- 1/26/2014 2:35:08 PM
You raise an important point. Many banks do not emphasize RAROC (Risk Adjusted Return On Capital) enough--even after this last in the financial meltdown series. I see Fannie Mae as an excellent example of ignoring risk managers (there is a sidebar in Chapter 2 of my book). I was there during the clean-up and interacted with those on the risk side.
As for AIG, Michael Lewis interviewed some of them afterwards and I was fortunate enough to speak with someone, who knew a little about it. It may be that there was a risk management issue at AIG--I am not familiar with that. I would like to learn more, please send references if possible.
Whether it is risk management or not, some decisions require a quant background; and quant teams should be led by a quant, who understands what they do and what they can do. I have written elsewhere how this failure got AIG into trouble.
- by Broadway0474, Blogger
- 1/26/2014 1:57:26 PM
@Randeroid, if by quants you mean "risk managers," then the problem at AIG was people ignored the risk managers. Too much money was to be made, and there's no way that big institutions like Lehmans or, gasp!, AIG would ever go under....
- 1/25/2014 9:52:18 AM
RE: You bring up some good points, but not on the "failures at AIG, rating agencies, etc.". Highly qualified quants were doing the quant there. The failures were not at all technical.
RESP: You are quite right! There were qualified quants building the models and the failures were not theirs. I should have been more specific (I value brevity). I wrote about this earlier in 'Three Good Reasons Why You Need A CAO.' At AIG and the rating agences, the failures were caused by off-topic managers making decisions that should have been made by quants. I view management of the quant team as technical (see book) and that is what I meant by 'examples of why non-quants should not try to do quant.' I am so sorry about any confusion. For the rating agencies, the quants were unjustly blamed for the failures of people, who would agree with the 'non-quant' conclusion in this survey. Correction appreciated.
- by Taavo, Prospector
- 1/25/2014 7:47:22 AM
You bring up some good points, but not on the "failures at AIG, rating agencies, etc.". Highly qualified quants were doing the quant there. The failures were not at all technical.
- by rscollica, Prospector
- 1/22/2014 5:27:22 PM
I would agree with you that analytic experts are not used enough and there aren't enough of them to go around in industry as a general rule. Here's why I think this is so. First, if a business problem has only a small impact of getting it incorrect, then one can live with some slop in the analytic model. However, if one is tackling a larger more severe business issue, say churn in the Telecom industry space, then a very small change can have a huge business impact (both positive and negative)! Second, I've seen a best practice in the banking industry where one pairs an MBA with domain knowledge with a "quant" and both of these work side-by-side to solve the business problem/project together. So, in order to assess the business situation of when do we need this sort of pairing, I would recommend both a qualitative and quantitative assessment. For the qualitative part, I think the domain expertise could bring the business requirements to the table in stating "what bad thing happens if we don't have this analytic model?" Answering that question should give a good idea of how impactful or not the situation might be. In the quantitative portion, the "quant" can assess the situation by quantifying a decision table such that if model output is incorrect, the cost/benefit might by X and if model is correct (to some level of accuracy) then the cost/benefit is Y. In this fashion, both of these assessments can be used to help prioritize which predictive analytics have the highest to lowest business impact. If you want to get more sophisticated, then resources to develop these can also be included. Food for thought
- 1/16/2014 3:55:31 PM
I think we do not use quants enough. As I understand this article, the basis of their conclusion is asking non-quants. Well, we live in a society with a low statistical literacy. Instead, I would point to the failures at AIG, the rating agencies, etc. for examples of why non-quants should not try to do quant. Corporations need better data analysis governance and they need to look for certifications while more user-friendly software and more access to data remove 'the safeties.'
- by BethSchultz, Blogger
- by @Deepmile_Chris, Prospector
- 1/16/2014 9:50:28 AM
I absolutely look for analysts who can go beyond the deductions and show the implications and opportunities. That means I need creative consultants. So I'm looking for quants who can consult or consultants who can do quantitative work. When I say I'm looking for a data analyst, what I'm really looking for is a software engineer who knows what I want to do with his or her skills.
I think we confuse the skills and competencies when we call for "predictive analytics". There's nothing predictive aboud analysis until we apply creative thinking. Quantitative analysis is phenomenology; it looks to the past. When we say "predictive analytics," all we're doing is embedding an assumption that past behaviors will be repeated in the future. Maybe it works out, but we're asking a lot of analysis unless we also fold in the creative and qualitative analysis.
- by kq4ym, Data Doctor
- 1/16/2014 8:31:59 AM
When analysis "for everyone" becomes more common it will surely be the non-quants who will rule for the cost saving involved most likely. But, until then it will have to be based on a cost-analysis of the benefits. But, as pointed out that's not always going to be an option because of the extra costs and skills needed in making the analysis. I would still think that there's going to always be a demand for the very skilled analysts and statistician types to lay the ground work for the more demanding and critiical taks, for a long time to come.