Estimating losses for prime mortgage books
I am struggling here with the modelling of prime and conventional mortgage losses. It matters because I am short Magic (see post). But it also matters to Fannie Mae et al.
Anyway think about a pool of mortgages which is following historical seasoning patters. For very high quality mortgages loss would be peaking (at low levels) at about year three. Subprime mortgages which are following historic seasoning patterns have losses that peak earlier.
Anyway – everything is tracking just fine – and then – for some reason the losses start ticking higher than predicted. Now you have to project what the ultimate losses will be.
By definition here you do not know. The losses are higher than expected – and they are diverging from projection. I am going to use the fashionable baseball analogy to talk about stages of the credit crisis from the perspective of bank-uncertainty.
First base: loans are turning sour at an increasing rate. They won’t get worse for ever (nothing does) – but they will get worse for an indeterminate amount of time. You have no real basis for estimating how long they are going to get worse for. You might remember the last credit cycle and (for your business) the deterioration lasted 15 months. So you model the losses assuming a “burn out period” of fifteen months. This is highly speculative and the confidence around your estimate will be low. You might think you know what is causing the sudden deterioration of credit vis trend. But you generally don’t. Your guess will be highly speculative too – though that will not stop media and pundits from speculating. If you were going to put an honest confident interval around your estimate it would be high. You can make investments on this basis – but you are shooting in the dark.
Second base: the loans are getting worse still – but they are getting worse at a lower rate. The new entrants into the 30-59 day delinquency bucket are falling and the serious delinquency buckets are rising – albeit at a much reduced rate. At this point you begin to get some confidence about the end losses. Your estimate should get more accurate because you can now see the “burn-out” rather than just having to guess it. You might get one or two write-backs but they are rare and they look desperate.
Third base: delinquency is falling across the board. The foreclosures are humming along – the courts have caught up with the reality on the ground for foreclosures and your estimates will be pretty accurate. You can still be wrong – but it is getting obvious by now what is going on. You even have enough loan files to work out what has gone wrong. [In stage one and two nobody really knows how much is fraud and how much is ordinary credit cycle.] Write-backs of losses are regularly occurring though are less frequent than further write-downs.
Home: the economy has caught up – some buckets are actually improving. There is still a backlog of real estate owned, but the property price has stabilised. Your estimates should be pretty good now. Write backs are more frequent than write downs.
And here is where the baseball analogy breaks down. Like baseball a lot of companies will be caught out (even insolvent) between bases. But unlike baseball you are never really home in this game. The economy could have a “double dip” recession – the so called W shaped recession. Certainty is not available.
Anyway in this schema I am fairly capable of modelling the credit losses once we get to the second base – and I am totally incompetent to model the credit losses on the first base. That is just reality. We have no real idea what is going on yet.
And at the moment we are clearly in the second base (or further) with respect to everything that was marketed as subprime 2006 and prior. If it is second lien marketed 2006 and prior we are probably in the third base because we know the loss severity too. (Loss severity = 100 percent!) And I feel pretty confident modelling it. I have even purchased a few stocks on that basis (notably Ambac which I have retained despite originally planning to sell some into the short squeeze).
But with prime mortgages we are at the first base. They are getting worse in every bucket at an increasing rate. It won’t happen forever – but if someone tells you about a model of these loans you know they are just making the key assumptions up. Maybe they are making it up with an educated guess (such as my 15 month burn out period). But they are still making it up.
And this is a problem for me because I have promised you an analysis of Fannie Mae’s credit losses – and I have also promised you that (at least on this blog) I will not make it up.
And I am not sure I can keep both promises.
I am short at least one stock because I guess the better part of their book will kill them. That stock is Magic. I think their subprime estimates are reasonable (though light) … but I sure that they do not have a handle on how the good parts of their book will behave and I am guessing the assumptions are optimistic.
However as I promised you an analysis of Fannie’s credit losses and I promised you that I wouldn’t make it up I have to come to a solution. The solution: get you to make it up!
So here are a few data series just for delinquency:
The two sequences all below 1% are Fannie and Freddie’s core book respectively. They are entirely mortgages below 80LTV and below the GSE threshold. This is the very best of the mortgage market. The books with secondary mortgage insurance are given in columns 2 and 4.
It should be observed that the good end of the credit market did not deteriorate at all until August last year. This is clear across a lot of the data I look at.
In August a lot of banks were “surprised” that the subprime crisis was spreading to their books. Well – give the management credit – if they had a true prime book and they were watching the data carefully – in other words if the management were everything you would hope they were – then they were surprised. Lesson to investors: do not call for the sacking of management who were surprised in August. [Be a little less forgiving of management who chose another month to be surprised.]
But it is pretty clear that this good stuff is getting worse at an accelerating rate. The last two months in particular have been bad.
And for the reasons explained that makes it very difficult to estimate where this deterioration is going to finish.
So – as I am not happy making it up can my readers have a go. What will be the delinquency for the insured and uninsured Fannie and Freddie books in twelve months? How about the Magic book?
Guesses anyone? Email or (preferably) in the comments.
John