Municipal Bond Ratings and Citizens’ Rights

1. Municipal Bond Ratings and Citizens’ Rights John Yinger The Maxwell School, Syracuse University, April 2013

2. Municipal Bond Ratings and Citizens' Rights The Role of Bond Ratings • Municipal bond ratings help investors by indicating the likelihood that a particular municipal bond will default. • Bond ratings also have significant consequences for the public interest, because a low rating raises the cost of borrowing money for public capital projects. • Despite this link to the public interest, however, bond ratings are not regulated. • This paper develops a regulatory framework to determine whether bond ratings reflect an appropriate balance between societal objectives and the legitimate business objectives of rating agencies.

3. Municipal Bond Ratings and Citizens' Rights Citizens’ Rights • This paper is concerned with the fair treatment of citizens in different jurisdictions. • Thanks to municipal bond ratings, citizens' must pay more for infrastructure in some jurisdictions than in others. • The question is whether this variation is to some degree “unfair,” in the sense that it is based on factors, such as the racial or ethnic composition of a jurisdiction, that businesses are not supposed to consider. • This paper draws on the nation’s civil rights laws to develop a regulatory framework for answering this question.

4. Municipal Bond Ratings and Citizens' Rights Municipal Bond Ratings • The 3 major agencies that rate municipal bonds—Fitch, Moody’s, and Standard and Poor’s (S&P)—use different methodologies, but they all say that their ratings indicate the likelihood that an investor will lose money because of default. • Ratings are based on factors that the rating agency believes are linked to the creditworthiness of the bond issue or issuer. • Many scholars have developed models for predicting ratings based on observable factors.

5. Municipal Bond Ratings and Citizens' Rights What Do Ratings Look Like?

6. Municipal Bond Ratings and Citizens' Rights Current Regulation of Ratings Agencies • In principle, these agencies are regulated by the Securities and Exchange Commission (SEC), but the only substantive regulation conducted by SEC has been the designation (in 1975!) of these three agencies as National Recognized Statistical Rating Organizations (NRSROs). • There are now 10 NRSROs, but Fitch, Moody’s, and S&P remain the only NRSROs that rate municipal bonds in the United States.

7. Municipal Bond Ratings and Citizens' Rights Credit Rating Agency Reform Act of 2006 • CRARA declares that “neither the Commission nor any State (or political subdivision thereof) may regulate the substance of credit ratings or the procedures and methodologies by which any nationally recognized statistical rating organization determines credit ratings.” • This paper develops societal standards for evaluating municipal bond ratings and designing regulation to ensure that these standards are met. • It is not clear to me whether the financial reform bill passed in 2010 changes this provision, but it does provide for new oversight of municipal bond ratings in the SEC.

8. Municipal Bond Ratings and Citizens' Rights A Framework from Civil Rights Legislation • The Fair Housing Act makes it illegal • for profit, to induce or attempt to induce any person to sell or rent any dwelling by representations regarding the entry or prospective entry into the neighborhood of a person or persons of a particular race, color, religion, sex, handicap, familial status, or national origin. • This type of discrimination, which is called redlining, is also prohibited by the Equal Credit Opportunity Act. • A municipal bond rating that reflects the racial or ethnic composition of the issuing jurisdiction involves behavior that is analogous to redlining.

9. Municipal Bond Ratings and Citizens' Rights Balancing Business and Societal Interests • Civil rights laws allow a business to base its decisions on relevant observable characteristic other than membership in a protected class. • In making a loan-approval decision, for example, a lender is allowed to consider observable factors that predict default, such as past credit history. • A credit rating agency should be allowed to base its ratings on any observable factors that do not undermine widely held views of fairness.

10. Municipal Bond Ratings and Citizens' Rights Bond Ratings and Borrowing Costs • Simonsen and Robbins (1996) examine TIC for 210 GO bonds issued by municipalities in Oregon in 1992 and 1993. They find that the difference in TIC between Baa and AA bonds was 0.40 percentage points. • Robbins and Simonsen (2007) examine TIC for 161 local government GO bonds in Missouri in 2004 and 2005. They find that AAA bonds have interest rates that are 0.4 percentage points lower than unrated bonds. • For revenue bonds issued between 1992 and 1994 by New Jersey and its authorities, Robbins (2002) finds that TIC is about 0.9 percentage points lower for uninsured AAA bonds than for Baa or unrated bonds.

11. Municipal Bond Ratings and Citizens' Rights The Logic of Bond Ratings • Bond ratings are linked to predicted probabilities of default, that is, of late payment or nonpayment of bond obligations. • Bond rating agencies build models of the determinants of default, • and then translate the predictions from these models into credit ratings; a higher predicted probability of default results in a lower rating.

12. Municipal Bond Ratings and Citizens' Rights The Default Regression • Let D stand for default (1 = default in the sample period; 0 = no observed default), Bi stand for the i-th bond characteristic, and Gj stand for the j-th characteristic of the issuing government and the jurisdiction it serves. • Then the agency can build a rating scheme system by estimating the following regression:

13. Municipal Bond Ratings and Citizens' Rights Predicting Default • The rating problem applies to prospective bond issues for which the B’s and the G’s are observed but D is not. • Using the estimated coefficients (weights) and the actual values of the B’s and the G’s, the agency can predict default.

14. Municipal Bond Ratings and Citizens' Rights The Rating Scheme • The agency then translates the probability of default into a rating through a monotonic, inverse function (i.e., higher predicted default = lower rating):

15. Municipal Bond Ratings and Citizens' Rights Add City Demographic Characteristics • With demographic traits, M, the equations are: • But this is equivalent to disparate-treatment discrimination (redlining) because it is based on M!

16. Municipal Bond Ratings and Citizens' Rights Avoiding “Discrimination” • To avoid the analog of disparate-treatment discrimination, use the same regression, but change the prediction and rating scheme to: • This approach accounts for M but does not include variation in M in the rating scheme.

17. Municipal Bond Ratings and Citizens' Rights Back-Door “Discrimination • A default prediction and rating scheme that appears to be group neutral just ignores M: • But this is analogous to disparate-impact discrimination.

18. Municipal Bond Ratings and Citizens' Rights Proving Disparate-Impact “Discrimination” • The standard theorem for omitted variable bias implies that • where γis the true coefficient of M and ri is the correlation between variable iand M. • Thus, the apparently group-neutral scheme (3) has weights that reflect minority status!

19. Municipal Bond Ratings and Citizens' Rights Regulatory Tool Based on Time-of-Issue Information • Data on defaults are not required; this tool builds on an expression for the unobserved default prediction by the rating agency, say D*: • This prediction feeds into a rating scheme, which can be observed. Let k=1 be the highest ranking (AAA). Then c in the following equation measures “discrimination”:

20. Municipal Bond Ratings and Citizens' Rights Problems with This Approach • The estimate of c is likely to be biased, upward or downward, if variables in B or G that are used by the rating agency are left out of the regulator’s regression. • This approach cannot determine whether the factors included in the regression are the factors actually used by the agency. • Indirect use of M, which is analogous to disparate-impact discrimination, can be hidden in the weights placed on other explanatory variables; this equation does not reveal whether the estimated weights accurately reflect the ability of an included variable to predict default.

21. Municipal Bond Ratings and Citizens' Rights Tools that Incorporate Default Information • These problems can be solved with additional information, namely, information on actual defaults. • If defaults are observed, the regulator can estimate the impact of B and G on default and then determine if a rating scheme is consistent with this relationship. • Rating schemes that reflect M after controlling for the impact of B and G on default involve “discrimination.”

22. Municipal Bond Ratings and Citizens' Rights The Special Case of General Obligation Bonds • GO bonds rarely default. • A recent study of GO bonds rated by Moody’s found one GO default from 1970 to 2000 and no GO defaults from 2001 to 2006. • Another study by S&P identifies 14 GO monetary defaults during the 1990s. Except for Orange County, CA, which defaulted on three short-term note deals in 1995, the amounts involved were small.

23. Municipal Bond Ratings and Citizens' Rights A Regulatory Shortcut • The fact that GO bonds do not default means that B and G do not predict default—and there is no need to control for them in evaluating a rating scheme. • An ordered logit regression of bond ratings on M determines whether there is “discrimination,” either disparate-treatment or disparate-impact. • One can control for the maximum probability of default consistent with the observed lack of default.

24. Municipal Bond Ratings and Citizens' Rights Table 2. Results for Standard and Poor's Large City Ratings, 2000 and 2007

25. Municipal Bond Ratings and Citizens' Rights Table 3. Results for Moody's Large City Ratings, 2000 and 2007

26. Municipal Bond Ratings and Citizens' Rights Table 4. Results for Fitch Large City Ratings, 2002 and 2007

27. Municipal Bond Ratings and Citizens' Rights Table 5. Results for Moody's City Ratings, 2002

28. Municipal Bond Ratings and Citizens' Rights Magnitudes • These effects are large. • With an estimated coefficient of 0.07, the average for percent black, the probability that a city will fall into one of the lower-rated categories increases by about 1.5 percentage points for every 1.0 percentage point increase in the black population. • Suppose 25% of bonds are rated Baa and the rest are rated AA. Then a 50 percentage point increase in the minority share would raise the probability of a Baa rating by 66.7 percentage points. • Using the interest-rate impact estimated by Simonsen and Robbins (0.4 percentage points), this population change would raise the expected interest rate by (0.667)∗(0.004) =0.002668, or $2,668 per million dollars of GO bonds

29. Municipal Bond Ratings and Citizens' Rights Conclusions • In the case of city GO bonds, the municipal bond ratings agencies appear to act in a manner that is analogous to redlining. • Although it does not violate existing civil rights laws, this type of behavior violates the societal norms that these laws express. • CRARA makes it illegal for the SEC to investigate this type of behavior—an obviously unsatisfactory situation.

30. Municipal Bond Ratings and Citizens' Rights Policy Recommendations • More regulation of municipal bond rating agencies is needed; the CRARA provision that proscribes the regulation of bond ratings should be repealed. • Moreover, new legislation should prohibit the use of a jurisdiction’s racial or ethnic composition in a bond rating, either directly or indirectly. • This regulation need not be intrusive and it certainly need not undermine the valuable service these agencies provide. • This legislation should give the rating agencies the right to appeal.