http://money.cnn.com/galleries/2007/fortune/0708/gallery.crisiscounsel.fortune/index.html?cnn=yes Article about how institutions are marking securities to to 'myth not market'.
While Buffet in my mind is one of the original trader-quants, I don't think he understands the way these models work. It is perfectly reasonable to mark to model, as long as the model is hedging at at least implied vol, or you keep the risk in your book balanced with other similar instruments, or both. The mathematics of this is well known. That is not to say there is no risk. But if you hedge every risk away, you can't make money. So you have to be exposed to some risk. The fact that those risks may come to fruition was probably well understood at inception, and is not a flaw in mark to model, but in the way the book was traded or a decision to what risks they chose to be exposed to. People love to blame models and quants for everything. 99% of the time the blame should be placed on misusing the model, or not understanding its weaknesses. You have to understand your tool to use it correctly. It's like blaming a gun or a car for killing a person. Cars and guns are driven by people. You still have to aim a car, apply brakes etc. If you are driving a sports car with those super slick tires, don't drive aggressively in the rain. A gun doesn't kill people. You have to point a gun and shoot. nitro
It's one thing to "mark to model". It's quite another for the underwriters of that garbage to be bidding that price in the open market, especially during a decline.
the problem is always the overshooting. you can trust models too much and blindly buy into cdo copula stuff, or can throw it out of the window as a whole. both is narrow minded. just think about it: is the alternative truly nonQuant at all? guessing the value of a cdo portfolio, assuming this is the "right" value? well if you buy into that you also assume the market was right in recent years when it underestimated the risk in cdos and alikes. at that time our models stated clearly: this is too optimistic. now the models say: at some spots in the current scenario the market is too pessimistic. quite reasaonable if you ask me. the true thing is that the correlation desks started with brand new people with little or no investment background. i guess the avg corr trader is below 35 and has not seen a credit crisis within his professional experience, not to speak within structured credit which got its big bang around 2000. so you have people who earned tons and tons in recent years and sure they trusted all the models that made their bonuses. and these "models" are not all the same. in 2005 if i recall correctly SP rated literally all cdo squares because their outer correlation of 0 showed the least risk, thus the highest rating for such very corr sensitive transactions. we modelled such deal and what was a brilliant AAA for SP was a A at best in our way of thinking ... with positive outer correlation. what i am trying to say is: take warren's statement with a good grain of salt ...
the thing is that with cdos and basel the power of the agencies is increasing dramatically. too dramatically. and they were not prepared. and i think they just start to be.
Good post. On a different note, I find it especially ironic to read articles written by journalists with absolutely no idea what they are writing about (but glowing with a lot of schadenfreude) when the are ripping the quants and proclaiming the "end of computer driven modelling". And it becomes a farce when the same journalists can't keep different quant funds apart although they may be run by the same institution, e.g. Renaissance Medaillon (high frequency market making, and even that is pure speculation) and Renaissance International Equities Fund (global quant equity 130/30).
Can you elaborate on that a little bit more? My understanding is that simply there is a divergence between what their models are valuing the securities at and what the market would bid at. Now, you are saying that they are incorrectly using the models to price their securities. Buffett is saying that no models would be able to accurately value these complex derivatives in current market conditions and only a mark-to-market valuation will suffice. So you are saying that their models can hedge the risk, but the issue is that these firms are exposed to risk and no model can accurately determine the extent of that exposure. Would appreciate your input, Nitro.
I loved that line by Buffet: "I wish I could mark my assents to models and not market". I guess if all these CDOs are marked to market, we will see a blood bath like never before. Let's hope it does not come to it. I learned something new about CDOs, I had no clue they are amrked to models, it is a fatasy world after all. redduke
in my opinion, the question mark to market vs mark to model is important when market and or model is "incorrect". in a orderly market, with valid models, marking to either should be fine. however, when one or both act incorrectly, prudent thing to do is to mark to the error that is the market, rather than the model because at the end of the day the cash balances that you have come from the market and not from the model. any model is only as good as the data that goes in. with most models using abstract ratings from agencies as inputs, there is a real possibility of divergence from reality. i do not think Buffet is saying one should not use models!!. Of course we should use models to identify and exploit market inefficiencies. It is just that self assessments are best done in a prudent manner.