Introduction to Calculating Risk

[Ari]

Hey everyone,
So I'm taking an online finance MOOC (very basic, basically just intro to finance kind of thing), and we came to the concept of risk. Qualitatively I understand the basic principles of risk, but quantitatively the lecturer basically just glossed over it saying that it's related to the variance of return, but kind of left it at that.

Any good resources for getting your feet wet with calculation of risk? E.g. let's say I wanted to calculate the risk of the S&P 500 (which could be a proxy for the riskiness of the overall market), how would I do that?

I'm wondering if calculation of risk is like Quantum Mechanics in the sense that you can either do it in an oversimplified way or a really complicated way.... i.e. no middle ground.

Thanks,
Ari

 

[qdogg]

From personal experience, I now have three different answers to this--

1) As a business school graduate and a CFA charterholder, the definition of "risk" (according to "Modern Portfolio Theory") is volatility
2) As a "certified FRM," the definition of risk (according to the "Global Association of Risk Professionals") is unexpected volatility
3) As a former risk quant who now works in the front office-- GARP is right, not the CFA Institute, but if someone with a math background ever tries to get you to believe that he knows how to compute "unexpected volatility," you should punch him in the face

good luck

 

[Alexandre Chretien]

There are several risk measures and even rules describing whether or not a measure of risk is coherent.

You can think of volatility as a measure of risk of a position but taking into account skewness and kurtosis is also interesting.

You can try reading about Value at Risk and Expected Shortfall. These are common risk measures.

 

[yetanotherquant]

You can try reading about Value at Risk and Expected Shortfall. These are common risk measures.

Yes. Moreover, calculation and reporting of VaR is required by the most of regulatory frameworks, so this alone is a good reason to learn VaR.

However, the fundamental (and generally unsolvable) problem is how to collate the pairs of expected (excess) return and the risk measure.
Since I study(ied) optimal trading strategies for retail investors I took maximum drawdown as the risk measure. I did my best to build a realistic model and simulated scenarios from it. Then I showed to a concrete person, which drawdown and which return s/he can expect. Usually, this person was able to decide for him/herself, which strategy would be optimal (or at least the best among presented). But the individuals have different risk aversion, so the answers we pretty different.

 

[zyz]

This is an interesting topic.

Is there any comprehensive list of risk measures by asset classes? While VaR and ES can be used in most of the cases, I think using market-specific exposure measures for different sorts of instruments that directly affect P&L is also very helpful.

For example:
Equities - betas, standard deviation;
Fixed Income - duration, convexity, DV01;
Options - Greeks;

Are there any other good risk measure that would show a better picture?

 

[Ken Abbott]

This is an interesting topic.

Is there any comprehensive list of risk measures by asset classes? While VaR and ES can be used in most of the cases, I think using market-specific exposure measures for different sorts of instruments that directly affect P&L is also very helpful.

For example:
Equities - betas, standard deviation;
Fixed Income - duration, convexity, DV01;
Options - Greeks;

Are there any other good risk measure that would show a better picture?

There are dozens. Maybe hundreds. Many are firm-specific. For example, some firms measure CSPV01 while others use PV1% or PV10% for credit spreads. Some firms normalize that measure at the 5y point to eliminate curve influence.

Bloomberg will have a decent, if not comprehensive list. Go to an Open Bloomberg and open the spreadsheet app.

 

[Tom Maloney]

There are dozens. Maybe hundreds. Many are firm-specific. For example, some firms measure CSPV01 while others use PV1% or PV10% for credit spreads. Some firms normalize that measure at the 5y point to eliminate curve influence.

Bloomberg will have a decent, if not comprehensive list. Go to an Open Bloomberg and open the spreadsheet app.

I think it's important to emphasize that while CSPV01, PV1%, PV10% are different analytics they all represent the same concept - what is the change of my asset's value when I wiggle the credit spread a little bit. Mathematically you can express a wiggle in different ways, but the concept is the same.

You'll generally see two different types of risk analytics:

1. Asset valuation sensitivity to the valuation function inputs (e.g. "the Greeks")
2. Asset valuation changes due to various market scenarios (VaR, Lehman, China deval, Oil shocks)

You could even argue that (1) is really just a special case of (2).

 

[Le Van]

I was reading some academic papers on systemic risk measures such as CoVaR that is able to quantify a single bank's contribution to systemic risk while VaR cannot. Is the idea of measuring systemic risk appealing to banks at all or is it only applicable to regulators? AFAIK IMF has started to use CoVaR in their stability assessment report.