Yes, The Bankers Lied
(CC photo of Wall Street sign via Wikimedia)
One of the lingering questions from the financial crash is whether the banks that sold junk subprime mortgage-backed securities were honestly deluded themselves, or consciously misrepresented their product. A new paper from the Center for Real Estate at the Columbia Business School suggests that, at least some of the time, they were lying.
Here’s how we know. During the 2000s, a very large share of residential mortgages were originated by relatively thinly-capitalized “mortgage banks” that held the mortgages only until they had a sufficient number to sell to a money center bank that, in turn, would package them into bonds secured by the underlying mortgage cash flows and sell them to investors. In the later years of the housing bubble, the mortgages underlying such securities were heavily concentrated in subprime properties, since they carried sufficiently stiff interest rates to make them attractive to investors after all the supply-chain players had pocketed their fees.
The study’s authors examine a massive $2 trillion database consisting of nearly all non-agency residential mortgaged-backed securities (ones not guaranteed by Fanny or Freddie) sold to investors between 2005 and 2007. The buyers were sophisticated investors, predominately institutions like pension funds, endowments, or middle-market banks. The authors then compared borrower data provided to investors with data on the same borrowers maintained by a large private credit agency. They focused on two specific data points – whether the home was actually the borrower’s primary residence, and the presence of a second lien. Mortgages on second homes or investor properties and first mortgages on homes encumbered by second mortgages are both historically among the more likely to default.
The researchers were able to pair data from the two databases across a very large sample of the loans. They were conservative in their definition of incorrect data. For example, they did not count a Home Equity Line of Credit (HELOC) as a second mortgage – although it is – and they did not label a property non-owner occupied unless the borrower listed a different primary residence for at least a full year after the mortgage closing.
The results showed that 13.6 percent of all the loans misrepresented one or the other of those two data points. More than a quarter of the loans with non-owner occupants were reported as primary residences, and 15 percent of loans with non-HELOC second mortgages were reported as lien-free. (It’s actually difficult for a bank not to notice a second lien.) As might be expected, default rates on such loans were significantly higher than defaults on loans made to owner-occupants and without second liens.
The authors then looked at two interest rates – the rate charged by the banks to the underlying borrowers, and the rates on those mortgages paid to the investors in the securities. On average, the underlying borrowers paid interest rates to the banks commensurate with the actual risk; but the banks paid their investors rates commensurate with represented risk. The natural inference is that the banks knew what they were doing.
The authors calculate that if the banks were forced to buy back their misrepresented loans – as virtually all the loan documents require them to do – it would cost them about $160 billion. That is probably a low estimate. In 2010, the Charles Schwab Corporation filed two California actions against a very wide swath of major banks engaged in creating and selling residential mortgage-backed securities. The suits were supported by detailed research on the underlying facts in a database of 75,000 securitized mortgages Schwab had purchased for their investors during the housing bubble. They looked at more variables than the Columbia researchers did, and found a total error rate at least twice as high. Like the Columbia researchers, the Schwab investigators found a consistent error-skew toward omitting unflattering information on the individual loans. (The case is still locked up in a procedural tango.)
In other words, some substantial portion of the bubble was created by fraud.