I am continually surprised by the overgeneralized predictions about eventual subprime mortgage credit losses lately being thrown around by people who ought to know better. As I’ve discussed here before, Egan Jones’ Sean Egan came up with his preposterous estimate that the guarantors need $200 billion in new capital based on . . . well, no research at all. Bill Gross says that eventual credit default swap losses will tally to $250 billion. How does he come up with that number? He doesn’t say.
In fact, it’s possible to subject these back-of-envelope predictions of Armageddon to a reasonableness test. How? By looking at the actual credit performance of individual mortgage-backed securities issued as the subprime market began to crack up. The most widely followed are the MBS that underlie the ABX indices issued over the past two years, the ABX 06-1, 06-2, 07-1, and 07-2. They represent (as was the plan when they were included in the indices) a meaningful cross section of subprime mortgages underwritten in those time periods.
And as you’ll see in a minute, the credit quality among the different bonds varies enormously. Despite what the talking heads on CNBC would have you believe, not every subprime mortgage written over the past two years is bound to default—and not every MBS issued over that period is headed for a crackup. There is a huge disparity that’s not reflected in the alarmists’ broad-brush predictions of calamity.
Let’s take a look. I’ve pulled out the best- and worst-performing of the 20 loan pools that make up the ABX 06-1 and the 06-2. First, the 06-1:
First moral of story: underwriting discipline makes a difference. And by all appearances, Well Fargo (WFC) is a superior subprime mortgage underwriter than its competitors, particularly WMC (now a unit of General Electric (GE)). Notably, Moody’s (MCO) (who’s apparently as panicked about what’s going on as anyone) acknowledged this very fact last fall when it broke out subprime originators into three tiers, from worst to best. Wells Fargo was, not surprisingly, ranked, among the best.
WMC, by contrast, was in the lowest tier, while Countrywide (CFC) was in the middle. According to Moody’s, delinquency rates of loans originated by Tier 3 lenders have historically run 2.4 times higher than those originated by Tier 1 lenders.
In the two pools we’re looking at here, the Wells Fargo-originated loans have shown a 30% lower delinquency rate than the WMC-Countrywide pool. Moreover, since the Wells pool has a much lower level of balances still outstanding, we estimate its lifetime losses will be only one-third of the mixed pool’s.
That is an enormous difference. But, again, it’s not a tough conclusion to come to if you’re willing to look at the underlying data. What’s more, the difference between the two pools hasn’t come about by coincidence. Underwriting rigor really does count. Take a look:
Notably, the average FICO on the WMC pool is higher than on the Wells pool—and doesn’t even count as subprime by some standards. But the WMC pool is a mess, just the same. The key to the outperformance of the Wells Fargo pool is just what you might expect: a much higher percentage of full-doc loans and much lower exposure to California.
This shouldn’t be a surprise. If you underwrite well, even if you’re underwriting a subprime mortgage, you will very likely get paid back.
My point is that the data for all 80 of these MBS (which were selected to be representative of the market as a whole, remember) is there for anyone who wants to look. Yes, the performance of some of the bonds is truly ugly. But others are holding up remarkably well. If you want to come up with a reasonable estimate of how high eventual losses will be, all you have to do is go through the securities one by one. Yes, that makes for a lot of numbers to crunch. But the analytical work isn’t really all that hard.
Now let’s look at the best and worst loan pools in the ABX 06-2:
Once again, the best-performing pool was also originated by Wells Fargo, and that the worst was originated by WMC. Sense a pattern?
This is why I go apoplectic when I read analysts’ sweeping estimates of huge losses, that were arrived at with no apparent consideration for the characteristics of individual bonds, such as who the underwriter is. The data is there. People ought to look at it.
For ourselves, we have looked at the bond-by-bond data, and we don’t arrive at estimates of cumulative losses from subprime mortgage lending that are anywhere near as high as the huge, back-of-the-envelope numbers that keep making their way into the headlines.
Granted, we might not be right in the end. But I prefer the rigor of our approach of looking at the granular data, and then updating every 30 days as new numbers come out, to the finger-in-the-wind methodology that others are apparently using. The big numbers might get the publicity. But the bottoms-up ones are far more likely to be right in the end.
Tom Brown is head of BankStocks.com.