Value at risk: Et risikomål set i lyset af den nuværende krise

Emil Sjøntoft

Student thesis: Master thesis


Value at Risk is defined as “the maximum loss that will occur within a significance level over a certain period” The overall question I want to answer with this thesis is: “Is Value at Risk a risk measure that can be used properly during the current financial crisis?”To answer this question with the thesis, I first go through the different practical aspect of building a Value at Risk (VaR) model. Then I look at the critical sides of using VaR. At last I go through VaR and how it has performed during the crisis, and the current discussion of the use of VaR. VaR has underperformed drastically during the last two years. That raises a discussion on the overall reliance on VaR’s ability to measure market risk. The discussion mainly evolves around the length of the periods of historical data, that should be used to calculate VaR. The longer the period, the more former extreme events, and the higher probability will VaR devote to extreme market movements. The shorter the periods, the faster the VaR model will adapt to changes in the market volatility, which will warn the user, that the risk might be increasing. The other part of the discussions concentrates about the ability to measure and forecast the volatility, and the use of the normal probability distribution to calculate VaR. If VaR fails to measure the market risk, and the actual loss reported by the banks is higher than VaR predicts more than 5 % of the time (with a 95 % confidence level), then it could indicate that VaR is misleading and therefore is a danger to the user. There are two main arguments for VaR being misleading. The first argument being that VaR doesn’t say anything about the risk hiding outside the confidence level. If VaR is calculated to be 2 mill Kr. within 95 % confidence, then 5 % of the times their will be a loss over 2mill Kr. The loss could be 2.1 mill Kr. or it could be 10 mill Kr. which is catastrophic. This also gives incentive for the traders to take positions with under 5 % probability for losses, which then will show that the traders make profit with very little risk, when indeed they are taking a lot of risk, because the risk with the probability under 5 % is very harmful. VaR is also misleading when it underestimates the risk, which it has shown to have done during 2007 and 2008 of the current crisis. The question is why VaR have been under estimating the market risk. There are three main reasons: VaR hasn’t reacted to the changes in volatility, the past isn’t a good indication of the future and the normal probability distribution can’t be used to describe financial returns. There are too many assumptions in the theory that don’t fit the real world markets. The volatility and correlation is forecasted from historical data, which has shown just as good as a pure guess. There is no reason to believe, that history will provide a reasonable picture of the future, because markets change and adjust to the past events, and if something is predictable, it will not be truly harmful, and visa versa. I think the main problem is that VaR is expressed with such accuracy. Leave the probabilities out of the calculation, because they are wrong and can’t be trusted, which is why they make more harm than good, if the management and investor believe blindly in the risk profile reported by VaR. I recommend that the risk managers perform scenario tests, with likely and unlikely scenarios. It could be fx.”if the LIBOR rises 1,5% within the next 1 month these positions will lose 0,5 mill kr.” but without attached probability that provide false security, instead the trader or investor will have their own opinion about the likelihood of such an event.

EducationsMSc in Finance and Accounting, (Graduate Programme) Final Thesis
Publication date2009
Number of pages75