arxiv: 2604.20877 · v1 · submitted 2026-04-11 · 💱 q-fin.RM · stat.AP· stat.ME

Recognition: unknown

When AAA Satisfies Nothing: Impossibility Theorems for Structured Credit Ratings

Marco Pollanen

Authors on Pith no claims yet

Pith reviewed 2026-05-10 16:35 UTC · model grok-4.3

classification 💱 q-fin.RM stat.APstat.ME

keywords credit ratingsstructured financeBayes theoremAAA ratingsbase ratesdiscriminationimpossibilityrisk management

0 comments

The pith

The AAA precision claim for structured products exceeded what pre-crisis information could support.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper applies Bayes' theorem to demonstrate that any asserted reliability target for a credit rating requires a minimum level of statistical discrimination between instruments that will repay and those that will not. At the low base rates characteristic of structured finance, a four-nines reliability target requires discrimination on the order of 10,000 to 1. Published credit-prediction research provides no affirmative evidence that discrimination of this magnitude was achievable with the data available before the crisis. In contrast, corporate AAA ratings remain feasible under the same logic because higher base rates and richer information lower the required discrimination. The realized shortfalls in structured ratings reached roughly 90,000-fold, indicating that the AAA label asserted more precision than the information environment could deliver.

Core claim

Bayes' theorem shows that any reliability target for a credit rating imposes a minimum discrimination requirement between instruments that repay and those that do not. For structured finance at low base rates, a four-nines target requires discrimination on the order of 10,000:1, while a three-nines target requires 1,000:1. No evidence exists in the credit-prediction literature that this level of discrimination was attainable with pre-crisis information. Realized shortfalls reached approximately 90,000-fold. The same framework explains why corporate AAA ratings were historically feasible and indicates ongoing tensions in modern collateralized loan obligations.

What carries the argument

Application of Bayes' theorem to derive the minimum statistical discrimination needed to achieve a given posterior reliability target at given base rates.

If this is right

AAA ratings for structured products were not supportable by the available pre-crisis information.
Corporate bond AAA ratings remain consistent with the information environment under this analysis.
Current CLO ratings may still face similar mismatches between precision targets and achievable discrimination.
The realized performance of rated structured products fell far short of the implied reliability benchmarks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Rating agencies could improve transparency by reporting the discrimination levels implicit in their ratings.
This approach could be extended to evaluate other predictive claims in finance where base rates are low.
Modern data and models might be tested to see if they now meet the required discrimination thresholds for AAA claims.
Investors in structured products should discount ratings that do not account for base-rate effects.

Load-bearing premise

The premise that AAA ratings assert a four-nines level of posterior reliability and that structured finance repayment base rates are low enough to necessitate extreme discrimination to reach that target.

What would settle it

A published pre-crisis credit model demonstrating at least 10,000:1 discrimination between repaying and defaulting structured instruments, based solely on data available at the time, would falsify the central impossibility result.

Figures

Figures reproduced from arXiv: 2604.20877 by Marco Pollanen.

read the original abstract

A credit rating of AAA asserts near-certainty of repayment. This paper asks whether the pre-crisis information environment could have supported that assertion for structured products. Bayes' theorem implies that any reliability target requires a minimum level of statistical discrimination between instruments that will repay and those that will not. At structured-finance base rates, a four-nines reliability target demands discrimination on the order of 10,000 to 1. A three-nines target demands 1,000 to 1. Nothing in the published credit-prediction literature provides an affirmative basis for believing that discrimination of this magnitude was achievable with the data available at rating time. Retrospectively, the realized system fell short of the four-nines benchmark by roughly 90,000-fold. The framework accommodates the historical feasibility of corporate AAA ratings, where high base rates and rich information produce low required discrimination. Illustrative calibrations for contemporary collateralized loan obligations suggest that material tension between the precision target and the information environment persists. The central implication is that the AAA precision claim itself likely exceeded what the available information could support.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Bayes theorem applied to structured ratings reveals a steep information requirement for AAA that pre-crisis models likely couldn't meet, though the result is sensitive to the chosen reliability target.

read the letter

The main thing to know is that this paper uses Bayes' theorem to calculate the level of discrimination needed to justify AAA ratings in structured finance, concluding that the required ratios were too high to be realistic given available data. It does a good job laying out how a high posterior reliability target combined with low default base rates forces a large likelihood ratio. The 10,000 to 1 figure for four-nines reliability is derived directly, and the paper nicely shows why corporate AAA was more feasible with higher base rates. The retrospective shortfall calculation and the application to modern CLOs add concrete illustrations. The argument is logically consistent on its own terms. What is new is the specific threshold derivation for structured products rather than a general critique. The soft spots are in the inputs. The paper takes a four-nines posterior as the AAA standard without much discussion of why that number or whether rating agencies or regulators viewed it that way. The structured finance base rates are asserted to be low enough for the 10,000:1 result, but without sensitivity analysis or clear data sources, small shifts in those values change the conclusion substantially. The claim about the credit-prediction literature not supporting such discrimination is stated but would benefit from more explicit comparison to existing models. This paper is for readers focused on the quantitative foundations of rating reliability and post-crisis analysis. Someone looking for a probabilistic explanation of rating limits will find it useful. It has enough structure to warrant peer review, as the core idea is worth testing and refining. I recommend sending it to referees with notes to strengthen the justification for the key parameters.

Referee Report

2 major / 2 minor

Summary. The manuscript applies Bayes' theorem to structured credit ratings, demonstrating that a four-nines posterior reliability target for AAA ratings requires likelihood ratios of approximately 10,000:1 at low base rates typical of structured finance. It concludes that the pre-crisis literature on credit prediction does not support the attainability of such discrimination levels with contemporaneous data, rendering the AAA assertion informationally unsupported. The analysis contrasts this with corporate bond AAA ratings, where higher base rates lower the required discrimination, and provides retrospective and forward-looking calibrations for CLOs.

Significance. If substantiated, this work supplies a principled quantitative argument for why AAA ratings on structured products were inherently problematic given the information environment. By formalizing the discrimination requirement via Bayes' theorem and linking it to empirical literature, it offers a framework that could inform rating standards and regulatory oversight. The paper merits credit for its transparent use of probabilistic reasoning, its consistency with the corporate rating case, and its illustrative application to current products. It contributes to the risk management literature by highlighting precision-information mismatches.

major comments (2)

[§2 (Bayes derivation)] §2 (Bayes derivation): The claim that a four-nines target demands discrimination on the order of 10,000:1 depends on the specific choice of posterior target t=0.0001 and base rate b without historical citation or sensitivity checks. The relation LR ≈ b / t (for small b) means that t=0.001 or a modestly higher b reduces the requirement by an order of magnitude, directly affecting whether the literature gap is decisive for the central impossibility claim.
[Literature review section] Literature review section: The assertion that 'nothing in the published credit-prediction literature provides an affirmative basis' for the required discrimination levels is load-bearing but lacks a systematic protocol or explicit reporting of the highest discrimination or AUC values found in pre-2008 studies on structured or asset-backed securities.

minor comments (2)

[Abstract] Abstract: The reference to 'illustrative calibrations for contemporary collateralized loan obligations' would be clearer with a parenthetical note on the data sources or parameter values employed.
Notation for base rates, posterior targets, and likelihood ratios should be defined explicitly at first use and used consistently.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for these constructive comments, which highlight opportunities to strengthen the robustness and transparency of our analysis. We address each major point below and will incorporate revisions accordingly.

read point-by-point responses

Referee: The claim that a four-nines target demands discrimination on the order of 10,000:1 depends on the specific choice of posterior target t=0.0001 and base rate b without historical citation or sensitivity checks. The relation LR ≈ b / t (for small b) means that t=0.001 or a modestly higher b reduces the requirement by an order of magnitude, directly affecting whether the literature gap is decisive for the central impossibility claim.

Authors: We agree that explicit sensitivity analysis and citations for parameter choices would improve the section. The four-nines target (t = 0.0001) reflects a standard interpretation of AAA as near-certainty of repayment, aligned with pre-crisis rating agency statements and regulatory expectations for such ratings. Base rates are taken from historical default frequencies in structured finance. In revision we will add a dedicated sensitivity subsection (or table) reporting required likelihood ratios across t = 0.001, 0.0001, and 0.00001 and across a documented range of base rates drawn from the cited historical sources. This will clarify that the order-of-magnitude gap with the empirical literature persists under plausible alternative calibrations. revision: yes
Referee: The assertion that 'nothing in the published credit-prediction literature provides an affirmative basis' for the required discrimination levels is load-bearing but lacks a systematic protocol or explicit reporting of the highest discrimination or AUC values found in pre-2008 studies on structured or asset-backed securities.

Authors: We accept that greater explicitness is warranted. Our literature review drew on major pre-2008 studies of credit-risk prediction for both corporate and structured/asset-backed securities; the highest AUCs reported in those works fall in the 0.65–0.85 range, which translate to likelihood ratios well below the 1,000:1–10,000:1 thresholds required by the Bayes relation at structured-finance base rates. To make this transparent we will expand the section to include (i) a brief description of the search protocol (databases, keywords, date restrictions) and (ii) a summary table listing the highest discrimination metrics (AUC, accuracy, or odds ratios) from the key studies examined. This will document that no pre-2008 paper supplied affirmative evidence for the discrimination levels needed to support the AAA claim. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation applies standard Bayes' theorem to explicit assumptions and external literature

full rationale

The paper states a reliability target t, invokes Bayes' theorem to obtain the required likelihood ratio LR ≈ (1-t)/t * b/(1-b) at base rate b, asserts that no published credit-prediction work demonstrates such discrimination was attainable pre-crisis, and notes that corporate AAA is feasible under higher b. None of these steps reduce by construction to fitted parameters, self-citations, or renamed inputs; the target t and base-rate values are stated as modeling choices rather than derived from the conclusion, and the literature-gap claim is external. The retrospective shortfall comparison inherits the same parametric structure but does not create self-definition or load-bearing self-reference.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on Bayes' theorem together with assumed base rates for structured products and a chosen high reliability target for AAA; these are not derived from first principles within the paper.

free parameters (2)

structured finance base rate
Prior probability of repayment/default used to compute the discrimination required for a given posterior reliability.
AAA reliability target
Posterior probability threshold (e.g., 99.99%) taken to define the meaning of AAA.

axioms (1)

standard math Bayes' theorem
Standard rule for updating probabilities with new information, invoked to relate base rates, discrimination, and reliability.

pith-pipeline@v0.9.0 · 5488 in / 1215 out tokens · 69471 ms · 2026-05-10T16:35:44.568381+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

27 extracted references · 18 canonical work pages

[1]

Acharya, V.V., Schnabl, P., and Suarez, G. (2013). Securitization without risk transfer. Journal of Financial Economics 107(3), 515--536. https://doi.org/10.1016/j.jfineco.2012.09.004

work page doi:10.1016/j.jfineco.2012.09.004 2013
[2]

Benmelech, E., and Dlugosz, J. (2009a). The alchemy of CDO credit ratings. Journal of Monetary Economics 56(5), 617--634. https://doi.org/10.1016/j.jmoneco.2009.04.006

work page doi:10.1016/j.jmoneco.2009.04.006 2009
[3]

Benmelech, E., and Dlugosz, J. (2009b). The credit rating crisis. NBER Macroeconomics Annual 24, 161--207. https://doi.org/10.1086/648293

work page doi:10.1086/648293
[4]

BlackRock. (2025). What are collateralized loan obligations (CLOs)? https://www.blackrock.com/us/financial-professionals/insights/what-are-clos

2025
[5]

Bolton, P., Freixas, X., and Shapiro, J. (2012). The credit ratings game. Journal of Finance 67(1), 85--112. https://doi.org/10.1111/j.1540-6261.2011.01708.x

work page doi:10.1111/j.1540-6261.2011.01708.x 2012
[6]

Casscells, W., Schoenberger, A., and Graboys, T.B. (1978). Interpretation by physicians of clinical laboratory results. New England Journal of Medicine 299(18), 999--1001. https://doi.org/10.1056/NEJM197811022991808

work page doi:10.1056/nejm197811022991808 1978
[7]

Cordell, L., Roberts, M.R., and Schwert, M. (2023). CLO performance. Journal of Finance 78(3), 1235--1278. https://doi.org/10.1111/jofi.13224

work page doi:10.1111/jofi.13224 2023
[8]

Cornaggia, J., Cornaggia, K.J., and Xia, H. (2017). Credit ratings across asset classes: A long-term perspective. Review of Finance 21(2), 465--509. https://doi.org/10.1093/rof/rfx002

work page doi:10.1093/rof/rfx002 2017
[9]

Coval, J., Jurek, J., and Stafford, E. (2009). The economics of structured finance. Journal of Economic Perspectives 23(1), 3--25. https://doi.org/10.1257/jep.23.1.3

work page doi:10.1257/jep.23.1.3 2009
[10]

DeMarzo, P.M. (2005). The pooling and tranching of securities: A model of informed intermediation. Review of Financial Studies 18(1), 1--35. https://doi.org/10.1093/rfs/hhi008

work page doi:10.1093/rfs/hhi008 2005
[11]

Demyanyk, Y., and Van Hemert, O. (2011). Understanding the subprime mortgage crisis. Review of Financial Studies 24(6), 1848--1880. https://doi.org/10.1093/rfs/hhp033

work page doi:10.1093/rfs/hhp033 2011
[12]

Donnelly, C., and Embrechts, P. (2010). The devil is in the tails: Actuarial mathematics and the subprime mortgage crisis. ASTIN Bulletin 40(1), 1--33. https://doi.org/10.2143/AST.40.1.2049222

work page doi:10.2143/ast.40.1.2049222 2010
[13]

Financial Crisis Inquiry Commission. (2011). The Financial Crisis Inquiry Report. U.S.\ Government Printing Office

2011
[14]

Gigerenzer, G. (2002). Calculated Risks: How to Know When Numbers Deceive You. Simon & Schuster

2002
[15]

Griffin, J.M., and Nickerson, J. (2023). Are CLO collateral and tranche ratings disconnected? Review of Financial Studies 36(6), 2319--2360. https://doi.org/10.1093/rfs/hhac078

work page doi:10.1093/rfs/hhac078 2023
[16]

Griffin, J.M., and Tang, D.Y. (2012). Did subjectivity play a role in CDO credit ratings? Journal of Finance 67(4), 1293--1328. https://doi.org/10.1111/j.1540-6261.2012.01748.x

work page doi:10.1111/j.1540-6261.2012.01748.x 2012
[17]

ICE Data Indices, LLC. (2026). ICE BofA AAA US Corporate Index option-adjusted spread [BAMLC0A1CAAA]. Retrieved from FRED, Federal Reserve Bank of St.\ Louis. https://fred.stlouisfed.org/series/BAMLC0A1CAAA

2026
[18]

Keys, B.J., Mukherjee, T., Seru, A., and Vig, V. (2010). Did securitization lead to lax screening? Evidence from subprime loans. Quarterly Journal of Economics 125(1), 307--362. https://doi.org/10.1162/qjec.2010.125.1.307

work page doi:10.1162/qjec.2010.125.1.307 2010
[19]

Khandani, A.E., Kim, A.J., and Lo, A.W. (2010). Consumer credit-risk models via machine-learning algorithms. Journal of Banking & Finance 34(11), 2767--2787. https://doi.org/10.1016/j.jbankfin.2010.06.001

work page doi:10.1016/j.jbankfin.2010.06.001 2010
[20]

Li, D.X. (2000). On default correlation: A copula function approach. Journal of Fixed Income 9(4), 43--54. https://doi.org/10.3905/jfi.2000.319253

work page doi:10.3905/jfi.2000.319253 2000
[21]

Moody's Investors Service. (2006). Default and recovery rates of corporate bond issuers, 1920--2005. Special Comment, January 2006

2006
[22]

National Association of Insurance Commissioners, Capital Markets Bureau. (2025). U.S.\ insurers' total collateralized loan obligation investments. https://content.naic.org/sites/default/files/capital-markets-special-report-clo-ye2024.pdf

2025
[23]

Neyman, J., and Pearson, E.S. (1933). On the problem of the most efficient tests of statistical hypotheses. Philosophical Transactions of the Royal Society of London, Series A 231, 289--337. https://doi.org/10.1098/rsta.1933.0009

work page doi:10.1098/rsta.1933.0009 1933
[24]

Palmer Square Capital Management. (2026). CLO indices statistics. https://www.palmersquarecap.com/clo-indices/palmer-square-clo-senior-debt-index

2026
[25]

S&P Global Ratings. (2024). CLO spotlight: Thirty years strong: U.S.\ CLO tranche defaults from 1994 through first-quarter 2024

2024
[26]

Skreta, V., and Veldkamp, L. (2009). Ratings shopping and asset complexity: A theory of ratings inflation. Journal of Monetary Economics 56(5), 678--695. https://doi.org/10.1016/j.jmoneco.2009.04.002

work page doi:10.1016/j.jmoneco.2009.04.002 2009
[27]

Structured Finance Association. (2020). Collateralized loan obligations: Balancing crucial business lending with financial stability concerns

2020