As I review some of the terms of mortgage backed securities, I am really struck by the complexity that the designers of these securities created. Too clever by half, I am afraid.
A $1 billion pool of mortgages went into a trust, and out came 20 tranches of different securiies. The top few were called A level, as they got first dibs on any payments. Then the middle 15 or so were "mezzanine" and they got second dibs on payments, but of course in order of M1, M2, M3.....
Then at the very bottom were the equity and unrated tranches who were last to be paid and first to incur losses.
But this was not the only complexity. There was a lock out period which meant that principal payments went only to the A tranches initially. There was something called overcollateralization, meaning that the face value of the MBS sold was less than the principal value of the mortgages. Overcollateralization was enhanced by capturing excess spread, the difference between the average interest rate on the mortgages vs. the average rate paid on the MBS. Then there were triggers that would be evaluated a couple years into the pool performance that would determine if this over-collateralization would be released to benefit the equity or the higher rated tranches.
Plus there was an interest rate swap on the whole pool, converting a fixed payment by the mortgage trust to a LIBOR payment from the swap counterparty.
There is simply no way that anyone can value such a complex scheme. No way. You've got default risk, prepayment risk, 20 different tranches, the triggering events, ....
But here is the real mystery. Why all this complexity? If you look at how Fannie Mae creates MBS, they throw everything into a pool and create just one class of MBS pass-through security out of it. Of course, it is easy for them, because they guarantee the payments, so everything is highly rated.
Therein lies the secret to what was going on. The creators of the subprime MBS were optimizing the tradeoff between the A rated MBS versus the lower-rated tranches. They were facing prices of MBS that were higher for high rated securities than for lower rated ones. And the relationship between price and rating was nonlinear, in that as rating went up, price went up by even more. This is what made it feasible to essentially put more risk on the lower class MBS, and less risk on the higher rated ones. If the relationship between price and rating was linear, then you cannot benefit from trading risk between the classes.
I can imagine that the underwriters would go to the rating agency and show them the deal, and the raters would say something like: "No, to get AAA you have to give the top tranches more security." Thereby came all the bells and whistles: reduce the size of the AAA tranche, increase the overcollateralization, create some trigger events that would benefit the AAA.
As I put this issue to one of my finance colleagues, he brought up what we call the Modigliani-Miller Theorem: that the value of the firm cannot be increased by moving claims to that fixed value between the debt and equity holders.
The application of that concept to MBS is simply that this trading of risk between the different tranches could not have created any value. What one tranche gained, another lost.
So why was it done, and to such an extent? Good question, and we are living with the complex consequences.
I think it was essentially taking advantage of errors in ratings and errors by the capital markets in pricing securities rated by the agencies.