2008 MBS Rating Failure
2008 MBS Rating Failure
In 2007-2008, mortgage-backed securities rated AAA by Moody's, S&P, and Fitch defaulted in unprecedented numbers. The simultaneous default of securities bearing the agencies' highest rating destroyed confidence in ratings for a decade and exposed fundamental flaws in the rating methodologies: reliance on historical house-price data from a non-representative period, correlation assumptions that broke down, and perverse incentives in the issuer-pays model.
Key takeaways
- MBS rated AAA were downgraded to junk status within 24 months of 2006-2007 origination
- Rating models assumed house prices never declined nationally and regional declines were independent
- The agencies failed to account for stated-income loans, minimal down-payment borrowers, and low documentation standards
- Mortgage originators had no skin in the game once mortgages were securitized, creating moral hazard
- The crisis permanently damaged confidence in agency ratings and led to regulatory changes
MBS structure and the rating challenge
Mortgage-backed securities bundle 2,000 to 10,000 mortgages into tranches. The safest tranches (senior, AAA) receive principal and interest first. Lower tranches (BBB, BB, unrated) absorb losses first. If 2 percent of mortgages default, senior tranches are unaffected. If 8 percent default, senior tranches are mostly unaffected. If 15 percent default, senior tranches experience losses.
Historically, residential mortgages defaulted at low rates. From 1980-2002, even in recessions, national default rates stayed under 1 percent. This historical data is what rating agencies fed into their models.
The models also assumed correlation independence: if California home prices fell, that had no bearing on Florida or Arizona. Regional diversification would protect against localized downturns. A mortgage pool with borrowers in 50 states would have naturally offsetting regional risk.
This framework wasn't unreasonable for traditional full-documentation, 20-percent-down mortgages. It was catastrophic for 2005-2006 origination featuring stated-income, 3-percent-down mortgages issued by brokers compensated on volume, not outcomes.
The origination shift and underwriting deterioration
In 2002-2006, mortgage origination shifted from banks (balance-sheet lenders, skin in the game) to non-bank mortgage brokers (originate-to-distribute, zero skin in the game). A broker originating a $300,000 mortgage earned a commission whether the borrower could afford it or not. The broker sold the mortgage to an investment bank within 30 days. The investment bank pooled mortgages and sold them to rating agencies for evaluation.
At each step, the incentive was to originate more mortgages, not better mortgages. A loan officer earning $2,000 per mortgage closes 20 mortgages per month by relaxing standards. One earning $2,000 per mortgage with full recourse liability for defaults closes 5 mortgages per month of unquestionable quality.
By 2006, stated-income mortgages represented 30-40 percent of origination in some markets. A borrower could claim $150,000 annual income with no verification. As long as they had a pulse and the house appreciated, the loan performed. But in a flat or declining market, stated-income borrowers walked away.
Rating model blindness to deterioration
Rating agencies didn't ignore the deterioration in underwriting standards — they didn't measure it. The models consumed loan-level data (LTV, FICO, loan purpose) but not origination channel or documentation level. An agency that separately rated stated-income vs. full-documentation mortgages would have downgraded stated-income pools significantly. Instead, agencies applied one model to all, failing to adjust for known quality deterioration.
Furthermore, the historical default data used to calibrate models came from 1980-2002. That period included the Savings & Loan crisis of 1989-1992, the 1991-1992 recession, and the 2001-2002 recession. Each of these produced default rates 2-4 times normal. But none of them produced national house-price declines. Every region recovered within 2-3 years.
The models thus incorporated "house prices always recover" as a hidden assumption. With stated-income borrowers and 3-percent-down mortgages, that assumption was lethal. If house prices fell, borrowers had no equity cushion. Default occurred quickly and in bulk.
Correlation breakdown and simultaneous downgrades
The independent correlation assumption broke down entirely. When housing markets peaked in 2006 and began declining in 2007, the decline was nearly simultaneous across all regions. California, Florida, Arizona, Nevada, Michigan — all saw 20-30 percent declines by 2009. The "diversification" was illusory.
Moreover, the regional decline triggered simultaneous default surges. As house prices fell, borrowers in underperforming regions realized they could walk away. Default clusters appeared in specific ZIP codes and then spread regionally. By mid-2008, mortgage default rates hit 4 percent nationally. By 2009, they hit 7-8 percent. By 2010, they exceeded 11 percent in some cohorts.
These default rates far exceeded what any AAA rating could tolerate. AAA-rated corporate bonds default at roughly 0.05 percent per year over 10 years — total cumulative default probability under 1 percent. Yet MBS pools with 7-8 percent annual default rates were rated AAA.
The agencies' response was to downgrade wholesale. In 2008-2009, they downgraded AAA-rated MBS to CCC, B, and even unrated within 12-24 months of issuance. These weren't mature securities showing unexpected deterioration — these were new originations where the inputs were wrong from day one.
The issuer-pays model and rating incentives
Rating agencies operated on the issuer-pays model. A mortgage securitizer (e.g., Countrywide) hired Moody's to rate their $500 million MBS issuance. Moody's earned $100,000-$500,000 depending on issuance size and complexity. If Moody's downgraded the MBS to BBB, Countrywide simply hired S&P or Fitch instead and published that rating.
This created obvious conflicts of interest. Agencies that were too strict would lose business to competitors. Agencies that were permissive would gain underwriting volume. From 2004-2006, all three major agencies competed on lenient ratings for MBS. Internal emails revealed later showed agency staff expressing doubt about models and ratings but being overruled by revenue-focused management.
An agency rating MBS strictly would see borrowers (mortgage originators) take their business elsewhere. An agency rating leniently would see issuance volume rise 40 percent year-over-year. The incentive structure was clear: lenient ratings won business.
Regulatory response and NRSRO designation changes
The 2008 crisis led directly to legislative and regulatory changes. The SEC tightened Nationally Recognized Statistical Rating Organization (NRSRO) registration, requiring agencies to disclose methodologies, track historical accuracy, and avoid conflicts of interest. The Dodd-Frank Act (2010) mandated that agencies conduct periodic back-testing of models against actual outcomes.
However, the root conflict — issuer-pays funding — remains. Governments have struggled to find alternatives. Investor-pays models (mutual funds or index providers pay for ratings) reduce issuance volume and limit rating availability. Government-pays models (the SEC or central banks rate bonds) are administratively expensive and political.
By 2015, agencies had lost enough credibility that many large investors ignored agency ratings entirely or treated them as floor estimates, not substitutes for analysis. Bond managers at sophistication-level funds now view agency ratings as one input among many — price signals from credit default swap markets, equity performance, and fundamental analysis matter more.
Lessons for investors: when to discount rating changes
The MBS crisis illustrates why investors shouldn't treat rating changes as gospel. Key lessons:
First, ratings are backward-looking. They reflect historical periods that may not repeat. A rating calibrated on 1980-2002 data is useless in a regime where house prices fall nationally.
Second, structural changes in origination or underwriting may not show up in rating models. Stated-income mortgages don't exist as a distinct variable in legacy models — they're just mortgages with worse credit metrics that the model treats as isolated credit issues.
Third, correlation breakdowns are common in stressed environments. Diversification that looks perfect in calm markets may evaporate when risk factors become correlated. Regional housing diversity proved useless when all regions declined simultaneously.
How it flows: the MBS rating cycle
Next
The MBS crisis revealed rating agencies' blindness and conflicts of interest. But it didn't end rating agencies' influence. Institutions still rely on ratings for regulatory capital requirements and investment mandates. The next article examines the explicit conflicts that persist in the issuer-pays model and the post-2008 reforms that attempted to mitigate them.