The geeky elephant in the room

I’ve seen a lot of stuff on the Web about the crisis on Wall Street, but mostly it’s people whipping their own pet horses so I’ll try to provide y’all a fair reckoning of what went down and why.

The first factor in the crisis is the easy money regime ruling the Fed over the last decade. They had a lot of reasons for shoveling cash out the door of the way they did (reasons I won’t get into here), but for every action there’s a consequence—and in the world of symbols, the reaction ain’t always equal and opposite like it is in the physical world. That’s factor one.

As I think we all know by now, lenders tried to put this cash to work by drumming up as much business as possible in the mortgage markets. There were powerful incentives for them to loosen underwriting standards beyond simple greed. Congress had mandated that banks focus at least some of their efforts towards "disadvantaged" communities, but the only way to comply with the law was to accept some sub–prime applications. That’s factor two.

These two factors put the chips on the table, but it was Wall Street who decided to shove the stack all-in. Like Mikey McD in the opening scene of Rounders, they thought they had a lock. But Fortuna disguised herself as Teddy KGB and stacked ‘em for their overconfidence. It’s not Greenspan or the bleeding-heart liberals who caused the current meltdown, it was the know-it-all math geeks making the big bucks on the Street.

Now mortgage lenders don’t particularly care about the borrower’s credit as long as they can sell the loan to someone else. Most of the companies that write mortgages sell them to the big Wall Street houses, who in turn smash the mortgages together into a bond and sell them on the Street. The factors that go into pricing the security generally have to do with the dependability of that cash flow in relationship to the prevailing rate of return that the street expects from bonds. For Fannie Mae bonds, that dependability was considered to be pretty high—after all, the taxpayer had pretty much guaranteed that they would make up any losses.

But the buyers wanted a clearer idea of the risk inherent in the mortgage-backed bonds they were buying. Some wanted to make more money for more risk, and some were willing to accept lower return for less risk. So the houses build financial products out of these mortgages using a process known as securitization. You can pretty much securitize anything as long as it has a cash flow associated with it. The bonds were split into what are called tranches, with the lowest-risk (and lowest return) bonds in the highest tranche, and the highest-risk (and highest return) bonds going into the sub-prime tranche. Each tranche had thousands of different mortgages in them, all with the same risk of default as determined by the math-types. This way, if any one mortgage defaulted, the bond itself was able to absorb the loss with the thousands of other mortgages in it that didn’t default.

What happened is the buyers of these sub-prime mortgages miscalculated in a very fundamental way the kind of risk they were getting paid to take by owning these bonds. The way they calculated the risk of the bonds was figuring out the average default rate on the underlying mortgages. Just as an example, let’s say the default rate on prime-rate mortgages has historically been zero. The return on that investment will be roughly the same as the interest rates paid by the mortgages. Then let’s say that the default rate on sub-prime mortgages has historically been two percent. So the premium paid by the sub-prime mortgages will factor that into the price and pay a return rate that compensates investors for that two percent default rate. All very nice and clean, yes?

Well, not exactly. The problem lies in the assumptions that the quants made when assessing the risk. Can you see what it is in the paragraph above? You know what it is. We’ve all heard and read the term so often that our eyes just glide over it without a thought. Past performance is no guarantee of future results. Historical averages fail to describe anything but the historical average. Every quant shop pays lip service to this fact, but when they assess risk profiles, they assume the opposite. Otherwise, there’s no way to use a Value at Risk (VAR) model because it uses historical daily price volatility as the key measure or risk.

The problem was made even worse because Wall Street forgot even the lip service that they paid to uncertainty. By treating their VAR models as gospel, they discovered that there was a way to make even more money! By using leverage! With the markets awash in cash due to the easy money described in point one, mortgage-backed security investors would borrow money at cheap rates and buy more mortgage bonds that returned a higher rate than what they had to pay on the loan. And bonus—they could use the mortgage bonds as collateral against the loan because the lenders were using the same VAR models! Hooray—free money! By the time the party ended, the whole financial system was on the hook: investors, lenders, and insurers.

What this crisis boils down to is that Wall Street has no idea how to assess the non-linear nature of market risk but they build a lot of prettied-up linear models to pretend that they’ve got everything covered. And in normal markets, they do. But when things go wrong, they go a lot wronger because of these normalized assumptions.

Again, I refer you to the Simon quote in the blog header.