Modelling Fannie Mae and Freddie Mac – Part III

In the previous post I showed you what had caused the big losses at Fannie and Freddie to date.  In short it was mark-to-market securities (on private label securities and the like) and interest rate hedging instruments that collapsed in value as interest rates went to zero.

I also showed that these losses were not likely to continue to be a drain on Fannie or Freddie. 

What matters now is the vast (multi-trillion dollar) books of traditional business that they guarantee – mom and pop mortgages by people with good credit and no fancy mortgage terms or liars loans.  These have not caused many actual losses to date (less than $6 billion at each company) but the provisions for losses from this business are large and getting larger. 

It’s the future losses we need to worry about.  And so now – unlike in the last post – I need to make estimates about the future to see what losses will be.  Casey Stengel argued you should “never make predictions, especially about the future”.  At the risk of being exposed as a fool I am going to breach Mr Stengel’s advice. 

However I have some tools for making these predictions.  The purpose of this post is to introduce readers to the tools…

Default curves that point to the sky

Both Fannie and Freddie publish default curves by vintage.  Here is the Freddie Mac curve from the last quarterly results…

And here is the Fannie Mae curve from the last quarterly…

(click these pictures for more detail…  you can find originals in the credit supplements both companies release…)

These curves show the cumulative default of a pool of mortgages (as a percentage of the original pool balance) over time. 

Note that Freddie Mac defaults are lower in all recent vintages than Fannie Mae.  I posted on that on this blog and asked for explanations.  The correct explanation was given (and I have since checked it).  Almost the entire difference is that Fannie Mae did considerably more business with Countrywide and IndyMac – and hence has a worse book of business in aggregate.

Also note that the curves for the 2006 and 2007 years in particular are very sour at both companies.  2005 is a bad year (but probably manageable) and 2004 and prior years will cause few problems.

Default not loss curves

These are default curves not loss curves.  To turn them into loss curves you would need to know the severity by vintage and neither Fannie nor Freddie publish enough information to work that out (though you can get some reasonable estimates from the published data). 

Take again the Freddie data.  The 2000 vintage pool (meaning all the mortgages guaranteed by Freddie Mac in 2000) has had a cumulative default of about 1.07 percent.  That means that 1.07 percent of the mortgages written in that year defaulted.  Prior to the current mortgage bust 2000 was considered a bad year of business.

Nonetheless 1.07 percent default did not cause any problems for Freddie because the severity (loss given default) was less than 10 percent.  The 1.1 percent default caused about 10 basis points of loss over a decade.  Given the guarantee fees were almost 20 basis points per year this pool of business was profitable.

The problem with the 2006 and 2007 vintages is not only is the cumulative default quite large but the severity will also be very high.  Fannie Mae severity during the last quarter was over 45 percent.  High defaults multiplied by high losses given defaults means big problems.

How big – and how you might model them is the subject of the next post.