Or why simple models survive.
The availability of conceptual equipment can matter even if the theory underpinning the equipment is not understood -software systems allow traders with only a rough grasp of the theory of options or of CDOs to calculate implied volatilities or base correlations- or not believed. Those who do understand the models that are used in such calculations frequently view them as oversimplifications. I have, for example, yet to interview a credit-derivatives trader who regards as adequate the ‘single-factor Gaussian copula’ model normally used in credit correlation calculations. Nevertheless, the simple models remain in wide use. More complex models face formidable barriers as communicative tools, because for full communication both parties must be using the same model, and that is seldom the case once one moves beyond simple models. Furthermore, the simple models typically have just one free parameter -’implied volatility’, for example- with the other parameters being either fixed by market convention (CDO pricing, for example, was often done assuming a recovery rate after default of 40 per cent, whatever the corpoaration that has issued the debt in question) or regarded as empirically observed facts. When numbers of free parameters are larger, or parameters do not have intuitive interpretations -as is often the case with more complex models- communication and negotiation become much harder.
In Material Markets: How Economic Agents are Constructed by Donald MacKenzie
Related:
Donald MacKenzie: End-of-the-World Trade
Felix Salmon: Recipe for Disaster: The Formula That Killed Wall Street
Paul Wilmott: Copulas and Cults
Eric Falkenstein: Don’t Blame The Quants, Felix
Was Wall Street killed by a formula?
“True or false: the CDS market is primarily a speculators’ market”
False, it’s primarily a hedging and arb market, there is some speculation for the simple reason that if you want to get rid of risk somebody else has to want it.
The market can expand if there is a simpler model or if some parameters are arbitrarily fixed by convention but that will make things worse in the long run and less informative/transparent in the short run. For ex if CDS become quoted upfront +running coupon as will be the case in the U.S. from march 20th, the quote is less informative and while it can be converted into a spread that requires “conventions” that are and are known to be false like recovery rates or flat yield curve.
er…”one-fact” s/b “one factor”
jck, I, for one, appreciate greatly your effort to make this stuff transparent.
Were you around when Goldman made their Black-Derman-Toy model freely available? That was a one-fact model, not terribly simple; and it didn’t stop the market from getting more complex.
For discussion/thought, let me pose “the problem”, perceived or otherwise, with CDS in this way:
True or false: the CDS market is primarily a speculators’ market, therefore, it does not matter what model is used, just what price was paid last.
Yes, of course.
But professions like engineering also live by simple models but the difference is that the education process emphasizes:
1) knowing the limits of the model practically and mathematically,
2) knowing how to identify when you are approaching or crossing or have crossed the limits,
3) knowing what likely consequences exist when you cross the limits, and
4) using empirical testing at many stages of the model usage to validate the assumptions that you are within the limits.
This is the rigor that makes an engineering education hell to go through but assures that planes aren’t falling out of the sky either regularly or all at once. If we were like economists every person would know at least one person killed by an iPod, a collapsing building or a glass of water.
Economics does none of these verifications with its models. Most models (especially of macroeconomic and financial “engineering” [it's a dire insult to engineers to use the term 'engineering' like this]) are essentially non-empirical and ideologically based/proven.
There is no permanent place for ugly mathematics
Milton Friedman had this to say..
http://members.shaw.ca/compilerpress1/Anno%20Friedman%20Positive.htm
I agree, models will still fail but the market structure as to be resilient enough to cope when that happens. I would tolerate OTC markets up to a certain notional size and force mandatory central clearing above that.
As far as I can see, this is an argument for centralized clearing, daily MTM, and standardized securities. While nice in theory, the bespoke model simply fails when markets go awry.