Saturday, September 27, 2008

Reverse Auction Questions

The design of any reverse auction to be used by Treasury is not at all clear. One big issue is how to define the range of mortgage-backed securities (MBS) to be included in any auction (I am assuming that they will be buying MBS). The complexity of the MBS market is rather overwhelming: In a typical $1 billion mortgage pool, there would be upwards of 15 different tranches of MBS created, from the top A level securities through mezzanine to even lower priority class securities. Suppose there was $2 trillion of subprime and alt-A mortgages securitized in this way. That would mean 2,000 different pools/trusts created and about 30,000 different classes of MBS securities!!

So obviously you cannot just run an auction where you say, submit offers to sell any MBS. You have to define the class of MBS you will buy more narrowly. Of course, if you go too narrow -- for example, just the M-2 class from the GSAMP Trust 2006-NC2 pool -- you may end up with too few holders and therefore too few potential sellers.

I suspect the Treasury will have to define acceptable securities for any auction by things like original offer date, original degree of credit subordination, current and/or original rating, plus all kinds of characteristics on the underlying mortgage pool. I think one can define classes narrowly enough so that the MBS are similar enough to think that one price can be efficiently put on all of them, while defining the class broadly enough so that the auction has effective competition.

Once the Treasury buys the MBS, what do they do with them? Like a used car dealer, I suspect they will want to "fix them up" so they can be sold at a higher price. What might that mean? Lots of things, including possibly re-pooling the MBS, creating what were called collateralized mortgage obligations. Put all the MBS from an auction into a new pool, and define new securities with claims to that pool of cash flows. I would love to see some simplification there: don't define many new classes, keep it simple and just create one or a couple different CDO tranches.

Then they have to structure an auction to re-sell the new securities. That will be the easy part.


agnes said...
This comment has been removed by a blog administrator.
Anonymous said...

Former student here:

There are more issues than the ones you just mentioned. Essentially, for RMBS, there are only so many characteristics that are available to the investing public through media like Bloomberg. The data includes pool statistics like LTV, FICO scores, geographic concentration of the pool, % of secondary residences in the pool, % of a tranche that is 30, 60, 90 overdue (and there is already a 30-60 day delay when this info gets to Bloomberg, so those delinquency numbers can change significantly. If you are a party to the PSA, you can get further info like prepayment speeds (also with a lag). The issue really is the timeliness and accuracy of the data. It's often not right, and a few months old. Hard to price a black box.

The biggest problem for repackaging is that the loan tapes (lists of mortgages in pools) are messy and often just flat wrong. These securities were slapped together so fast that often a single loan can appear in multiple RMBS tranches, which should be impossible, unless the cash flow has distinct characteristics that allow for an easy split (fixed for a period, then floating, for instance). Hard to divvy up a single cashflow twice...but it happens. A lot.

These auctions can/will be a disaster. Look for any place that runs a loss mitigation arm to do well, like Clayton Holdings. There will be plenty of cleanup jobs created in the mortgage industry if the fine print actually needs to be sorted properly. Chances are that it will not be sorted properly, and all sorts of servicing horror stories will ensue, not to mention bonds that aren't receiving proper cash flows.

CMOs are just like CDOs--extra slicing and dicing to make chicken $hit into chicken salad. It won't work.

One of my favorite auction theory economists is Paul Milgrom. I am sure he will opine on this. He's a well-tapped expert at govt autcions.

I enjoy your blog, even if you do come off as a UP/UV right-wing nutjob sometimes.

I am eagerly awaiting the house passing the bill, as I watch my dollar-cost-averaged, index-invested 401(k) melt away (on paper). Ouch.

Robert G. Hansen said...

Anonymous provided some good information on the information problems with MBS. I was looking at the data on Bloomberg for one mortgage pool, and was impressed at the breadth of information that seemed to be available. Knowledge of prepayments is critical, of course, and it sounds like that is available only to some parties. Also the errors in the data and lags are scary and contribute to information asymmetries. All of which makes the old adverse selection/winners curse problem real and severe.

There will be horror stories, for sure. But those seem to be more a function of the shoddy workmanship up front than from the mistakes that will surely be made in trying to sort it all out.

Harry Schell said...

Maybe this is a journey best not taken...

Use FDIC to inject cash and match the solvent with insolvent banks.

Use the bank "workout departments to deal with defaults and related properties.

No new rules or complex government agency to spool up and staff. And then find a new lease on life when its original mission is complete.

Congress will find oversight harder, but this is good, so far as I can see.

And I have been broke, but survived, so I know a little about being backed up to the wall.