strange idea

dehaan · Post by **dehaan** » Mon Nov 30, 2009 11:06 pm

Hi,

First, a disclaimer. I do not realy like the terminology of mathematical phynance; I find it a bit misleading and arcane; nevertheless I use it. I happend to work during last 10 years as a quantitative analyst, mainly for risk management and phynancial engineering departments of different phynancial entities. What follows next is my recent experience of emloyee of a bank.

i was asked to quickly quantitatively model a ruther simple investment strategy, the aim being to quantify its risks via Monte Carlo simulations. Once done I showed the results to my boss, saying that the random evolution of the risk factor was modeled in the simplest possible way, as lognormal diffusion with historical drift and historical volatility. The historical drift being very close to zero, 10^-6, I made it zero and took it away from the equation, thus reducing the quantity of summations and reducing the time of the Excell calculations. The boss made big eyes and bent his head to the left shoulder, looking tenderly. "Why did not you use the rsik-neutral probability?", said he. "Because we are estimating the risks, we ain't doing no pricing" said I. "But you should have to.", said he. "Why?!!! For Christ sake, why???!!!!!", cries me. "In this way you would COVER yourself, protecting yourself from possible reproach-that you did not do it- from the bigger bosses", calmly says he. "But we are no risk neutral investor: for us the risk - the volatilities, the vars, the esxpected shortfalls - do matter; we do not realy care about the expectations: this being the only care of the people who do pricing; we are doing risk management here! Therefore why the hell should I do this under risk-neutral measure????" cries me. "I came from the parts of the world, where there were no free lunches", continues my boss. "But we have got no any free lunch in these simulations! Look: there are scenarios - states of the phycking World - where we have got negative returns-and scenarios where we have got positive returns - there is no free lunch here! We do take risks, and our estimation of risks do depend on our view point (models, parameters, etc.)", does not stop me. "This is something which would be extremely difficult to explain to my boss, who is used to see only one number. But once we say that we are doing it in the risk-neutral way, he is probably to become impressed and would probably ask no more questions. This is in your interests: for you ll have less work to do in December."

The subject matter of this conversation still comes back to me. How in the world can we do risk simulation - risk budgeting - ALM - Markowitz - Black-Litterman etc in the risk neutral world....? how the idea of this became possible? I do not know what to say....Is the idea sane or simply mad?

sfca · Post by **sfca** » Tue Dec 01, 2009 12:15 am

You may not be as right as you think you are. First of all, you should give him the benefit of a doubt that if he is your boss, he may actually know some things. And, its fine to express a difference of opinion, but you probably should not argue with the person responsible for your paycheck. You work for him so your job is to do things for him, not undermine him. With respect to producing risk neutral estimates for superiors, he may actually have very good reasons. For one, it will appear as an arm's length number unbiased by his own views. He can say that is the market's expectations, instead of his own estimates that others may suspect are biased by his interest in managing the portfolio a certain way. Your boss may be contrained by knowing that it would take too much effort to explain finance to suspicious managers. In addition, you say he does look at scenario analysis so its not as if he is clueless about the potential values. I agree with you that risk neutral valuation can be a poor metric in many cases. It may be that you are looking at this from a micro/tactical standpoint while he is looking at this from a macro/strategic/management standpoint. There is definitely information missing from your story, but my point is just that there may be a different way to look at this than the right/wrong question on RN valuation. He may be doing management.

pj · Post by pj » Tue Dec 01, 2009 11:32 am

> He may be doing management.

I think it's precisely the point.

In my first job, I was helping to build

a historical market model for

insurance companies.

Historical, meaning the drift.

The insurers do not hedge, they buy and hold.

My boss, tired of the explanations about historical versus risk neutral measures (he was firmly in favour of historical), simply put in the option to use the risk neutral drift.

No one was using that option and everyone was happy.

My 2 cents.

nnja · Post by **nnja** » Tue Dec 01, 2009 2:25 pm

Your problem isn't a question of whether you should do risk management in the real world or the risk neutral measure. Your real problem is that you weren't really asked to quantify risks, you were asked to produce filler so that your boss's boss could put a mental check in the due diligence box. Sometimes that is necessary, like for government regulations, accounting stuff, etc. But I share your discomfort with it in this case.

However, your boss might know that there is more to this story. Maybe his boss isn't really considering the strategy seriously; he's just scratching an intellectual itch. Then the desire to satisfy the big boss while doing as little work as possible makes some sense.

So while you may be technically right (the best kind of right /sarcasm), I don't think that you are dealing with a technical question here. Maybe this thread should be in Careers?

dehaan · Post by **dehaan** » Wed Dec 02, 2009 3:47 am

Thank you guys; very interesting. I will certainly avoid telling details because I am not suicide at this moment, just slightly paranoid.

nnja: I will certainly avoid being sarctastic. Initially I was not sarcastic: I was furious. And since my fury was not taken seriously, i grew sarcastic. Bad idea, i agree.

I have just never ever seen this before. All risk managment softwares that i worked with before and that produced MC simulation of the market, did the simulation of the risk factors under historical measure - plus risk neutral valuation if needed (full repricing of options, etc.) in all "historically" simulated future states of the world. I d very much like to see a formal argument in favour of producing risk numbers under RN measure, or, for that matter, under historical measure.

Risk Neutrality is a category of the future, a forward looking concept. One can try to back-test it, however, if one wishes. But taken as it is, it means that an investor bases his decisions only on mathematical expectations. In other words, mathematical expectations being the only criterion allowing him to discriminate between lotteries aka decisions. If a RN neutral investor sees that the expectations of all lotteries he is dealing with are the same, he would call the situation arbitrage free and would be completely indifferent wrt which one to prefer: the one with a lot of volatility will be the same as the one with no volatility, as far as they have the same expected return. So why does such an investor should care about risk? I think this is incoherent.

As a matter of fact, even if one bases his decisions on the forward prices (unbiased market estimations), everybody knows that at forward's maturity the market could be so far away from the intial forward that one'd be biting nails.

LongTheta · Post by **LongTheta** » Wed Dec 02, 2009 10:23 am

dehaan, How's your short bunds position?

sfca · Post by **sfca** » Wed Dec 02, 2009 6:12 pm

dehaan, one could easily argue that all if those risk management softwares are using historical metrics only then they may not be any better than RN. We just went through some extreme market values and situations and those numbers should be judged accordingly. The economy slowly coming out of a business cycle is not going to have the same characteristics as one in a recession or tipping into a recession. If you are just doing some kind of short term VAR numbers it may not matter. But if you are doing longer term analysis, for example, you could easily generate a negative 2% federal funds rate target in 2010 using recent historical trends. I would argue that some analysts use historical numbers to hide from doing economics in the same way that one can hide behind RN. Either way, someone who knows both economics and finance should be involved if a risk management system is to give good numbers. Its really unfortunate that economics has gotten such a bad reputation from both arrogant clueless academics and the collection of sell side twits at the investment banks (come on guys, their own damn traders don't take positions on what they say so why the hell would you point to them as proof economists can't forecast? They are there to entertain the customers and not to do anything else).

[edited to remove some minor swear words and again for clarity]

Mandark · Post by **Mandark** » Thu Dec 03, 2009 5:08 pm

I remember going to a presentation by AIG and JPMorgan at the end of 04 on Managed Tranche Notes (I still have the handouts). They used historical risk simulations to show that the odds of any shortfall for the AAA tranche was so infinitesimal as not to be worth considering. Nonetheless, you could earn 45bps over AAA corporates by investing. Simple risk neutral analysis of the spread tells you that you've got a greater than 2% odds of something bad happening over the 5 year life. Moral is that historical risk measures can be very, very wrong.

Personally, I like to see risk-simulations done RN so as to avoid any hindsight bias. History, I think, is best handled with scenarios. If either shows significant risk, you know you're in trouble. That doesn't mean you're safe, but at least it will alert you to situations where you are definitely not.

One more point, when you use historical, the guys managing the portfolio will tack on risks that the market has priced in but that haven't occurred in the past. They're not dumb. They know that by doing this they can get extra return while not showing the true risks and thereby increase their expected bonuses. Using RN keeps people honest.

dehaan · Post by **dehaan** » Thu Dec 03, 2009 11:48 pm

Unbiased? What about backtesting RN risk measures?

Honest? What does it mean wrt to this problem?

Is there any other formal, solid justification of using RN to see the real risks?

sfca · Post by **sfca** » Fri Dec 04, 2009 1:52 am

I think your question about justifications was answered. The use of RN valuation can reduce model risk. RN valuation will avoid judgements about the relevant historical periods to choose for historical simulations and can limit the risk from poor economic forecasts. The devil you know is better than the devil you don't know.

Second, I've been across the table from OCC examiners and they expected and were very happy to see RN valuation for some risk reports. In that case they were the most important consumer so the solid justification is that you are giving the consumer what he wants.

Personally, I prefer economics to RN to measure risk, but I work for my manager and he does not work for me. So, if my manager asked me to use tarot cards I would eagerly get out my Rider Waite deck and start shuffling.