are all ratings created equal? the impact of issuer size...
TRANSCRIPT
THE JOURNAL OF FINANCE • VOL. LXVII, NO. 6 • DECEMBER 2012
Are All Ratings Created Equal?The Impact of Issuer Size on the Pricing
of Mortgage-Backed Securities
JIE (JACK) HE, JUN (QJ) QIAN, and PHILIP E. STRAHAN∗
ABSTRACT
Initial yields on both AAA-rated and non-AAA rated mortgage-backed security (MBS)tranches sold by large issuers are higher than yields on similar tranches sold by smallissuers during the market boom years of 2004 to 2006. Moreover, the prices of MBSsold by large issuers drop more than those sold by small issuers, and the differencesare concentrated among tranches issued during 2004 to 2006. These results suggestthat investors price the risk that large issuers received more inflated ratings thansmall issuers, especially during boom periods.
AS THE MOST SEVERE financial and economic crisis since the Great Depressionunfolds, scholars, practitioners, and regulators have been studying its causesand possible cures to prevent similar crises in the future. At the center of thecrisis is the growth of the mortgage-backed securities (MBS) market, whichboth fueled and was fueled by the housing market boom. In this paper, we studythe role of the three main rating agencies—Moody’s, Standard & Poor’s (S&P),and Fitch—in the expansion of the MBS market. We test whether conflicts ofinterest may have played a role in the growth of MBS, and whether the marketpriced plausible measures of the intensity of conflicts related to issuer size andregulatory status.
Rating agencies play an important role in fixed income securities markets,in part because they have access to private information. Access to such infor-mation is protected from regulations such as Reg-FD, and ratings themselves
∗He is at the Terry College of Business, University of Georgia; Qian and Strahan are at theCarroll School of Management, Boston College and are affiliated with the Wharton FinancialInstitutions Center. Strahan is also affiliated with the NBER. We appreciate helpful comments fromCam Harvey (Editor); an Associate Editor; two anonymous referees; Efraim Benmelech; PatrickBolton; Jess Cornaggia; Gerard Hoberg; Christopher James; Brian Quinn; Amit Seru; Joel Shapiro;Richard Stanton; Dragon Tang; James Vickery; and seminar/session participants at Boston College,Brigham Young University, DePaul University, Federal Reserve Bank of New York, London Schoolof Economics, Northwestern University, Queen’s University (Canada), Simon Fraser University,University of Florida, University of Maryland, American Economic Association meetings (Denver),China International Conference in Finance (Beijing), European Finance Association meetings(Frankfurt), the 21st Conference on Financial Economics and Accounting (College Park), and theNBER conference on securitization. We thank Calvin Chau, Hugh Kirkpatrick, Sailu Li, YingzhenLi, and Chenying Zhang for excellent research assistance and Boston College and University ofGeorgia for financial support.
2097
2098 The Journal of Finance R©
are incorporated into regulations of many financial institutions. Abundant ev-idence shows that credit ratings contain information not imbedded in pricesfor corporate bonds, and Jorion, Zhu, and Shi (2005) show that the impact ofrating changes on stock prices becomes stronger after Reg-FD. Ratings are alsoshown to be an important determinant of corporate decisions such as capitalstructure (e.g., Kisgen (2006)).
Rating agencies, however, have come under criticism for practices that mayhave spurred undue expansion and then collapse of the MBS market. Manycritics emphasize a potential conflict in the way agencies structure their fees.Instead of being compensated by “consumers” (e.g., institutional investors) forproducing high-quality ratings, agencies are paid by issuers. The conflict ofinterest hypothesis thus posits that rating agencies may grant more favorableratings to issuers who may be able to bring, or potentially take away, substan-tial current and future business. In addition, regulations contingent on ratingsmay further distort the incentives of both issuers and rating agencies, sinceholding highly rated MBS securities lowers the burden of capital requirements.
The risk of lost reputation weighs against potential conflicts of interest forthe rating agencies. As recent theoretical work shows, however, several forcesmay have tilted toward rating “inflation,” especially for large MBS issuers. Un-like the corporate bonds market, a small number of large issuers of MBS bringmany deals to the rating agencies and thus may have greater bargaining powerthan large bond issuers (e.g., Frenkel (2010)). Perverse incentives of the ratingagencies worsen during market booms, when the short-term benefits of addi-tional rating business net of potential reputational costs are the highest (e.g.,Bar-Isaac and Shapiro (2010), Bolton, Freixas, and Shapiro (2012)). Moreover,more complicated MBS tranches were packaged and sold during 2004 to 2006,thereby increasing ratings disagreement. Disagreement increases issuers’ in-centive to “shop for better ratings,” even if each rating agency truthfully reportsits findings, because an issuer can purchase and report the most favorable rat-ing(s) after receiving preliminary opinions from multiple agencies. Shoppingmay thus lead to inflated ratings (e.g., Mathis, McAndrews, and Rochet (2009),Skreta and Veldkamp (2009)). To summarize, the booming housing and MBSmarkets between 2004 and 2006, with the associated growth in revenues forrating agencies and increased complexity of deals, may have worsened conflictsof interest and pushed toward leniency. These observations provide the basisof our empirical tests.
We match price histories, initial yields, and ratings from Moody’s, S&P, andFitch for a large sample of privately issued (i.e., not backed by government-sponsored enterprises, or GSEs) MBS between 2000 and 2006 with informationon the market share of issuers.1 We also obtain information on the character-istics of the tranches (e.g., size of principal amount, weighted average life,
1Throughout most of our sample period there were just four Nationally Recognized StatisticalRatings Organizations (NRSROs)—Moody’s, S&P, Fitch, and DBRS, which achieved NRSRO statusin 2003. However, DBRS focused almost exclusively on the corporate bond market (Kisgen andStrahan (2010)).
Are All Ratings Created Equal? 2099
geographical distribution of the underlying mortgages), deals (e.g., number oftranches, degree of subordination), as well as issuers (e.g., issuer type andrating at the issuance date). Our tests are based on cross-sectional differencesbetween tranches sold by large issuers versus small issuers (where issuer sizeis based on the issuing institutions’ one-year lagged annual market share),between tranches sold by regulated versus less-regulated issuers, and betweentranches sold in market boom years versus earlier years.
The fraction of a deal financed at AAA (highest possible rating) offers a sim-ple measure of the aggressiveness with which a deal has been rated. Using asmall set of collateralized debt obligations (CDOs) issued between 1997 and2007 for which they have the full set of collateral characteristics, Griffin andTang (2012) document that a major rating agency increased the fraction ratedAAA by 12% on average, relative to what was implied by their quantitativemodels, consistent with ratings inflation. Our study differs from theirs in thatwe explore a much broader set of deals, including all residential and commer-cial MBS available from SDC and Bloomberg. Our approach allows us to testhow issuer characteristics (e.g., issuer size and regulatory status) and marketconditions affected market expectations about the integrity of the rating pro-cess. The disadvantage, however, is that we have less ability to control fully forcollateral quality. Thus, rather than model the fraction rated AAA, we focus onmarket prices.
If rating agencies treat different kinds of MBS issuers equally, yields onsimilarly rated securities ought to be similar; that is, they ought not to reflectissuer characteristics. Conversely, if investor concerns about ratings inflationworsen with issuer size, then yields may reflect such concerns. In our primaryset of tests we thus compare initial yields (ex ante credit quality) of tranchessold by large versus small issuers, conditional on the credit rating. This yieldspread is about 10% higher on tranches sold by large issuers than on similarlyrated tranches issued by small issuers during market boom years.2 The effectis similar in both AAA and non-AAA markets, suggesting that investors areskeptical even of tranches receiving the highest possible rating. The estimatedcoefficients translate into an increase in yields of about 15 basis points (bps),relative to a mean spread of 147 bps, for large-issuer tranches. We find nosignificant difference in yield spreads, however, during non-boom years. Thisresult implies that investors recognize that potential conflicts of interest mayworsen during booms, leading to compromise in the rating process, and, as aresult, demand a price discount on the large-issuer tranches. These results arerobust to the inclusion of issuer fixed effects.
The credit rating process, beyond conflicts related to the issuer-pay fee struc-ture, may also have been distorted by financial institutions’ attempts to exploit“regulatory arbitrage” opportunities. For example, banks could reduce requiredcapital by transforming mortgages (held in the banking book) to highly rated
2In empirical tests we split our sample into AAA and non-AAA tranches; for tranches with morethan one rating, we define them as having an AAA rating only if all of the (reported) ratings areAAA (or equivalent).
2100 The Journal of Finance R©
MBS held in the trading book (Acharya and Richardson (2009)). In addition, inJuly 2004 the U.S. bank and thrift regulators exempted depository institutionsfrom FASB rule “Fin 46,” which had forced consolidation of most securitizedassets onto the balance sheet in the aftermath of the Enron scandal. This rul-ing allowed depositories to create “shadow banks,” off-balance sheet conduitsholding long-term securitized assets financed with short-term asset-backedcommercial paper (ABCP). These structures reduced the capital requirementto zero, while leaving all of the risk with the issuing banks, which typically pro-vided the conduits with liquidity guarantees to facilitate the sale of the ABCPs(Acharya, Schnabl, and Suarez (2012)). Following this decision, the ABCP mar-ket boomed, with outstandings rising from about $600 billion in July 2004 toits peak of $1.2 trillion by the summer of 2007. We find that MBS issued by de-positories following the July 2004 decision had yields about 10% to 17% higherthan tranches sold by less regulated institutions. We also find higher yields onAAA-rated tranches of more complex deals, proxied by the number of tranches(Furfine (2010)), as well as a trend increase in deal complexity during 2004to 2006. Overall, both increasing deal complexity and regulatory arbitrage didseem to have distorted the rating process and markets (Opp, Opp, and Harris(2012)). Controlling for both effects, however, changes neither the magnitudenor the significance of the effect of issuer size on yields.
We also obtain a number of interesting results on how the market pricesMBS tranches. For example, more ratings equate to lower yields. This ef-fect is most pronounced among small issuers in the AAA market. Specifically,AAA tranches issued by small issuers have yields about 19% higher whenrated by one agency compared to similar tranches rated by all three agen-cies. This suggests that investors price the risk that issuers shopped for thebest rating when tranches have fewer than three ratings. By shopping, anissuer could censor out pessimistic ratings, thus reducing the number of rat-ings observed by investors.3 Consistent with this incentive, we also find thattranches issued where rating agencies disagree have initial yield spreads thatare 10% higher than those of tranches receiving the same rating across multipleagencies.
In the secondary set of tests, we also examine the ex post performance of MBSsecurities by looking at price changes between origination and April 2009. BothAAA- and non-AAA-rated tranches sold by larger issuers in the boom performworse than similar tranches sold by smaller issuers—during boom years, pricesfor these large-issuer tranches drop about 10% more than prices for similartranches sold by small issuers. (This result is robust to the inclusion of issuerfixed effects for the non-AAA rated tranches only.) In addition, we find that pricechanges are attenuated slightly when we control for the initial yield, suggestingthat markets rationally incorporate concerns about the rating process into exante pricing.
3As pointed out by Opp, Opp, and Harris (2012), ratings shopping need not lead to inflation ifrating agencies account for this behavior and rate more conservatively than they otherwise wouldsuch that each rating is unbiased conditional on being the best rating chosen.
Are All Ratings Created Equal? 2101
Our paper contributes to the literature on the role of credit ratings in thefinancial crisis. Prior work examines lending practices as a potential cause forthe run-up in house prices (e.g., Mian and Sufi (2009), Keys et al. (2010),Loutskina and Strahan (2010)). Several papers empirically examine creditratings in structured finance markets (e.g., Adelino (2009), Benmelech andDlugosz (2009a, 2009b), Nadauld and Sherlund (2009), Ashcraft, Gold-smith-Pinkham, and Vickery (2010), Demiroglu and James (2010), Griffin andTang (2012)). These studies find that ratings are not always accurate mea-sures of default risk; nor are they a sufficient statistic for risk. Adelino (2009)shows that yield spreads add incremental explanatory power beyond ratingsin forecasting defaults. Griffin and Tang (2012) document flaws in how rat-ing agencies use their internal models, and Ashcraft, Goldsmith-Pinkham, andVickery (2010) show that simple observable measures of collateral risk forecastdefault conditional on the credit rating in a sample of Alt-A and subprime MBS.Our paper is the first to test for incentive problems related to issuer size, andto test whether the market incorporates concerns about the integrity of therating process into ex ante pricing and ex post performance.
Prior research also examines conflicts of interest facing financial institutionssuch as investment banks (e.g., Kisgen, Qian, and Song (2009)) and subprimelenders (e.g., Alexander et al. (2002)), but studies of conflicts facing ratingagencies focus mainly on the corporate bond market (e.g., Becker and Milbourn(2010), Bongaerts, Cremers, and Goetzmann (2012)). Our work shows thatconflicts may be exacerbated in new and booming markets such as MBS, andalso that investor wariness of this problem affects prices.
The rest of the paper is organized as follows. In Section I we review theevolution of the MBS markets and discuss our hypotheses and tests. We thenintroduce our data on MBS securities in Section II and present results fromour empirical tests in Section III. We conclude in Section IV.
I. Overview of Credit Ratings and MBS Markets
Prior research documents that rating agencies play a key role in the tra-ditional corporate bond market. Credit ratings are perhaps more importantin the recently developed markets for structured finance products, includingMBS securities, for several reasons. For one, the cash flows and risks of cor-porate bonds are tied to the performance and prospects of one company. Bycontrast, structured finance involves a complicated securitization process, withpooling and tranching of credit-sensitive assets. For a fixed collateral pool (inthe case of MBS these would be home mortgages), structured finance sepa-rates payments to investors into prioritized claims called “tranches,” whichabsorb losses from the underlying portfolio following seniority. Hence, rat-ings depend on the quality of the collateral as well as the seniority and de-gree of subordination of the tranche. While securitization has revolutionizedfixed income markets and brought billions of dollars to investment banks,for many investors this process can be opaque and tainted by asymmetric
2102 The Journal of Finance R©
information and moral hazard problems.4 To the extent that uninformed in-vestors trust the rating agencies to assess these complicated securities, creditratings likely play a more important role in the structured finance product mar-ket than in the corporate bond market, where independent research is morefeasible.
There is also strong demand among various types of institutional investors.For pension fund managers focusing on the fixed income markets and seekinghigh returns but constrained by the level of risk, highly rated MBS tranchesoffer an ideal vehicle. The securitization process described above can producemany more AAA-rated tranches than the fraction of AAA-rated corporate bonds(just 1% of which are AAA rated). The pooling and tranching process eliminatesmost of the idiosyncratic risk of the underlying assets, while the remainingsystematic risk leads to higher expected returns (Coval, Jurek, and Stafford(2009a)). For banks, broker dealers, and insurance companies, credit ratingsaffect the amount of capital needed to hold in reserve. Seemingly safe AAA-rated structured finance products also expand the supply of collateral to backrepurchase agreements that many money market mutual funds use to managetheir liquidity risk (Gorton and Metrick (2011)). Moreover, Fannie Mae andFreddie Mac purchased huge volumes of AAA-rated structured MBS that theycould finance at below-market borrowing rates due to their special status asgovernment-sponsored enterprises.
For rating agencies, the new fixed income products emerging out of thegrowth of structured finance provide substantial revenue potential beyond theirtraditional market of corporate bonds. The total volume of originations of sub-prime mortgages, for example, rose from $65 billion in the late 1990s to over$600 billion in 2006. In the case of Moody’s, profits tripled between 2002 and2006. At the peak of the market, Moody’s disclosed that 44% of its revenuescame from rating structured finance products, exceeding the 32% earned fromrating corporate bonds. There is also direct evidence that rating agencies offerprice discounts for large and frequent issuers of corporate bonds.5 It is naturalto expect that such practice also exists in dealing with large issuers of struc-tured finance products including MBS. As pointed out above, issuance is morehighly concentrated in structured finance, with large financial institutions suchas banks and investment banks being key players. This concentration impliesthat some large issuers have substantial bargaining power as they can bring,and certainly take away, rating business. The confluence of tremendous new
4See, for example, Coval, Jurek, and Stafford (2009b) for a review of structured finance, andAshcraft and Schuermann (2008) for a review of potential problems of the securitization process.See Keys et al. (2010) for evidence that securitization led to lax screening by lenders.
5According to S&P’s disclosure reports (including rating fee structure) in 2008, corporate is-suers typically pay “up to 4.25 bps for most transactions” and the minimum fee is $67,500. Inaddition, “S&P will consider alternative fee arrangements for large volume issuers and othercompanies that want multi-year ratings services agreements” (Standard and Poor’s (2008)). SeeBecker and Milbourn (2010) for more details on the practice of rating agencies in the corporate bondmarket.
Are All Ratings Created Equal? 2103
revenue flows in the late 2000s with significant bargaining power of largeissuers thus worsened the conflict of interest problem inherent in the agencies’issuer-pay fee structure.6
Given these potential conflicts, we test whether investors priced the riskthat credit rating agencies favored large issuers over small ones, and whetherthis effect grew stronger as the market boomed. Ratings shopping may alsocompromise the integrity of the rating process. Issuers sometimes receive pre-liminary opinions to determine whether to purchase a rating. Shoppers willtend to censor out pessimistic ratings, thus leading to inflated purchased andobserved ratings, regardless of whether rating agencies truthfully convey theirown information. The direct impact of ratings shopping is not observable, sinceissuers are not required to disclose all the contacts they have made with rat-ing agencies (see Sangiorgi and Spatt (2010) for more details). We do, however,control for the potential effects of shopping by including the number of reportedratings and rating disagreement among multiple agencies. Finally, given thesignificant benefits of packaging and holding highly rated MBS securities, weexamine whether ratings-based regulations further alter the incentives of bothissuers and rating agencies. For example, institutions facing tighter regula-tions may securitize their assets more aggressively, which leads to differencesin deal structure, collateral quality, and pricing.
We build a large sample of non-GSE-backed MBS tranches issued duringthe period 2000 to 2006, matched to characteristics of their issuers. As dis-cussed above, we take a valuation from outside approach to examine ourmain hypotheses—whether and when investors and markets recognize poten-tial problems in the rating process. For example, investors may have initiallyfailed to distinguish the credit quality of similarly rated tranches based onissuer size. Later on, as the housing market began to unwind, investors mayhave begun to recognize the difference in these two groups and adjusted yieldsaccordingly.
We conduct both an ex ante price test and an ex post performance test. First,we examine whether investors and the market recognize potential ratings infla-tion when they price tranches at issuance, conditional on the credit rating. Wecompare the yields (at issuance) on securities sold by large versus small issuers.If the market believes that large issuers receive more favorable treatment fromrating agencies due to the reasons discussed above, then their tranches oughtto have higher credit risk (due to more aggressive subordination structuresand/or riskier underlying collateral) and thus command higher initial yields.Second, we study the post-issuance performance of these two groups of secu-rities by looking at their (cumulative) price changes between origination andApril 2009. If large issuers enjoy favorable ratings and the market does not
6Rating agencies have also been criticized for using models that tend to overestimate the like-lihood of rising and high levels of housing prices, and thus underestimate the default risk of MBSsecurities. Our focus is not on the accuracy of these rating models per se, but rather on whetherand how the market prices MBS securities issued by large versus small issuers differently due tothe possible conflict of interest problem.
2104 The Journal of Finance R©
fully price this into initial yields, then the securities they sell ought to performworse than otherwise similar securities sold by small issuers when the marketworsens in 2007. Taken together, these two sets of results should give us amuch better idea of how the adverse incentive problem may affect the qualityof ratings during one of the worst crises in history.
II. Data and Methods
We begin the process of data compilation with the Securities Data Corpora-tion (SDC) database, which provides a large sample of tranches of privatelyissued (i.e., non-GSE) MBS deals. For each deal, SDC provides basic informa-tion on asset/collateral types (mortgage, credit card, auto loans, bonds, etc.),the number of tranches, as well as the issuer(s) and bookrunner(s). For otherdeal and tranche characteristics, including initial and subsequent ratings andprices, principal amount, coupon type and rate, and maturity (weighted av-erage life, and whether the tranche is paid off prior to April 2009), we relyon manually collected data from Bloomberg. Our sample includes MBS dealsoriginated and issued in 2000 through 2006, and we follow the prices of thesedeals through April of 2009.
A. Empirical Models
We estimate two sets of models relating issuer size and market conditions to(1) yield spreads at issuance and (2) price changes from the issuance date toApril 2009. The key explanatory variables are the lagged market share of theissuer (Issuer Share) and its interaction with HOT, defined as the fraction ofthe total principal amount of all tranches issued in a given year over the totalamount issued across all years. To summarize analytically,
Log of Yield Spreadi, j,t = β1 Issuer Sharek,t−1 + γ 1 Issuer Sharek,t−1
× Hott + Initial Rating, Fraction AAA
(Level of Subordination),
Collateral and Issuer controls + e1
i, j,t
(1)
Price Changei, j,t = β2 Issuer Sharek,t−1 + γ 2 Issuer Sharek,t−1 × Hott
+ Initial Rating, Fraction AAA (Level of Subordination),
Log of Yield Spread, Collateral and Issuer controls
+ e2
i, j,t.
(2)
The data vary by year (t), issuer (k), deal (i), and tranche (j). In analyzingex ante pricing (Equation 1), estimated at the tranche level, we control for
Are All Ratings Created Equal? 2105
variables related to deal structure in some specifications (e.g., Fraction AAA,or, for non-AAA tranches, Level of Subordination). In Equation (2), we intro-duce Log of Yield Spread as an additional regressor in some models, since thisvariable is predetermined relative to the ex post outcome. We also report equa-tions (1) and (2) in their reduced forms—that is, without including FractionAAA (and other deal structure terms) in (1) and without deal structure termsand Log of Yield Spread in (2)—to estimate the total impact of issuer size onyields and price changes.
In all of our tests, we include issuance-year fixed effects, and we double-cluster for all tranches sold by the same issuer and in the same year tobuild standard errors.7 Note that by including the issuance-year effects,we absorb the direct effect of HOT, which has only time variation butno cross-sectional variation; hence, we only report its interaction with is-suer size. We also report all of our models with and without issuer fixedeffects.
B. Variable Construction and Summary Statistics
We obtain ratings from the largest three rating agencies, Moody’s, S&P, andFitch. More tranches are rated by S&P than Moody’s or Fitch, but even Fitchrates over half of the tranches. Each of the three agencies rates around 60%of all the tranches AAA, but the AAA-rated tranches are larger and constituteabout 90% of the total amount of financing.
B.1. Dependent Variables
Table I, Panel A, reports summary statistics for the overall sample. We havetwo sets of market-based variables to measure ex ante pricing and ex postperformance. Log of Yield Spread equals the natural log of the yield spreadof a tranche at issuance. For a tranche with a floating coupon rate, the yieldspread is defined as the fixed markup in bps over the reference rate specifiedat issuance (e.g., the one-month LIBOR rate). For a tranche with a fixed orvariable coupon rate, the yield spread equals the difference between the initialcoupon rate and the yield on a Treasury security whose maturity is closestto the tranche’s weighted-average life. The mean yield spread is 147 bps overthe whole sample; since there are on average about 15 tranches per deal,the sample for this variable has more than 65,000 observations (only about2/3 of these observations end up in the regressions due to missing values onother dimensions). Price Change equals the percentage change in the price of
7To estimate the double-clustered standard errors by issuer and cohort year, we use theStata code “cgmreg.ado,” downloaded from Doug Miller’s website: http://www.econ.ucdavis.edu/faculty/dlmiller/statafiles/. This program is used to run OLS and perform multidimensional clus-tering as described in Cameron, Gelbach, and Miller (2006).
2106 The Journal of Finance R©
Table ISummary Statistics of MBS and Issuer Characteristics
This table reports summary statistics of privately issued MBS sold between 2000 and 2006 FractionAAA and Number of Tranches in a Deal are deal-level variables; all others vary at the tranche level.Variables are defined in Section II.B of the text. ∗∗∗, ∗∗, and ∗ in Panel C indicate significance at the1%, 5%, and 10% levels, respectively, for either a two-sample t-test (on means) or a non-parametricmedian test (on medians) between characteristics of MBS sold by big and small issuers in a giventime period.
Variable N Mean Median Std p25 p75
Panel A: Overall Summary Statistics
Fraction AAA (in%) 5,548 88.73 93.79 13.08 82.97 97.18Initial yield spread (in bps) 65,895 147.07 120.00 447.63 51.49 185.00Price change (in%) 9,299 −14.90 −0.81 28.36 −24.01 0.03Issuer market share 85,272 0.05 0.04 0.04 0.03 0.07Hot MBS market 86,635 0.17 0.19 0.06 0.11 0.25Principal amount (in millions) 86,625 65.12 14.35 172.16 3.47 51.50Weighted average life (in
years)70,484 5.63 4.90 3.36 3.28 7.26
Fra. of Colla. in troubledstates (in%)
80,536 45.51 45.70 16.51 34.70 54.90
Herfindahl index of collateral 80,536 0.34 0.33 0.09 0.29 0.36Initial rating 77,261 2.02 1.00 1.42 1.00 3.00Issuer rating 77,198 2.90 2.67 0.91 2.44 3.11Number of tranches in a deal 5,910 14.66 14.00 10.33 8.00 19.00Level of subordination (in%) 83,655 93.00 97.00 12.00 93.00 99.00
Panel B: Description of Ratings Distribution
Rating Agencies
Number of tranches in sample Moody’s S&P Fitch
AAA 30,390 38,169 21,930Non-AAA 21,090 26,858 12,148Total 51,480 65,027 34,078
Number Initial Ratings Freq. Percent
1 15,031 19.452 51,136 66.193 11,094 14.36Total 77,261 100.00
Same Originator Servicer Freq. Percent
0 (“different”) 14,872 35.491 (“same”) 27,034 64.51Total 41,906 100.00
Rating disagreement Freq. Percent
0 (“same”) 54,063 86.881 (“different”) 8,167 13.12Total 62,230 100.00
(Continued)
Are All Ratings Created Equal? 2107
Table I—Continued
Panel B: Description of Ratings Distribution
Types of Issuers Freq. Percent
Commercial Banks 33,670 38.86Investment Banks 19,077 22.02Thrifts 17,388 20.07Finance Companies 8,013 9.25Others 8,487 9.80Total 86,635 100.00
Deals with Floating Tranches Freq. Percent
0 (“no”) 1,626 27.511 (“yes”) 4,284 72.49Total 5,910 100.00
Panel C: Distribution by Issuance Year and Issuer Size
2000–2003 2004–2006
Small Big Diff Small Big Diff
Principal amount N 7,835 18,605 25,164 35,021(in 000s) Mean 68.06 54.15 13.91∗∗∗ 74.44 63.59 10.85∗∗∗
Median 17.50 12.00 5.50∗∗∗ 16.56 12.75 3.81∗∗∗
Std 158.07 133.53 206.51 165.76Weighted average N 6,804 15,829 19,916 27,935
life (in years) Mean 5.90 6.68 −0.78∗∗∗ 4.99 5.44 −0.45∗∗∗
Median 5.13 5.78 −0.65∗∗∗ 4.79 4.76 0.03∗
Std 3.53 3.97 2.6 3.28Fra. of Colla. in N 7,027 17,436 23,114 32,959
troubled states Mean 41.58 44.63 −3.05∗∗∗ 45.20 47.04 −1.84∗∗∗
Median 43.50 46.4 −2.90∗∗∗ 44.30 46.40 −2.10∗∗∗
Std 18.08 15.72 17.11 15.93Herfindahl index N 7,027 17,436 23,114 32,959
of collateral Mean 0.34 0.36 −0.02∗∗∗ 0.33 0.34 −0.01∗∗∗
Median 0.33 0.34 −0.01∗∗∗ 0.31 0.32 −0.01∗∗∗
Std 0.10 0.10 0.09 0.08Initial rating N 6,852 16,352 22,834 31,223
Mean 1.87 1.71 0.16∗∗∗ 2.24 2.04 0.20∗∗∗
Median 1.00 1.00 0.00∗∗∗ 1.67 1.00 0.67∗∗∗
Std 1.30 1.24 1.46 1.47Issuer rating N 5,797 18,605 17,766 35,030
Mean 2.75 2.85 −0.10∗∗∗ 2.74 3.02 −0.28∗∗∗
Median 2.50 2.89 −0.39∗∗∗ 2.44 2.67 −0.23∗∗∗
Std 1.02 0.67 0.95 0.96Fraction AAA N 771 1,376 1,544 1,857
Mean 90.25 92.11 −1.86∗∗∗ 85.67 88.13 −2.46∗∗∗
Median 94.19 96.52 −2.33∗∗∗ 88.41 93.92 −5.51∗∗∗
Std 11.21 11.97 12.26 14.50
(Continued)
2108 The Journal of Finance R©
Table I—Continued
Panel C: Distribution by Issuance Year and Issuer Size
2000–2003 2004–2006
Small Big Diff Small Big Diff
Number of N 849 1,523 1,595 1,943tranches in Mean 9.23 12.22 −2.99∗∗∗ 15.78 18.03 −2.25∗∗∗
a deal Median 7.00 9.00 −2.00∗∗∗ 15.00 16.00 −1.00∗∗∗
Std 8.93 10.96 7.85 10.74Level of N 7,260 17,587 24,687 34,121
subordination Mean 94.49 94.57 −0.08 91.61 91.99 −0.38∗∗∗
Median 97.62 97.51 0.11∗∗∗ 95.72 96.57 −0.85∗∗∗
Std 8.50 10.31 11.12 13.55
Num InitialRatings 2000–2003 2004–2006
Small issuer 1 16.70% 15.36%2 67.67% 62.37%3 15.63% 22.27%
Total 100.00% 100.00%
Big issuer 1 21.90% 21.78%2 71.16% 66.05%3 6.94% 12.18%
Total 100.00% 100.00%
SameOriginatorServicer 2000–2003 2004–2006
Small issuer 0 (“different”) 36.31% 56.57%1 (“same”) 63.69% 43.43%Total 100.00% 100.00%
Big issuer 0 (“different”) 17.54% 27.14%1 (“same”) 82.46% 72.86%Total 100.00% 100.00%
RatingDisagreement 2000–2003 2004–2006
Small issuer 0 (“same”) 94.90% 78.84%1 (“different”) 5.10% 21.16%Total 100.00% 100.00%
Big issuer 0 (“same”) 96.61% 86.27%1 (“different”) 3.39% 13.73%Total 100.00% 100.00%
(Continued)
Are All Ratings Created Equal? 2109
Table I—Continued
Panel C: Distribution by Issuance Year and Issuer Size
Rating Disagreement 0 (“same”) 1 (“different”) Total
Small 82.50% 17.50% 100.00%Big 89.82% 10.18% 100.00%
χ2 test of diff. 703.36 p-value = 0.00
Deals with Floating Tranches 2000–2003 2004–2006
Small issuer 0 (“no”) 29.68% 14.61%1 (“yes”) 70.32% 85.39%
Total 100.00% 100.00%
Big issuer 0 (“no”) 46.68% 22.13%1 (“yes”) 53.32% 77.87%
Total 100.00% 100.00%
an MBS tranche between issuance and April 2009 (or its payoff date).8 Thissample is considerably smaller than the yield sample because Bloombergonly provides pricing history for the larger deals.9 About 45% of the 9,299tranches for which we have information on pricing history are paid off early andbefore the crisis, so the median price drop is only 0.8% while the mean drop isabout 15%.
B.2. Issuer Characteristics
Our key explanatory variable of interest, Issuer Share, equals the number ofMBS deals sold by an issuer over the total number of deals sold by all issuersin the previous year (using alternative measures of issuer market share basedon the principal amounts generates very similar results). We denote marketboom years through the continuous variable HOT, which varies from 5% in2000 to its peak of 25% in 2006. We are interested in testing whether the effectof issuer size changes when markets boom, so we introduce the interactionvariable Issuer Share × HOT.
Since the value of implicit recourse to investors may increase with issuerreputation, we control for Issuer Rating, equal to the numerical score for therating of the issuer at the issuance date (AAA = 1; AA+ = 1.67, AA = 2,AA– = 2.33, and so on); the mean issuer rating is A. In our tests we also dif-ferentiate issuer types (Panel B, Table I), and include an indicator equal to
8Note that these bonds do not experience bankruptcy when the underlying assets become dis-tressed due to the legal status of the Special Purpose Vehicles. Instead, actual and expected futurecash flows fall, leading to a decline in the prices.
9Comparing the subsample of tranches with pricing information with the whole sample, wecan see that large tranches (principal amount) are more likely to have price information fromBloomberg, which reports prices as the midquote (bid–ask) from security dealers.
2110 The Journal of Finance R©
one for banks and thrifts, which face tighter regulatory capital requirementsthan other MBS issuers such as finance companies (e.g., GMAC) or investmentbanks (e.g., Bear Stearns, Lehman, etc.). If regulatory arbitrage encouragesregulated banks to securitize their assets more aggressively, then there may bedifferences in deal structure, collateral quality, and pricing. We also interactthe regulatory indicator with a time indicator equal to one after July 2004,when the regulators exempted banks and thrifts from FASB rule FIN46 byallowing them to move assets into securitized conduits financed with ABCP.This regulatory decision led to a doubling of this financing mechanism—an in-crease of about $600 billion in the outstanding amount—over just three years.We also construct Same Originator Servicer, an indicator set to one if the orig-inator and the servicer of the tranche are owned by the same firm and zerootherwise. Same Originator Servicer is only available for a subset of our data;hence we estimate our models with an additional indicator, Missing Origina-tor Servicer, equal to one if the information on originator and servicer is notavailable.
B.3. Deal Structure
Table I, Panel A, also reports summary statistics for Fraction AAA, equalto the total principal amount of all the AAA tranches in an MBS deal dividedby the total principal amount of all the rated tranches in the deal. Among the5,548 deals for which we have information on the principal amount of all thetranches, an average of 89% of the dollar value is rated AAA (median is 94%).Initial Rating equals a numerical score based on the average of the ratings atranche received at issuance. In the regressions, we estimate the AAA-ratedsample separately from the sample of non-AAA tranches, and in the lattersample control for the rating with separate indicators for each distinct cate-gory based on the average score across ratings. This non-parametric strategyallows us to avoid imposing any functional relationship between the rating andpricing. As our main measure of deal structure, we add Level of Subordination(Panel A) for each tranche, defined as the dollar-weighted fraction of tranchesin the same deal that have a rating the same as or better than the giventranche.10 For example, for a hypothetical $100 million deal with $80 millionin the AAA tranche, $10 million in the BBB tranche, and another $10 millionin the B tranche, Level of Subordination would equal 80% for AAA, 90% forBBB, and 100% for B. This variable increases as the amount of protection for agiven tranche by lower-rated tranches decreases; this variable equals FractionAAA for the AAA-rated tranches.
Opp, Opp, and Harris (2012) show theoretically and Furfine (2010) empiri-cally that more complex deals may lead to greater ratings inflation. To controlfor this mechanism, we add the log of the number of tranches within the deal.
10We are only able to observe tranches that receive ratings and are sold to investors. Thus, wecannot control for additional support provided by sponsors in unrated equity tranches, for example.
Are All Ratings Created Equal? 2111
We also control for deals with floating rate coupon tranches with an indicatorvariable. In addition, we control in some models for the number of ratings ona deal, using an indicator equal to one for deals with one rating and anotherequal to one for deals with two ratings. Issuers can pressure rating agencies bysoliciting a preliminary opinion before deciding whether to purchase a rating.Hence, they may drop lower ratings after shopping their product to an agency.Thus, deals with just one or two ratings are more likely to have been shoppedthan those with three. Some deals with two or three ratings may also havebeen shopped, forcing the ratings to converge, but we do observe some trancheswith multiple ratings where the agencies disagree. We control for this effectby adding another indicator for deals with more than one rating in which theratings differ.
B.4. Collateral
We include a number of control variables to capture characteristics of theunderlying collateral. From Panel A, Principal Amount equals the dollar valueof the tranche; its distribution is highly skewed, with a mean of $65 million andmedian of $14 million. Weighted-Average Life, equal to the expected timing ofpayments of principal of a tranche, is also skewed with the mean of 5.6 years.11
Fraction of Collateral in Troubled States equals the fraction of collateral orig-inated in Arizona, California, Florida, and Nevada. This variable measuresthe degree of exposure to areas that experienced the highest rise leading upto the crisis followed by the largest decline during the crisis.12 HerfindahlIndex of Collateral (HHI) measures geographical concentration of the collat-eral pool, equal to the sum of the squared shares of the collateral within adeal across each of the top five states (with the largest amount of mortgages),with the aggregation of all the other states as the sixth category. This vari-able controls, albeit crudely, for the degree of correlation across loans within agiven pool.
B.5. Sample Description
Table I, Panel B, describes the ratings distribution. Moody’s and S&P bothhave similar market presence, rating more than 51,000 tranches, while Fitchrates more than 34,000. The majority of tranches receive two (66%) or three(14%) ratings, while almost 20% of the tranches have only one rating. Amongtranches with two or three ratings, we observe disagreement about 13% ofthe time. For about 65% of the tranches, the same financial institution acts as
11Note that this is not the same as duration, which measures the weighted-average time tomaturity based on the relative present values of cash flows as weights (see, e.g., chapter 27 ofSaunders and Cornett (2007) for more details).
12We realize that the importance of this variable may be obvious only in hindsight, althoughsome analysts were concerned about overheated regional markets in real time; nevertheless, all ofour key findings are robust to the exclusion of this variable from our models.
2112 The Journal of Finance R©
both originator and servicer. Commercial banks are the most prevalent issuers,with about 39% of the deals, followed by investment banks (22%), thrifts (20%),finance companies (9%), and others (10%).
Panel C of Table I sorts the tranches into cohorts based on issuance year andissuer size. For these simple comparisons, “Big” refers to issuers with marketshare in the top 10% among all issuers (of a given year), and “Small” refersto all others. Not surprisingly, the volume of tranches, in terms of principalamount, is much greater during the housing market boom of 2004 to 2006.In our regressions below, we compare the characteristics of the two groups ofMBS tranches issued by large versus small issuers across this boom periodversus the earlier sample period (2000 to 2003) by interacting Market Share ofIssuers with the (continuous) variable HOT as defined above. We report resultsexcluding the tranches issued in 2007, as the housing and MBS markets clearlyentered into a new regime as compared to the previous boom period.13
From Panel C, tranches sold by small issuers appear to be larger in sizeand shorter in terms of weighted-average life, which tend to be safer, thanthose sold by large issuers. Tranches sold by small issuers also have less ex-posure to troubled states and are better diversified (lower Herfindahl index).The numerical values of ratings indicate that tranches sold by small issuersreceive worse ratings (e.g., Initial Rating has a higher mean and median) thanthose from large issuers, especially during the boom years of 2004 to 2006. Onthe other hand, small issuers themselves tend to have slightly better ratingsthan large issuers at the issuance date. MBS deals sold by large issuers alsohave less subordination—that it, a greater fraction of the deal receiving an AAArating—than those sold by small issuers. Further, MBS deals put together byboth small and large issuers have a significantly larger number of tranchesduring the boom period (more complexity), but deals from large issuers havemore tranches than those from small issuers during both periods.
Tranches from small issuers are less likely to have a single rating and morelikely to have ratings from all three agencies than tranches sold by large is-suers. Perhaps not surprisingly, there is more disagreement (defined only fortranches with multiple ratings) during the boom years, given the large volumeof risky deals sold in this period. But, as with levels of subordination, the gapin disagreement widens during the boom. During 2004 to 2006, for example,tranches sold by small issuers received different ratings 21% of the time, com-pared to just 14% of the time for large-issuer tranches. These comparisons sug-gest that large issuers shopped deals across agencies more aggressively than
13According to the financial crisis timeline of the Federal Reserve Bank in St. Louis, in Febru-ary 2007 Freddie Mac announced that it will no longer buy the most risky subprime mortgageand mortgage-related securities; in April 2007 New Century Financial Corp., a leading subprimemortgage lender, filed for Ch. 11 bankruptcy; in June 2007 S&P and Moody downgraded over 100bonds backed by second-lien subprime mortgages, and Bear Stearns informed investors that it wassuspending redemptions from its High-Grade Structured Credit Strategies Enhanced LeverageFund. All of these events suggest that the housing and MBS markets began to deteriorate in early2007. When we include the 2007 observations in pooled regressions, we obtain qualitatively similarresults.
Are All Ratings Created Equal? 2113
smaller issuers. Finally, large issuers are more likely to act not only as theoriginator of the deal, but also the servicer, who is responsible for collectinginterest payments after issuance. Small issuers, on the other hand, are morelikely to sell deals with different servicers from the originators. This differencemay in part reflect economies of scale at large mortgage banks such as Wash-ington Mutual (WaMu). However, servicers may be unwilling to accept theirrole for tranches with high default risks; thus, having a different servicer fromoriginator may provide a check and balance system when issuing the security.
Overall, these simple comparisons indicate that the quality of tranches is-sued by small issuers appears to be better than those sold by large issuers,despite receiving lower ratings on average. Moreover, large issuers seem toshop more for ratings—they are more likely to have one rating, and when theydo have multiple ratings these ratings are more likely to agree. This differenceis stronger during the boom years.
Table II reports the top 10 issuers in each year of our sample period. Theranking for an institution in a given year is based on the number of dealsissued during the year and information collected by SDC.14 While the list oftop 10 issuers changes over time, most if not all institutions on the lists arethe well-known, largest institutions involved in various aspects of housing andsubprime lending.15 Interestingly, the top six issuers in 2006, Countrywide, GM(through its finance arm GMAC), Bear Stearns, Lehman Brothers, IndyMac,and WaMu all failed during the ensuing crisis. Moreover, Citigroup, the ninthlargest issuer, received a large capital injection through the TARP program.The bottom row illustrates that the MBS market is highly concentrated amonglarge issuers, in that the top 10 issuers account for 55% to 68% of all thenewly issued securities each year over our sample period. As discussed above,the dominance of large issuers implies that they have considerable bargainingpower over rating agencies.
III. Regressions Results
Tables III and IV report the main results on initial yields, estimated atthe tranche level. In Table III, we regress the yield spread at issuance oncharacteristics of the deals and tranches, the issuer, and the market. Table IVreports similar regressions using samples split by issuer size to test whetherdeal characteristics are priced differently across issuer types. We then examineprice changes in Table V (Equation 2), also estimated at the tranche level, andwe introduce Initial Yield Spread as a regressor.
14Note that in Table II issuer rankings and market share are based on the number of deals (notweighted by deal size) sold in the current year, whereas in regression models below we use laggedmarket share (from the previous year).
15We also rank bookrunners, or lead underwriters of the MBS securities, in each year. This listreflects the largest underwriters of structured finance products during this period, and overlapswith the list of largest issuers. We find (not reported) that the impact of ratings on the performanceof tranches mostly comes through large issuers, not bookrunners.
2114 The Journal of Finance R©
Tab
leII
Top
10Is
suer
sof
Pri
vate
lyS
old
MB
S(b
yN
um
ber
ofD
eals
)T
his
tabl
esh
ows
the
top
10is
suer
sof
priv
atel
yso
ldM
BS
sort
edby
the
nu
mbe
rof
deal
sin
each
year
ofth
esa
mpl
epe
riod
.Th
ela
stro
wsh
ows
the
tota
lmar
ket
shar
eof
the
top
10is
suer
sin
term
sof
the
nu
mbe
rof
deal
sth
eyis
sue
ina
give
nye
ar.
Ran
k20
0020
0120
0220
0320
0420
0520
06
1G
ener
alM
otor
sG
ener
alM
otor
sG
ener
alM
otor
sG
ener
alM
otor
sC
oun
tryw
ide
Cou
ntr
ywid
eC
oun
tryw
ide
2C
oun
tryw
ide
Cre
dit
Su
isse
Cou
ntr
ywid
eC
oun
tryw
ide
Gen
eral
Mot
ors
Gen
eral
Mot
ors
Gen
eral
Mot
ors
3W
ells
Far
goC
oun
tryw
ide
WaM
uL
ehm
anB
ear
Ste
arn
sL
ehm
anB
ear
Ste
arn
s4
JPM
orga
nC
has
eW
ells
Far
goC
redi
tS
uis
seC
redi
tS
uis
seL
ehm
anB
ear
Ste
arn
sL
ehm
an5
Ban
kof
Am
eric
aL
ehm
anL
ehm
anB
ear
Ste
arn
sB
ank
ofA
mer
ica
Ban
kof
Am
eric
aIn
dyM
ac6
Cit
igro
up
Cit
igro
up
Ban
kof
Am
eric
aB
ank
ofA
mer
ica
Cre
dit
Su
isse
Indy
Mac
WaM
u7
Gre
enP
oin
tB
ank
ofA
mer
ica
Bea
rS
tear
ns
WaM
uU
BS
Cre
dit
Su
isse
Gol
dman
Sac
hs
8L
ehm
anB
ear
Ste
arn
sW
ells
Far
goW
ells
Far
goM
orga
nS
tan
ley
Gol
dman
Sac
hs
JPM
orga
nC
has
e9
GE
JPM
orga
nC
has
eM
orga
nS
tan
ley
UB
SW
ells
Far
goM
erri
llL
ynch
Cit
igro
up
10C
onse
coW
aMu
UB
SJP
Mor
gan
Ch
ase
Gol
dman
Sac
hs
Mor
gan
Sta
nle
yC
redi
tS
uis
seM
arke
tsh
are
55.4
0%68
.12%
66.6
2%64
.25%
60.0
7%59
.17%
59.4
5%
Are All Ratings Created Equal? 2115
Tab
leII
IR
egre
ssio
nof
MB
SY
ield
Sp
read
toIs
suer
Sh
are
Th
ista
ble
repo
rts
OL
Sre
gres
sion
sof
the
Log
ofY
ield
Spr
ead
ofpr
ivat
ely
issu
edM
BS
tran
ches
onis
suer
mar
ket
shar
e,tr
anch
e-le
velc
har
acte
rist
ics,
oth
eris
suer
char
acte
rist
ics,
and
mar
ket
con
diti
ons.
Th
esa
mpl
ein
clu
des
all
tran
ches
for
wh
ich
we
can
obse
rve
init
ial
cou
pon
s(o
rfi
xed
spre
ads)
onB
loom
berg
orig
inat
edbe
twee
n20
00an
d20
06th
atre
ceiv
eda
rati
ng
from
Moo
dy’s
,S
&P,
orF
itch
.V
aria
bles
are
defi
ned
inS
ecti
onII
.Bof
the
text
.In
itia
lR
atin
gC
ateg
ory
Du
mm
ies
inP
anel
sC
and
Dar
ea
set
ofdu
mm
ies
toin
dica
teea
chle
vel
ofth
eav
erag
era
tin
gsa
give
ntr
anch
ere
ceiv
edat
issu
ance
,af
ter
we
con
vert
the
rati
ngs
into
an
um
eric
alva
lue
byse
ttin
gA
AA
=1,
AA
+=
1.67
,A
A=
2,A
A–
=2.
33,
and
soon
,an
dth
enta
keth
ear
ith
met
icav
erag
esof
the
rati
ngs
the
tran
che
rece
ives
.E
ach
regr
essi
onin
clu
des
sepa
rate
inte
rcep
tsfo
rco
upo
nty
pes
(floa
tin
g,fi
xed,
orva
riab
le).
Sta
nda
rder
rors
are
clu
ster
edby
both
coh
ort-
year
and
issu
er(t
wo-
way
clu
ster
ing)
.T-s
tati
stic
sar
ein
pare
nth
eses
.Pan
els
Aan
dB
pres
ent
resu
lts
for
AA
Atr
anch
eson
ly;P
anel
sC
and
Dpr
esen
tre
sult
sfo
rn
on-A
AA
tran
ches
only
.Pan
els
Aan
dC
don
oth
ave
issu
erfi
xed
effe
cts,
and
Pan
els
Ban
dD
hav
eis
suer
fixe
def
fect
s.∗∗
∗ ,∗∗
,an
d∗
indi
cate
sign
ifica
nce
atth
e1%
,5%
,an
d10
%le
vels
,res
pect
ivel
y.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
:AA
ATr
anch
esO
nly
,No
Issu
erF
ixed
Eff
ects
Issu
erS
har
e0.
2891
−1.2
301∗∗
0.22
47−1
.207
3∗∗0.
1904
−1.1
930∗∗
0.20
60−1
.120
7∗∗
(0.5
3)(−
2.35
)(0
.43)
(−2.
21)
(0.3
6)(−
2.24
)(0
.40)
(−2.
12)
HO
T∗
Issu
erS
har
e–
9.89
87∗∗
∗–
9.33
10∗∗
∗–
9.02
03∗∗
∗–
8.64
05∗∗
∗
–(3
.20)
–(2
.97)
–(2
.87)
–(2
.81)
Ban
kan
dT
hri
ft−0
.016
3−0
.040
7−0
.042
9−0
.065
7−0
.045
7−0
.067
7−0
.046
2−0
.067
2(−
0.35
)(−
0.91
)(−
1.09
)(−
1.46
)(−
1.18
)(−
1.54
)(−
1.17
)(−
1.51
)B
ank
and
Th
rift
∗P
ost
July
040.
0962
∗∗0.
1116
∗∗∗
0.14
17∗∗
∗0.
1559
∗∗∗
0.14
31∗∗
∗0.
1569
∗∗∗
0.14
46∗∗
∗0.
1576
∗∗∗
(2.2
0)(2
.61)
(3.2
7)(3
.50)
(3.2
5)(3
.42)
(3.1
6)(3
.37)
Lev
elof
Su
bord
inat
ion
––
0.67
36∗∗
∗0.
6718
∗∗∗
0.66
45∗∗
∗0.
6629
∗∗∗
0.24
170.
2825
––
(4.7
9)(4
.71)
(4.3
1)(4
.26)
(0.9
5)(1
.04)
HO
T∗
Lev
elof
Su
bord
inat
ion
––
––
––
2.35
592.
1199
––
––
––
(1.4
7)(1
.31)
Log
ofP
rin
cipa
l−0
.020
8∗∗−0
.021
0∗∗−0
.004
1−0
.004
4−0
.002
8−0
.003
2−0
.002
2−0
.002
6(−
2.20
)(−
2.48
)(−
0.46
)(−
0.49
)(−
0.32
)(−
0.35
)(−
0.24
)(−
0.28
)L
ogof
Wei
ghte
dA
vera
geL
ife
0.00
820.
0081
0.01
060.
0105
0.01
080.
0106
0.01
100.
0108
(0.0
8)(0
.08)
(0.1
1)(0
.11)
(0.1
1)(0
.11)
(0.1
1)(0
.11)
Log
ofN
um
ber
ofTr
anch
es–
–0.
1499
∗∗∗
0.14
91∗∗
∗0.
1515
∗∗∗
0.15
05∗∗
∗0.
1523
∗∗∗
0.15
13∗∗
∗
––
(5.3
2)(4
.88)
(5.1
2)(4
.75)
(5.1
4)(4
.80)
(Con
tin
ued
)
2116 The Journal of Finance R©
Tab
leII
I—C
onti
nu
ed
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elA
:AA
ATr
anch
esO
nly
,No
Issu
erF
ixed
Eff
ects
Fra
.of
Col
la.i
nTr
oubl
edS
tate
s0.
0043
∗∗0.
0042
∗∗0.
0031
∗∗0.
0031
∗∗0.
0029
∗∗0.
0029
∗0.
0029
∗∗0.
0029
∗
(2.5
7)(2
.56)
(2.2
9)(2
.16)
(2.0
3)(1
.94)
(2.0
5)(1
.95)
Her
fin
dah
lIn
dex
ofC
olla
tera
l−0
.324
1−0
.327
7−0
.305
6∗−0
.309
1∗−0
.284
1∗−0
.288
5∗−0
.287
0∗−0
.290
9∗
(−1.
61)
(−1.
60)
(−1.
82)
(−1.
78)
(−1.
71)
(−1.
67)
(−1.
74)
(−1.
71)
Sam
eO
rigi
nat
orS
ervi
cer
0.09
21∗∗
∗0.
0802
∗∗∗
0.07
17∗∗
∗0.
0605
∗∗∗
0.07
12∗∗
∗0.
0605
∗∗∗
0.06
79∗∗
∗0.
0580
∗∗∗
(3.3
3)(2
.83)
(3.1
9)(3
.06)
(3.0
2)(2
.97)
(2.9
2)(2
.80)
Mis
sin
gO
rigi
nat
orS
ervi
cer
0.05
310.
0500
0.03
000.
0272
0.02
700.
0245
0.02
650.
0241
(1.5
6)(1
.39)
(1.0
9)(0
.98)
(0.9
4)(0
.87)
(0.9
8)(0
.89)
Issu
erR
atin
g0.
0099
0.00
500.
0201
0.01
530.
0218
0.01
720.
0217
0.01
73(0
.53)
(0.2
6)(1
.08)
(0.8
2)(1
.17)
(0.9
0)(1
.17)
(0.9
2)O
ne
Init
ialR
atin
g–
––
–0.
0945
0.08
910.
0940
0.08
89–
––
–(1
.57)
(1.4
7)(1
.58)
(1.5
2)Tw
oIn
itia
lRat
ings
––
––
0.07
750.
0749
0.07
700.
0745
––
––
(1.0
6)(1
.03)
(1.0
5)(1
.03)
Coh
ort-
Year
Fix
edE
ffec
tsYe
sYe
sYe
sYe
sYe
sYe
sYe
sYe
sO
bser
vati
ons
25,1
2925
,129
25,1
2925
,129
25,1
2925
,129
25,1
2925
,129
R2
0.71
70.
718
0.72
90.
729
0.72
90.
730
0.73
00.
730
Join
tW
ald
test
sof
“On
eIn
itia
l3.
433.
804.
594.
52R
atin
gs”
and
“Tw
oIn
itia
l(0
.18)
(0.1
5)(0
.10)
(0.1
0)R
atin
gs”
(p-v
alu
e)
Are All Ratings Created Equal? 2117
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
:AA
ATr
anch
esO
nly
,wit
hIs
suer
Fix
edE
ffec
ts
Issu
erS
har
e0.
6988
0.38
320.
3982
0.01
670.
4240
0.07
370.
3096
0.04
96(1
.19)
(0.7
4)(0
.91)
(0.0
5)(1
.00)
(0.2
1)(0
.72)
(0.1
3)H
OT
∗Is
suer
Sh
are
–2.
6230
∗∗∗
–3.
1694
∗∗–
2.90
37∗
–2.
1836
–(2
.72)
–(2
.04)
–(1
.71)
–(1
.18)
Ban
kan
dT
hri
ft∗
Pos
tJu
ly04
0.13
97∗∗
∗0.
1411
∗∗0.
1717
∗∗∗
0.17
35∗∗
0.17
30∗∗
∗0.
1747
∗∗0.
1760
∗∗∗
0.17
72∗∗
(2.7
5)(2
.53)
(3.0
2)(2
.45)
(3.0
1)(2
.39)
(3.0
1)(2
.35)
Lev
elof
Su
bord
inat
ion
––
0.64
88∗∗
∗0.
6498
∗∗∗
0.64
63∗∗
∗0.
6472
∗∗∗
0.21
500.
2285
––
(3.9
4)(3
.85)
(3.7
2)(3
.56)
(0.7
3)(0
.72)
HO
T∗
Lev
elof
Su
bord
inat
ion
––
––
––
2.39
512.
3243
––
––
––
(1.3
2)(1
.26)
Log
ofP
rin
cipa
l−0
.018
8∗∗−0
.018
8∗∗−0
.003
6−0
.003
5−0
.002
6−0
.002
6−0
.002
1−0
.002
1(−
2.34
)(−
2.14
)(−
0.46
)(−
0.44
)(−
0.33
)(−
0.32
)(−
0.26
)(−
0.25
)L
ogof
Wei
ghte
dA
vera
geL
ife
0.01
040.
0104
0.01
200.
0119
0.01
230.
0122
0.01
240.
0123
(0.1
0)(0
.09)
(0.1
2)(0
.11)
(0.1
2)(0
.12)
(0.1
2)(0
.12)
Log
ofN
um
ber
ofTr
anch
es–
–0.
1397
∗∗∗
0.13
97∗∗
∗0.
1406
∗∗∗
0.14
06∗∗
∗0.
1418
∗∗∗
0.14
18∗∗
∗
––
(5.6
4)(4
.93)
(5.4
2)(4
.95)
(5.5
6)(5
.01)
Fra
.of
Col
la.i
nTr
oubl
edS
tate
s0.
0028
0.00
280.
0021
0.00
220.
0020
0.00
200.
0020
0.00
20(1
.63)
(1.6
3)(1
.37)
(1.3
8)(1
.24)
(1.2
6)(1
.26)
(1.2
8)H
erfi
nda
hlI
nde
xof
Col
late
ral
−0.1
974
−0.2
012
−0.1
935
−0.1
982
−0.1
767
−0.1
813
−0.1
802
−0.1
836
(−0.
92)
(−0.
94)
(−1.
03)
(−0.
98)
(−0.
93)
( −0.
90)
(−0.
97)
(−0.
94)
Sam
eO
rigi
nat
orS
ervi
cer
0.10
99∗∗
∗0.
1088
∗∗∗
0.09
12∗∗
∗0.
0899
∗∗∗
0.09
07∗∗
∗0.
0895
∗∗∗
0.08
75∗∗
∗0.
0867
∗∗∗
(5.3
5)(3
.31)
(6.2
9)(5
.13)
(6.4
2)(5
.17)
(5.4
3)(4
.76)
Mis
sin
gO
rigi
nat
orS
ervi
cer
0.06
54∗
0.06
53∗
0.04
320.
0431
0.04
110.
0411
0.04
030.
0403
(1.8
6)(1
.79)
(1.4
6)(1
.45)
(1.3
6)(1
.33)
(1.3
8)(1
.40)
Issu
erR
atin
g0.
0774
∗∗∗
0.07
40∗∗
∗0.
0691
∗∗∗
0.06
49∗∗
∗0.
0731
∗∗∗
0.06
92∗∗
∗0.
0697
∗∗∗
0.06
68∗∗
∗
(3.3
0)(2
.76)
(2.9
5)(2
.60)
(3.0
2)(2
.72)
(2.8
4)(2
.61)
On
eIn
itia
lRat
ing
––
––
0.07
200.
0707
0.07
110.
0701
––
––
(1.0
0)(0
.98)
(1.0
0)(0
.99)
(Con
tin
ued
)
2118 The Journal of Finance R©
Tab
leII
I—C
onti
nu
ed
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elB
:AA
ATr
anch
esO
nly
,wit
hIs
suer
Fix
edE
ffec
ts
Two
Init
ialR
atin
gs–
––
–0.
0562
0.05
560.
0551
0.05
47–
––
–(0
.70)
(0.6
9)(0
.69)
(0.6
8)C
ohor
t-Ye
arF
ixed
Eff
ects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Issu
erF
ixed
Eff
ects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Obs
erva
tion
s25
,129
25,1
2925
,129
25,1
2925
,129
25,1
2925
,129
25,1
29R
20.
725
0.72
50.
735
0.73
50.
735
0.73
60.
736
0.73
6
Join
tW
ald
test
sof
“On
eIn
itia
l1.
761.
641.
711.
79R
atin
gs”
and
“Tw
oIn
itia
l(0
.42)
(0.4
4)(0
.43)
(0.4
1)R
atin
gs”
(p-v
alu
e)
Pan
elC
:Non
-AA
ATr
anch
esO
nly
,No
Issu
erF
ixed
Eff
ects
Issu
erS
har
e0.
3102
−1.0
447∗∗
∗0.
3885
−1.0
788∗∗
∗0.
3516
−1.1
417∗∗
∗0.
3649
−1.1
266∗∗
∗
(1.0
0)(−
2.58
)(1
.22)
(−2.
58)
(1.0
8)(−
2.77
)(1
.15)
(−2.
81)
HO
T∗
Issu
erS
har
e–
7.91
47∗∗
∗–
8.56
19∗∗
∗–
8.74
21∗∗
∗–
8.73
11∗∗
∗
–(4
.51)
–(4
.43)
–(4
.92)
–(4
.95)
Ban
kan
dT
hri
ft−0
.117
9∗∗−0
.129
9∗∗−0
.119
6∗∗−0
.132
5∗∗−0
.122
6∗∗−0
.135
6∗∗−0
.123
8∗∗−0
.136
8∗∗
(−1.
98)
(−1.
98)
(−2.
09)
(−2.
43)
(−2.
17)
(−2.
38)
(−2.
18)
(−2.
23)
Ban
kan
dT
hri
ft∗
Pos
tJu
ly04
0.16
69∗∗
∗0.
1759
∗∗∗
0.18
26∗∗
∗0.
1925
∗∗∗
0.18
14∗∗
∗0.
1910
∗∗∗
0.18
40∗∗
∗0.
1936
∗∗∗
(2.8
3)(2
.91)
(3.1
6)(3
.73)
(3.2
7)(3
.83)
(3.3
4)(3
.81)
Lev
elof
Su
bord
inat
ion
––
0.29
21∗∗
∗0.
2979
∗∗∗
0.27
64∗∗
∗0.
2820
∗∗∗
−0.1
496
−0.1
427
––
(7.2
1)(7
.27)
(7.3
4)(7
.68)
(−0.
80)
(−0.
75)
HO
T∗
Lev
elof
Su
bord
inat
ion
––
––
––
2.23
27∗
2.22
62∗
––
––
––
(1.7
5)(1
.72)
Log
ofP
rin
cipa
l0.
0335
∗0.
0334
0.03
51∗
0.03
490.
0424
∗∗0.
0422
∗∗0.
0427
∗∗0.
0425
∗∗
(1.6
6)(1
.56)
(1.7
1)(1
.56)
(2.4
2)(2
.08)
(2.2
4)(2
.22)
Log
ofW
eigh
ted
Ave
rage
Lif
e−0
.082
1−0
.083
5−0
.092
6−0
.094
2∗−0
.087
5−0
.089
9−0
.089
2−0
.091
6(−
1.42
)(−
1.42
)(−
1.64
)(−
1.68
)(−
1.58
)(−
1.60
)(−
1.58
)(−
1.53
)L
ogof
Nu
mbe
rof
Tran
ches
––
0.02
620.
0268
0.01
930.
0198
0.02
080.
0213
––
(1.0
8)(1
.12)
(0.8
7)(0
.92)
(0.9
5)(0
.97)
Are All Ratings Created Equal? 2119
Pan
elC
:Non
-AA
ATr
anch
esO
nly
,No
Issu
erF
ixed
Eff
ects
Fra
.of
Col
la.i
nTr
oubl
edS
tate
s0.
0010
0.00
100.
0008
0.00
080.
0007
0.00
070.
0007
0.00
07(0
.84)
(0.8
4)(0
.67)
(0.6
0)(0
.53)
(0.5
4)(0
.55)
(0.5
3)H
erfi
nda
hlI
nde
xof
Col
late
ral
0.01
980.
0078
0.00
04−0
.012
8−0
.001
3−0
.015
5−0
.002
3−0
.016
5(0
.19)
(0.0
8)(0
.00)
(−0.
12)
(−0.
01)
(−0.
15)
(−0.
02)
(−0.
16)
Sam
eO
rigi
nat
orS
ervi
cer
−0.0
347
−0.0
422∗
−0.0
330
−0.0
410∗∗
−0.0
308
−0.0
392∗
−0.0
314
−0.0
397∗
(−1.
50)
(−1.
86)
(−1.
53)
(−2.
02)
(−1.
41)
(−1.
85)
(−1.
42)
(−1.
83)
Mis
sin
gO
rigi
nat
orS
ervi
cer
−0.0
453∗∗
−0.0
474∗∗
−0.0
423∗∗
∗−0
.044
4∗∗∗
−0.0
392∗∗
∗−0
.041
4∗∗−0
.039
5∗∗∗
−0.0
417∗∗
(−2.
33)
(−2.
34)
(−2.
74)
(−2.
61)
(−2.
60)
(−2.
47)
(−2.
64)
(−2.
54)
Issu
erR
atin
g0.
0194
0.01
750.
0187
0.01
670.
0172
0.01
520.
0174
0.01
53(1
.50)
(1.2
6)(1
.56)
(1.3
7)(1
.54)
(1.3
0)(1
.54)
(1.3
5)O
ne
Init
ialR
atin
g–
––
–0.
0669
0.06
760.
0666
0.06
73–
––
–(1
.46)
(1.5
0)(1
.44)
(1.4
9)Tw
oIn
itia
lRat
ings
––
––
0.04
80∗∗
∗0.
0438
∗∗0.
0483
∗∗∗
0.04
41∗∗
––
––
(2.7
9)(2
.17)
(2.6
0)(2
.43)
Rat
ing
Dis
agre
emen
t–
––
–0.
1055
∗∗∗
0.10
77∗∗
∗0.
1051
∗∗∗
0.10
73∗∗
∗
––
––
(4.4
0)(4
.57)
(4.3
6)(4
.46)
Coh
ort-
Year
Fix
edE
ffec
tsYe
sYe
sYe
sYe
sYe
sYe
sYe
sYe
sIn
itia
lRat
ing
Cat
egor
yD
um
mie
sYe
sYe
sYe
sYe
sYe
sYe
sYe
sYe
sO
bser
vati
ons
19,1
7919
,179
19,1
3319
,133
19,1
3319
,133
19,1
3319
,133
R2
0.60
20.
603
0.60
50.
606
0.60
90.
610
0.61
00.
611
Join
tW
ald
test
sof
“On
eIn
itia
l8.
30∗∗
∗6.
73∗∗
∗8.
00∗∗
∗7.
56∗∗
∗
Rat
ings
”an
d“T
wo
Init
ial
(0.0
2)(0
.03)
(0.0
2)(0
.02)
Rat
ings
”(p
-val
ue)
Pan
elD
:Non
-AA
ATr
anch
esO
nly
,wit
hIs
suer
Fix
edE
ffec
ts
Issu
erS
har
e0.
9659
0.26
880.
8724
0.13
970.
8600
0.05
630.
7970
0.02
62(1
.64)
(0.5
3)(1
.51)
(0.3
0)(1
.55)
(0.1
4)(1
.46)
(0.0
7)H
OT
∗Is
suer
Sh
are
–5.
9191
∗∗∗
–6.
2108
∗∗∗
–6.
8432
∗∗∗
–6.
5864
∗∗∗
–(2
.83)
–(3
.39)
–(4
.20)
–(3
.71)
Ban
kan
dT
hri
ft∗
Pos
tJu
ly04
0.17
17∗∗
∗0.
1732
∗∗∗
0.18
84∗∗
∗0.
1901
∗∗∗
0.18
80∗∗
∗0.
1897
∗∗∗
0.19
13∗∗
∗0.
1928
∗∗∗
(2.9
0)(3
.11)
(3.3
2)(3
.61)
(3.4
3)(3
.82)
(3.4
4)(3
.86)
(Con
tin
ued
)
2120 The Journal of Finance R©T
able
III—
Con
tin
ued
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pan
elD
:Non
-AA
ATr
anch
esO
nly
,wit
hIs
suer
Fix
edE
ffec
ts
Lev
elof
Su
bord
inat
ion
––
0.30
56∗∗
∗0.
3066
∗∗∗
0.29
26∗∗
∗0.
2937
∗∗∗
−0.0
995
−0.0
820
––
(4.2
6)(4
.17)
(4.1
7)(4
.22)
(−0.
53)
(−0.
41)
HO
T∗
Lev
elof
Su
bord
inat
ion
––
––
––
2.06
051.
9734
––
––
––
(1.5
6)(1
.51)
Log
ofP
rin
cipa
l0.
0326
∗0.
0328
∗0.
0331
∗0.
0333
∗0.
0409
∗∗0.
0412
∗∗0.
0412
∗∗0.
0414
∗∗
(1.6
7)(1
.73)
(1.7
5)(1
.68)
(2.3
6)(2
.24)
(2.3
3)(2
.01)
Log
ofW
eigh
ted
Ave
rage
Lif
e−0
.086
7−0
.086
8−0
.094
4∗−0
.094
6−0
.093
3∗−0
.093
9∗−0
.095
4∗−0
.095
9∗
(−1.
57)
(−1.
56)
(−1.
68)
(−1.
64)
(−1.
71)
(−1.
70)
(−1.
76)
(−1.
79)
Log
ofN
um
ber
ofTr
anch
es0.
0121
0.01
250.
0024
0.00
260.
0039
0.00
41(0
.44)
(0.4
2)(0
.09)
(0.1
0)(0
.15)
(0.1
3)F
ra.o
fC
olla
.in
Trou
bled
Sta
tes
0.00
020.
0002
0.00
000.
0001
−0.0
002
−0.0
001
−0.0
002
−0.0
001
(0.1
3)(0
.17)
(0.0
3)(0
.07)
(−0.
13)
(−0.
08)
( −0.
13)
(−0.
08)
Her
fin
dah
lIn
dex
ofC
olla
tera
l0.
0977
0.08
310.
0841
0.06
880.
0876
0.07
060.
0857
0.06
95(0
.91)
(0.7
4)(0
.77)
(0.6
1)(0
.84)
(0.6
7)(0
.82)
(0.6
0)S
ame
Ori
gin
ator
Ser
vice
r0.
0001
−0.0
020
0.00
19−0
.000
2−0
.000
8−0
.003
2−0
.001
2−0
.003
4(0
.01)
(−0.
16)
(0.2
1)(−
0.02
)(−
0.08
)(−
0.29
)(−
0.11
)(−
0.32
)M
issi
ng
Ori
gin
ator
Ser
vice
r−0
.023
8−0
.023
5−0
.020
9−0
.020
6−0
.022
0−0
.021
7−0
.022
5−0
.022
2(−
1.42
)(−
1.36
)(−
1.41
)(−
1.13
)(−
1.37
)(−
1.15
)(−
1.40
)(−
1.22
)Is
suer
Rat
ing
0.06
28∗∗
∗0.
0586
∗∗∗
0.05
93∗∗
∗0.
0549
∗∗∗
0.05
78∗∗
∗0.
0530
∗∗∗
0.05
64∗∗
∗0.
0519
∗∗∗
(3.2
8)(3
.49)
(3.0
5)(3
.37)
(2.7
5)(3
.00)
(2.6
4)(2
.85)
On
eIn
itia
lRat
ing
––
––
0.08
95∗∗
0.09
03∗∗
0.08
88∗∗
0.08
96∗∗
––
––
(2.1
8)(2
.24)
(2.0
7)(2
.03)
Two
Init
ialR
atin
gs–
––
–0.
0415
∗∗0.
0397
∗0.
0415
∗∗0.
0397
∗
––
––
(2.0
4)(1
.94)
(2.0
5)(1
.90)
Rat
ing
Dis
agre
emen
t–
––
–0.
1104
∗∗∗
0.11
18∗∗
∗0.
1098
∗∗∗
0.11
11∗∗
∗
––
––
(4.2
0)(4
.35)
(4.1
2)(4
.22)
Coh
ort-
Year
Fix
edE
ffec
tsYe
sYe
sYe
sYe
sYe
sYe
sYe
sYe
sIs
suer
Fix
edE
ffec
tsYe
sYe
sYe
sYe
sYe
sYe
sYe
sYe
sIn
itia
lRat
ing
Cat
egor
yD
um
mie
sYe
sYe
sYe
sYe
sYe
sYe
sYe
sYe
sO
bser
vati
ons
19,1
7919
,179
19,1
3319
,133
19,1
3319
,133
19,1
3319
,133
R2
0.61
30.
613
0.61
60.
616
0.62
00.
621
0.62
10.
621
Join
tW
ald
test
sof
“On
eIn
itia
l7.
89∗∗
7.55
∗∗7.
70∗∗
7.26
∗∗
Rat
ings
”an
d“T
wo
Init
ial
(0.0
2)(0
.02)
(0.0
2)(0
.03)
Rat
ings
”(p
-val
ue)
Are All Ratings Created Equal? 2121
Table IVRegression of MBS Yield Spread to Tranche Characteristics on
Samples Split by Issuer SizeThis table reports OLS regressions of the Log of Yield Spread on tranche-level characteristics,various issuer characteristics, and market conditions for samples split by issuer market share.“Big issuer” means that the market share of the issuer falls into the top 10% of the market sharedistribution in a given year, and “Small issuers” refers to the rest of issuers in that year. Each re-gression includes separate intercepts for coupon types (floating, fixed, or variable). Standard errorsare clustered by both cohort-year and issuer (two-way clustering). T-statistics are in parentheses.Panel A presents regression results for different subsamples and Panel B compares the averageyield spread for tranches actually sold by big issuers with the predicted yield spread for those sametranches using the small-issuer coefficients. ∗∗∗, ∗∗, and ∗ indicate significance at the 1%, 5%, and10% levels, respectively.
Panel A: Regression Coefficients
AAA Tranches Non-AAA Tranches
Big issuers Small issuers Big issuers Small issuers(1) (2) (3) (4)
Bank and Thrift 0.0114 −0.2011∗∗∗ −0.0580 −0.2323∗∗∗(0.45) (−3.92) (−0.93) (−3.83)
Bank and Thrift ∗ Post July 04 0.0806∗ 0.3237∗∗∗ 0.1217∗ 0.3007∗∗∗(1.68) (6.58) (1.65) (6.14)
Level of Subordination 0.7159∗∗∗ 0.6100∗∗∗ 0.2957∗∗∗ 0.2210∗∗∗(3.54) (6.88) (3.88) (3.97)
Log of Principal −0.0022 −0.0065 0.0465∗∗ 0.0348(−0.23) (−0.49) (2.25) (1.64)
Log of Weighted Average Life −0.0082 0.0388 −0.0876 −0.0907(−0.08) (0.36) (−1.58) (−1.21)
Log of Number of Tranches 0.1321∗∗∗ 0.1718∗∗∗ −0.0091 0.0573(5.12) (4.69) (−0.35) (1.51)
Fra. of Colla. in Troubled States 0.0042∗ 0.0009 0.0009 −0.0010(1.89) (0.74) (0.71) (−0.56)
Herfindahl Index of Collateral −0.3479 −0.1239 −0.0270 0.1052(−1.43) (−0.97) (−0.17) (0.68)
Same Originator Servicer 0.0859∗∗∗ 0.0008 −0.0122 −0.0979∗∗(2.93) (0.02) (−0.52) (−2.53)
Missing Originator Servicer 0.0372 0.0141 −0.0420 −0.0222(0.82) (0.66) (−1.53) (−1.00)
Issuer Rating 0.0224 −0.0022 0.0008 0.0239(1.13) (−0.12) (0.05) (1.41)
One Initial Rating 0.0331 0.1876∗∗∗ 0.0964 0.0912(0.44) (3.04) (1.64) (1.63)
Two Initial Ratings 0.0267 0.1515∗∗∗ 0.0662∗ 0.0431∗∗(0.34) (2.71) (1.80) (2.20)
Ratings disagreement – – 0.0979∗∗∗ 0.1190∗∗∗– – (3.18) (5.66)
Cohort-Year Fixed Effects Yes Yes Yes YesInitial Rating Dummies – – Yes YesObservations 15,622 9,507 11,380 7,753R2 0.735 0.727 0.602 0.633
(Continued)
2122 The Journal of Finance R©
Table IV—Continued
Panel B: Actual and Predicted Mean Basis-Point Spreads for Large-Issuer Tranches (PredictedValues Based on Small-Issuer Coefficients)
AAA Tranches Non-AAA Tranches
Actual Predicted Actual Predicted
Year (1) (2) (3) (4)2000–2003 176 171 182 1762004–2006 98 83 115 103
A. Yield Spread at Issuance
Does the market price the risk of agency problems—the risk of large-issuerdeals? If larger issuers exert greater bargaining power, yield spreads shouldrise with issuer size conditional on the credit rating. Since the credit ratingideally acts as a sufficient statistic for risk (absent agency problems), it is lessimportant to condition on the full set of collateral characteristics in this set-ting compared to modeling the fraction rated AAA, as in other studies (e.g.,Ashcraft, Goldsmith-Pinkham, and Vickery (2010), Griffin and Tang (2012)).Thus, we compare how initial yields vary with issuer size controlling for thedistribution of ratings (ratings indicators, the number of ratings, and a dis-agreement indicator). Since most of the securities are priced and sold at par,initial yield spreads gauge the market’s assessment of ex ante credit quality(i.e., risk).
Figure 1 presents suggestive evidence by plotting initial yield spreads fortranches sold by large versus small issuers. “Big issuer” indicates that themarket share falls into the top 10% of the market share distribution in agiven year, while “Small issuer” refers to the other issuers in the same year.As mentioned earlier, for a tranche with a floating coupon rate, yield spreadis the fixed markup (in bps) over the benchmark rate; for a tranche with afixed or variable coupon rate, yield spread is the difference between the initialcoupon rate and the yield of a Treasury security whose maturity is closest tothe tranche’s weighted average life. Tranches are sorted by their issuance year(cohort), and we plot the median initial yield spread for each cohort of the twogroups of tranches during 2000 to 2006. Figure 1 shows that yields on tranchessold by large issuers consistently exceed yields from small issuers, with theaverage difference about 18 bps. The gap in the yield spreads is the largestduring the market boom period of 2004 to 2006, with the difference in 2004over 37 bps.
Table III tests whether the patterns in Figure 1 hold up after controlling forthe initial rating of the tranche using a full set of indicators for each uniquevalue of the average rating. Columns 1 and 2 control for collateral and is-suer characteristics (reduced forms), and in the subsequent columns we addvariables related to deal structure: Level of Subordination, One Initial Ratingand Two Initial Rating indicators (to test for shopping), Rating Disagreement
Are All Ratings Created Equal? 2123T
able
VR
egre
ssio
nof
MB
SP
rice
Ch
ange
toIs
suer
Sh
are
Th
ista
ble
repo
rts
OL
Sre
gres
sion
sof
the
chan
gein
the
pric
eof
priv
atel
yis
sued
MB
Str
anch
eson
issu
erm
arke
tsh
are,
oth
eris
suer
and
tran
che-
leve
lch
arac
teri
stic
san
dm
arke
tco
ndi
tion
s.T
he
sam
ple
incl
ude
sal
ltra
nch
esfo
rw
hic
hw
eca
nob
serv
epr
ices
onB
loom
berg
orig
inat
edbe
twee
n20
00an
d20
06th
atre
ceiv
edat
leas
ton
era
tin
gfr
omM
oody
’s,S
&P,
orF
itch
.Th
ede
pen
den
tva
riab
leis
the
perc
enta
gech
ange
inth
epr
ice
ofa
tran
che
betw
een
issu
ance
and
Apr
il20
09(o
rit
spa
yoff
date
).S
eeS
ecti
onII
.Bfo
rot
her
vari
able
defi
nit
ion
s.E
ach
regr
essi
onin
clu
des
sepa
rate
inte
rcep
tsfo
rco
upo
nty
pes
(floa
tin
g,fi
xed,
orva
riab
le).
Sta
nda
rder
rors
are
clu
ster
edby
both
coh
ort-
year
and
issu
er(t
wo-
way
clu
ster
ing)
.T-s
tati
stic
sar
ein
pare
nth
eses
.P
anel
sA
and
Bpr
esen
tre
sult
sfo
rA
AA
tran
ches
only
;Pan
els
Can
dD
show
resu
lts
for
non
-AA
Atr
anch
eson
ly.P
anel
sA
and
Cdo
not
hav
eis
suer
fixe
def
fect
s,an
dP
anel
sB
and
Dh
ave
issu
erfi
xed
effe
cts.
∗∗∗ ,
∗∗,a
nd
∗in
dica
tesi
gnifi
can
ceat
the
1%,5
%,a
nd
10%
leve
ls,r
espe
ctiv
ely.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Pan
elA
:AA
ATr
anch
esO
nly
,No
Issu
erF
ixed
Eff
ects
Issu
erS
har
e−0
.376
70.
2962
0.41
70∗∗
0.42
74∗∗
0.39
79∗∗
0.40
24∗∗
0.18
32(−
1.60
)(1
.58)
(2.3
7)(2
.36)
(2.1
5)(2
.50)
(0.9
5)H
OT
∗Is
suer
Sh
are
–−5
.255
5∗∗
−5.9
520∗
∗∗−5
.772
1∗∗∗
−5.8
203∗
∗∗−5
.392
6∗∗
−3.2
683∗
∗
–(−
2.12
)(−
2.73
)(−
2.80
)(−
2.74
)(−
2.57
)(−
2.09
)B
ank
and
Th
rift
−0.0
087
0.00
490.
0058
0.00
760.
0087
0.00
510.
0043
(−0.
79)
(0.3
6)(0
.43)
(0.5
1)(0
.59)
(0.3
3)(0
.20)
Ban
kan
dT
hri
ft∗
Pos
tJu
ly04
0.07
37∗
0.06
30∗
0.05
04∗
0.04
64∗
0.04
34∗
0.04
280.
0291
(1.8
9)(1
.86)
(1.8
0)(1
.66)
(1.6
6)(1
.57)
(0.9
5)L
evel
ofS
ubo
rdin
atio
n–
–−0
.130
9∗−0
.134
9∗∗
−0.0
328
−0.0
997
−0.0
713
––
(−1.
91)
(−2.
12)
(−0.
47)
(−1.
34)
(−1.
05)
HO
T∗
Lev
elof
Su
bord
inat
ion
––
––
−0.5
736
−0.2
901
−0.4
069
––
––
(−1.
04)
(−0.
52)
(−0.
88)
Log
ofY
ield
Spr
ead
––
––
––
−0.0
096
––
––
––
(−0.
28)
HO
T∗
Log
ofY
ield
Spr
ead
––
––
––
−0.5
516∗
∗∗
––
––
––
(−4.
29)
Log
ofP
rin
cipa
l0.
0002
−0.0
011
−0.0
014
−0.0
017
−0.0
019
0.00
35−0
.001
3(0
.02)
(−0.
11)
(−0.
15)
(−0.
19)
(−0.
21)
(0.3
9)(−
0.15
)L
ogof
Wei
ghte
dA
vera
geL
ife
−0.1
490∗
∗∗−0
.149
0∗∗∗
−0.1
494∗
∗∗−0
.149
4∗∗∗
−0.1
495∗
∗∗−0
.149
6∗∗∗
−0.0
944∗
∗
(−4.
73)
(−4.
68)
(−4.
59)
(−4.
58)
(−4.
40)
(−4.
06)
(−2.
50)
Log
ofN
um
ber
ofTr
anch
es−0
.025
7−0
.027
6−0
.027
3−0
.032
4−0
.037
3(−
1.01
)(−
1.09
)(−
1.09
)(−
1.14
)(−
1.19
)F
ra.o
fC
olla
.in
Trou
bled
Sta
tes
−0.0
002
−0.0
002
−0.0
001
−0.0
001
−0.0
001
−0.0
002
−0.0
002
(−0.
22)
(−0.
16)
(−0.
15)
(−0.
15)
(−0.
12)
(−0.
14)
(−0.
15)
Her
fin
dah
lIn
dex
ofC
olla
tera
l0.
0641
0.07
060.
0888
0.08
770.
0837
0.08
200.
0756
(0.5
2)(0
.59)
(0.7
8)(0
.76)
(0.7
1)(0
.59)
(0.6
6)
(Con
tin
ued
)
2124 The Journal of Finance R©
Tab
leV
—C
onti
nu
ed
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Pan
elA
:AA
ATr
anch
esO
nly
,No
Issu
erF
ixed
Eff
ects
Sam
eO
rigi
nat
orS
ervi
cer
−0.0
564∗
∗−0
.047
4−0
.044
0−0
.040
8−0
.040
2−0
.040
1−0
.032
5(−
2.29
)(−
1.53
)(−
1.49
)(−
1.31
)(−
1.29
)(−
1.32
)(−
1.02
)M
issi
ng
Ori
gin
ator
Ser
vice
r0.
0146
0.01
240.
0125
0.01
150.
0112
0.01
080.
0046
(0.5
6)(0
.45)
(0.4
7)(0
.43)
(0.4
3)(0
.37)
(0.1
7)Is
suer
Rat
ing
−0.0
050
0.00
080.
0008
0.00
170.
0023
−0.0
013
−0.0
045
(−0.
97)
(0.2
2)(0
.20)
(0.4
4)(0
.46)
(−0.
19)
(−0.
64)
On
eIn
itia
lRat
ing
––
–−0
.052
8−0
.050
6−0
.058
7−0
.058
5–
––
(−1.
40)
(−1.
32)
(−1.
61)
(−1.
13)
Two
Init
ialR
atin
gs–
––
−0.0
142
−0.0
142
−0.0
225
−0.0
194
––
–(−
0.79
)(−
0.79
)(−
1.22
)(−
1.01
)C
ohor
t-Ye
arF
ixed
Eff
ects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Obs
erva
tion
s3,
602
3,60
23,
602
3,60
23,
602
3,06
53,
065
R2
0.47
50.
480
0.48
40.
486
0.48
70.
495
0.52
2
Pan
elB
:AA
ATr
anch
esO
nly
,wit
hIs
suer
Fix
edE
ffec
ts
Issu
erS
har
e0.
0202
0.13
490.
2144
0.19
320.
1768
0.03
95−0
.076
6(0
.07)
(0.3
7)(0
.58)
(0.5
3)(0
.49)
(0.1
4)(−
0.29
)H
OT
∗Is
suer
Sh
are
–−1
.469
5−1
.885
8−1
.829
6−1
.826
60.
0390
1.39
46–
(−0.
69)
(−0.
94)
(−0.
87)
(−0.
85)
(0.0
2)(0
.87)
Ban
kan
dT
hri
ft∗
Pos
tJu
ly04
0.06
250.
0634
0.05
150.
0511
0.04
970.
0409
0.04
42(1
.34)
(1.3
5)(1
.15)
(1.1
7)(1
.15)
(0.9
7)(1
.07)
Lev
elof
Su
bord
inat
ion
––
−0.0
957
−0.1
018
−0.0
616
−0.0
848
−0.0
484
––
(−1.
27)
(−1.
41)
(−0.
91)
(−1.
52)
(−0.
92)
HO
T∗
Lev
elof
Su
bord
inat
ion
––
––
−0.2
281
−0.1
660
−0.3
293
––
––
(−0.
42)
(−0.
30)
(−0.
73)
Log
ofY
ield
Spr
ead
––
––
––
0.00
19–
––
––
–(0
.06)
HO
T∗
Log
ofY
ield
Spr
ead
––
––
––
−0.5
812∗
∗∗
––
––
––
(−4.
58)
Log
ofP
rin
cipa
l0.
0001
0.00
000.
0004
0.00
020.
0002
0.00
800.
0032
(0.0
1)(0
.00)
(0.0
4)(0
.02)
(0.0
1)(0
.61)
(0.2
5)L
ogof
Wei
ghte
dA
vera
geL
ife
−0.1
502∗
∗∗−0
.150
1∗∗∗
−0.1
499∗
∗∗−0
.150
1∗∗∗
−0.1
501∗
∗∗−0
.146
9∗∗∗
−0.0
950∗
∗∗
(−4.
43)
(−4.
44)
(−4.
23)
(−4.
25)
(−4.
11)
(−3.
67)
(−2.
60)
Are All Ratings Created Equal? 2125
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Pan
elB
:AA
ATr
anch
esO
nly
,wit
hIs
suer
Fix
edE
ffec
ts
Log
ofN
um
ber
ofTr
anch
es–
–−0
.007
9−0
.009
0−0
.008
8−0
.010
6−0
.015
6–
–(−
0.51
)(−
0.56
)(−
0.54
)(−
0.53
)(−
0.68
)F
ra.o
fC
olla
.in
Trou
bled
Sta
tes
0.00
050.
0005
0.00
050.
0005
0.00
050.
0007
0.00
04(0
.45)
(0.4
2)(0
.45)
(0.4
3)(0
.43)
(0.6
8)(0
.41)
Her
fin
dah
lIn
dex
ofC
olla
tera
l0.
1125
0.11
490.
1160
0.11
470.
1127
0.11
450.
1242
(0.8
3)(0
.84)
(0.8
7)(0
.86)
(0.8
1)(0
.73)
(0.8
9)S
ame
Ori
gin
ator
Ser
vice
r−0
.013
9−0
.013
6−0
.013
2−0
.011
9−0
.011
5−0
.008
0−0
.002
5(−
0.46
)(−
0.46
)(−
0.44
)(−
0.37
)(−
0.35
)(−
0.21
)(−
0.07
)M
issi
ng
Ori
gin
ator
Ser
vice
r0.
0065
0.00
630.
0083
0.00
860.
0085
0.00
610.
0045
(0.2
8)(0
.26)
(0.3
4)(0
.34)
(0.3
3)(0
.21)
(0.1
8)Is
suer
Rat
ing
0.02
19∗
0.02
19∗
0.02
28∗
0.02
26∗
0.02
30∗
0.02
57∗
0.02
85∗∗
(1.7
3)(1
.73)
(1.7
9)(1
.88)
(1.7
1)(1
.82)
(2.1
4)O
ne
Init
ialR
atin
g–
––
−0.0
358
−0.0
350
−0.0
383
−0.0
374
––
–(−
0.85
)(−
0.75
)(−
0.88
)(−
0.74
)Tw
oIn
itia
lRat
ings
––
–−0
.005
6−0
.005
5−0
.014
6−0
.013
1–
––
(−0.
26)
(−0.
24)
(−0.
62)
(−0.
55)
Coh
ort-
Year
Fix
edE
ffec
tsYe
sYe
sYe
sYe
sYe
sYe
sYe
sIs
suer
Fix
edE
ffec
tsYe
sYe
sYe
sYe
sYe
sYe
sYe
sO
bser
vati
ons
3,60
23,
602
3,60
23,
602
3,60
23,
065
3,06
5R
20.
514
0.51
50.
516
0.51
70.
517
0.53
10.
555
Pan
elC
:Non
-AA
ATr
anch
esO
nly
,No
Issu
erF
ixed
Eff
ects
Issu
erS
har
e−0
.050
21.
1878
∗∗∗
1.18
96∗∗
∗1.
2137
∗∗∗
1.16
07∗∗
∗0.
3590
0.39
08(−
0.15
)(3
.41)
(3.6
0)(4
.06)
(3.3
9)(0
.79)
(1.0
3)H
OT
∗Is
suer
Sh
are
–−9
.228
7∗∗∗
−9.0
623∗
∗∗−9
.448
8∗∗∗
−8.9
954∗
∗∗−6
.697
9∗∗∗
−6.1
320∗
∗∗
–(−
4.83
)(−
4.98
)(−
6.04
)(−
5.45
)(−
2.71
)(−
2.92
)B
ank
and
Th
rift
−0.0
238
−0.0
165
−0.0
164
−0.0
157
−0.0
157
−0.0
029
0.00
01(−
0.86
)(−
0.67
)(−
0.65
)(−
0.64
)(−
0.64
)(−
0.14
)(0
.00)
Ban
kan
dT
hri
ft∗
Pos
tJu
ly04
−0.0
445
−0.0
507
−0.0
483
−0.0
484
−0.0
469
−0.0
581∗
∗−0
.059
6∗∗
(−1.
21)
(−1.
52)
(−1.
41)
(−1.
45)
(−1.
37)
(−2.
22)
(−2.
31)
Lev
elof
Su
bord
inat
ion
––
0.02
960.
0312
−0.0
809
−0.1
477
−0.2
626
––
(0.7
2)(0
.67)
(−0.
43)
(−0.
86)
(−1.
28)
HO
T∗
Lev
elof
Su
bord
inat
ion
––
––
0.62
800.
8871
1.42
12–
––
–(0
.67)
(0.9
1)(1
.31)
(Con
tin
ued
)
2126 The Journal of Finance R©
Tab
leV
—C
onti
nu
ed
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Pan
elC
:Non
-AA
ATr
anch
esO
nly
,No
Issu
erF
ixed
Eff
ects
Log
ofY
ield
Spr
ead
––
––
––
0.12
41∗∗
––
––
––
(2.1
4)H
OT
∗L
ogof
Yie
ldS
prea
d–
––
––
–−0
.582
8∗∗
––
––
––
(−2.
40)
Log
ofP
rin
cipa
l−0
.020
5∗−0
.021
8∗−0
.020
3∗−0
.018
9−0
.018
7−0
.026
4∗−0
.028
5∗∗
(−1.
75)
(−1.
75)
(−1.
66)
(−1.
50)
(−1.
49)
(−1.
90)
(−2.
17)
Log
ofW
eigh
ted
Ave
rage
Lif
e−0
.167
1∗∗∗
−0.1
622∗
∗∗−0
.161
9∗∗∗
−0.1
558∗
∗−0
.156
5∗∗
−0.1
518∗
∗−0
.157
3∗∗
(−2.
76)
(−2.
72)
(−2.
67)
(−2.
51)
(−2.
52)
(−2.
08)
(−2.
27)
Log
ofN
um
ber
ofTr
anch
es–
–−0
.002
7−0
.000
00.
0003
0.00
350.
0075
––
(−0.
12)
(−0.
00)
(0.0
1)(0
.16)
(0.3
6)F
ra.o
fC
olla
.in
Trou
bled
Sta
tes
0.00
120.
0013
0.00
130.
0012
0.00
120.
0012
0.00
12(1
.17)
(1.2
9)(1
.34)
(1.2
7)(1
.15)
(1.2
8)(1
.28)
Her
fin
dah
lIn
dex
ofC
olla
tera
l−0
.223
3−0
.208
5−0
.207
1−0
.200
1−0
.197
5−0
.177
3−0
.217
5(−
1.12
)(−
1.10
)(−
1.07
)(−
1.04
)(−
1.02
)(−
0.96
)(−
1.15
)S
ame
Ori
gin
ator
Ser
vice
r−0
.048
7∗∗
−0.0
419∗
−0.0
415∗
∗−0
.044
7∗∗
−0.0
451∗
∗−0
.025
7−0
.024
0(−
1.99
)(−
1.91
)(−
2.00
)(−
2.11
)(−
2.17
)(−
1.28
)(−
1.30
)M
issi
ng
Ori
gin
ator
Ser
vice
r0.
0200
0.01
660.
0169
0.01
670.
0170
0.01
080.
0122
(1.4
5)(1
.27)
(1.1
9)(1
.15)
(1.0
2)(0
.60)
(0.6
8)Is
suer
Rat
ing
−0.0
026
−0.0
027
−0.0
033
−0.0
042
−0.0
042
−0.0
104
−0.0
128
(−0.
25)
(−0.
33)
(−0.
35)
(−0.
50)
(−0.
49)
(−1.
01)
(−1.
15)
On
eIn
itia
lRat
ing
––
–0.
0425
0.04
010.
0606
∗∗0.
0588
∗∗
––
–(1
.33)
(1.1
7)(2
.46)
(2.5
5)Tw
oIn
itia
lRat
ings
––
–0.
0077
0.00
710.
0152
0.01
56–
––
(0.3
8)(0
.35)
(1.2
4)(0
.98)
Rat
ing
Dis
agre
emen
t–
––
−0.0
008
−0.0
013
0.00
720.
0050
––
–(−
0.05
)(−
0.08
)(0
.52)
(0.3
1)C
ohor
t-Ye
arF
ixed
Eff
ects
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Init
ialR
atin
gC
ateg
ory
Du
mm
ies
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Obs
erva
tion
s2,
624
2,62
42,
614
2,61
42,
614
2,33
02,
330
R2
0.61
00.
616
0.61
60.
617
0.61
80.
563
0.57
2
Are All Ratings Created Equal? 2127
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Pan
elD
:Non
-AA
ATr
anch
esO
nly
,wit
hIs
suer
Fix
edE
ffec
ts
Issu
erS
har
e−0
.368
20.
5423
0.57
500.
6779
∗0.
6293
−0.4
490
−0.5
414
(−0.
84)
(1.4
4)(1
.50)
(1.7
5)(1
.50)
(−0.
96)
(−1.
12)
HO
T∗
Issu
erS
har
e–
−9.9
492∗
∗∗−9
.924
4∗∗∗
−10.
3939
∗∗∗
−9.8
894∗
∗∗−9
.354
0∗∗∗
−7.9
951∗
∗∗
–(−
3.06
)(−
2.98
)(−
3.50
)(−
3.39
)(−
3.71
)(−
3.24
)B
ank
and
Th
rift
∗P
ost
July
04−0
.068
0∗−0
.058
2∗−0
.058
8−0
.060
9∗−0
.058
2−0
.057
9∗−0
.058
5∗∗
(−1.
93)
(−1.
64)
(−1.
53)
(−1.
67)
(−1.
47)
(−1.
86)
(−2.
05)
Lev
elof
Su
bord
inat
ion
––
−0.0
157
−0.0
127
−0.1
212
−0.1
869
−0.3
232
––
(−0.
31)
(−0.
24)
(−0.
58)
(−1.
12)
(−1.
63)
HO
T∗
Lev
elof
Su
bord
inat
ion
––
––
0.61
700.
9290
1.61
75–
––
–(0
.59)
(1.0
5)(1
.60)
Log
ofY
ield
Spr
ead
––
––
––
0.13
18∗∗
––
––
––
(2.1
1)H
OT
∗L
ogof
Yie
ldS
prea
d–
––
––
–−0
.579
7∗∗
––
––
––
(−2.
29)
Log
ofP
rin
cipa
l−0
.017
8−0
.016
5−0
.016
1−0
.014
0−0
.013
8−0
.017
1−0
.018
5(−
1.58
)(−
1.47
)(−
1.51
)(−
1.23
)(−
1.24
)(−
1.25
)(−
1.45
)L
ogof
Wei
ghte
dA
vera
geL
ife
−0.1
213
−0.1
217
−0.1
249
−0.1
173
−0.1
174
−0.1
367
−0.1
341
(−1.
43)
(−1.
45)
(−1.
43)
(−1.
36)
(−1.
41)
(−1.
43)
(−1.
43)
Log
ofN
um
ber
ofTr
anch
es–
–−0
.017
1−0
.015
6−0
.015
70.
0023
0.00
34–
–(−
0.72
)(−
0.69
)(−
0.76
)(0
.11)
(0.1
6)F
ra.o
fC
olla
.in
Trou
bled
Sta
tes
0.00
24∗∗
0.00
23∗∗
0.00
23∗∗
0.00
24∗∗
0.00
23∗∗
0.00
160.
0016
(2.1
0)(2
.08)
(2.0
9)(2
.14)
(2.0
2)(1
.27)
(1.3
1)
(Con
tin
ued
)
2128 The Journal of Finance R©
Tab
leV
—C
onti
nu
ed
(1)
(2)
(3)
(4)
(5)
(6)
(7)
Pan
elD
.Non
-AA
ATr
anch
esO
nly
,wit
hIs
suer
Fix
edE
ffec
ts
Her
fin
dah
lIn
dex
ofC
olla
tera
l−0
.201
5−0
.184
5−0
.189
8−0
.182
8−0
.179
2−0
.094
1−0
.137
9(−
0.92
)(−
0.82
)(−
0.85
)(−
0.81
)(−
0.79
)(−
0.45
)(−
0.66
)S
ame
Ori
gin
ator
Ser
vice
r−0
.023
9−0
.021
1−0
.020
3−0
.020
7−0
.022
0−0
.006
6−0
.007
1(−
0.96
)(−
1.06
)(−
1.00
)(−
0.99
)(−
1.04
)(−
0.33
)(−
0.37
)M
issi
ng
Ori
gin
ator
Ser
vice
r0.
0237
0.02
160.
0233
0.02
200.
0223
0.01
320.
0141
(1.3
3)(1
.12)
(1.1
6)(1
.09)
(1.0
4)(0
.66)
(0.7
5)Is
suer
Rat
ing
0.02
30∗∗
0.01
94∗∗
0.01
660.
0174
0.01
770.
0216
0.01
52(2
.26)
(1.9
9)(1
.53)
(1.6
0)(1
.62)
(1.4
4)(0
.95)
On
eIn
itia
lRat
ing
––
–0.
0527
0.04
980.
0789
∗∗∗
0.07
08∗∗
––
–(1
.34)
(1.1
1)(2
.89)
(2.4
9)Tw
oIn
itia
lRat
ings
––
–0.
0115
0.01
080.
0254
∗0.
0234
––
–(0
.62)
(0.5
6)(1
.88)
(1.6
0)R
atin
gD
isag
reem
ent
––
–0.
0114
0.01
100.
0169
0.01
31–
––
(0.5
8)(0
.54)
(0.8
6)(0
.59)
Coh
ort-
Year
Fix
edE
ffec
tsYe
sYe
sYe
sYe
sYe
sYe
sYe
sIs
suer
Fix
edE
ffec
tsYe
sYe
sYe
sYe
sYe
sYe
sYe
sIn
itia
lRat
ing
Cat
egor
yD
um
mie
sYe
sYe
sYe
sYe
sYe
sYe
sYe
sO
bser
vati
ons
2,62
42,
624
2,61
42,
614
2,61
42,
330
2,33
0R
20.
642
0.64
60.
646
0.64
70.
647
0.60
30.
611
Are All Ratings Created Equal? 2129
Figure 1. Initial yield spreads of privately sold MBS (sorted by issuing year and is-suer market share). This figure shows the median initial yield spreads of privately issued MBStranches sorted by issuing year (cohort) and issuer market share. The sample includes all tranchesfor which we observe initial coupons (or fixed spreads) on Bloomberg originated between 2000 and2006 that received at least one rating from Moody’s, S&P, or Fitch. For a tranche with floatingcoupon, the yield spread is defined as the fixed markup over the benchmark rate specified at is-suance (e.g., the one-month LIBOR rate). For a tranche with fixed or variable coupon, yield spreadis defined as the difference between the initial coupon rate and the yield on a Treasury securitywhose maturity is closest to the tranche’s weighted average life. Issuer market share is calculatedas the number of deals originated by the issuer divided by the total number of deals in the currentyear. “Big issuer” means that the market share of the issuer falls into the top 10% of the marketshare distribution in that year, and “Small issuer” refers to the rest of issuers in that year.
indicator (equal to one for tranches with multiple ratings that disagree), andLog of Number of Tranches (a proxy for deal complexity). The dependent vari-able equals the natural log of the yield spread calculated at the issuance date.We split the sample into AAA-rated tranches (AAA-rated by all ratings) versusall non-AAA rated tranches, and we also include dummy variables for coupontypes (floating, fixed, or variable; not reported in tables).
Panel A, Table III, reports the yield results for AAA-rated tranches withoutissuer fixed effects. This model exploits cross-sectional as well as within-issuertime variation in market share. The yield on tranches sold by large issuers ison average higher than that on tranches sold by small issuers during boomyears. The coefficient from the baseline model (Column 1, Panel A) is positivebut not significant, but in Column 2 we find a strong interaction between issuersize and market conditions. In a hot year such as 2006 (when HOT = 0.25),the yield spread would be about 13% higher for an issuer with a 10% marketshare (such as Countrywide or GM) relative to a very small issuer (–1.23×0.1+ 9.8987×0.25×0.1 = 12.6%). This effect translates into a 19 bp increase inyields, somewhat smaller than the unconditional comparisons in Figure 1.
2130 The Journal of Finance R©
Level of Subordination enters the yield spread regressions with a very strongpositive coefficient, although its inclusion does little to the effects of issuer size;nor does it interact with market conditions (HOT). Increasing this variable fromthe 25th to the 75th percentile of its distribution would increase yield spreadsby about 10%. This makes sense because Level of Subordination representsthe degree of leverage in the tranche, so greater leverage implies greater riskand thus higher yields. We also find some evidence that regulatory distortionsaffected the ratings process. During the post-July 2004 period, tranches issuedby banks and thrifts had yield spreads 10% to 15% higher than those sold byless regulated entities. This may reflect their greater incentive to securitizemore aggressively to lower the effect of regulatory capital requirements.
Panel B of Table III reports the same set of models but includes issuerfixed effects. We find that the effects of issuer size (through interactions withHOT) are somewhat smaller but remain statistically significant. Magnitudesare only slightly smaller, despite the large decline in the interaction term,because the linear term switches sign. The yield spread would be about 10%higher for an issuer with a 10% market share relative to a very small issuer(0.38×0.1 + 2.623×0.25×0.1 = 10.4%), versus 13% from the model withoutfixed effects. The effects of subordination and the regulatory indicators arealso similar. Interestingly, Issuer Rating enters the fixed effects model witha positive coefficient, suggesting that declines in an issuer’s credit standingare priced into the deals that they sell, perhaps because the value of implicitrecourse falls as issuer credit quality declines (Gorton and Souleles (2006)).This effect only emerges in the fixed effects specification, however. We also findthat the yield on tranches for which the same institution acts as originator andservicer is higher than that on tranches with different originator and servicer,and this result is robust to the inclusion of issuer fixed effects.
We also obtain a number of interesting results on how the market prices cer-tain MBS tranches. For example, we find that the yield on AAA-rated tranchesincluded in deals with a greater number of tranches is higher in both panels, in-dicating that investors are suspicious of the quality of more complicated deals.The coefficient in Column 3, Panel A, implies that, as the number of tranchesin a deal doubles (the 25th percentile of this variable is eight while the 75th
percentile is 19), the yield on the AAA tranche increases by a little more than10%. This result is consistent with theories of ratings inflation based on regu-lation arbitrage (Opp, Opp, and Harris (2012)) and asset complexity (Mathis,McAndrews, and Rochet (2009), Skreta and Veldkamp (2009)). Controlling forthe effect of deal complexity, however, does not change the link from issuersize to yield spreads. We also find that tranches with more underlying mort-gages originated from troubled states (AZ, CA, FL, and NV) have higher yields(though not significant with issuer fixed effects). Interestingly, we find thatbetter-diversified AAA-rated deals, as measured by the cross-state HHI, havehigher yields (again only without fixed effects). This result supports the modelof Coval, Jurek, and Stafford (2009a), who show that AAA-rated structured-finance deals with a high degree of diversification act like economic catastro-phe bonds that would default only under dire economic scenarios. Thus, such
Are All Ratings Created Equal? 2131
bonds must offer high yields to compensate investors for bearing systematicrisk.
For the non-AAA-rated tranches (Panels C and D), we find similar resultsfor issuer size as in the AAA market. The magnitudes are a bit smaller inthe models without fixed effects, and a bit larger in the models with fixedeffects. Increasing issuer share from very small to 10% during a hot year wouldincrease yield spreads by almost 10% (–1.04 × 0.1 + 7.91 × 0.25 × 0.1 = 9.8%)based on Column 2 of Panel C. In the models with fixed effects the magnitudeincreases to 17% (0.27 × 0.1 + 5.92 × 0.25 × 0.1 = 17.5%). The variables relatedto ratings shopping enter the non-AAA market with similar magnitude butgreater statistical significance compared to the AAA market. We find thattranches with one rating have yields about 7% to 9% higher than those withall three ratings (the omitted group) and the tranches with two ratings haveyields about 4% to 5% higher than the omitted group. Rating disagreementalso enters the model very significantly—tranches with disagreement have11% higher spreads—suggesting that there may be a large payoff to ratingsshopping, since shopping could conceal the lower rating from investors.16
Table IV, Panel A, separates our sample into two groups based on issuer size(as in Table I), and into another two groups based on AAA versus non-AAA(as in Table III). We report just one specification for each of the four samples(without issuer fixed effects), and we do not include the issuer size variableitself since we have “controlled” for this factor through the sample split. Thisapproach allows us to test whether other issuer as well as deal and tranchecharacteristics have different effects on yield depending on issuer size. Theresults suggest, first, that the effect of the level of subordination is consistentacross the samples. For tranches sold by both large and small issuers, wefind that more aggressive subordination leads to higher yields; moreover, themagnitude of this effect is similar across both samples (compare Column 1with 2 and Column 3 with 4). We also find that the effects of regulatory statusand deal complexity carry through to both samples. More complex deals—thosewith more tranches—have higher yields in both AAA samples, and the yieldsare higher for deals issued by banks and thrifts after 2004 for both samples.
We do find one significant difference across the two samples within the AAAmarket: the effect of multiple ratings is much more pronounced for tranchessold by small issuers. For these tranches, having three ratings is associatedwith much lower yields, on the order of 15% to 19% lower. We have argued thatlarge issuers have more bargaining power when dealing with rating agenciesthan small issuers due to their greater market share. Hence, even if a tranchesold by a large issuer has multiple ratings, the market may remain suspicious ofits quality—results from Table III confirm this as yields are higher on tranchessold by large issuers even after controlling for the number of ratings. On theother hand, when a tranche sold by a small issuer has only one or two ratings,the market can perhaps draw clearer inferences that the tranche has been
16Rating Disagreement is undefined for the AAA sample because we only include tranches ratedAAA by all the agencies that rated the tranche.
2132 The Journal of Finance R©
shopped—because it is doubtful that the issuer can push any one agency toalter the rating—and demand a greater price discount. Consistent with thisinterpretation, the frequency of having three ratings is much higher amongsmall issuers than large ones (recall Table I). Put slightly differently, investorsare in a better position to judge whether small issuers have shopped theirproducts (based on having one or two ratings) and punish them accordingly;hence it makes sense that small issuers are more likely to purchase multipleratings, even when those ratings disagree.
In Panel B of Table IV, we illustrate the effect of issuer size on yields usingthe split-sample results. We compare the average yield for tranches actuallysold by large issuers with the predicted yield for those same tranches usingthe coefficients from the small-issuer sample. The strategy offers a robustnesstest to our baseline analysis in Table III, where we constrain all of the effectsof issuer characteristics to be the same across all tranches. We find resultsthat are slightly larger in magnitude to the more parsimonious model. Duringthe non-boom period (2000 to 2003), average yields for large-issuer tranchesare just slightly higher than what would be predicted had those tranches beensold by small issuers: 176 versus 171 bps for AAA-rated and 182 versus 176for non-AAA rated. During the boom years (2004 to 2006), however, the gapwidens to about 15 bps. In the AAA market, the average yield for tranches soldby large issuers was 98 bps, compared to 83 bps that would have been predictedhad those tranches been sold by small issuers (an increase of 18%). We find asimilar gap in the non-AAA market. Thus, even allowing all of the coefficientsto differ, we continue to find strong evidence that investors priced the risk thatotherwise-similar tranches carry greater risk when sold by large issuers.
B. Ex Post Price Performance
Figure 2 presents simple unconditional graphical evidence on our secondtest—price change after issuance—for the two groups of securities. For all thetranches, the initial price is set at par—$100 per $100 face value, or very closeto $100. We group tranches by their issuance year (cohort); Figure 2(a) plotsthe median cumulative price change for all tranches in the 2000 to 2003 cohortsfrom the first month after issuance until April 2009 (or the last reported price),while Figure 2(b) plots the median price change for the 2004 to 2006 cohorts.Prices in all the cohorts from both figures remain more or less flat during thefirst few years after issuance, but begin to drop early in 2007. From Figure2(b), prices of tranches issued during the market boom period of 2004, 2005,and 2006 and by large issuers dropped by 54% from the issuance date, ascompared to a 37% drop by small issuers, a difference of 17 percentage pointsbetween these two groups.
Table V reports regressions testing whether the patterns in Figure 2 continueto hold after adding control variables. As in Table III, we start with reducedform models that control for collateral and issuer characteristics, and includethe full set of credit rating indicators for non-AAA tranches (Columns 1 and2). We then add deal structure variables (Columns 3 to 6) and, at last, we
Are All Ratings Created Equal? 2133
Figure 2. Monthly price changes of privately sold MBS (sorted by issuer market shareand issuing year). This figure shows the change in the monthly price of privately issued MBStranches sorted by issuer market share, calculated as the number of deals sold by the issuer dividedby the total number of deals in the current year. “Big issuer” means that the market share of theissuer falls into the top 10% of the market share distribution in that year, and “Small issuer”refers to the rest of issuers in that year. The sample includes all tranches for which we can observeprices on Bloomberg originated between 2000 and 2006. The price history starts from the monthof issuance to the month that the security stops trading or April 2009, whichever comes first. (a)Median prices for tranches originated between 2000 and 2003, and (b) median prices for tranchesoriginated between 2004 and 2006.
2134 The Journal of Finance R©
add the ex ante log yield spread variable (Column 7). Adding the yield spreadallows us to test the extent to which the market priced the risk of large-issuerdeals. That is, if the market prices this risk, then the effect of issuer sizeought to be attenuated or even eliminated in its ability to predict outcomes.The dependent variable is price change, for which we have one observation pertranche, calculated as the percentage change between the price during the firstmonth after issuance and the final price as of April 2009 (if available) or the lastavailable monthly price otherwise. As noted earlier, the sample is considerablysmaller than our sample of initial yields (Tables III and IV) because Bloombergonly provides a pricing history on a subset of the tranches.17
As in Tables III and IV, Table V again separates results into AAA-ratedtranches with and without issuer fixed effects (Panels A and B) versus all othertranches (Panels C and D). We find a negative and significant impact of issuersize for both samples during boom years, although not in the AAA samplewith issuer fixed effects. The coefficients from the baseline models suggestthat tranches sold by large issuers fell by about 10 percentage points morethan those sold by small issuers in the AAA market during boom years (fromPanel A, Column 2: 0.296 × 0.1 − 5.26 × 0.25 × 0.1 = −10%), and 11 percentagepoints more in the lower-rated tranches (from Panel C, Column 2: 1.188 × 0.1− 9.23 × 0.25 × 0.1 = −11.2%).18
The results also suggest that market prices incorporate the ex post riskof a bad outcome, but only during the boom years. In both AAA and non-AAA rated tranches, the interaction between Log of Yield Spread with HOT isnegative and significant. In Columns 6 and 7, we estimate the same samplewith and without the yield variable to judge the extent to which adding pricingattenuates the effect of issuer size on ex post outcomes. The results suggest asmall attenuation, but only in the AAA-rated sample.
C. Discussion
Overall, the results across Tables III and IV suggest that the market pricesthe risk of large-issuer sponsored deals, conditional on the credit rating andduring the housing boom. The positive effect of issuer size during the boomis robust to including controls for regulatory arbitrage, to unobserved hetero-geneity across issuers, and to various dimensions of deal structure, including
17About 45% of the 9,299 tranches for which we have information on pricing history arepaid off early and before the crisis. Once they are paid off, the ratings are withdrawn and re-ported price series stop. In robustness tests (results shown in the Internet Appendix, availableat http://www.afajof.org/supplements.asp) we estimate our ex post performance regressions with-out tranches that are paid off early. The results are similar to those reported in Table V for theAAA market; for non-AAA tranches, the signs and statistical significance remain in most models,although the magnitudes fall.
18We have also estimated the effect of issuer size for subsamples based on whether a deal isabove or below the median geographical concentration and also whether a deal is above or belowthe median fraction of collateral in troubled states. We find that in each of these subsamples the expost performance is worse among cohorts sold by large issuers during hot markets. These resultsare available in the Internet Appendix.
Are All Ratings Created Equal? 2135
the level of subordination. The level of subordination itself is also strongly cor-related with yields; more aggressively structured deals require higher yieldseven conditioning on the rating, and this result is consistent and has similarmagnitude for tranches sold by both large and small issuers. Yet adding thisvariable does little to the effect of issuer size on yield. In fact, the magnitudesare almost completely unaffected by all of the controls (other than the issuerfixed effects).
Do higher yields for large-issuer MBS indicate that ratings were intention-ally inflated more for them than for smaller issuers? This inference makessense theoretically because greater deal flow enhances large issuers’ bargain-ing power. But it is also possible that large issuers were merely better able toconstruct pools that would seem safer according to the models used to build theratings. Our results do not allow us to separate these two possibilities becausewe do not look specifically at the rating process itself. In contrast, Griffin andTang (2012) do present evidence that a major rating agency altered model re-sults to inflate ratings on a small sample of CDOs, though they do not link thedegree of inflation to the size of issuers. Based on our evidence, we can concludethat investors perceived higher risk in large-issuer sold deals (conditional onthe rating). The ex post performance in Table V suggests that this pricing wassensible.
This discussion also raises the question of whether ratings inflation evenmatters. If investors can price risks, why do we care? First, not all investorsare sophisticated, so inflated ratings may lead to misallocation of risk across thefinancial system (i.e., inefficient risk sharing). Second, even for sophisticatedinvestors, the fact that the ratings process could be compromised is a problemas they may not know how much to trust the ratings, or how much informationis in fact embedded in the ratings. Third, and perhaps most importantly, theregulation of large and “too big to let fail” financial institutions depended on theaccuracy of credit ratings. Ratings inflation allowed these regulated firms toincrease leverage beyond what was justified by their risks, thereby making thefinancial system as a whole more vulnerable to small shocks. The costs of thisexcess leverage became all too clear during the financial crisis. Efforts towardregulatory reform, such as the Dodd–Frank Act, reasonably recognize the dan-ger of placing too much weight on credit ratings in devising regulations. Theproblem that Dodd–Frank does not address, however, is finding an alternativemeans to assess risk for regulatory compliance at a reasonable cost.19
IV. Conclusions
Our paper tests whether issuer size affected the pricing of MBS, one of thelargest and fastest growing credit markets. Rating agencies play a crucial role
19Stulz (2008) points out that knowing whether risks are correctly assessed and the nature ofthe mistakes is key for risk management. Coffee (2010) discusses how regulations should dependless on ratings; he also discusses potential problems in moving away from the issuer-pay model inpractice.
2136 The Journal of Finance R©
in the corporate bond market, and were a key part of the rise and fall of thehousing and MBS markets. It is perhaps not surprising to see that analyticalmodels used by rating agencies were imperfect. Many sophisticated investorsand policy makers systematically underestimated default risk in housing, par-ticularly the risk that the whole U.S. housing market would decline simulta-neously. Our findings, however, suggest that investors suspected that mistakeswere systematically correlated with issuer size and market conditions: yieldson tranches sold by large issuers were higher conditional on the credit ratingand other controls. For both AAA- and non-AAA-rated tranches sold by largeissuers, their prices drop more than similar tranches sold by smaller issuerswhen the housing bubble began to unravel. These performance differences areconcentrated among deals packaged and sold during the market boom years of2004 through 2006. We also find that ratings-based regulations and regulatoryarbitrage of financial institutions distorted the rating process, but controllingfor these effects does not change our main results on issuer size. Overall, weconclude that there is a robust relation between issuer size and the marketprices of mortgage-backed securities conditional on ratings. Conflicts betweenthe interests of issuers (who pay for ratings) versus those of investors (whoconsume ratings) may explain this relationship.
Initial submission: 24 January 2011; Final version received: 3 July 2012Editor Campbell Harvey
REFERENCESAcharya, Viral, and Matthew Richardson, 2009, Causes of the financial crisis, Critical Review 21,
195–210.Acharya, Viral, Philipp Schanbl, and Gustavo Suarez, 2012, Securitization without risk transfer,
Journal of Financial Economics, forthcoming.Adelino, Manuel, 2009, How much do investors rely on ratings? The case of mortgage-backed
securities, Working paper, Dartmouth College.Alexander, William, Scott Grimshaw, Grant McQueen, and Barret Slade, 2002, Some loans are
more equal than others: Third-party origination and defaults in the subprime mortgage in-dustry, Real Estate Economics 30, 667–697.
Ashcraft, Adam, Paul Goldsmith-Pinkham, and James Vickery, 2010, MBS ratings and the mort-gage credit boom, Working paper, Federal Reserve Bank of New York.
Ashcraft, Adam, and Til Schuermann, 2008, Understanding the securitization of subprime mort-gage credit, Foundations and Trends in Finance 2, 191–209.
Bar-Isaac, Heski, and Joel Shapiro, 2010, Ratings quality over the business cycle, Working paper,Oxford University.
Becker, Bo, and Todd Milbourn, 2010, How did increased competition affect credit ratings? Journalof Financial Economics 101, 493–514.
Benmelech, Efraim, and Jennifer Dlugosz, 2009a, The alchemy of CDO credit ratings, Journal ofMonetary Economics 56, 617–634.
Benmelech, Efraim, and Jennifer Dlugosz, 2009b, The credit rating crisis, Working paper, HarvardUniversity.
Bolton, Patrick, Xavier Freixas, and Joel Shapiro, 2012, The credit ratings game, Journal ofFinance 67, 85–112.
Bongaerts, D., Martijn Cremers, and William Goetzmann, 2012, Tiebreaker: Certification andmultiple ratings, Journal of Finance 67, 112–152.
Are All Ratings Created Equal? 2137
Cameron, Colin, Jonah Gelbach, and Douglas Miller, 2006, Robust inference with multi-way clus-tering, NBER Technical Working Paper No. 327.
Coffee, John C. Jr., 2010, Ratings reform: The good, the bad, and the ugly, Working paper, ColumbiaUniversity Law School.
Coval, Joshua, Jakub Jurek, and Erik Stafford, 2009a, Economic catastrophe bonds, AmericanEconomic Review 99, 628–666.
Coval, Joshua, Jakub Jurek, and Erik Stafford, 2009b, The economics of structured finance, Jour-nal of Economic Perspectives 23, 3–25.
Demiroglu, Cem, and Christopher James, 2010, Works of friction: Originator-sponsor affiliationand losses in mortgage-backed securities, Working paper, University of Florida.
Frenkel, Sivan, 2010, Repeated interaction and rating inflation: A model of double reputation,Working paper, Tel Aviv University.
Furfine, Craig, 2010, Deal complexity, loan performance and the pricing of commercial mortgage-backed securities, Working paper, Northwestern University.
Gorton, Gary, and Andrew Metrick, 2011, Securitized banking and the run on repo, Journal ofFinancial Economics 104, 425–451.
Gorton, Gary, and Nick Souleles, 2006, Special purpose vehicles and securitization, in Mark Careyand Rene Stulz, eds.: The Risks of Financial Institutions (University of Chicago Press, Chicago,IL).
Griffin, John, and Dragon Tang, 2012, Did subjectivity play a role in CDO credit ratings? Journalof Finance 67, 1293–1328.
Jorion, Philippe, Liu Zhu, and Charles Shi, 2005, Informational effects of regulation FD: Evidencefrom rating agencies, Journal of Financial Economics 76, 309–330.
Keys, Benjamin J., Tanmoy K. Mukherjee, Amit Seru, and Vikrant Vig, 2010, Did securitizationlead to lax screening? Evidence from sub-prime loans, Quarterly Journal of Economics 125,307–362.
Kisgen, Darren, 2006, Credit rating and capital structure, Journal of Finance 61, 1035–1072.Kisgen, Darren, Jun Qian, and Weihong Song, 2009, Are fairness opinions fair? The case of mergers
and acquisitions, Journal of Financial Economics 91, 179–207.Kisgen, Darren, and Philip E. Strahan, 2010, Do regulations based on credit ratings affect a firm’s
cost of capital? Review of Financial Studies 23, 4324–4347.Loutskina, Elena, and Philip E. Strahan, 2010, Informed and uninformed investment in housing:
The downside of diversification, Review of Financial Studies 24, 1447–1480.Mathis, J., Jerome McAndrews, and Jean-Charles Rochet, 2009, Rating the raters: Are reputation
concerns powerful enough to discipline rating agencies? Journal of Monetary Economics 56,657–674.
Mian A., and Amir Sufi, 2009, The consequences of mortgage credit expansion: Evidence from the2007 mortgage crisis, Quarterly Journal of Economics 124, 1449–1496.
Nadauld, Taylor, and Shane Sherlund, 2009, The role of the securitization process in expansionof subprime credit, Working paper, Brigham Young University and Federal Reserve Board ofGovernors.
Opp, Christian, Marcus Opp, and Milton Harris, 2012, Rating agencies in the face of regulation,Journal of Financial Economics, forthcoming.
Sangiorgi, Francesco, and Chester Spatt, 2010, Equilibrium credit ratings and policy, Workingpaper, Carnegie Mellon University.
Saunders, Anthony, and Marcia M. Cornett, 2007, Financial Institutions Management, 6th edition(McGraw-Hill, New York, NY).
Skreta, Vasiliki, and Laura Veldkamp, 2009, Ratings shopping and asset complexity: A theory ofratings inflation, Journal of Monetary Economics 56, 678–695.
Standard and Poor’s, 2008, Standard & Poor’s Rating Services U.S. Rating Fees Disclosure, pressrelease, (http://www2.standardandpoors.com/spf/pdf/fixedincome/RatingsFees2008.pdf).
Stulz, Rene M., 2008, Risk management failures: What are they and when do they happen? Journalof Applied Corporate Finance 20, 58–67.