Amidst the rising angst among financial services investors last week comes some fresh data that indicates subprime losses won’t be as bad as everyone seems to think.

Ironic, I know. The data in question come from the June servicer reports, filed last week, for the loans that make up the ABX subprime loan indices. The numbers support our belief that the leading indicators of subprime credit quality are improving (yes, improving) and that the soaring loss forecasts being thrown around by some observers, most notably the rating agencies, will almost surely turn out to be too pessimistic.

Refresher: for months, we’ve analyzed credit trends in the subprime mortgage market by looking at the specific loan trusts that make up the ABX, the index designed to be a proxy for the market overall. We’ve looked in particular at the four indices that cover the period of loosest underwriting, the ABX 06-1, 06-2, 07-1, and 07-2.

Realized losses from the trusts are still relatively low, in large part because the indices are still so new, and because it can take awhile for a problem loan to become a realized loss. So while some estimates on cumulative losses run to 30% or more, most of those losses have yet to occur. Here’s where realized losses stood at the end of May:

Again, the vast majority of losses haven’t happened yet. We get to our estimate of eventual cumulative losses via a three-step process:

First, tally realized losses to date.

Second, estimate losses from loans that are already past due, assuming current roll rates and severity don’t change.

Finally, estimate losses from performing loans that are still current, using an estimated default rate and estimated severity.

We estimate the majority of losses in all four ABX indices will come from loans currently 60 or more days past due, in foreclosure, or already repossessed.  Cumulative losses from this group of loans won’t likely be higher than we expect, given the very high roll rates we assume (mid-90%) and 55% severity.

So that leaves us with the estimated losses from loans that are still performing. The key to forecasting losses here are prepayment speeds, the pace of new delinquencies, the rate at which delinquent loans roll from early-stage buckets to later-stage ones, and severity.

Let’s take those four items in order. First, prepayment speeds. They have been low so far this year, and stable. We don’t envision prepay speeds accelerating any time soon, but if they do, cumulative loss estimates would fall.

Next is the pace of new delinquencies--the rate at which good loans go bad.  This number has improved steadily all year, and has provided the biggest reason for encouragement on eventual losses. New delinquencies can be measured several ways; we’ll look at two. 

The first four charts show loans 31 to 60 days past due for the four ABX indices at issue.  The 30-day delinquency buckets, which can be highly volatile month-to-month, are the earliest indication of future problems.

Note the dollar amount of loans 31 to 60 days delinquent in the 06-1 index has declined for eight straight months, and is now 34% below the peak set last October. That’s clearly a positive trend! Take a look:



 In the other three indices, loans 31 to 60 days past due peaked in January and are down by 12% to 15%:

Alternatively, you can look at loans 31 to 60 days past dues as a percentage of the trust’s balance at the start of the month. We call this the inflow rate, and show the rates for the four indices in the charts below.








The story’s the same. Notice the 06-1’s inflow rate peaked in November last year, the inflow rate for the 07-1 and 07-2 peaked in January, and the 06-2 inflow rate peaked in February.  Those, too, are positive trends!

The third predictor of cumulative losses is roll rates. We find the monthly roll rate data from Clayton Holdings, a mortgage surveillance firm, to be more indicative of what’s happening than the data from the servicer reports.  The Clayton numbers show roll rates improving, but in our forecast of cumulative losses, we assume they remain at the current high levels.

Severity is the last piece of the puzzle in estimating cumulative losses.  We assume severity will average 55% going forward.  It will most likely rise some over the coming 18 months, but fall after that.  Each 10-percentage-point change in severity raises our cumulative loss assumption by 3 to 4 percentage points, depending on which index we’re looking at.

To sum up, we believe most forecasts of cumulative subprime mortgage losses are much too high because they overestimate losses from loans originated in 2006 and 2007. Having reviewed the May loan performance data put out by loan servicers, we are more confident in that view than ever. The key piece of data: new problem loans, which spiked last fall, have been declining for months.

Over the short term, emotion can overwhelm analysis in influencing stock prices--both up and down.  In the long run, though, good analysis that leads to the correct fundamental conclusion will almost inevitably be financially rewarding.  Based on the data, I continue to believe subprime mortgage cumulative loss estimates are way too high!

What do you think? Let me know!