Sunday, May 2, 2010

Reg AB Proposed Changes

Last month the SEC released a 667 page proposal that I've avoided reading because of the length and the proverbial stack of other reading material on my desk. Its come up in a number of conversations (mostly with attorneys) since, so I started today... some bullets below.

Off the Shelf Registration is no longer based on ratings from NRSROs, and can be eligible with

  1. a signature from the depositor's CEO stating that the assets in the pool have "characteristics that provide a reasonable basis to believe that they will produce.... cash flows to service any payments due and payable on the securities described in the prospectus."
  2. Risk retention by the sponsor of "each tranche" of the securitization - so, are they proposing the sponsor holds 5% of each tranche here? I don't think that works. In all fairness, I don't think retention works no matter how it is applied (sponsor simply values close to zero, passes costs along to investor, and there is nothing but upside). This really is ineffective though - the sponsor gladly takes on the 5% retention at the fattest part of the stack at the top, again values the lower retention pieces at zero, passes along costs to investor. The top of the stack goes on their balance sheet at par, probably stays there for the life of the loan paying out coupons that more than make up for the lower retention pieces.
    In the above example, based on a Subprime deal, the sponsor retains 5% of each tranche. A sponsor is going to look at this and say, well, the only classes really ever going to take a loss are M1 and down, so we're *really* only retaining 1.05%, the other 4.95% is top-of-the-stack investment quality tranches that we'd love to own anyway. They would be wrong - the highest rated class (3A1) is currently rated BBB- by S&P, nothing else is better than CCC or D, but nonetheless, it isn't going to have a mind altering change on how a sponsor treats the deal quality.
  3. Reps & Warranties reviewed periodically by a third party, funded by the obligated party

Other changes
  1. Extended filing deadlines to "allow investors with additional time to analyze transaction-specific information" to promote independent analysis of ABS by investors rather than reliance on credit ratings (if investors are still relying on credit ratings after what we've just been through, then they deserve to lose their money). Oh, and it's like 5 days...
  2. No more prosups, preliminary prospectus and final prospectus for each transaction
  3. Asset/loan level data via Python
  4. Disclosure of any reps/warranty or repurchase/replace breaches by the originators or sponsor in the past 3 years
  5. my favorite "additional information regarding originators and sponsors"
  6. Rule 144A and RegD may have contributed to the lack of access and sufficient time to evaluate investments - 144A & RegD are the definition of QIBs and AIs - if they felt like they didn't fully understand or didn't have enough time to analyze, THEY SHOULD NOT HAVE INVESTED IN THE FIRST PLACE. I have very little sympathy for this "poor me" investor tactic.
  7. 430D - substantially all information omitted from the preliminary prospectus would have to be filed in a final at least 5 days before closing.
Too long, stopped reading. I'll pick it back up though. Some of you will recall the idiotic requirements that made it through in the first Reg AB rounds, so I'm sure there will be some nuggets this time around as well.

Also worth noting that Cryptome had a copy of the rules too, dated as of 5/3 (tomorrow). I thought there might be something there, but didn't see anything different in their version.

1 comment:

crabsofsteel said...

Revisions look pretty dumb to me. Let's take a pretty bad example: a CDO backed by bottom level subprime mezz bonds with a 5 year option to auction the collateral. Nothing in the revisions prevents stuff like this. Not only that, instead of having to retain the 15% if non-AAA I couldn't sell, now I only have to retain 5%, so it makes things worse! Also, any deal structure which requires Python and XML to be properly described probably shouldn't be allowed to exist.