All the Devils Are Here [75]
• Another example: Moody’s was initially more conservative on securitizations in cases where, in addition to the first lien, there was a second-lien mortgage. But that was a problem because S&P had a different, looser standard: it concluded in 2001 that as long as second-lien loans were attached to no more than 20 percent of the mortgages in the pool, it would treat the entire pool as if it didn’t have additional risk. “The other agencies took the same position shortly thereafter,” Richard Bitner, a former subprime lender, later told the Financial Crisis Inquiry Commission. He added, “The rating agencies effectively gave birth to the subprime piggyback mortgage.” Those were subprime mortgages in which the homeowner avoided putting up any cash and got two loans—one for the mortgage itself and another for the down payment.
The great advantage issuers had in seeking triple-A ratings is that they rarely needed all three agencies to be involved in any one deal. Investors liked having two agencies rate a deal, but nobody cared about having all three involved. So issuers could play the agencies off each other. They didn’t really care which rating agencies bestowed the rating. All that mattered was the rating itself. “The triple-A was the brand, not Moody’s,” says a former Moody’s structured finance managing director.
Like everyone else utilizing risk models, the rating agencies used the mathematics of probability theory to arrive at their ratings. A given mortgage-backed deal might contain as many as ten thousand mortgages. As every investor is taught, diversification spreads risk, so one question was, how diversified were the mortgages? If they all came from California, they were less diversified than if some were from California, some from Idaho, and some from Connecticut. The working assumption was if home prices dropped in California, they would remain stable, and even keep rising, in other parts of the country. The Wall Street term for spreading risk this way—and there are more complex variants—is correlation. Correlation is essentially a way of describing, in numerical terms, the likelihood that if one security defaults, others would default in tandem. Zero correlation means that one default would have no effect on anything else in the security; 100 percent correlation means that if one defaults, everything else would, too. The closer the mortgage-backed security came to zero correlation, the greater the percentage of tranches that could be labeled triple-A. Underwriters often added credit enhancements to boost the percentage of triple-A tranches.
One obvious flaw of this approach is that nowhere in the process was anyone required to conduct real-world due diligence about the underlying mortgages. As the SEC later noted, “There is no requirement that a rating agency verify the information contained in RMBS loan portfolios presented to it for rating.” (RMBS stands for residential mortgage-backed security.) A second problem is that the rating agency models were built on a series of assumptions. One assumption was that if housing prices declined, the declines would not be severe. Another was that the housing market in California was indeed uncorrelated with the housing market in Connecticut. And then there was the fact that assumptions could be changed. If the bankers didn’t like the outcome of the analysis, maybe a little rejiggering might be in order.
For instance, UBS banker Robert Morelli, upon hearing that S&P might be revising its RMBS ratings, sent an e-mail to an S&P analyst. “Heard your ratings could be 5 notches back of moddys [sic] equivalent,” he wrote. “Gonna kill your resi biz. May force us to do moodyfitch only . . .” Internally, the rating agencies had a term for this: ratings shopping. Even Clarkson acknowledged that it took place. “There is a lot of rating shopping