MARKETS REQUIRE ACCURATE DATA...:

My Manhattan Project: How I helped build the bomb that blew up Wall Street. (Michael Osinski, Mar 29, 2009, New York Magazine)

I wrote the software that turned mortgages into bonds.

Because of the news, you probably know more about this than you ever wanted to. The packaging of heterogeneous home mortgages into uniform securities that can be accurately priced and exchanged has been singled out by many critics as one of the root causes of the mess we’re in. I don’t completely disagree. But in my view, and of course I’m inescapably biased, there’s nothing inherently flawed about securitization. Done correctly and conservatively, it increases the efficiency with which banks can loan money and tailor risks to the needs of investors. Once upon a time, this seemed like a very good idea, and it might well again, provided banks don’t resume writing mortgages to people who can’t afford them. Here’s one thing that’s definitely true: The software proved to be more sophisticated than the people who used it, and that has caused the whole world a lot of problems.

The first collateralized mortgage obligation, or CMO, was created in 1983 by First Boston and Salomon Brothers, but it would be years before computer technology advanced sufficiently to allow the practice to become widespread. Massive databases were required to track every mortgage in the country. You needed models to create the intricate network of bonds based on the homeowners’ payments, models to predict prepayment rates, and models to predict defaults. You needed the Internet to sail these bonds back and forth across the world, massaging their content to fit an investor’s needs at a moment’s notice. Add to all this the complacency, greed, entitlement, and callous stupidity that characterized banks in post-2001 America, and you have a recipe for disaster. [...]

At Lehman, I began a thirteen-year effort to streamline the process of securitizing home mortgages, as well as other forms of debt. That was 1988, around the time of the savings-and-loan crisis. Remember that one? Lenders had gone nuts with, what else, real estate, and as they went bust, the government was stepping into the breach. Mortgage securitization was the answer. Retail lenders could make the loan, take a fee, then sell the mortgage to an investment bank. The bank, after bundling thousands of the mortgages together, could, through a little software magic, issue bonds based on that bundle of loans. Now, an investor does not want a single person’s mortgage, much the same as you may not want to underwrite your sibling’s purchase of an overpriced McMansion. But when 1,000 similar loans are combined, and the U.S. government, through Freddie Mac and Fannie Mae, absorbs the default risk, you now have a nifty little AAA-rated piece of paper paying one or two points above Treasury bills. And if the value of the loans is in excess of the limit set by the government agencies, your savvy friends on Wall Street can create a class of subordinated bonds that will absorb all the defaults in the deal. With friends like these … [...]

CMOs became more complicated, my job was to make everything seem simple—to, in effect, mask the complexity that would’ve made the bonds difficult to trade. We invented a language for mortgage-backed bonds. I called it BondTalk. [...]

Up until that point, almost all my securitization work had involved prime mortgages—those mortgages given to people who had an extremely high probability of paying them back. When a client wanted me to enhance my software to include “subprime” debt, well, that was something new, and I have to admit, I was kind of excited. This would greatly enlarge my universe of clients, because the subprime market was then split among many smaller players, each of whom needed my software.

I quickly learned how fishy this world could be. A client I knew who specialized in auto loans invited me up to his desk to show me how to structure subprime debt. Eager to please, I promised I could enhance my software to model his deals in less than a month. But when I glanced at the takeout in the deal, I couldn’t believe my eyes. Normally, in a prime-mortgage deal, an investment bank makes only a tiny margin. But this deal had two whole percentage points of juice! Looking at the underlying loans, I was shocked.

“Who’s paying 16 percent for a car loan?” I asked. The current loan rate was then around 8 percent.

“Oh, people who have defaulted on loans in the past. That’s why they’re called subprime,” he informed me. I had known this guy off and on for years. He was an intelligent, articulate, pleasant fellow. He and his wife came to my house for dinner. He had the comfortable manner of someone who had been to good schools—he was not one of the “dudes” trying to jam bonds into a Palm Beach widow’s account. (Those guys were also my clients.)

“But if they defaulted on loans at 8, how can they ever pay back a loan at 16 percent?” I asked.

“It doesn’t matter,” he confided. “As long as they pay for a while. With all that excess spread, we can make a ton. If they pay for three years, they will cure their credit and re-fi at a lower rate.”

That never happened.

...but the debt merchants required complexity and obfuscation.

Posted by Orrin Judd at March 31, 2009 8:58 PM