tag:blogger.com,1999:blog-5975524006824862804.post7283715115486346017..comments2024-02-10T02:23:08.475-08:00Comments on Paul's Pontifications: Why the banks collapsed, and how a paper on Haskell programming can help stop it happening next timePaul Johnsonhttp://www.blogger.com/profile/07353083601285449293noreply@blogger.comBlogger12125tag:blogger.com,1999:blog-5975524006824862804.post-64022993811652291072009-04-22T15:36:00.000-07:002009-04-22T15:36:00.000-07:00LexiFi sell a commercial implementation of this wr...<A HREF="http://www.lexifi.com/" REL="nofollow">LexiFi</A> sell a commercial implementation of this written in <A HREF="http://www.ffconsultancy.com/products/ocaml_journal/" REL="nofollow">OCaml</A>.Jon Harrophttps://www.blogger.com/profile/11059316496121100950noreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-76515017244795746202008-10-15T15:16:00.000-07:002008-10-15T15:16:00.000-07:00Regulations are programs for an open system: ther...Regulations are programs for an open system: therefore completely insane, conceptually incommensurate with reality, etc.<BR/><BR/>We don't need more laws (ditto programs for open systems), we need only a few, all variations on "Seller beware, you have the responsibility to provide the buyer with all of the information they need to make a good decision, to do it in a way they can understand, and to check that they in fact did understand it." and "size of the liability / responsibility is proportional to the size of the possible loss by the buyer".<BR/><BR/>That would lower the influence of lawyers and lower the cost of doing business, at least for honest businesses, thereby giving them an advantage.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-66174315501103160942008-09-29T14:14:00.000-07:002008-09-29T14:14:00.000-07:00This is a revolutionary idea. That's why it will ...This is a revolutionary idea. That's why it will require a revolution to implement it.s9https://www.blogger.com/profile/08326725206530909278noreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-71208147197059063912008-09-28T16:34:00.000-07:002008-09-28T16:34:00.000-07:00Credit Default Swaps issued by highly leveraged op...Credit Default Swaps issued by highly leveraged operators among themselves is semantic nonsense. <BR/><BR/>If Haskell could be used to demonstrate that this system is nothing but a house of cards, then the government has no choice but to pay attention and regulate, or firewall this sector from the rest of the economy. <BR/><BR/>However, I doubt such assessments of value can come from a pure specification, without some subjective assignment of what real-world values actually are.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-53849124223423134442008-09-27T15:55:00.000-07:002008-09-27T15:55:00.000-07:00"Thats why I want the global model to be owned by ..."Thats why I want the global model to be owned by the regulators, not the banks"<BR/><BR/>The current crisis does not exist solely because of fraud. There is an argument that we simply don't understand how to model many of the financial instruments. Buffet has warned about them, Mandelbrot (of fractals fame) has warned about the underlying mathematics and Taleb is now known for explaining how we underestimate risk.<BR/><BR/>Taleb and Mandelbrot argue that we don't even know how to accurately describe the movement of simple stock prices (not predict them, just statistically and accurately describe them).<BR/><BR/>Besides, regulators already audit risk models of depository banks. Doesn't help much when the underlying models underestimate risk.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-33671850413065371832008-09-27T12:16:00.000-07:002008-09-27T12:16:00.000-07:00You start off by pointing out that lying about mor...You start off by pointing out that lying about mortgage applications was a factor, but your proposed solution does not explicitly address this epidemic of lying. If companies lie to the government about their client's incomes, no amount of bogus models will allow the government to correctly model what's going on.<BR/><BR/>Clearly, any solution needs to involve enforcing existing laws against fraud (and mis-selling) far more vigorously.Robinhttps://www.blogger.com/profile/05420921538604886480noreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-32577082974807183922008-09-27T11:31:00.000-07:002008-09-27T11:31:00.000-07:00The issue is really how these assumptions are used...<I>The issue is really how these assumptions are used to either push a model in the wrong or right way.</I><BR/><BR/>Thats why I want the global model to be owned by the regulators, not the banks. The banks would be required to disclose their contracts in a computer-readable form, but its up to the regulator to put all those contracts together into the model.Paul Johnsonhttps://www.blogger.com/profile/07353083601285449293noreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-52014014769510152882008-09-27T09:08:00.000-07:002008-09-27T09:08:00.000-07:00I think one of the biggest things you are missing ...I think one of the biggest things you are missing is that many of these contracts events or contracts are nearly possible to model by hand. Clearly spelling them out and putting them into a computer to automatically automate is a much larger problem. In many cases the only way to model things is by making assumptions to bring the state space down to something that can be mathematically understood. The issue is really how these assumptions are used to either push a model in the wrong or right way. In the case of the mortgages, the assumption was simply that prices would continue up. If you had put this into a computer, it would have had the same effect as if a human decided it. It almost seems like you are proposing a system that can automatically detect new problems with contracts before those problems happen.<BR/><BR/>Although interesting, I don't think Haskell is the solution to this problem. The solution is to somehow regulate coporate greed. (AND actually attempt to model some of the crazy assumptions they come up with).<BR/><BR/>This isn't the only time that assumptions have caused people to lose a lot of money, and it won't be the last. No amount of modeling will ever fix this problemphubabahttps://www.blogger.com/profile/10396008730638774605noreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-77046846160344311242008-09-27T02:15:00.000-07:002008-09-27T02:15:00.000-07:00Yes, this is indeed a technocratic solution. Fina...Yes, this is indeed a technocratic solution. Finance is technocratic by nature. A pure "free market" won't work, but on the other hand having politicians run the banks won't work either (they will simply "lend" money to uncompetitive companies that happen to employ lots of voters). So you need to let the banks look for the profit on one hand, but keep a tight rein on their risk-taking on the other. The problem is how to give the regulators the necessary visibility into what the banks are actually doing.Paul Johnsonhttps://www.blogger.com/profile/07353083601285449293noreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-23904731643145767232008-09-26T19:47:00.000-07:002008-09-26T19:47:00.000-07:00The contracts paper is indeed very interesting, bu...The contracts paper is indeed very interesting, but I don't think it would've helped in the current crisis. The paper does talk about doing risk valuations, but financial firms already measure risk. Their risk models can be fairly simple (historical VaR) to quite complex, requiring teams of PhDs. If the the computer is given wrong assumptions about correlations, probability distributions, liquidity, the result will be wrong as well. Garbage in, garbage out.<BR/><BR/>The paper is still awesome, it just won't save the world :)falconhttps://www.blogger.com/profile/12277230403055479892noreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-68958829758674899882008-09-26T17:26:00.000-07:002008-09-26T17:26:00.000-07:00I agree with the first commenter: this is a very i...I agree with the first commenter: this is a very interesting (from a theory/tech point of view), but very technocratic. The value in these sorts of technologies is, as you say, detecting these sorts of "bottom valued" constructions, where a single mis-step will make the whole thing crash and burn. But tying any regulatory oversight to any particular formalism is just giving them another target.<BR/><BR/>Shifting the goal posts so the companies find ways to lie to a regulator with honest mistakes in formal languages probably won't fix the problem -- the incentives are still there and (if it works anything like most other areas of corporate regulation in the US) the disincentives will likely be pitifully inadequate for the purpose of deterrence. I doubt that I will every be dissuaded from my belief that the solution to these sorts of problems is broad regulatory power backed with enormous personal fines and jail terms for traders, managers, executives, and others who are culpable or ought to have an obligation to prevent it (along with regulations to prevent the corps from paying fines, etc. for them).<BR/><BR/>On the "pro" side, any move like this would buoy the market for type theorists and the like.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-5975524006824862804.post-29736812946534984742008-09-26T15:11:00.000-07:002008-09-26T15:11:00.000-07:00I think this is genuinely interesting and very tec...I think this is genuinely interesting and very technocratic. However, I think the finance world is too big to regulate efficiently and instead it would be best to manage incentives rather than try to model risk for an entire industry. Nonetheless, everyone should be using Haskell more. It could really save the world :DAnonymousnoreply@blogger.com