Rating Agencies Escape Financial Reform

Congress must not be too upset with the rating agencies for their performance leading up to the financial crisis after all. On Tuesday, the conference committee determining the fate of the financial regulation bill decided to kill an amendment created by Sen. Al Franken (D-MN) passed in the Senate. This is a huge disappointment, as it could have helped to prevent the conflict-of-interest that currently exists in the ratings industry.

With the Franken amendment gone, the only rule still alive in the bill that seeks reform the rating agencies attempts to shift more liability to them when their ratings are incorrect. This probably won't hold up in court, however, as the agency assessments are protected by the First Amendment. You would need a constitutional amendment to change that.

The Franken amendment would have required all new asset- and mortgage-backed securities to receive a rating by an assigned credit rating agency. That assignment would have been done by a group consisting of mostly investors -- the ones who care most about these ratings. Instead, the measure will be replaced with yet another study. Franken's provision passed the Senate easily, 64-35 in May. So what happened?

Obviously, the financial industry lobbyists won out. But that's not how the members of Congress tell it. Here's their explanation, which borders on hilarious, via the New York Times:

But some of the chief sponsors of the regulatory overhaul, including Representative Barney Frank, Democrat of Massachusetts and chairman of the Financial Services Committee, and Senator Christopher J. Dodd, Democrat of Connecticut and chairman of the banking committee, had opposed the idea. They warned of practical difficulties in the measure's plan to make random assignments of rating agencies.

The emphasis is mine. The article goes on to quote Dodd complaining, "It's complicated." What market is more complicated -- credit ratings or derivatives? Clearly, the latter, yet Congress' bill certainly feels free to regulate the daylights out of that industry. So what's the problem with ratings reform?

In fact, the rater assignment 'problem' Frank and Dodd claim to worry about isn't that complicated. The bill calls for a random process that takes into consideration agency performance and workload capability. In other words, rating agencies with a better track record will get more business, but smaller rating agencies shouldn't be overwhelmed, so the number of deals they get would be limited. Here's an example of how you could do it:

First, obtain loss probability matrices from all agencies based on initial ratings. For example (and I just made these up for explanatory purposes):

AAA: 0%
AA: 0.5%
A: 1%
BBB: 3%
BB: 5%
B: 10%

Then you take each asset-backed deal that the agency has rated over the past 10 years and determine a performance score based on this matrix. You could do this by using the following equation (n = # of deals rated):

Score = Summation [ Abs. Value {(Deal_1 Loss) - (Exp. Rating_1 Loss) } + . . . + n ] / n

Okay, this is a little complicated for those who don't indulge in much math, but it looks worse than it is. All you do is add up the absolute value of the difference of each bond's actual loss from its expected loss according to the initial rating, and divide that sum by the number of deals. The lower a score, the better an agency's accuracy. A perfect score would be zero.

Then, you take the mean and standard deviation of the agencies' scores. There's your baseline. Next, create a population based on some equation that uses these statistics. I used the following equation, but there's some discretion here (depending on how you want to do the weighting):

Population Units Per Agency = 100 - [(Score - Mean) / Standard Deviation]*20

Then you have a population you can take a random selection from, weighted by rating agency success rate. Finally, determine the deals per analyst ratio in a quarter. I used two-thirds, but agencies can tell what their ratio is on a case-by-case basis. That way, when an agency's plate is full, it can be taken out of the equation and the remaining deals can be taken on by remaining firms. You would update all of the inputs into this system quarterly.

Here's a chart with this an example of this data, based on a world with four rating agencies:

rating agency reform 2010-06r.PNG

For example, a deal is submitted to the new board. You use a random number generator. Out pops 217. It gets assigned to Agency F (its population bounds go from 120 to 232), assuming it hasn't already reached its capacity of 60 deals this quarter. Another deal is submitted. This time the random number is 322. That one goes to Agency S (bounds 233 to 324). Etc.

So is that complicated? A little. But it isn't too complicated. The system would have been created by investors sitting on the new board. If they can model a mortgage-backed security, then they can figure out a little system that uses performance-weighted random selection to assign a rater to each new deal. I'm no statistician, and I could do it.

Now that the relative ease of creating a system like the Franken proposal calls for has been proven, can we please have it back?

Update: Just heard from a source on the hill that there's some positive movement towards automatically adopting Franken's system after the two-year study. So perhaps not all is lost. Time will tell.