Value at risk

From Chapter 13 of The Analyst's Handbook, by Matt Lang and Bree Lang

Suppose that the treasurer of a student fund is asked how much the fund could lose on a bad day. An answer of "all of it" is technically correct and of no use to anyone, so what the question is really after is the size of a loss that would be unusual without being a catastrophe. Value at risk (VaR) is the usual way of putting a number on that.

A one-day VaR at 95% is the loss that the fund should do better than on 95 days out of 100. The simplest way to estimate it assumes that the fund's daily returns follow a normal curve:

VaR = μ − z × σ

Here μ is the average daily return, σ is the standard deviation of the daily returns and z is the number of standard deviations that leaves the chosen share of days below the line (1.645 at 95% and 2.326 at 99%).

Imagine a fund with an average daily return of 0.3300% and a daily standard deviation of 0.5116%. Its VaR at 95% is 0.3300% − 1.645 × 0.5116% = −0.51%. This means that on 19 days out of 20 the fund should do better than a loss of 0.51%. If the fund is worth $11,352, that is about $58.

The answer depends heavily on the confidence level we pick. With the same two inputs, the VaR at 90% is −0.33% and the VaR at 99% is −0.86%. In other words, the same history can be made to say that a routine bad day costs a third of a percent or most of a full percent, and a VaR quoted without its confidence level has not told us much.

There are two things a VaR figure does not say. First, it says nothing about the twentieth day. The VaR marks where the bad days begin and not how deep they go, and the measure that covers the second question is the conditional VaR (CVaR), which averages every loss beyond the line.

Second, the normal curve tends to understate how often markets have very bad days. One alternative is the historical VaR, which sorts the actual daily returns and reads off the fifth-worst out of every hundred. The historical method has a weakness of its own, however, because it can never report a day worse than the worst day in the window. A fund with sixty calm days behind it will be told that its worst case is one of those sixty days, which is comforting and probably wrong.

Because of this, it is worth showing more than one method and paying attention when they disagree. Chapter 13 of the textbook covers all three, along with stress tests.

Read Chapter 13: Stress tests, VaR, and CVaR

Related topics: Maximum drawdown · Correlation and diversification · All topics

The Analyst's Handbook is Sifter's free textbook for student investors, written by Matt Lang and Bree Lang. Sifter is free for students. This page as markdown.