Credit Ratings and the Cost of Municipal Financing

Credit Ratings and the Cost of Municipal Financing Abstract A common belief held among researchers and policy makers is that regulatory reliance has inflated market demand for credit ratings, despite their decreasing informational value. Advances in information technology, coupled with reputation losses following the subprime crisis, renew the question of whether investors still rely on ratings to assess credit risk. Using Moody’s 2010 scale recalibration, which was unrelated to changing issuer fundamentals, we find that ratings still matter to investors and to issuers—apart from any regulatory implications. Our results commend improved disclosure to mitigate mechanistic reliance on ratings and inefficiencies due to rating standards that vary across asset classes. Received October 9, 2015; editorial decision June 7, 2017 by Editor Andrew Karolyi. The purpose of this paper is to test whether credit rating agencies (CRAs) remain relevant as information intermediaries in the modern information environment. There is a massive literature documenting correlation between credit ratings and securities prices; however, these papers commonly suffer an endogeneity problem.1 Specifically, it is difficult to determine whether investors respond directly to credit ratings, or if investors and CRAs merely observe and react to the same information about issuer fundamentals.2 Kliger and Sarig (2000) address this problem by showing that markets reacted when Moody’s Investors Service (Moody’s) added modifiers to its ratings scale in 1982. Much has changed since 1982. The speed and cost of information processing have exponentially increased and decreased, respectively, because of advancements in information technology and the advent of the internet. These developments have likely resulted in at least some investors who previously relied on CRAs to begin performing their own credit risk analyses. Even the less ambitious investors now have alternative sources of information—including market prices of credit default swaps (CDS)—that were not available in 1982. Further, both Moody’s and Standard and Poor’s (S&P) suffered significant loss of reputation capital as a result of the inaccurate ratings they produced in the run-up to the recent financial crisis. Although the ratings most relevant during the financial crisis were those of structured finance products, some question CRA viability more broadly. For example, during a post-crisis congressional hearing, Congressman Christopher Shays argued, “They have no brand, they have no credibility whatsoever. I can’t imagine any investor trusting them.”3 For these reasons, we revisit the question of whether credit rating agencies still matter. There are several more recent papers (reviewed in Section 1) indicating that credit ratings continue to have real economic effects. However, the evidence in these papers suggests that ratings now matter primarily (if not exclusively) due to their regulatory implications. Because ratings affect investment standards and capital requirements, they affect the value of securities to institutional investors, even if these investors do not rely on ratings to evaluate credit risk. The regulatory implications of ratings are particularly acute in markets dominated by regulated investors, including the more commonly studied corporate bond market. The unanswered research question we tackle here is whether investors continue to rely on credit ratings for information about credit risk, apart from the confounding effects of ratings’ regulatory implications. We find robust evidence that they do. We examine the impact of Moody’s recalibration of its municipal bond (muni) rating scale in the spring of 2010, after the dust of the recent financial crisis had settled. Historically, the criteria Moody’s used to assign municipal bond ratings was based on how likely the municipality was to require financial support from higher levels of government. These criteria were unique to municipal bonds; bonds in all other asset classes are rated according to their expected losses. When Moody’s recalibrated its ratings for municipal debt, it applied the expected loss criteria it uses for all other asset classes. This change in criteria resulted in upgrades of zero to four notches on $2.2 trillion of municipal debt. Importantly, Moody’s designed the recalibration to be uncorrelated with changes in issuer fundamentals. By changing the criteria, Moody’s provided a new perspective on the bonds’ credit risk, but the fundamentals of the bonds did not change. This event allows us to overcome the endogeneity challenge faced by most prior studies. Unlike the corporate bond market studied previously, the muni market is dominated by unregulated retail investors.4 By focusing on retail trades in the muni market, we avoid confounding regulatory effects. Because our setting involves rating changes that do not result from changes in fundamentals, and because the transactions in our sample are not executed by regulated investors, we are able to cleanly identify investor reliance on ratings to assess credit risk. An important feature of the recalibration is that not all munis were upgraded. Municipal issuers that were already “well calibrated” to the global scale for other asset classes serve as our control group in a difference-in-differences framework. These bonds provide reasonable benchmarks for how the prices on upgraded bonds would have behaved in the absence of Moody’s recalibration. Because credit ratings on insured bonds reflect the credit quality of the insurer, we include only uninsured bonds in our analyses. (Roughly 60% of the $\$$ 2.2 trillion sample munis are uninsured.) Our sample consists of roughly equally-sized treatment and control groups: $\$$640 billion of uninsured munis experienced upgrades due to recalibration, and $\$$601 billion did not. We find robust evidence that investors reacted to this event. Controlling for bond characteristics and a host of fixed effects, we find that upgraded bonds experience a decrease in credit spreads of 19 to 33 basis points (bp) relative to non-upgraded bonds. This is the main result in the paper. This reliance is economically meaningful, and our results are robust to a wide variety of alternative specifications. Although the municipal bond market is a natural setting to test whether investors rely on ratings to assess risk, we take seriously the possibility that our results could still reflect some regulation-based demand. Although our secondary market results focus exclusively on retail-size trades, we impose as a robustness check an additional filter that removes retail transactions for issues with any level of holdings by insurance companies facing ratings-based capital charges. Results and conclusions are unchanged. We make use of various institutional features to further test for regulation-based demand. For example, most ratings-based regulations distinguish between broad rating categories rather than individual notches within those categories. We exploit this feature by comparing results among bonds with equal-sized upgrades that do and do not cross into new broad rating categories. These tests provide at most limited evidence of regulation-based demand. Some regulations also use a “lowest rating binds” criterion, which employs the lower of Moody’s and Standard & Poor’s (S&P) ratings. We exploit this feature by comparing the impact of the recalibration among upgraded bonds with ratings from S&P that remain, versus become, lower than ratings from Moody’s. If the rating from S&P was lower than the rating from Moody’s before the recalibration and remained lower after Moody’s upgrades, then the recalibration should have no regulatory bite. However, if the Moody’s rating leapfrogged the S&P rating (and thus the S&P rating becomes the lower rating), then this upgrade should have regulatory implications. However, we find no differential impact of the recalibration on these two groups of upgraded bonds. Overall, the cross-sectional analysis of upgrades based on their likely regulatory effects provides corroborating evidence that investors’ reaction to Moody’s recalibration reflects primarily a reliance on ratings to price risk rather than an increase in regulation-based demand. We explore several other alternative explanations for our results. For example, Harris and Piwowar (2006) show that municipal bond liquidity increases with credit quality. If investor perception of bonds’ credit quality increases when ratings are upgraded, then the price changes we observe could reflect lower liquidity premiums. A consequence of this hypothesis is that upgraded bonds should experience permanent increases in liquidity. This is not what we find. Although upgraded bonds’ trading volume increases immediately after the recalibration relative to non-upgraded bonds, this increase is transitory. We find that upgraded bonds’ trading volume in the period three to six months after the recalibration is statistically indistinguishable from that prior to the recalibration. This finding indicates that upgraded bonds do not experience permanent increases in liquidity, despite their permanently higher ratings. Another possible explanation for our results is that the control group, the non-upgraded bonds, coincidentally experienced a decrease in returns around the time of the recalibration. If this is true, then our results could still obtain even if the market did not bid up the prices of the upgraded bonds. We find no evidence of this effect. If anything, the returns of the non-upgraded group are slightly higher (although not significantly so) around the recalibration in comparison to their own returns in a 180-day period preceding the recalibration. Next, we examine the possibility that our results could reflect shifts in demand for particular levels of governments’ bonds around the recalibration. For example, state-level issuers received some of the largest upgrades. If investors happened to experience an increase in demand for state-level (as opposed to county, city, or other levels of government) bonds around the time of the recalibration, then our results could reflect that demand shift instead of a response to the rating changes. We address this possibility by including in our regressions issuer-level-of-government fixed effects that vary before and after the recalibration. Our results are fully robust. Finally, we consider the possibility that our results could reflect differential changes in the fundamentals of the upgraded and non-upgraded groups. Moody’s (2010) states that the recalibration does not reflect changes in credit risk and that any ratings under review prior to the recalibration remained under review. Still, we address this potential concern by comparing S&P’s ratings to Moody’s ratings around the time of the recalibration. If the recalibration did indeed reflect changes in fundamentals, then we should observe S&P eventually change its ratings in a pattern similar to Moody’s. We start by constructing ratings transition matrices for bonds rated by both Moody’s and S&P around the recalibration. We find little similarity in the shape of S&P’s and Moody’s transition matrices. We also examine a time series of the two raters’ average muni ratings. We observe a sharp increase in Moody’s ratings relative to S&P in 2010 that remains fully intact through the end of our data availability. Overall, although we cannot disprove the possibility that the upgraded and non-upgraded bonds’ fundamentals were changing differently around the time of the recalibration, we find no evidence to support this possibility. We turn next to the primary market to test whether market reliance on ratings has real economic effects. We conduct this analysis at the issuer level, sorting issuers by whether their outstanding bonds were upgraded as a result of Moody’s recalibration. Using a multivariate difference-in-differences approach, we find that spreads on new issues by upgraded issuers decrease by 15 to 22 bp, relative to the control group. This magnitude is comparable to what we find in the secondary market, and it indicates that our findings are economically meaningful. The product of $\$$ 640 billion (the face value of uninsured municipal debt upgraded during recalibration) and 15 bp (our most conservative estimate of the recalibration effect on offer yields) is $\$$960 million. This back-of-the-envelope calculation provides an estimate of aggregate excess interest paid annually (in 2010 dollars) by U.S. taxpayers due to Moody’s previous dual-class rating system. We also observe that upgraded issuers see a larger increase in issuance volume than non-upgraded issuers in the years following the recalibration. This finding demonstrates that ratings have real economic effects; lower borrowing costs increase municipal borrowing (and presumably increase municipal investment)5. We extend our primary market analysis to test whether investors rely more heavily on credit ratings when the amount and quality of alternative sources of information are low. For example, we employ the issuer level of government as a proxy for issuer size and opacity. Consistent with an information effect, we find the results of the recalibration are weakest among state-level issuers and strongest among cities. The recalibration effect is also stronger among issuers in states with more opaque accounting practices, in states identified as more corrupt, and among municipalities without ratings from S&P. Combined, these additional results indicate that ratings are more influential when investors have less alternative information. Overall, our contribution to the literature is original evidence that investors rely on credit ratings to assess risk, and that this reliance is greatest among opaque issuers for which investors lack alternative sources of information. Our results bring new evidence to bear regarding the classic question of whether and how security prices depend on credit ratings. The results also shed light on how investors process information in the municipal bond market, a multi-trillion-dollar market that is relatively opaque and is beginning to receive greater attention from researchers. 1. Institutional Background and Literature Review 1.1 Moody’s dual class ratings Moody’s uses its Global Scale when rating corporate bonds, sovereign debt, and structured finance products. These bonds are rated according to their expected losses. Expected loss is the product of probability of default and loss given default. Historically, Moody’s rated municipal bonds according to separate criteria. Moody’s assigned municipal ratings based on how likely a municipality is to require extraordinary support from a higher level of government in order to avoid default; Moody’s (2007, 2). This changed in the spring of 2010, when Moody’s recalibrated its municipal bond rating criteria to match that of the Global Scale. Moody’s (2010, 1) clarifies that the recalibration is intended to enhance the comparability of ratings across asset classes, not to indicate a change in credit quality: Our benchmarking $$\ldots$$ will result in an upward shift for most state and local government long-term municipal ratings by up to three notches. The degree of movement will be less for some sectors $$\ldots$$ which are largely already aligned with ratings on the global scale. Market participants should not view the recalibration of municipal ratings as ratings upgrades, but rather as a recalibration of the ratings to a different scale. $$\ldots$$ [The recalibration] does not reflect an improvement in credit quality or a change in our opinion. Importantly for our study, Moody’s (2010) indicates that any ratings under review for upgrade or downgrade prior to recalibration would remain under review—not lumped into these massive ratings changes. As such, our sample does not include any natural upgrades associated with improving issuer fundamentals that would contaminate the estimates generated by our tests. The timeline of Moody’s recalibration is as follows. In 2008, Moody’s revealed its intention to recalibrate its municipal bond rating scale.6 This announcement, however, contained no information regarding which bonds’ ratings would change or by how much. On March 16, 2010, Moody’s announced the particulars of the recalibration. On this date, Moody’s published a white paper containing its Primary Algorithm. The Primary Algorithm indicated which bonds’ ratings would be upgraded and by how much. The two characteristics that Moody’s used to recalibrate muni ratings were preexisting ratings and sectors. Moody’s categorizes bonds into four sectors. Figure A.1 in the Internet Appendix reproduces the Primary Algorithm. The actual recalibration was enacted on four dates. The first recalibration date was April 16, 2010, one month after the publication of the Primary Algorithm. The second, third, and fourth recalibration dates were April 23, May 1, and May 7, 2010, respectively. We provide details in Section 2 on the number and par values of bonds that were upgraded (and not upgraded) on each date. We test whether the market reacted to this event. If the market does not rely on Moody’s ratings to assess risk, then we should see no reaction. However, the change in criteria applied by Moody’s potentially allows the market to learn new information about the bonds. Consider the following analogy. Imagine a student who earns a B on a mostly qualitative exam. If the student soon after earns an A on a mostly quantitative exam in the same class, then the professor will update her opinion on the student’s aptitude. Yet the student’s aptitude does not change from one exam to the next. What changes are the criteria used to evaluate aptitude. In the same way, the market might react to Moody’s recalibration even though the bonds’ fundamentals do not change around the time of the recalibration. What changes are the criteria used to evaluate credit risk. Credit risk is not one-dimensional. If we find a response to the recalibration, then we can infer that investors updated their views about the bonds’ credit risk after Moody’s evaluated the bonds under the new criteria. Just as the professor takes a more favorable view of the student’s aptitude when using criteria that reward quantitative ability, the market might change its view of munis when Moody’s uses criteria based on expected losses. 1.2 Credit ratings and financial regulation Financial regulators have historically relied on credit ratings to establish capital requirements and prudent investment guidelines. This regulatory reliance on ratings dates to at least a ruling by the U.S. Comptroller of the Currency in 1931. Under Rule 5b-3 of the Investment Company Act, the United States Securities and Exchange Commission (SEC) treated Aaa-rated bonds as equivalent to Treasuries. Pension fund investment guidelines established by the Employee Retirement Income Security Act (ERISA) and bank capital requirements established by the Basel Committee on Banking Supervision have likewise been ratings-based. Under the Standardized Approach in Basel II, single A-rated munis carry a higher charge (20% risk weight) than Aa- or Aaa-rated munis (0% risk weight).7 Capital charges established by the National Association of Insurance Commissioners (NAIC) range from 3.39% to 19.5% for speculative grade (SG) bonds compared with 0.30% to 0.96% for investment grade (IG) bonds. This body of regulation creates incentives for regulated investors to respond to ratings, irrespective of whether they rely on ratings to evaluate risk. Indeed, the state of the credit ratings literature suggests that ratings matter primarily due to their regulatory implications. For example, Ellul, Jotikasthira, and Lundblad (2011) document fire sales by insurance companies when bonds in their portfolios are downgraded from IG to SG. The liquidity premiums associated with these sales indicate that the sales were attributable to capital charges rather than any information communicated by the downgrade. Becker and Ivashina (2015) further document regulatory arbitrage by insurance companies chasing yield in a ratings-based regulatory environment. Because the NAIC sets capital charges based on ratings, savvy insurance companies circumvent their regulatory capital charges by over-allocating capital to the bonds with the highest credit risk within a particular credit rating category. Stanton and Wallace (2013) similarly conclude that overinvestment in high-risk commercial mortgage-backed securities (CMBS) is attributable to such regulatory arbitrage in a ratings-based environment, and Cornaggia, Cornaggia, and Hund (2017) simulate potential regulatory arbitrage among banks subject to Basel II capital requirements. Opp, Opp, and Harris (2013) provide a formal model and conclude that regulatory implications of ratings are of first-order concern for marginal investors. In addition to official regulation, Chen et al. (2014) document the reliance on credit ratings in private investment mandates, asset management policies, and informal procedures that employ ratings to restrict holdings by mutual funds and investment advisors. Perhaps Ekins and Calabria (2012, 1) summarize the evolution of CRA relevance most succinctly in their conclusion: “Government regulatory use of credit ratings inflated the market demand for NRSRO ratings, despite the decreasing informational value of credit ratings.”8 Because of the regulation-based demand for ratings, recent studies that document market reaction to ratings changes or other real economic effects (e.g., Kisgen 2012; Almeida et al. 2017; and Begley 2015) cannot conclude that investors rely on ratings for information. Results from these studies may instead reflect changes in regulatory compliance costs. In fact, Almeida et al. (2017) specifically focus on ratings-based Basel II capital requirements in their study of the real effects of sovereign debt downgrades. One benefit of our setting is that the retail investors who dominate the muni market are subject to none of the aforementioned regulations. Any reaction among retail investors must therefore reflect an information effect. We are the first, to our knowledge, to directly test the extent to which market participants rely on ratings to assess credit risk apart from their need to manage regulatory capital charges and comply with other ratings-based regulations. 1.3 Modern relevance of CRAs as information intermediaries It is not obvious that modern investors rely on credit ratings to assess risk. For one thing, ratings are too coarse to fully reflect differences in credit quality across all rated securities; see Goel and Thakor (2015). Second, traditional ratings are designed to be stable over time and not to reflect real-time changes in credit quality (Moody’s 2006). Third, the conflicts of interest in the issuer-pays CRA compensation structure are well known.9 Fourth, improvements in information technology provide investors more-granular and more-timely credit risk metrics than traditional credit ratings (e.g., Cornaggia and Cornaggia 2013). Finally, the CRAs lost significant reputation capital due to their role in the subprime crisis. Finally, although Kliger and Sarig (2000) demonstrate a causal impact of ratings on security prices, these authors use an event from 1982. In 1982, investors lacked internet access, paid for long-distance telephone service, and received hard copies of Moody’s investment manuals in the mail. As such, it is unclear whether traditional CRAs matter as much as they did in 1982. We are the first, to our knowledge, to disentangle both the regulatory and endogeneity problems faced by existing literature in the modern information era. 1.4 Muni market information environment Another novel feature of our paper is the ability to exploit cross-sectional variation in the quality of information available to investors in municipal bonds. Other papers focus primarily on corporate securities that trade in relatively liquid, transparent markets and therefore may not be generalized to markets with less transparent issuers. Although the municipal finance market is large ( $\$$ 3.77 trillion in March 2014), this market is opaque and decentralized. Unlike corporations, state and local governments are not subject to the registration and reporting requirements of the SEC. Therefore, financial disclosure by municipalities is often less reliable, less comparable, and less timely than information released by corporations; however, the information quality varies widely across municipal issuers (see Ingram, Brooks, and Copeland 1983; Gore 2004; and Cuny 2016). This cross-sectional variation in the information environment across muni markets allows for additional tests of investor reliance on ratings to assess credit risk. 2. Data Collection and Sample Description 2.1 The recalibration event Our data consist of ratings from Moody’s and S&P, bond market transaction prices and volume from the Municipal Securities Rulemaking Board (MSRB), and issue/issuer characteristics from Ipreo. From Moody’s, we collect ratings data on every bond issue by a state or local government that had a “Change in Scale” rating action on April 16, April 23, May 1, or May 7 in 2010, as well as the ratings on all past and future issues by the same issuers. Because market perception of insured bonds reflects the credit quality of the monolines, we focus on uninsured bonds in our empirical analyses. Table 1 presents the number of issues and cumulative par value of recalibrated investment-grade munis.10 Panel A contains all bonds with a “Change in Scale” rating action. Panel B contains the uninsured bonds from which we draw our sample. Table 1 Number and par values of recalibrated bonds Panel A: All bonds All “Change in Scale” rating actions “Change in Scale” results in an upgrade “Change in Scale” results in no change in rating Recalibration date $$N$$ bonds Total par $$N$$ bonds Total par $$N$$ bonds Total par April 16, 2010 213,260 $\$$932.8 billion 190,144 $\$$812.5 billion 23,116 $\$$120.4 billion April 23, 2010 201,962 $\$$312.9 billion 186,946 $\$$281.2 billion 15,016 $\$$31.7 billion May 1, 2010 124,053 $\$$249.9 billion 108,046 $\$$199.4 billion 16,007 $\$$50.6 billion May 7, 2010 105,855 $\$$715.2 billion 24,221 $\$$67.4 billion 81,634 $\$$647.8 billion Sum 645,130 $\$$2,210.8 billion 509,357 $\$$1,360.5 billion 135,773 $\$$850.5 billion Panel B: Uninsured bonds April 16, 2010 90,621 $\$$566.3 billion 72,213 $\$$466.0 billion 18,408 $\$$100.3 billion April 23, 2010 55,891 $\$$96.8 billion 42,769 $\$$70.5 billion 13,122 $\$$26.4 billion May 1, 2010 54,021 $\$$117.2 billion 40,550 $\$$72.3 billion 13,471 $\$$44.9 billion May 7, 2010 65,510 $\$$461.2 billion 8,944 $\$$31.5 billion 56,566 $\$$429.6 billion Sum 266,043 $\$$1,241.5 billion 164,476 $\$$640.3 billion 101,567 $\$$601.2 billion Panel A: All bonds All “Change in Scale” rating actions “Change in Scale” results in an upgrade “Change in Scale” results in no change in rating Recalibration date $$N$$ bonds Total par $$N$$ bonds Total par $$N$$ bonds Total par April 16, 2010 213,260 $\$$932.8 billion 190,144 $\$$812.5 billion 23,116 $\$$120.4 billion April 23, 2010 201,962 $\$$312.9 billion 186,946 $\$$281.2 billion 15,016 $\$$31.7 billion May 1, 2010 124,053 $\$$249.9 billion 108,046 $\$$199.4 billion 16,007 $\$$50.6 billion May 7, 2010 105,855 $\$$715.2 billion 24,221 $\$$67.4 billion 81,634 $\$$647.8 billion Sum 645,130 $\$$2,210.8 billion 509,357 $\$$1,360.5 billion 135,773 $\$$850.5 billion Panel B: Uninsured bonds April 16, 2010 90,621 $\$$566.3 billion 72,213 $\$$466.0 billion 18,408 $\$$100.3 billion April 23, 2010 55,891 $\$$96.8 billion 42,769 $\$$70.5 billion 13,122 $\$$26.4 billion May 1, 2010 54,021 $\$$117.2 billion 40,550 $\$$72.3 billion 13,471 $\$$44.9 billion May 7, 2010 65,510 $\$$461.2 billion 8,944 $\$$31.5 billion 56,566 $\$$429.6 billion Sum 266,043 $\$$1,241.5 billion 164,476 $\$$640.3 billion 101,567 $\$$601.2 billion This table displays the number and total par value of municipal bonds for which Moody’s issued a “Change in Scale” rating action between April 16, 2010, and May 7, 2010. Panel A includes all bonds rated by Moody’s. Panel B restricts the sample in Panel A to uninsured bonds. We collect ratings data on bonds issued by state or local governments from Moody’s. Table 1 Number and par values of recalibrated bonds Panel A: All bonds All “Change in Scale” rating actions “Change in Scale” results in an upgrade “Change in Scale” results in no change in rating Recalibration date $$N$$ bonds Total par $$N$$ bonds Total par $$N$$ bonds Total par April 16, 2010 213,260 $\$$932.8 billion 190,144 $\$$812.5 billion 23,116 $\$$120.4 billion April 23, 2010 201,962 $\$$312.9 billion 186,946 $\$$281.2 billion 15,016 $\$$31.7 billion May 1, 2010 124,053 $\$$249.9 billion 108,046 $\$$199.4 billion 16,007 $\$$50.6 billion May 7, 2010 105,855 $\$$715.2 billion 24,221 $\$$67.4 billion 81,634 $\$$647.8 billion Sum 645,130 $\$$2,210.8 billion 509,357 $\$$1,360.5 billion 135,773 $\$$850.5 billion Panel B: Uninsured bonds April 16, 2010 90,621 $\$$566.3 billion 72,213 $\$$466.0 billion 18,408 $\$$100.3 billion April 23, 2010 55,891 $\$$96.8 billion 42,769 $\$$70.5 billion 13,122 $\$$26.4 billion May 1, 2010 54,021 $\$$117.2 billion 40,550 $\$$72.3 billion 13,471 $\$$44.9 billion May 7, 2010 65,510 $\$$461.2 billion 8,944 $\$$31.5 billion 56,566 $\$$429.6 billion Sum 266,043 $\$$1,241.5 billion 164,476 $\$$640.3 billion 101,567 $\$$601.2 billion Panel A: All bonds All “Change in Scale” rating actions “Change in Scale” results in an upgrade “Change in Scale” results in no change in rating Recalibration date $$N$$ bonds Total par $$N$$ bonds Total par $$N$$ bonds Total par April 16, 2010 213,260 $\$$932.8 billion 190,144 $\$$812.5 billion 23,116 $\$$120.4 billion April 23, 2010 201,962 $\$$312.9 billion 186,946 $\$$281.2 billion 15,016 $\$$31.7 billion May 1, 2010 124,053 $\$$249.9 billion 108,046 $\$$199.4 billion 16,007 $\$$50.6 billion May 7, 2010 105,855 $\$$715.2 billion 24,221 $\$$67.4 billion 81,634 $\$$647.8 billion Sum 645,130 $\$$2,210.8 billion 509,357 $\$$1,360.5 billion 135,773 $\$$850.5 billion Panel B: Uninsured bonds April 16, 2010 90,621 $\$$566.3 billion 72,213 $\$$466.0 billion 18,408 $\$$100.3 billion April 23, 2010 55,891 $\$$96.8 billion 42,769 $\$$70.5 billion 13,122 $\$$26.4 billion May 1, 2010 54,021 $\$$117.2 billion 40,550 $\$$72.3 billion 13,471 $\$$44.9 billion May 7, 2010 65,510 $\$$461.2 billion 8,944 $\$$31.5 billion 56,566 $\$$429.6 billion Sum 266,043 $\$$1,241.5 billion 164,476 $\$$640.3 billion 101,567 $\$$601.2 billion This table displays the number and total par value of municipal bonds for which Moody’s issued a “Change in Scale” rating action between April 16, 2010, and May 7, 2010. Panel A includes all bonds rated by Moody’s. Panel B restricts the sample in Panel A to uninsured bonds. We collect ratings data on bonds issued by state or local governments from Moody’s. The recalibration event of 2010 followed the monolines’ loss of their Aaa ratings in June 2008.11 We thus consider the extent to which the composition of the muni market (insured versus uninsured issues) may have changed around the time of the recalibration. We find that the resulting change in the proportion of insured/uninsured munis occurred more than two years prior to the recalibration event. We report this evidence in the Internet Appendix; see Figure A.1. In March 2010, Moody’s Primary Algorithm advertised a zero- to three-notch upgrade associated with the recalibration. Table 2 reports the actual migration matrix. As in Table 1, Panel A contains all bonds and Panel B contains only the uninsured bonds from which we draw our sample. The proportion of bonds upgraded varies by initial rating. Other than Aaa-rated bonds, which by definition cannot upgrade, no initial rating level retained more than 50% of its original bonds. Among uninsured bonds, 54% rated Aa1 upgraded to Aaa. No other bonds reached the Aaa level. Approximately 57% of bonds originally in the A categories migrated into Aa categories, and 64% of bonds in Baa categories migrated into the A range. Only 11 bonds were upgraded more than three notches (from A3 to Aa2). For the 9,714 issuers with multiple bonds outstanding at the time of recalibration, we examine (but do not tabulate) the within-issuer ratings distributions before and after the recalibration and find that these distributions remain largely intact.12 Table 2 Ratings migration matrix for Moody’s “Change in Scale” rating actions Panel A: All bonds Rating before recalibration Aaa Aa1 Aa2 Aa3 A1 A2 A3 Baa1 Baa2 Baa3 Sum Rating after recalibration Aaa 47,917 27,164 75,081 Aa1 20,412 70,503 45 50 91,010 Aa2 19,405 114,519 74,029 11 207,964 Aa3 11,802 23,403 79,338 36 114,579 A1 10,997 12,304 58,818 22,040 104,159 A2 9,930 5,901 1,570 12,246 29,647 A3 7,591 1,617 341 2,334 11,883 Baa1 2,707 159 2,764 5,630 Baa2 3,072 153 3,225 Baa3 1,952 1,952 Sum 47,917 47,576 89,908 126,366 108,479 101,572 72,357 27,934 15,818 7,203 645,130 Panel B: Uninsured bonds Rating after recalibration Aaa 46,828 20,404 67,232 Aa1 17,579 40,536 5 29 58,149 Aa2 14,204 43,229 14,620 11 72,064 Aa3 6,413 7,009 14,098 16 27,536 A1 4,321 3,560 9,838 4,525 22,244 A2 4,333 1,418 598 2,245 8,594 A3 3,575 449 87 1,042 5,153 Baa1 1,502 81 614 2,197 Baa2 1,758 74 1,832 Baa3 1,042 1,042 Sum 46,828 37,983 54,740 49,647 25,979 21,991 14,858 7,074 4,171 2,772 266,043 Panel A: All bonds Rating before recalibration Aaa Aa1 Aa2 Aa3 A1 A2 A3 Baa1 Baa2 Baa3 Sum Rating after recalibration Aaa 47,917 27,164 75,081 Aa1 20,412 70,503 45 50 91,010 Aa2 19,405 114,519 74,029 11 207,964 Aa3 11,802 23,403 79,338 36 114,579 A1 10,997 12,304 58,818 22,040 104,159 A2 9,930 5,901 1,570 12,246 29,647 A3 7,591 1,617 341 2,334 11,883 Baa1 2,707 159 2,764 5,630 Baa2 3,072 153 3,225 Baa3 1,952 1,952 Sum 47,917 47,576 89,908 126,366 108,479 101,572 72,357 27,934 15,818 7,203 645,130 Panel B: Uninsured bonds Rating after recalibration Aaa 46,828 20,404 67,232 Aa1 17,579 40,536 5 29 58,149 Aa2 14,204 43,229 14,620 11 72,064 Aa3 6,413 7,009 14,098 16 27,536 A1 4,321 3,560 9,838 4,525 22,244 A2 4,333 1,418 598 2,245 8,594 A3 3,575 449 87 1,042 5,153 Baa1 1,502 81 614 2,197 Baa2 1,758 74 1,832 Baa3 1,042 1,042 Sum 46,828 37,983 54,740 49,647 25,979 21,991 14,858 7,074 4,171 2,772 266,043 This table displays migration matrices on underlying ratings for municipal bonds for which Moody’s issued a “Change in Scale” rating action. Panel A includes all bonds rated by Moody’s. Panel B restricts the sample in Panel A to uninsured bonds. The horizontal axis represents bonds’ ratings before the first recalibration date (April 16, 2010), and the vertical axis represents the bonds’ ratings after the fourth and final recalibration date (May 7, 2010). We collect ratings data on bonds issued by state or local governments from Moody’s. Table 2 Ratings migration matrix for Moody’s “Change in Scale” rating actions Panel A: All bonds Rating before recalibration Aaa Aa1 Aa2 Aa3 A1 A2 A3 Baa1 Baa2 Baa3 Sum Rating after recalibration Aaa 47,917 27,164 75,081 Aa1 20,412 70,503 45 50 91,010 Aa2 19,405 114,519 74,029 11 207,964 Aa3 11,802 23,403 79,338 36 114,579 A1 10,997 12,304 58,818 22,040 104,159 A2 9,930 5,901 1,570 12,246 29,647 A3 7,591 1,617 341 2,334 11,883 Baa1 2,707 159 2,764 5,630 Baa2 3,072 153 3,225 Baa3 1,952 1,952 Sum 47,917 47,576 89,908 126,366 108,479 101,572 72,357 27,934 15,818 7,203 645,130 Panel B: Uninsured bonds Rating after recalibration Aaa 46,828 20,404 67,232 Aa1 17,579 40,536 5 29 58,149 Aa2 14,204 43,229 14,620 11 72,064 Aa3 6,413 7,009 14,098 16 27,536 A1 4,321 3,560 9,838 4,525 22,244 A2 4,333 1,418 598 2,245 8,594 A3 3,575 449 87 1,042 5,153 Baa1 1,502 81 614 2,197 Baa2 1,758 74 1,832 Baa3 1,042 1,042 Sum 46,828 37,983 54,740 49,647 25,979 21,991 14,858 7,074 4,171 2,772 266,043 Panel A: All bonds Rating before recalibration Aaa Aa1 Aa2 Aa3 A1 A2 A3 Baa1 Baa2 Baa3 Sum Rating after recalibration Aaa 47,917 27,164 75,081 Aa1 20,412 70,503 45 50 91,010 Aa2 19,405 114,519 74,029 11 207,964 Aa3 11,802 23,403 79,338 36 114,579 A1 10,997 12,304 58,818 22,040 104,159 A2 9,930 5,901 1,570 12,246 29,647 A3 7,591 1,617 341 2,334 11,883 Baa1 2,707 159 2,764 5,630 Baa2 3,072 153 3,225 Baa3 1,952 1,952 Sum 47,917 47,576 89,908 126,366 108,479 101,572 72,357 27,934 15,818 7,203 645,130 Panel B: Uninsured bonds Rating after recalibration Aaa 46,828 20,404 67,232 Aa1 17,579 40,536 5 29 58,149 Aa2 14,204 43,229 14,620 11 72,064 Aa3 6,413 7,009 14,098 16 27,536 A1 4,321 3,560 9,838 4,525 22,244 A2 4,333 1,418 598 2,245 8,594 A3 3,575 449 87 1,042 5,153 Baa1 1,502 81 614 2,197 Baa2 1,758 74 1,832 Baa3 1,042 1,042 Sum 46,828 37,983 54,740 49,647 25,979 21,991 14,858 7,074 4,171 2,772 266,043 This table displays migration matrices on underlying ratings for municipal bonds for which Moody’s issued a “Change in Scale” rating action. Panel A includes all bonds rated by Moody’s. Panel B restricts the sample in Panel A to uninsured bonds. The horizontal axis represents bonds’ ratings before the first recalibration date (April 16, 2010), and the vertical axis represents the bonds’ ratings after the fourth and final recalibration date (May 7, 2010). We collect ratings data on bonds issued by state or local governments from Moody’s. In Table 3, we track post-recalibration rating actions (upgrades, downgrades, or affirmations) for uninsured bonds through the first year after recalibration and, for completeness, through the end of our data collection. The summary statistics in this table shed light on the permanence of Moody’s recalibration. Recalibrated bonds in our sample are subsequently upgraded (downgraded) at most two (three) notches in the year following recalibration. The average recalibrated bond is downgraded 0.019 notches in the subsequent year. We see some evidence that bonds with larger upgrades during recalibration experience larger subsequent downgrades. Among the bonds upgraded three notches, the average recalibrated bond is downgraded 0.113 notches in the subsequent year. Because the vast majority of recalibrated bonds had no subsequent rating action (and thus do not appear in Table 3), the majority of subsequent rating actions are affirmations, and the magnitudes of any downgrades are small relative to the preceding upgrades due to recalibration, we conclude that the recalibration event was permanent. Table 3 Subsequent rating actions after recalibration Panel A: Rating differences $$N$$ bonds Mean SD Min 25% Median 75% Max All bonds with rating updates after recalibration 22,788 –0.021 0.217 $$-$$3 0 0 0 3 Bonds with rating updates within one year 20,469 –0.019 0.200 $$-$$3 0 0 0 2 $$\quad$$ after recalibration Panel B: Rating differences split by size of upgrade due to recalibration All bonds with rating updates after recalibration $$\quad$$ No change 9,077 0.006 0.172 $$-$$2 0 0 0 3 $$\quad$$ 1 notch 10,160 –0.018 0.188 $$-$$3 0 0 0 1 $$\quad$$ 2 notch 2,909 –0.098 0.329 $$-$$2 0 0 0 1 $$\quad$$ 3 notch 642 –0.112 0.391 $$-$$2 0 0 0 2 Bonds with rating updates within one year after recalibration $$\quad$$ No change 7,761 0.006 0.144 $$-$$2 0 0 0 2 $$\quad$$ 1 notch 9,391 –0.011 0.165 $$-$$3 0 0 0 1 $$\quad$$ 2 notch 2,688 –0.098 0.319 $$-$$2 0 0 0 1 $$\quad$$ 3 notch 629 –0.113 0.393 $$-$$2 0 0 0 2 Panel A: Rating differences $$N$$ bonds Mean SD Min 25% Median 75% Max All bonds with rating updates after recalibration 22,788 –0.021 0.217 $$-$$3 0 0 0 3 Bonds with rating updates within one year 20,469 –0.019 0.200 $$-$$3 0 0 0 2 $$\quad$$ after recalibration Panel B: Rating differences split by size of upgrade due to recalibration All bonds with rating updates after recalibration $$\quad$$ No change 9,077 0.006 0.172 $$-$$2 0 0 0 3 $$\quad$$ 1 notch 10,160 –0.018 0.188 $$-$$3 0 0 0 1 $$\quad$$ 2 notch 2,909 –0.098 0.329 $$-$$2 0 0 0 1 $$\quad$$ 3 notch 642 –0.112 0.391 $$-$$2 0 0 0 2 Bonds with rating updates within one year after recalibration $$\quad$$ No change 7,761 0.006 0.144 $$-$$2 0 0 0 2 $$\quad$$ 1 notch 9,391 –0.011 0.165 $$-$$3 0 0 0 1 $$\quad$$ 2 notch 2,688 –0.098 0.319 $$-$$2 0 0 0 1 $$\quad$$ 3 notch 629 –0.113 0.393 $$-$$2 0 0 0 2 This table displays summary statistics on the rating actions (upgrades, downgrades, or affirmations) subsequent to recalibration for uninsured municipal bonds. A bond’s rating must update again after a “Change in Scale” rating action for inclusion in this table. We report the difference in the new rating and the rating produced by the recalibration, measured in notches. A positive (negative) difference indicates a subsequent upgrade (downgrade). Zero difference indicates that the recalibrated rating was subsequently affirmed. The sample ends in October 2012. We translate Moody’s 21-point alphanumeric scale into a numeric scale such that Aaa $$=$$ 21, Aa1 $$=$$ 20, ..., C $$=$$ 1. We collect ratings data on bonds issued by state or local governments from Moody’s. Table 3 Subsequent rating actions after recalibration Panel A: Rating differences $$N$$ bonds Mean SD Min 25% Median 75% Max All bonds with rating updates after recalibration 22,788 –0.021 0.217 $$-$$3 0 0 0 3 Bonds with rating updates within one year 20,469 –0.019 0.200 $$-$$3 0 0 0 2 $$\quad$$ after recalibration Panel B: Rating differences split by size of upgrade due to recalibration All bonds with rating updates after recalibration $$\quad$$ No change 9,077 0.006 0.172 $$-$$2 0 0 0 3 $$\quad$$ 1 notch 10,160 –0.018 0.188 $$-$$3 0 0 0 1 $$\quad$$ 2 notch 2,909 –0.098 0.329 $$-$$2 0 0 0 1 $$\quad$$ 3 notch 642 –0.112 0.391 $$-$$2 0 0 0 2 Bonds with rating updates within one year after recalibration $$\quad$$ No change 7,761 0.006 0.144 $$-$$2 0 0 0 2 $$\quad$$ 1 notch 9,391 –0.011 0.165 $$-$$3 0 0 0 1 $$\quad$$ 2 notch 2,688 –0.098 0.319 $$-$$2 0 0 0 1 $$\quad$$ 3 notch 629 –0.113 0.393 $$-$$2 0 0 0 2 Panel A: Rating differences $$N$$ bonds Mean SD Min 25% Median 75% Max All bonds with rating updates after recalibration 22,788 –0.021 0.217 $$-$$3 0 0 0 3 Bonds with rating updates within one year 20,469 –0.019 0.200 $$-$$3 0 0 0 2 $$\quad$$ after recalibration Panel B: Rating differences split by size of upgrade due to recalibration All bonds with rating updates after recalibration $$\quad$$ No change 9,077 0.006 0.172 $$-$$2 0 0 0 3 $$\quad$$ 1 notch 10,160 –0.018 0.188 $$-$$3 0 0 0 1 $$\quad$$ 2 notch 2,909 –0.098 0.329 $$-$$2 0 0 0 1 $$\quad$$ 3 notch 642 –0.112 0.391 $$-$$2 0 0 0 2 Bonds with rating updates within one year after recalibration $$\quad$$ No change 7,761 0.006 0.144 $$-$$2 0 0 0 2 $$\quad$$ 1 notch 9,391 –0.011 0.165 $$-$$3 0 0 0 1 $$\quad$$ 2 notch 2,688 –0.098 0.319 $$-$$2 0 0 0 1 $$\quad$$ 3 notch 629 –0.113 0.393 $$-$$2 0 0 0 2 This table displays summary statistics on the rating actions (upgrades, downgrades, or affirmations) subsequent to recalibration for uninsured municipal bonds. A bond’s rating must update again after a “Change in Scale” rating action for inclusion in this table. We report the difference in the new rating and the rating produced by the recalibration, measured in notches. A positive (negative) difference indicates a subsequent upgrade (downgrade). Zero difference indicates that the recalibrated rating was subsequently affirmed. The sample ends in October 2012. We translate Moody’s 21-point alphanumeric scale into a numeric scale such that Aaa $$=$$ 21, Aa1 $$=$$ 20, ..., C $$=$$ 1. We collect ratings data on bonds issued by state or local governments from Moody’s. We further examine the permanence of Moody’s recalibration by testing whether new bonds issued after recalibration have the same, higher ratings generated by the recalibration.13 Of upgraded municipalities that issue bonds in both the year before and the year after their recalibrations, the average rating on their outstanding bonds changed from 17.8 ($$\approx$$Aa3) to 19.2 ($$\approx$$Aa2) as a result of the recalibration. The standard deviation of these ratings declined from 1.4 notches to 1.1 notches. These numbers appear in Table A.4 in the Internet Appendix. Importantly, the ratings on new bonds (issued in the year after recalibration) exhibit a similar average (19.1 $$\approx $$ Aa2) and standard deviation (1.9 notches) as the recalibrated bonds. We thus conclude that Moody’s applied its recalibrated ratings standards to new issues going forward after the recalibration, not just to bonds outstanding at that time. 2.2 Secondary market data We examine bond returns and trading volume in secondary markets around recalibration. We gather secondary market trading data from the MSRB Electronic Municipal Market Access (EMMA) database. The MSRB reports all trades of municipal bonds in the EMMA database, excluding both reversals and cancellations from historical data. The transaction data include prices, dollar volume, trade time, and whether the transaction was “Customer bought,” “Customer sold,” or “Inter-dealer trade.” The database does not distinguish between retail and institutional customers. In order to focus on retail investors, we restrict the sample based on trade size. Following Dick-Nielsen, Feldhutter, and Lando (2012), we attribute trade sizes less than or equal to $\$$ 100,000 to retail investors. (All results are robust to an alternative $\$$50,000 cutoff.) Because retail investors dominate the muni market, this restriction reduces the sample size by only 30%. For transactions involving a “Customer,” the yield is also reported. The reported yield is the lower of the yield-to-call and the yield-to-maturity. Municipal bond dealers include discount brokerages, full service brokerages, municipal advisers, and investment banks.14 Finally, we measure the percentage of each bond’s principal held by insurance companies at the time of Moody’s recalibration using Annual Fraternal, Life, Property, Separate Accounts, and Title & Health Data from the National Association of Insurance Commissioners (NAIC). 2.3 New issues data We gather data on new issues from the Ipreo i-Deal database including offer yield, sale date, maturity date, par value, coupon rate, as well as information on insurance and other support. For the same reason we exclude insured bonds in tests that use secondary market data, we exclude new issues that carry insurance in our tests based on primary market data. Panel B of Figure A.2 (Internet Appendix) displays the dollar volume of insured and uninsured municipal bond issues by month from April 2009 to April 2011. We focus on this time period for two reasons. First, our multivariate regressions, which we explain below, use data from this time period. Second, it spans the Build America Bonds (BAB) program, which ran from April 2009 to December 2010.15 Uninsured bonds dominate this market over this time period. 2.4 Matching treasury securities We measure credit spreads in several ways for completeness. We calculate spreads by subtracting duration-matched Treasury yields.16 For robustness, we calculate duration two ways. We estimate Spread to Treasury$$_{1}$$ with duration calculated using each bond’s time to maturity, regardless of whether the bond is callable. We estimate Spread to Treasury$$_{2}$$ with duration calculated by substituting the callable bonds’ call dates in lieu of their maturity dates. Because munis are generally tax exempt, we also examine after-tax spreads. Spread to after-tax Treasury$$_{1}$$ (Spread to after-tax Treasury$$_{2})$$ is similar to Spread to Treasury$$_{1}$$ (Spread to Treasury$$_{2})$$ except we use an after-tax yield on the duration-matched Treasury security. Because BABs are generally taxable, we employ after-tax yields of BABs (rather than raw yields) for both after-tax spread variables. We assume a marginal tax rate of 35%. 3. Empirical Results 3.1 Price impact of ratings recalibration We begin by studying the secondary market return behavior for outstanding munis. This analysis allows us to focus on a narrow window around recalibration dates, which should limit the influence of other contemporaneous events.17 We estimate market impact in two ways. First, for each recalibration date, we calculate CARs from 60 trading days prior to the recalibration date to 60 trading days after, using our control group (i.e., zero-notch “Change in Scale” upgrades) as the benchmark. Second, we examine changes in secondary market yields (and credit spreads), controlling for a host of issue characteristics in a multivariate difference-in-differences regression framework. We supplement this second set of tests with placebo analyses that use alternative event dates. We employ all observations for which we have sufficient data in each analysis. Details on sample construction and number of observations employed in each figure and table are available in the Internet Appendix (Table A.1). 3.1.1 CARs. In Figure 1, we plot CARs for portfolios of upgraded bonds with 95% confidence intervals around each recalibration date. April 16 is the cleanest event of the four, as there is no overlap with the other three dates until trading day $$+$$5. This date also features the most upgrades of the four, giving the tests the most power. We calculate returns by trade-weighting as described by Bessembinder et al. (2009).18 In particular, we calculate the daily price, $$P_{t}$$, as: \begin{align} P_{t}=\sum\limits_{i=1}^N {\frac{\it tradesize}{\sum\nolimits_{j=1}^N {\it tradesize}_{j}}{price}_{i}} \end{align} (1) on days with at least one trade, and the most recent daily price on days with no trades. After trade-weighting the prices, we define returns, $$R_{t}$$, as: \begin{align} R_{t}=\frac{P_{t+1}-P_{t}}{P_{t}} \end{align} (2) Figure 1 View largeDownload slide Cumulative abnormal returns around recalibration dates This figure displays cumulative abnormal returns of outstanding, uninsured municipal bonds with ratings recalibrated by Moody’s. We include only bonds purchased by customers (as opposed to dealers) on at least six days during 120-day trading windows centered on the bonds’ recalibration dates. Panel A (B, C, D) shows cumulative abnormal returns for bonds recalibrated on April 16 (April 23, May 1, May 7), 2010. In each panel, the solid vertical bar indicates March 16, 2010, the day the recalibration was announced, and the dashed vertical bar indicates the respective implementation date. We split the bonds by the size of the upgrade (one, two, or three notches). The comparison group in each panel consists of municipal bonds that were recalibrated on the indicated recalibration date, but not upgraded (i.e., the zero-notch group). Dotted lines represent 95% confidence intervals. We gather secondary market trading data from the MSRB Electronic Municipal Market Access (EMMA) database. Figure 1 View largeDownload slide Cumulative abnormal returns around recalibration dates This figure displays cumulative abnormal returns of outstanding, uninsured municipal bonds with ratings recalibrated by Moody’s. We include only bonds purchased by customers (as opposed to dealers) on at least six days during 120-day trading windows centered on the bonds’ recalibration dates. Panel A (B, C, D) shows cumulative abnormal returns for bonds recalibrated on April 16 (April 23, May 1, May 7), 2010. In each panel, the solid vertical bar indicates March 16, 2010, the day the recalibration was announced, and the dashed vertical bar indicates the respective implementation date. We split the bonds by the size of the upgrade (one, two, or three notches). The comparison group in each panel consists of municipal bonds that were recalibrated on the indicated recalibration date, but not upgraded (i.e., the zero-notch group). Dotted lines represent 95% confidence intervals. We gather secondary market trading data from the MSRB Electronic Municipal Market Access (EMMA) database. We form equal-weighted portfolios for zero-, one-, two-, and three-notch upgrades and calculate cumulative returns over the 120-trading-day window for each portfolio. We then calculate CARs for each upgraded portfolio by subtracting the zero-notch portfolio’s cumulative return. We observe significant positive CARs in Figure 1 for portfolios of upgraded bonds for each of the first three recalibration dates. In each panel, post-recalibration CARs are around 50 bp for munis upgraded one notch. This effect generally increases in the magnitude of the recalibration, with larger CARs for munis upgraded two or three notches. Only the fourth and final recalibration date shows no significant effect. This non-result is not surprising, given that most of the rating recalibrations on the fourth date were zero-notch changes, as shown in Table 1. Relative to the earlier three dates, only a small percentage of bonds received upgrades, thus limiting the power of the test. Figure 1 also reveals abnormal returns between the release of the Primary Algorithm and the first recalibration date. The goal of our paper is to establish whether investors rely on ratings to assess credit risk. The abnormal returns preceding the first recalibration date shed some light on the reasons why they might do so. Because CRAs are exempt from Regulation Fair Disclosure (FD), it is possible that credit ratings contain nonpublic information. However, even if CRAs merely compile public information, it may be rational for investors to rely on ratings—depending on the costs associated with analyzing this information. Figure 1 suggests that at least some portion of investor reliance reflects the costs associated with credit analysis. Finally, we consider that the drift in returns following the recalibration dates in Figure 1 may reflect illiquidity since these are averages across bonds and many do not trade on a given day. To focus more precisely on trades around the recalibrations, we perform univariate tests after calculating bond returns using the change in price from the last trade in the 30 days before Moody’s published the Primary Algorithm to the first trade in the 30 days after the bonds’ ratings change. (Results are robust to using the first trade on or after the recalibration dates.) As in Figure 1, we sort recalibrated bonds into equal-weighted portfolios based on upgrade magnitude. We then compare the average return to the portfolios of upgraded bonds to the control group (the portfolio of zero-notch upgrades). We perform this analysis across each of the four recalibration dates. As in Figure 1, we observe significant increases in returns of upgraded bonds relative to non-upgraded bonds for the first three recalibration dates.19 We do not tabulate these results to conserve space. 3.1.2 Multivariate regressions. We expand our analysis of credit rating price impact to a multivariate difference-in-differences setting in Table 4. This approach provides a more rigorous test of the patterns observed in Figure 1. These ordinary least squares (OLS) regressions employ yields and credit spreads as dependent variables, averaged across transactions by bond during either the 30 days before Moody’s published the Primary Algorithm or the 30 days after the bond’s rating was recalibrated. We include in the regressions uninsured bonds with trading in both periods. We omit data in the intermediate period between when Moody’s published the Primary Algorithm and when the bonds’ ratings were actually recalibrated, inclusive of the rating change dates. Graphically, we define the pre- and post-event periods for the four recalibration dates as seen in Figure 2. This approach allows us to sidestep the issue of whether investors responded to the publication of the Primary Algorithm versus the actual rating changes. Moody’s white papers (such as the publication containing the Primary Algorithm) are obscure sources of information for most retail investors. Investors may not know when Moody’s publishes a white paper, let alone digest it promptly upon release. They may only become aware of rating changes after they actually take place. Either way, omitting data from the intermediate period allows a clean comparison of prices before the recalibration was detailed to after it took effect. Figure 2 View largeDownload slide Event timeline Figure 2 View largeDownload slide Event timeline In Table 4, Raw yield is the lower of the yield-to-call and the yield-to-maturity. Each of the credit spread variables is defined in Section 2.4. Panel A captures the effect of recalibration on bond yields and credit spreads with Upgrade, an indicator variable taking a value of one if the issuer of the bond experienced an upgrade on any of its outstanding bonds during any of the recalibration events and zero if the issuer’s bonds experienced only zero-notch “Change in Scale” upgrades. Post recalibration is an indicator variable taking a value of one if the observation is from the 30 days after the bond’s recalibration date. This variable takes a value of zero if the observation is from the 30 days prior to the publication of Moody’s Primary Algorithm. Table 4 Secondary market regressions Panel A: Recalibration effect captured with upgrade indicator (upgrade) Raw yield Spread to Treasury$$_{{1}}$$ Spread to Treasury$$_{{2}}$$ Spread to after-tax Treasury$$_{{1}}$$ Spread to after-tax Treasury$$_{{2}}$$ Spread to after-tax Treasury$$_{{2}}$$ (1) (2) (3) (4) (5) (6) Post recalibration $$\times$$ Upgrade –0.06 –0.33 –0.28 –0.23 –0.19 –0.21 (0.07) (0.08)*** (0.08)*** (0.08)*** (0.07)*** (0.07)*** Post recalibration –0.13 0.27 0.15 0.13 0.06 0.07 (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.02)** (0.02)** Upgrade –0.16 –0.03 –0.33 –0.10 –0.37 –0.26 (0.17) (0.16) (0.22) (0.16) (0.23) (0.34) Par 0.07 0.08 0.06 0.08 0.06 0.14 (0.04)* (0.04)** (0.04) (0.04)** (0.04) (0.05)** Duration$$_{\mathrm{1}}$$ 0.18 –0.03 0.04 (0.02)*** (0.01)*** (0.01)*** Duration$$_{\mathrm{2}}$$ –0.13 –0.04 –0.04 (0.02)*** (0.02)** (0.04) Coupon 0.01 –0.01 –0.01 –0.01 –0.01 –0.02 (0.06) (0.04) (0.04) (0.05) (0.05) (0.07) Outstanding bonds –0.03 –0.05 0.00 –0.04 0.01 –0.02 (0.04) (0.04) (0.05) (0.04) (0.05) (0.06) GO 0.23 0.10 –0.11 0.14 –0.05 –0.09 (0.47) (0.35) (0.53) (0.40) (0.56) (0.72) BAB 1.47 1.28 1.25 –0.64 –0.74 –1.43 (0.17)*** (0.14)*** (0.19)*** (0.12)*** (0.14)*** (0.22)*** Callable 0.05 –0.24 0.59 –0.12 0.69 0.16 (0.13) (0.13)* (0.16)*** (0.11) (0.15)*** (0.50) Negotiated 0.00 0.05 –0.16 0.05 –0.14 0.00 (0.15) (0.12) (0.17) (0.12) (0.15) (0.17) Issue rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Yes Moody’s sector FE? Yes Yes Yes Yes Yes Yes NAIC filter? No No No No No Yes Adjusted $$R^{\mathrm{2}}$$ 0.48 0.45 0.46 0.41 0.47 0.49 $$N$$ 1,910 1,910 1,910 1,910 1,910 852 Post recalibration $$\times$$ Notches –0.07 –0.18 –0.16 –0.14 –0.12 –0.12 (0.03)** (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.07)*** Post recalibration –0.10 0.27 0.16 0.15 0.08 0.07 (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.02)** (0.02)** Panel B: Recalibration effect captured with upgrade magnitude (notches) Notches –0.09 0.00 –0.11 –0.05 –0.13 –0.12 (0.09) (0.08) (0.10) (0.08) (0.10) (0.13) Par 0.07 0.08 0.06 0.08 0.06 0.14 (0.04) (0.04)** (0.04) (0.04)** (0.04) (0.05)** Duration$$_{1}$$ 0.18 –0.03 0.04 (0.02)*** (0.01)*** (0.01)*** Duration$$_{2}$$ –0.13 –0.03 –0.04 (0.02)*** (0.02)** (0.04) Coupon 0.02 0.00 0.00 –0.01 0.00 –0.01 (0.07) (0.05) (0.05) (0.05) (0.05) (0.08) Outstanding bonds –0.03 –0.05 0.00 –0.04 0.00 –0.03 (0.04) (0.04) (0.06) (0.04) (0.06) (0.07) GO 0.24 0.10 –0.11 0.14 –0.05 –0.12 (0.47) (0.35) (0.51) (0.39) (0.55) (0.67) BAB 1.45 1.27 1.22 –0.66 –0.77 –1.48 (0.17)*** (0.14)*** (0.19)*** (0.13)*** (0.14)*** (0.24)*** Callable 0.04 –0.24 0.59 –0.13 0.69 0.16 (0.13) (0.13)* (0.16)*** (0.11) (0.15)*** (0.48) Negotiated 0.00 0.05 –0.18 0.04 –0.16 –0.03 (0.14) (0.12) (0.17) (0.12) (0.15) (0.18) Issue rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Yes Moody’s sector FE? Yes Yes Yes Yes Yes Yes NAIC filter? No No No No No Yes Adjusted $$R^{\mathrm{2}}$$ 0.58 0.45 0.46 0.42 0.58 0.49 $$N$$ 1,910 1,910 1,910 1,910 1,910 852 Panel A: Recalibration effect captured with upgrade indicator (upgrade) Raw yield Spread to Treasury$$_{{1}}$$ Spread to Treasury$$_{{2}}$$ Spread to after-tax Treasury$$_{{1}}$$ Spread to after-tax Treasury$$_{{2}}$$ Spread to after-tax Treasury$$_{{2}}$$ (1) (2) (3) (4) (5) (6) Post recalibration $$\times$$ Upgrade –0.06 –0.33 –0.28 –0.23 –0.19 –0.21 (0.07) (0.08)*** (0.08)*** (0.08)*** (0.07)*** (0.07)*** Post recalibration –0.13 0.27 0.15 0.13 0.06 0.07 (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.02)** (0.02)** Upgrade –0.16 –0.03 –0.33 –0.10 –0.37 –0.26 (0.17) (0.16) (0.22) (0.16) (0.23) (0.34) Par 0.07 0.08 0.06 0.08 0.06 0.14 (0.04)* (0.04)** (0.04) (0.04)** (0.04) (0.05)** Duration$$_{\mathrm{1}}$$ 0.18 –0.03 0.04 (0.02)*** (0.01)*** (0.01)*** Duration$$_{\mathrm{2}}$$ –0.13 –0.04 –0.04 (0.02)*** (0.02)** (0.04) Coupon 0.01 –0.01 –0.01 –0.01 –0.01 –0.02 (0.06) (0.04) (0.04) (0.05) (0.05) (0.07) Outstanding bonds –0.03 –0.05 0.00 –0.04 0.01 –0.02 (0.04) (0.04) (0.05) (0.04) (0.05) (0.06) GO 0.23 0.10 –0.11 0.14 –0.05 –0.09 (0.47) (0.35) (0.53) (0.40) (0.56) (0.72) BAB 1.47 1.28 1.25 –0.64 –0.74 –1.43 (0.17)*** (0.14)*** (0.19)*** (0.12)*** (0.14)*** (0.22)*** Callable 0.05 –0.24 0.59 –0.12 0.69 0.16 (0.13) (0.13)* (0.16)*** (0.11) (0.15)*** (0.50) Negotiated 0.00 0.05 –0.16 0.05 –0.14 0.00 (0.15) (0.12) (0.17) (0.12) (0.15) (0.17) Issue rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Yes Moody’s sector FE? Yes Yes Yes Yes Yes Yes NAIC filter? No No No No No Yes Adjusted $$R^{\mathrm{2}}$$ 0.48 0.45 0.46 0.41 0.47 0.49 $$N$$ 1,910 1,910 1,910 1,910 1,910 852 Post recalibration $$\times$$ Notches –0.07 –0.18 –0.16 –0.14 –0.12 –0.12 (0.03)** (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.07)*** Post recalibration –0.10 0.27 0.16 0.15 0.08 0.07 (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.02)** (0.02)** Panel B: Recalibration effect captured with upgrade magnitude (notches) Notches –0.09 0.00 –0.11 –0.05 –0.13 –0.12 (0.09) (0.08) (0.10) (0.08) (0.10) (0.13) Par 0.07 0.08 0.06 0.08 0.06 0.14 (0.04) (0.04)** (0.04) (0.04)** (0.04) (0.05)** Duration$$_{1}$$ 0.18 –0.03 0.04 (0.02)*** (0.01)*** (0.01)*** Duration$$_{2}$$ –0.13 –0.03 –0.04 (0.02)*** (0.02)** (0.04) Coupon 0.02 0.00 0.00 –0.01 0.00 –0.01 (0.07) (0.05) (0.05) (0.05) (0.05) (0.08) Outstanding bonds –0.03 –0.05 0.00 –0.04 0.00 –0.03 (0.04) (0.04) (0.06) (0.04) (0.06) (0.07) GO 0.24 0.10 –0.11 0.14 –0.05 –0.12 (0.47) (0.35) (0.51) (0.39) (0.55) (0.67) BAB 1.45 1.27 1.22 –0.66 –0.77 –1.48 (0.17)*** (0.14)*** (0.19)*** (0.13)*** (0.14)*** (0.24)*** Callable 0.04 –0.24 0.59 –0.13 0.69 0.16 (0.13) (0.13)* (0.16)*** (0.11) (0.15)*** (0.48) Negotiated 0.00 0.05 –0.18 0.04 –0.16 –0.03 (0.14) (0.12) (0.17) (0.12) (0.15) (0.18) Issue rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Yes Moody’s sector FE? Yes Yes Yes Yes Yes Yes NAIC filter? No No No No No Yes Adjusted $$R^{\mathrm{2}}$$ 0.58 0.45 0.46 0.42 0.58 0.49 $$N$$ 1,910 1,910 1,910 1,910 1,910 852 This table displays OLS regression results with various measures of secondary market yields and spreads as the dependent variable. The dependent variables appear at the top of each column. The dependent variables are defined in the legend of Table A.2 and are averaged across transactions by bond either during the 30-day window before Moody’s published the Primary Algorithm or after the bond’s recalibration date. We include in the regressions uninsured bonds with trading in both 30-day periods before Moody’s published the Primary Algorithm and after the bond’s recalibration date. Panel A captures the effect of recalibration on bond yields and spreads with Upgrade, an indicator variable taking a value of one if the bond experienced an upgrade as a result of its recalibration and zero if the bond experienced no change in ratings. Post recalibration is an indicator variable taking a value of one if the observation is from the 30-day window after the bond’s recalibration date and zero if the observation is from the 30-day period prior to the publication of Moody’s Primary Algorithm. We exclude data from the day of the bond’s recalibration. Other control variables are defined in the legend of Table A.2. Issue rating pre-recalibration FE are fixed effects based on the bond’s rating before the bond’s recalibration. Notches represents the change in the rating around the recalibration in terms of number of notches. Panel B captures the effect of recalibration on changes in average yields and spreads with Notches. Issuer level of government FE are fixed effects based on whether the issuer is a state, county, city, or other. Issuer state FE are fixed effects based on the issuer’s state. Moody’s sector FE are fixed effects based on the four sectors into which Moody’s classifies bonds in its Primary Algorithm. We measure the percentage of each bond’s principal held by insurance companies at the time of Moody’s recalibration using Annual Fraternal, Life, Property, Separate Accounts, and Title & Health Data from the National Association of Insurance Commissioners (NAIC). Column (6) replicates column (5) after excluding bonds with any nonzero holdings by insurance companies. Standard errors are in parentheses below coefficient estimates. We cluster standard errors by issuer. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 4 Secondary market regressions Panel A: Recalibration effect captured with upgrade indicator (upgrade) Raw yield Spread to Treasury$$_{{1}}$$ Spread to Treasury$$_{{2}}$$ Spread to after-tax Treasury$$_{{1}}$$ Spread to after-tax Treasury$$_{{2}}$$ Spread to after-tax Treasury$$_{{2}}$$ (1) (2) (3) (4) (5) (6) Post recalibration $$\times$$ Upgrade –0.06 –0.33 –0.28 –0.23 –0.19 –0.21 (0.07) (0.08)*** (0.08)*** (0.08)*** (0.07)*** (0.07)*** Post recalibration –0.13 0.27 0.15 0.13 0.06 0.07 (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.02)** (0.02)** Upgrade –0.16 –0.03 –0.33 –0.10 –0.37 –0.26 (0.17) (0.16) (0.22) (0.16) (0.23) (0.34) Par 0.07 0.08 0.06 0.08 0.06 0.14 (0.04)* (0.04)** (0.04) (0.04)** (0.04) (0.05)** Duration$$_{\mathrm{1}}$$ 0.18 –0.03 0.04 (0.02)*** (0.01)*** (0.01)*** Duration$$_{\mathrm{2}}$$ –0.13 –0.04 –0.04 (0.02)*** (0.02)** (0.04) Coupon 0.01 –0.01 –0.01 –0.01 –0.01 –0.02 (0.06) (0.04) (0.04) (0.05) (0.05) (0.07) Outstanding bonds –0.03 –0.05 0.00 –0.04 0.01 –0.02 (0.04) (0.04) (0.05) (0.04) (0.05) (0.06) GO 0.23 0.10 –0.11 0.14 –0.05 –0.09 (0.47) (0.35) (0.53) (0.40) (0.56) (0.72) BAB 1.47 1.28 1.25 –0.64 –0.74 –1.43 (0.17)*** (0.14)*** (0.19)*** (0.12)*** (0.14)*** (0.22)*** Callable 0.05 –0.24 0.59 –0.12 0.69 0.16 (0.13) (0.13)* (0.16)*** (0.11) (0.15)*** (0.50) Negotiated 0.00 0.05 –0.16 0.05 –0.14 0.00 (0.15) (0.12) (0.17) (0.12) (0.15) (0.17) Issue rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Yes Moody’s sector FE? Yes Yes Yes Yes Yes Yes NAIC filter? No No No No No Yes Adjusted $$R^{\mathrm{2}}$$ 0.48 0.45 0.46 0.41 0.47 0.49 $$N$$ 1,910 1,910 1,910 1,910 1,910 852 Post recalibration $$\times$$ Notches –0.07 –0.18 –0.16 –0.14 –0.12 –0.12 (0.03)** (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.07)*** Post recalibration –0.10 0.27 0.16 0.15 0.08 0.07 (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.02)** (0.02)** Panel B: Recalibration effect captured with upgrade magnitude (notches) Notches –0.09 0.00 –0.11 –0.05 –0.13 –0.12 (0.09) (0.08) (0.10) (0.08) (0.10) (0.13) Par 0.07 0.08 0.06 0.08 0.06 0.14 (0.04) (0.04)** (0.04) (0.04)** (0.04) (0.05)** Duration$$_{1}$$ 0.18 –0.03 0.04 (0.02)*** (0.01)*** (0.01)*** Duration$$_{2}$$ –0.13 –0.03 –0.04 (0.02)*** (0.02)** (0.04) Coupon 0.02 0.00 0.00 –0.01 0.00 –0.01 (0.07) (0.05) (0.05) (0.05) (0.05) (0.08) Outstanding bonds –0.03 –0.05 0.00 –0.04 0.00 –0.03 (0.04) (0.04) (0.06) (0.04) (0.06) (0.07) GO 0.24 0.10 –0.11 0.14 –0.05 –0.12 (0.47) (0.35) (0.51) (0.39) (0.55) (0.67) BAB 1.45 1.27 1.22 –0.66 –0.77 –1.48 (0.17)*** (0.14)*** (0.19)*** (0.13)*** (0.14)*** (0.24)*** Callable 0.04 –0.24 0.59 –0.13 0.69 0.16 (0.13) (0.13)* (0.16)*** (0.11) (0.15)*** (0.48) Negotiated 0.00 0.05 –0.18 0.04 –0.16 –0.03 (0.14) (0.12) (0.17) (0.12) (0.15) (0.18) Issue rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Yes Moody’s sector FE? Yes Yes Yes Yes Yes Yes NAIC filter? No No No No No Yes Adjusted $$R^{\mathrm{2}}$$ 0.58 0.45 0.46 0.42 0.58 0.49 $$N$$ 1,910 1,910 1,910 1,910 1,910 852 Panel A: Recalibration effect captured with upgrade indicator (upgrade) Raw yield Spread to Treasury$$_{{1}}$$ Spread to Treasury$$_{{2}}$$ Spread to after-tax Treasury$$_{{1}}$$ Spread to after-tax Treasury$$_{{2}}$$ Spread to after-tax Treasury$$_{{2}}$$ (1) (2) (3) (4) (5) (6) Post recalibration $$\times$$ Upgrade –0.06 –0.33 –0.28 –0.23 –0.19 –0.21 (0.07) (0.08)*** (0.08)*** (0.08)*** (0.07)*** (0.07)*** Post recalibration –0.13 0.27 0.15 0.13 0.06 0.07 (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.02)** (0.02)** Upgrade –0.16 –0.03 –0.33 –0.10 –0.37 –0.26 (0.17) (0.16) (0.22) (0.16) (0.23) (0.34) Par 0.07 0.08 0.06 0.08 0.06 0.14 (0.04)* (0.04)** (0.04) (0.04)** (0.04) (0.05)** Duration$$_{\mathrm{1}}$$ 0.18 –0.03 0.04 (0.02)*** (0.01)*** (0.01)*** Duration$$_{\mathrm{2}}$$ –0.13 –0.04 –0.04 (0.02)*** (0.02)** (0.04) Coupon 0.01 –0.01 –0.01 –0.01 –0.01 –0.02 (0.06) (0.04) (0.04) (0.05) (0.05) (0.07) Outstanding bonds –0.03 –0.05 0.00 –0.04 0.01 –0.02 (0.04) (0.04) (0.05) (0.04) (0.05) (0.06) GO 0.23 0.10 –0.11 0.14 –0.05 –0.09 (0.47) (0.35) (0.53) (0.40) (0.56) (0.72) BAB 1.47 1.28 1.25 –0.64 –0.74 –1.43 (0.17)*** (0.14)*** (0.19)*** (0.12)*** (0.14)*** (0.22)*** Callable 0.05 –0.24 0.59 –0.12 0.69 0.16 (0.13) (0.13)* (0.16)*** (0.11) (0.15)*** (0.50) Negotiated 0.00 0.05 –0.16 0.05 –0.14 0.00 (0.15) (0.12) (0.17) (0.12) (0.15) (0.17) Issue rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Yes Moody’s sector FE? Yes Yes Yes Yes Yes Yes NAIC filter? No No No No No Yes Adjusted $$R^{\mathrm{2}}$$ 0.48 0.45 0.46 0.41 0.47 0.49 $$N$$ 1,910 1,910 1,910 1,910 1,910 852 Post recalibration $$\times$$ Notches –0.07 –0.18 –0.16 –0.14 –0.12 –0.12 (0.03)** (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.07)*** Post recalibration –0.10 0.27 0.16 0.15 0.08 0.07 (0.02)*** (0.02)*** (0.02)*** (0.02)*** (0.02)** (0.02)** Panel B: Recalibration effect captured with upgrade magnitude (notches) Notches –0.09 0.00 –0.11 –0.05 –0.13 –0.12 (0.09) (0.08) (0.10) (0.08) (0.10) (0.13) Par 0.07 0.08 0.06 0.08 0.06 0.14 (0.04) (0.04)** (0.04) (0.04)** (0.04) (0.05)** Duration$$_{1}$$ 0.18 –0.03 0.04 (0.02)*** (0.01)*** (0.01)*** Duration$$_{2}$$ –0.13 –0.03 –0.04 (0.02)*** (0.02)** (0.04) Coupon 0.02 0.00 0.00 –0.01 0.00 –0.01 (0.07) (0.05) (0.05) (0.05) (0.05) (0.08) Outstanding bonds –0.03 –0.05 0.00 –0.04 0.00 –0.03 (0.04) (0.04) (0.06) (0.04) (0.06) (0.07) GO 0.24 0.10 –0.11 0.14 –0.05 –0.12 (0.47) (0.35) (0.51) (0.39) (0.55) (0.67) BAB 1.45 1.27 1.22 –0.66 –0.77 –1.48 (0.17)*** (0.14)*** (0.19)*** (0.13)*** (0.14)*** (0.24)*** Callable 0.04 –0.24 0.59 –0.13 0.69 0.16 (0.13) (0.13)* (0.16)*** (0.11) (0.15)*** (0.48) Negotiated 0.00 0.05 –0.18 0.04 –0.16 –0.03 (0.14) (0.12) (0.17) (0.12) (0.15) (0.18) Issue rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Yes Moody’s sector FE? Yes Yes Yes Yes Yes Yes NAIC filter? No No No No No Yes Adjusted $$R^{\mathrm{2}}$$ 0.58 0.45 0.46 0.42 0.58 0.49 $$N$$ 1,910 1,910 1,910 1,910 1,910 852 This table displays OLS regression results with various measures of secondary market yields and spreads as the dependent variable. The dependent variables appear at the top of each column. The dependent variables are defined in the legend of Table A.2 and are averaged across transactions by bond either during the 30-day window before Moody’s published the Primary Algorithm or after the bond’s recalibration date. We include in the regressions uninsured bonds with trading in both 30-day periods before Moody’s published the Primary Algorithm and after the bond’s recalibration date. Panel A captures the effect of recalibration on bond yields and spreads with Upgrade, an indicator variable taking a value of one if the bond experienced an upgrade as a result of its recalibration and zero if the bond experienced no change in ratings. Post recalibration is an indicator variable taking a value of one if the observation is from the 30-day window after the bond’s recalibration date and zero if the observation is from the 30-day period prior to the publication of Moody’s Primary Algorithm. We exclude data from the day of the bond’s recalibration. Other control variables are defined in the legend of Table A.2. Issue rating pre-recalibration FE are fixed effects based on the bond’s rating before the bond’s recalibration. Notches represents the change in the rating around the recalibration in terms of number of notches. Panel B captures the effect of recalibration on changes in average yields and spreads with Notches. Issuer level of government FE are fixed effects based on whether the issuer is a state, county, city, or other. Issuer state FE are fixed effects based on the issuer’s state. Moody’s sector FE are fixed effects based on the four sectors into which Moody’s classifies bonds in its Primary Algorithm. We measure the percentage of each bond’s principal held by insurance companies at the time of Moody’s recalibration using Annual Fraternal, Life, Property, Separate Accounts, and Title & Health Data from the National Association of Insurance Commissioners (NAIC). Column (6) replicates column (5) after excluding bonds with any nonzero holdings by insurance companies. Standard errors are in parentheses below coefficient estimates. We cluster standard errors by issuer. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. We control for issue characteristics including par value, duration, coupon rate, and dummy variables indicating whether the issues are callable, BABs, issued through a negotiated process, and whether the bonds are General Obligation (GO) or revenue bonds. We define all variables and report summary statistics in the Internet Appendix (Table A.2). Issue rating pre-recalibration FE are fixed effects based on ratings before recalibration. Issuer state FE control for unobserved heterogeneity at the state level. Issuer level of government FE are fixed effects based on Moody’s classification.20Moody’s sector FE are fixed effects based on the four sectors into which Moody’s classifies bonds in its Primary Algorithm.21 We cluster standard errors by issuer. The coefficients on Post recalibration in Panel A indicate that the average Raw yield is lower (13 bp, significant at 5%) and the average credit spreads are higher (6 to 27 bp, significant at 5% or 1%) in the later time period. Because yields and spreads are time varying, we cannot infer any impact of the recalibration from this observation. The interaction term reported in the top row captures the important difference-in-differences. The treatment group’s yields declined by 6 bp, and credit spreads declined by 19 to 33 bp (significant at 5% or 1%) relative to the control group. Because we focus on retail-size trades, we believe that we have a clean test of the muni market reaction to new information communicated by the recalibrated ratings, free from any confounding regulatory compliance effects. As a robustness test, we replicate Table 4 after excluding issues with any holdings by insurance companies. We believe that excluding retail transactions for bonds that have any level of insurance company holdings is a very stringent filter.22 However, despite the reduction in sample size, Table 4 results remain significant across each of the dependent variables. To conserve space, we tabulate this robustness test in column (6) only for the dependent variable employed in column (5). In lieu of the binary indicator Upgrade, Table 4, Panel B, captures the effect of rating recalibration on average yields and credit spreads with a discrete variable. Notches is the difference between Rating post-recalibration and Rating pre-recalibration. As suggested by Figure 1, we find that the recalibration effect is larger in magnitude for municipalities whose bonds experienced larger upgrades. For each notch outstanding bonds are upgraded during recalibration, ex post Raw yield is 7 bp lower compared with the control group. For each notch of upgrade, spreads are lower by 12 to 18 bp. This result implies that spreads fall 36 to 54 bp for bonds upgraded three notches. As in Panel A, results are robust in column (6) to the highly stringent exclusion of transactions for bonds with any level of insurance company holdings. We replicate Table 4 under a wide variety of alternative specifications. We do not tabulate these results in the interest of conserving space. We find the results are fully robust to including fixed effects for rating levels in the post-recalibration period instead of the pre-recalibration period. The results are also fully robust to clustering standard errors by a host of alternative dimensions, such as by bond, issuer’s state, the four recalibration dates, pre- or post-recalibration period, issuer’s state $$\times$$ pre- or post-recalibration period, issuer’s state $$\times$$ the bond’s rating in the pre- or post-recalibration period, the four recalibration dates $$\times$$ the bond’s rating in the pre- or post-recalibration period, and the four recalibration dates $$\times$$ the issuer’s state. We further replicate each of these specifications using Notches instead of Upgrade to capture the recalibration effect and obtain results similar to those reported in Table 4, Panel B. 3.2 Alternative hypotheses The results thus far indicate that investors reacted to Moody’s recalibration. This section considers several alternative explanations for this finding. 3.2.1 Do the results reflect differences in returns that already existed before the recalibration? A potential concern regarding the analysis thus far is that our results may reflect differences in returns that already existed before Moody’s recalibration. We address this concern with a placebo analysis. We replace each of the four recalibration dates with false dates drawn randomly from the period 180 to 30 days before the first actual recalibration date. If the spreads on the upgraded bonds were already drifting upward relative to the non-upgraded bonds prior to the recalibration, then we should see results similar to our baseline when using event dates in the year prior to implementation. We replicate the regression in Table 4, Panel A, column (5), using the false dates. We complete this process a total of 10,000 times. Each time, we collect the $$t-$$statistic for the coefficient on Post recalibration$$\times$$Upgrade. Figure 3 plots a distribution of these $$t-$$statistics. The $$t$$-statistic from Table 4, Panel A, column (5), is –2.6 and lies in the lower 1% of the placebo distribution. This analysis demonstrates the uniqueness of the market reaction to the actual implementation dates. Figure 3 View largeDownload slide Histogram of placebo test results This figure displays a histogram of $$t$$-statistics from 10,000 placebo regressions in which we replicate the regression in Table 4, Panel A, column (5), after randomly replacing each of the four implementation dates with false dates drawn randomly from the period 180 to 30 days before the announcement date, March 16, 2010. The $$t$$-statistics come from the coefficient on Post recalibration$$\times$$Upgrade. The vertical dashed line represents the $$t$$-statistic from the true regression in Table 4, Panel A, column (5). Figure 3 View largeDownload slide Histogram of placebo test results This figure displays a histogram of $$t$$-statistics from 10,000 placebo regressions in which we replicate the regression in Table 4, Panel A, column (5), after randomly replacing each of the four implementation dates with false dates drawn randomly from the period 180 to 30 days before the announcement date, March 16, 2010. The $$t$$-statistics come from the coefficient on Post recalibration$$\times$$Upgrade. The vertical dashed line represents the $$t$$-statistic from the true regression in Table 4, Panel A, column (5). 3.2.2 Do the results reflect changes in liquidity?. Harris and Piwowar (2006) show that municipal bond liquidity increases with credit quality. If rating upgrades resulting from the recalibration are associated with perceived improvements in credit quality, then liquidity for upgraded bonds should increase, leading to lower returns due to lower liquidity premiums. We test this possibility in Table 5. Table 5 regressions employ Daily volume as the dependent variable, defined as the natural logarithm of the average daily trading volume for customers buying munis during the period before or after the bond’s recalibration date. Control variables are the same as those employed in Table 4. The sample in Table 5 consists of 7,341 bonds with at least one customer transaction in the 90 days before the publication of Moody’s Primary Algorithm and the 90 days after the recalibration dates, resulting in 14,682 observations. (See Table A.1 for sample reconciliation.) Table 5 The effect of recalibrated ratings on secondary market trading volume (1) (2) Upgrade $$\times$$ Post recalibration 3.34 (0.29)*** Upgrade $$\times$$ Post recalibration with delay 0.14 (0.20) Upgrade –0.15 –0.14 (0.17) (0.14) Post recalibration –5.40 (0.24)*** Post recalibration with delay –6.88 (0.07)*** Par –0.12 –0.11 (0.04)*** (0.02)** Duration$$_{2}$$ 0.11 0.10 (0.01)*** (0.01)*** Coupon –0.02 0.05 (0.04) (0.02)** Outstanding bonds 0.03 0.06 (0.07) (0.03)** GO 0.91 0.04 (0.15)*** (0.06) BAB 0.53 0.10 (0.17)*** (0.09) Callable –0.26 –0.18 (0.12)** (0.07)*** Negotiated 0.15 0.17 (0.13) (0.06)*** Issue rating pre-recal. FE? Yes Yes Issuer level of govt. FE? Yes Yes Issuer state FE? Yes Yes Moody’s sector FE? Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.35 0.75 $$N$$ 14,682 14,682 (1) (2) Upgrade $$\times$$ Post recalibration 3.34 (0.29)*** Upgrade $$\times$$ Post recalibration with delay 0.14 (0.20) Upgrade –0.15 –0.14 (0.17) (0.14) Post recalibration –5.40 (0.24)*** Post recalibration with delay –6.88 (0.07)*** Par –0.12 –0.11 (0.04)*** (0.02)** Duration$$_{2}$$ 0.11 0.10 (0.01)*** (0.01)*** Coupon –0.02 0.05 (0.04) (0.02)** Outstanding bonds 0.03 0.06 (0.07) (0.03)** GO 0.91 0.04 (0.15)*** (0.06) BAB 0.53 0.10 (0.17)*** (0.09) Callable –0.26 –0.18 (0.12)** (0.07)*** Negotiated 0.15 0.17 (0.13) (0.06)*** Issue rating pre-recal. FE? Yes Yes Issuer level of govt. FE? Yes Yes Issuer state FE? Yes Yes Moody’s sector FE? Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.35 0.75 $$N$$ 14,682 14,682 This table displays OLS regression results with Daily volume as the dependent variable. This variable is the natural logarithm of the average daily trading volume for customers buying municipal bonds during the 90-day window centered on the bond’s recalibration date. We calculate trading volume using secondary market trading data from the Municipal Securities Rulemaking Board (MSRB). The sample consists of 7,341 uninsured bonds with at least one transaction by a customer in the 90 days prior to the publication of Moody’s Primary Algorithm and the 90 days after the recalibration date (resulting in 14,682 observations). Upgrade is an indicator variable taking a value of one if the bond experienced an upgrade during the recalibration event and zero if the bond experienced no change in ratings. Post recalibration is an indicator variable taking a value of one if the observation is from the window [0,89] days relative to the bond’s recalibration date and zero if the observation is from the window [–90,–1] days prior to the publication of Moody’s Primary Algorithm. Post recalibration with delay is an indicator variable taking a value of one if the observation is from the window [90,179] days relative to the bond’s recalibration date and zero if the observation is from the window [–90,-1] days prior to the publication of Moody’s Primary Algorithm. Issue rating pre-recalibration FE are fixed effects based on the rating of the bond before its recalibration date. Issuer level of government FE are fixed effects based on whether the issuer is a state, county, city, or other. Issuer state FE are fixed effects based on the issuer’s state. Moody’s sector FE are fixed effects based on the four sectors into which Moody’s classifies bonds in its Primary Algorithm. Standard errors. We cluster standard errors by issuer. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 5 The effect of recalibrated ratings on secondary market trading volume (1) (2) Upgrade $$\times$$ Post recalibration 3.34 (0.29)*** Upgrade $$\times$$ Post recalibration with delay 0.14 (0.20) Upgrade –0.15 –0.14 (0.17) (0.14) Post recalibration –5.40 (0.24)*** Post recalibration with delay –6.88 (0.07)*** Par –0.12 –0.11 (0.04)*** (0.02)** Duration$$_{2}$$ 0.11 0.10 (0.01)*** (0.01)*** Coupon –0.02 0.05 (0.04) (0.02)** Outstanding bonds 0.03 0.06 (0.07) (0.03)** GO 0.91 0.04 (0.15)*** (0.06) BAB 0.53 0.10 (0.17)*** (0.09) Callable –0.26 –0.18 (0.12)** (0.07)*** Negotiated 0.15 0.17 (0.13) (0.06)*** Issue rating pre-recal. FE? Yes Yes Issuer level of govt. FE? Yes Yes Issuer state FE? Yes Yes Moody’s sector FE? Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.35 0.75 $$N$$ 14,682 14,682 (1) (2) Upgrade $$\times$$ Post recalibration 3.34 (0.29)*** Upgrade $$\times$$ Post recalibration with delay 0.14 (0.20) Upgrade –0.15 –0.14 (0.17) (0.14) Post recalibration –5.40 (0.24)*** Post recalibration with delay –6.88 (0.07)*** Par –0.12 –0.11 (0.04)*** (0.02)** Duration$$_{2}$$ 0.11 0.10 (0.01)*** (0.01)*** Coupon –0.02 0.05 (0.04) (0.02)** Outstanding bonds 0.03 0.06 (0.07) (0.03)** GO 0.91 0.04 (0.15)*** (0.06) BAB 0.53 0.10 (0.17)*** (0.09) Callable –0.26 –0.18 (0.12)** (0.07)*** Negotiated 0.15 0.17 (0.13) (0.06)*** Issue rating pre-recal. FE? Yes Yes Issuer level of govt. FE? Yes Yes Issuer state FE? Yes Yes Moody’s sector FE? Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.35 0.75 $$N$$ 14,682 14,682 This table displays OLS regression results with Daily volume as the dependent variable. This variable is the natural logarithm of the average daily trading volume for customers buying municipal bonds during the 90-day window centered on the bond’s recalibration date. We calculate trading volume using secondary market trading data from the Municipal Securities Rulemaking Board (MSRB). The sample consists of 7,341 uninsured bonds with at least one transaction by a customer in the 90 days prior to the publication of Moody’s Primary Algorithm and the 90 days after the recalibration date (resulting in 14,682 observations). Upgrade is an indicator variable taking a value of one if the bond experienced an upgrade during the recalibration event and zero if the bond experienced no change in ratings. Post recalibration is an indicator variable taking a value of one if the observation is from the window [0,89] days relative to the bond’s recalibration date and zero if the observation is from the window [–90,–1] days prior to the publication of Moody’s Primary Algorithm. Post recalibration with delay is an indicator variable taking a value of one if the observation is from the window [90,179] days relative to the bond’s recalibration date and zero if the observation is from the window [–90,-1] days prior to the publication of Moody’s Primary Algorithm. Issue rating pre-recalibration FE are fixed effects based on the rating of the bond before its recalibration date. Issuer level of government FE are fixed effects based on whether the issuer is a state, county, city, or other. Issuer state FE are fixed effects based on the issuer’s state. Moody’s sector FE are fixed effects based on the four sectors into which Moody’s classifies bonds in its Primary Algorithm. Standard errors. We cluster standard errors by issuer. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. The interaction term Upgrade$$\times$$Post recalibration reported in the top row of column (1) indeed indicates an increase in trading volume after the ratings changes. Significant at 1%, this result adds further evidence that muni investors responded to Moody’s scale recalibration. Column (1) indicates that the liquidity of the upgraded bonds improved relative to the non-upgraded bonds. In light of this finding, one might infer that the results in Figure 1 and Table 4 reflect changes in liquidity, rather than investors responding to credit ratings for information purposes. However, this interpretation implies that upgraded bonds exhibit a permanent increase in liquidity relative to non-upgraded bonds. We test this possibility in column (2) after substituting Post recalibration with delay for Post recalibration. Post recalibration with delay is an indicator variable taking a value of one if the observation is from the window [90,179] days relative to the bond’s recalibration date and zero if the observation is from the window [–90,–1] days prior to the publication of Moody’s Primary Algorithm. This approach allows us to test whether average daily trading volume remains elevated for upgraded bonds over the long term. We find it does not. The coefficient on Upgrade$$\times$$Post recalibration with delay is insignificant, indicating the trading volume of upgraded bonds is statistically indistinguishable from its pre-recalibration level. 3.2.3 Are the results driven by a sudden decline in returns for non-upgraded bonds? Our results could obtain if the returns for the non-upgraded bonds happened to be abnormally low around the time of the recalibration. We address this possibility two ways. First, we provide more detail regarding the cumulative returns of the non-upgraded bonds underlying Figure 1. Figure A.2 in the Internet Appendix displays separate plots for bonds upgraded zero, one, two, and three notches for each of the four recalibration dates. These plots generally show that the cumulative returns for the non-upgraded bonds behave very similarly to those of the upgraded bonds in the days leading up to the publication of Moody’s Primary Algorithm. In some cases, the cumulative returns for the non-upgraded bonds are actually higher than those of the soon-to-be-upgraded bonds. Second, we compare the changes in spreads of the non-upgraded bonds around the time of the recalibration and in the period preceding Moody’s publication of its Primary Algorithm. Figure 4 displays a histogram of placebo average changes in spreads for the non-upgraded bonds drawing random dates from the 180-day period prior to when the bonds’ ratings were recalibrated. This figure effectively reproduces the dependent variable from Table 4, Panel A, column (5), for the non-upgraded bonds many times with placebo recalibration dates. If the returns of the non-upgraded bonds were abnormally low around the time of the recalibration, then we should see the true change in Spread to after-tax Treasury$$_{2}$$ lie outside the placebo distribution. This is not what we observe. If anything, the changes in spreads for the non-upgraded bonds are slightly higher (although not significantly so) around the time of the recalibration in comparison to their own recent history. Figure 4 View largeDownload slide Histogram of placebo average change in spread for the control group Each of the four recalibration events include zero-notch upgrades, and these non-upgraded bonds compose our control group. Comparing spreads in the 30 days before versus 30 days after the bonds are recalibrated, we observe an average change in spread of 0.07139%. The vertical dashed line indicates this amount. This figure reflects similar calculations after we randomly draw placebo dates from the 180-day period prior to when the bonds were recalibrated. For example, for a bond that received a zero-notch upgrade on April 16, 2010, we randomly draw a placebo date from the 180-day period before the publication of Moody’s Primary Algorithm (March 16, 2010) and compute the change in spread from the 30-day window before the placebo date to the spread during the window 31 to 60 days after the placebo date. This gap between the two periods accounts for the 30-day gap between the publication of Moody’s Primary Algorithm and April 16, 2010. For a bond that received a zero-notch upgrade on April 23, 2010, we randomly draw a placebo date from the 180-day period before the publication of Moody’s Primary Algorithm (March 16, 2010) and compute the change in spread from the 30-day window before the placebo date to the spread in the window 38 to 67 days after the placebo date. We do this for all non-upgraded bonds across the four recalibration dates and compute the average. We repeat this process 10,000 times and construct the displayed histogram. Figure 4 View largeDownload slide Histogram of placebo average change in spread for the control group Each of the four recalibration events include zero-notch upgrades, and these non-upgraded bonds compose our control group. Comparing spreads in the 30 days before versus 30 days after the bonds are recalibrated, we observe an average change in spread of 0.07139%. The vertical dashed line indicates this amount. This figure reflects similar calculations after we randomly draw placebo dates from the 180-day period prior to when the bonds were recalibrated. For example, for a bond that received a zero-notch upgrade on April 16, 2010, we randomly draw a placebo date from the 180-day period before the publication of Moody’s Primary Algorithm (March 16, 2010) and compute the change in spread from the 30-day window before the placebo date to the spread during the window 31 to 60 days after the placebo date. This gap between the two periods accounts for the 30-day gap between the publication of Moody’s Primary Algorithm and April 16, 2010. For a bond that received a zero-notch upgrade on April 23, 2010, we randomly draw a placebo date from the 180-day period before the publication of Moody’s Primary Algorithm (March 16, 2010) and compute the change in spread from the 30-day window before the placebo date to the spread in the window 38 to 67 days after the placebo date. We do this for all non-upgraded bonds across the four recalibration dates and compute the average. We repeat this process 10,000 times and construct the displayed histogram. 3.2.4 Are the results driven by shifts in demand for certain levels of governments’ bonds? Table A.3 in the Internet Appendix replicates the specification in Table 4, Panel A, column (5), after splitting the sample by issuer level of government. This table shows that the secondary market results are driven primarily by state-level issuers and issuers at other levels of government. The results are weaker for county- and city-level issuers. We dig deeper into these patterns to investigate whether the main results in Table 4 could be driven by shifts in demand for bonds issued by certain levels of government, rather than investors responding to changes in ratings. We control for this possibility directly in Table A.4 in the Internet Appendix. Here, we replicate Table 4 after including the issuer level of government fixed effects that vary by pre- and post-treatment figures. For example, instead of including a fixed effect for the state of California, as we do in Table 4, Table A.4 includes one fixed effect for California in the pre-treatment period and another for California in the post-treatment period. Under this specification, our main results are virtually unchanged, indicating our results do not reflect a change in taste among investors for bonds from different levels of government during the time of the recalibration. 3.2.5 Are the results driven by differential changes in fundamentals for the upgraded and non-upgraded bonds? Moody’s claims that the recalibration was intentionally uncorrelated with changes in issuer fundamentals. However, it is possible that Moody’s selected bonds for upgrades that were experiencing strengthening fundamentals. We believe this explanation is unlikely for two reasons. First, as mentioned above, Moody’s indicated that any bonds under review going into the recalibration would remain under review after the recalibration. In other words, bonds that Moody’s believed were due for natural rating changes would receive proper consideration for those changes after the recalibration had concluded. Second, we examine the behavior of S&P’s ratings around the time of Moody’s recalibration. If Moody’s recalibration reflects changes in fundamentals, then we should expect S&P to eventually enact rating changes that correlate with those of Moody’s. Table A.7 in the Internet Appendix provides detailed transition matrices of S&P’s ratings from before the recalibration to a year after. We observe no clear evidence that S&P responded to Moody’s recalibration. Most bonds rated by S&P retain their ratings at least one year after the recalibration. For example, of the 26,582 bonds with AA$$+$$ ratings from S&P, 25,155 of them still have AA$$+$$ ratings a year after the recalibration. A similar, consistent pattern appears at each rating level. Further, we examine the long-run behavior of Moody’s and S&P’s ratings in Figure 5. In this figure, we compute average ratings on uninsured bonds with ratings from both Moody’s and S&P over a four-year period centered on Moody’s recalibration. We separate these bonds based on whether they were upgraded as a result of Moody’s recalibration and report average ratings by month for these subgroups of bonds. Figure 5 View largeDownload slide Average Moody’s and Standard & Poor’s credit ratings through time This figure displays average credit ratings for uninsured municipal bonds rated by Moody’s and Standard & Poor’s from a four-year period centered on Moody’s recalibration. We separate the bonds by whether their ratings were upgraded as a result of Moody’s recalibration. The vertical dashed line represents April 2010. We convert Moody’s and Standard & Poor’s 21-point alphanumeric scales into numbers. Ratings are increasing in credit quality, such that Aaa/AAA $$=$$ 21, Aa1/AA$$+ =$$ 20,..., C/C $$=$$ 1. We collect credit ratings from Moody’s website and Standard & Poor’s RatingsDirect. Figure 5 View largeDownload slide Average Moody’s and Standard & Poor’s credit ratings through time This figure displays average credit ratings for uninsured municipal bonds rated by Moody’s and Standard & Poor’s from a four-year period centered on Moody’s recalibration. We separate the bonds by whether their ratings were upgraded as a result of Moody’s recalibration. The vertical dashed line represents April 2010. We convert Moody’s and Standard & Poor’s 21-point alphanumeric scales into numbers. Ratings are increasing in credit quality, such that Aaa/AAA $$=$$ 21, Aa1/AA$$+ =$$ 20,..., C/C $$=$$ 1. We collect credit ratings from Moody’s website and Standard & Poor’s RatingsDirect. Among upgraded bonds (plotted with solid lines in Figure 5), we observe that Moody’s ratings are lower than S&P’s in 2008 and 2009 but then converge due to recalibration and remain comparable to S&P’s ratings through the end of our data availability. Among the control group that did not upgrade due to recalibration (plotted with dashed lines), we observe that Moody’s ratings are higher than S&P’s by approximately one notch over the entire four-year period. The stability in Moody’s ratings after recalibration affirms our conclusion that the recalibration was a permanent change in ratings. 3.3 Evidence from new issues We consider next the extent to which market reliance on credit ratings has real economic effects. Specifically, we examine the impact on new debt pricing and new debt issuance. In light of the secondary market results above, we predict that upgraded municipalities face lower borrowing costs in the primary market as a result of their higher ratings. To the extent that previously higher borrowing costs (associated with the more stringent rating standards) discouraged municipal investment, we predict an increase in borrowing among the upgraded municipalities. We test both predictions in a difference-in-differences framework. In the preceding tests of secondary market prices and trading volume, our treatment and control groups consist of recalibrated bonds. In this section, we define treatment and control groups at the municipality level.23 Specifically, the treatment group contains municipalities whose outstanding bonds were recalibrated up at least one notch. The control group contains municipalities whose outstanding bonds had a “Change in Scale” rating action of zero notches. In order to analyze effects at the issuer level, we calculate the average ratings of all outstanding bonds for each issuer before and after recalibration dates. (The modal within-issuer standard deviation of credit ratings is zero notches.) We refer to these averages as Issuer rating pre-recalibration and Issuer rating post-recalibration, respectively. We examine first the impact on new issue pricing in multivariate regressions, with results reported in Table 6. We calculate the average characteristics of new, uninsured issues for each issuer in the year before and, separately, the year after the issuer’s outstanding bonds were recalibrated. We compute the change in each bond characteristic for each issuer. This approach results in one observation per issuer. Dependent variables are changes in yields and spreads. Raw yield is the offer yield in all analyses of new issues. Each of the credit spread variables is defined in Section 2.4. We control for changes in issue characteristics and report summary statistics for this issuer-level sample in Table A.5.24 In order to capture variation in the liquidity of issuers’ bonds, we also control for the change in the number of bonds outstanding. Fixed effects are similar to those in Table 4. Table 6 Issuer-level primary market regressions Panel A: Recalibration effect captured with upgrade indicator (upgrade) $$\Delta $$ Raw yield $$\Delta $$ Spread to Treasury$$_{\mathrm{1}}$$ $$\Delta $$ Spread to Treasury$$_{\mathrm{2}}$$ $$\Delta $$ Spread after-tax Treasury$$_{\mathrm{1}}$$ $$\Delta $$ Spread after-tax Treasury$$_{\mathrm{2}}$$ (1) (2) (3) (4) (5) Upgrade –0.15 –0.19 –0.22 –0.18 –0.21 (0.06)** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Par 0.04 0.15 0.21 0.03 0.08 (0.04) (0.04)*** (0.04)*** (0.03) (0.04)** $$\Delta $$ Duration$$_{\mathrm{1}}$$ 0.20 –0.06 0.02 (0.01)*** (0.01)*** (0.01)* $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.09 0.00 (0.02)*** (0.01) $$\Delta $$ Coupon 0.46 0.29 0.35 0.31 0.36 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.15 –0.09 –0.04 –0.09 –0.03 (0.13) (0.14) (0.14) (0.12) (0.13) $$\Delta $$ GO –0.33 –0.36 –0.36 –0.32 –0.32 (0.09)*** (0.09)*** (0.09)*** (0.08)*** (0.09)*** $$\Delta $$ BAB 0.62 0.63 0.66 –0.70 –0.68 (0.06)*** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Callable 0.11 0.01 0.23 0.06 0.27 (0.06)* (0.06) (0.06)*** (0.05) (0.05)*** $$\Delta $$ Negotiated –0.06 0.02 0.35 0.31 0.36 (0.07) (0.07) (0.03)*** (0.03)*** (0.03)*** Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.69 0.40 0.41 0.37 0.37 $$N$$ 865 865 865 865 865 Panel B: Recalibration effect captured with upgrade magnitude (notches) Notches –0.11 –0.13 –0.14 –0.11 –0.13 (0.04)*** (0.04)*** (0.04)*** (0.04)*** (0.04)*** $$\Delta $$ Par 0.04 0.15 0.21 0.03 0.09 (0.04) (0.04)*** (0.04)*** (0.03) (0.04)** $$\Delta $$ Duration$$_{\mathrm{1}}$$ 0.20 –0.06 0.02 (0.01)*** (0.01)*** (0.01) $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.09 –0.00 (0.02)*** (0.01) $$\Delta $$ Coupon 0.46 0.29 0.35 0.31 0.36 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.16 –0.10 –0.05 –0.10 –0.04 (0.13) (0.14) (0.14) (0.12) (0.13) $$\Delta $$ GO –0.30 –0.33 –0.32 –0.29 –0.28 (0.09)*** (0.09)*** (0.10)*** (0.08)*** (0.09)*** $$\Delta $$ BAB 0.61 0.61 0.64 –0.71 –0.70 (0.06)*** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Callable 0.11 0.01 0.23 0.06 0.28 (0.06)* (0.06) (0.06)*** (0.05) (0.05)*** $$\Delta $$ Negotiated –0.06 0.02 –0.07 0.05 –0.03 (0.07) (0.07) (0.08) (0.07) (0.07) Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.69 0.40 0.41 0.37 0.36 $$N$$ 865 865 865 865 865 Panel A: Recalibration effect captured with upgrade indicator (upgrade) $$\Delta $$ Raw yield $$\Delta $$ Spread to Treasury$$_{\mathrm{1}}$$ $$\Delta $$ Spread to Treasury$$_{\mathrm{2}}$$ $$\Delta $$ Spread after-tax Treasury$$_{\mathrm{1}}$$ $$\Delta $$ Spread after-tax Treasury$$_{\mathrm{2}}$$ (1) (2) (3) (4) (5) Upgrade –0.15 –0.19 –0.22 –0.18 –0.21 (0.06)** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Par 0.04 0.15 0.21 0.03 0.08 (0.04) (0.04)*** (0.04)*** (0.03) (0.04)** $$\Delta $$ Duration$$_{\mathrm{1}}$$ 0.20 –0.06 0.02 (0.01)*** (0.01)*** (0.01)* $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.09 0.00 (0.02)*** (0.01) $$\Delta $$ Coupon 0.46 0.29 0.35 0.31 0.36 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.15 –0.09 –0.04 –0.09 –0.03 (0.13) (0.14) (0.14) (0.12) (0.13) $$\Delta $$ GO –0.33 –0.36 –0.36 –0.32 –0.32 (0.09)*** (0.09)*** (0.09)*** (0.08)*** (0.09)*** $$\Delta $$ BAB 0.62 0.63 0.66 –0.70 –0.68 (0.06)*** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Callable 0.11 0.01 0.23 0.06 0.27 (0.06)* (0.06) (0.06)*** (0.05) (0.05)*** $$\Delta $$ Negotiated –0.06 0.02 0.35 0.31 0.36 (0.07) (0.07) (0.03)*** (0.03)*** (0.03)*** Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.69 0.40 0.41 0.37 0.37 $$N$$ 865 865 865 865 865 Panel B: Recalibration effect captured with upgrade magnitude (notches) Notches –0.11 –0.13 –0.14 –0.11 –0.13 (0.04)*** (0.04)*** (0.04)*** (0.04)*** (0.04)*** $$\Delta $$ Par 0.04 0.15 0.21 0.03 0.09 (0.04) (0.04)*** (0.04)*** (0.03) (0.04)** $$\Delta $$ Duration$$_{\mathrm{1}}$$ 0.20 –0.06 0.02 (0.01)*** (0.01)*** (0.01) $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.09 –0.00 (0.02)*** (0.01) $$\Delta $$ Coupon 0.46 0.29 0.35 0.31 0.36 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.16 –0.10 –0.05 –0.10 –0.04 (0.13) (0.14) (0.14) (0.12) (0.13) $$\Delta $$ GO –0.30 –0.33 –0.32 –0.29 –0.28 (0.09)*** (0.09)*** (0.10)*** (0.08)*** (0.09)*** $$\Delta $$ BAB 0.61 0.61 0.64 –0.71 –0.70 (0.06)*** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Callable 0.11 0.01 0.23 0.06 0.28 (0.06)* (0.06) (0.06)*** (0.05) (0.05)*** $$\Delta $$ Negotiated –0.06 0.02 –0.07 0.05 –0.03 (0.07) (0.07) (0.08) (0.07) (0.07) Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.69 0.40 0.41 0.37 0.36 $$N$$ 865 865 865 865 865 This table displays OLS regression results with changes in various measures of offer yields and spreads as the dependent variable. The dependent variables appear at the top of each column. The dependent variables are defined in the legend of Table A.5. Panel A captures the effect of recalibration on changes in issuers’ yields and credit spreads with Upgrade, an indicator variable taking a value of one if the issuer experienced an upgrade on its outstanding bonds during any of the recalibration events and zero if the issuer’s bonds experienced no change in ratings. Other control variables are defined in the legend of Table A.5. We calculate the average rating of all outstanding bonds for each issuer before (Issuer rating pre-recalibration) and after (Issuer rating post-recalibration) the recalibration dates. Notches represents the change in this number rounded to the nearest whole number. Panel B captures the effect of recalibration on changes in issuers’ yields and credit spreads with Notches. Issuer rating pre-recalibration FE are fixed effects based on the average rating of all outstanding bonds for each issuer before its recalibration date. Issuer level of government FE are fixed effects based on whether the issuer is a state, county, city, or other. Issuer state FE are fixed effects based on the issuer’s state. Standard errors are in parentheses below coefficient estimates. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 6 Issuer-level primary market regressions Panel A: Recalibration effect captured with upgrade indicator (upgrade) $$\Delta $$ Raw yield $$\Delta $$ Spread to Treasury$$_{\mathrm{1}}$$ $$\Delta $$ Spread to Treasury$$_{\mathrm{2}}$$ $$\Delta $$ Spread after-tax Treasury$$_{\mathrm{1}}$$ $$\Delta $$ Spread after-tax Treasury$$_{\mathrm{2}}$$ (1) (2) (3) (4) (5) Upgrade –0.15 –0.19 –0.22 –0.18 –0.21 (0.06)** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Par 0.04 0.15 0.21 0.03 0.08 (0.04) (0.04)*** (0.04)*** (0.03) (0.04)** $$\Delta $$ Duration$$_{\mathrm{1}}$$ 0.20 –0.06 0.02 (0.01)*** (0.01)*** (0.01)* $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.09 0.00 (0.02)*** (0.01) $$\Delta $$ Coupon 0.46 0.29 0.35 0.31 0.36 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.15 –0.09 –0.04 –0.09 –0.03 (0.13) (0.14) (0.14) (0.12) (0.13) $$\Delta $$ GO –0.33 –0.36 –0.36 –0.32 –0.32 (0.09)*** (0.09)*** (0.09)*** (0.08)*** (0.09)*** $$\Delta $$ BAB 0.62 0.63 0.66 –0.70 –0.68 (0.06)*** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Callable 0.11 0.01 0.23 0.06 0.27 (0.06)* (0.06) (0.06)*** (0.05) (0.05)*** $$\Delta $$ Negotiated –0.06 0.02 0.35 0.31 0.36 (0.07) (0.07) (0.03)*** (0.03)*** (0.03)*** Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.69 0.40 0.41 0.37 0.37 $$N$$ 865 865 865 865 865 Panel B: Recalibration effect captured with upgrade magnitude (notches) Notches –0.11 –0.13 –0.14 –0.11 –0.13 (0.04)*** (0.04)*** (0.04)*** (0.04)*** (0.04)*** $$\Delta $$ Par 0.04 0.15 0.21 0.03 0.09 (0.04) (0.04)*** (0.04)*** (0.03) (0.04)** $$\Delta $$ Duration$$_{\mathrm{1}}$$ 0.20 –0.06 0.02 (0.01)*** (0.01)*** (0.01) $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.09 –0.00 (0.02)*** (0.01) $$\Delta $$ Coupon 0.46 0.29 0.35 0.31 0.36 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.16 –0.10 –0.05 –0.10 –0.04 (0.13) (0.14) (0.14) (0.12) (0.13) $$\Delta $$ GO –0.30 –0.33 –0.32 –0.29 –0.28 (0.09)*** (0.09)*** (0.10)*** (0.08)*** (0.09)*** $$\Delta $$ BAB 0.61 0.61 0.64 –0.71 –0.70 (0.06)*** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Callable 0.11 0.01 0.23 0.06 0.28 (0.06)* (0.06) (0.06)*** (0.05) (0.05)*** $$\Delta $$ Negotiated –0.06 0.02 –0.07 0.05 –0.03 (0.07) (0.07) (0.08) (0.07) (0.07) Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.69 0.40 0.41 0.37 0.36 $$N$$ 865 865 865 865 865 Panel A: Recalibration effect captured with upgrade indicator (upgrade) $$\Delta $$ Raw yield $$\Delta $$ Spread to Treasury$$_{\mathrm{1}}$$ $$\Delta $$ Spread to Treasury$$_{\mathrm{2}}$$ $$\Delta $$ Spread after-tax Treasury$$_{\mathrm{1}}$$ $$\Delta $$ Spread after-tax Treasury$$_{\mathrm{2}}$$ (1) (2) (3) (4) (5) Upgrade –0.15 –0.19 –0.22 –0.18 –0.21 (0.06)** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Par 0.04 0.15 0.21 0.03 0.08 (0.04) (0.04)*** (0.04)*** (0.03) (0.04)** $$\Delta $$ Duration$$_{\mathrm{1}}$$ 0.20 –0.06 0.02 (0.01)*** (0.01)*** (0.01)* $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.09 0.00 (0.02)*** (0.01) $$\Delta $$ Coupon 0.46 0.29 0.35 0.31 0.36 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.15 –0.09 –0.04 –0.09 –0.03 (0.13) (0.14) (0.14) (0.12) (0.13) $$\Delta $$ GO –0.33 –0.36 –0.36 –0.32 –0.32 (0.09)*** (0.09)*** (0.09)*** (0.08)*** (0.09)*** $$\Delta $$ BAB 0.62 0.63 0.66 –0.70 –0.68 (0.06)*** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Callable 0.11 0.01 0.23 0.06 0.27 (0.06)* (0.06) (0.06)*** (0.05) (0.05)*** $$\Delta $$ Negotiated –0.06 0.02 0.35 0.31 0.36 (0.07) (0.07) (0.03)*** (0.03)*** (0.03)*** Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.69 0.40 0.41 0.37 0.37 $$N$$ 865 865 865 865 865 Panel B: Recalibration effect captured with upgrade magnitude (notches) Notches –0.11 –0.13 –0.14 –0.11 –0.13 (0.04)*** (0.04)*** (0.04)*** (0.04)*** (0.04)*** $$\Delta $$ Par 0.04 0.15 0.21 0.03 0.09 (0.04) (0.04)*** (0.04)*** (0.03) (0.04)** $$\Delta $$ Duration$$_{\mathrm{1}}$$ 0.20 –0.06 0.02 (0.01)*** (0.01)*** (0.01) $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.09 –0.00 (0.02)*** (0.01) $$\Delta $$ Coupon 0.46 0.29 0.35 0.31 0.36 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.16 –0.10 –0.05 –0.10 –0.04 (0.13) (0.14) (0.14) (0.12) (0.13) $$\Delta $$ GO –0.30 –0.33 –0.32 –0.29 –0.28 (0.09)*** (0.09)*** (0.10)*** (0.08)*** (0.09)*** $$\Delta $$ BAB 0.61 0.61 0.64 –0.71 –0.70 (0.06)*** (0.06)*** (0.06)*** (0.05)*** (0.06)*** $$\Delta $$ Callable 0.11 0.01 0.23 0.06 0.28 (0.06)* (0.06) (0.06)*** (0.05) (0.05)*** $$\Delta $$ Negotiated –0.06 0.02 –0.07 0.05 –0.03 (0.07) (0.07) (0.08) (0.07) (0.07) Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.69 0.40 0.41 0.37 0.36 $$N$$ 865 865 865 865 865 This table displays OLS regression results with changes in various measures of offer yields and spreads as the dependent variable. The dependent variables appear at the top of each column. The dependent variables are defined in the legend of Table A.5. Panel A captures the effect of recalibration on changes in issuers’ yields and credit spreads with Upgrade, an indicator variable taking a value of one if the issuer experienced an upgrade on its outstanding bonds during any of the recalibration events and zero if the issuer’s bonds experienced no change in ratings. Other control variables are defined in the legend of Table A.5. We calculate the average rating of all outstanding bonds for each issuer before (Issuer rating pre-recalibration) and after (Issuer rating post-recalibration) the recalibration dates. Notches represents the change in this number rounded to the nearest whole number. Panel B captures the effect of recalibration on changes in issuers’ yields and credit spreads with Notches. Issuer rating pre-recalibration FE are fixed effects based on the average rating of all outstanding bonds for each issuer before its recalibration date. Issuer level of government FE are fixed effects based on whether the issuer is a state, county, city, or other. Issuer state FE are fixed effects based on the issuer’s state. Standard errors are in parentheses below coefficient estimates. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. As in Panel A of Table 4, Panel A of Table 6 captures the effect of recalibration on changes in new issue pricing with the indicator variable Upgrade. Controlling for pre-recalibration rating level, issuer state, and level of government, along with average changes in new issue characteristics, we find that upgraded municipalities face lower financing costs, relative to the control group, following recalibration. Differences in Raw yield average 15 bp, and differences in credit spreads range 18 to 22 bp. Each of these differences is significant at 5% or 1%. Panel B employs the discrete variable Notches, defined here as the difference between the Issuer rating post-recalibration and Issuer rating pre-recalibration. Controlling for the same factors—including rating level—we find that each notch of upgrade results in a reduction in yield (spread) of 11 bp (11 to 14 bp). This result implies that a municipality whose bonds were upgraded three notches would enjoy spreads on new issues that are 33 to 42 bp lower than a similar municipality whose ratings were not upgraded. Overall, we conclude from Table 6 that investors’ reliance on ratings to price bonds, combined with Moody’s dual-class rating system, was costly for U.S. taxpayers. Over $\$$ 640 billion in uninsured municipal debt was upgraded during the recalibration (Table 1), and our most conservative estimate of the average impact on new issue yields is 15 bp. The product of $\$$640 billion and 15 bp is $\$$960 million. This figure is an estimate of aggregate annual excess interest (in 2010 dollars). For context, the average cost to build a new elementary school is approximately $\$$7 million (in 2013 dollars).25 Table 6 results are robust to a wide variety of untabulated alternative specifications. These specifications include employing fixed effects for rating levels in the post-recalibration period instead of the pre-recalibration period. We also cluster standard errors by issuer state, Issuer rating pre-recalibration, and Issuer rating post-recalibration. We include linear controls for Issuer rating pre-recalibration and Issuer rating post-recalibration in lieu of rating fixed effects. Results in all of these specifications are similar in significance and magnitude to those reported in Table 6. In Table A.6, we split the sample based on rating categories in the pre- and post-recalibration periods. We find the results are robust across most categories. The noteworthy exception is the group recalibrated out of the Baa-range, for which we lack sufficient observations to estimate the model. (See the Internet Appendix for details.) We examine next the extent to which upgraded municipalities capitalized on their lower borrowing costs. We find evidence that the recalibration expanded municipal debt capacity. Figure 6 shows that, relative to issuance levels in the year before the recalibration, issuers of both upgraded and non-upgraded bonds actually issue less debt in the year after the recalibration. However, this decline in issuance activity is less pronounced for the upgraded issuers. Panel A shows that upgraded issuers experience a significantly higher increase in issuance volume compared with non-upgraded issuers two years after the recalibration. Panels B and C split the sample in Panel A by whether the bonds are GO (Panel B) or revenue bonds (Panel C). Panel B shows broadly similar patterns to those in Panel A. After two years, upgraded issuers of GO bonds issue marginally significantly more bonds than non-upgraded issuers. A stronger pattern emerges in Panel C. This panel shows that upgraded issuers of revenue bonds experience a significantly larger increase in their issuance activity compared with non-upgraded bonds in the first year after the recalibration. Overall, to the extent that increased municipal borrowing results in increased municipal investment, our results show that market reliance on credit ratings effects real investment as well as taxpayer financing costs. Figure 6 View largeDownload slide Bond issuance for upgraded and non-upgraded issuers This figure displays average issue volume of upgraded and non-upgraded issuers before and after the recalibration. For each issuer, we compute the total par value of bonds issued in the year before the recalibration. We normalize this value to one and compute total par value of new issues relative to this amount in the years after recalibration. The sample includes only municipalities that issue at least one bond in the year before, the year after, and two years after recalibration. Panel A displays issuance of all bond types. Panels B and C restrict the sample to General Obligation bonds and Revenue bonds, respectively. Dotted lines represent 95% confidence intervals. The data come from the Ipreo i-Deal new issues database. Figure 6 View largeDownload slide Bond issuance for upgraded and non-upgraded issuers This figure displays average issue volume of upgraded and non-upgraded issuers before and after the recalibration. For each issuer, we compute the total par value of bonds issued in the year before the recalibration. We normalize this value to one and compute total par value of new issues relative to this amount in the years after recalibration. The sample includes only municipalities that issue at least one bond in the year before, the year after, and two years after recalibration. Panel A displays issuance of all bond types. Panels B and C restrict the sample to General Obligation bonds and Revenue bonds, respectively. Dotted lines represent 95% confidence intervals. The data come from the Ipreo i-Deal new issues database. 3.4 Variation in information environment The SEC Office of Investor Education and Advocacy warns, “While some investors find it helpful to consider credit ratings when making an investment decision, it is important that you not rely solely on credit ratings when deciding whether to purchase municipal bonds.”26 To test the extent to which muni investors consider other sources of information, we employ a variety of proxies that capture cross-sectional variation in municipalities’ information environments. We further examine proxies for issuer opacity and corruption, assuming that information provided by corrupt municipalities is less reliable. We also consider whether S&P rated the bonds. If markets rely exclusively on ratings, then the price reaction to an upgrade should not vary by the availability of other information. In Table 7, we revisit the multivariate regression model reported in column (5) of Table 6, Panel A, using Spread to after-tax Treasury$$_{2}$$ as the dependent variable. (Results below are robust to using Raw yield or the other three measures of spreads. We report results with Spread to after-tax Treasury$$_{2}$$ because we believe it is our most precise measure of muni pricing.) Here, we include proxies for issuer opacity and corruption as independent variables. We include summary statistics for these variables in Table A.5. In Table 7, we include in columns (1) through (3) indicator variables for the issuer level of government. We conjecture that investors have better information about the financial health of larger, more transparent issuers (states) than they have for counties and cities. The variable State in column (1) compares states to the combined set of counties, cities, and other municipal issuers. The interaction term Upgrade$$\times$$State indicates that the recalibration effect documented in Table 6 is muted by 27 bp among state issuers relative to non-state issuers. In contrast, the interaction term Upgrade$$\times$$City in column (3) indicates that the recalibration effect is exacerbated among cities by an average of 17 bp relative to non-city issuers. To the extent that states (cites) are relatively transparent (opaque), these results suggest that muni market reliance on credit ratings is greatest in the weakest information environments. Table 7 Issuer-level primary market regressions with information environment analysis Level of government comparison (1) (2) (3) (4) (5) (6) (7) (8) Upgrade $$\times$$ State 0.27 (0.13)** Upgrade $$\times$$ County 0.07 (0.13) Upgrade $$\times$$ City –0.17 (0.10)* Upgrade $$\times$$ Opaque –0.16 (0.09)* Upgrade $$\times$$ Not rated by S&P –0.15 (0.08)* Upgrade $$\times$$ Corrupt$$_{\rm Risk \ index \ from \ SII}$$ –0.06 (0.11) Upgrade $$\times$$ Corrupt$$_{\rm Convictions \ 2010}$$ –0.10 (0.05)** Upgrade $$\times$$ Corrupt$$_{\rm Convictions \ 2010-2012}$$ –0.11 (0.05)** Upgrade –0.33 –0.26 –0.17 –0.16 –0.17 –0.22 –0.22 –0.21 (0.08)*** (0.07)*** (0.06)*** (0.06)*** (0.06)*** (0.09)** (0.05)*** (0.05)*** State –0.15 (0.09) County 0.03 (0.11) City –0.04 (0.10) Opaque –0.04 (0.09) Not rated by S&P 0.10 (0.07) Corrupt$$_{\rm Risk \ index \ from \ SII}$$ 0.02 (0.09) Corrupt$$_{\rm Convictions \ 2010}$$ 0.08 (0.04)** Corrupt$$_{\rm Convictions \ 2010-2012}$$ 0.08 (0.04)** $$\Delta $$ Par 0.11 0.10 0.11 0.09 0.08 0.09 0.09 0.09 (0.04)*** (0.04)*** (0.04)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.00 –0.00 –0.00 –0.01 0.00 –0.01 –0.01 –0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) $$\Delta $$ Coupon 0.30 0.30 0.29 0.28 0.24 0.29 0.29 0.29 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.02)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds 0.01 –0.00 0.01 –0.06 0.01 –0.07 –0.06 –0.06 (0.15) (0.15) (0.15) (0.14) (0.11) (0.14) (0.14) (0.15) $$\Delta $$ GO –0.21 –0.22 –0.23 –0.28 –0.21 –0.29 –0.29 –0.29 (0.08)*** (0.08)*** (0.08)*** (0.08)*** (0.06)*** (0.08)*** (0.08)*** (0.08)*** $$\Delta $$ BAB –0.52 –0.52 –0.53 –0.49 –0.46 –0.48 –0.48 –0.48 (0.05)*** (0.05)*** (0.05)*** (0.05)*** (0.04)*** (0.05)*** (0.05)*** (0.05)*** $$\Delta $$ Callable 0.12 0.12 0.13 0.17 0.13 0.18 0.18 0.18 (0.05)** (0.05)** (0.05)** (0.05)*** (0.04)*** (0.05)*** (0.05)*** (0.05)*** $$\Delta $$ Negotiated –0.10 –0.09 –0.10 –0.08 –0.07 –0.08 –0.08 –0.08 (0.07) (0.07) (0.07) (0.07) (0.05) (0.07) (0.07) (0.07) Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? No No No Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes No No No No Adjusted $$R^{\mathrm{2}}$$ 0.27 0.26 0.27 0.24 0.31 0.24 0.24 0.24 $$N$$ 865 865 865 865 865 865 865 865 Level of government comparison (1) (2) (3) (4) (5) (6) (7) (8) Upgrade $$\times$$ State 0.27 (0.13)** Upgrade $$\times$$ County 0.07 (0.13) Upgrade $$\times$$ City –0.17 (0.10)* Upgrade $$\times$$ Opaque –0.16 (0.09)* Upgrade $$\times$$ Not rated by S&P –0.15 (0.08)* Upgrade $$\times$$ Corrupt$$_{\rm Risk \ index \ from \ SII}$$ –0.06 (0.11) Upgrade $$\times$$ Corrupt$$_{\rm Convictions \ 2010}$$ –0.10 (0.05)** Upgrade $$\times$$ Corrupt$$_{\rm Convictions \ 2010-2012}$$ –0.11 (0.05)** Upgrade –0.33 –0.26 –0.17 –0.16 –0.17 –0.22 –0.22 –0.21 (0.08)*** (0.07)*** (0.06)*** (0.06)*** (0.06)*** (0.09)** (0.05)*** (0.05)*** State –0.15 (0.09) County 0.03 (0.11) City –0.04 (0.10) Opaque –0.04 (0.09) Not rated by S&P 0.10 (0.07) Corrupt$$_{\rm Risk \ index \ from \ SII}$$ 0.02 (0.09) Corrupt$$_{\rm Convictions \ 2010}$$ 0.08 (0.04)** Corrupt$$_{\rm Convictions \ 2010-2012}$$ 0.08 (0.04)** $$\Delta $$ Par 0.11 0.10 0.11 0.09 0.08 0.09 0.09 0.09 (0.04)*** (0.04)*** (0.04)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.00 –0.00 –0.00 –0.01 0.00 –0.01 –0.01 –0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) $$\Delta $$ Coupon 0.30 0.30 0.29 0.28 0.24 0.29 0.29 0.29 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.02)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds 0.01 –0.00 0.01 –0.06 0.01 –0.07 –0.06 –0.06 (0.15) (0.15) (0.15) (0.14) (0.11) (0.14) (0.14) (0.15) $$\Delta $$ GO –0.21 –0.22 –0.23 –0.28 –0.21 –0.29 –0.29 –0.29 (0.08)*** (0.08)*** (0.08)*** (0.08)*** (0.06)*** (0.08)*** (0.08)*** (0.08)*** $$\Delta $$ BAB –0.52 –0.52 –0.53 –0.49 –0.46 –0.48 –0.48 –0.48 (0.05)*** (0.05)*** (0.05)*** (0.05)*** (0.04)*** (0.05)*** (0.05)*** (0.05)*** $$\Delta $$ Callable 0.12 0.12 0.13 0.17 0.13 0.18 0.18 0.18 (0.05)** (0.05)** (0.05)** (0.05)*** (0.04)*** (0.05)*** (0.05)*** (0.05)*** $$\Delta $$ Negotiated –0.10 –0.09 –0.10 –0.08 –0.07 –0.08 –0.08 –0.08 (0.07) (0.07) (0.07) (0.07) (0.05) (0.07) (0.07) (0.07) Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? No No No Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes No No No No Adjusted $$R^{\mathrm{2}}$$ 0.27 0.26 0.27 0.24 0.31 0.24 0.24 0.24 $$N$$ 865 865 865 865 865 865 865 865 This table displays OLS regression results with $$\Delta $$Spread to after-tax Treasury$$_{2}$$ as the dependent variable. This variable is defined in the legend of Table A.5. Upgrade is an indicator variable taking a value of one if the issuer experienced an upgrade on its outstanding bonds during any of the recalibration events and zero if the issuer’s bonds experienced no change in ratings. State (County, City) is an indicator variable taking a value of one if the issuer is a state (county, city) and zero if the issuer is any other level of government. Opaque is an indicator taking a value of one if the issuer is, or is located in, a state with above-median opacity. We measure opacity with a state-level index developed by U.S. PIRG that evaluates the 50 states by the extent to which they provide online access to government spending data. The opacity index is from 2010. Not rated by S&P is an indicator variable taking a value of one if none of the issuer’s bonds were rated by S&P in the year prior to Moody’s recalibration and zero otherwise. Corrupt$$_{Risk \ index \ from \ SII}$$ is an indicator variable taking a value of one if the issuer is, or is located within, a state with above-median corruption risk. We measure corruption risk with a state-level corruption risk index developed by the State Integrity Investigation. The corruption risk index is a snapshot from 2013. Corrupt$$_{Convictions \ 2010}$$ is the number of public officials convicted of corruption in the issuer’s state in 2010 divided by the state’s population in the same year, standardized to follow a mean-zero, unit-variance distribution. Corrupt$$_{Convictions \ 2010-2012}$$ is the average number of public officials convicted of corruption in the issuer’s state from 2010 to 2012 divided by the state’s average population over the same time period, standardized to follow a mean-zero, unit-variance distribution. Convictions data are available on an annual basis from the U.S. Department of Justice’s website. Control variables are defined in the legend of Table A.5. Issuer rating pre-recalibration FE and Issuer rating post-recalibration FE are fixed effects based on the average rating of all outstanding bonds for each issuer before and after the recalibration dates, respectively. Issuer level of government FE are fixed effects based on whether the issuer is a state, county, city, or other. Issuer state FE are fixed effects based on the issuer’s state. Standard errors are in parentheses below coefficient estimates. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 7 Issuer-level primary market regressions with information environment analysis Level of government comparison (1) (2) (3) (4) (5) (6) (7) (8) Upgrade $$\times$$ State 0.27 (0.13)** Upgrade $$\times$$ County 0.07 (0.13) Upgrade $$\times$$ City –0.17 (0.10)* Upgrade $$\times$$ Opaque –0.16 (0.09)* Upgrade $$\times$$ Not rated by S&P –0.15 (0.08)* Upgrade $$\times$$ Corrupt$$_{\rm Risk \ index \ from \ SII}$$ –0.06 (0.11) Upgrade $$\times$$ Corrupt$$_{\rm Convictions \ 2010}$$ –0.10 (0.05)** Upgrade $$\times$$ Corrupt$$_{\rm Convictions \ 2010-2012}$$ –0.11 (0.05)** Upgrade –0.33 –0.26 –0.17 –0.16 –0.17 –0.22 –0.22 –0.21 (0.08)*** (0.07)*** (0.06)*** (0.06)*** (0.06)*** (0.09)** (0.05)*** (0.05)*** State –0.15 (0.09) County 0.03 (0.11) City –0.04 (0.10) Opaque –0.04 (0.09) Not rated by S&P 0.10 (0.07) Corrupt$$_{\rm Risk \ index \ from \ SII}$$ 0.02 (0.09) Corrupt$$_{\rm Convictions \ 2010}$$ 0.08 (0.04)** Corrupt$$_{\rm Convictions \ 2010-2012}$$ 0.08 (0.04)** $$\Delta $$ Par 0.11 0.10 0.11 0.09 0.08 0.09 0.09 0.09 (0.04)*** (0.04)*** (0.04)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.00 –0.00 –0.00 –0.01 0.00 –0.01 –0.01 –0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) $$\Delta $$ Coupon 0.30 0.30 0.29 0.28 0.24 0.29 0.29 0.29 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.02)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds 0.01 –0.00 0.01 –0.06 0.01 –0.07 –0.06 –0.06 (0.15) (0.15) (0.15) (0.14) (0.11) (0.14) (0.14) (0.15) $$\Delta $$ GO –0.21 –0.22 –0.23 –0.28 –0.21 –0.29 –0.29 –0.29 (0.08)*** (0.08)*** (0.08)*** (0.08)*** (0.06)*** (0.08)*** (0.08)*** (0.08)*** $$\Delta $$ BAB –0.52 –0.52 –0.53 –0.49 –0.46 –0.48 –0.48 –0.48 (0.05)*** (0.05)*** (0.05)*** (0.05)*** (0.04)*** (0.05)*** (0.05)*** (0.05)*** $$\Delta $$ Callable 0.12 0.12 0.13 0.17 0.13 0.18 0.18 0.18 (0.05)** (0.05)** (0.05)** (0.05)*** (0.04)*** (0.05)*** (0.05)*** (0.05)*** $$\Delta $$ Negotiated –0.10 –0.09 –0.10 –0.08 –0.07 –0.08 –0.08 –0.08 (0.07) (0.07) (0.07) (0.07) (0.05) (0.07) (0.07) (0.07) Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? No No No Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes No No No No Adjusted $$R^{\mathrm{2}}$$ 0.27 0.26 0.27 0.24 0.31 0.24 0.24 0.24 $$N$$ 865 865 865 865 865 865 865 865 Level of government comparison (1) (2) (3) (4) (5) (6) (7) (8) Upgrade $$\times$$ State 0.27 (0.13)** Upgrade $$\times$$ County 0.07 (0.13) Upgrade $$\times$$ City –0.17 (0.10)* Upgrade $$\times$$ Opaque –0.16 (0.09)* Upgrade $$\times$$ Not rated by S&P –0.15 (0.08)* Upgrade $$\times$$ Corrupt$$_{\rm Risk \ index \ from \ SII}$$ –0.06 (0.11) Upgrade $$\times$$ Corrupt$$_{\rm Convictions \ 2010}$$ –0.10 (0.05)** Upgrade $$\times$$ Corrupt$$_{\rm Convictions \ 2010-2012}$$ –0.11 (0.05)** Upgrade –0.33 –0.26 –0.17 –0.16 –0.17 –0.22 –0.22 –0.21 (0.08)*** (0.07)*** (0.06)*** (0.06)*** (0.06)*** (0.09)** (0.05)*** (0.05)*** State –0.15 (0.09) County 0.03 (0.11) City –0.04 (0.10) Opaque –0.04 (0.09) Not rated by S&P 0.10 (0.07) Corrupt$$_{\rm Risk \ index \ from \ SII}$$ 0.02 (0.09) Corrupt$$_{\rm Convictions \ 2010}$$ 0.08 (0.04)** Corrupt$$_{\rm Convictions \ 2010-2012}$$ 0.08 (0.04)** $$\Delta $$ Par 0.11 0.10 0.11 0.09 0.08 0.09 0.09 0.09 (0.04)*** (0.04)*** (0.04)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Duration$$_{\mathrm{2}}$$ –0.00 –0.00 –0.00 –0.01 0.00 –0.01 –0.01 –0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) $$\Delta $$ Coupon 0.30 0.30 0.29 0.28 0.24 0.29 0.29 0.29 (0.03)*** (0.03)*** (0.03)*** (0.03)*** (0.02)*** (0.03)*** (0.03)*** (0.03)*** $$\Delta $$ Outstanding bonds 0.01 –0.00 0.01 –0.06 0.01 –0.07 –0.06 –0.06 (0.15) (0.15) (0.15) (0.14) (0.11) (0.14) (0.14) (0.15) $$\Delta $$ GO –0.21 –0.22 –0.23 –0.28 –0.21 –0.29 –0.29 –0.29 (0.08)*** (0.08)*** (0.08)*** (0.08)*** (0.06)*** (0.08)*** (0.08)*** (0.08)*** $$\Delta $$ BAB –0.52 –0.52 –0.53 –0.49 –0.46 –0.48 –0.48 –0.48 (0.05)*** (0.05)*** (0.05)*** (0.05)*** (0.04)*** (0.05)*** (0.05)*** (0.05)*** $$\Delta $$ Callable 0.12 0.12 0.13 0.17 0.13 0.18 0.18 0.18 (0.05)** (0.05)** (0.05)** (0.05)*** (0.04)*** (0.05)*** (0.05)*** (0.05)*** $$\Delta $$ Negotiated –0.10 –0.09 –0.10 –0.08 –0.07 –0.08 –0.08 –0.08 (0.07) (0.07) (0.07) (0.07) (0.05) (0.07) (0.07) (0.07) Issuer rating pre-recal. FE? Yes Yes Yes Yes Yes Yes Yes Yes Issuer level of govt. FE? No No No Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes No No No No Adjusted $$R^{\mathrm{2}}$$ 0.27 0.26 0.27 0.24 0.31 0.24 0.24 0.24 $$N$$ 865 865 865 865 865 865 865 865 This table displays OLS regression results with $$\Delta $$Spread to after-tax Treasury$$_{2}$$ as the dependent variable. This variable is defined in the legend of Table A.5. Upgrade is an indicator variable taking a value of one if the issuer experienced an upgrade on its outstanding bonds during any of the recalibration events and zero if the issuer’s bonds experienced no change in ratings. State (County, City) is an indicator variable taking a value of one if the issuer is a state (county, city) and zero if the issuer is any other level of government. Opaque is an indicator taking a value of one if the issuer is, or is located in, a state with above-median opacity. We measure opacity with a state-level index developed by U.S. PIRG that evaluates the 50 states by the extent to which they provide online access to government spending data. The opacity index is from 2010. Not rated by S&P is an indicator variable taking a value of one if none of the issuer’s bonds were rated by S&P in the year prior to Moody’s recalibration and zero otherwise. Corrupt$$_{Risk \ index \ from \ SII}$$ is an indicator variable taking a value of one if the issuer is, or is located within, a state with above-median corruption risk. We measure corruption risk with a state-level corruption risk index developed by the State Integrity Investigation. The corruption risk index is a snapshot from 2013. Corrupt$$_{Convictions \ 2010}$$ is the number of public officials convicted of corruption in the issuer’s state in 2010 divided by the state’s population in the same year, standardized to follow a mean-zero, unit-variance distribution. Corrupt$$_{Convictions \ 2010-2012}$$ is the average number of public officials convicted of corruption in the issuer’s state from 2010 to 2012 divided by the state’s average population over the same time period, standardized to follow a mean-zero, unit-variance distribution. Convictions data are available on an annual basis from the U.S. Department of Justice’s website. Control variables are defined in the legend of Table A.5. Issuer rating pre-recalibration FE and Issuer rating post-recalibration FE are fixed effects based on the average rating of all outstanding bonds for each issuer before and after the recalibration dates, respectively. Issuer level of government FE are fixed effects based on whether the issuer is a state, county, city, or other. Issuer state FE are fixed effects based on the issuer’s state. Standard errors are in parentheses below coefficient estimates. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Column (4) employs the indicator variable Opaque, which takes the value of one if the issuer is, or is located in, a state with above-median opacity according to a state-level index developed by U.S. Public Interest Research Group.27 Consistent with variation in market reliance on ratings across information environments, the interaction term Upgrade$$\times$$Opaque indicates that the recalibration effect is stronger among issuers from states with below-median disclosure. We consider in column (5) whether the upgraded issuers are also rated by S&P. Butler (2008) argues that nonrated bonds are harder for underwriters to sell. Although all of the bonds in our sample are rated by Moody’s, the information provided by Moody’s should be especially important when S&P does not provide a rating. Indeed, the interaction term Upgrade$$\times$$Not rated by S&P indicates that the market reliance on Moody’s ratings is stronger (15 bp significant at 10%) among the set for which no guidance is provided by S&P. Columns (6) through (8) bifurcate the sample based on measures of corruption risk. Butler, Fauver, and Mortal (2009) document that higher state corruption is associated with greater credit risk and bond yields. Ang et al. (Forthcoming) further document that higher state corruption is associated with greater value destruction by advanced refunding. We posit that corrupt states have less reliable disclosure and test whether market participants therefore rely more heavily on ratings when pricing new issues from these states. Column (6) employs a measure of corruption risk provided by the State Integrity Investigation (SII).28 Columns (7) and (8) consider the number of public officials convicted of corruption, obtained from the U.S. Department of Justice. In column (7), we focus on corruption convictions in the year of the recalibration event. In column (8), we consider the conviction rate over the 2010–12 window. We use convictions data from years shortly after the recalibration event because convictions represent the culmination of the legal process (e.g., arrests and indictments that occurred in earlier years). Therefore, convictions from 2010 to 2012 may reflect corruption that came to light around the time of the recalibration. We standardize each proxy to follow a mean-zero, unit-variance distribution. The direct effect of corruption risk in columns (7) and (8) is an 8-bp increase in credit spreads, significant at 5%. The interaction terms in these specifications indicate that Moody’s recalibration effect is stronger (significant at 5%) among municipal issuers from corrupt states. This result suggests a certification role of the CRA in a manner similar to the underwriter certification role in an initial public offering (IPO) of a previously private and therefore opaque firm. Overall, results from Table 7 support the conclusion that market reliance on credit ratings appears greatest in the weakest information environments. Because bond prices reflect information not captured in the credit rating, our results suggest some level of investor due diligence even in a market dominated by retail investors. 3.5 Regulatory effects of credit ratings Because our research setting is the municipal bond market, a market dominated by unregulated investors, we conclude that the market reaction to Moody’s recalibration cannot be attributed to an increase in demand for highly rated securities among regulated investors. However, many observers feel that ratings matter primarily, if not only, because of their regulatory implications. We attempt to provide direct tests for this explanation in this section. We start by categorizing upgrades based on their likely regulatory implications. Specifically, reserve requirements and other ratings-based regulations are typically written around broad rating categories (e.g., Aaa, Aa, A, Baa) rather than notches (e.g., Aa1, Aa2, Aa3). In order to isolate the effect of a regulatory threshold from magnitude of upgrade, we examine one-notch upgrades separately from two-notch upgrades.29 We first restrict our treatment group to one-notch upgrades and separate bonds upgraded into a new broad rating category (e.g., from A1 to Aa3) from those remaining in the original broad category (e.g., A2 to A1). Separately, we restrict our treatment group to two-notch upgrades and separate bonds upgraded into a new broad rating category (e.g., from A2 to Aa3) from those that remain in the original broad rating category (e.g., A3 to A1). In either case, the upgrades crossing a broad rating category should have greater regulatory implications than upgrades of the same magnitude that do not cross a threshold. We test the regulatory hypothesis in Table 8. In Panel A, we compare changes in secondary market spreads (relative to the control group) in a multivariate framework with fixed effects (as in Table 4). We find no differential impact of the recalibration among upgraded bonds that do and do not cross into new broad rating categories. These non-results corroborate the idea that our main results do not reflect changes in regulation-based demand. Table 8 The effect of recalibrated ratings on regulation-based demand Panel A: Secondary market regressions 0 or 1 notches 0 or 2 notches (1) (2) Post recal. $$\times$$ Upgrade $$\times$$ Reg. status change 0.02 0.03 (0.04) (0.08) Post recalibration $$\times$$ Upgrade –0.07 0.02 (0.03)** (0.06) Upgrade $$\times$$ Regulatory status change –0.05 –0.26 (0.17) (0.34) Post recalibration 0.10 0.10 (0.01)*** (0.01)*** Upgrade –0.26 0.47 (0.15) (0.28) Par 0.07 0.08 (0.04) (0.04)* Duration$$_{\mathrm{2}}$$ 0.11 0.10 (0.01)*** (0.01)*** Coupon 0.05 0.09 (0.04) (0.07) Outstanding bonds –0.03 0.00 (0.04) (0.05) GO –0.18 –0.41 (0.14) (0.19)** BAB –1.11 –1.18 (0.11)*** (0.16)*** Callable 0.49 0.34 (0.11)*** (0.14) Negotiated 0.18 0.09 (0.12) (0.13) Issue rating pre-recalibration FE? Yes Yes Issuer level of government FE? Yes Yes Issuer state FE? Yes Yes Moody’s sector FE? Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.60 0.60 $$N$$ 2,356 1,618 Panel A: Secondary market regressions 0 or 1 notches 0 or 2 notches (1) (2) Post recal. $$\times$$ Upgrade $$\times$$ Reg. status change 0.02 0.03 (0.04) (0.08) Post recalibration $$\times$$ Upgrade –0.07 0.02 (0.03)** (0.06) Upgrade $$\times$$ Regulatory status change –0.05 –0.26 (0.17) (0.34) Post recalibration 0.10 0.10 (0.01)*** (0.01)*** Upgrade –0.26 0.47 (0.15) (0.28) Par 0.07 0.08 (0.04) (0.04)* Duration$$_{\mathrm{2}}$$ 0.11 0.10 (0.01)*** (0.01)*** Coupon 0.05 0.09 (0.04) (0.07) Outstanding bonds –0.03 0.00 (0.04) (0.05) GO –0.18 –0.41 (0.14) (0.19)** BAB –1.11 –1.18 (0.11)*** (0.16)*** Callable 0.49 0.34 (0.11)*** (0.14) Negotiated 0.18 0.09 (0.12) (0.13) Issue rating pre-recalibration FE? Yes Yes Issuer level of government FE? Yes Yes Issuer state FE? Yes Yes Moody’s sector FE? Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.60 0.60 $$N$$ 2,356 1,618 Panel B: Issuer-level primary market regressions Criteria used to define Regulatory status change: i. i. ii. i. $$\bigcap$$ ii. i. $$\bigcup$$ ii. Number of notches issuers are upgraded: 0 or 1 0 or 2 0 or 1 0 or 1 0 or 1 (1) (2) (3) (4) (5) Upgrade $$\times$$ Reg. status change –0.15 0.11 –0.04 –0.27 –0.10 (0.07)** (0.14) (0.08) (0.16) (0.06) Upgrade –0.14 –0.24 –0.17 –0.15 –0.11 (0.05)*** (0.14)* (0.08)** (0.09)* (0.06)* $$\Delta $$ Par 0.08 0.13 0.01 –0.03 0.08 (0.04)** (0.05)*** (0.05) (0.06) (0.04)** $$\Delta $$ Duration$$_{2}$$ –0.02 –0.03 0.06 0.04 0.00 (0.02) (0.02) (0.02)*** (0.02)* (0.02) $$\Delta $$ Coupon 0.40 0.45 0.29 0.43 0.36 (0.03)*** (0.04)*** (0.04)*** (0.06)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.02 –0.06 –0.13 –0.59 –0.01 (0.14) (0.18) (0.26) (0.33)* (0.14) $$\Delta $$ GO –0.29 –0.21 –0.35 –0.23 –0.36 (0.11)*** (0.12)* (0.12)*** (0.16) (0.10)*** $$\Delta $$ BAB –0.69 –0.76 –0.70 –0.64 –0.69 (0.06)*** (0.09)*** (0.08)*** (0.10)*** (0.06)*** $$\Delta $$ Callable 0.31 0.32 0.25 0.34 0.28 (0.06)*** (0.08)*** (0.08)*** (0.10)*** (0.06)*** $$\Delta $$ Negotiated –0.04 0.01 –0.08 –0.04 –0.04 (0.08) (0.10) (0.11) (0.14) (0.08) Constant 0.44 0.38 0.53 0.60 0.44 (0.06)*** (0.08)*** (0.08)*** (0.10)*** (0.06)*** Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{2}$$ 0.38 0.40 0.32 0.38 0.35 $$N$$ 704 428 500 318 773 Panel B: Issuer-level primary market regressions Criteria used to define Regulatory status change: i. i. ii. i. $$\bigcap$$ ii. i. $$\bigcup$$ ii. Number of notches issuers are upgraded: 0 or 1 0 or 2 0 or 1 0 or 1 0 or 1 (1) (2) (3) (4) (5) Upgrade $$\times$$ Reg. status change –0.15 0.11 –0.04 –0.27 –0.10 (0.07)** (0.14) (0.08) (0.16) (0.06) Upgrade –0.14 –0.24 –0.17 –0.15 –0.11 (0.05)*** (0.14)* (0.08)** (0.09)* (0.06)* $$\Delta $$ Par 0.08 0.13 0.01 –0.03 0.08 (0.04)** (0.05)*** (0.05) (0.06) (0.04)** $$\Delta $$ Duration$$_{2}$$ –0.02 –0.03 0.06 0.04 0.00 (0.02) (0.02) (0.02)*** (0.02)* (0.02) $$\Delta $$ Coupon 0.40 0.45 0.29 0.43 0.36 (0.03)*** (0.04)*** (0.04)*** (0.06)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.02 –0.06 –0.13 –0.59 –0.01 (0.14) (0.18) (0.26) (0.33)* (0.14) $$\Delta $$ GO –0.29 –0.21 –0.35 –0.23 –0.36 (0.11)*** (0.12)* (0.12)*** (0.16) (0.10)*** $$\Delta $$ BAB –0.69 –0.76 –0.70 –0.64 –0.69 (0.06)*** (0.09)*** (0.08)*** (0.10)*** (0.06)*** $$\Delta $$ Callable 0.31 0.32 0.25 0.34 0.28 (0.06)*** (0.08)*** (0.08)*** (0.10)*** (0.06)*** $$\Delta $$ Negotiated –0.04 0.01 –0.08 –0.04 –0.04 (0.08) (0.10) (0.11) (0.14) (0.08) Constant 0.44 0.38 0.53 0.60 0.44 (0.06)*** (0.08)*** (0.08)*** (0.10)*** (0.06)*** Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{2}$$ 0.38 0.40 0.32 0.38 0.35 $$N$$ 704 428 500 318 773 Panel A displays OLS regression results with Spread to after-tax Treasury$$_{2}$$ as the dependent variable. This variable is defined in the legend of Table A.2 and is averaged across transactions by bond either during the 30-day window before Moody’s published the Primary Algorithm or after the bond’s recalibration date. We require the bonds to have trading in both 30-day periods before Moody’s published the Primary Algorithm and after the bond’s recalibration date to be included in the regressions. Post recalibration is an indicator variable taking a value of one if the observation is from the 30-day window after the bond’s recalibration date and zero if the observation is from the 30-day period prior to Moody’s publication of the Primary Algorithm. We exclude data from the day of the bond’s recalibration. Upgrade is an indicator variable taking a value of one if the bond experienced an upgrade during the recalibration event and zero if the bond experienced no change in ratings. Regulatory status change is an indicator variable taking a value of zero for bonds that were not upgraded as a result of recalibration or were upgraded but remained within the same broad rating category. Other control variables are defined in the legend of Table A.2. Issue rating pre-recalibration FE are fixed effects based on the bond’s rating before the bond’s recalibration. Issuer level of government FE are fixed effects based on whether the issuer is a state, county, city, or other. Issuer state FE are fixed effects based on the issuer’s state. Moody’s sector FE are fixed effects based on the four sectors into which Moody’s classifies bonds in its Primary Algorithm. We obtain municipal bond prices from the Municipal Securities Rulemaking Board (MSRB). We cluster standard errors by issuer. Panel B displays OLS regression results with $$\Delta $$Spread to after-tax Treasury$$_{2}$$ as the dependent variable. This variable is defined in the legend of Table A.5. For this panel, Upgrade is an indicator variable taking a value of one if the issuer experienced an upgrade on its outstanding bonds during any of the recalibration events and zero if the issuer’s bonds experienced no change in ratings. In column 1 (column 2) Regulatory status change is an indicator variable taking a value of one if the issuer’s bonds changed regulatory status as a result of recalibration and zero if not. We use two criteria to identify issuers whose bonds change regulatory status. For criteria i., Regulatory status change is an indicator variable taking a value of one if the issuer’s bonds were upgraded one notch as a result of recalibration and the rating change crossed into a new broad rating category. Under criteria i, Regulatory status change takes a vale of zero for issuers whose bonds were not upgraded as a result of recalibration or whose bonds were upgraded one notch but remained within the same broad rating category. For criteria ii, Regulatory status change is an indicator variable taking a value of one if the issuer’s bonds had more pessimistic ratings from Moody’s than S&P prior to recalibration and equal to or more optimistic than S&P after recalibration. Under criteria ii, Regulatory status change takes a value of zero for issuers whose bonds were not upgraded as a result of recalibration, upgraded issuers whose bonds have more pessimistic ratings from Moody’s than S&P after recalibration, or upgraded issuers whose bonds had equal or more optimistic ratings from Moody’s than S&P before recalibration. The regression in column (4) defines Regulatory status change as the intersection of criteria i and ii. The regression in column (5) defines Regulatory status change as the union of criteria i and ii. Control variables are defined in the legend of Table A.5. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. Table 8 The effect of recalibrated ratings on regulation-based demand Panel A: Secondary market regressions 0 or 1 notches 0 or 2 notches (1) (2) Post recal. $$\times$$ Upgrade $$\times$$ Reg. status change 0.02 0.03 (0.04) (0.08) Post recalibration $$\times$$ Upgrade –0.07 0.02 (0.03)** (0.06) Upgrade $$\times$$ Regulatory status change –0.05 –0.26 (0.17) (0.34) Post recalibration 0.10 0.10 (0.01)*** (0.01)*** Upgrade –0.26 0.47 (0.15) (0.28) Par 0.07 0.08 (0.04) (0.04)* Duration$$_{\mathrm{2}}$$ 0.11 0.10 (0.01)*** (0.01)*** Coupon 0.05 0.09 (0.04) (0.07) Outstanding bonds –0.03 0.00 (0.04) (0.05) GO –0.18 –0.41 (0.14) (0.19)** BAB –1.11 –1.18 (0.11)*** (0.16)*** Callable 0.49 0.34 (0.11)*** (0.14) Negotiated 0.18 0.09 (0.12) (0.13) Issue rating pre-recalibration FE? Yes Yes Issuer level of government FE? Yes Yes Issuer state FE? Yes Yes Moody’s sector FE? Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.60 0.60 $$N$$ 2,356 1,618 Panel A: Secondary market regressions 0 or 1 notches 0 or 2 notches (1) (2) Post recal. $$\times$$ Upgrade $$\times$$ Reg. status change 0.02 0.03 (0.04) (0.08) Post recalibration $$\times$$ Upgrade –0.07 0.02 (0.03)** (0.06) Upgrade $$\times$$ Regulatory status change –0.05 –0.26 (0.17) (0.34) Post recalibration 0.10 0.10 (0.01)*** (0.01)*** Upgrade –0.26 0.47 (0.15) (0.28) Par 0.07 0.08 (0.04) (0.04)* Duration$$_{\mathrm{2}}$$ 0.11 0.10 (0.01)*** (0.01)*** Coupon 0.05 0.09 (0.04) (0.07) Outstanding bonds –0.03 0.00 (0.04) (0.05) GO –0.18 –0.41 (0.14) (0.19)** BAB –1.11 –1.18 (0.11)*** (0.16)*** Callable 0.49 0.34 (0.11)*** (0.14) Negotiated 0.18 0.09 (0.12) (0.13) Issue rating pre-recalibration FE? Yes Yes Issuer level of government FE? Yes Yes Issuer state FE? Yes Yes Moody’s sector FE? Yes Yes Adjusted $$R^{\mathrm{2}}$$ 0.60 0.60 $$N$$ 2,356 1,618 Panel B: Issuer-level primary market regressions Criteria used to define Regulatory status change: i. i. ii. i. $$\bigcap$$ ii. i. $$\bigcup$$ ii. Number of notches issuers are upgraded: 0 or 1 0 or 2 0 or 1 0 or 1 0 or 1 (1) (2) (3) (4) (5) Upgrade $$\times$$ Reg. status change –0.15 0.11 –0.04 –0.27 –0.10 (0.07)** (0.14) (0.08) (0.16) (0.06) Upgrade –0.14 –0.24 –0.17 –0.15 –0.11 (0.05)*** (0.14)* (0.08)** (0.09)* (0.06)* $$\Delta $$ Par 0.08 0.13 0.01 –0.03 0.08 (0.04)** (0.05)*** (0.05) (0.06) (0.04)** $$\Delta $$ Duration$$_{2}$$ –0.02 –0.03 0.06 0.04 0.00 (0.02) (0.02) (0.02)*** (0.02)* (0.02) $$\Delta $$ Coupon 0.40 0.45 0.29 0.43 0.36 (0.03)*** (0.04)*** (0.04)*** (0.06)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.02 –0.06 –0.13 –0.59 –0.01 (0.14) (0.18) (0.26) (0.33)* (0.14) $$\Delta $$ GO –0.29 –0.21 –0.35 –0.23 –0.36 (0.11)*** (0.12)* (0.12)*** (0.16) (0.10)*** $$\Delta $$ BAB –0.69 –0.76 –0.70 –0.64 –0.69 (0.06)*** (0.09)*** (0.08)*** (0.10)*** (0.06)*** $$\Delta $$ Callable 0.31 0.32 0.25 0.34 0.28 (0.06)*** (0.08)*** (0.08)*** (0.10)*** (0.06)*** $$\Delta $$ Negotiated –0.04 0.01 –0.08 –0.04 –0.04 (0.08) (0.10) (0.11) (0.14) (0.08) Constant 0.44 0.38 0.53 0.60 0.44 (0.06)*** (0.08)*** (0.08)*** (0.10)*** (0.06)*** Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{2}$$ 0.38 0.40 0.32 0.38 0.35 $$N$$ 704 428 500 318 773 Panel B: Issuer-level primary market regressions Criteria used to define Regulatory status change: i. i. ii. i. $$\bigcap$$ ii. i. $$\bigcup$$ ii. Number of notches issuers are upgraded: 0 or 1 0 or 2 0 or 1 0 or 1 0 or 1 (1) (2) (3) (4) (5) Upgrade $$\times$$ Reg. status change –0.15 0.11 –0.04 –0.27 –0.10 (0.07)** (0.14) (0.08) (0.16) (0.06) Upgrade –0.14 –0.24 –0.17 –0.15 –0.11 (0.05)*** (0.14)* (0.08)** (0.09)* (0.06)* $$\Delta $$ Par 0.08 0.13 0.01 –0.03 0.08 (0.04)** (0.05)*** (0.05) (0.06) (0.04)** $$\Delta $$ Duration$$_{2}$$ –0.02 –0.03 0.06 0.04 0.00 (0.02) (0.02) (0.02)*** (0.02)* (0.02) $$\Delta $$ Coupon 0.40 0.45 0.29 0.43 0.36 (0.03)*** (0.04)*** (0.04)*** (0.06)*** (0.03)*** $$\Delta $$ Outstanding bonds –0.02 –0.06 –0.13 –0.59 –0.01 (0.14) (0.18) (0.26) (0.33)* (0.14) $$\Delta $$ GO –0.29 –0.21 –0.35 –0.23 –0.36 (0.11)*** (0.12)* (0.12)*** (0.16) (0.10)*** $$\Delta $$ BAB –0.69 –0.76 –0.70 –0.64 –0.69 (0.06)*** (0.09)*** (0.08)*** (0.10)*** (0.06)*** $$\Delta $$ Callable 0.31 0.32 0.25 0.34 0.28 (0.06)*** (0.08)*** (0.08)*** (0.10)*** (0.06)*** $$\Delta $$ Negotiated –0.04 0.01 –0.08 –0.04 –0.04 (0.08) (0.10) (0.11) (0.14) (0.08) Constant 0.44 0.38 0.53 0.60 0.44 (0.06)*** (0.08)*** (0.08)*** (0.10)*** (0.06)*** Issuer level of govt. FE? Yes Yes Yes Yes Yes Issuer state FE? Yes Yes Yes Yes Yes Adjusted $$R^{2}$$ 0.38 0.40 0.32 0.38 0.35 $$N$$ 704 428 500 318 773 Panel A displays OLS regression results with Spread to after-tax Treasury$$_{2}$$ as the dependent variable. This variable is defined in the legend of Table A.2 and is averaged across transactions by bond either during the 30-day window before Moody’s published the Primary Algorithm or after the bond’s recalibration date. We require the bonds to have trading in both 30-day periods before Moody’s published the Primary Algorithm and after the bond’s recalibration date to be included in the regressions. Post recalibration is an indicator variable taking a value of one if the observation is from the 30-day window after the bond’s recalibration date and zero if the observation is from the 30-day period prior to Moody’s publication of the Primary Algorithm. We exclude data from the day of the bond’s recalibration. Upgrade is an indicator variable taking a value of one if the bond experienced an upgrade during the recalibration event and zero if the bond experienced no change in ratings. Regulatory status change is an indicator variable taking a value of zero for bonds that were not upgraded as a result of recalibration or were upgraded but remained within the same broad rating category. Other control variables are defined in the legend of Table A.2. Issue rating pre-recalibration FE are fixed effects based on the bond’s rating before the bond’s recalibration. Issuer level of government FE are fixed effects based on whether the issuer is a state, county, city, or other. Issuer state FE are fixed effects based on the issuer’s state. Moody’s sector FE are fixed effects based on the four sectors into which Moody’s classifies bonds in its Primary Algorithm. We obtain municipal bond prices from the Municipal Securities Rulemaking Board (MSRB). We cluster standard errors by issuer. Panel B displays OLS regression results with $$\Delta $$Spread to after-tax Treasury$$_{2}$$ as the dependent variable. This variable is defined in the legend of Table A.5. For this panel, Upgrade is an indicator variable taking a value of one if the issuer experienced an upgrade on its outstanding bonds during any of the recalibration events and zero if the issuer’s bonds experienced no change in ratings. In column 1 (column 2) Regulatory status change is an indicator variable taking a value of one if the issuer’s bonds changed regulatory status as a result of recalibration and zero if not. We use two criteria to identify issuers whose bonds change regulatory status. For criteria i., Regulatory status change is an indicator variable taking a value of one if the issuer’s bonds were upgraded one notch as a result of recalibration and the rating change crossed into a new broad rating category. Under criteria i, Regulatory status change takes a vale of zero for issuers whose bonds were not upgraded as a result of recalibration or whose bonds were upgraded one notch but remained within the same broad rating category. For criteria ii, Regulatory status change is an indicator variable taking a value of one if the issuer’s bonds had more pessimistic ratings from Moody’s than S&P prior to recalibration and equal to or more optimistic than S&P after recalibration. Under criteria ii, Regulatory status change takes a value of zero for issuers whose bonds were not upgraded as a result of recalibration, upgraded issuers whose bonds have more pessimistic ratings from Moody’s than S&P after recalibration, or upgraded issuers whose bonds had equal or more optimistic ratings from Moody’s than S&P before recalibration. The regression in column (4) defines Regulatory status change as the intersection of criteria i and ii. The regression in column (5) defines Regulatory status change as the union of criteria i and ii. Control variables are defined in the legend of Table A.5. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively. In Panel B, we compare the primary market credit spreads at the issuer level (relative to the control group with zero-notch upgrades) in a multivariate framework, as in Table 6. The direct effect Regulatory status change drops from the regressions because this variable does not vary by issuer from the pre-recalibration period to the post-recalibration period. We also exclude Issuer rating pre-recalibration FE because this characteristic correlates with whether upgraded issuers change regulatory status. (For example, upgraded issuers with pre-recalibration ratings of Aa1, Moody’s highest rating in the Aa category, necessarily change regulatory status.) In this panel, column (1) shows a greater impact of Moody’s recalibration among issuers whose bonds were upgraded one notch across a regulatory threshold relative to those with one notch upgrades that remained within a broad rating category. This 15-bp difference is significant at 5% and is the only evidence we have to support a regulatory channel by which credit ratings impact muni prices. Column (2) shows no significant result among the two-notch upgrades. Given (limited) evidence to support the regulatory channel in column (1), we further test for regulatory effects among this sample. Specifically, some ratings-based regulations include “lowest rating binds” provisions. We test for effects of such regulations by adjusting the criteria for Regulatory status change. We refer to the criteria used thus far as “criteria i.” Under “criteria ii,” the variable Regulatory status change takes a value of one if the issuer’s bonds had more pessimistic ratings from Moody’s than S&P prior to recalibration and equal to or more optimistic than S&P after recalibration. Regulatory status change takes a value of zero for issuers whose bonds were not upgraded as a result of recalibration, upgraded issuers whose bonds still have more pessimistic ratings from Moody’s than S&P after recalibration, or upgraded issuers whose bonds had equal or more optimistic ratings from Moody’s than S&P before recalibration. We introduce Regulatory status change under criteria ii to our baseline regression and interact it with Upgrade. We focus on issuers whose bonds were upgraded zero notches or one notch because this sample has more observations than the sample of issuers upgraded zero or two notches. If the effect of the recalibration is stronger among upgraded issuers whose bonds experienced a change in regulatory status, then the coefficient on this interaction term should be negative. Although the sign is negative, the coefficient on Upgrade$$\times$$Regulatory status change is insignificant. The result appears in column (3) in Panel B. We conduct two additional tests to address the possibility that criteria i and criteria ii overlap. First, we redefine Regulatory status change as the intersection of criteria i and ii. This approach allows us to test for regulatory effects among a select group of issuers whose bonds almost certainly changed regulatory status. Column (4) in Panel C contains the results. The coefficient on Upgrade$$\times$$Regulatory status change is large (–27 bp) but insignificant. Next, we redefine Regulatory status change as the union of criteria i and ii. This approach liberalizes the definition of which issuers’ bonds changed regulatory status. The coefficient on Upgrade$$\times$$Regulatory status change is negative but insignificant. Overall, Table 8 provides at most limited support for a regulatory effect of the recalibration, no matter how we define a change in regulatory status. 4. Conclusion Our contribution to the literature is evidence that credit ratings affect securities prices because investors rely on them to assess credit risk, not just because ratings affect regulatory compliance costs. This is the core and classic function of credit ratings. Despite advances in information technology and a rise in alternative sources for credit risk analysis, we show that investors still rely on ratings in the modern era. However, in contrast to an exclusive reliance on ratings, we also find that investors rely most heavily on credit ratings in the poorest information environments. Finally, we document that investor reliance on credit ratings has real economic effects for issuers and taxpayers. This evidence is important for several reasons. First, the premise that investors take ratings at face value underlies a host of high-profile lawsuits following the recent financial crisis, including cases brought against Moody’s and S&P by the U.S. Department of Justice. Second, our results indicate that efforts to reduce regulatory reliance on ratings (FSB Principle I and Dodd-Frank Section 939A) will not negate the role of credit rating agencies in the economy. We conclude that regulators concerned about mechanistic reliance on ratings—and municipal issuers concerned about the fairness of rating standards that vary across asset classes—should focus on increasing and standardizing disclosure requirements across state and local governments. Such disclosure improvements should further support regulatory efforts to protect retail investors participating in muni markets.30 Our results from the municipal market suggest similar benefits to increased transparency among currently opaque special-purpose entities issuing structure finance products. Finally, we estimate that Moody’s more stringent municipal rating scale, coupled with investor reliance on ratings to price risk, cost taxpayers an aggregate $\$$ 960 million annually in excess interest. This estimate is conservative as it reflects only the aggregate excess interest of our sample of uninsured bonds. A more comprehensive estimate would include any excess insurance premiums paid by municipalities choosing to purchase bond insurance rather than pay higher interest rates. This estimate is timely as the U.S. U.S. Securities and Exchange Commission (2011) considers the Dodd-Frank mandate to standardize credit ratings across all rated securities (see § 938). This estimate should also be of interest to the states, cities, and public agencies that have called for an end to the dual-class rating system. For example, the State of Connecticut recently filed (and settled) a lawsuit against the rating agencies, stating, “The credit rating agencies and bond insurers have enjoyed enormous profits, at the expense of taxpayers, as a result of this deceptive dual rating system. The harm to taxpayers across the country is real and substantial.”31. Acknowledgements We thank two anonymous referees, Matt Billett, Alex Butler, Richard Cantor, Igor Cunha, Jennifer Dlugosz, Miguel Ferreira, Rick Green, John Griffin, Andrew Karolyi, Laura Levenstein, Alfred Medioli, Michael Piwowar, Matt Spiegel, Merxe Tudela, Anjan Thakor, Charles Trzcinka, and audience members at American University, Bank of Canada, George Mason University, George Washington University, Georgetown University, Indiana University, Penn State University, Texas Tech University, the University of Amsterdam, the University of Georgia, University of Southern California, the 2013 Bond Buyer Brandeis University Municipal Finance conference, the 2014 NBER Workshop on Credit Rating Agencies, and the 2016 Chicago Financial Institutions Conference for helpful comments. Any errors belong to the authors. Supplementary data can be found on The Review of Financial Studies web site. Footnotes 1 Prior papers documenting a correlation between credit ratings and security prices include Hettenhouse and Sartoris (1976), Weinstein (1977), Pinches and Singleton (1978), Ingram, Brooks, and Copeland (1983), Holthausen and Leftwich (1986), Hand, Holthausen, and Leftwich (1992), Goh and Ederington (1993), Hite and Warga (1997), Ederington and Goh (1998), Dichev and Piotroski (2001), Alp (2013), and Cornaggia and Cornaggia (2013). 2 Liu and Thakor (1984, 345) are the first to recognize this problem: “Unfortunately, few ‘clean’ empirical tests have been performed. The problem is obvious. A state bond’s yield is likely to depend on its rating as well as the state’s economic characteristics. But the rating itself is related to these characteristics” More recently, regarding mortgage-backed securities’ ratings, Ashcraft et al. (2011, 116) explain, “The main empirical challenge to be overcome is identifying the causal effect of ratings on security prices, holding security fundamentals fixed.” 3 House Hearing “Credit Rating Agencies and the Financial Crisis,” transcript available from the U.S. Government Printing Office at https://www.gpo.gov/fdsys/pkg/CHRG–110hhrg51103/html/CHRG–110hhrg51103.htm. 4 In 2010, households held $\$$1.87 trillion of the $\$$3.77 trillion municipal debt market. Mutual funds are a distant second, holding 14%, followed by money market funds (10%), property-casualty insurance companies (9%), and U.S.-chartered depository institutions (7%) (www.federalreserve.gov/releases/z1/current/z1.pdf). 5 Adelino, Cunha, and Ferriera (Forthcoming) document an increase in government employment along with other real effects of this shock to the supply of municipal credit. 6 See “Moody’s to Recalibrate Its U.S. Municipal Bond Ratings to the Company’s Global Rating Scale,” September 2, 2008. 7 See paragraphs 53 and 58 of the Basel Capital Standards, available at http://www.bis.org/publ/bcbs128.pdf. 8 The SEC certifies certain credit rating agencies whose ratings are useful for regulatory compliance as Nationally Recognized Statistical Rating Organizations (NRSROs). 9 See Partnoy (1999), Skreta and Veldkamp (2009), Bolton, Freixas, and Shapiro (2012), Griffin and Tang (2012), He, Qian, and Strahan (2012), Bar-Isaac and Shapiro (2013), Bongaerts (2013), Griffin, Nickerson, and Tang (2013), Fulghieri, Strobl, and Xia (2014), Xia (2014), Bruno, Cornaggia, and Cornaggia (2016), Cornaggia, Cornaggia, and Xia (2016), and Cornaggia, Cornaggia, and Hund (2017). 10 Less than 1% of “Change in Scale” actions were associated with bonds that had speculative grade ratings, and none of these actions resulted in an upgrade. We discard these bonds for ease of presentation of the transition matrices and other results reported by rating. Including these observations does not alter any of our results. 11 S&P downgraded MBIA, Inc., and Ambac Financial Group, Inc., two notches to AA on June 6, 2008. Moody’s downgraded MBIA (Ambac) five (three) notches to A2 (Aa3) on June 19, 2008; Reuter’s (http://www.reuters.com/article/2008/06/05/bonds-insurers-sandp-idUSN0519442220080605) and Dow Jones (http://www.marketwatch.com/story/moodys-downgrades-aaa-rating-of-ambac-mbia). 12 The within-issuer standard deviation of bond ratings averages 0.202 notches (0.206 notches) prior to (following) recalibration. The similarity of these standard deviations indicates that Moody’s did not, for example, upgrade the lowest- or highest-rated bonds for each municipality. Rather, it indicates Moody’s generally shifted upward the entire distribution of ratings for each issuer. 13 We numerically transform Moody’s rating scale ascending in credit quality (Aaa $$=$$ 21, Aa1 $$=$$ 20, ... , C $$=$$ 1). 14 See http://www.msrb.org/BDRegistrants.aspx for a current list of MSRB registered broker-dealers. 15 See http://money.cnn.com/2010/12/22/news/economy/build_america_bonds/. 16 All results are robust to matching Treasuries based on maturity rather than duration. We match by duration because munis often have partial amortization or call provisions that reduce the effective maturity. We compute benchmark Treasury yields as follows. First, we gather STRIPS data from Bloomberg. We keep coupon STRIPS and discard principal STRIPS, because unlike coupon STRIPS, which derive their prices from a wide range of securities, principal STRIPS are identified with a single security (and thus may be influenced by idiosyncratic factors such as premiums for cheapest-to-deliver bonds). For each sample bond, we select the two STRIPS with the closest durations (one above and one below) and linearly interpolate to determine the benchmark. Because these zero-coupon yields are determined by no-arbitrage conditions, they can be directly observed and are not model-specific. 17 Illiquidity in fixed income markets typically complicates abnormal bond return calculations. However, during April and May of 2010, more than $\$$300 billion of munis changed hands in more than one million transactions. 18 Using corporate bond trading data, these authors show that calculating abnormal returns using trade-weighted prices increases the power of the test and reduces Type 1 errors relative to using end-of-day prices. 19 These results are further robust to controlling for the length of time prior to and following the recalibration dates that the bonds trade. For example, the return for a bond that trades on days immediately before Moody’s published the Primary Algorithm and immediately after its rating changes reflects fewer general, systematic changes in the muni market than a bond that trades on days 30 days before Moody’s published the Primary Algorithm and 30 days after its rating changes. We address this concern by replicating the analysis after adjusting bond returns by a municipal bond index. Specifically, we subtract from the return of each bond the return on the S&P Municipal Bond Index over the same time horizon. For example, we subtract the index return from day $$-1$$ to $$+$$1 ($$-30$$ to $$+$$30) for bonds that trade on days –1 and $$+$$1 ($$-30$$ and $$+$$30). We then test for differences in these adjusted returns among upgraded versus non-upgraded bonds. The results do not change. 20 Moody’s classifies municipal issuers into four groups: states, counties, cities, and other. The “state” category includes state agencies such as the California Housing Authority. The “other” category includes school districts and other special tax districts. Some examples from this category are the Birmingham Waterworks Sewer Board in Alabama, the City of Columbus School District in Ohio, and the Las Vegas Valley Water District in Nevada. 21 We sort bonds into sectors based on the field “Sector” in Moody’s data. 22 For example, this filter would screen out a retail-sized trade for an issue held 99% by retail investors and 1% by institutional investors and for which there were no institutional trades. Because we have no reason to believe that institutional holdings create any regulatory-based demand among retail investors, this robustness test imposes a very conservative hurdle for sample inclusion. 23 Untabulated results at the bond level produce qualitatively similar results to the issuer-level analysis that follows. By distilling the bond-level observations into one observation per issuer, we avoid the concern that oversampling could prop up the statistical significance of the results. 24 The Internet Appendix also contains a detailed sample reconciliation in Table A.20. 25 Reed Construction Data (http://www.reedconstructiondata.com/rsmeans/models/elementary-school/). 26 SEC Pub. No. 134 is available at https://www.sec.gov/investor/alerts/municipalbondsbulletin.pdf. 27 The U.S. PIRG ranks the 50 states according to the extent to which they provide online access to government spending data. The opacity index is from the year of Moody’s recalibration (2010). Although we use new issue data from one year before to one year after recalibration, we assume this measure of opacity for any particular state is similar through the sample period. Opacity scores are available at http://www.uspirg.org/reports/usp/following-money-2011. 28 The corruption risk index is a snapshot from 2013. We assume the corruption risk in 2013 is correlated with such risk over our sample period. Corruption risk measures are available at http://www.stateintegrity.org/your_state. 29 All else equal, a two-notch upgrade is more likely to cross into a new broad rating category than a one-notch upgrade. We discard upgrades greater than two notches because they necessarily cross into new broad categories. 30 See SEC Commissioner Aguilar’s February 2015 comments at http://www.sec.gov/news/statement/making-municipal-securities-market-more-transparent-liquid-fair.html. 31 See “House’s Frank says municipal ratings add unfair costs,” Bloomberg, March 12, 2008: http://www.bloomberg.com/apps/news?pid=newsarchive&sid=a2Mx_BNUsDZA&refer=home See also press release from the Connecticut Attorney General’s Office, July 30, 2008, available at http://www.ct.gov/ag/cwp/view.asp?a=2795&q=420390. References Adelino, M., Cunha, I. and Ferriera. M. A. Forthcoming . The economic effects of public financing: Evidence from municipal bond ratings recalibration. Review of Financial Studies . https://doi.org/10.1093/rfs/hhx049 . Almeida, H., Cunha, I. Ferreira, M. A. and Restrepo. F. 2017 . The real effects of credit ratings: Using sovereign downgrades as a natural experiment. Journal of Finance 72 : 249 – 90 . Google Scholar CrossRef Search ADS Alp, A. 2013 . Structural shifts in credit ratings. Journal of Finance 68 : 2435 – 70 . Google Scholar CrossRef Search ADS Ang, A., Green, R. C. Longstaff, F. A. and Xing. Y. Forthcoming. Advance refunding of municipal bonds. Journal of Finance . Ashcraft, A., Goldsmith-Pinkham, P. Hull, P. and Vickery. J. 2011 . Credit ratings and security prices in the subprime MBS market. American Economic Review: Papers & Proceedings 101 : 115 – 19 . Google Scholar CrossRef Search ADS Bar-Isaac, H., and Shapiro. J. 2013 . Ratings quality over the business cycle. Journal of Financial Economics 108 : 62 – 78 . Google Scholar CrossRef Search ADS Becker, B., and Ivashina, V. 2015 . Reaching for yield in the bond market. Journal of Finance 70 : 1863 – 1902 . Google Scholar CrossRef Search ADS Begley, T. A., 2015 . The real costs of corporate credit ratings. Working paper , Washington University. Google Scholar CrossRef Search ADS Bessembinder, H., Kahle, K. M. Maxwell, W. F. and Xu. D. 2009 . Measuring abnormal bond performance. Review of Financial Studies 22 : 4219 – 58 . Google Scholar CrossRef Search ADS Bolton, P., Freixas, X. and Shapiro. J. 2012 . The credit ratings game. Journal of Finance 67 : 85 – 112 . Google Scholar CrossRef Search ADS Bongaerts, D. 2013 . Can alternative business models discipline credit rating agencies? Working Paper, Erasmus University . Google Scholar CrossRef Search ADS Bruno, V., Cornaggia, J. and Cornaggia. K. J. 2016 . Does regulatory certification affect the information content of credit ratings? Management Science 62 : 1578 – 97 . Google Scholar CrossRef Search ADS Butler, A. W. 2008 . Distance still matters: Evidence from municipal bond underwriting. Review of Financial Studies 21 : 763 – 84 . Google Scholar CrossRef Search ADS Butler, A. W., Fauver, L. and Mortal. S. 2009 . Corruption, political connections, and municipal finance. Review of Financial Studies 22 : 2673 – 2705 . Google Scholar CrossRef Search ADS Chen, Z., Lookman, A. A. Schürhoff, N. and Seppi. D. J. 2014 . Rating-based investment practices and bond market segmentation. Review of Asset Pricing Studies 4 : 162 – 205 . Google Scholar CrossRef Search ADS Cornaggia, J., and Cornaggia. K. J. 2013 . Estimating the costs of issuer-paid credit ratings. Review of Financial Studies 26 : 2229 – 69 . Google Scholar CrossRef Search ADS Cornaggia, J., Cornaggia, K. J. and Hund. J. 2017 . Credit ratings across asset classes. Review of Finance 21 : 465 – 509 . Google Scholar CrossRef Search ADS Cornaggia, J., Cornaggia, K. J. and Xia. H. 2016 . Revolving doors on Wall Street. Journal of Financial Economics 120 : 400 – 419 . Google Scholar CrossRef Search ADS Cuny, C. 2016 . Voluntary disclosure incentives: Evidence from the municipal bond market. Journal of Accounting and Economics 62 : 87 – 102 . Google Scholar CrossRef Search ADS Dichev, I. D., and Piotroski. J. D. 2001 . The long-run stock returns following bond ratings changes. Journal of Finance 41 : 173 – 203 . Google Scholar CrossRef Search ADS Dick-Nielsen, J., Feldhutter, P. and Lando. D. 2012 . Corporate bond liquidity before and after the onset of the subprime crisis. Journal of Financial Economics 103 : 471 – 92 . Google Scholar CrossRef Search ADS Ederington, L. H., and Goh. J. C. 1998 . Bond rating agencies and stock analysts: Who knows what when? Journal of Financial and Quantitative Analysis 33 : 569 – 85 . Google Scholar CrossRef Search ADS Ekins, E. M., and Calabria. M. A. 2012 . Regulation, market structure, and role of the credit rating agencies. Policy Analysis 704 : 1 – 36 . Ellul, A., Jotikasthira, C. and Lundblad. C. 2011 . Regulatory pressure and fire sales in the corporate bond markets. Journal of Financial Economics 101 : 96 – 620 . Google Scholar CrossRef Search ADS Financial Stability Board. 2010 . Principles for reducing reliance on CRA ratings. http://www.financialstabilityboard.org/publications/r_121105b.pdf. Financial Stability Board. 2012 . Roadmap and workshop for reducing reliance on CRA ratings. http://www.financialstabilityboard.org/wp-content/uploads/r_121105b.pdf. Fulghieri, P., Strobl, G. and Xia. H. 2014 . The economics of solicited and unsolicited credit ratings. Review of Financial Studies 27 : 484 – 515 . Google Scholar CrossRef Search ADS Goel, A. M., and Thakor. A. V. 2015 . Information reliability and welfare: A theory of coarse credit ratings. Journal of Financial Economics 115 : 541 – 57 . Google Scholar CrossRef Search ADS Goh, J. C., and Ederington. L. H. 1993 . Is a bond rating downgrade good news, bad news or no news to stockholders? Journal of Finance 48 : 2001 – 8 . Google Scholar CrossRef Search ADS Gore, A. K. 2004 . The effects of GAAP regulation and bond market interaction on local government disclosure. Journal of Accounting and Public Policy 23 : 23 – 52 . Google Scholar CrossRef Search ADS Griffin, J. M., Nickerson, J. and Tang. D. Y. 2013 . Rating shopping or catering? An examination of the response to competitive pressure for CDO ratings. Review of Financial Studies 26 : 2270 – 2310 . Google Scholar CrossRef Search ADS Griffin, J. M., and Tang. D. 2012 . Did subjectivity play a role in CDO credit ratings? Journal of Finance 67 : 1293 – 1328 . Google Scholar CrossRef Search ADS Hand, J. R. M., Holthausen, R. W. and Leftwich. R. W. 1992 . The effect of bond rating agency announcements on bond and stock prices. Journal of Finance 47 : 733 – 52 . Google Scholar CrossRef Search ADS Harris, L., and Piwowar. M. 2006 . Secondary trading costs in the municipal bond market. Journal of Financial Economics 61 : 1361 – 97 . He, J., Qian, J. and Strahan. P. E. 2012 . Are all ratings created equal? The impact of issuer size on the pricing of mortgage-backed securities. Journal of Finance 67 : 2097 – 2137 . Google Scholar CrossRef Search ADS Hettenhouse, G. W., and Sartoris. W. L. 1976 . An analysis of the information value of bond rating changes. Quarterly Review of Economics and Business 16 : 65 – 78 . Hite, G., and Warga. A. 1997 . The effect of bond-rating changes on bond price performance. Financial Analysts Journal 53 : 35 – 51 . Google Scholar CrossRef Search ADS Holthausen, R. W., and Leftwich. R. W. 1986 . The effect of bond rating changes on common stock prices. Journal of Financial Economics 17 : 57 – 89 . Google Scholar CrossRef Search ADS Ingram, R. W., Brooks, L. E. and Copeland. R. M. 1983 . The information content of municipal bond rating changes: A note. Journal of Finance 38 : 997 – 1003 . Google Scholar CrossRef Search ADS Kisgen, D. J. 2012 . The real and financial effects of credit ratings: Evidence from Moody’s adjustments. Working Paper , Boston College . Google Scholar CrossRef Search ADS Kliger, D., and Sarig. O. 2000 . The information value of bond ratings. Journal of Finance 55 : 2879 – 2903 . Google Scholar CrossRef Search ADS Liu, P., and Thakor. A. V. 1984 . Interest yields, credit ratings, and economic characteristics of state bonds: An empirical analysis: Note. Journal of Money, Credit, and Banking 16 : 344 – 51 . Google Scholar CrossRef Search ADS Moody’s Investors Service. 2006 . Analyzing the tradeoff between ratings accuracy and stability. Moody’s Investors Service. 2007 . The U.S. municipal bond rating scale: Mapping to the global rating scale and assigning global scale ratings to municipal obligations. Moody’s Investors Service. 2010 . Recalibration of Moody’s U.S. municipal ratings to its global rating scale. Opp, C., Opp, M. and Harris. M. 2013 . Rating agencies in the face of regulation. Journal of Financial Economics 108 : 46 – 61 . Google Scholar CrossRef Search ADS Partnoy, F. 1999 . The Siskel and Ebert of financial markets: two thumbs down for the credit rating agencies. Washington University Law Quarterly 77 : 619 – 712 . Pinches, G. E., and Singleton. J. C. 1978 . The adjustment of stock prices to bond rating changes. Journal of Finance 33 : 29 – 44 . Google Scholar CrossRef Search ADS Skreta, V., and Veldkamp. L. 2009 . Ratings shopping and asset complexity: A theory of ratings inflation. Journal of Monetary Economics 56 : 678 – 95 . Google Scholar CrossRef Search ADS Stanton, R., and Wallace. N. 2013 . CMBS subordination, ratings inflation, and regulatory-capital arbitrage. Working Paper , University of California at Berkeley . Google Scholar CrossRef Search ADS Thakor, A. V., and Liu. P. 1984 . Interest yields, credit ratings, and economic characteristics of state bonds: An empirical analysis: Note. Journal of Money, Credit, and Banking 16 : 344 – 51 . Google Scholar CrossRef Search ADS U.S. Securities and Exchange Commission. 2011 . Summary report of commission staff’s examinations of each NRSRO as required by Section 15E(p)(3)(c) of the Securities Exchange Act of 1934. Weinstein, M. I. 1977 . The effect of a rating change announcement on bond price. Journal of Financial Economics 5 : 329 – 50 . Google Scholar CrossRef Search ADS Xia, H. 2014 . Can investor-paid credit rating agencies improve the information quality of issuer-paid rating agencies? Journal of Financial Economics 111 : 450 – 68 . Google Scholar CrossRef Search ADS © The Author(s) 2017. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Review of Financial Studies Oxford University Press

Credit Ratings and the Cost of Municipal Financing

Loading next page...
 
/lp/ou_press/credit-ratings-and-the-cost-of-municipal-financing-AdY1weQ850
Publisher
Oxford University Press
Copyright
© The Author(s) 2017. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.
ISSN
0893-9454
eISSN
1465-7368
D.O.I.
10.1093/rfs/hhx094
Publisher site
See Article on Publisher Site

Abstract

Abstract A common belief held among researchers and policy makers is that regulatory reliance has inflated market demand for credit ratings, despite their decreasing informational value. Advances in information technology, coupled with reputation losses following the subprime crisis, renew the question of whether investors still rely on ratings to assess credit risk. Using Moody’s 2010 scale recalibration, which was unrelated to changing issuer fundamentals, we find that ratings still matter to investors and to issuers—apart from any regulatory implications. Our results commend improved disclosure to mitigate mechanistic reliance on ratings and inefficiencies due to rating standards that vary across asset classes. Received October 9, 2015; editorial decision June 7, 2017 by Editor Andrew Karolyi. The purpose of this paper is to test whether credit rating agencies (CRAs) remain relevant as information intermediaries in the modern information environment. There is a massive literature documenting correlation between credit ratings and securities prices; however, these papers commonly suffer an endogeneity problem.1 Specifically, it is difficult to determine whether investors respond directly to credit ratings, or if investors and CRAs merely observe and react to the same information about issuer fundamentals.2 Kliger and Sarig (2000) address this problem by showing that markets reacted when Moody’s Investors Service (Moody’s) added modifiers to its ratings scale in 1982. Much has changed since 1982. The speed and cost of information processing have exponentially increased and decreased, respectively, because of advancements in information technology and the advent of the internet. These developments have likely resulted in at least some investors who previously relied on CRAs to begin performing their own credit risk analyses. Even the less ambitious investors now have alternative sources of information—including market prices of credit default swaps (CDS)—that were not available in 1982. Further, both Moody’s and Standard and Poor’s (S&P) suffered significant loss of reputation capital as a result of the inaccurate ratings they produced in the run-up to the recent financial crisis. Although the ratings most relevant during the financial crisis were those of structured finance products, some question CRA viability more broadly. For example, during a post-crisis congressional hearing, Congressman Christopher Shays argued, “They have no brand, they have no credibility whatsoever. I can’t imagine any investor trusting them.”3 For these reasons, we revisit the question of whether credit rating agencies still matter. There are several more recent papers (reviewed in Section 1) indicating that credit ratings continue to have real economic effects. However, the evidence in these papers suggests that ratings now matter primarily (if not exclusively) due to their regulatory implications. Because ratings affect investment standards and capital requirements, they affect the value of securities to institutional investors, even if these investors do not rely on ratings to evaluate credit risk. The regulatory implications of ratings are particularly acute in markets dominated by regulated investors, including the more commonly studied corporate bond market. The unanswered research question we tackle here is whether investors continue to rely on credit ratings for information about credit risk, apart from the confounding effects of ratings’ regulatory implications. We find robust evidence that they do. We examine the impact of Moody’s recalibration of its municipal bond (muni) rating scale in the spring of 2010, after the dust of the recent financial crisis had settled. Historically, the criteria Moody’s used to assign municipal bond ratings was based on how likely the municipality was to require financial support from higher levels of government. These criteria were unique to municipal bonds; bonds in all other asset classes are rated according to their expected losses. When Moody’s recalibrated its ratings for municipal debt, it applied the expected loss criteria it uses for all other asset classes. This change in criteria resulted in upgrades of zero to four notches on $2.2 trillion of municipal debt. Importantly, Moody’s designed the recalibration to be uncorrelated with changes in issuer fundamentals. By changing the criteria, Moody’s provided a new perspective on the bonds’ credit risk, but the fundamentals of the bonds did not change. This event allows us to overcome the endogeneity challenge faced by most prior studies. Unlike the corporate bond market studied previously, the muni market is dominated by unregulated retail investors.4 By focusing on retail trades in the muni market, we avoid confounding regulatory effects. Because our setting involves rating changes that do not result from changes in fundamentals, and because the transactions in our sample are not executed by regulated investors, we are able to cleanly identify investor reliance on ratings to assess credit risk. An important feature of the recalibration is that not all munis were upgraded. Municipal issuers that were already “well calibrated” to the global scale for other asset classes serve as our control group in a difference-in-differences framework. These bonds provide reasonable benchmarks for how the prices on upgraded bonds would have behaved in the absence of Moody’s recalibration. Because credit ratings on insured bonds reflect the credit quality of the insurer, we include only uninsured bonds in our analyses. (Roughly 60% of the $\$$ 2.2 trillion sample munis are uninsured.) Our sample consists of roughly equally-sized treatment and control groups: $\$$640 billion of uninsured munis experienced upgrades due to recalibration, and $\$$601 billion did not. We find robust evidence that investors reacted to this event. Controlling for bond characteristics and a host of fixed effects, we find that upgraded bonds experience a decrease in credit spreads of 19 to 33 basis points (bp) relative to non-upgraded bonds. This is the main result in the paper. This reliance is economically meaningful, and our results are robust to a wide variety of alternative specifications. Although the municipal bond market is a natural setting to test whether investors rely on ratings to assess risk, we take seriously the possibility that our results could still reflect some regulation-based demand. Although our secondary market results focus exclusively on retail-size trades, we impose as a robustness check an additional filter that removes retail transactions for issues with any level of holdings by insurance companies facing ratings-based capital charges. Results and conclusions are unchanged. We make use of various institutional features to further test for regulation-based demand. For example, most ratings-based regulations distinguish between broad rating categories rather than individual notches within those categories. We exploit this feature by comparing results among bonds with equal-sized upgrades that do and do not cross into new broad rating categories. These tests provide at most limited evidence of regulation-based demand. Some regulations also use a “lowest rating binds” criterion, which employs the lower of Moody’s and Standard & Poor’s (S&P) ratings. We exploit this feature by comparing the impact of the recalibration among upgraded bonds with ratings from S&P that remain, versus become, lower than ratings from Moody’s. If the rating from S&P was lower than the rating from Moody’s before the recalibration and remained lower after Moody’s upgrades, then the recalibration should have no regulatory bite. However, if the Moody’s rating leapfrogged the S&P rating (and thus the S&P rating becomes the lower rating), then this upgrade should have regulatory implications. However, we find no differential impact of the recalibration on these two groups of upgraded bonds. Overall, the cross-sectional analysis of upgrades based on their likely regulatory effects provides corroborating evidence that investors’ reaction to Moody’s recalibration reflects primarily a reliance on ratings to price risk rather than an increase in regulation-based demand. We explore several other alternative explanations for our results. For example, Harris and Piwowar (2006) show that municipal bond liquidity increases with credit quality. If investor perception of bonds’ credit quality increases when ratings are upgraded, then the price changes we observe could reflect lower liquidity premiums. A consequence of this hypothesis is that upgraded bonds should experience permanent increases in liquidity. This is not what we find. Although upgraded bonds’ trading volume increases immediately after the recalibration relative to non-upgraded bonds, this increase is transitory. We find that upgraded bonds’ trading volume in the period three to six months after the recalibration is statistically indistinguishable from that prior to the recalibration. This finding indicates that upgraded bonds do not experience permanent increases in liquidity, despite their permanently higher ratings. Another possible explanation for our results is that the control group, the non-upgraded bonds, coincidentally experienced a decrease in returns around the time of the recalibration. If this is true, then our results could still obtain even if the market did not bid up the prices of the upgraded bonds. We find no evidence of this effect. If anything, the returns of the non-upgraded group are slightly higher (although not significantly so) around the recalibration in comparison to their own returns in a 180-day period preceding the recalibration. Next, we examine the possibility that our results could reflect shifts in demand for particular levels of governments’ bonds around the recalibration. For example, state-level issuers received some of the largest upgrades. If investors happened to experience an increase in demand for state-level (as opposed to county, city, or other levels of government) bonds around the time of the recalibration, then our results could reflect that demand shift instead of a response to the rating changes. We address this possibility by including in our regressions issuer-level-of-government fixed effects that vary before and after the recalibration. Our results are fully robust. Finally, we consider the possibility that our results could reflect differential changes in the fundamentals of the upgraded and non-upgraded groups. Moody’s (2010) states that the recalibration does not reflect changes in credit risk and that any ratings under review prior to the recalibration remained under review. Still, we address this potential concern by comparing S&P’s ratings to Moody’s ratings around the time of the recalibration. If the recalibration did indeed reflect changes in fundamentals, then we should observe S&P eventually change its ratings in a pattern similar to Moody’s. We start by constructing ratings transition matrices for bonds rated by both Moody’s and S&P around the recalibration. We find little similarity in the shape of S&P’s and Moody’s transition matrices. We also examine a time series of the two raters’ average muni ratings. We observe a sharp increase in Moody’s ratings relative to S&P in 2010 that remains fully intact through the end of our data availability. Overall, although we cannot disprove the possibility that the upgraded and non-upgraded bonds’ fundamentals were changing differently around the time of the recalibration, we find no evidence to support this possibility. We turn next to the primary market to test whether market reliance on ratings has real economic effects. We conduct this analysis at the issuer level, sorting issuers by whether their outstanding bonds were upgraded as a result of Moody’s recalibration. Using a multivariate difference-in-differences approach, we find that spreads on new issues by upgraded issuers decrease by 15 to 22 bp, relative to the control group. This magnitude is comparable to what we find in the secondary market, and it indicates that our findings are economically meaningful. The product of $\$$ 640 billion (the face value of uninsured municipal debt upgraded during recalibration) and 15 bp (our most conservative estimate of the recalibration effect on offer yields) is $\$$960 million. This back-of-the-envelope calculation provides an estimate of aggregate excess interest paid annually (in 2010 dollars) by U.S. taxpayers due to Moody’s previous dual-class rating system. We also observe that upgraded issuers see a larger increase in issuance volume than non-upgraded issuers in the years following the recalibration. This finding demonstrates that ratings have real economic effects; lower borrowing costs increase municipal borrowing (and presumably increase municipal investment)5. We extend our primary market analysis to test whether investors rely more heavily on credit ratings when the amount and quality of alternative sources of information are low. For example, we employ the issuer level of government as a proxy for issuer size and opacity. Consistent with an information effect, we find the results of the recalibration are weakest among state-level issuers and strongest among cities. The recalibration effect is also stronger among issuers in states with more opaque accounting practices, in states identified as more corrupt, and among municipalities without ratings from S&P. Combined, these additional results indicate that ratings are more influential when investors have less alternative information. Overall, our contribution to the literature is original evidence that investors rely on credit ratings to assess risk, and that this reliance is greatest among opaque issuers for which investors lack alternative sources of information. Our results bring new evidence to bear regarding the classic question of whether and how security prices depend on credit ratings. The results also shed light on how investors process information in the municipal bond market, a multi-trillion-dollar market that is relatively opaque and is beginning to receive greater attention from researchers. 1. Institutional Background and Literature Review 1.1 Moody’s dual class ratings Moody’s uses its Global Scale when rating corporate bonds, sovereign debt, and structured finance products. These bonds are rated according to their expected losses. Expected loss is the product of probability of default and loss given default. Historically, Moody’s rated municipal bonds according to separate criteria. Moody’s assigned municipal ratings based on how likely a municipality is to require extraordinary support from a higher level of government in order to avoid default; Moody’s (2007, 2). This changed in the spring of 2010, when Moody’s recalibrated its municipal bond rating criteria to match that of the Global Scale. Moody’s (2010, 1) clarifies that the recalibration is intended to enhance the comparability of ratings across asset classes, not to indicate a change in credit quality: Our benchmarking $$\ldots$$ will result in an upward shift for most state and local government long-term municipal ratings by up to three notches. The degree of movement will be less for some sectors $$\ldots$$ which are largely already aligned with ratings on the global scale. Market participants should not view the recalibration of municipal ratings as ratings upgrades, but rather as a recalibration of the ratings to a different scale. $$\ldots$$ [The recalibration] does not reflect an improvement in credit quality or a change in our opinion. Importantly for our study, Moody’s (2010) indicates that any ratings under review for upgrade or downgrade prior to recalibration would remain under review—not lumped into these massive ratings changes. As such, our sample does not include any natural upgrades associated with improving issuer fundamentals that would contaminate the estimates generated by our tests. The timeline of Moody’s recalibration is as follows. In 2008, Moody’s revealed its intention to recalibrate its municipal bond rating scale.6 This announcement, however, contained no information regarding which bonds’ ratings would change or by how much. On March 16, 2010, Moody’s announced the particulars of the recalibration. On this date, Moody’s published a white paper containing its Primary Algorithm. The Primary Algorithm indicated which bonds’ ratings would be upgraded and by how much. The two characteristics that Moody’s used to recalibrate muni ratings were preexisting ratings and sectors. Moody’s categorizes bonds into four sectors. Figure A.1 in the Internet Appendix reproduces the Primary Algorithm. The actual recalibration was enacted on four dates. The first recalibration date was April 16, 2010, one month after the publication of the Primary Algorithm. The second, third, and fourth recalibration dates were April 23, May 1, and May 7, 2010, respectively. We provide details in Section 2 on the number and par values of bonds that were upgraded (and not upgraded) on each date. We test whether the market reacted to this event. If the market does not rely on Moody’s ratings to assess risk, then we should see no reaction. However, the change in criteria applied by Moody’s potentially allows the market to learn new information about the bonds. Consider the following analogy. Imagine a student who earns a B on a mostly qualitative exam. If the student soon after earns an A on a mostly quantitative exam in the same class, then the professor will update her opinion on the student’s aptitude. Yet the student’s aptitude does not change from one exam to the next. What changes are the criteria used to evaluate aptitude. In the same way, the market might react to Moody’s recalibration even though the bonds’ fundamentals do not change around the time of the recalibration. What changes are the criteria used to evaluate credit risk. Credit risk is not one-dimensional. If we find a response to the recalibration, then we can infer that investors updated their views about the bonds’ credit risk after Moody’s evaluated the bonds under the new criteria. Just as the professor takes a more favorable view of the student’s aptitude when using criteria that reward quantitative ability, the market might change its view of munis when Moody’s uses criteria based on expected losses. 1.2 Credit ratings and financial regulation Financial regulators have historically relied on credit ratings to establish capital requirements and prudent investment guidelines. This regulatory reliance on ratings dates to at least a ruling by the U.S. Comptroller of the Currency in 1931. Under Rule 5b-3 of the Investment Company Act, the United States Securities and Exchange Commission (SEC) treated Aaa-rated bonds as equivalent to Treasuries. Pension fund investment guidelines established by the Employee Retirement Income Security Act (ERISA) and bank capital requirements established by the Basel Committee on Banking Supervision have likewise been ratings-based. Under the Standardized Approach in Basel II, single A-rated munis carry a higher charge (20% risk weight) than Aa- or Aaa-rated munis (0% risk weight).7 Capital charges established by the National Association of Insurance Commissioners (NAIC) range from 3.39% to 19.5% for speculative grade (SG) bonds compared with 0.30% to 0.96% for investment grade (IG) bonds. This body of regulation creates incentives for regulated investors to respond to ratings, irrespective of whether they rely on ratings to evaluate risk. Indeed, the state of the credit ratings literature suggests that ratings matter primarily due to their regulatory implications. For example, Ellul, Jotikasthira, and Lundblad (2011) document fire sales by insurance companies when bonds in their portfolios are downgraded from IG to SG. The liquidity premiums associated with these sales indicate that the sales were attributable to capital charges rather than any information communicated by the downgrade. Becker and Ivashina (2015) further document regulatory arbitrage by insurance companies chasing yield in a ratings-based regulatory environment. Because the NAIC sets capital charges based on ratings, savvy insurance companies circumvent their regulatory capital charges by over-allocating capital to the bonds with the highest credit risk within a particular credit rating category. Stanton and Wallace (2013) similarly conclude that overinvestment in high-risk commercial mortgage-backed securities (CMBS) is attributable to such regulatory arbitrage in a ratings-based environment, and Cornaggia, Cornaggia, and Hund (2017) simulate potential regulatory arbitrage among banks subject to Basel II capital requirements. Opp, Opp, and Harris (2013) provide a formal model and conclude that regulatory implications of ratings are of first-order concern for marginal investors. In addition to official regulation, Chen et al. (2014) document the reliance on credit ratings in private investment mandates, asset management policies, and informal procedures that employ ratings to restrict holdings by mutual funds and investment advisors. Perhaps Ekins and Calabria (2012, 1) summarize the evolution of CRA relevance most succinctly in their conclusion: “Government regulatory use of credit ratings inflated the market demand for NRSRO ratings, despite the decreasing informational value of credit ratings.”8 Because of the regulation-based demand for ratings, recent studies that document market reaction to ratings changes or other real economic effects (e.g., Kisgen 2012; Almeida et al. 2017; and Begley 2015) cannot conclude that investors rely on ratings for information. Results from these studies may instead reflect changes in regulatory compliance costs. In fact, Almeida et al. (2017) specifically focus on ratings-based Basel II capital requirements in their study of the real effects of sovereign debt downgrades. One benefit of our setting is that the retail investors who dominate the muni market are subject to none of the aforementioned regulations. Any reaction among retail investors must therefore reflect an information effect. We are the first, to our knowledge, to directly test the extent to which market participants rely on ratings to assess credit risk apart from their need to manage regulatory capital charges and comply with other ratings-based regulations. 1.3 Modern relevance of CRAs as information intermediaries It is not obvious that modern investors rely on credit ratings to assess risk. For one thing, ratings are too coarse to fully reflect differences in credit quality across all rated securities; see Goel and Thakor (2015). Second, traditional ratings are designed to be stable over time and not to reflect real-time changes in credit quality (Moody’s 2006). Third, the conflicts of interest in the issuer-pays CRA compensation structure are well known.9 Fourth, improvements in information technology provide investors more-granular and more-timely credit risk metrics than traditional credit ratings (e.g., Cornaggia and Cornaggia 2013). Finally, the CRAs lost significant reputation capital due to their role in the subprime crisis. Finally, although Kliger and Sarig (2000) demonstrate a causal impact of ratings on security prices, these authors use an event from 1982. In 1982, investors lacked internet access, paid for long-distance telephone service, and received hard copies of Moody’s investment manuals in the mail. As such, it is unclear whether traditional CRAs matter as much as they did in 1982. We are the first, to our knowledge, to disentangle both the regulatory and endogeneity problems faced by existing literature in the modern information era. 1.4 Muni market information environment Another novel feature of our paper is the ability to exploit cross-sectional variation in the quality of information available to investors in municipal bonds. Other papers focus primarily on corporate securities that trade in relatively liquid, transparent markets and therefore may not be generalized to markets with less transparent issuers. Although the municipal finance market is large ( $\$$ 3.77 trillion in March 2014), this market is opaque and decentralized. Unlike corporations, state and local governments are not subject to the registration and reporting requirements of the SEC. Therefore, financial disclosure by municipalities is often less reliable, less comparable, and less timely than information released by corporations; however, the information quality varies widely across municipal issuers (see Ingram, Brooks, and Copeland 1983; Gore 2004; and Cuny 2016). This cross-sectional variation in the information environment across muni markets allows for additional tests of investor reliance on ratings to assess credit risk. 2. Data Collection and Sample Description 2.1 The recalibration event Our data consist of ratings from Moody’s and S&P, bond market transaction prices and volume from the Municipal Securities Rulemaking Board (MSRB), and issue/issuer characteristics from Ipreo. From Moody’s, we collect ratings data on every bond issue by a state or local government that had a “Change in Scale” rating action on April 16, April 23, May 1, or May 7 in 2010, as well as the ratings on all past and future issues by the same issuers. Because market perception of insured bonds reflects the credit quality of the monolines, we focus on uninsured bonds in our empirical analyses. Table 1 presents the number of issues and cumulative par value of recalibrated investment-grade munis.10 Panel A contains all bonds with a “Change in Scale” rating action. Panel B contains the uninsured bonds from which we draw our sample. Table 1 Number and par values of recalibrated bonds Panel A: All bonds All “Change in Scale” rating actions “Change in Scale” results in an upgrade “Change in Scale” results in no change in rating Recalibration date $$N$$ bonds Total par $$N$$ bonds Total par $$N$$ bonds Total par April 16, 2010 213,260 $\$$932.8 billion 190,144 $\$$812.5 billion 23,116 $\$$120.4 billion April 23, 2010 201,962 $\$$312.9 billion 186,946 $\$$281.2 billion 15,016 $\$$31.7 billion May 1, 2010 124,053 $\$$249.9 billion 108,046 $\$$199.4 billion 16,007 $\$$50.6 billion May 7, 2010 105,855 $\$$715.2 billion 24,221 $\$$67.4 billion 81,634 $\$$647.8 billion Sum 645,130 $\$$2,210.8 billion 509,357 $\$$1,360.5 billion 135,773 $\$$850.5 billion Panel B: Uninsured bonds April 16, 2010 90,621 $\$$566.3 billion 72,213 $\$$466.0 billion 18,408 $\$$100.3 billion April 23, 2010 55,891 $\$$96.8 billion 42,769 $\$$70.5 billion 13,122 $\$$26.4 billion May 1, 2010 54,021 $\$$117.2 billion 40,550 $\$$72.3 billion 13,471 $\$$44.9 billion May 7, 2010 65,510 $\$$461.2 billion 8,944 $\$$31.5 billion 56,566 $\$$429.6 billion Sum 266,043 $\$$1,241.5 billion 164,476 $\$$640.3 billion 101,567 $\$$601.2 billion Panel A: All bonds All “Change in Scale” rating actions “Change in Scale” results in an upgrade “Change in Scale” results in no change in rating Recalibration date $$N$$ bonds Total par $$N$$ bonds Total par $$N$$ bonds Total par April 16, 2010 213,260 $\$$932.8 billion 190,144 $\$$812.5 billion 23,116 $\$$120.4 billion April 23, 2010 201,962 $\$$312.9 billion 186,946 $\$$281.2 billion 15,016 $\$$31.7 billion May 1, 2010 124,053 $\$$249.9 billion 108,046 $\$$199.4 billion 16,007 $\$$50.6 billion May 7, 2010 105,855 $\$$715.2 billion 24,221 $\$$67.4 billion 81,634 $\$$647.8 billion Sum 645,130 $\$$2,210.8 billion 509,357 $\$$1,360.5 billion 135,773 $\$$850.5 billion Panel B: Uninsured bonds April 16, 2010 90,621 $\$$566.3 billion 72,213 $\$$466.0 billion 18,408 $\$$100.3 billion April 23, 2010 55,891 $\$$96.8 billion 42,769 $\$$70.5 billion 13,122 $\$$26.4 billion May 1, 2010 54,021 $\$$117.2 billion 40,550 $\$$72.3 billion 13,471 $\$$44.9 billion May 7, 2010 65,510 $\$$461.2 billion 8,944 $\$$31.5 billion 56,566 $\$$429.6 billion Sum 266,043 $\$$1,241.5 billion 164,476 $\$$640.3 billion 101,567 $\$$601.2 billion This table displays the number and total par value of municipal bonds for which Moody’s issued a “Change in Scale” rating action between April 16, 2010, and May 7, 2010. Panel A includes all bonds rated by Moody’s. Panel B restricts the sample in Panel A to uninsured bonds. We collect ratings data on bonds issued by state or local governments from Moody’s. Table 1 Number and par values of recalibrated bonds Panel A: All bonds All “Change in Scale” rating actions “Change in Scale” results in an upgrade “Change in Scale” results in no change in rating Recalibration date $$N$$ bonds Total par $$N$$ bonds Total par $$N$$ bonds Total par April 16, 2010 213,260 $\$$932.8 billion 190,144 $\$$812.5 billion 23,116 $\$$120.4 billion April 23, 2010 201,962 $\$$312.9 billion 186,946 $\$$281.2 billion 15,016 $\$$31.7 billion May 1, 2010 124,053 $\$$249.9 billion 108,046 $\$$199.4 billion 16,007 $\$$50.6 billion May 7, 2010 105,855 $\$$715.2 billion 24,221 $\$$67.4 billion 81,634 $\$$647.8 billion Sum 645,130 $\$$2,210.8 billion 509,357 $\$$1,360.5 billion 135,773 $\$$850.5 billion Panel B: Uninsured bonds April 16, 2010 90,621 $\$$566.3 billion 72,213 $\$$466.0 billion 18,408 $\$$100.3 billion April 23, 2010 55,891 $\$$96.8 billion 42,769 $\$$70.5 billion 13,122 $\$$26.4 billion May 1, 2010 54,021 $\$$117.2 billion 40,550 $\$$72.3 billion 13,471 $\$$44.9 billion May 7, 2010 65,510 $\$$461.2 billion 8,944 $\$$31.5 billion 56,566 $\$$429.6 billion Sum 266,043 $\$$1,241.5 billion 164,476 $\$$640.3 billion 101,567 $\$$601.2 billion Panel A: All bonds All “Change in Scale” rating actions “Change in Scale” results in an upgrade “Change in Scale” results in no change in rating Recalibration date $$N$$ bonds Total par $$N$$ bonds Total par $$N$$ bonds Total par April 16, 2010 213,260 $\$$932.8 billion 190,144 $\$$812.5 billion 23,116 $\$$120.4 billion April 23, 2010 201,962 $\$$312.9 billion 186,946 $\$$281.2 billion 15,016 $\$$31.7 billion May 1, 2010 124,053 $\$$249.9 billion 108,046 $\$$199.4 billion 16,007 $\$$50.6 billion May 7, 2010 105,855 $\$$715.2 billion 24,221 $\$$67.4 billion 81,634 $\$$647.8 billion Sum 645,130 $\$$2,210.8 billion 509,357 $\$$1,360.5 billion 135,773 $\$$850.5 billion Panel B: Uninsured bonds April 16, 2010 90,621 $\$$566.3 billion 72,213 $\$$466.0 billion 18,408 $\$$100.3 billion April 23, 2010 55,891 $\$$96.8 billion 42,769 $\$$70.5 billion 13,122 $\$$26.4 billion May 1, 2010 54,021 $\$$117.2 billion 40,550 $\$$72.3 billion 13,471 $\$$44.9 billion May 7, 2010 65,510 $\$$461.2 billion 8,944 $\$$31.5 billion 56,566 $\$$429.6 billion Sum 266,043 $\$$1,241.5 billion 164,476 $\$$640.3 billion 101,567 $\$$601.2 billion This table displays the number and total par value of municipal bonds for which Moody’s issued a “Change in Scale” rating action between April 16, 2010, and May 7, 2010. Panel A includes all bonds rated by Moody’s. Panel B restricts the sample in Panel A to uninsured bonds. We collect ratings data on bonds issued by state or local governments from Moody’s. The recalibration event of 2010 followed the monolines’ loss of their Aaa ratings in June 2008.11 We thus consider the extent to which the composition of the muni market (insured versus uninsured issues) may have changed around the time of the recalibration. We find that the resulting change in the proportion of insured/uninsured munis occurred more than two years prior to the recalibration event. We report this evidence in the Internet Appendix; see Figure A.1. In March 2010, Moody’s Primary Algorithm advertised a zero- to three-notch upgrade associated with the recalibration. Table 2 reports the actual migration matrix. As in Table 1, Panel A contains all bonds and Panel B contains only the uninsured bonds from which we draw our sample. The proportion of bonds upgraded varies by initial rating. Other than Aaa-rated bonds, which by definition cannot upgrade, no initial rating level retain