Subprime Mortgage Loss Trends - and Why Many Cumulative Loss Estimates Are Likely Way Too High

In the format of refuting the conventional wisdom:

Subprime delinquencies are still rising.

True, but beside the point. Yes, it’s a fact that the delinquency rates are indeed rising for each of the four ABX indices that represent the worst-performing subprime vintages, and that seriously delinquent loans are still increasing (as the charts below illustrate). But, delinquency rates are lagging indicators of credit trends.

The reason why has to do with a quirk in the number’s calculation. To figure a loan pool’s delinquency rate, you divide delinquent loans in the pool by the pool’s current outstanding balances rather than its original size. That can distort things: As loans are paid down (yes, that happens) and principal amortizes, the denominator in the equation drops relatively rapidly. Up in the numerator, meanwhile, delinquent loans tend to stick around for months; it takes up to a year for delinquent loan to become a foreclosure and realized loss. Result: as the denominator falls, the delinquency rate can keep rising even in the absence of any newly delinquent loans at all. As I say, this is not a leading indicator of changes in credit trends.

Instead, I focus on two numbers that are much better predictors of changes in credit quality: a) the monthly inflow, in dollars, of new problem loans (defined as loans 30 days to 60 days past due) and b) the rate at which delinquent loans move from early-stage delinquency buckets to later-stage buckets—so-called “roll rates.” And, both those numbers have improved steadily in all four ABX indices so far this year.

The decline in 30-day delinquencies will lead to an eventual decline in the widely watched, but misleading, 60-plus-day delinquency rate. (It has to. Loans can’t jump from current to past 60 days past due in one fell swoop.) And in fact, that’s in the process of happening now. Take a look at the four charts below. They show monthly changes in 60-plus-day delinquencies, in dollars, in the four key ABX indices. In ABX 06-1, seriously delinquent dollars have increased at a slower rate for five straight months, in 06-2, the rate of increase has slowed for four consecutive months, in 07-1, it has slowed for eight straight months, and in 07-2, the rate of increase has fallen for all seven months the index has been in existence.

My point: the early indicator of problems, 30-to-60-day past due loans, have been declining. With roll-rates stable to improving, this will first lead to a decline in 60-plus-day past-due loan dollars, and, finally, to a decline in the widely watched 60-plus-day delinquent loans as a percentage of current balances. As this plays out, observers who’ve lately raised their estimates of cumulative losses to the moon (rating agencies, this means you) will have to bring them down.


Bank loan loss reserves as a percentage of nonperforming assets are at their lowest level since 1993, and nonperforming assets are rising rapidly.

It makes no sense to measure reserve adequacy by comparing a bank’s loss reserve to its nonperforming assets, or to criticize the bank when this ratio declines. There are several reasons why. First, the loan loss reserve is built mechanically to cover future losses to be generated by all the bank’s loans, both performing and nonperforming. The principal driver in the loss reserve calculation is a rolling average of historical loss rates, by loan type. Unsecured consumer loans, to name one, never show up in the nonperforming-asset category, but are an important driver of the reserve calculation, nonetheless. The loan loss reserve does not exist to solely cover future losses from nonperforming loans! Second, nonperforming loans and assets are typically secured. Over the past 30 years the banking industry’s historical loss rate from nonperforming loans has been less than 20%!

These facts alone should suggest that, as nonperforming loans rise, the reserve-to-nonperforming-loan ratio can, should, and does decline.

Third, a bank takes a partial writedown on a loan the moment it deems the principal balance to be impaired. Take, for instance, a $10 million loan secured by inventory originally valued at $11 million. If this loan becomes nonperforming and the bank estimates it will recover just $8 million from inventory’s the eventual sale, it will write down (charge off) by $2 million and categorize the other $8 million as nonperforming until the collateral is ultimately sold. It would be crazy to expect, and GAAP accounting prevents, a bank to keep an $8 million reserve for the balance.

Finally, the nonperforming assets of a bank consist of its nonperforming loans as well as its foreclosed assets. The auditors for the vast majority of banks do not allow future losses on assets owned through foreclosure to be absorbed through the loan loss reserve. It’s not hard to understand why. The loan loss reserve is just that: a reserve to cover losses from loans. Foreclosed assets are not loans.


All subprime mortgage loans that are more than 60 days past due will be written off.

No, they won’t be. It’s of course difficult for a subprime borrower to recover once he’s missed two monthly mortgage payments, but that does happen. What’s more, lately it’s happening more and more. In our most recent piece we show that currently 6% to 8% of loans more than 90 days past due become “cured,” and that 8% of the loans in foreclosure cure as well. These cure rates are well below historic levels, they are meaningful and rising, nonetheless.

Further, the federal government’s effort to help delinquent borrowers get back on track seems to be having some success, as well. In particular, the government-facilitated Hope Now alliance between subprime mortgage underwriters and servicers last week reported that, in a survey of roughly 603,000 subprime loans set to reset between January and April of this year, 5% have already been modified, and fully 45% were paid in full when the homeowner refinanced the loan or sold the property. Just 0.3% have entered foreclosure.

The credit performance of subprime mortgage loans underwritten in 2006 and 2007 has been awful. But people who are trying to estimate how bad things will eventually get should rely on real data, rather depend on some exaggerated “gut feels.”


It’s too early to buy the financials.

There is no telling how high the fear factor over financials will go in the short run. Still, a deep dive into the credit problems of the industry overall, as well as at individual companies, indicates to us that the problems are manageable for almost all companies and the stocks are deeply undervalued.