors will prefer higher ratings. Under this mechanism CRAs earn zero profits if, and only if, their predictions are not inflated, whereas issuers only purchase ratings if those do not underestimate their creditworthiness. The ratings produced in such a market will reflect all available information about the creditworthiness of issuers and their bonds.
Because we do not live in a world of perfect foresight, even under ideal incentives the CRAs will be prone to making mistakes. To address this problem we introduce a parameter—alpha—allowing CRAs to choose the desired degree to exposure to liability. The smaller alpha is, the more mistakes CRAs are allowed to make without suffering losses (and the more economic profits they can make if their ratings are accurate). By announcing alpha to investors, the CRAs commit to a certain degree of confidence in their own forecasting models and they are expected to compete on this credibility variable, along with the rating level, in order to attract issuers. In theory, CRAs may opt out of liability even entirely, unless they want their ratings to be relevant for regulation. In order to avoid that the market equilibrium is again determined by regulated investors’ demand for high ratings, we require regulation to set a minimum alpha as a precondition for rating to generate regulatory benefits. Otherwise, the key feature of alpha is its contractibility. Being a standardized commitment device supported by an enforceable strict liability rule, the alpha generated by the market can be as low as to keep CRAs in business and as high as to make ratings informative even for naïve investors.
Lastly, we introduce corrections to protect CRAs from systemic risk, which we define simply as correlated defaults. Because ratings are uninformative about systemic risk, CRAs should not be liable when defaults depend on it. This outcome is easily achieved for corporate bonds, whose defaults are strongly correlated only in the medium to long term: the CRAs liability for these bonds should expire three months (the typical “watchlist” timing) after the production or the confirmation of a rating. Defaults of structured finance products, however, can be correlated also in the short run and the exposure to systemic risk is particularly severe for the higher tranches (for which the liability exposure is also higher). Therefore, we recommend that CRAs liability for such products be conditional on rating inflation being observed for the population of rated assets over a certain period, for instance one year. Importantly, this period can be extended backwards by a public authority, for instance up to the average maturity of the products in question, upon declaration of financial crisis status. This solution is countercyclical, because it rewards the CRAs that were more conservative in rating structured finance products during economic booms.
The full paper is available here.