Description
Debt instruments the agencies rate may include government bonds, corporate bonds, CDs, municipal bonds, preferred stock, and collateralized securities, such as mortgage-backed securities and CDOs.
Credit Rating Agencies: An Alternative Model
Pragyan Deb† and Gareth Murphy‡ November 2009
Abstract We explore the con?ict of interest which arises in the credit ratings industry due to the current issuer-pay model. We argue that the most e?cient way of aligning the incentives of rating agencies is to return to an investor-pay system. Careful analysis of the reforms currently being discussed suggests that, by themselves, they will have insu?cient impact if the current issuer-pay model is maintained. While a return to the pre-1970s investor-pay model would solve the con?ict of interest problem, the issue of free-riding amongst some investors is likely to result in insu?cient revenues for the rating agencies. We argue that although free-riding is a problem, the increasing use of ratings by institutions coupled with the rise in the speed of information di?usion and predominance of electronic trading venues over the last few decades would ensure that there are investors willing to subscribe to ratings. This investor-pay revenue could be supplemented by a government subsidy to ensure that su?cient resources are available to the rating agencies. To fund the government subsidy, we propose that a small tax would be levied on issuers or at the point of issue. A limited number of rating licenses which provided a ‘right to rate’ which would be auctioned (just like 3G auctions) and the auction winners would be entitled to a portion of the tax pool which is paid in arrears and linked to the share of the investor-pay market that they manage to achieve. Such a system would align the incentives of the rating agencies with investors, would ensure a commercially viable ratings agency industry and would have negligible impact on primary issuance markets.
Keywords: Competition, Reputation, Financial Regulation, Auction, Industrial Organisation JEL Classi?cations: C7, D4, G2, K2, L1, L9
† London School of Economics and Financial Markets Group, E-mail: [email protected] ‡ Bank of England, E-mail: [email protected]
1
Introduction
The credit rating industry aims to o?er investors valuable information about ?rms in need of ?nancing. Due to asymmetric information between the ?rms and the investors, credit ratings often have a pivotal impact on the ?rms’ ?nancing outcomes. However since the early 1970s, rating agencies have relied on an issuer-pay model, creating a con?ict of interest - the largest source of income for the rating agencies are the fees paid by the very ?rms that the rating agencies are supposed to impartially rate.1 This tempts rating agencies to rate better than what fundamentals suggest, as many have pointed out during the recent subprime crisis. In this paper, we explore the con?ict of interest which arises in the credit rating industry due to the issuer-pay model. The rest of this section reviews the current business model of rating agencies and summarises the academic literature highlighting the problems with the current structure. Section 2 examines the proposals currently being discussed to reform the rating agencies, and suggests that by themselves, they will have an insu?cient impact if the current issuer-pay model is maintained. In Section 3, we propose an alternative structure for the industry - an investor-pay model, supplemented by a government subsidy to ensure su?cient resources are available to the rating agencies. Section 4 concludes.
1.1
Business Models: Past and Present
Before the 1970s, the ratings industry operated under an investor-pay model. Investors subscribed to ratings released by the agencies, and these subscription revenues were the main source of income for the rating agencies. However, owing to the public good nature of ratings and the increase in free-riding due to the spread of the photocopier, rating agencies switched to the current issuer-pay model. In 1970s Moody’s and Fitch switched to the issuer pay model with S&P following a few years later2 . As the market stands today, S&P and Moody’s (which account for around 80% of the market) rate and make public all SEC-registered ?rms, whether the rating is requested or not3 . If the issuer does not request a rating, the rating agency issues a rating based on publicly available information. If issuer requests a rating, it pays the rating fees and provides the rating agency with private information about the ?rm. Typically an issuer gets a higher rating after having solicited a rating from a rating agency.4
1 2
SEC, 2008, p.9 S&P switched to issuer pay model for municipal bonds in 1968 3 FITCH only does solicited ratings 4 There are 2 possible explanations for this -
1
The combined revenue from ratings of the big-three rating agencies - S&P, Moody’s and FITCH stood at US$3.7 billion in 2008. This, when compared with total rated bond issuance of US$3.9 trillion5 implies average revenue from rating in the order of 10bps. This is important for calibrating the subsidies in the investor-pay model outlined in section 3 since it suggests that any tax imposed to fund the subsidy is likely to be small and thus have a potentially negligible impact on the market.
Name Standard & Poors Ratings Services Moodys Investors Service, Inc. Fitch, Inc. A.M. Best Company, Inc. DBRS Ltd. Egan-Jones Rating Company Japan Credit Rating Agency, Ltd. LACE Financial Corp. Rating and Investment Information, Inc. Realpoint LLC Comments Largest player Second largest after S&P Much smaller, but part of big-three Specialises in the insurance market Focus on Canada Investor-pay Focus on Japan Investor-pay, specialises in ?nancial institutions Focus on Japan Specialises in structured ?nance market
Table 1: NRSRO designated rating agencies in US, as of September 2008
Apart from the big-three rating agencies, seven other rating agencies are recognised as Nationally Recognized Statistical Rating Organizations (NRSROs) in the US. However, they are comparatively tiny and they restrict their activities to particular market segments. An interesting trend in the last few years has been the emergence of investor-pay rating agencies, such as Egan-Jones Rating Company, LACE Financial and Rapid Ratings. Furthermore, as Firgure 1 shows, Moody’s issuer-pay revenues fell from 89% of total revenues in 1999 to 72% in 2008. These trends indicate that the investor-pay business model may be viable, although the issuer-pay model is by far dominant and remains the main source of revenue for rating agencies.
1.2
Problems with the current system
The current issuer-pay system creates a con?ict of interest for the rating agencies since their main source of revenue, i.e the rating fee, comes from the issuers that the rating agencies are supposed to impartially rate. An analysis of the annual reports of the big-three rating agencies suggests that issuer-pay rating revenues account for roughly 84% of Moody’s total revenues (average of last 5 years), while the corresponding amount for S&P and FITCH is 72% and 85% respectively.
• Sample selection bias - only ?rms with favorable information choose to solicit ratings. • Rating agencies are biased and improve the rating if they get fees.
Both ?gures were down sharply due to the subprime crisis. In 2007, ratings revenues stood at US$4.9 billion, while bond issuance was US$4.7 trillion.
5
2
US$ million 2500
2000
Other Revenue Ratings Revenue
1500
1000
500 86% 89% 86% 87% 83% 81% 80% 82% 83% 81% 72%
0 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Figure 1: Moody’s ratings revenue (issuer-pay) has fallen over time
While rating agencies have long claimed that this con?ict is unlikely to bias their ratings out of concern for their reputation, the recent sub-prime crisis has clearly shown that this is not the case. Every rating agency has its own published fee structure, but there is widespread use of negotiated rates for frequent issuers. In general, frequent, large issuers tend to establish long term relationships with the rating agencies, making the task of regulating the relationship much more di?cult. Thus, even if issuers are forced to pay rating agencies for initial analysis,6 it is unlikely to ensure proper incentives for rating agencies when the issuer-rater relationship is viewed as a long term relationship as opposed to a one-o? transaction. Since the crisis, a growing body of research has highlighted the weaknesses of the current model. Rochet, Mathis, and Mc Andrews (2008) show that reputational concerns are not enough to solve the con?ict of interest problem. In equilibrium, rating agencies are likely to behave laxly, i.e. rate bad projects as good and are prone to reputation cycles. While Rochet, Mathis, and Mc Andrews (2008) look at the con?ict in the context of a monopolistic rating agency, Camanho, Deb, and Liu (2009) show that the often sighted solution of introducing more competition is likely
6
as proposed under the Cuomo Plan
3
to aggravate the con?ict further and lead to increased ratings in?ation. This result stems from the fact that increasing the number of rating agencies with the aim of increasing competition, reduces an individual rating agencies reward from maintaining reputation. If the overall size of the ratings market remains the same, then with increased competition, each rating agency has a smaller share of the market. In such a situation, the bene?ts of maintaining reputation (and giving honest ratings) falls and ratings in?ation increases. Camanho, Deb, and Liu (2009) show that in general, moving from a monopolistic to a duopolistic setting will increase ratings in?ation and aggravate the lax behaviour of rating agencies. Becker and Milbourn (2008) measure empirically the relationship between competition amongst rating agencies on reputation building. They show that increased competition7 leads to ratings in?ation and less informative ratings. Thus, reputational concerns are not su?cient to discipline rating agencies. It is the case that this result depends critically on the assumption that the overall market size remains the same. However, increasing the size of the ratings market by requiring issuers to approach multiple rating agencies is also not likely to help. Bolton, Freixas, and Shapiro (2009) show that with multiple ratings, issuers have more opportunity to rate-shop8 and in the presence of naive investors,9 monopoly is superior to duopoly in terms of total ex-ante investor welfare. Therefore, in order to enhance competition, if we allow for a large number of rating agencies, more and more investors will ?nd it di?cult to infer rate-shopping and e?ectively the fraction of naive investors in the market would rise. At the same time, with more rating agencies, the issuers ability to rate shop would increase thereby decreasing ex-ante investor welfare. Skreta and Veldkamp (2008) do not consider the strategic behavior of rating agencies but explore the interaction between rate-shopping, complexity of the security10 and competition. They show that the intensity for rate-shopping increases with the complexity of the security and competition between rating agencies ampli?es this problem. Both these papers argue for a ban on rate-shopping. However, such a ban11 coupled with enhanced competition in the form of larger number of rating agencies will decrease the market size for each rating agency, make existing clients more valuable and thus reduce their incentive to maintain their reputation. Therefore, increasing the number of players while banning rateshopping is likely to aggravate ratings in?ation as highlighted in Camanho, Deb, and Liu (2009).
in the form of increased market share for FITCH get multiple ratings from di?erent rating agencies and make public only favourable ratings 9 investors who are unable to infer the extent of rate-shopping. 10 a key factor for structured products 11 which will be very di?cult to enforce given the long term relationship between the issuer and the rating agency
8 7
4
2
Proposed Reforms
There exists general consensus in policy circles on the need to strengthen the regulation of credit rating agencies, ‘including measures to promote robust policies and procedures that manage and disclose con?icts of interest, di?erentiate between structured and other products, and otherwise strengthen the integrity of the ratings process’.12 However most of the reforms currently being discussed attempt to increase regulatory oversight and disclosure norms in order to tweak the current issuer-pay system and make it more transparent without attempting to eliminate the inherent con?ict of interest outlined above. In what follows, we look at some of the major reform proposals and argue that, by themselves they will have an insu?cient impact. However, when combined with an investor-pay model (as proposed in Section 3), they are likely to be highly e?ective in enhancing transparency and e?ciency in the ratings industry. We look at the following reforms proposals 1. Lowering Barriers to Entry 2. Holding Rating Agencies Legally Liable for their Ratings 3. Taxpayer Funded Credit Rating Agency 4. Ban on Rate-Shopping and Consulting Services 5. Make Ratings more Informative
2.1
Lowering Barriers to Entry
Moody’s and S&P have dominated the credit ratings industry for years and together with FITCH control around 95% of the market. The ratings industry has large and often prohibitive barriers to entry. While some of these barriers such as reputation, scale economies and network e?ects13 are inherent in the business, others are a result of regulatory restrictions and insu?cient disclosure requirements. It has been suggested that if information is made more widely available, it might reduce entry barriers to this industry, increase competition and serve to alleviate the incentive problems in this industry. Typically, in the case of solicited ratings, issuers provide private information to rating
12 13
US Department of the Treasury - Financial Regulatory Reform: A New Foundation (2009) investors desire for consistency of rating categories across issuers limits number of players
5
agencies. If this information is revealed to the wider market, then it could incentivise other players such as insurance, pension and mutual funds, brokerages etc. to use this information in their own internal reports and investment decisions. They can then act as investor-raters making their reports public, providing the wider investor community with an additional source of information - a de-facto rating . This will enhance the level of competition in the ratings industry. However, as noted earlier, merely increasing competition in the ratings business is likely to be counterproductive. However, in this case, since the funds and brokerages undertaking research have a stake in their own ratings, they will not have the same con?ict of interest that plague the rating agencies. In essence, it will be similar to having incumbent rating agencies have a stake (legal or economic) in their ratings. However, the potential for market manipulation by the investor-raters needs to be taken into account. For example, investor-raters may be tempted to sell their stake just before a downgrade or issue biased ratings in order to move the markets in a particular direction. At this stage, a distinction needs to be drawn between brokerages, which are likely to have much wider coverage but less stake in their reports14 and mutual funds which would be much more selective on their coverage, but are likely to have a much larger stake in their reports.15 In general, wider coverage would mean having a lower stake in their reports. Thus, if the investorraters indeed morph into de-facto rating agencies and cover most asset classes, their stake in their reports would become marginal and they will end up with incentives very similar to the incumbent rating agencies. Even if these issues are resolved, we have another potential problem. Issuers may not be willing to make their private information public. While they may be willing to disclose the information to a rating agency in con?dence, they may undermine their competitive position if the same information is made available to the public and thus their competitors. Thus we might get 2 di?erent cases. In the ?rst case, if disclosure norms are not mandatory, then pure rating agencies will have access to additional private information provided by the issuers vis-a-vis investor-raters. Thus investors have to decide between 2 di?erent kinds of rating agencies • Pure Rating Agencies, that face a con?ict of interest and potentially issue biased ratings, but have access to private information of the issuers. • Investor-Raters, that face the right incentives, but are informationally constrained.
14 15
which is primarily used as a marketing product assuming their investment decisions are veri?ably based on these reports
6
Thus, this kind of a mechanism will be of limited bene?t and will make the job of the investors much more di?cult. The other alternative is mandatory disclosure of all private information. However, such a policy may not be in the interest of the issuers. In the case of corporate issuers, it risks exposing their company secrets to competitors. In the case of structured products, complete disclosure would result in lower pro?ts and thus decrease the incentive to innovate. The gains from ?nancial innovation will be capped because full disclosure will allow competitors to copy the product very quickly. Over time, this may sti?e innovation and may have detrimental welfare consequences. Furthermore, such blanket disclosure requirements will be very hard to implement and are likely to create ine?cient market structures. For example, if disclosure requirements are imposed on one part of the market (for example on exchange traded products), then it would encourage issuers to go for privately placed OTC instruments. Over time, the OTC market will expand and end up becoming the dominant part of the system.
2.2
Holding Rating Agencies Legally Liable for their Ratings
Currently, credit ratings are treated as ‘opinions’ and thus rating agencies cannot be held legally liable for their ratings which are protected under free speech. However, if rating agencies are made liable for the quality of their ratings,16 it would go a long way in resolving the con?ict of interest. Currently, the only factor restraining rating agencies is the concern for reputation, but if this channel is strengthened by forcing rating agencies to have a legal or ?nancial stake in their ratings, rating agencies would no longer ?nd it pro?table to in?ate ratings. This would have the same e?ect as having investor-raters in the market (see section 2.1) However, given the nature of the ratings industry, it would be very di?cult to force rating agencies to have a stake in their ratings. Given the capital structure of rating agencies and the size of the market that needs to be rated, a ?nancial stake in their ratings is not feasible.17 While a ?nancial or a legal penalty is possible in the event of ex-post poor quality rating, it would be very hard to implement. For the system of penalties for poor ratings to work e?ectively, it would require authorities (regulatory or judicial) to be able to distinguish between poor ratings arising out of rating agency bias or bad luck. If it is the former, then a penalty would be warranted, but in case of the latter, the penalty would lead to ine?ciencies.
like auditors unlike banks and other ?nancial institutions, rating agencies do not hold any capital. Hence the US proposal for ‘joint lability’
17 16
7
It is very unlikely that authorities would be able to make such a distinction with relative certainty, making a blanket penalty the only feasible option. But a blanket penalty every time the quality of ratings falls below a threshold would do little to solve the incentive problem of rating agencies. Such a blanket penalty would only serve to make rating agencies excessively conservative. This is because rating agencies, fearing the penalty would prefer to err on the side of caution and give lower ratings. Therefore such a proposal would essentially replace the upward bias in ratings with a conservative bias, leading to less informative ratings and possible capital misallocation. Furthermore, the US ‘joint liability’ proposal18 would make the entry barriers to the industry even more prohibitive. This is because under such a system, any new potential entrant may become liable for all ratings issued by other players. As James H. Gellert of Rapid Ratings International, Inc. puts it,19
“Why would one want to become an NRSRO joining a group dominated by three players with an iceberg of lawsuits looming on their horizon? That would be like swimming towards the Titanic.”
2.3
Taxpayer Funded Credit Rating Agency
A taxpayer funded rating agency would not be driven by commercial motives and thus such an agency is not likely in?ate ratings. Since investors are rational, they will take this into account and thus the taxpayer funded agency will have perfect reputation. If the taxpayer funded agency is just as e?cient as the private rating agency, it will capture the entire market. This is because given its perfect reputation and same cost to the issuer as private rating agency, an issuer with a good project will always prefer to approach the taxpayer funded rating agency for rating. The only case where an issuer will approach the private rating agency is when it had a bad project and hopes that the private rating agency will give it a good rating. But this will be fully anticipated by the market and the private rating agencies rating will be useless. Thus in this situation, the presence of the taxpayer funded rating agency will lead to the elimination of private rating agencies from the market. On the other had, the taxpayer funded rating agency might have a perfect reputation, but it may be less e?cient in identifying good and bad projects. Thus for the investors, the trade-o?
See EuroWeek (2009) and Financial Times (2009b) for details of the US proposals to make rating agencies jointly liable for their ratings 19 See Gellert (2009)
18
8
between private and public rating agency would boil down to a choice between between incentives and ability. While the taxpayer funded rating agency will give unbiased ratings, it would be less e?cient. On the other hand, the private rating agency would give biased ratings but will be more e?cient at identifying projects. This system by itself, will not solve the con?ict of interest of the private rating agencies. However, it might have a limited disciplining e?ect on private rating agencies by setting an upper limit for ratings in?ation. This is because the presence of a taxpayer funded rating agency may set a lower bound for reputation of the private rating agency. If the reputation of the private rating agency falls below the threshold, all issuers may approach the taxpayer funded rating agency (same mechanism as above) and the private rating agency may be eliminated from the market.
2.4
Ban Rate-Shopping and Consulting Services
Rate-shopping refers to issuers approaching multiple rating agencies for rating and selectively disclosing only those ratings that are favorable. Often issuers play one rating agency against the other in order to get favourable ratings. To quote former chief of Moody’s, Tom McGuire,20
“The banks pay only if [the ratings agency] delivers the desired rating. . . If Moody’s and a client bank don’t see eye-to-eye, the bank can either tweak the numbers or try its luck with a competitor. . . ”
Furthermore, rating agencies routinely provide ‘creative suggestions’ to issuers regarding the design of structured products in order to improve ratings. The recent subprime crisis, provides clear evidence of this. Instead of being neutral players, rating agencies were working actively with investment banks on the design of complex securities, such as CDOs and MBS. As a former Moody’s expert in securitization says “Every agency has a model available to bankers that allows them to run the numbers until they get something they like and send it in for a rating”.
Or, as Chris Flanagan, the subprime analyst at JPMorgan claimed
“Gaming is the whole thing. . . Banks were gaming the ratings and designing only the securities that were blessed by the rating agency.”
20
New York Times Magazine, Triple-A-Failure, April 27, 2008
9
This is of particular concern for complex structured products since small super?cial changes to the asset pool or the nature of the contract can have disproportionate e?ects on ratings. Thus, such ‘consultations’ can potentially amount to data mining,21 severely underling the ratings methodology. Given this situation, it has been suggested that requiring rating agencies to disclose all ratings irrespective of the wishes of the issuers as well as forcing rating agencies to make public any initial consultation they may have with issuers can put a stop to rate-shopping. However, under the issuer-pay system, given the complex nature of interaction between the issuers and the rating agencies, it would be very di?cult to enforce such a regulation. As is clear form the opposition of rating agencies to any such move,22 rating agencies are unlikely to actively cooperate in the implementation of such regulation. Thus, such a move would require constant regulatory supervision, making enforcement costly and di?cult. Furthermore, as pointed out in Section 1.2, such a ban would decrease the overall size of the ratings market (by reducing the incidence of multiple ratings), reducing rating agencies incentives from maintaining reputation and thus resulting in more ratings in?ation.23
2.5
More Informative Ratings
It has been argued that as it stands, ratings do not provide su?cient information to investors to make informed decisions. The current system of rating buckets are too coarse and provide insu?cient information to investors. The ratings bucket of di?erent rating agencies have di?erent meanings and they often hide important details such as variance, likelihood of default and/or loss severity in the event of default. This is particularly true in case of structured products which are extremely complex and investors need more information to make informed decisions. For example, under the probability of default based ratings approach (employed by S&P and FITCH) the rating agency calculates the likelihood of income stream of a CDO tranche falling short of its contractual value and determines the rating by comparing this likelihood with historically calculated default probability. Similarly, under the expected loss based approach used by Moody’s, percentage expected losses are calculated and compared with historical benchmark information. Both models provide an overall expected rating, but hide details regarding the distribution of losses etc. which are vital for informed investor decision.
21 Since every statistical model has some Type II error, i.e. failure to reject a null (high rating) when it is false, the issuer can get the desired rating by repeatedly tweaking the product. Thus through consultations, issuers can ensure that their products get a hight rating, even if it is not warranted 22 See Watson (2008a) and Watson (2008b) 23 See Camanho, Deb, and Liu (2009)
10
While increasing the information content of ratings is a step in the right direction, the e?ect of such a move on ratings bias is uncertain. Under the issuer-pay model, rating agencies will ?nd it optimal to in?ate ratings, even if the content of the rating is more informative. However, it is possible that more information will make it easier for investors to detect the bias and thus make it a little less pronounced. But fundamentally, making rating more e?ective is unlikely to solve the incentive problems of the issuer-pay model. However, as explained in the following section, making ratings more informative while switching to the alternative investor-pay model can mitigate the free rider problem to a large extent.
Proposal Current System Lower Barriers to Entry Holding Rating Agencies Legally Liable for their Ratings Taxpayer Funded Credit Rating Agency Ban Rate-Shopping & Consulting Services More Informative Ratings Investor-Pay Investor-Pay with subsidy
a b
Incentive Aligned No No No Yes No No Yes Yes
Commercially RateViable Shopping Yes Yes Yes No Yesa , Nob Yes Yes No Yes Yes Yes Yes No Yes No No
Direction of Rating Bias In?ation In?ation, potential market manipulation Conservative In?ation In?ation In?ation None None
if taxpayer funded rating agency is less e?cient assume taxpayer-funded rating agency is equally e?cient Table 2: Impact of Proposed Reforms
3
An Alternative Model
The analysis in section 2 suggests that the reforms currently being discussed will be insu?cient to solve the incentive problems inherent in the issuer-pay model. Thus we propose a return to the pre1970s investor-pay model. The investor-pay model is the most e?ective way of aligning incentives of rating agencies and investors and eliminating the con?ict of interest of rating agencies. However, it is also prone to the problem of free-riding, resulting in insu?cient resources for research and analysis and thus poor quality ratings. As the rating agencies point out, free riding was one of the main reasons behind the switch to an issuer-pay system in the 1970s. While it is true that free riding is a big issue, the increasing use of ratings by institutions, coupled with the sharp rise in the speed of information di?usion in the markets and the predominance of electronic trading venues over the last few decades has increased the speed of price response to 11
market events to only a few seconds.24 This is likely to ensure that there are investors willing to subscribe to a rating agency as opposed to waiting for selected parts of the information to leak through the market. The delay involved in free riding might be too costly for some investors. Thus, investors that require ratings information quickly will be willing to pay in order to avoid delays inherent in free riding. Investors such as large traders and brokerages, pension and mutual funds, insurance companies and other institutions that have structures in place to act on the information quickly will be willing to subscribe to the ratings as opposed to free riding and waiting for the information to arrive with a lag. Given the high speed of information di?usion, the opportunity cost of waiting is likely to be too high for such investors since the market would have already moved in response to the ratings information. On the other hand, investors who cannot trade immediately such as households and small retail investors, will have a low opportunity cost and they would likely choose to free ride. However, in equilibrium, there would always be some investors willing to subscribe to the ratings. In this context, regulation aimed at making ratings more informative25 is extremely important as it can potentially mitigate the free riding problem to a large extent. Even if investors are able to learn the broad ratings bracket through free riding, it is unlikely that they will get access to the necessary details fast enough if they try to free ride. They would need to subscribe to get reports, access to analysts etc. Thus, as ratings become more and more informative, the opportunity cost of free riding increases, mitigating the free rider problem to a large extent. More formally, assume that investors using credit rating for their investment decisions can choose to be subscribers or free riders. If they choose to be free riders, they do not have to pay the rating agencies subscription fee (s). However free riding involves a delay and we assume that the the opportunity cost of free riding for the investor concerned is ? . As a modeling device, we assume that there exists a continuum of investors with ? lying between a and b. Assuming a continuum of investors makes the model tractable26 . Intuitively, a represents the opportunity cost for a investor who would not act on the ratings information (and is thus close to 0) while b is the highest possible opportunity cost (representing institutional investors who would trade on rating information immediately). Thus, each investor decides whether to free ride or subscribe depending on ? . If ? > s, then the investor is better o? subscribing to the rating agency as the opportunity cost of free riding is too high. On the other hand, if ? < s, the investor prefers to free ride and wait for the information
24 25
[[[[[quote papers on e?ect of media on speed of information di?usion]]]]] See section 2.5 26 The result holds for any possible distributional assumptions
12
Free Riders
Subscribers
a
s
?
b
Figure 2: Investor Choice - Free Rider or Subscriber
b?s b?a .
to arrive with a lag. The fraction of investors subscribing to the rating agency is given by ? =
Note that the fraction of subscribers is a function of the fees charged by the rating agency and the information content of ratings. As the information content of the ratings increases (i.e. ? increases), the fraction of subscribing investors rises, thus mitigating the free rider problem. Also, as the subscription fee falls, the the fraction of subscribing investors rises. Therefore it is likely that there will always be a fraction of investors (?) who ?nd it pro?table to subscribe to the rating agencies reports. Assuming a unit mass of investors and a per-subscriber costs of rating c, the pro?ts of the rating agency takes the following form:
? = ?s ? c The problem arises when the fee charged is relatively modest and/or the fraction ? is small. In this case the investor-pay system may not generate enough revenues for the rating agencies, i.e. ? < 0. In such a situation, the rating agencies will not survive or will be forced to cut costs and compromise on the quality of their research. To avoid this scenario, it might be necessary to establish a supplementary source of revenue for the rating agencies. Such an alternate source of revenue can take the form of a government subsidy. In order to provide rating agencies with the correct incentives, the subsidy should be proportional to its subscription revenues. Such a subsidy will incentivise the rating agencies to maximise their investor-pay revenues, thus aligning incentives. The new pro?t function becomes
? = ?s ? c + ?s? An attractive feature of this structure is that it does not require costly regulatory information as once the level of subsidy is optimally determined, the distribution of the subsidy amongst the rating agencies is based on their respective market shares. This market determined distribution system is much more desirable than the regulator trying to determine the share of each rating 13
agency on a case by case basis. This system is dynamic in the sense that it re?ects the markets (investors) best guess about the relative usefulness of di?erent rating agencies ratings. If a particular rating agency does not perform up to standards, its market share is likely to fall over time, and with it the subsidy. Thus, this system provides a market determined measure of performance as opposed to a pre-set metric which can result in regulatory arbitrage and is open to abuse. Another big advantage of this system is that it makes rate-shopping much more di?cult for issuers. This is because with an investor-pay system, it is not in the interest of rating agencies to hide (or not make public) an unfavorable report. Thus, with an investor-pay system, any ban on rate-shopping will be much easier to implement since rating agencies would themselves have a strong incentive to comply with it. A potential problem with the above system is that the rating agencies incentives will be aligned only with fraction ? of the investors that subscribe to their ratings. If the incentives of the remaining (1 ? ?) free-riding investors are di?erent, then it might result in a con?ict of interest27 . This is particularly true if the di?erences in incentives are unknown or change in an unpredictable way, then it might turn out to be a potential problem. While we address this issue further in section 3.3, at this stage its important to point out that making credit ratings more informative should mitigate this to a large extent (see section 2.5). This is because, even if the incentives of the subscribing and free riding investors are not perfectly aligned, they are unlikely to be in direct con?ict with each other. They might have slightly di?erent preferences, but it is unlikely that they can gain at each others expense - as in the case of issuers vs investors where there is a clear and direct con?ict28 . Furthermore, under this system, the capital structure and revenue stream of the rating agencies are clear and open to scrutiny. Thus, if it would be easy to determine if a particular rating agency is under the in?uence of a special interest group. As discussed in section 3.3, other rating agencies, competing directly for market share, would do all they can to highlight this fact in order to win over subscribers. Finally, a potent safety feature of this system is that the free-riding subscribers would always have an option to subscribe to the ratings, thus aligning incentives limiting the potential damage from misaligned incentives.
For example, pension funds with a mandate to invest in only AAA securities might want more securities to be rated AAA 28 For example, a pension fund demanding AAA rated bonds, would still prefer bonds of a higher quality. They may prefer the rating brackets to be broadened to increase their scope and scale of operations, but they would still prefer to invest in bonds that do not default
27
14
3.1
Optimal Subsidy
From the regulatory perspective, determining the actual level of subsidy is not a trivial matter. Given the revenue stream of the rating agency, the subsidy should be such that the rating agency at least breaks even. Note that the break-even condition allows the rating agencies to make reasonable pro?ts in order to be viable and able to respond and invest to dynamic market conditions. We set pro?ts to 0 to simplify the exposition: ? = ?s ? c + ?s? = 0 c or ? ? = ?1 ?s In order to optimally determine ? ? , the regulator needs detailed and accurate information on the rating agency’s costs as well as demand, supply and optimal pricing of the ratings. While this is possible (as in the case of other regulated industries like electricity, gas, water etc.), this is likely to impose a heavy burden on the regulator. Furthermore, ever if the regulator could optimally set ? ? , since the subsidy is designed to ensure that the rating agencies break even in equilibrium, it would deter rating agencies from competing for market share. Given this problem, we need a mechanism which imposes less informational burden on the regulator and preserves the incentives of the rating agencies to compete amongst themselves for investor-pay revenues and market share. In order to ensure the latter, the subsidy should be ?xed a-priori allowing rating agencies to maximise their pro?t without worrying about an o?setting reduction in the subsidy. Thus the rating agency should reap the rewards from ex-post increase in market share arising from improvement in the ratings and overall quality of service, innovation, cost reduction etc. Since each rating agency is best suited to determine its own cost structure and expected revenues, we propose an auction mechanism, similar to the 3G auctions, to determine the optimal subsidy rate for the rating agencies. Each rating agency determines the minimum level of subsidy it requires to break even ?i = ci ?1 ?i s i
If the auction process is properly designed, then the most e?cient rating agencies, i.e. the ones that are con?dent of getting higher subscription revenues from the investors and have lower costs will win the auction. After the auction, the rating agencies will continue to compete since the subsidy is contingent on the investor-pay revenues.
15
The auction process would involve giving rating agencies the ‘right to rate’ and get the subsidy for a ?xed number of years. The optimal period would depend on the length of time the rating agencies require to recoup their costs and make reasonable pro?ts. Thus, the period would be a critical factor in determining the attractiveness of the auction process and thus the degree of competition. For the sake of this exposition, we assume that 5 years is optimal. Furthermore, in order to maintain the quality of ratings and make the ratings market more dynamic, we propose repeat annual auctions on a rolling basis. In the annual actions, the rating agency that has the lowest market share and has completed its guaranteed 5 year period will have to compete in the auction once again for the right to rate. This will ensure that there is enough competition in the market and incumbent rating agencies are forced to continuously compete and maintain best possible rating standards. Thus this mechanism would use the market to determine and punish the most ine?cient rating agency as opposed to having the regulator measure performance base on a particular metric.
3.2
The Auction Mechanism
A well designed auction is the method most likely to allocate resources to those who can use them most valuably since it forces business to put their ‘money where their mouths are’ when they submit bids.An auction can therefore extract and use information otherwise unavailable to the government.29 However, it is vital that the auction is properly designed and tailored to the particular context. We look at the literature on 3G licence auction for motivation on optimal auction design for the ratings business. Broadly, the optimal auction mechanism should achieve 3 main objectives: • It should allocate the ‘right to rate’ to the most e?cient rating agencies • It should promote competition • It should minimises the subsidy burden One of the key challenges in designing the auction for the ratings industry is the big advantage of incumbents30 over potential new entrants. As discussed earlier, the ratings industry is plagued by barriers to entry, both economic and regulatory. In addition to the large set-up and ?xed costs to entry, incumbents also have the advantage of established customer bases and brand-name
1 For example, the successful UK 3G auction yielded about 22.5 billion or 2 2 % of GNP and resulted in a competitive telecom market in the UK 30 particularly the big 3 rating agencies - S&P, Moody’s and FITCH 29
16
recognition. Thus, in order to make the auction process competitive, it is vital that the auction design provides su?cient incentives to new entrants. One of the key parameters to achieve this is the number of licences. Since there are three dominant incumbents in the market, the number of licences on o?er should be greater than three. Otherwise any new entrant would be deterred from participating in the auction process rendering it ine?cient.31 However, in the context of the ratings industry, given its current structure and the barriers to entry that already exists, giving out more than three licences, coupled with the ?xed term guarantee to recoup costs, might actually lead to a decline in entry barriers and foster competition. The other potential problem with the auction mechanism is that the incumbent rating agencies might feel forced to win a new licence in order to avoid a sharp reduction the value of their previous investments. The proposal to require the worst performing rating agency to participate in the auction process after the ?xed guaranteed term would lead to disruptions and can act as a disincentive for rating agencies from participating in the auction. However, such a mechanism is vital to ensure e?ciency and continued competition in the ratings industry. The incumbents’ disincentive from repeat auctions must be balanced with the need to maintain a dynamic system.32 While the exact length of time a rating agency needs to recoup costs is debatable, some kind of penalty for poor performance is necessary to maintain market e?ciency. In any case, after the ?xed term, the worst performing incumbent rating agency will still be allowed to compete in the auction process. We believe that the risk of losing the ‘right to rate’ is no di?erent from the general penalty for poor performance in the market. Risk is part of any business and rating agencies are free to take this into account while competing in the bidding process. In principle, the number of licenses can be changed to re?ect the prevailing market conditions Broadly, we have 3 distinct auction mechanisms: • Simultaneous Ascending Design equivalent to Second Price Auction • Sealed Bid Auction equivalent to First Price Auction • Hybrid Anglo-Dutch Design
For example, the Netherland 3G auction of July 2000 performed poorly because there were 5 licences with 5 incumbents, deterring potential new entrants 32 Repeat auctions can be made contingent of the market share of the worst performing rating agency falling below some pre-determined threshold. However, such a system risks cartelisation of the industry, with incumbents not competing with each other in order to sustain super-normal pro?ts.
31
17
combines element of both We analyse these 3 types of auctions with a view to see which is best suited for the ratings industry. Simultaneous Ascending Design In the simultaneous ascending auction, or the English Auction, participants bid openly against one another, with each subsequent bid higher than the previous bid. The auction ends when no participant is willing to bid further, at which point the highest bidder pays their bid. The distinguishing feature of this auction type is that the current highest bid is always available to potential bidders. Thus, in terms of outcomes, its equivalent to the sealed bid, second price auction. In the context of allocating multiple, equivalent licences, the auction ends when the number of participants remaining are equal to the number of licences available. Thus, the subsidy is equal to the level required to sustain the least e?cient rating agency.33 While this type of auction ensures that the most e?cient rating agencies win the auction and also the subsidy is kept at a minimum, it discourages competition and entry. In an ascending auction, there is a strong presumption that the ?rm that values winning the most will be the eventual winner, because even if it is outbid at an early stage, it can eventually top any opposition. Thus, dominant incumbents with established structures and client base can outbid potential entrants at any stage. This can create a situation where potential entrants are discouraged to even participate in the bidding process even in the presence of modest bidding costs. This results in insu?cient competition during the auction and can result in high levels of subsidy due to cartelisation of incumbents. Thus, in the presence of strong incumbents, this kind of auction often leads to insu?cient competition. Sealed Bid Auction In the sealed bid auction or more formally the ?rst price sealed bid auction, all bidders simultaneously submit sealed bids so that no bidder knows the bid of any other participant. Bidders make a single ‘best and ?nal’ o?er. The highest bidder, in this case the rating agency requiring the lowest subsidy, wins and the subsidy equals its own bid. From the perspective of encouraging more entry, the merit of a sealed-bid auction is that the outcome is much less certain than in an ascending auction. An advantaged incumbent will probably win a sealed-bid auction, but it must make its single ?nal o?er in the face of uncertainty about its rivals’ bids, and because it wants to get a bargain, its sealed-bid will not be the lowest
33
with more e?cient rating agencies making a pro?t
18
subsidy it could be pushed to in an ascending auction. So ‘weaker’ bidders have at least some chance of victory, even when they would surely lose an ascending auction Because sealed-bid auctions are more attractive to entrants, they may also discourage consortia from forming. If the strong rating agencies form a consortium, they may simply attract others into the bidding in the hope of beating the consortium. So incumbent rating agencies are more likely to bid independently in a sealed-bid auction, making this auction much more competitive. Thus the greatest advantage of this type of auction is that it gives weak bidders a chance (a ‘hope and dream’ in the words of one frustrated potential entrant) which can attract more bidders and results in more competition in the bidding process. However, this very fact is also its greatest shortcoming - because it allows bidders with lower valuations (higher costs in this context) to sometimes beat opponents with higher values (and so encourages entry) it is more likely to lead to ine?cient34 outcomes. Furthermore, sealed bids do not allow bidders to gather information on the business plans of their rivals by observing who is staying in and who is getting out as the price rises. They therefore make it impossible for bidders to re?ne their valuations of the licences on the basis of this information. Thus, this type of auction is only recommended when the most important objective is to encourage potential bidders to entry.35 Hybrid Anglo-Dutch Design The Anglo-Dutch design tries to marry the bene?ts of both type of auctions. In an AngloDutch auction for one object, the price rises until all but two bidders quit and the last two bidders then make a sealed bids. With 5 licences to sell, the subsidy would fall until only 6 competitors remained. The surviving bidders would then be committed to bid at or below this level of subsidy in a sealed-bid auction in which the four lowest bidders are awarded a licence. There are two versions of the Anglo-Dutch design; one in which each winner is committed to receiving its own bid, and one in which each winner is committed to paying the ?fth-highest winning bid. While the former is more likely to encourage competition, the latter is more e?cient in term of extracting private information of bidder and incentivising rating agencies to reveal their costs truthfully. The sealed-bid stage of the Anglo-Dutch design encourages competition by giving a chance to weaker players. Just as in the standard sealed-bit auction, weaker bidders have a chance of winning the auction and unlike the ascending price auction be outbid by the dominant incumbents. This encourages entry into the auction. However, as noted above, this also leads to ine?ciencies. The Anglo-Dutch design tries to overcome this by having an ascending auction as the ?st stage.
34 35
rating agency with higher costs winning the auction process As in the successful Danish 3G Auction of September 2001
19
This ensures that the auction does not result in very high levels of subsidies because of underbidding by all players. However, by having a sealed-bid stage, it encourages weaker participants to stay in the auction in the hope of winning in the last stage. Essentially, this design tries to balance the twin goals of encouraging competition, while at the same time ensuring that there is an e?cient outcome, i.e a high cost rating agency does not win the auction and the subsidy burden is minimised. The table below summarises the features of the 3 types of auctions Auction Type Simultaneous Ascending Design Sealed Bid Auction Hybrid Anglo-Dutch Design Entry Lowest Highest Moderate E?ciency Highest Lowest Moderate
Table 3: Comparision of di?erent Auction Types
Given that the initial auction is likely to o?er more licences than the number of incumbents, we believe that the Anglo-Dutch design is most appropriate in this setting. It balances the twin goals of encouraging entry while at the same time ensuring that the most e?cient rating agencies have a greater chance of winning, thus enhancing overall e?ciency and keeping the subsidy burden low. For the later annual auctions, we recommend the sealed bid auction. This is because in the annual rounds, an incumbent and potential new entrants must compete for a single licence, with 4 other incumbents already part of the industry. In such a situation, more incentives are necessary for new entrants and the sealed bid auction is most appropriate.
3.3
Industry after the Auction
After the auction, the industry would consist of a given number of rating agencies, competing with each other for market share. The pro?ts of an individual rating agency would depend critically on its investor-pay revenues, since the subsidy from the government would be linked to this revenue, thereby aligning the incentives of the rating agencies with the investors. A critical factor in determining the attractiveness of this design is the nature of competition in the industry after the auction and the mechanisms in place to prevent the rating agencies from coming under the in?uence of a particular class of investors. Insu?cient competition and cartelisation can lead to moral hazard problems for the industry leading to high cost and poor quality ratings. However, by its very design, the repeat auction mechanism ensures competition. The threat of new entry is likely to force the rating agency with the smallest market share to break the cartel and increase e?ort in order to gain gain market
20
share. This would in turn induce all other rating agencies to exert e?ort as no rating agency would want to risk competing in the auction and loose its ‘right to rate’. The more likely situation is intense competition and a race to the bottom. If the rating agencies are homogenous, then this is the likely outcome with each rating agency undercutting the other resulting in a ‘Bertrand Equilibrium’ with subscription fees close to 0. However, ratings are di?erentiated and we look at the Industrial Organisation (IO) literature to outline the nature of competition and di?erentiation in the ratings industry. Rating agencies can di?erentiate each other across two dimensions - the quality of their ratings and their coverage and specialisation. In the IO literature, the former is classi?ed as Vertical Di?erentiation while the latter is known as Horizontal Di?erentiation. Under both these settings, rating agencies would make positive pro?ts, have di?erent market shares and would cater to investors with di?erent preferences.36
Vertical Di?erentiation refers to di?erentiation on the basis of the quality of ratings. Rating agencies can choose to provide high quality rating and charge a higher price or they can provide lower quality ratings for a lower price. Investors choose between the di?erent rating agencies based on their own preferences i.e quality vs price. The market share of a rating agency would ultimately depend on the preferences of investors as well as the market segment the rating agency targets. The IO literature suggest that in general, rating agencies would ?nd it optimal to di?erentiate themselves as opposed to competing directly with each other by catering to the same market.37
Horizontal Di?erentiation
refers to specialisation in di?erent products. In this setting, dif-
ferent rating agencies would choose to specialise in di?erent ?nancial instruments and regions. For example, while one rating agency may specialise in corporate bonds the other may choose to specialise in structured products. Investors would choose di?erent rating agencies on the basis of their individual preferences. Thus an investor primarily investing in corporate bonds would go for the rating agency specialising in corporate bonds while an investor investing in structured products would subscribe to the rating agency specialising in it. Once again, the IO literature suggest rating agencies would choose to di?erentiate and cater to di?erent segments of the market38 Thus the ratings market can be thought o? as a two-dimensional space, with rating agencies choosing where to position themselves in order to di?erentiate themselves from their rivals (see
36 The current market structure shows some evidence of such di?erentiation, with the smaller rating agencies specialising in particular sectors or regions. See Table 1 for details. 37 See Gabszewicz and Thisse (1979, 1980), Gabszewicz and Thisse (1980), Gabszewicz, Shaked, Sutton, and Thisse (1981), Bonanno (1986) and Gal-Or (1983) 38 See Hotelling (1929) and Salop (1979)
21
?gure 3). The vertical axis shows quality while the horizontal axis represents specialisation in di?erent products. Investors choose the rating agency based on their preference for specialisation and quality. Since ratings are di?erentiated, the rating agencies have some market power and make pro?ts in equilibrium.
Differences in quality of ratings
Rating Agency
Vertical
Horizontal
Specialisation in different products
Figure 3: Market for Ratings - di?erentiated by quality and specialisation
Note that the distribution of investors in the two-dimensional space need not be uniform. For example, the proportion of investors demanding very high quality, more expensive ratings might be much higher than those demanding lower quality, cheaper ratings. The rating agency targeting the high quality segment of the market will in general have a higher market share. However, while such a position would be sought after by all rating agencies, once a particular rating agency positions itself in this position, the other rating agencies would ?nd it optimal to cater to a di?erent segment of the market. This is because rating agencies would prefer to relax price competition through product di?erentiation.39 If it so happens that most of the clients of one particular rating agency belong to a particular class (say pension funds), then some other rating agency would position itself to target a di?erent segment of the market. In so doing, they would continue to serve their may economic purpose
39
see Shaked and Sutton (1982)
22
- to facilitate collective research instead of each investor being forced to conduct research on a individual basis. The often sighed danger of investor-pay rating agencies coming under undue in?uence of a particular class of investors can be minimises further by requiring rating agencies to make public the distribution of their investor-pay revenues across the population of investors. Since rating agencies would already be providing information on their total investor-pay revenues to the regulators in order to get the subsidy, this should not impose undue burden on the rating agencies. Such disclosers would make any possible ratings bias clear to the investors. In case the market is unable to correct the potential biases, such data would also allow regulators to identify any concentrations or de?cits in the market and take corrective measures.
4
Conclusion
Since the credit-crisis, widespread concern has arisen about the functioning and business model of rating agencies, both at an academic level as well as in policy circles. There exists general consensus on the need to strengthen the regulation of rating agencies. However, the proposals currently being discussed are insu?cient to resolve the con?ict of interest inherent in the issuer-pay business model of rating agencies. As Branson Davies puts it,40
“The rating agencies are so compromised that no amount of regulation can, in my view, make up for the the fundamental ?aws in their incentive structures which is simply a re?ection of a ?awed business model.”
We look at the proposals currently being discussed and reach a similar conclusion. As is clear from Table 2, the proposed reforms, by themselves, will have an insu?cient impact if the current issuer-pay model is maintained. However, drawing on various strands of the academic literature, we propose an alternative model for the industry which: • Aligns the incentives of the rating agencies • Puts a stop to rate-shopping • Leads to unbiased ratings • Makes the ratings industry commercially viable
40
see Davies (2008)
23
• Results in a market driven, competitive ratings industry We propose a return to an investor-pay model supplemented by a subsidy from the government. In order to keep the industry market-oriented and minimise the regulatory burden, we propose an auction mechanism to provide the ‘right to rate’ to the most e?cient rating agencies. We further propose repeat annual actions in order to provide a market-based ‘stick’ to discipline the rating agencies and ensure dynamism and competition in the industry. Furthermore, our model addresses two key criticisms of the investor-pay solution to the ratings conundrum. It creates a supplementary source of revenue for the rating agencies, mitigating the free riding problem. At the same time, our model ensures that rating agencies are not unduly in?uenced by a particular class of investors with their own agendas by forcing rating agencies to compete for investor-pay revenues by di?erentiating themselves and catering to all classes of investors.
24
References
Credit rating agencies: developments and policy issues. ECB Monthly Bulletin, pages 107 – 117, May 2009. A. Barr. S&P proposes changes to ratings process. http://www.
marketwatch.com/news/story/sp-plans-changes-ratings-cuomo/story.aspx?guid= {3A7B5846-1E5E-48D0-ACCB-96A22E992ADF}, February 2008. T. Beck. Reforming the Credit Rating Agency Industry. http://crisistalk.worldbank.org/ 2008/12/reforming-the-c.html, December 2008. B. Becker and T. T. Milbourn. Reputation and Competition: Evidence from the Credit Rating Industry. Harvard Business School Finance Working Paper, (09-051), October 2008. K. Binmore and P. Klemperer. The biggest auction ever: the sale of the British 3G telecom licences. Economic Journal, pages 74–96, 2002. P. Bolton, X. Freixas, and J. Shapiro. The Credit Ratings Game. Working Paper, 2009. G. Bonanno. Vertical di?erentiation with Cournot competition. Economic Notes, 15(2):68–91, 1986. N. Camanho, P. Deb, and Z. Liu. Credit Rating and Competition. SSRN eLibrary, 2009. C. d’Aspremont, J. Gabszewicz, and J. Thisse. On Hotelling “stability in competition”. Econometrica, 47(5):1145–1150, 1979. B. Davies. The case for central bank liquidity provision as a public-private partnership. FMG Special Papers SP182, Financial Markets Group, Oct. 2008. A. Dixit and J. Stiglitz. Monopolistic competition and optimum product diversity. The American Economic Review, pages 297–308, 1977. N. Economides. Minimal and maximal product di?erentiation in Hotelling’s duopoly. Economics Letters, 21(1):67–71, 1986. R. Engelbrecht-Wiggans. Auctions and bidding models: A survey. Management Science, 26(2): 119–142, 1980. EuroWeek. Say no to collective punishment for rating agencies.
http://www.euroweek.com/Article.aspx?ArticleID=2305725, September 2009.
25
Financial
Times.
Rating
agencies.
http://www.ft.com/cms/s/3/
83059044-995b-11de-ab8c-00144feabdc0.html?ftcamp=rss, September 2009a. Financial Times. Rating agency pressures build. http://www.ft.com/cms/s/0/9c859182-ae2211de-87e7-00144feabdc0.html, October 2009b. J. Gabszewicz and J. Thisse. Price competition, quality and income disparities. Journal of Economic Theory, 20(3):340–359, 1979. J. Gabszewicz and J. Thisse. Entry (and exit) in a di?erentiated industry. Journal of Economic Theory, 22(2):327–338, 1980. J. Gabszewicz, A. Shaked, J. Sutton, and J. Thisse. Price competition among di?erentiated products: a detailed study of a Nash equilibrium. STICERD-Theoretical Economics Paper Series, 1981. E. Gal-Or. Quality and quantity competition. The Bell Journal of Economics, pages 590–600, 1983. J. H. Gellert. Testimony Concerning: Transforming Credit Rating Agencies. United States House of Representatives, Committee on Financial Services, September 2009. H. Hotelling. Stability in competition. The economic journal, pages 41–57, 1929. N. Jenkinson. Ratings in structured ?nance: what went wrong and what can be done to address shortcomings? BIS-CGFS Papers, 32, July 2008. P. Klemperer. Some observations on the British and German 3G telecom auctions. P. Klemperer. What really matters in auction design. Journal of Economic Perspectives, pages 169–189, 2002a. P. Klemperer. How (not) to run auctions: The European 3G telecom auctions. European Economic Review, 46(4-5):829–845, 2002b. K. Lancaster. The economics of product variety: A survey. Marketing Science, pages 189–206, 1990. D. Monderer and M. Tennenholtz. K-price auctions. Games and Economic Behavior, 31(2): 220–244, 2000. D. Pal. Does cournot competition yield spatial agglomeration? Economics Letters, 60(1):49 – 53, 1998. ISSN 0165-1765. 26
J. C. Rochet, J. Mathis, and J. Mc Andrews. Rating the Raters. Banque de France – Toulouse School of Economics Conference on “Macroeconomics and Liquidity”, 2008. S. Salop. Monopolistic competition with outside goods. The Bell Journal of Economics, pages 141–156, 1979. A. Shaked and J. Sutton. Relaxing price competition through product di?erentiation. The Review of Economic Studies, 49(1):3–13, 1982. A. Shaked and J. Sutton. Product di?erentiation and industrial structure. The Journal of Industrial Economics, pages 131–146, 1987. D. L. Singer. Ratings Agencies 2007 = Equity Analysts 2000? http://bigpicture.typepad. com/comments/2007/08/ratings-agencie.html, August 2007. V. Skreta and L. Veldkamp. Ratings Shopping and Asset Complexity: A Theory of Ratings In?ation. 2008. J. Sutton. Sunk costs and market structure: Price competition, advertising, and the evolution of concentration. Mit Press, 1995. J. Tirole. The theory of industrial organization. MIT press, 1993. A. Toth. A new pricing scheme for credit rating agencies: a suggestion. Bank of England Note, January 2009. US Department of the Treasury - Financial Regulatory Reform: A New Foundation. http: //www.financialstability.gov/roadtostability/regulatoryreform.html, August 2009. R. Watson. SIFMA/ESF response to IOSCO technical committee consultation report on the role of credit rating agencies in structured ?nance markets. April 2008a. R. Watson. ESF/SIFMA response to CESR consultation paper on the role of credit rating agencies in structured ?nance. March 2008b. M. Zelmer. Reforming the Credit-Rating Process. Bank of Canada Financial System Review, December
p 51–57, 2007.
27
doc_373323561.pdf
Debt instruments the agencies rate may include government bonds, corporate bonds, CDs, municipal bonds, preferred stock, and collateralized securities, such as mortgage-backed securities and CDOs.
Credit Rating Agencies: An Alternative Model
Pragyan Deb† and Gareth Murphy‡ November 2009
Abstract We explore the con?ict of interest which arises in the credit ratings industry due to the current issuer-pay model. We argue that the most e?cient way of aligning the incentives of rating agencies is to return to an investor-pay system. Careful analysis of the reforms currently being discussed suggests that, by themselves, they will have insu?cient impact if the current issuer-pay model is maintained. While a return to the pre-1970s investor-pay model would solve the con?ict of interest problem, the issue of free-riding amongst some investors is likely to result in insu?cient revenues for the rating agencies. We argue that although free-riding is a problem, the increasing use of ratings by institutions coupled with the rise in the speed of information di?usion and predominance of electronic trading venues over the last few decades would ensure that there are investors willing to subscribe to ratings. This investor-pay revenue could be supplemented by a government subsidy to ensure that su?cient resources are available to the rating agencies. To fund the government subsidy, we propose that a small tax would be levied on issuers or at the point of issue. A limited number of rating licenses which provided a ‘right to rate’ which would be auctioned (just like 3G auctions) and the auction winners would be entitled to a portion of the tax pool which is paid in arrears and linked to the share of the investor-pay market that they manage to achieve. Such a system would align the incentives of the rating agencies with investors, would ensure a commercially viable ratings agency industry and would have negligible impact on primary issuance markets.
Keywords: Competition, Reputation, Financial Regulation, Auction, Industrial Organisation JEL Classi?cations: C7, D4, G2, K2, L1, L9
† London School of Economics and Financial Markets Group, E-mail: [email protected] ‡ Bank of England, E-mail: [email protected]
1
Introduction
The credit rating industry aims to o?er investors valuable information about ?rms in need of ?nancing. Due to asymmetric information between the ?rms and the investors, credit ratings often have a pivotal impact on the ?rms’ ?nancing outcomes. However since the early 1970s, rating agencies have relied on an issuer-pay model, creating a con?ict of interest - the largest source of income for the rating agencies are the fees paid by the very ?rms that the rating agencies are supposed to impartially rate.1 This tempts rating agencies to rate better than what fundamentals suggest, as many have pointed out during the recent subprime crisis. In this paper, we explore the con?ict of interest which arises in the credit rating industry due to the issuer-pay model. The rest of this section reviews the current business model of rating agencies and summarises the academic literature highlighting the problems with the current structure. Section 2 examines the proposals currently being discussed to reform the rating agencies, and suggests that by themselves, they will have an insu?cient impact if the current issuer-pay model is maintained. In Section 3, we propose an alternative structure for the industry - an investor-pay model, supplemented by a government subsidy to ensure su?cient resources are available to the rating agencies. Section 4 concludes.
1.1
Business Models: Past and Present
Before the 1970s, the ratings industry operated under an investor-pay model. Investors subscribed to ratings released by the agencies, and these subscription revenues were the main source of income for the rating agencies. However, owing to the public good nature of ratings and the increase in free-riding due to the spread of the photocopier, rating agencies switched to the current issuer-pay model. In 1970s Moody’s and Fitch switched to the issuer pay model with S&P following a few years later2 . As the market stands today, S&P and Moody’s (which account for around 80% of the market) rate and make public all SEC-registered ?rms, whether the rating is requested or not3 . If the issuer does not request a rating, the rating agency issues a rating based on publicly available information. If issuer requests a rating, it pays the rating fees and provides the rating agency with private information about the ?rm. Typically an issuer gets a higher rating after having solicited a rating from a rating agency.4
1 2
SEC, 2008, p.9 S&P switched to issuer pay model for municipal bonds in 1968 3 FITCH only does solicited ratings 4 There are 2 possible explanations for this -
1
The combined revenue from ratings of the big-three rating agencies - S&P, Moody’s and FITCH stood at US$3.7 billion in 2008. This, when compared with total rated bond issuance of US$3.9 trillion5 implies average revenue from rating in the order of 10bps. This is important for calibrating the subsidies in the investor-pay model outlined in section 3 since it suggests that any tax imposed to fund the subsidy is likely to be small and thus have a potentially negligible impact on the market.
Name Standard & Poors Ratings Services Moodys Investors Service, Inc. Fitch, Inc. A.M. Best Company, Inc. DBRS Ltd. Egan-Jones Rating Company Japan Credit Rating Agency, Ltd. LACE Financial Corp. Rating and Investment Information, Inc. Realpoint LLC Comments Largest player Second largest after S&P Much smaller, but part of big-three Specialises in the insurance market Focus on Canada Investor-pay Focus on Japan Investor-pay, specialises in ?nancial institutions Focus on Japan Specialises in structured ?nance market
Table 1: NRSRO designated rating agencies in US, as of September 2008
Apart from the big-three rating agencies, seven other rating agencies are recognised as Nationally Recognized Statistical Rating Organizations (NRSROs) in the US. However, they are comparatively tiny and they restrict their activities to particular market segments. An interesting trend in the last few years has been the emergence of investor-pay rating agencies, such as Egan-Jones Rating Company, LACE Financial and Rapid Ratings. Furthermore, as Firgure 1 shows, Moody’s issuer-pay revenues fell from 89% of total revenues in 1999 to 72% in 2008. These trends indicate that the investor-pay business model may be viable, although the issuer-pay model is by far dominant and remains the main source of revenue for rating agencies.
1.2
Problems with the current system
The current issuer-pay system creates a con?ict of interest for the rating agencies since their main source of revenue, i.e the rating fee, comes from the issuers that the rating agencies are supposed to impartially rate. An analysis of the annual reports of the big-three rating agencies suggests that issuer-pay rating revenues account for roughly 84% of Moody’s total revenues (average of last 5 years), while the corresponding amount for S&P and FITCH is 72% and 85% respectively.
• Sample selection bias - only ?rms with favorable information choose to solicit ratings. • Rating agencies are biased and improve the rating if they get fees.
Both ?gures were down sharply due to the subprime crisis. In 2007, ratings revenues stood at US$4.9 billion, while bond issuance was US$4.7 trillion.
5
2
US$ million 2500
2000
Other Revenue Ratings Revenue
1500
1000
500 86% 89% 86% 87% 83% 81% 80% 82% 83% 81% 72%
0 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Figure 1: Moody’s ratings revenue (issuer-pay) has fallen over time
While rating agencies have long claimed that this con?ict is unlikely to bias their ratings out of concern for their reputation, the recent sub-prime crisis has clearly shown that this is not the case. Every rating agency has its own published fee structure, but there is widespread use of negotiated rates for frequent issuers. In general, frequent, large issuers tend to establish long term relationships with the rating agencies, making the task of regulating the relationship much more di?cult. Thus, even if issuers are forced to pay rating agencies for initial analysis,6 it is unlikely to ensure proper incentives for rating agencies when the issuer-rater relationship is viewed as a long term relationship as opposed to a one-o? transaction. Since the crisis, a growing body of research has highlighted the weaknesses of the current model. Rochet, Mathis, and Mc Andrews (2008) show that reputational concerns are not enough to solve the con?ict of interest problem. In equilibrium, rating agencies are likely to behave laxly, i.e. rate bad projects as good and are prone to reputation cycles. While Rochet, Mathis, and Mc Andrews (2008) look at the con?ict in the context of a monopolistic rating agency, Camanho, Deb, and Liu (2009) show that the often sighted solution of introducing more competition is likely
6
as proposed under the Cuomo Plan
3
to aggravate the con?ict further and lead to increased ratings in?ation. This result stems from the fact that increasing the number of rating agencies with the aim of increasing competition, reduces an individual rating agencies reward from maintaining reputation. If the overall size of the ratings market remains the same, then with increased competition, each rating agency has a smaller share of the market. In such a situation, the bene?ts of maintaining reputation (and giving honest ratings) falls and ratings in?ation increases. Camanho, Deb, and Liu (2009) show that in general, moving from a monopolistic to a duopolistic setting will increase ratings in?ation and aggravate the lax behaviour of rating agencies. Becker and Milbourn (2008) measure empirically the relationship between competition amongst rating agencies on reputation building. They show that increased competition7 leads to ratings in?ation and less informative ratings. Thus, reputational concerns are not su?cient to discipline rating agencies. It is the case that this result depends critically on the assumption that the overall market size remains the same. However, increasing the size of the ratings market by requiring issuers to approach multiple rating agencies is also not likely to help. Bolton, Freixas, and Shapiro (2009) show that with multiple ratings, issuers have more opportunity to rate-shop8 and in the presence of naive investors,9 monopoly is superior to duopoly in terms of total ex-ante investor welfare. Therefore, in order to enhance competition, if we allow for a large number of rating agencies, more and more investors will ?nd it di?cult to infer rate-shopping and e?ectively the fraction of naive investors in the market would rise. At the same time, with more rating agencies, the issuers ability to rate shop would increase thereby decreasing ex-ante investor welfare. Skreta and Veldkamp (2008) do not consider the strategic behavior of rating agencies but explore the interaction between rate-shopping, complexity of the security10 and competition. They show that the intensity for rate-shopping increases with the complexity of the security and competition between rating agencies ampli?es this problem. Both these papers argue for a ban on rate-shopping. However, such a ban11 coupled with enhanced competition in the form of larger number of rating agencies will decrease the market size for each rating agency, make existing clients more valuable and thus reduce their incentive to maintain their reputation. Therefore, increasing the number of players while banning rateshopping is likely to aggravate ratings in?ation as highlighted in Camanho, Deb, and Liu (2009).
in the form of increased market share for FITCH get multiple ratings from di?erent rating agencies and make public only favourable ratings 9 investors who are unable to infer the extent of rate-shopping. 10 a key factor for structured products 11 which will be very di?cult to enforce given the long term relationship between the issuer and the rating agency
8 7
4
2
Proposed Reforms
There exists general consensus in policy circles on the need to strengthen the regulation of credit rating agencies, ‘including measures to promote robust policies and procedures that manage and disclose con?icts of interest, di?erentiate between structured and other products, and otherwise strengthen the integrity of the ratings process’.12 However most of the reforms currently being discussed attempt to increase regulatory oversight and disclosure norms in order to tweak the current issuer-pay system and make it more transparent without attempting to eliminate the inherent con?ict of interest outlined above. In what follows, we look at some of the major reform proposals and argue that, by themselves they will have an insu?cient impact. However, when combined with an investor-pay model (as proposed in Section 3), they are likely to be highly e?ective in enhancing transparency and e?ciency in the ratings industry. We look at the following reforms proposals 1. Lowering Barriers to Entry 2. Holding Rating Agencies Legally Liable for their Ratings 3. Taxpayer Funded Credit Rating Agency 4. Ban on Rate-Shopping and Consulting Services 5. Make Ratings more Informative
2.1
Lowering Barriers to Entry
Moody’s and S&P have dominated the credit ratings industry for years and together with FITCH control around 95% of the market. The ratings industry has large and often prohibitive barriers to entry. While some of these barriers such as reputation, scale economies and network e?ects13 are inherent in the business, others are a result of regulatory restrictions and insu?cient disclosure requirements. It has been suggested that if information is made more widely available, it might reduce entry barriers to this industry, increase competition and serve to alleviate the incentive problems in this industry. Typically, in the case of solicited ratings, issuers provide private information to rating
12 13
US Department of the Treasury - Financial Regulatory Reform: A New Foundation (2009) investors desire for consistency of rating categories across issuers limits number of players
5
agencies. If this information is revealed to the wider market, then it could incentivise other players such as insurance, pension and mutual funds, brokerages etc. to use this information in their own internal reports and investment decisions. They can then act as investor-raters making their reports public, providing the wider investor community with an additional source of information - a de-facto rating . This will enhance the level of competition in the ratings industry. However, as noted earlier, merely increasing competition in the ratings business is likely to be counterproductive. However, in this case, since the funds and brokerages undertaking research have a stake in their own ratings, they will not have the same con?ict of interest that plague the rating agencies. In essence, it will be similar to having incumbent rating agencies have a stake (legal or economic) in their ratings. However, the potential for market manipulation by the investor-raters needs to be taken into account. For example, investor-raters may be tempted to sell their stake just before a downgrade or issue biased ratings in order to move the markets in a particular direction. At this stage, a distinction needs to be drawn between brokerages, which are likely to have much wider coverage but less stake in their reports14 and mutual funds which would be much more selective on their coverage, but are likely to have a much larger stake in their reports.15 In general, wider coverage would mean having a lower stake in their reports. Thus, if the investorraters indeed morph into de-facto rating agencies and cover most asset classes, their stake in their reports would become marginal and they will end up with incentives very similar to the incumbent rating agencies. Even if these issues are resolved, we have another potential problem. Issuers may not be willing to make their private information public. While they may be willing to disclose the information to a rating agency in con?dence, they may undermine their competitive position if the same information is made available to the public and thus their competitors. Thus we might get 2 di?erent cases. In the ?rst case, if disclosure norms are not mandatory, then pure rating agencies will have access to additional private information provided by the issuers vis-a-vis investor-raters. Thus investors have to decide between 2 di?erent kinds of rating agencies • Pure Rating Agencies, that face a con?ict of interest and potentially issue biased ratings, but have access to private information of the issuers. • Investor-Raters, that face the right incentives, but are informationally constrained.
14 15
which is primarily used as a marketing product assuming their investment decisions are veri?ably based on these reports
6
Thus, this kind of a mechanism will be of limited bene?t and will make the job of the investors much more di?cult. The other alternative is mandatory disclosure of all private information. However, such a policy may not be in the interest of the issuers. In the case of corporate issuers, it risks exposing their company secrets to competitors. In the case of structured products, complete disclosure would result in lower pro?ts and thus decrease the incentive to innovate. The gains from ?nancial innovation will be capped because full disclosure will allow competitors to copy the product very quickly. Over time, this may sti?e innovation and may have detrimental welfare consequences. Furthermore, such blanket disclosure requirements will be very hard to implement and are likely to create ine?cient market structures. For example, if disclosure requirements are imposed on one part of the market (for example on exchange traded products), then it would encourage issuers to go for privately placed OTC instruments. Over time, the OTC market will expand and end up becoming the dominant part of the system.
2.2
Holding Rating Agencies Legally Liable for their Ratings
Currently, credit ratings are treated as ‘opinions’ and thus rating agencies cannot be held legally liable for their ratings which are protected under free speech. However, if rating agencies are made liable for the quality of their ratings,16 it would go a long way in resolving the con?ict of interest. Currently, the only factor restraining rating agencies is the concern for reputation, but if this channel is strengthened by forcing rating agencies to have a legal or ?nancial stake in their ratings, rating agencies would no longer ?nd it pro?table to in?ate ratings. This would have the same e?ect as having investor-raters in the market (see section 2.1) However, given the nature of the ratings industry, it would be very di?cult to force rating agencies to have a stake in their ratings. Given the capital structure of rating agencies and the size of the market that needs to be rated, a ?nancial stake in their ratings is not feasible.17 While a ?nancial or a legal penalty is possible in the event of ex-post poor quality rating, it would be very hard to implement. For the system of penalties for poor ratings to work e?ectively, it would require authorities (regulatory or judicial) to be able to distinguish between poor ratings arising out of rating agency bias or bad luck. If it is the former, then a penalty would be warranted, but in case of the latter, the penalty would lead to ine?ciencies.
like auditors unlike banks and other ?nancial institutions, rating agencies do not hold any capital. Hence the US proposal for ‘joint lability’
17 16
7
It is very unlikely that authorities would be able to make such a distinction with relative certainty, making a blanket penalty the only feasible option. But a blanket penalty every time the quality of ratings falls below a threshold would do little to solve the incentive problem of rating agencies. Such a blanket penalty would only serve to make rating agencies excessively conservative. This is because rating agencies, fearing the penalty would prefer to err on the side of caution and give lower ratings. Therefore such a proposal would essentially replace the upward bias in ratings with a conservative bias, leading to less informative ratings and possible capital misallocation. Furthermore, the US ‘joint liability’ proposal18 would make the entry barriers to the industry even more prohibitive. This is because under such a system, any new potential entrant may become liable for all ratings issued by other players. As James H. Gellert of Rapid Ratings International, Inc. puts it,19
“Why would one want to become an NRSRO joining a group dominated by three players with an iceberg of lawsuits looming on their horizon? That would be like swimming towards the Titanic.”
2.3
Taxpayer Funded Credit Rating Agency
A taxpayer funded rating agency would not be driven by commercial motives and thus such an agency is not likely in?ate ratings. Since investors are rational, they will take this into account and thus the taxpayer funded agency will have perfect reputation. If the taxpayer funded agency is just as e?cient as the private rating agency, it will capture the entire market. This is because given its perfect reputation and same cost to the issuer as private rating agency, an issuer with a good project will always prefer to approach the taxpayer funded rating agency for rating. The only case where an issuer will approach the private rating agency is when it had a bad project and hopes that the private rating agency will give it a good rating. But this will be fully anticipated by the market and the private rating agencies rating will be useless. Thus in this situation, the presence of the taxpayer funded rating agency will lead to the elimination of private rating agencies from the market. On the other had, the taxpayer funded rating agency might have a perfect reputation, but it may be less e?cient in identifying good and bad projects. Thus for the investors, the trade-o?
See EuroWeek (2009) and Financial Times (2009b) for details of the US proposals to make rating agencies jointly liable for their ratings 19 See Gellert (2009)
18
8
between private and public rating agency would boil down to a choice between between incentives and ability. While the taxpayer funded rating agency will give unbiased ratings, it would be less e?cient. On the other hand, the private rating agency would give biased ratings but will be more e?cient at identifying projects. This system by itself, will not solve the con?ict of interest of the private rating agencies. However, it might have a limited disciplining e?ect on private rating agencies by setting an upper limit for ratings in?ation. This is because the presence of a taxpayer funded rating agency may set a lower bound for reputation of the private rating agency. If the reputation of the private rating agency falls below the threshold, all issuers may approach the taxpayer funded rating agency (same mechanism as above) and the private rating agency may be eliminated from the market.
2.4
Ban Rate-Shopping and Consulting Services
Rate-shopping refers to issuers approaching multiple rating agencies for rating and selectively disclosing only those ratings that are favorable. Often issuers play one rating agency against the other in order to get favourable ratings. To quote former chief of Moody’s, Tom McGuire,20
“The banks pay only if [the ratings agency] delivers the desired rating. . . If Moody’s and a client bank don’t see eye-to-eye, the bank can either tweak the numbers or try its luck with a competitor. . . ”
Furthermore, rating agencies routinely provide ‘creative suggestions’ to issuers regarding the design of structured products in order to improve ratings. The recent subprime crisis, provides clear evidence of this. Instead of being neutral players, rating agencies were working actively with investment banks on the design of complex securities, such as CDOs and MBS. As a former Moody’s expert in securitization says “Every agency has a model available to bankers that allows them to run the numbers until they get something they like and send it in for a rating”.
Or, as Chris Flanagan, the subprime analyst at JPMorgan claimed
“Gaming is the whole thing. . . Banks were gaming the ratings and designing only the securities that were blessed by the rating agency.”
20
New York Times Magazine, Triple-A-Failure, April 27, 2008
9
This is of particular concern for complex structured products since small super?cial changes to the asset pool or the nature of the contract can have disproportionate e?ects on ratings. Thus, such ‘consultations’ can potentially amount to data mining,21 severely underling the ratings methodology. Given this situation, it has been suggested that requiring rating agencies to disclose all ratings irrespective of the wishes of the issuers as well as forcing rating agencies to make public any initial consultation they may have with issuers can put a stop to rate-shopping. However, under the issuer-pay system, given the complex nature of interaction between the issuers and the rating agencies, it would be very di?cult to enforce such a regulation. As is clear form the opposition of rating agencies to any such move,22 rating agencies are unlikely to actively cooperate in the implementation of such regulation. Thus, such a move would require constant regulatory supervision, making enforcement costly and di?cult. Furthermore, as pointed out in Section 1.2, such a ban would decrease the overall size of the ratings market (by reducing the incidence of multiple ratings), reducing rating agencies incentives from maintaining reputation and thus resulting in more ratings in?ation.23
2.5
More Informative Ratings
It has been argued that as it stands, ratings do not provide su?cient information to investors to make informed decisions. The current system of rating buckets are too coarse and provide insu?cient information to investors. The ratings bucket of di?erent rating agencies have di?erent meanings and they often hide important details such as variance, likelihood of default and/or loss severity in the event of default. This is particularly true in case of structured products which are extremely complex and investors need more information to make informed decisions. For example, under the probability of default based ratings approach (employed by S&P and FITCH) the rating agency calculates the likelihood of income stream of a CDO tranche falling short of its contractual value and determines the rating by comparing this likelihood with historically calculated default probability. Similarly, under the expected loss based approach used by Moody’s, percentage expected losses are calculated and compared with historical benchmark information. Both models provide an overall expected rating, but hide details regarding the distribution of losses etc. which are vital for informed investor decision.
21 Since every statistical model has some Type II error, i.e. failure to reject a null (high rating) when it is false, the issuer can get the desired rating by repeatedly tweaking the product. Thus through consultations, issuers can ensure that their products get a hight rating, even if it is not warranted 22 See Watson (2008a) and Watson (2008b) 23 See Camanho, Deb, and Liu (2009)
10
While increasing the information content of ratings is a step in the right direction, the e?ect of such a move on ratings bias is uncertain. Under the issuer-pay model, rating agencies will ?nd it optimal to in?ate ratings, even if the content of the rating is more informative. However, it is possible that more information will make it easier for investors to detect the bias and thus make it a little less pronounced. But fundamentally, making rating more e?ective is unlikely to solve the incentive problems of the issuer-pay model. However, as explained in the following section, making ratings more informative while switching to the alternative investor-pay model can mitigate the free rider problem to a large extent.
Proposal Current System Lower Barriers to Entry Holding Rating Agencies Legally Liable for their Ratings Taxpayer Funded Credit Rating Agency Ban Rate-Shopping & Consulting Services More Informative Ratings Investor-Pay Investor-Pay with subsidy
a b
Incentive Aligned No No No Yes No No Yes Yes
Commercially RateViable Shopping Yes Yes Yes No Yesa , Nob Yes Yes No Yes Yes Yes Yes No Yes No No
Direction of Rating Bias In?ation In?ation, potential market manipulation Conservative In?ation In?ation In?ation None None
if taxpayer funded rating agency is less e?cient assume taxpayer-funded rating agency is equally e?cient Table 2: Impact of Proposed Reforms
3
An Alternative Model
The analysis in section 2 suggests that the reforms currently being discussed will be insu?cient to solve the incentive problems inherent in the issuer-pay model. Thus we propose a return to the pre1970s investor-pay model. The investor-pay model is the most e?ective way of aligning incentives of rating agencies and investors and eliminating the con?ict of interest of rating agencies. However, it is also prone to the problem of free-riding, resulting in insu?cient resources for research and analysis and thus poor quality ratings. As the rating agencies point out, free riding was one of the main reasons behind the switch to an issuer-pay system in the 1970s. While it is true that free riding is a big issue, the increasing use of ratings by institutions, coupled with the sharp rise in the speed of information di?usion in the markets and the predominance of electronic trading venues over the last few decades has increased the speed of price response to 11
market events to only a few seconds.24 This is likely to ensure that there are investors willing to subscribe to a rating agency as opposed to waiting for selected parts of the information to leak through the market. The delay involved in free riding might be too costly for some investors. Thus, investors that require ratings information quickly will be willing to pay in order to avoid delays inherent in free riding. Investors such as large traders and brokerages, pension and mutual funds, insurance companies and other institutions that have structures in place to act on the information quickly will be willing to subscribe to the ratings as opposed to free riding and waiting for the information to arrive with a lag. Given the high speed of information di?usion, the opportunity cost of waiting is likely to be too high for such investors since the market would have already moved in response to the ratings information. On the other hand, investors who cannot trade immediately such as households and small retail investors, will have a low opportunity cost and they would likely choose to free ride. However, in equilibrium, there would always be some investors willing to subscribe to the ratings. In this context, regulation aimed at making ratings more informative25 is extremely important as it can potentially mitigate the free riding problem to a large extent. Even if investors are able to learn the broad ratings bracket through free riding, it is unlikely that they will get access to the necessary details fast enough if they try to free ride. They would need to subscribe to get reports, access to analysts etc. Thus, as ratings become more and more informative, the opportunity cost of free riding increases, mitigating the free rider problem to a large extent. More formally, assume that investors using credit rating for their investment decisions can choose to be subscribers or free riders. If they choose to be free riders, they do not have to pay the rating agencies subscription fee (s). However free riding involves a delay and we assume that the the opportunity cost of free riding for the investor concerned is ? . As a modeling device, we assume that there exists a continuum of investors with ? lying between a and b. Assuming a continuum of investors makes the model tractable26 . Intuitively, a represents the opportunity cost for a investor who would not act on the ratings information (and is thus close to 0) while b is the highest possible opportunity cost (representing institutional investors who would trade on rating information immediately). Thus, each investor decides whether to free ride or subscribe depending on ? . If ? > s, then the investor is better o? subscribing to the rating agency as the opportunity cost of free riding is too high. On the other hand, if ? < s, the investor prefers to free ride and wait for the information
24 25
[[[[[quote papers on e?ect of media on speed of information di?usion]]]]] See section 2.5 26 The result holds for any possible distributional assumptions
12
Free Riders
Subscribers
a
s
?
b
Figure 2: Investor Choice - Free Rider or Subscriber
b?s b?a .
to arrive with a lag. The fraction of investors subscribing to the rating agency is given by ? =
Note that the fraction of subscribers is a function of the fees charged by the rating agency and the information content of ratings. As the information content of the ratings increases (i.e. ? increases), the fraction of subscribing investors rises, thus mitigating the free rider problem. Also, as the subscription fee falls, the the fraction of subscribing investors rises. Therefore it is likely that there will always be a fraction of investors (?) who ?nd it pro?table to subscribe to the rating agencies reports. Assuming a unit mass of investors and a per-subscriber costs of rating c, the pro?ts of the rating agency takes the following form:
? = ?s ? c The problem arises when the fee charged is relatively modest and/or the fraction ? is small. In this case the investor-pay system may not generate enough revenues for the rating agencies, i.e. ? < 0. In such a situation, the rating agencies will not survive or will be forced to cut costs and compromise on the quality of their research. To avoid this scenario, it might be necessary to establish a supplementary source of revenue for the rating agencies. Such an alternate source of revenue can take the form of a government subsidy. In order to provide rating agencies with the correct incentives, the subsidy should be proportional to its subscription revenues. Such a subsidy will incentivise the rating agencies to maximise their investor-pay revenues, thus aligning incentives. The new pro?t function becomes
? = ?s ? c + ?s? An attractive feature of this structure is that it does not require costly regulatory information as once the level of subsidy is optimally determined, the distribution of the subsidy amongst the rating agencies is based on their respective market shares. This market determined distribution system is much more desirable than the regulator trying to determine the share of each rating 13
agency on a case by case basis. This system is dynamic in the sense that it re?ects the markets (investors) best guess about the relative usefulness of di?erent rating agencies ratings. If a particular rating agency does not perform up to standards, its market share is likely to fall over time, and with it the subsidy. Thus, this system provides a market determined measure of performance as opposed to a pre-set metric which can result in regulatory arbitrage and is open to abuse. Another big advantage of this system is that it makes rate-shopping much more di?cult for issuers. This is because with an investor-pay system, it is not in the interest of rating agencies to hide (or not make public) an unfavorable report. Thus, with an investor-pay system, any ban on rate-shopping will be much easier to implement since rating agencies would themselves have a strong incentive to comply with it. A potential problem with the above system is that the rating agencies incentives will be aligned only with fraction ? of the investors that subscribe to their ratings. If the incentives of the remaining (1 ? ?) free-riding investors are di?erent, then it might result in a con?ict of interest27 . This is particularly true if the di?erences in incentives are unknown or change in an unpredictable way, then it might turn out to be a potential problem. While we address this issue further in section 3.3, at this stage its important to point out that making credit ratings more informative should mitigate this to a large extent (see section 2.5). This is because, even if the incentives of the subscribing and free riding investors are not perfectly aligned, they are unlikely to be in direct con?ict with each other. They might have slightly di?erent preferences, but it is unlikely that they can gain at each others expense - as in the case of issuers vs investors where there is a clear and direct con?ict28 . Furthermore, under this system, the capital structure and revenue stream of the rating agencies are clear and open to scrutiny. Thus, if it would be easy to determine if a particular rating agency is under the in?uence of a special interest group. As discussed in section 3.3, other rating agencies, competing directly for market share, would do all they can to highlight this fact in order to win over subscribers. Finally, a potent safety feature of this system is that the free-riding subscribers would always have an option to subscribe to the ratings, thus aligning incentives limiting the potential damage from misaligned incentives.
For example, pension funds with a mandate to invest in only AAA securities might want more securities to be rated AAA 28 For example, a pension fund demanding AAA rated bonds, would still prefer bonds of a higher quality. They may prefer the rating brackets to be broadened to increase their scope and scale of operations, but they would still prefer to invest in bonds that do not default
27
14
3.1
Optimal Subsidy
From the regulatory perspective, determining the actual level of subsidy is not a trivial matter. Given the revenue stream of the rating agency, the subsidy should be such that the rating agency at least breaks even. Note that the break-even condition allows the rating agencies to make reasonable pro?ts in order to be viable and able to respond and invest to dynamic market conditions. We set pro?ts to 0 to simplify the exposition: ? = ?s ? c + ?s? = 0 c or ? ? = ?1 ?s In order to optimally determine ? ? , the regulator needs detailed and accurate information on the rating agency’s costs as well as demand, supply and optimal pricing of the ratings. While this is possible (as in the case of other regulated industries like electricity, gas, water etc.), this is likely to impose a heavy burden on the regulator. Furthermore, ever if the regulator could optimally set ? ? , since the subsidy is designed to ensure that the rating agencies break even in equilibrium, it would deter rating agencies from competing for market share. Given this problem, we need a mechanism which imposes less informational burden on the regulator and preserves the incentives of the rating agencies to compete amongst themselves for investor-pay revenues and market share. In order to ensure the latter, the subsidy should be ?xed a-priori allowing rating agencies to maximise their pro?t without worrying about an o?setting reduction in the subsidy. Thus the rating agency should reap the rewards from ex-post increase in market share arising from improvement in the ratings and overall quality of service, innovation, cost reduction etc. Since each rating agency is best suited to determine its own cost structure and expected revenues, we propose an auction mechanism, similar to the 3G auctions, to determine the optimal subsidy rate for the rating agencies. Each rating agency determines the minimum level of subsidy it requires to break even ?i = ci ?1 ?i s i
If the auction process is properly designed, then the most e?cient rating agencies, i.e. the ones that are con?dent of getting higher subscription revenues from the investors and have lower costs will win the auction. After the auction, the rating agencies will continue to compete since the subsidy is contingent on the investor-pay revenues.
15
The auction process would involve giving rating agencies the ‘right to rate’ and get the subsidy for a ?xed number of years. The optimal period would depend on the length of time the rating agencies require to recoup their costs and make reasonable pro?ts. Thus, the period would be a critical factor in determining the attractiveness of the auction process and thus the degree of competition. For the sake of this exposition, we assume that 5 years is optimal. Furthermore, in order to maintain the quality of ratings and make the ratings market more dynamic, we propose repeat annual auctions on a rolling basis. In the annual actions, the rating agency that has the lowest market share and has completed its guaranteed 5 year period will have to compete in the auction once again for the right to rate. This will ensure that there is enough competition in the market and incumbent rating agencies are forced to continuously compete and maintain best possible rating standards. Thus this mechanism would use the market to determine and punish the most ine?cient rating agency as opposed to having the regulator measure performance base on a particular metric.
3.2
The Auction Mechanism
A well designed auction is the method most likely to allocate resources to those who can use them most valuably since it forces business to put their ‘money where their mouths are’ when they submit bids.An auction can therefore extract and use information otherwise unavailable to the government.29 However, it is vital that the auction is properly designed and tailored to the particular context. We look at the literature on 3G licence auction for motivation on optimal auction design for the ratings business. Broadly, the optimal auction mechanism should achieve 3 main objectives: • It should allocate the ‘right to rate’ to the most e?cient rating agencies • It should promote competition • It should minimises the subsidy burden One of the key challenges in designing the auction for the ratings industry is the big advantage of incumbents30 over potential new entrants. As discussed earlier, the ratings industry is plagued by barriers to entry, both economic and regulatory. In addition to the large set-up and ?xed costs to entry, incumbents also have the advantage of established customer bases and brand-name
1 For example, the successful UK 3G auction yielded about 22.5 billion or 2 2 % of GNP and resulted in a competitive telecom market in the UK 30 particularly the big 3 rating agencies - S&P, Moody’s and FITCH 29
16
recognition. Thus, in order to make the auction process competitive, it is vital that the auction design provides su?cient incentives to new entrants. One of the key parameters to achieve this is the number of licences. Since there are three dominant incumbents in the market, the number of licences on o?er should be greater than three. Otherwise any new entrant would be deterred from participating in the auction process rendering it ine?cient.31 However, in the context of the ratings industry, given its current structure and the barriers to entry that already exists, giving out more than three licences, coupled with the ?xed term guarantee to recoup costs, might actually lead to a decline in entry barriers and foster competition. The other potential problem with the auction mechanism is that the incumbent rating agencies might feel forced to win a new licence in order to avoid a sharp reduction the value of their previous investments. The proposal to require the worst performing rating agency to participate in the auction process after the ?xed guaranteed term would lead to disruptions and can act as a disincentive for rating agencies from participating in the auction. However, such a mechanism is vital to ensure e?ciency and continued competition in the ratings industry. The incumbents’ disincentive from repeat auctions must be balanced with the need to maintain a dynamic system.32 While the exact length of time a rating agency needs to recoup costs is debatable, some kind of penalty for poor performance is necessary to maintain market e?ciency. In any case, after the ?xed term, the worst performing incumbent rating agency will still be allowed to compete in the auction process. We believe that the risk of losing the ‘right to rate’ is no di?erent from the general penalty for poor performance in the market. Risk is part of any business and rating agencies are free to take this into account while competing in the bidding process. In principle, the number of licenses can be changed to re?ect the prevailing market conditions Broadly, we have 3 distinct auction mechanisms: • Simultaneous Ascending Design equivalent to Second Price Auction • Sealed Bid Auction equivalent to First Price Auction • Hybrid Anglo-Dutch Design
For example, the Netherland 3G auction of July 2000 performed poorly because there were 5 licences with 5 incumbents, deterring potential new entrants 32 Repeat auctions can be made contingent of the market share of the worst performing rating agency falling below some pre-determined threshold. However, such a system risks cartelisation of the industry, with incumbents not competing with each other in order to sustain super-normal pro?ts.
31
17
combines element of both We analyse these 3 types of auctions with a view to see which is best suited for the ratings industry. Simultaneous Ascending Design In the simultaneous ascending auction, or the English Auction, participants bid openly against one another, with each subsequent bid higher than the previous bid. The auction ends when no participant is willing to bid further, at which point the highest bidder pays their bid. The distinguishing feature of this auction type is that the current highest bid is always available to potential bidders. Thus, in terms of outcomes, its equivalent to the sealed bid, second price auction. In the context of allocating multiple, equivalent licences, the auction ends when the number of participants remaining are equal to the number of licences available. Thus, the subsidy is equal to the level required to sustain the least e?cient rating agency.33 While this type of auction ensures that the most e?cient rating agencies win the auction and also the subsidy is kept at a minimum, it discourages competition and entry. In an ascending auction, there is a strong presumption that the ?rm that values winning the most will be the eventual winner, because even if it is outbid at an early stage, it can eventually top any opposition. Thus, dominant incumbents with established structures and client base can outbid potential entrants at any stage. This can create a situation where potential entrants are discouraged to even participate in the bidding process even in the presence of modest bidding costs. This results in insu?cient competition during the auction and can result in high levels of subsidy due to cartelisation of incumbents. Thus, in the presence of strong incumbents, this kind of auction often leads to insu?cient competition. Sealed Bid Auction In the sealed bid auction or more formally the ?rst price sealed bid auction, all bidders simultaneously submit sealed bids so that no bidder knows the bid of any other participant. Bidders make a single ‘best and ?nal’ o?er. The highest bidder, in this case the rating agency requiring the lowest subsidy, wins and the subsidy equals its own bid. From the perspective of encouraging more entry, the merit of a sealed-bid auction is that the outcome is much less certain than in an ascending auction. An advantaged incumbent will probably win a sealed-bid auction, but it must make its single ?nal o?er in the face of uncertainty about its rivals’ bids, and because it wants to get a bargain, its sealed-bid will not be the lowest
33
with more e?cient rating agencies making a pro?t
18
subsidy it could be pushed to in an ascending auction. So ‘weaker’ bidders have at least some chance of victory, even when they would surely lose an ascending auction Because sealed-bid auctions are more attractive to entrants, they may also discourage consortia from forming. If the strong rating agencies form a consortium, they may simply attract others into the bidding in the hope of beating the consortium. So incumbent rating agencies are more likely to bid independently in a sealed-bid auction, making this auction much more competitive. Thus the greatest advantage of this type of auction is that it gives weak bidders a chance (a ‘hope and dream’ in the words of one frustrated potential entrant) which can attract more bidders and results in more competition in the bidding process. However, this very fact is also its greatest shortcoming - because it allows bidders with lower valuations (higher costs in this context) to sometimes beat opponents with higher values (and so encourages entry) it is more likely to lead to ine?cient34 outcomes. Furthermore, sealed bids do not allow bidders to gather information on the business plans of their rivals by observing who is staying in and who is getting out as the price rises. They therefore make it impossible for bidders to re?ne their valuations of the licences on the basis of this information. Thus, this type of auction is only recommended when the most important objective is to encourage potential bidders to entry.35 Hybrid Anglo-Dutch Design The Anglo-Dutch design tries to marry the bene?ts of both type of auctions. In an AngloDutch auction for one object, the price rises until all but two bidders quit and the last two bidders then make a sealed bids. With 5 licences to sell, the subsidy would fall until only 6 competitors remained. The surviving bidders would then be committed to bid at or below this level of subsidy in a sealed-bid auction in which the four lowest bidders are awarded a licence. There are two versions of the Anglo-Dutch design; one in which each winner is committed to receiving its own bid, and one in which each winner is committed to paying the ?fth-highest winning bid. While the former is more likely to encourage competition, the latter is more e?cient in term of extracting private information of bidder and incentivising rating agencies to reveal their costs truthfully. The sealed-bid stage of the Anglo-Dutch design encourages competition by giving a chance to weaker players. Just as in the standard sealed-bit auction, weaker bidders have a chance of winning the auction and unlike the ascending price auction be outbid by the dominant incumbents. This encourages entry into the auction. However, as noted above, this also leads to ine?ciencies. The Anglo-Dutch design tries to overcome this by having an ascending auction as the ?st stage.
34 35
rating agency with higher costs winning the auction process As in the successful Danish 3G Auction of September 2001
19
This ensures that the auction does not result in very high levels of subsidies because of underbidding by all players. However, by having a sealed-bid stage, it encourages weaker participants to stay in the auction in the hope of winning in the last stage. Essentially, this design tries to balance the twin goals of encouraging competition, while at the same time ensuring that there is an e?cient outcome, i.e a high cost rating agency does not win the auction and the subsidy burden is minimised. The table below summarises the features of the 3 types of auctions Auction Type Simultaneous Ascending Design Sealed Bid Auction Hybrid Anglo-Dutch Design Entry Lowest Highest Moderate E?ciency Highest Lowest Moderate
Table 3: Comparision of di?erent Auction Types
Given that the initial auction is likely to o?er more licences than the number of incumbents, we believe that the Anglo-Dutch design is most appropriate in this setting. It balances the twin goals of encouraging entry while at the same time ensuring that the most e?cient rating agencies have a greater chance of winning, thus enhancing overall e?ciency and keeping the subsidy burden low. For the later annual auctions, we recommend the sealed bid auction. This is because in the annual rounds, an incumbent and potential new entrants must compete for a single licence, with 4 other incumbents already part of the industry. In such a situation, more incentives are necessary for new entrants and the sealed bid auction is most appropriate.
3.3
Industry after the Auction
After the auction, the industry would consist of a given number of rating agencies, competing with each other for market share. The pro?ts of an individual rating agency would depend critically on its investor-pay revenues, since the subsidy from the government would be linked to this revenue, thereby aligning the incentives of the rating agencies with the investors. A critical factor in determining the attractiveness of this design is the nature of competition in the industry after the auction and the mechanisms in place to prevent the rating agencies from coming under the in?uence of a particular class of investors. Insu?cient competition and cartelisation can lead to moral hazard problems for the industry leading to high cost and poor quality ratings. However, by its very design, the repeat auction mechanism ensures competition. The threat of new entry is likely to force the rating agency with the smallest market share to break the cartel and increase e?ort in order to gain gain market
20
share. This would in turn induce all other rating agencies to exert e?ort as no rating agency would want to risk competing in the auction and loose its ‘right to rate’. The more likely situation is intense competition and a race to the bottom. If the rating agencies are homogenous, then this is the likely outcome with each rating agency undercutting the other resulting in a ‘Bertrand Equilibrium’ with subscription fees close to 0. However, ratings are di?erentiated and we look at the Industrial Organisation (IO) literature to outline the nature of competition and di?erentiation in the ratings industry. Rating agencies can di?erentiate each other across two dimensions - the quality of their ratings and their coverage and specialisation. In the IO literature, the former is classi?ed as Vertical Di?erentiation while the latter is known as Horizontal Di?erentiation. Under both these settings, rating agencies would make positive pro?ts, have di?erent market shares and would cater to investors with di?erent preferences.36
Vertical Di?erentiation refers to di?erentiation on the basis of the quality of ratings. Rating agencies can choose to provide high quality rating and charge a higher price or they can provide lower quality ratings for a lower price. Investors choose between the di?erent rating agencies based on their own preferences i.e quality vs price. The market share of a rating agency would ultimately depend on the preferences of investors as well as the market segment the rating agency targets. The IO literature suggest that in general, rating agencies would ?nd it optimal to di?erentiate themselves as opposed to competing directly with each other by catering to the same market.37
Horizontal Di?erentiation
refers to specialisation in di?erent products. In this setting, dif-
ferent rating agencies would choose to specialise in di?erent ?nancial instruments and regions. For example, while one rating agency may specialise in corporate bonds the other may choose to specialise in structured products. Investors would choose di?erent rating agencies on the basis of their individual preferences. Thus an investor primarily investing in corporate bonds would go for the rating agency specialising in corporate bonds while an investor investing in structured products would subscribe to the rating agency specialising in it. Once again, the IO literature suggest rating agencies would choose to di?erentiate and cater to di?erent segments of the market38 Thus the ratings market can be thought o? as a two-dimensional space, with rating agencies choosing where to position themselves in order to di?erentiate themselves from their rivals (see
36 The current market structure shows some evidence of such di?erentiation, with the smaller rating agencies specialising in particular sectors or regions. See Table 1 for details. 37 See Gabszewicz and Thisse (1979, 1980), Gabszewicz and Thisse (1980), Gabszewicz, Shaked, Sutton, and Thisse (1981), Bonanno (1986) and Gal-Or (1983) 38 See Hotelling (1929) and Salop (1979)
21
?gure 3). The vertical axis shows quality while the horizontal axis represents specialisation in di?erent products. Investors choose the rating agency based on their preference for specialisation and quality. Since ratings are di?erentiated, the rating agencies have some market power and make pro?ts in equilibrium.
Differences in quality of ratings
Rating Agency
Vertical
Horizontal
Specialisation in different products
Figure 3: Market for Ratings - di?erentiated by quality and specialisation
Note that the distribution of investors in the two-dimensional space need not be uniform. For example, the proportion of investors demanding very high quality, more expensive ratings might be much higher than those demanding lower quality, cheaper ratings. The rating agency targeting the high quality segment of the market will in general have a higher market share. However, while such a position would be sought after by all rating agencies, once a particular rating agency positions itself in this position, the other rating agencies would ?nd it optimal to cater to a di?erent segment of the market. This is because rating agencies would prefer to relax price competition through product di?erentiation.39 If it so happens that most of the clients of one particular rating agency belong to a particular class (say pension funds), then some other rating agency would position itself to target a di?erent segment of the market. In so doing, they would continue to serve their may economic purpose
39
see Shaked and Sutton (1982)
22
- to facilitate collective research instead of each investor being forced to conduct research on a individual basis. The often sighed danger of investor-pay rating agencies coming under undue in?uence of a particular class of investors can be minimises further by requiring rating agencies to make public the distribution of their investor-pay revenues across the population of investors. Since rating agencies would already be providing information on their total investor-pay revenues to the regulators in order to get the subsidy, this should not impose undue burden on the rating agencies. Such disclosers would make any possible ratings bias clear to the investors. In case the market is unable to correct the potential biases, such data would also allow regulators to identify any concentrations or de?cits in the market and take corrective measures.
4
Conclusion
Since the credit-crisis, widespread concern has arisen about the functioning and business model of rating agencies, both at an academic level as well as in policy circles. There exists general consensus on the need to strengthen the regulation of rating agencies. However, the proposals currently being discussed are insu?cient to resolve the con?ict of interest inherent in the issuer-pay business model of rating agencies. As Branson Davies puts it,40
“The rating agencies are so compromised that no amount of regulation can, in my view, make up for the the fundamental ?aws in their incentive structures which is simply a re?ection of a ?awed business model.”
We look at the proposals currently being discussed and reach a similar conclusion. As is clear from Table 2, the proposed reforms, by themselves, will have an insu?cient impact if the current issuer-pay model is maintained. However, drawing on various strands of the academic literature, we propose an alternative model for the industry which: • Aligns the incentives of the rating agencies • Puts a stop to rate-shopping • Leads to unbiased ratings • Makes the ratings industry commercially viable
40
see Davies (2008)
23
• Results in a market driven, competitive ratings industry We propose a return to an investor-pay model supplemented by a subsidy from the government. In order to keep the industry market-oriented and minimise the regulatory burden, we propose an auction mechanism to provide the ‘right to rate’ to the most e?cient rating agencies. We further propose repeat annual actions in order to provide a market-based ‘stick’ to discipline the rating agencies and ensure dynamism and competition in the industry. Furthermore, our model addresses two key criticisms of the investor-pay solution to the ratings conundrum. It creates a supplementary source of revenue for the rating agencies, mitigating the free riding problem. At the same time, our model ensures that rating agencies are not unduly in?uenced by a particular class of investors with their own agendas by forcing rating agencies to compete for investor-pay revenues by di?erentiating themselves and catering to all classes of investors.
24
References
Credit rating agencies: developments and policy issues. ECB Monthly Bulletin, pages 107 – 117, May 2009. A. Barr. S&P proposes changes to ratings process. http://www.
marketwatch.com/news/story/sp-plans-changes-ratings-cuomo/story.aspx?guid= {3A7B5846-1E5E-48D0-ACCB-96A22E992ADF}, February 2008. T. Beck. Reforming the Credit Rating Agency Industry. http://crisistalk.worldbank.org/ 2008/12/reforming-the-c.html, December 2008. B. Becker and T. T. Milbourn. Reputation and Competition: Evidence from the Credit Rating Industry. Harvard Business School Finance Working Paper, (09-051), October 2008. K. Binmore and P. Klemperer. The biggest auction ever: the sale of the British 3G telecom licences. Economic Journal, pages 74–96, 2002. P. Bolton, X. Freixas, and J. Shapiro. The Credit Ratings Game. Working Paper, 2009. G. Bonanno. Vertical di?erentiation with Cournot competition. Economic Notes, 15(2):68–91, 1986. N. Camanho, P. Deb, and Z. Liu. Credit Rating and Competition. SSRN eLibrary, 2009. C. d’Aspremont, J. Gabszewicz, and J. Thisse. On Hotelling “stability in competition”. Econometrica, 47(5):1145–1150, 1979. B. Davies. The case for central bank liquidity provision as a public-private partnership. FMG Special Papers SP182, Financial Markets Group, Oct. 2008. A. Dixit and J. Stiglitz. Monopolistic competition and optimum product diversity. The American Economic Review, pages 297–308, 1977. N. Economides. Minimal and maximal product di?erentiation in Hotelling’s duopoly. Economics Letters, 21(1):67–71, 1986. R. Engelbrecht-Wiggans. Auctions and bidding models: A survey. Management Science, 26(2): 119–142, 1980. EuroWeek. Say no to collective punishment for rating agencies.
http://www.euroweek.com/Article.aspx?ArticleID=2305725, September 2009.
25
Financial
Times.
Rating
agencies.
http://www.ft.com/cms/s/3/
83059044-995b-11de-ab8c-00144feabdc0.html?ftcamp=rss, September 2009a. Financial Times. Rating agency pressures build. http://www.ft.com/cms/s/0/9c859182-ae2211de-87e7-00144feabdc0.html, October 2009b. J. Gabszewicz and J. Thisse. Price competition, quality and income disparities. Journal of Economic Theory, 20(3):340–359, 1979. J. Gabszewicz and J. Thisse. Entry (and exit) in a di?erentiated industry. Journal of Economic Theory, 22(2):327–338, 1980. J. Gabszewicz, A. Shaked, J. Sutton, and J. Thisse. Price competition among di?erentiated products: a detailed study of a Nash equilibrium. STICERD-Theoretical Economics Paper Series, 1981. E. Gal-Or. Quality and quantity competition. The Bell Journal of Economics, pages 590–600, 1983. J. H. Gellert. Testimony Concerning: Transforming Credit Rating Agencies. United States House of Representatives, Committee on Financial Services, September 2009. H. Hotelling. Stability in competition. The economic journal, pages 41–57, 1929. N. Jenkinson. Ratings in structured ?nance: what went wrong and what can be done to address shortcomings? BIS-CGFS Papers, 32, July 2008. P. Klemperer. Some observations on the British and German 3G telecom auctions. P. Klemperer. What really matters in auction design. Journal of Economic Perspectives, pages 169–189, 2002a. P. Klemperer. How (not) to run auctions: The European 3G telecom auctions. European Economic Review, 46(4-5):829–845, 2002b. K. Lancaster. The economics of product variety: A survey. Marketing Science, pages 189–206, 1990. D. Monderer and M. Tennenholtz. K-price auctions. Games and Economic Behavior, 31(2): 220–244, 2000. D. Pal. Does cournot competition yield spatial agglomeration? Economics Letters, 60(1):49 – 53, 1998. ISSN 0165-1765. 26
J. C. Rochet, J. Mathis, and J. Mc Andrews. Rating the Raters. Banque de France – Toulouse School of Economics Conference on “Macroeconomics and Liquidity”, 2008. S. Salop. Monopolistic competition with outside goods. The Bell Journal of Economics, pages 141–156, 1979. A. Shaked and J. Sutton. Relaxing price competition through product di?erentiation. The Review of Economic Studies, 49(1):3–13, 1982. A. Shaked and J. Sutton. Product di?erentiation and industrial structure. The Journal of Industrial Economics, pages 131–146, 1987. D. L. Singer. Ratings Agencies 2007 = Equity Analysts 2000? http://bigpicture.typepad. com/comments/2007/08/ratings-agencie.html, August 2007. V. Skreta and L. Veldkamp. Ratings Shopping and Asset Complexity: A Theory of Ratings In?ation. 2008. J. Sutton. Sunk costs and market structure: Price competition, advertising, and the evolution of concentration. Mit Press, 1995. J. Tirole. The theory of industrial organization. MIT press, 1993. A. Toth. A new pricing scheme for credit rating agencies: a suggestion. Bank of England Note, January 2009. US Department of the Treasury - Financial Regulatory Reform: A New Foundation. http: //www.financialstability.gov/roadtostability/regulatoryreform.html, August 2009. R. Watson. SIFMA/ESF response to IOSCO technical committee consultation report on the role of credit rating agencies in structured ?nance markets. April 2008a. R. Watson. ESF/SIFMA response to CESR consultation paper on the role of credit rating agencies in structured ?nance. March 2008b. M. Zelmer. Reforming the Credit-Rating Process. Bank of Canada Financial System Review, December
p 51–57, 2007.27
doc_373323561.pdf