Friday, May 30, 2008

Bloomberg's Weird Numbers

by Tanta on 5/30/2008 07:18:00 AM

Forgive me for once again falling into despair over the media's inability to report sensibly and critically on foreclosure and delinquency numbers. I should be immune by now. If you are wiser than I, just skip to the next post. If you still cradle to your wounded heart the battered but indomitable belief that even media outlets like Bloomberg can learn to spot the flaws in a reported statistic, and that there is a point to doing this, click the link below.

***********

The headline: "New Overdue Home Loans Swamp Effort to Fix Mortgages in Default." We will take this as a promise that the article is going to demonstrate something about the relationship between newly delinquent loans and workout efforts.

The lede:

May 30 (Bloomberg) -- Newly delinquent mortgage borrowers outnumbered people who caught up on their overdue payments by two to one last month, a sign that nationwide efforts to help homeowners avoid default may be failing.

In April, 73,880 homeowners with privately insured mortgages fell more than 60 days late on payments, compared with 39,584 who got back on track, a report today from the Washington-based Mortgage Insurance Companies of America said.
The last of eighteen paragraphs:
Last month's 54 percent "cure ratio" among defaulted mortgages compares with 80 percent a year earlier and 87 percent in March. The comparison may not be valid because one lender changed the way it calculated defaults and cures reported to the insurers.
So we start with an eye-popping number, and then only at the very end do we note that this number may mean much less than meets the eye. This is, in fact, what MICA said in its data release:
WASHINGTON, D.C. May 30, 2008 – Mortgage Insurance Companies of America (MICA) today released its monthly statistical report for April which includes a one-time adjustment to the number of defaults and cures and also notes an 11.7% increase in new insurance written year-over-year.

As a result of a major lender’s change to its methodology for recording delinquencies, and to how it reports them to MICA’s members, there was a sharp increase, to 73,880, in reported defaults in April. The increase includes both newly reported defaults for the month, as well as previously unreported defaults by this lender.

MICA’s members reported 39,584 cures in April. This statistic also reflects the above noted change in reporting defaults.
I assumed when I read this that somebody--a large somebody, since it significantly impacts the data--switched over from the OTS method to the MBA method of delinquency reporting. I do not know if this is the case or not. Before I published this article, however, I might have called MICA for a comment. In any case I might have been more cautious with headlining a number that is described as a "one-time adjustment" to the data collection. Burying that in the last paragraph is . . . disingenuous.

I'm also a touch troubled by the statement that "Last month's 54 percent 'cure ratio' among defaulted mortgages compares with 80 percent a year earlier and 87 percent in March." That is literally true. However, the cure rate in December of 2007 was 54.1% and in January of 2008 was 51.4%. Could there be some seasonality in these numbers? Another confounding factor besides new delinquencies?

So what about the second half of the claim?
"Modifications are not occurring nearly at the numbers necessary to stem the foreclosure crisis," Allen Fishbein, housing director for the Consumer Federation of America in Washington, said in a May 19 interview. "People are still going into foreclosure when, with a writedown on existing principal, they could still stay in their homes."

In the first two months of 2008, lenders modified loans for 114,000 borrowers while starting 346,000 foreclosures, according to a study by the Durham, North Carolina-based Center for Responsible Lending. In April, 22 percent of the homes in the foreclosure process had been taken over by lending banks; a year earlier, that figure was 15 percent, according to Irvine, California-based data provider RealtyTrac.
Did you assume, when you read that second paragraph, that the 114,000 modifications were exclusive of (not the same loans as) the 346,000 foreclosure starts? It seems you were supposed to assume that. But is is true? "Foreclosure start" simply means that a legally-required preliminary filing (a Notice of Default, Notice of Intent, or Lis Pendens, depending on the state and the type (judicial or non-judicial) of foreclosure) has been made. That is a "start" because in most jurisdictions it will be another 90 to 180 days, or even more in some states, until the auction can be scheduled, the home sold, and the foreclosure "completed." My own view is that the "best practice" is to work hard to negotiate a modification, if possible, in the early days of delinquency before starting the foreclosure process. However, that is not always possible, and it is also "best practice" to continue to attempt reasonable workouts during the foreclosure process all the way up the day before sale, if necessary. There are certainly cases in which a borrower simply cannot be brought to talk to the servicer until the initial FC filing galvanizes him into it. All of this means that it is impossible to look simply at modifications completed in a period compared to foreclosures started in a period and conclude that the starts will never get a mod or that the mods were not effected after the FC start.

Besides that, where is the data to back up the idea that a 30% ratio of modifications to foreclosure starts is poor performance? I am personally not sure that much more than 30% of recent vintage loans can be saved. Back out fraud, flippers and speculators, and borrowers whose loan balances would have to be reduced by half in order to get a workable payment--which would most likely exceed the cost to the investor of a foreclosure--and 30% doesn't sound so shabby.

As far as the second claim--the increase from 15% to 22% of homes in foreclosure "taken over by lending banks," I'm prepared to read that literally. There is no jurisdiction in which a foreclosed home must be purchased by the lender at the foreclosure sale; all jurisdictions require public auctions in which third parties can bid. An increase in REO (lenders "winning" the auction) does not necessarily mean an increase in completed foreclosures; it can mean that fewer third parties care to bid on foreclosed homes. All the data I have seen recently suggests that this is the case: buyers are still wary of further price declines, and lenders are still bidding higher than potential RE investors. One therefore expects the FC-to-REO numbers to increase. But they can do that even in the absence of an increase in total foreclosures. In order for this statistic to mean much, we have to know how much of the increase is due to more foreclosures, and how much due to fewer third-party bidders.

So put these dubious statistics together--the rest of the Bloomberg article is basically filler--and you get anomalous data on new delinquencies, ambiguous data on modification-to-foreclosure-starts, and a claim about REO rates substituting for a claim about foreclosure completion rates. How about taking back that headline, Bloomberg?

You know, last year I might have had some more sympathy for these reporters. We were just newly into the whole problem and a lot of concepts--delinquency reporting methodology, foreclosure processes, various ways of reporting "cures" and "starts"--were all new to everybody except industry insiders and a handful of totally Nerdly blog readers. But surely by now we can have moved the ball forward a couple of yards? I am here to affirm that if you have been reporting on "the foreclosure crisis" for a year or more and you still can't ask basic questions about the press releases you read, you aren't doing your job.

I, Too, Need To Know Why These Numbers Are So Squirrelly