Whose definition of subprime?

There is no standard definition of subprime. I used to think of the world in terms of “Household International Subprime” and “Conseco Financial Subprime”. The guys at Household didn’t think you could ever write loans at the Conseco level profitably over a cycle (discussion in year 2000). The HI credit modellers I talked to were well aware of the risks of the business – but thought it was unlikely that there would be severe stress on it outside periods of large unemployment. (They were wrong!)

Nonetheless the year 2000 distinction between Household and Conseco could be given with FICO scores:

· Household Subprime was FICO620 to 680, and

· Conseco subprime was below 620.

I think the guys at HI were more-or-less right. The HI business might be profitable on average over long periods and the Conseco business was hopeless at the outset. [Profitable over long periods does not make it a good business.]

After that conversation the world changed and everyone started doing Conseco subprime. There were plenty of issuers who worked with FICOs of 575 and below. I was short several – not because I thought a blow-up was inevitable –just that that sort of business cannot be profitable over a cycle.

High FICO defaults – or why FICOs were misleading

FICOs have proved not to be a great indicator of default.

There are deals done with a weighted average FICO of 710 which are defaulting very badly – see the FICO in Mish’s bad deal for instance. These deals consisted almost entirely of refinances – often cash-out refinances. A person with very poor credit could show as having good credit if they always repaid their last loan. They can achieve this by sequential cash-out-refinances. (A rolling loan gathers no loss.)

Deals with average FICOs above 700 that are behaving that way usually contain a very large number of cash-out-refis.

Why I mention

One of my games is to look at the definition of subprime that people were using (particularly prior to the recent credit crisis). Lots of companies wanted to deny they did subprime loans – so all they did was define subprime to be a credit notch below where they were. Defining the loan as Alt-A doesn’t make the bad credit go away – but it made you look safe. [See IndyMac for a company that denied doing bad credit by defining good as what they did.]

Anyway here are two definitions that stand out for me. The first is from MGIC – a mortgage insurer. The second from Ambac.

Here is MGIC’s definition (2006 annual report):

A-minus and subprime credit loans are written through the bulk channel. A-minus loans have FICO scores of 575-619, as reported to MGIC at the time a commitment to insure is issued, and subprime loans have FICO scores of less than 575. [MGIC defined prime loans as having a FICO above 620.]

To me this definition was astounding. The whole of Households year 2000 business would be prime by MGIC’s definition.

Ambac drank the poison too. It drank less poison but was less aware that it was doing it. The word FICO never appears in the 2006 annual which is full of soothing words about how they had reduced their underwritings in the subprime area. By the 2007 annual report they included a FICO definition:

FICO scores range from 300 to 850. Though there are no industry standard definitions, generally FICO scores are as follows: prime (FICO score over 710), mid-prime (FICO score between 640 and 710) and sub-prime (FICO score below 640).

The entire MGIC midprime business and some prime business is subprime by Ambac’s definition.

I knew in advance that MGIC’s underwriting standards were worse than Ambac. Much much worse. I was short MGIC and had sold out of Ambac. [I always wanted to be long Ambac as much as anything because I liked the CEO. I just did not like the credit cycle and some deals they were underwriting - so I sold my position well before the top.]

MGIC will pay almost 2 billion in claims this year. They are paying less claims than they anticipated – but not for any good reason – it is just that state legislatures are passing bills to slow down foreclosure and the courts are jammed and there are delays.

Ambac pays about 20 million in claims a month (admittedly rising).

Ambac has many times the claims paying capacity of MGIC and a small fraction of the claims rate.

But MGIC is still writing business and Ambac has almost ceased.

They might both be bust – indeed I express no opinion about the insurance companies. But if MGIC remains solvent and Ambac fails then the world is a very strange place indeed. [I will explore MGIC in more detail in a future post. Suffice to say I have no position long or short and the world might indeed be strange.]