Introduction

The main activity of credit rating agencies (CRAs) is providing investors and regulators with certifications of the quality of financial assets. Any lender is interested in knowing what is the likelihood that the borrower will honor his debt, and ratings serve exactly this need. In a nutshell, ratings are informed opinions on the probability that the lender (issuer of the financial asset) will not repay the borrower (investor). In other words, ratings are an estimate of the probability of default (PD) of a given bond. In some cases, ratings also account for other factors, like the expected magnitude of the losses associated with a possible default of the issuer (loss given default (LGD)). When ratings are not accurate, they can be either inflated or deflated. A rating is inflated when the CRA overestimates the creditworthiness of the rated bond. Instead, a rating is deflated when the creditworthiness of the issuer is underestimated. CRAs can potentially be useful actors on financial markets because they help reducing information asymmetries between the issuers of the rated assets and regulators and investors. However, over the last decade CRAs have been at the center of a very heated debate both at the policy level and on the scientific literature as, allegedly, they have played a crucial role in the recent financial crisis. In particular, many scholars and policy makers have argued that CRAs issued inflated ratings thus contributing to the explosion and the propagation of the financial crisis (White 2010). This entry investigates how much truth there is behind these accusations, what are the reasons that might have lead credit rating agencies to inflate their ratings, and what are the proposals advanced by the law and economics literature to induce CRAs to issue accurate ratings. Two preliminary caveats are required. First, ratings can be divided in two broad categories: solicited ratings and unsolicited ratings. Solicited ratings are requested by the issuer that pays a fee to be rated by the CRA and provides the CRA with relevant information. Instead, unsolicited ratings are spontaneously issued by the CRA that does not receive any fee and are generally based on information available to the public. Because solicited and unsolicited ratings are associated with drastically different incentives for the parties involved, these two kinds of ratings cannot be analyzed together. This entry focuses only on solicited ratings as the fees collected issuing this kind of ratings generate the vast majority of CRAs revenues. Second, not all solicited ratings are equal. Ratings of structured finance products are markedly different – and way more problematic – than ratings of corporate bonds.

History of the Market for Ratings

In 1909 John Moody published the first publicly available bond ratings. Other firms soon engaged in this practice creating the market for ratings. At the time, these ratings were sold to investors who paid to have an overview of the creditworthiness of a number of issuers. This is a business model currently referred to as investor-pays model. However, a number of reforms and technological innovations completely transformed the landscape of the market for ratings. A detailed analysis lies beyond the scope of this entry, but a quick overview of the most relevant changes is useful to understand what the problems in the market for ratings are.

Starting from the 1930s, regulators began to grant credit rating agencies an increasingly fundamental role in the functioning of financial markets. For example, in 1936 bank regulators issued a decree that was aimed at preventing banks from investing in “junk bonds,” as defined by “recognized ratings manuals.” As the only recognized ratings manuals were Moody’s, Poor’s, Standard, and Fitch, regulators de facto gave to the judgments of these four rating agencies (later to become three when Standard and Poor’s merged into Standard & Poor’s) the status of law (White 2010). Insurance and pension regulators adopted very similar rules (White 2010). In the 1970s, Moody’s, Standard & Poor’s, and Fitch relevance on financial markets was further increased by the combined effect of two decisions of the Security and Exchange Commission (SEC). On the one hand, the SEC imposed that the minimum capital requirements of brokers-dealers had to be tied to the riskiness of their asset portfolio, and ratings were to be used to determine the level of risk. On the other hand, afraid that smaller CRAs not constrained by reputational concerns could issue inflated ratings to attract customers, the SEC dictated that only ratings issued by “nationally recognized statistical rating organization” (NRSRO) could influence the minimum capital requirements. Moody’s, Standard & Poor’s, and Fitch were the only CRAs that obtained the status of NRSRO. Combined with a number of other regulations, the effect of these reforms was to grant Moody’s, Standard & Poor’s, and Fitch a quasi-regulatory power.

Another piece of the puzzle is the change in the business model that CRAs undertook in the 1970s. Potential free riding problems associated with the diffusion of fast photocopy machines undermined the investor-pays model. In particular, because it was becoming easy, quick, and cheap to disseminate information, rating agencies feared that the content of their ratings could have circulated also among investors that did not pay the relative fee. Therefore, instead of exacting a payment from investors who wanted to see the ratings of a given issuer, CRAs started to require a fee from the issuer that they had to rate. Nowadays, 95% of CRAs revenues derive from the fees collected from issuers (Partnoy 1999). Last, over the last decades rating agencies started to rate structured finance products that were becoming more and more complex.

Summarizing, three main factors characterized the evolution of the market for ratings:

  • An increased regulatory relevance of ratings that granted CRAs a quasi-regulatory power.

  • The adoption of an issuer-pays model.

  • CRAs started rating complex structured finance products.

Failures in the Market for Ratings

In the wake of the crisis, referring to CRAs the Nobel Prize winner 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. (Krugman 2010)

To put it differently, CRAs rate their clients, and hence they could be inclined to cater to the needs of the latter to attract more business (Darcy 2009). This view has been extremely influential in the literature, in the policy debate, and in the media, but it overlooks the insights of reputational capital theory (e.g., Choi 1998). According to this theory, “the only reason that rating agencies are able to charge fees at all is because the public has enough confidence in the integrity of these ratings to find them of value in evaluating the riskiness of investments” (Macey 1998). Therefore, absent other market failures, reputational sanctions would discipline CRAs’ behavior preventing them from inflating ratings. In this vein, the literature has looked beyond the conflict of interest and identified three other market failures that altered the functioning of the market for ratings.

First, reputational capital theory is grounded on the idea that investors are sophisticated enough to determine when ratings are inflated. However, if a large enough fraction of investors is Naive and cannot identify inaccurate ratings, reputational sanctions become largely ineffective (Bolton et al. 2012). Second, CRAs collect their fee only when they publish the ratings, and hence issuers could contact multiple rating agencies and request publication only for the most favorable rating received (Dennis 2009). This practice of shopping for the most favorable rating can result in rating inflation, especially for complex assets (Skreta and Veldkamp 2009). Third, as high ratings are associated with regulatory benefits, issuers might be interested in purchasing good ratings, regardless of whether investors trust the ratings. Thus, when the regulatory benefits attached to high ratings are sufficiently relevant, a rating agency “finds it profitable to stop acquiring any information and merely facilitates regulatory arbitrage through rating inflation” (Opp et al. 2013).

Therefore, there is a combination of market failures that, associated with the issuer-pays model, induces CRAs to inflate their ratings. And indeed, while the effective contribution of rating inflation to the financial crisis is still disputed (Gorton and Ordoñez (2014) argue that the contribution was likely to be limited), a large part of the literature finds that ratings, especially of structured finance products, were inflated (e.g., Calomiris 2009).

Fixing the Market for Rating

The issuer-pays model, combined with the possibility of shopping for the most favorable rating, the regulatory benefits attached to high ratings, and the naivety of some investors, creates incentives for the credit rating agencies to inflate their ratings. As there is such a complex web of market failures, inducing CRAs to issue accurate ratings is no easy task. Moreover, not all ratings are equal, and ratings of complex structured finance products create more concerns than the traditional ratings of corporate bonds. The main reasons are that (i) structured finance products are more complex and rating inflation can be more severe for complex bonds (Skreta and Veldkamp 2009) and (ii) many structured finance products behave as economic catastrophe bonds, that is, these financial assets are less resistant to economic downturns and their defaults are highly correlated (Coval et al. 2009). Any proposal that aims at improving the functioning of the market for ratings should account for these differences.

The market for ratings is an oligopoly with high barriers to entry dominated by Moody’s, Standard & Poor’s, and Fitch, and hence an obvious solution to ameliorate CRAs incentives could be increasing the competition in the market (Hill 2003). Nevertheless, empirical research shows that under the status quo, competition worsens the quality of ratings (Becker and Milbourn 2011). In presence of rating shopping more competition negatively affects the quality of ratings because the issuer has more choices when searching for the most favorable rating (Becker and Milbourn 2011). Alternatively, as regulatory benefits attached to high ratings neutralize – or at least reduce the impact of – reputational sanctions, one obvious solution to improve the quality of the ratings is reducing their regulatory relevance (Flannery et al. 2010). The Dodd-Frank Act takes exactly this path, but it seems unlikely that the implemented reforms will suffice to eliminate both direct and indirect regulatory benefits derived from relying on ratings (Hill 2010). And indeed, the European regulator explicitly remarked that there are no perfect substitutes for ratings, and hence ratings are bound to have some regulatory value (Pacces and Romano 2015). As noted by Coffee (2011), regulators decided to assign regulatory value to ratings because they have limited information and cannot develop reliable measures of risk. In other words, ratings are a precious component of regulation, provided that they are accurate, because there is no guarantee that alternative solutions to identify excessive risk will not prove to be even more problematic. Therefore, reforms should attempt to improve CRAs’ incentives while preserving – at least partially – the role played by ratings in financial regulation. Another possible reform is forcing CRAs to abandon the current issuer-pays model in favor of different business models (Mathis et al. 2009) or even introducing some sort of public funding for CRAs (Listokin and Taibleson 2010). However, the information contained in ratings has the nature of a public good because it can easily be disseminated among investors, and hence alternatives to the issuer-pays model are generally considered unworkable (Partnoy 1999; Coffee 2011). Another proposal is to pay rating agencies with the debt that they rate (Listokin and Taibleson 2010). In this vein, CRAs would be punished when issuing overoptimistic ratings because they receive debt that is worth less than they claim. Last, a path that has been widely explored by the law and economics literature is to make the liability threat faced by CRAs issuing inaccurate ratings more credible. In fact, for many years CRAs have been de facto immune to liability claims (Coffee 2006; Partnoy 2006), and in the United States they were even put under the umbrella of the First Amendment on the freedom of speech (Deats 2010). Generally, the literature considers a negligence rule as the most appropriate to induce credit rating agencies to issue accurate ratings. Under this rule, rating agencies are asked to compensate the investor only when they have been negligent in formulating their rating. In Europe, the United States, and Australia the liability of CRAs is largely based on this logic. There are, however, a number of problems with this approach (Pacces and Romano 2015). First, it is extremely hard for courts to identify the optimal level of care (Coffee 2004), because ratings are complex and prospective judgments that necessarily involve at least some subjectivity on the part of the raters. Determining when there was negligence in the formulation of this prospective judgment has been defined a Serbonian Bog by the literature (Coffee 2004). Imprecise and uncertain standards of care are notoriously associated with an increase in transaction costs and more unpredictability of courts’ behavior. Second, CRAs are not responsible for all the losses associated to a default. For example, CRAs did not lower Enron rating below investment grade until only a few days before the bankruptcy, thus fueling accusations of negligent behavior on their part (Frost 2007). However, while CRAs can be held liable for not detecting Enron’s problems, they are certainly not liable for the fact that Enron went bankrupt. Therefore, they should be asked to compensate only a fraction of the harm associated with Enron’s bankruptcy. Identifying which fraction of the harm is attributable to CRAs conduct is extremely challenging, if not impossible (Pacces and Romano 2015). If the liability threat is reinforced via a negligence rule, the combined effect of these two problems might be that rating agencies become exceedingly conservative in their judgments or even refuse to rate risky securities. These problems would be especially severe because, on the one hand, risky assets are exactly those for which ratings are more needed. On the other hand, CRAs can play a beneficial role only if they issue accurate ratings, not if they issue deflated ratings. Empirical evidence and theoretical studies suggest that this risk is concrete. The Dodd-Frank Act significantly increased CRAs’ liability exposure, and Dimitrov et al. (2015) found that as a consequence the informative content of corporate ratings further worsened. At a theoretical level, Goel and Thakor (2011) show that when the liability threat is severe, credit rating agencies have an incentive to issue deflated ratings, because it is unlikely that courts will hold them liable for conservative ratings. To put it differently, the expected liability faced by CRAs issuing inflated ratings is larger than that faced by CRAs issuing deflated ratings.

The problems of a strict liability rule are as severe. On the one hand, if CRAs are asked to cover all the losses whenever an issuer they rated defaults, they would bankrupt almost immediately as the default of a single large issuer might cause losses that exceed the assets of a CRA. Moreover, under a strict liability rule, the injurer (here the CRA) acts as a de facto insurer (Priest 1987). It is common wisdom that only uncorrelated risks can be insured (Priest 1987), whereas defaults – especially of structured finance products – are highly correlated and concentrate during economic crises (Coval et al. 2009). Pacces and Romano (2015) attempt to cope with this problem proposing a less intrusive form of strict liability that relies mainly on market forces. Introducing a damage cap based on objective factors and corrections to shield CRAs from the risk of correlated defaults, Pacces and Romano (2015) argue that a modified regime of strict liability might induce CRAs to issue ratings that are as accurate as the available forecasting techniques allow.

In conclusion, the market for ratings is characterized by multiple market failures, and hence the literature is still struggling to find effective solutions.

Future Research

How to prevent future malfunctioning in the market for ratings is still an open issue. In the coming years quantitative studies might attempt to disentangle the size of the impact of each market failure and to understand how these market failures interact with each other. For example, regulatory benefits and investors’ naivety reduce the effect of reputational sanctions, but how their effects are related is more obscure. The relationship between the effects of these two failures might be additive (e.g., if each reduces the magnitude of reputational sanctions by 10%, then the joint effect is 20%), or the effects of the two market failures could stand in more complex relationships (e.g., the presence of naïve investors magnifies the effect on reputational sanctions of the regulatory benefits, thus producing a joint effect larger than 20%). And indeed, we have seen that the issuer-pays model in itself is not necessarily problematic, but it creates perverse incentives when combined with other market failures. A clearer and more quantitative understating of the interactions among market failures in the market for ratings might help improving the regulatory regime of rating agencies.

Another important question that awaits an answer is to which extent regulatory reliance on rating agencies is motivated. In turn, answering this question implies that alternative solutions and their possible limitations are explored. Flannery et al. (2010) attempt exactly this task, but more studies are warranted.

Cross-References