CREDIT RATINGS AND THE EVOLUTION OF THE MORTGAGE-BACKED SECURITIES MARKET

Transcription

1 CREDIT RATINGS AND THE EVOLUTION OF THE MORTGAGE-BACKED SECURITIES MARKET Jie (Jack) He Jun QJ Qian Philip E. Strahan University of Georgia Boston College Boston College & NBER Current version: November 2010 Abstract We examine whether rating agencies (Moody s, S&P, and Fitch) reward large issuers of mortgagebacked securities, who bring substantial business, by granting them unduly favorable ratings. Tranches sold by large issuers are more likely to have only one rating, and when they have multiple ratings they tend to be the same. The initial yield on the AAA-rated tranches sold by large issuers is higher than that on similar tranches sold by small issuers, especially during the market boom years of For both AAA and non-aaa rated tranches sold by large issuers, their prices drop more than those sold by small issuers, and the differences are concentrated during We conclude that large issuers are more likely to shop for better ratings, and the market is aware of this incentive and skeptical of the quality of ratings, especially during booming periods. JEL Classifications: G2, G1, L1. Keywords: Credit ratings, mortgage-backed securities, tranche, conflict of interest, yield. We appreciate helpful comments from Patrick Bolton, Joel Shapiro, Richard Stanton, Dragon Tang, James Vickery, and seminar/session participants at Boston College, Federal Reserve Bank of New York, Queen s University (Canada), China International Conference in Finance (Beijing), European Finance Association meetings (Frankfurt), and the NBER conference on securitization. We thank Calvin Chau, Hugh Kirkpatrick, Sailu Li, Yingzhen Li, and Chenying Zhang for excellent research assistance and Boston College for financial support. We are responsible for all the remaining errors.

2 I. Introduction As the most severe financial and economic crisis since the Great Depression unfolds, scholars, practitioners, and regulators have been studying its causes and possible cures to prevent similar crises in the future. At the center of the crisis is the explosive growth of the mortgagebacked securities (MBS) market, which is both fueled by and fueling the housing market boom. In this paper, we study an important piece of the evolution of the MBS market the rating agencies, Moody s, S&P, and Fitch (the largest three agencies), in particular and their role in the expansion of the MBS market. Specifically, we examine whether conflicts of interest lie behind the growth of MBS, and whether and when the market began to realize this incentive. We ask, in particular, did the rating agencies grant large MBS issuers, who brought substantial business, unduly favorable ratings? Rating agencies play an important role in all fixed income securities markets, in part because they have access to private information on the securities and issuers and this privilege is protected from regulations such as Reg-FD. 1 However, rating agencies have recently been sharply criticized for their practice that is behind the rise and fall of the MBS market. Many criticisms are based on the fact that the rating industry faces a potential conflict in their fees/income structure. Instead of being compensated and rewarded by the consumers (e.g., institutional investors) for producing high-quality ratings, the agencies are paid by issuers of fixed income securities. 2 The conflict of interest hypothesis thus stipulates that rating agencies give more favorable ratings to large issuers because they bring, and could potentially take away, more business. 1 Abundant evidence shows credit ratings contain information not imbedded in prices for corporate bonds, and Jorion, Zhu, and Shi (2005) show that the impact of rating changes on stock prices becomes stronger after Reg-FD. Ratings are also shown to be an important determinant for corporate decisions such as capital structure (Kisgen, 2006). 2 Rating agencies have also been criticized for using models that tend to overestimate the likelihood of rising and high levels of housing prices, and thus underestimate the default risk of MBS securities (e.g., Coval, Jurek, and Stafford, 2009). Our focus is not on the accuracy of these rating models per se, but rather on whether and how the market prices MBS securities issued by large vs. small issuers differently due to the possible conflict of interest problem. 1

3 The countervailing factor for the conflict of interest problem is the reputation of rating agencies, an important asset in the market with a few dominant players. As recent theoretical work (e.g., Frenkel, 2010) shows, the balance of forces may have tilted toward rating inflation for large issuers in the new MBS market for much of the 2000s, as compared to the mature corporate bond market, for two reasons. First, the booming housing and MBS markets generate a significant new revenue source for rating agencies, and second, unlike the corporate bond market, a small number of large issuers of MBS brought many deals to the ratings agencies and thus they have much greater bargaining power than other issuers. Theoretical research (e.g., Bolton, Freixas, and Shapiro, 2009; Bar-Isaac and Shapiro, 2010) also shows that the perverse incentive of the rating agencies is the strongest during market booms because the benefits in terms of additional rating business generated (net of potential reputation loss) are the highest. These facts and arguments provide the basis of our empirical tests. We match a large sample of MBS tranches sold between 2000 and 2006 with information on the price, initial yield (at issuance), and rating history (from Moody s, S&P and Fitch) to their issuers. 3 Around sixty percent of the tranches are rated AAA, the highest possible grade, by each of the three rating agencies. We also obtain information on the characteristics of the tranches (e.g., size of principal amount, weighted average life, geographical distribution of the underlying mortgages) as well as those of the issuers (e.g., issuer rating at the issuance date). We compare tranches sold by large issuers vs. small issuers, where issuer size is based on the issuing institutions (one-year) lagged annual market shares. We also differentiate market boom years, during which the principal amount issued accounts for a large fraction of the total amount issued over the entire sample period, from other years. 3 Throughout most of our sample there were just four Nationally Recognized Statistical Ratings Organizations (NRSROs) Moody s, S&P, Fitch, and DBRS, who achieved NRSRO status in However, DBRS focused almost exclusively on the corporate bond market (Kisgen and Strahan, forthcoming). 2

4 In our first set of tests we examine whether large issuers are more likely to shop for better ratings. By its very nature, rating shopping tends to be hidden from view; yet, it ought to influence the distribution and information content of ratings that are revealed to investors (and thus observable in our dataset). Shoppers will tend to censor out pessimistic ratings, thus reducing the number of ratings observed empirically and, at the same time, reducing the likelihood of ratings disagreements. We argue that large issuers, due to greater bargaining power, are more apt to exert their influence and bargaining power when talking to different rating agencies. Consistent with this hypothesis, we find that tranches sold by large issuers are less likely to have multiple ratings (at issuance) than similar tranches sold by small issuers. Moreover, when these tranches do have multiple ratings, the ratings are more likely to be the same as compared to the ratings on similar tranches sold by small issuers. Together these results suggest that large issuers censor out low ratings after shopping their product to all the rating agencies, thus making it less likely for investors to observe ratings disagreement. After identifying ratings shopping as the main channel through which large issuers can influence the rating process, we next examine whether investors and the market recognize this incentive when they try to price these tranches. We first examine whether the market differentiates the ex ante credit quality of tranches sold by large vs. small issuers by comparing their initial yields (at issuance, over a benchmark). For AAA-rated tranches, the initial yield on tranches issued by large issuers is 7.4% higher than that of similar tranches issued by small issuers, suggesting that the market is skeptical of the quality of these tranches receiving the highest possible grade. For both AAA- and non-aaa rated tranches, the yield spread between otherwise similar tranches issued by large and small issuers is greater during the market booming period. This implies that the market is also aware that the potential conflict of interest problem facing rating agencies is more severe during booms, leading to further compromise in the rating process, and 3

5 accordingly demands a steeper (price) discount on these tranches. Consistent with our first set of results on ratings shopping, we also find that non-aaa rated tranches that receive a smaller number of ratings and those tranches with multiple ratings that are different have higher yield spreads. We also obtain a number of interesting results on how the market prices MBS tranches. For example, tranches with a greater fraction of their underlying mortgages originated from troubled states (Arizona, California, Florida, and Nevada) have higher yields. Among AAA-rated tranches, those with a more concentrated mortgage portfolio have lower yields than those with a more diversified mortgage portfolio. This is consistent with the predictions of Coval, Jurek, and Stafford (2009), who argue that more diversified MBS tranches, and especially AAA-rated tranches, should have more systemic risk and thus require a higher yield. In our final set of tests, we examine the ex post performance of these two groups of securities by looking at price changes between origination and April, Both AAA- and non- AAA rated tranches sold by larger issuers performed worse than similar tranches sold by smaller issuers. The underperformance is more pronounced for non-aaa rated tranches than AAA-rated ones prices for these tranches sold by large issuers drop 11.5% more than similar tranches sold by small issuers; and the differences in price drop between tranches sold by large and small issuers are concentrated during the market boom years of 2004 through Combining these results with results on initial yields, we conclude that the market incorporates concerns about the ratings process into ex ante pricing (yields) more for AAA-rated tranches than for non-aaa rated tranches. Since the market has already demanded a higher initial yield for AAA-rated tranches sold by large issuers, these tranches do not underperform similar tranches sold by small issuers ex post by a large margin. On the other hand, while investors do not demand a deep discount initially for non-aaa tranches sold by large issuers, these tranches perform significantly worse, ex post, than those issued by small issuers. 4

6 Our paper complements and contributes to the growing literature on the anatomy of housing and MBS markets as well as incentive problems of financial service industries. Prior work (e.g., Keys et al. 2009; Mian and Sufi, 2009; Loutskina and Strahan, 2009) has examined lending practices of banks and excessive credit supply as potential causes for the run-up in housing and related markets. There are several recent papers (e.g., Ashcraft, Goldsmith-Pinkham, and Vickrey, 2009; Benmelech and Dlugosz, 2009a, 2009b; Adelino, 2009; Griffin and Tang, 2009) empirically examining credit ratings in the broad structured finance (including MBS) markets. These papers find that while ratings provide useful information beyond what is incorporated in prices, they are not always accurate measures for default risk; nor are they a sufficient statistic for risk. For example, market measures such as yield spreads add incremental explanatory power beyond the ratings in forecasting defaults. Our paper is the first to test for incentive problems of rating agencies related to issuer size in the MBS markets. Griffin and Tang (2009) examine possible flaws of rating agencies internal models and Ashcraft et al. (2009) examine the pooling and tranching process of mortgages into MBS. By contrast, we take a valuation from outside approach and examine whether and how the market incorporates their concerns of the rating process into pricing by looking at ex ante and ex post measures (yields and price performance). While prior research has examined conflict of interest facing financial institutions such as investment banks (e.g., Kisgen, Qian, and Song, 2009), limited empirical research on rating agencies has focused on how conflict of interest affects the rating process in the corporate bond market (e.g., Becker and Milbourn, 2009). Our work thus contributes to the literature by showing that the conflict of interest problem can be a lot worse in the MBS market, the channel through which large issuers can affect the rating process and whether investors are aware of this problem in their pricing of securities. The rest of the paper is organized as follows. In Section II we briefly review the evolution of the MBS markets and discuss our hypotheses and explain how we construct our tests. We then 5

7 introduce our data on MBS securities in Section III and present results from our empirical tests in Section IV. We conclude in Section V. II. Overview of Credit Ratings and MBS Markets Prior research has documented that rating agencies play important roles in the traditional corporate bond market. Credit ratings are perhaps more important in the recently developed markets for structured finance products including MBS securities for several reasons. First, there are important differences in how cash flows and risk are evaluated between structured finance products and traditional fixed income products such as corporate bonds. The cash flows and risks of corporate bonds are generally tied to the performance and prospects of one company. By contrast, structured finance products involve a complicated securitization process, and in particular, pooling and tranching (multiple times) of a large number of credit-sensitive assets. For a fixed collateral pool (in the case of MBS these would be home mortgages), structured finance separates the pool into prioritized claims/tranches, which absorb losses from the underlying portfolio following seniority, and receive different ratings before being sold to investors. While securitization has revolutionized fixed income markets and brought billions of dollars of revenues to Wall Street, for many investors, even some institutional investors, this process can be opaque and filled with problems of asymmetric information and moral hazard. 4 To the extent that uninformed investors trust the rating agencies as the expert in objectively assessing the risks of these complicated securities, it is reasonable to assume that credit ratings, especially the coveted AAA rating, plays a much more important role in helping these investors make investment decisions than the case of corporate bonds, where independent research is more feasible. 4 See, e.g., Coval et al. (2009) for a review of structured finance, and Ashcraft and Schuermann (2008) for a review of potential problems of the securitization process. See Keys et al (2009) for evidence that securitization led to lax screening by lenders. 6

8 Second, for many institutional investors, especially those focusing on the fixed income markets and seeking high yield/return investment opportunities but constrained by the level of risk exposure (ratings), highly rated MBS tranches offer ideal investment instruments. The securitization process described above can produce many more AAA-rated tranches (among the population of all tranches) than the fraction of AAA-rated corporate bonds (among all corporate bonds, which is about 1 percent). The pooling and tranching process eliminates most of the idiosyncratic risk of underlying assets, while the remaining systemic risk (and this risk can be high if there is unexpected economy-wide shocks) leads to higher expected returns. For banks, broker dealers, and insurance companies, for example, credit ratings affected the amount of capital needed to hold in reserve for purchases of rated securities. Hence, the demand to buy structured products varied mechanically with the credit rating. Third, for rating agencies, the various types of new fixed income products provide tremendous new rating businesses beyond the traditional markets. In the case of Moody s, as the total volume of originations of subprime mortgages rose from $65 billion in the late 1990s to over $600 billion in 2006, Moody s profits tripled between 2002 and In 2006, the peak of the housing and related financial markets, according to their disclosure, 44 percent of Moody s revenues came from rating structured finance products, exceeding the 32 percent of revenues from rating corporate bonds. There is also direct evidence that rating agencies offer price discounts for large and frequent issuers of corporate bonds. 5 It is natural to assume that such practice also exists in dealing with large issuers of structured finance products including MBS. As pointed out above, an important difference between corporate bond and MBS markets is that the issuance of MBS 5 According to S&P s disclosure reports (including rating fee structure) in 2008, S&P stated that corporate issuers typically pay up to 4.25 basis points for most transactions and that the minimum fee is $67,500. In addition, S&P will consider alternative fee arrangements for large volume issuers and other companies that want multi-year ratings services agreements (Standard and Poor s 2008). Also see Becker and Milbourn (2009) for more details on the practice of rating agencies in the corporate bond market. 7

9 securities is highly concentrated among large issuers large financial institutions such as banks and investment banks as compared to large issuers of corporate bonds (i.e., large corporations). The concentration of the issuance of MBS securities among a few large players implies that large issuers have substantial bargaining power over rating agencies as they can bring, and certainly take away, rating businesses. Hence, as predicted by recent theory work (e.g., Frenkel, 2010), the tremendous new revenue source in the MBS market, along with significant bargaining power of large issuers, indicates that conflict of interest may be a more powerful force, relative to potential reputation loss of rating agencies, in shaping the process of assigning ratings in the MBS and other structured finance products markets. To summarize, given the unique nature of the MBS markets and their close relationships with rating agencies, it is reasonable to assume that the conflict of interest problem, present in all fixed income markets, is perhaps more pronounced in the MBS market. In addition, recent theories (e.g., Bolton, Freixas, and Shapiro, 2009; Bar-Isaac and Shapiro, 2010) argue that rating agencies incentive to grant large issuers unduly favorable ratings is stronger during market booming periods. This is because during these periods, the benefits of this practice, in terms of additional rating businesses generated, far outweigh the costs of potential reputational loss (and hence loss of future rating businesses), in part due to a lower likelihood of being caught (since a larger fraction of investors during these periods are uninformed). To test these hypotheses, we examine and compare two groups of MBS tranches issued during the period : those issued by the largest issuers (top 10% in terms of market share in a given year) and those issued by the rest of the (much smaller) issuers. Our main hypothesis is that credit rating agencies (Moody s, S&P, and Fitch) may have favored large issuers because these issuers bring in more rating businesses and revenues. Moreover, this incentive is stronger during market booming periods. In the context of structured finance and a given pool of mortgages, more 8

10 favorable ratings imply a greater fraction of financing in the highly-rated tranches (i.e. the AAA slice), which implies greater risk across all tranches within a deal. We also examine whether and when investors recognize this behavior. For example, investors may have initially failed to distinguish the credit quality of similarly-rated tranches based on issuer size. Later on, as the housing market began to unwind, investors may have begun to recognize the difference in these two groups and adjusted yields accordingly. To test our hypotheses, we conduct three sets of tests on a large sample of MBS tranches matched with characteristics of issuers. First, we study ratings shopping by issuers a main channel through which large issuers can affect the rating process. The hypothesis is that large issuers can put pressure on different rating agencies in order to obtain the best possible rating for their product. Moreover, they would drop lower ratings after shopping their product to all the rating agencies. This hypothesis would imply that tranches sold by large issuers tend to have fewer ratings (at issuance) than similar tranches sold by small issuers; when these tranches do have multiple ratings, the ratings are more likely to be the same as inferior rating(s) have been censored. If evidence supports the ratings-shopping hypothesis, we want to examine whether and when investors and the market recognize this incentive and the flawed rating process when they price tranches sold by large issuers vs. those sold by small issuers. We look at both an ex ante and an ex post measure. We first compare the initial yields (at issuance) on securities sold by large vs. small issuers. If the market believes securities sold by large issuers have higher credit risk, initial yields ought to be higher. In our final set of tests, we study the post-issuance performance of these two groups of securities by looking at their (cumulative) price changes between origination and April, If large issuers enjoy favorable ratings and the market does not fully price this in the initial price/yield, then securities they sell ought to perform worse than otherwise similar securities sold by small issuers. Taken together, these three sets of results should give us a much better idea regarding 9

11 just how the adverse incentive problem affects the quality of ratings during one of the worst crises in history. III. Data We begin the process of data compilation with the Securities Data Corporation (SDC) database, which provides a large sample of tranches of privately-issued (i.e. non GSE) MBS deals. For each deal, SDC provides the basic information on asset/collateral types (mortgage, credit card, auto loans, bonds, etc), the number of tranches, as well as information on the issuer(s) and bookrunner(s). For other deal and tranche characteristics, including initial and subsequent ratings and prices, principal amount, coupon type and rate, and maturity (weighted average life, and whether the tranche is paid off prior to April 2009), we rely on manually collected data from Bloomberg. Our sample includes MBS deals originated and issued in 2000 through Variable Construction and Summary Statistics Table 1, Panel A describes the whole sample. We obtain ratings from the largest three rating agencies, Moody s, S&P, and Fitch. There are more tranches rated by S&P than Moody s and Fitch in our sample, and each of the three agencies rate around 60% of all the tranches AAA. Our main independent variable of interest, market share of issuer, is calculated as the number of tranches sold by an issuer over the total number of tranches sold in the previous year (using alternative measures of issuer market share based on the principal amounts yields very similar results). We denote market boom years through the variable HOT, defined as the fraction of total principal amount of tranches issued in a given year over the total amount of all years. We are interested in testing whether the effect of issuer size changes when markets boom, so we will introduce the interaction variable, market share of issuer * HOT, into some of our specifications. 10

12 We build a number of control variables to capture characteristics of a tranche and its underlying collateral that may be correlated with outcomes. Principal amount equals the size of the tranche; its distribution is highly skewed, with the mean $61 million and median only $12 million. Weighted average life, equal to the expected timing of payments of principal of a tranche, is also skewed with the mean 5.8 years. 6 Fraction of collateral in troubled states equals the fraction of collateral originated in Arizona, California, Florida, and Nevada. It measures the degree of exposure to areas that experienced the highest rise leading up to the crisis followed by the largest drop during the crisis. 7 HHI of Collateral equals the sum of the squared shares of the collateral within a deal across each of the top five states (with the largest amount of mortgages), with the aggregation of the other states as the sixth category. Initial Rating equals a numerical score based on the average of the ratings a tranche received at issuance: we set AAA = 1, AA = 2, and so on; hence a higher score implies a worse rating. Similarly, we construct the issuer rating equal to the same numerical score for the issuer at the issuance date. (Issuer rating is only available for a subset of our sample; hence we estimate models with and without this variable.) Last, we construct Same Originator Servicer, equal to a dummy set to 1 if the originator and the servicer of the deal are owned by the same firm and 0 otherwise. (Same Originator Servicer is currently available for a subset of our data. We are in the process of collecting for this variable for the remaining tranches.) As indicated earlier, we have two sets of market-based measures that serve as our main measures of ex ante risk and ex post performance. Initial Yield Spread equals the difference (in basis points) between the initial coupon rate on a tranche and the yield of a Treasury security whose maturity is closest to the tranche s weighted average life. Price change equals the percentage 6 Note that this is not the same as duration that measures the weighted-average time to maturity based on the relative present values of cash flows as weights (see, e.g., Ch. 27 of Saunders and Cornett, 2008, for more details). 7 We realize that the importance of this variable may be obvious only in hindsight, although some analysts were concerned about overheated regional markets in real time; nevertheless, all of our key findings are robust to the exclusion of this variable from our models. 11

13 change in the price of an MBS tranche between issuance and April 2009 (or its payoff date). About 45% of the 9,299 tranches that we have information on pricing history are paid off early and before the crisis, and so the median price drop is only 0.8% while the mean drop is about 15%. 8 We also have two variables to capture ratings shopping the number of ratings and an indicator for ratings disagreement (=1 if there is some disagreement; 0 otherwise). The majority of tranches receive two (71%) or three ratings (14%), while about 15% of the tranches have only one rating. Among tranches with two or three ratings, we observe disagreement about 13% of the time. Unconditional Comparisons: Large v. Small Issuers Panel B of Table 1 sorts the tranches by cohorts based on issuance years and also based on issuer size. Big issuers are those with their market shares in the top 10% among all issuers (of a given year), and Small issuers refer to all the rest of the issuers (in that year). Not surprisingly, the volume of tranches, in terms of principal amount, is much greater during the housing market boom of For our empirical tests below we compare the characteristics of the two groups of MBS tranches issued by large vs. small issuers across this boom period vs. the earlier sample period ( ). We conduct these comparisons by interacting market share of issuers with the (continuous) variable HOT as defined in Panel A. We report results excluding the tranches issued in 2007, as the housing and MBS markets clearly entered into a new regime as compared to the previous booming period. 9 8 Comparing the subsample of tranches with pricing information with the whole sample, we can see that large tranches (principal amount) are more likely to have price information from Bloomberg, which reports prices as the mid-quote (bid-ask) from security dealers. 9 According to the financial crisis timeline from the Federal Reserve Bank in St. Louis, in February 2007, Freddie Mac announces that it will no longer buy the most risky subprime mortgage and mortgage-related securities; in April 2007, New Century Financial Corp., a leading subprime mortgage lender, files for Chapter 11 bankruptcy; in June 2007, S&P and Moody downgrade over 100 bonds backed by second-lien subprime mortgages, and Bear Stearns informs investors that it is suspending redemptions from its High-Grade Structured Credit Strategies Enhanced Leverage Fund. All of these events suggest that the housing and MBS markets began to deteriorate in early When we include the 2007 observations in pooled regressions we obtain qualitatively similar results. 12

14 From Panel B, tranches sold by small issuers appear to be larger in size and shorter in terms of weighted average life, which tend to be safer, than those sold by large issuers. Tranches sold by small issuers also have less exposure to troubled states and are better diversified (lower HHI). The numerical values of ratings indicate that tranches sold by small issuers receive worse ratings (e.g., initial rating has a higher mean and median) than those from large issuers, especially during the boom years of On the other hand, small issuers themselves tend to have slightly better ratings than large issuers at the issuance date. Tranches from small issuers are also less likely to have only one rating and more likely to have ratings from all three agencies than tranches sold by large issuers. Perhaps not surprisingly, there is much greater rating disagreement among tranches with multiple ratings during the boom years, given the large volume sold in this period. But the multiple ratings (most with two ratings) of tranches sold by large issuers are less likely to differ from each other (difference is statistically significant at 1%). For example, during tranches sold by small issuers received different ratings 20.6% of the time, compared to just 12.4% of the time for large-issuer tranches. Finally, large issuers are more likely to act as both originator (issuer) and servicer, who collect interest payments after issuance. Small issuers, on the other hand, are more likely to get another institution to act as servicer. This difference may in part reflect economies of scale at large mortgage banks such as Washington Mutual (WaMu). However, servicers may be unwilling to accept their role for tranches with high default risks; thus, having a different servicer from issuer may provide a check and balance system when issuing the security. Overall, the preliminary evidence from Panel B appears to indicate that the quality of tranches issued by small issuers is better than those sold by large issuers during both subsample periods, despite receiving lower ratings. The evidence is also consistent with the hypotheses that large issuers are more like to shop for better ratings, and rating agencies grant them more favorable 13

15 ratings due to their conflict of interest and the bargaining power of the large issuers. Panel C of Table 1 presents results from simple, pair-wise correlation tests. Consistent with evidence from Panel B, market share of issuer is negatively correlated with issue size, positively correlated with expected life, and positively correlated with collateral concentration. In addition, issuer size is negatively correlated with the number of initial ratings (of a tranche) and rating disagreement. Large issuers also tend to have worse ratings (correlation coefficient is 0.28) but tranches sold by these issuers tend to receive better ratings. Finally, tranches that are larger in size, shorter in expected life, tranches with a smaller number of ratings and without ratings disagreement tend to have better initial ratings. Table 2 reports the top five issuers in each year of our sample period. The ranking for an institution in a given year is based on the number of deals issued during the year and information collected by SDC. We also derive issuer rankings based on the dollar amount issued in each year and this alternative procedure yields very similar rankings (in each year). While the list of top five issuers changes over time, most if not all institutions on the lists are the well-known, largest institutions involved in various aspects of housing and subprime lending. 10 Interestingly, each of the top five issuers in 2006, Countrywide, GM (through its finance arm GMAC), Bear Stearns, Lehman Brothers, and IndyMac, failed during the ensuing crisis! The bottom row of the table illustrates that the MBS market is highly concentrated among large issuers, in that the top five issuers account for 37.7% to 47.1% of all the newly issued securities each year over our sample period. As discussed above, the dominance of large issuers implies that they have considerable bargaining power over rating agencies as they can bring and take away tremendous amount of rating business. 10 We also rank bookrunners, or lead underwriters of the MBS securities, in each year. This list reflects the largest underwriters of structured finance products during this period, and overlaps with the list of largest issuers. We find (not reported) that the impact of ratings on the performance of tranches mostly comes through large issuers, not bookrunners. 14

16 IV. Results We conduct three sets of tests relating issuer size and market condition to: 1) ratings shopping; 2) ex ante initial yield; and, 3) ex post price change. The key explanatory variable is the market share of an issuer and its interaction with HOT. As controls, we include the log of the principal amount and the log of weighted average life of the tranche, which measures the prepayment risk and the expected timing of receiving payments of the principal. We also include the fraction of collateral in troubled states and collateral concentration (HHI). We estimate our performance models (i.e. yield and price change models) separately for the AAA-rated tranches and non-aaa rated; for the non-aaa subsamples, we control for the tranche rating averaged across the agencies that rated the tranche. We also estimated models with and without issuer ratings (since we lose data when we include this variable). In all of our tests, we include cohort (i.e. issuance year) fixed effects, and we cluster for all tranches issued by the same issuer in a given year to build standard errors. Note that by including the cohort-year effects we absorb the direct effect of HOT, which has only time variant but no cross-sectional variation, so we only report its interaction with issuer size. Ratings Shopping Ratings shopping, while of necessity hidden from view, ought to influence the distribution of ratings that are revealed to investors and thus observable in our dataset. Shopping implies that pessimistic rating(s) will not be revealed to market participants; such ratings will be censored because the issuer will not be willing to pay for them, thus leaving only the more optimistic rating(s). Thus, shopping in general should reduce the number of ratings; it ought to increase the generosity or optimism of ratings from the perspective of the issuer; and, when multiple ratings are evident, it should increase the probability of agreement. If larger issuers have more bargaining power, the payoff to ratings shopping should be greater for them than for smaller issuers. This 15

17 would imply two things: first, we ought to see a positive relationship between the prevalence of ratings shopping and issuer size. Hence, we tests how the number of ratings (Table 3) and the level of ratings disagreement (Table 4) vary with issuer size. Second, the threat of shopping should be more potent for larger issuers. So, even conditional on the level of ratings shopping (proxied by the number of ratings and the level of ratings disagreement), we ought to observe that larger issuers receive more favorable ratings (see below). To test these ideas, we regress the number of ratings (Table 3) and an indicator for ratings disagreement (Table 4) on Market share of issuer, HOT, and the set of controls outlined earlier. For the number of ratings, we estimate both OLS and Ordered Probits (because the dependent variable ranges from 1 to 3). For the disagreement indicator, we first condition the sample on having at least two ratings. We then report both a linear probability model (OLS) as well as a Probit. In this second model, we also control for the number of initial ratings, which equals either 2 or 3 in this sample, to remove any mechanical relationship between the number of ratings and the probability of disagreement. We find, first, consistent evidence that the expected number of ratings declines with issuer size. This result is statistically significant in both the OLS and the Ordered Probit models, and the result holds in models with and without issuer ratings. We do not find, however, that the effect differs with market conditions that is, the interaction between HOT and issuer market share equals zero (columns 2, 4, 6 and 8). (Note that the direct effect of market conditions is absorbed by the cohort-year fixed effects.) Second, we find that when there is more than one rating, disagreement decreases with issuer size. Unconditionally, the likelihood of disagreement equals 9.44% for large issuers compared to 17.47% for small issuers (Table 1). This increase is similar in magnitude to what we would infer 16

18 from the OLS regressions in columns 1 and 3 of Table 4 comparing an issuer with a 10% market share with one having a market share below 1%. Thus, MBS deals sold by large issuers have fewer ratings that are more similar. Together these two facts point to ratings shopping large issuers with greater market power shop more than small issuers because their bargaining power is greater. Yield at Issuance As noted, ratings shopping changes not only the distribution of ratings revealed to investors but also affects the information content of ratings. If larger issuers shop more due to stronger bargaining power, ratings ought to be more favorable for them even after controlling for the number of ratings and the level of ratings disagreement. To test this idea, we now compare how initial yields vary with issuer size. Since most of the securities are priced and sold at par, initial yield gauges the market s assessment of ex ante credit quality (that is, risk). Figure 1 presents preliminary evidence on the comparison of initial yield spreads for tranches sold by large vs. small issuers. The yield spread is calculated at the issuance (with initial price) date. Specifically, for a tranche with a floating coupon rate, yield spread is defined as the fixed mark-up, in basis points (bps), over the reference rate specified at issuance (e.g. the 1-month LIBOR rate). For a tranche with a fixed or variable coupon rate, yield spread is defined as the difference between the initial coupon rate and the yield of a corresponding Treasury security whose maturity is closest to the tranche s weighted average life. Tranches with all ratings are again sorted and grouped by their issuance year (cohort), and we plot the median initial yield spread for each cohort of the two groups of tranches during Figure 1 shows that yields on tranches sold by large issuers consistently exceed yields brought by small issuers by about 20 basis points. Table 5 tests whether the patterns in Figure 1 hold up after controlling for ratings and other 17

19 deal characteristics. The dependent variable equals the natural log of one plus the yield spread calculated at issuance (with initial price) date. In all the regressions, we control for the same set of characteristics as before, plus we control for the distribution of ratings (i.e. the level of the rating, the number of ratings, and an indicator for ratings disagreement). We split the sample into AAArated tranches and all non-aaa rated tranches, and we also include dummy variables for coupon types (floating, fixed, or variable). For AAA-rated tranches, the yield on tranches sold by large issuers is on average higher than that on tranches sold by small issuers. The coefficient from the baseline model (column 1) suggests that the yield would be about 7.4% higher for an issuer with 10% market share relative to a very small issuer. This effect is a bit smaller in magnitude to what we would estimate from Figure 2 (20 basis points): since the average spread is 147 basis points (recall Table 1), a 7.4% increase equals about 10 basis points. Also in contrast to the Figure, the effect is stronger during boom years (columns 2 & 4); at the 75 th percentile of HOT, the coefficients would suggest that largeissuer tranches have yields about 13.5% higher than small-issuer tranches. We also find a negative coefficient on the number of ratings, although this effect is not statistically significant. (Ratings disagreement is undefined for this sample because we only include tranches rated AAA by all agencies that rated the tranche.) For the non-aaa rated tranches, we find a very strong effect of ratings shopping particularly rating disagreement as well as a significant residual effect of issuer size that becomes more evident during boom years. The coefficients (columns 5-8) suggest that yields are 16% higher when the agencies disagree over the rating, which emphasizes the incentive for issuers to avoid revealing such disagreement to the market. Given this incentive, and given our earlier result that large issuers are less likely to sell deals over which the ratings diverge, it should not be surprising that the market prices the size of the issuer. Specifically, we find a significant interaction between 18

20 our HOT market variable and the market share of issuer (columns 6 & 8). At the 75 th percentile of HOT, the coefficient suggests that the pricing of large-issuer tranches exceeds the pricing of smallissuer tranches by about 5%. Moreover, the significantly negative sign of the number of ratings suggests that the market sees through the issuers incentive to censor bad ratings and thus demands a higher yield for tranches with fewer ratings. Finally, we also obtain a number of interesting results on how the market prices certain characteristics of MBS tranches. For example, tranches with a greater fraction of their underlying mortgages originated from troubled states (Arizona, California, Florida, and Nevada) have higher yields, and this effect is stronger for AAA-rated tranches. Interestingly, we also find that betterdiversified deals, as measured by the cross-state HHI, have higher yields. This result supports the model of Coval, Jurak, and Stafford (2009), who show that AAA-rated deals with a high degree of diversification act like economic catastrophe bonds that would default only under dire economic scenarios. Thus, such bonds must offer high yields to compensate investors for their high level of systematic risk. Ex Post Price Performance Figure 2 presents some simple, unconditional graphical evidence on our third test price change after issuance for the two groups of securities. Once again, for all the tranches, the initial price is set at par $100 per $100 face value, or very close to $ We group tranches by their issuance year (cohort) and then plot the median cumulative price change for all tranches from the first month after issuance until April 2009 (or the last reported price). The figure compares price changes for tranches sold by Big Issuers (defined as those in the top 10% in a given year) versus 11 As indicated earlier, about 45% of the 9,299 tranches that we have information on pricing history are paid off early and before the crisis. Once they are paid off, the ratings are withdrawn and reported price series stop. Note that these bonds do not experience bankruptcy when the underlying assets become distressed due to the special legal status of the Special Purpose Vehicles. Instead, actual and expected future cash flows fall, leading to a decline in the price. 19

21 those sold by other ( Small Issuers ). Prices in all the cohorts from both figures remain more or less flat during the first few years after issuance (except for the 2007 cohort), but begin to drop early in From Figure 2, prices of tranches issued during the market booming period of 2004, 2005, and 2006 and by large issuers dropped more during than those issued by small issuers (and in the same cohort). The gap is particularly large for the 2006 cohort, where prices dropped by 60% for the large-issuer bonds vs. 40% for the small-issuer bonds. Table 6 uses regressions to test whether the patterns in Figure 2 continue to hold after controlling for deal/tranche characteristics and other factors that may affect ex post performance. The dependent variable is price change, for which we have one observation per tranche, calculated as the percentage change between the price during the first month after issuance and the final price as of April, 2009 (if available) or the last available monthly price otherwise. As noted earlier, the sample is considerably smaller than our sample on initial yields (Table 5) because Bloomberg only provides a pricing history on some of the tranches. As before, we control for log of total principal with the tranche and the weighted average life of the collateral, the number of ratings and a ratings disagreement indicator, the fraction of collateral in troubled states, the concentration across states (HHI), the rating of the tranche and the rating of the issuer, as well as year fixed effects. We also include an indicator equal to one if the issuer also acts as the servicer of the deal. 12 As in all of the models, we cluster across all tranches within a common cohort-year/issuer. As in Table 5, Table 6 again separates results into the AAA-rated tranches (columns 1-4) vs. all other tranches (columns 5-8). We find a negative and significant impact of issuer size for both set of tranches, with larger effects for the non-aaa rated tranches. This latter comparison makes sense because the lower-rated tranches are more sensitive to changes in collateral value and default 12 The indicator for originator to equal servicer is currently only available for a subset of our data. We will add this variable to the initial yield analysis in a future draft. As of this writing we are continuing the hand-collection of this item. 20

22 patterns in the underlying mortgages. We also find that the impact of issuer size is greater during the boom years, as suggested by the significant interaction between HOT and market share of issuer (columns 2 & 4, 6 & 8). This interaction reflects the patterns evident across cohort years in Figure 2. The coefficients from the baseline models suggest that tranches sold by large issuers fell by about 3 to 6 percentage points more than those sold by small issuers in the AAA-rated market, and 7 to 11.5 percentage points more in the lower-rated tranches. The impact of issuer size is especially striking during the boom years. At the 75 th percentile of HOT, the coefficients in column 8 suggest that large-issuer tranches fell by 17% more than small-issuer tranches. The results also suggest that tranches with more ratings performed better, as did deals in which the servicing rights that were sold to third parties (i.e., when the Same Originator Servicer indicator equals 0). This last result may suggest that the sale of servicing rights is more difficult for deals that may be suspect in quality to sophisticated investors. Combining the results on price performance with results on initial yields, we conclude that the market incorporates concerns about the rating process into ex ante pricing (yields) more for AAA-rated tranches than for non-aaa rated tranches. Since the market has already demanded a higher initial yield for AAA-rated tranches sold by large issuers (columns 1-4 in Table 5), these tranches do not underperform similar tranches sold by small issuers ex post by a large margin. On the other hand, while investors do not demand a deep discount initially for non-aaa tranches sold by large issuers, these tranches perform significantly worse, ex post, than those issued by small issuers (columns 5-8 in Table 6). Robustness Checks We have estimated three sets of unreported robustness tests. First, one might be concerned that our results reflect unobserved issuer differences that may be correlated with size and thus could explain links from issuer size to outcomes. Notice that even if this were true, the rating agencies 21

23 ought to capture heterogeneity across issuers and account for these differences in assigning ratings; so the alternative hypothesis would suggest a lesser mistake but not mistake free by the rating agencies when assigning ratings. To try to rule this out, we have estimated our models with issuerlevel fixed effects. These effects sweep out time-invariant common factors within an issuer and thus reduce concerns about unobserved heterogeneity. The problem with this strategy is that much of the variation in issuer size is constant over time, and the issuer effects will take out this variation. (Recall Table 2: Countrywide and GM appear in the Top 5 in every year.) When we do control for issuer effects, we obtain stronger results on the negative impact of market share of issuer on price performance. The results on ratings shopping and ex ante pricing are weaker with fixed effects, however, although sign patterns remain the same in most instances. Second, we have also estimated our ex ante yield tests and ex post price performance tests separately for Moody s and S&P. The idea is to test whether the two largest rating agencies appear to behave differently in setting ratings. Perhaps one of the agencies serves the interest of large issuers and the other does not. These results are quite similar across the two agencies, and both samples are consistent with the idea that large issuers get more favorable ratings on the deals that they sell. This should not be too surprising given our results on ratings shopping. Shopping creates pressure on the agencies not only to assign generous ratings, but it also creates incentive for ratings to converge, thus eliminating differences in behavior across the agencies. Third, we have replaced the numerical ratings with indicator variables for each rating. This non-parametric strategy does not impose the restriction that the difference between rating notches remains constant as one moves across the rating distribution. The disadvantage of this approach is that we cannot test whether, for example, issuers with better ratings receive better pricing when they sell deals. In these models, we obtain similar effects of our variables of interest. 22