As in, the kind of thing the CRO of a hedge fund would do–I’m not talking from the point of view of a hedge fund investor. I’m looking for a book or good review article on the subject that covers the tools and techniques used in the industry across the main strategies. Any suggestions? Ta.
Ask a question like that and someone will start spouting a bunch of statistical quantitative stuff! I’ll try to help, tho’: Risk management = tell your boss ahead of time the likelihood of certain losses based on current portfolio, history & some other inputs. Risk management in action = If the portfolio goes down, sell just like everybody else. Run.
I’ve often liked the analytical content of risk management, but it strikes me as a politically difficult position to hold. If things are going well with the fund, people don’t like you because you tell them to pull back on their investments because the risk to the portfolio or firm is too high. You are unpopular. If things go badly, then people turn around and blame you, the risk manager for not doing your job, because, oops… they took too much risk. A difficult payoff…
chadwick, if things go wrong those risk management geniuses have figured out a way to cover their butt by way of saying, “we estimate with 98% certainty that a loss of more than 50% is likely to occur more than 2% of the time” hmmm Ok whatever … thanks.
bchadwick Wrote: ------------------------------------------------------- > I’ve often liked the analytical content of risk > management, but it strikes me as a politically > difficult position to hold. If things are going > well with the fund, people don’t like you because > you tell them to pull back on their investments > because the risk to the portfolio or firm is too > high. You are unpopular. If things go badly, > then people turn around and blame you, the risk > manager for not doing your job, because, oops… > they took too much risk. > > A difficult payoff… Nah. Good traders like the risk manager. The risk manager is like an editor. S/he may not be anything like as good at trading as the trader, but a different perpective is valuable to everyone. Good trades blow up sometimes and it’s nobody’s fault. When you’re making money, a good risk manager is trying to ferret out other company risks (market risk is only part of the job) and that keeps everyone safe and happy.
sometimes risk management is compared to brakes. why do cars have breaks? so, that they will be able to go faster. something tells me that denali has practical trading experience. am I wrong?
denali, I don’t know of a good book and/or article on this, unfortunately. Essentially, the risk manager will be monitoring limits (in terms of sensitivities, VaR limits, capital limits…), checking out individual trading books to understand positions, spending time to keep on top of the markets. As Joey mentioned above, market risk is only a part of the job - someone needs to keep an eye on liquidity risk, counterparty credit, operational risk… Trading background, knowledge of the markets and trading strategies helps tremendously. I do not know about the statistical quantitative stuff mentioned above, maybe in an quant strategy place but in general, I do not think you need to be a quant to be a risk manager. But you should be strong enough to voice an unpopular opinion repeatedly. On the cynical side, I heard CEOs love CROs because a CRO is the natural person to fire when trades blow up. One more layer of protection for the CEO … I don’t know how much truth there is to it.
take the FRM.
FrankArabia Wrote: ------------------------------------------------------- > take the FRM. I did, three years ago, and back then there wasn’t really anything on risk management within a hedge fund–it was more focused on the banking context. Has this changed? I guess what I’m asking is, what are the metrics that a CRO will be looking at (so aside from counterparty due diligence, op risk etc.). FourCastles mentioned, for example, VaR, but few of the HFs I’ve seen calculate a VaR.
denali, VaR is one of the basic measures HF have to measure. It’s possible that they don’t use it as much as measures alternative (or complimentary) to VaR such as CVaR (expected shortfall), etc because VaR has serious weaknesses. Currently 30% of FRM exam is Market Risk Measurement and Management: http://www.garp.com/frmexam/frmexamfaq.asp
There’s much more to understanding risks of a position that calculating its VaR. The usual sensitivities to the underlying risk factors are usually calculated (delta/gamma/vega for equities, duration/convexity/DV01 for interest rates) but also FX exposure, credit exposures to individual counterparties, sensitivities to spreads. Various stress test scenarios are usually simulated for market conditions such as market crash, liquidity crisis, spreads blowout (flight to quality), currency crisis. Various liquidity measures will be calculated, such as survival horizon. VaR by itself is too high level. Remember, it assumes ‘normal’ evolution of risk factors. In a true crisis, correlations go to one and your VaR is not very useful, so you want to know much more about the positions.
Fun stuff!
maratikus Wrote: ------------------------------------------------------- > Fun stuff! which is why I left the world of risk.
>VaR by itself is too high level. Remember, it assumes ‘normal’ evolution of risk factors. In a true crisis, correlations go to one and your VaR is not very useful, so you want to know much more about the positions. That’s always been my problem with risk management - people seem to calculate their 99% max loss in 7 days to 8 significant figures, whereas we know that the time you actually need risk management is when there are scary shocks. I’m not saying risk management is a Bad Thing, just that many people seem to think its utility is greater than it really is. Speaking as a non-risk manager, obviously… 
chrismaths Wrote: ------------------------------------------------------- > >VaR by itself is too high level. Remember, it > assumes ‘normal’ evolution of risk factors. In a > true crisis, correlations go to one and your VaR > is not very useful, so you want to know much more > about the positions. > > That’s always been my problem with risk management > - people seem to calculate their 99% max loss in 7 > days to 8 significant figures, whereas we know > that the time you actually need risk management is > when there are scary shocks. > > I’m not saying risk management is a Bad Thing, > just that many people seem to think its utility is > greater than it really is. > > Speaking as a non-risk manager, obviously… Speaking as the risk manager, I think that’s nonsense. I know tons of senior risk guys and everyone of them is way, way more sophisticated than that. Further, this business about only needing risk management when there are scary shocks is just not reality. Risk management prepares you for scary shocks and keeps you out of all kinds of trouble. It’s not about calculating VaR.
Joey, I’m sure you and your mates do a wonderful job, and manage risks in ways so clever I can’t even begin to comprend them - and as you have outlined you use far more sophisticated models than simply a VaR. But the map vs territory problem remains, in spite of the complexity of the models. Also, as the complexity increases, the people who ultimately using the risk information (the traders, CIOs etc) understand the assumptions underlying the measures, and therefore their limitations, less and less. Then someone gets a bright idea to gear up, etc etc. It’s the little bit of knowledge being very dangerous. Ever since I was taught about standard deviation as a kid, I’ve never managed to shake the impression that it was a nasty hack - it always seemed that MAD was a much better measure of what it was really trying to measure, and accuracy was sacrificed for mathematical neatness. As stdev is a parameter in a large amount of statistics, I never really felt that comfortable with confidence intervals/hypothesis testing etc, even through my maths degree. I’ve always thought the outliers were more important than the bit in the middle, but just learnt the techniques along with everyone else. As a chap with a PhD in stats, you possess a far deeper understanding of the limitations/advantages of each technique than I do however!
FourCastles Wrote: ------------------------------------------------------- > VaR by itself is too high level. Remember, it > assumes ‘normal’ evolution of risk factors. In a > true crisis, correlations go to one and your VaR > is not very useful, so you want to know much more > about the positions. I think that’s overstating VaR’s scope. My understanding is that it’s little more than rank-ordering returns and identifying the appropriate tail. It’s agnostic as to how the return sample is created. For example you could sample historical returns from Aug-Sep 1998, apply those to your portfolio, and calculate a VaR. You’d indeed see some abnormal correlations in the process.
DarienHacker, you are suggesting looks more like stress testing. VaR, CVaR and stress testing compliment each other. Each has its advantages and disadvantages and each can be calculated using multiple approaches.
chrismaths, you are not giving RMs enough credit, I am afraid 
(ehm, a euphemism for ‘you’re talking out of your a$$, buddy’). Var is only one of (many) tools a risk manager uses. It is naive to think RMs do nothing but wait for their morning reports full of VaRs and then run around screaming. Any RM who has been through a single crisis realizes, of course, the limitation of not only VaR measures, but also of sensitivities and other tools.