Value at Risk

[karlmarx]

i encounter the term 'value at risk' every now and then when reading annual reports and even on tv. i only know it is used in risk management, and big traders rely heavily on this. but i have no idea how it is calculated or what its reading mean. can anybody shed some light on what VaR means or how it is measured? thanks

[cyclone]

Value at Risk (RaV) is defined as the maximum potential change in value of a portfolio with a given probability over a certain time period.

Value at Risk measures the the potential loss in value of a portofolio over a defined period (t) for a given probability (x).

If the VaR on an asset is $ 10 million at a one-day, 5% probability level, there is only a 5% chance that the value of the asset will drop more than $ 10 million over any given day.
Or we can expect a loss of $ 10 million or more on one day in 20 days.

[arthur]

Supposedly to be the best intro book for this topic.

http://www.amazon.com/Value-Risk-Benchma...ep_dpt_2#_

But if my memories didn't fail me, many financial risk mgrs complained (upon hindsight, of cos!) that this system broke down (ie. didn't anticipate the risks) when the 08 crisis hit all the banks.

[freedom]

VaR is a statistical measurement. And like all statistical measurement, it would fail miserably when black swan kinda events happened such as 08 GFC.

IIRC, a lot of VaRs are calculated with 90 - 95% statistical confidence only.

[brattzz]

more and more terms leads to more confusion...

[karlmarx]

i asked because i came across this term more than a few times when reading noble's AR. they claim to have a VaR of less than 1% of their equity. i'm not sure what that means. i'm trying to understand how they lost so much money on cotton one quarter ago. would the default by cotton farmers be considered such a 'black swan' event?

[freedom]

IIRC, Noble said the cotton default was 3-sigma event, which should be beyond the VaR control of Noble Group. I can't be certain, need some statistics expert help

[shanrui_91]

1 sigma is equals to 1 standard deviation, 3 sigma equals 3 std deviation. Using the normal distribution (bell curve), 3 std deviation from the Mean is equals to 0.26% chance of occuring

[Spidey]

VAR is a model that disregard the worst outcomes. For example, the 95% means that the model disregards the 10 worst day in the market. Also it assumes that there is no trending, i.e. it assumes consecutive 10 down days cannot happen. VAR does not say anything about what happen if the 10 worst days happen consecutively. For example, VAR of 1% does not describe what will be the loss if the 10 worst days of cotton price move against you. The loss can be 30% in the 10 days, for example.

In simple terms, what they are saying is, “oh don’t worry, this river is on the average only 4 feet deep, that 8 feet hole is just an outlier and it does not exist in the world of statistics, or VAR.”

VAR is usually use for position sizing, i.e. how much to trade. In addition to VAR, you need to know what is their cut loss discipline, i.e. do they cut loss at 1% or 5% or 20%.