Tuesday, February 02, 2010
The Future of Risk Management and Financial Ratings
I recently came across a thought-provoking table purporting to compare the present and future of risk management and financial ratings (2009, Ontonix). Most of the predictions are illuminating. However, I doubt that financial risk analysts will migrate away from statistics and Monte Carlo simulation (MCS) models toward "model-free" methodologies (whatever that means). In fact, I predict that the use of statistics and MCS methods in financial risk analysis will expand in the coming years. Nevertheless, I find the juxtapostion of the conventional and future to be parallactically interesting and intriguing.
Reference: "The Present and Future of Risk Management and Rating," (2009), Ontonix Complexity Management.
2 comments:
"Model-free" means that you don't try to impose a mathematical model on top of some data, i.e. a line, a parabola, etc. You leave the data "as is" and you don't manipulate it. What you do is extract some of its features (entropy, "amount of structure", attractors, etc.) pretty much like Nature does. Nature doesn't build models but our education, for some reason, makes us think in terms of functions and statistics. I am not sure if the methods that have created this global mess will continue to thrive.
Thanks for the explanation of what was meant by "model-free." I'm still pondering the "model-free" perspective, but if I understand you correctly, a "model-free" understanding views data nominatively while rejecting a realist or conceptualist view of the world. In epistemic terms, the "model-free" approach views data iconically, and therefore rejects alternative analogic or symbolic frameworks. Said still a third way, the "model-free" approach is to undertake a quasi-empirical view of the data, making logico-deductive and intuitive-inductive methods essentially irrelevant. While I understand and respect the modernity implied by "model-free" thinking, I am always suspicious of the thinker who summarily rejects logic and intuition out of hand. In particular, I would point out that quasi-empirical methods are only useful for verification and discovery of theory, and fail to contribute functionally to explanation, systematization, or communication. Thus, I would urge analysts to triangulate their methods and thinking around the problem through the concomitant use of symbolic, analogic, and iconic models and frameworks, while allowing pragmatism to be the final judge. Now, I will concede one point, and that is a "model-free" review of data in advance of analysis is essential, and it certainly appears that financial analysts, managers, and policy-makers were derelict in reviewing the data prior to the financial meltdown. Thank you for the opportunity to respond and comment. You still have me thinking…
Post a Comment