Comment by dxbydt | Hacker News Reader

Comment by dxbydt | Hacker News Reader

The DoJ lawsuit is over S&P's ratings model for CDOs & CLOs. Portfolio loss distributions are computed using expected loss distributions constructed with historical asset correlations, via large MC simulations. Each ratings agency uses different metrics - moody's using the GCorr ( global correlation matrix), S&P uses creditpro etc, which changes the rating transitions. There's no foolproof way to do this - its a very seat-of-the-pants applied multivariate stats problem, where "bad inputs" skew the SDR ( scenario default rate ) which then drives the ultimate cdo rating. When you have 7 tranches on average & you have considerable discretion over what your put into each tranche, you can essentially "buy" a rating.

Most of the money that went all-in was pension funds. So in a very real perverse way, the public was entirely responsible for this mess. The public bids up house prices - this skews up the asset correlations of the CDO models. The house prices keep going up, & the public parks its profits & savings in pension funds. The pension funds then buy up the very CDOs! You have a nice positive feedback loop that eventually collapses on itself.

Here's a good primer we used back in school ( http://bit.ly/14AHP5q ). start on page 13.

It's perverse to blame the public for bad financial models created by ratings agencies whose sole professional responsibility is to accurately model risk. "But the public caused an asset bubble we didn't anticipate" sounds like the Wall Street Banker version of "my dog ate my homework." Adjusting the models when they produce incorrect answers is their job.

> bad financial models created by ratings agencies

Impersonal words like "ratings agency" actually consists of a bunch of really smart math, stat & physics phds ( & the occasional finance phd who didn't get into the buyside IBs & hedgefunds :) The ratings are spit out by models. The models ar driven by simulations upon giant matrices whose input is historical correlations. When the inputs fuck up, the outputs will get hosed too.

> Adjusting the models when they produce incorrect answers

Nobody adjusted the models when they produced incorrect answers because the answers weren't incorrect. They weren't the answers you like in hindsight, but they were correct insofar as the model was vetted & the input was correct.

Lemme give you a crazy, absolute bizarre scenario. Lets say there is massive acceptance of sexting pics among US youth starting tomorrow. Youtube is flooded with penises & vaginas. Then the churches start filing lawsuits against google because youtube is actively promoting pornography. You see what I'm getting at. For youtube to be relevant, it has to cater to the youth & not actively censor content especially stuff that has mainstream acceptance among youth. At the same time, it becomes liable. Rock & hard place. This is the exact sort of crazy bizarre stuff that happened during the financial crisis. The models are predictive, not normative. House prices actually went up, year after year, & the cdo ratings simply reflected that. Then people got burnt & are deciding the models are responsible! If the models priced normatively, the investors would simply look elsewhere for different models that priced predictively.

Lemme give you a crazy, absolute bizarre scenario. Lets say there is massive acceptance of sexting pics among US youth starting tomorrow. Youtube is flooded with penises & vaginas. Then the churches start filing lawsuits against google because youtube is actively promoting pornography. You see what I'm getting at. For youtube to be relevant, it has to cater to the youth & not actively censor content especially stuff that has mainstream acceptance among youth

Yeah, that's pretty crazy and actually self-serving and disingenuous. If ratings labelled themselves "free-expression platforms for Math PhD's who don't guarantee nothin' but garbage-in, garbage-out" well then they'd have no obligation beyond that of youtube. But they signed up, actually put pen-to-paper and got money, for the responsibility of being "the adult in the room". You seem to think that their failure to actually be that is just a terribly unfortunate coincidence that they should have no responsibility for, despite their claiming that it earlier.

> Nobody adjusted the models when they produced incorrect answers because the answers weren't incorrect. They weren't the answers you like in hindsight, but they were correct insofar as the model was vetted & the input was correct.

I can't believe you are seriously arguing this. According to you, ratings agencies have no responsibility for the accuracy of their models as long as they fit historical data. By this logic, it's "correct" to rate Greek bonds at AAA right now because they have never defaulted before (this is a simple and completely accurate model that fits past data). Also see: http://xkcd.com/605/

Ratings agencies aren't paid to make models, they are paid to make predictions. When those predictions are wrong, it's on them. It's not reasonable to give a AAA rating that is based on an assumption that housing prices will rise at 10%/year forever, just because they have for the last ten years.

Rating agencies aren't paid to be a cultural zeitgeist that analyzes and summarizes the way financial markets are right now. They aren't paid to "go with the flow" and accept assumptions uncritically. They are paid to predict what will happen in the future, and to make especially sure that the most confident ratings (which grannies depend on to make sure they don't lose their retirement) are only given to financial instruments that make the most conservative assumptions.

I'm just astounded that you so freely deflect blame for this onto the public. I hope that this isn't the general sentiment inside the financial industry.

That's a good overview and a great link, but I have to ask why the ratings agencies wouldn't be expected to take positive feedback loops into account? After all, this phenomenon has been discussed since at least the days of Dutch tulip mania.

We had an asset bubble with a painful pop right at the turn of the century; and while it appears the Fed and many economic pundits such as Paul Krugman advised pumping up the housing market a bit to compensate (since houses have considerable utility independent of their asset performance), housing bubbles aren't exactly unknown either. I got caught in the early 90s UK housing bubble as a callow youth and luckily had the good sense (or risk aversion) to sit the recent bubble out because there was just way too much easy cash sloshing about to believe that it could last.

I have some understanding of economics and a basic grasp of finance, so I'm no expert, but how many standard deviations away from the long-term average do you need before something triggers the 'too good to be true' warning? While 'the market can remain irrational longer than you can remain solvent,' and everyone hates a party pooper, I feel that when ratings agencies say they are helpless in the face of public irrationality, it's like inviting consumers to buy a car with no brakes.

When you have 7 tranches on average & you have considerable discretion over what your put into each tranche, you can essentially "buy" a rating.

Yes, I believe that was what was happening and it wasn't something the rating agencies became aware of after that the fact.

There's agency which is tasked with testing the goodness of milk. Let's the some milk suppliers have figured out a way of doctoring the milk so that it appeared to be good while actually being bad or even poisonous. Suppose the agency was aware of this doctoring and instead did nothing.

If that agency happened to promise the goodness of said milk, would that agency be liable, despite the milk "honestly" passing their tests?

http://en.wikipedia.org/wiki/2008_Chinese_milk_scandal

Should we feel like because the figures of various sorts used to rate CDOs etc were "poisoned" in 2005-2008 and the rating agencies were "unlucky" enough to make a lot of money knowing certifying a lot of BS, that these agencies shouldn't be punished for their good, "honest" work?

"Don't hate the player, hate the game and keep your hands off my ill-gotten gains..."

Reply on news.ycombinator.com