Arnold Kling  

What is a High Mortgage Default Rate?

Swagel on Mortgage Finance... Budget Arithmetic...

Some further points.

1. This article says

OFHEO has identified as a benchmark the 30- year, fixed-rate mortgages originated in Arkansas, Louisiana, Mississippi and Oklahoma in 1983 and 1984, just before oil prices collapsed. During the first five to six years into the lives of these loans, house prices in the region fell about 16 percent. As a result, the 1983 and 1984 vintage mortgages purchased from that region by the two companies experienced high levels of failure. OFHEO's research shows that these mortgages defaulted at an average 14.9 percent in their first 10 years, leading to dollar losses equaling 63.3 percent of the original principal balance of the loans

Note the phrase 1983 and 1984 vintage mortgages. Today, those would be the equivalent of mortgages originated in 2006 and 2007. Older mortgages would tend to default at a much lower rate, because the underlying houses experienced appreciation before prices began falling.

2. David Min's chart shows the rate of serious delinquency for mortgages as of Q2 2010. For "prime" mortgages, this is 7 percent.

3. As far as I can tell, Min's data includes all vintages of mortgages, not just the 2006 and 2007 vintages that should have been most vulnerable to house price declines. For example, a 2001 mortgage should have experienced significant underlying appreciation on the home, so that the borrower's equity would still be positive today.

4. As far as I can tell, Min's data is a snapshot of mortgages that were delinquent as of Q2 2010. So that if a loan went into foreclosure in 2008 or 2009, it would not count in his data. What we really want is the cumulative default rate on mortgage loans that were made in the boom years. If I am right that the 7 percent delinquency rate is only a snapshot, and not a cumulative number, then Brad DeLong's arithmetic is beside the point, or to use his word, silly.

5. The answer to the question I posed to Brad is that the guarantee fee is about 20 basis points. The agencies need to keep losses below 20 basis points in order to break even.

6. The present value of 20 basis points of guarantee fee is obtained by multiplying 20 by the DV01, which means dollar value of one basis point. Twenty years ago, the DV01 for a 30-year mortgage was between 4 and 5. Assuming it has not changed since the days when I was doing default cost analysis, the present value of the guarantee fee is about 100 basis points, or one percent.

7. If you lose 50 percent of the loan amount in a default (note the OFHEO estimate of 63 percent), that means that you cannot afford more than a 2 percent default rate. Otherwise, you lose more than 1 percent in present value, which exceeds the DV01 of the guarantee fee.

8. Even with a default rate higher than 2 percent, Freddie and Fannie could have survived. They were making profits on their portfolio holdings (the difference between the interest rates on mortgages and the rates on agency debt), and they held capital, which by itself was supposed to enable them to survive a severe downturn in house prices. But Freddie and Fannie blew through their profits on their portfolios as well as their capital.

9. It would be interesting to try to simulate Freddie and Fannie's performance assuming the same path of house prices but a mix of loans purchased in 2004-2007 that looks more like the loans purchased in 1994-1997 in terms of loan-to-value ratio, loan purpose, and credit score. I bet that Freddie and Fannie would have come through the housing crash just fine under such a simulation.

10. It would be interesting to try to simulate what the housing market as a whole would have done without Freddie and Fannie's support for high-risk lending. My bet is that the boom and the bust would have been a good deal milder.

Comments and Sharing

COMMENTS (7 to date)
Lord writes:

All is vanity. Allow unqualified lending and no lending is safe, no margin can prevent losses, no limit can be placed on defaults. Allow unqualified lending and all such attempts only lead to more unqualified lending, larger booms and crashes, and more severe crises. Safe lending cannot be made safer, only unsafe lending can be prevented making safe lending safe. While data can be collected on safe and unsafe lending, they are not independent.

mark writes:

"As far as I can tell, Min's data includes all vintages of mortgages, not just the 2006 and 2007 vintages that should have been most vulnerable to house price declines."

The GSE's break this out. For example, as of Q210, Freddie Mac's serious delinquency rates for single family guaranteed mortgages for the 06 and 07 vintages were, respectively, 9.9% and 11.1%. (Slide 17)

On the previous slide, it shows that nonperformng single family loans (all vintages) were 6.3% of total single family loans, and produced an 83 bp loss in Q22010.

mark writes:

Slide 20 of the link in my comment above gives you the salient credit statistics of each vintage. 2007 just jumps off the page - Alt A loans, 7% of the total portfolio but 31% of that vintage; FICO less than 620, 4% of the total portfolio but 27% of the 07 vintage; IO loans 6% of the total portfolio but 41% of 07's loans.


MARK writes:

You can compare the data above about 07 with the portfolio as at year end 06, summarized by S&P in a 12/29/06 credit rating:

"The remainder of the portfolio includes adjustable-rate mortgages (ARMs), 7%; multifamily loans, 4%; interest-only mortgages, 4%; option-ARMs, 2%; balloon mortgages, 2%; and other loans, 1%."

"Credit losses have been virtually zero in the multifamily portfolios. Delinquencies in the single-family portfolio continue to be very
low. At Sept. 30, 2006, the noncredit enhanced single-family portfolio had a delinquency rate of only 22 basis points (bps). At July 31, 2006, the total delinquency rate for the single-family portfolio was 52 bps, and management's estimates of net charge-offs were only $95 million through August, or only 1 bp."

"The mortgage portfolio also has a high FICO average score with 88% of the portfolio having a FICO score of at least 660."

Further they quantify the guarantee fee at an average of 16bps.

At that date, Freddie's debt / equity ratio was 96.63 / 3.37.

And the next year, they embarked on a credit risk binge sufficient to wipe that equity out, all by its lonesome.

Patrick R. Sullivan writes:

Thanks for that info, Mark. Game, set, match.

mark writes:

The figures on slide 20 may reflect the portion of the 07 vintage that remained in the portfolio as of Q2 2010 and not apply to the entire vintage. Good loans in that vintage may have been refi'd already, reducing the denominator for the vintage. Can't tell from the slide. I am confident however that, in the 07 vintage, each of their (a) FICO less than 620, (b) Alt A and (c) IO tranches was larger than their equity capital as at 1/1/07. If memory serves, they guaranteed or bought something like 40B of subprime in the second half of 07 and early 08 - with equity capital of 37B or so.

David Min writes:

Dr. Kling, I think you are addressing a fundamentally different issue than the one I was addressing. You are saying that the GSEs should not have taken on as much risk as they did. I don't think anyone disputes that.

The particular question I am addressing is whether Pinto's equation of loans that have a 7-10% serious delinquency rate with loans that have a 30% serious delinquency rate is justified. As I have pointed out, you can't get to the "GSEs caused the crisis" conclusion without that critical (and fundamentally flawed) step. Their subprime/Alt-A purchases were an example of "too little too late" and moreover, their Alt-A purchases made it harder for them to meet their affordable housing goals. In this context, it might have been useful for you to note that PLS had a 45% and 42% serious delinquency rate (a rate that includes BOTH 90 day delinquencies AND foreclosures, in response to your point 4) for years 2006 and 2007, as compared to the 9-11% for Freddie Mac.

So if you want to say that Fannie and Freddie need major reforms because they took on an unacceptably high level of risk, particularly for taxpayer backed entities, welcome to the club. I don't think there's anyone who will disagree with you, certainly not anyone I've met who's studied the data even a little bit. On the other hand, if you want to argue that Pinto's "high risk" loan categories (which you are apparently agreeing are comparable to prime conforming loans) are equivalent to PLS subprime, or even PLS generally, you are completely ignoring all the data.

Comments for this entry have been closed
Return to top