The Impact of the Dodd Frank Act - PowerPoint Presentation

1. The Impact of the Dodd Frank Act on the Determinants of Credit Rating QualityPaul Klumpes, Aalborg University Business School, DenmarkIliya Komarev, Montpellier Business School, FranceKonstantinos Eleftheriou, University of Piraeus, Greece Anne-Lise Ronsse, EDHEC, FranceInternational Conference on Finance, Banking and AccountingMontpellier 09/09/2023

2. OverviewMotivationInstitutional BackgroundPrior LiteratureHypothesesResearch Method Empirical TestsDiscussion of resultsConclusion

3. 1. Institutional MotivationCredit rating agencies (CRAs) under political and economic pressure due to:failure to detect bankruptcies leading to Credit Rating Agency Reform Act of 2006, leading to Dodd-Frank Wall Street Reform and Protection Act of 2010 (DFA),Introduced legal liability for credit rating agency judgements, replacing the former journalistic opinion as the basis for credit rating assessmentsImposed obligation on credit rating agencies to disclose their rating methodologiesHowever, maintaining the issuer-pays model raises questions about conflict of interest (i.e., lack of independence, important ancillary services) and the impact behavioural bias on rating grades.

4. 1. Theoretical MotivationCognitive psychology insights into decision making under uncertainty (post- versus pre-DFA)Prospect theory (Kahneman and Tversky, 1979)Investors switch from risk aversion to loss aversion under ambiguityAnchoring and adjustment (Tversky and Kahneman, 1973)Prior experience results in overreliance on a mean in judging credit qualityTherefore, Increased reliance on cash flow measures of credit qualityConservatism in adjustment of credit rating quality over time

5. 2. Institutional Background (1/2)The role of the credit rating agencies was criticized in relation to the bankruptcy of Enron in 2001SOX required the SEC to provide a report within 18 months concerning the role of credit rating agencies in securities markets. The SEC report (2004) found several deficiencies in the rating process; no regulation because the rating process was not legally binding. The Credit Rating Agency Reform Act of 2006, obliged credit rating agencies to disclose publicly their procedures and methodologies, and to provide the SEC with audited financial statements.

6. 2. Institutional Background (2/2)Financial crisis of 2007-2009 involving defaults/bailouts of several financial institutions despite high-level credit ratingsCongress passed the DFA, following criticism on the role of credit rating agencies in facilitating the financial crisisCreation of an Office of Credit Ratings, empowered to conduct yearly review CRA’s rating methodologies. Requirements related to; disclosure of credit rating methodology and internal controls,ensuring continuing professional education and training standards.Revocation of the previous legislative protection regarding CRA’s liability against legal action: rating grades are no longer “opinions” but binding expert statements.

7. 3. Prior Empirical Literature (1/2)Ratings are typically analysed from the perspective of the credit issuing firm, not the CRA:- Hovakimian et al. (2008) – firm behave strategically to achieve target credit ratings (and target debt levels).Based on this insight, empirical research has established that firms seek to influence the grading through accounting manipulations: - Alissa et al. (2013) find that issuers manipulate ratings trough earnings and real activities managements; manipulations are more important for threshold firms.- Jung et al. (2013) find that firms in the extremes of a credit ‘notch’ manage the volatility of earning to achieve better grade.

8. 3. Prior Empirical Literature (2/2)The impact of Dodd-Frank ActDimitrov et al. (2015) find that post-DFA, the CRAs issue more downward biased ratings and more false warnings, leading overall to less informative ratings.

9. 4. Hypotheses (1/3)Consistent with the predictions of Tversky and Kahneman (1973) and the “anchor and adjustment” heuristic, the passage of the DFA will result in conservatism bias in credit rating decisions;H1a. Pre-DFA there is a statistically significant negative association between determining or changing the credit rating grade and prior variation in credit rating gradeH1b. Post-DFA there is NOT a statistically significant association between determining or changing the credit rating grade and prior variation in credit rating grade

10. 4. Hypotheses (2/3)Consistent with the predictions of Kahneman and Tversky (1979) and prospect theory,H2a. Pre DFA, there is a statistically significant association between determining the credit rating grade and earnings levelH2b. Post DFA, there is a statistically significant association between determining the credit rating grade and cash flow level

11. 4. Hypotheses (3/3)Consistent with the predictions of Kahneman and Tversky (1979) and prospect theory,H3b. Pre-DFA there is a statistically significant association between changes in the credit rating grade and changes in earningsH3b. Post DFA there is a statistically significant association between changes in the credit rating grade and changes in cash flow

12. 5. Research Method (1/4)Random effects panel probit: (1)Pooled ordered probit: (2)Where,i denotes the firm, t denotes the timeΦ(.) is the standard normal cumulative distribution function; are independent and identically distributed ; κ is a set of cutpoints with K denoting the number of possible outcomes (K = 3 for Eq. (2), respectively); y is the dependent binary or ordinal variable (RATE and RATE” in Eq. (1) and CHANGE and RATE’ in Eq. (2)). x is a vector of dependent variables defined by two alternative specifications: (see next slide) 

13. 5. Research Method (2/4)Furthermore,x is a vector of explanatory variables defined by two alternative ways: x = [SDCR, G, H, Z] in the case of the accounting earnings-based model and x = [SDCR, G, L, Z] in the case of the net operating cash flow model, Where,G = [MTB, TANG, RDIND, SGA, SIZE] is a vector of standard credit rating determinants, according to Hovakimian et al. (2008); Z is a vector of dummy variables controlling for the industry fixed effects, clustered into mining, manufacturing, utilities, and retail (Fama and French, 1997); H = [PROFIT, OPRISK] is a vector of earnings-based credit rating determinants and L = [OPCFO, CFO, CHETA] is a vector of cash flow-based credit rating determinants.

14. 5. Research Method (3/4)Definition of independent variables :SDCR, volatility of credit ratings, proxied by the standard deviation of rating over prior five years.MTB, the ratio of a firm’s market value of assets to total assets, where the market value of assets is total assets minus book equity plus market equity. TANG, the ratio of a firm’s net property, plant, and equipment to total assets.RDIND, a binary variable set equal to one if a firm reports R&D expenses, and zero otherwise. SGA, the ratio of a firm’s selling, general, and administrative expenses to sale. SIZE, the natural logarithm of sales.PROFIT, the ratio of a firm’s operating income over lagged total assets.OPRISK, the standard deviation of a firm’s operating income, scaled by lagged total assets over the previous five fiscal years.CFO, the net cash flow from operations over total debt.OPCFO, the standard deviation of a firm’s net operating cash flow, scaled by lagged total assets over the previous five fiscal years.CHETA, the total of cash and cash equivalents over total assets.

15. 5. Research Method (4/4)Data and data sourcesCompustat data for both annual credit ratings (S&P) and company financials.Large companies, belonging to S&P 500, in continuous existence from 2000 to 2015430 companies for the period from 2002 to 2017 (16 years). Unbalanced panel of 6,846 company-year observations.

16. 6. Empirical Tests & Results (1/5)Earnings Model of Credit Rating Quality GradesEquation (1) Period before Dodd-Frank ActPeriod after Dodd-Frank ActVariablesCoefficientsMarginal effectsCoefficientsMarginal effectssdcr-1.306***-0.085***-0.758***-0.053*** (0.182)(0.012)(0.254)(0.019)………..profit2.4660.1623.4570.240 (1.786)(0.116)(2.369)(0.164)oprisk-4.703-0.308-1.242-0.086 (4.822)(0.320)(4.567)(0.320)Constant-16.63*** -15.79***  (2.454) (2.617) Robust standard errors are reported in parentheses. *** p<0.01, ** p<0.05, * p<0.1

17. 6. Empirical Tests & Results (2/5)Cash Flow Model of Credit Rating Quality GradesEquation (1) Period before Dodd-Frank ActPeriod after Dodd-Frank ActVariablesCoefficientsMarginal effectsCoefficientsMarginal effectssdcr-1.328***-0.085***-0.788***-0.053*** (0.188)(0.013)(0.268)(0.019)……….opcfo-5.637-0.362-3.602-0.240 (5.212)(0.334)(5.656)(0.383)cfo4.606***0.296***3.482*0.232* (1.476)(0.098)(2.056)(0.135)cheta0.3820.0254.896**0.326*** (1.673)(0.107)(1.902)(0.122)Constant-16.96*** -16.77***  (2.508) (2.888) Robust standard errors are reported in parentheses. *** p<0.01, ** p<0.05, * p<0.1

18. 6. Empirical Tests & Results (3/5)Earnings Model of Incremental Credit Ratings ChangesEquation (2)Robust standard errors are reported in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Period before Dodd-Frank ActPeriod after Dodd-Frank Act CoefficientsMarginal effectsCoefficientsMarginal effects VariablesDowngradeNo changeUpgradeDowngradeNo changeUpgrade sdcr-0.161***0.030***-0.007**-0.023***-0.0030.00050.0001-0.0006  (0.058)(0.011)(0.003)(0.008)(0.061)(0.009)(0.002)(0.011) ……..profit3.751***-0.691***0.160***0.531***4.440***-0.647***-0.149***0.796***  (0.717)(0.133)(0.046)(0.107)(0.606)(0.093)(0.043)(0.111) oprisk-2.467*0.454*-0.105*-0.349*-1.3890.2020.047-0.249  (1.276)(0.235)(0.060)(0.181)(1.292)(0.187)(0.046)(0.232) 

19. 6. Empirical Tests & Results (4/5)Cash Flow Model of Incremental Credit Rating ChangesEquation (1) Period before Dodd-Frank ActPeriod after Dodd-Frank Act CoefficientsMarginal effectsCoefficientsMarginal effectsVariablesDowngradeNo changeUpgradeDowngradeNo changeUpgradesdcr-0.191***0.036***-0.008**-0.027***  0.0030.001-0.004 (0.059)(0.011)(0.004)(0.008)(0.060)(0.009)(0.002)(0.011)……………OPCFO-2.0410.382-0.0895-0.292-3.696***0.540***0.124**-0.664*** (1.665)(0.311)(0.0752)(0.239)(1.282)(0.187)(0.0551)(0.231)CFO1.304*-0.244*0.0572*0.187*3.243***-0.474***-0.109***0.583*** (0.750)(0.140)(0.0345)(0.109)(0.659)(0.0986)(0.0360)(0.120)CHETA0.174-0.03260.007650.02501.033*-0.151*-0.03470.186* (0.599)(0.112)(0.0263)(0.0857)(0.606)(0.0886)(0.0223)(0.109)Robust standard errors are reported in parentheses. *** p<0.01, ** p<0.05, * p<0.1

20. 6. Empirical Tests & Results (5/5)Robustness Tests1. Sensitivity to rating grade definitions - we use of Threshold grade (e.g., Brown et al., 2015)Results are significantly similar2. Sensitivity to assessment methodology – use of Moody’s more “conservative” credit ratingsResults are significantly similar3. Test for endogeneity (lagged independent variables)Results are significantly similar4. Comparison regression coefficients pre- and post- Dodd Frank ActNo significant differences in these two periods, except for the coefficients of PROFIT significantly different pre- and post-Dodd Frank Act when threshold grade are used.

21. 7. Discussion of ResultsHypothesis H1a – Supported for both setting the grade and changes in the gradeHypothesis H1b – Supported for changes in the grade but rejected for setting the gradeHypothesis H2a – Rejected for setting the gradeHypothesis H2b – Supported for setting the gradeHypothesis H3a – Supported for changing the gradeHypothesis H3b – Supported for changing the grade

22. 8.ConclusionThe DFA of 2010 was expected CRAs to establish more objective assessment of US debt issuersThe passage of the DFA resulted in the followingAn increased reliance on an anchoring and adjustment heuristic, causing CRAs to be more conservative in changing the gradeA change in judgemental focus from an earnings orientation to cash flow orientationHowever, the results are not entirely consistent with our predictionsAfter DFA, there is no association between determining the credit rating grade and prior variation in credit rating gradeBefore DFA, there is no association between determining the credit rating grade and earnings Overall, the expected impact of the DFA is questionable. This conclusion is subject to several caveats of the study: survivorship bias in our sample, only large S&P 500 firms covered, alternative econometric specifications...

23. Thank you!QuestionsSuggestions