Observers looking for an early sign of a turn in credit quality in the subprime mortgage market have begun to obsess over two numbers in particular: the value of the ABX subprime mortgage indices and the past-60-day delinquency plus foreclosure rate on the mortgage-backed securities that underlie the ABX. Very sensible. There’s just one problem: improvements in both numbers are apt to give a lagging rather than a leading indication of the timing and magnitude of the peak in subprime losses. If you’re waiting for them to give you a signal that it’s time to buy, you’ll end up waiting way too long.
First, about the ABX. It’s become pretty clear that there’s a persistent price disconnect--sometimes a very wide one—between the value of the ABX and the underlying securities the index is supposed to represent. Many believe (yours truly included) that the disconnect has come about as a result of the high level of speculative trading activity surrounding the indices themselves. The heavy speculative trading has overwhelmed the ability of arbitrageurs to reconcile the indices’ estimated value against their underlying securities (which are highly illiquid, in any event). Over time, of course, these disconnects should resolve themselves. Until then, the huge volatility in the ABX should persist—which means it may not be the great indicator of changes in credit quality that people seem to think it is.
Just as misleading, in my view, is mortgage-watchers’ other favorite number: the past-60-day delinquency plus foreclosure rate of ABX pools. Thanks to the mechanics of how the rate is calculated, it will turn down, I expect, long after underlying credit problems peak.
Take a look at Charts 1 and 2, below, and you’ll get the picture. They show the over-60-day delinquency plus foreclosure rates for the ABX 06-1 and 06-2 indices. Both are obviously rising fast, and are already at high levels. By now, just about 30% of the remaining outstanding balance of the 20 loan pools that make up the ABX 06-1 are either more than 60 days delinquent or are in foreclosure. That’s nearly double what it was at mid-year. No wonder, you say, the rating agencies can’t keep up with changes to their cumulative lifetime loss estimates on these securities.
But hold on. The delinquency rate ought to be a lagging indicator of credit quality, not a leading one. Why? Arithmetic. The way the calculation works, the denominator is constantly falling as loan principal is paid down and loans are refinanced. But the delinquencies in the numerator stick around until the foreclosure process is completed. In most states, that can take a year or more. Result? Delinquency rates can seem to worsen as the overall portfolio shrinks, even though the amount of delinquent loans in dollar terms may have stopped rising.
So the widely published past-60-day delinquency plus foreclosure rate can be misleading.
I believe the key to predicting the cumulative lifetime loss of a portfolio is to look at 1) the roll-rates of problem loans from early-stage delinquency, to later-stage, to foreclosure, and 2) the rate of in-flows, measured in dollars, of newly delinquent loans. Which is to say, you want to watch for how much formerly good stuff is turning bad, and how much of the bad stuff is curing. (Severity assumptions, which depend on the ultimate sale price of the foreclosed property, are also important. But at this point, using 45% to 50%, they’re not likely to change significantly.)
Go back to Charts 1 and 2. Remember, the over-60-day delinquency and foreclosure rate is calculated using a denominator that declines each month as principal is paid down and loans are paid off (it still does happen!). Now look at Charts 3 and 4, below. They show the absolute dollars of loans 30 to 89 days delinquent in the 20 loan pools that make up each ABX index. It’s a materially different story: since October, the dollar value of loans in the early-stage delinquency buckets in both the 06-1 and the 06-2 have been relatively stable.
In fact, 19 of the 20 loan pools that make up the 06-1 index showed lower delinquencies (measured in dollars) in January than they did in December. Of these 19 pools, five saw delinquencies peak in October, four in November, and nine in December. In the less-seasoned 06-2 index, seven of the 20 loan pools showed lower dollar delinquencies in January than in December.
Does that mean that credit quality has bottomed? Not necessarily. There may be some seasonal factors at work (though I doubt it). Then again, the dollarized numbers tell a completely different story that than does the rising delinquency rate does.
Let’s take a step back and look at what happened in 2007. The inflow of new delinquent loans grew tremendously throughout the year, roll rates worsened, and severity of ultimate loss after foreclosure increased. All three factors led to significant increases in estimated lifetime losses in each individual loan pool and, collectively, for each index.
But with the 2006 loans, things are changing. The increase in new delinquent loans has slowed, as has the deterioration in roll rates. (Severity is still increasing, though.) As I analyze our loss forecast by loan pool and compare them to the loss forecasts published by the sell-side and the rating agencies, I come away with lower cumulative lifetime loss assumptions. Why? I don’t see the performing part of the remaining outstanding balances generating the magnitude of losses that the rating agencies and the investment banks assume in their “stress case.”
In about two weeks, the next set of monthly servicer reports on the portfolios will come out, so we’ll have more data to analyze. Given the long delay between initial delinquency and ultimate foreclosure, you can count on a further rise in the past-60-day delinquency rate. If you want a better sense of what’s happening at the margin, though, I suggest you focus on the dollar amounts of delinquent loans in the early-delinquent bucket (that is, 30-to-60 days) and any changes in roll rates.
Back in 1990, my then-partner Frank DeSantis and I encouraged banks to disclose the quarterly inflow of new non-accrual loans, not just total nonperforming assets. This gave us confidence to predict the subsequent peak in total nonperforming assets, losses, and the trough in earnings. Leading indicators will obviously be the key in this credit cycle as well. If you’re just looking at the delinquency rate, you’re looking in the wrong place.