Stress test measures the behavior of the portfolio under various market scenarios. For example, if the price of oil falls, a portfolio with more allocations to oil dependent companies will be more affected than portfolios with less energy exposure. This dependency to oil prices can be measured many ways: the simplest way is to perform a historical regression of returns from all the assets against oil and infer the portfolio sensitivity to oil, or historical beta. For the sake of argument, let’s assume that Beta is 0.2, i.e. a 1% move on oil prices (up or down) will result in a 0.2% move (up or down) in the portfolio. Then you can extrapolate that information to any shock magnitude: a -10% oil move results on a -2% portfolio move and so on.
We don’t use this simplistic approach for various reasons. For one, it is predicated solely on historical information. Also, it assumes that on extreme scenarios the relationship between the portfolio and oil will be preserved: we know that the 0.2 relationship will not be constant and correlations change during more extreme scenarios. And finally it assumes a linear relationship between all securities and oil, introducing serious errors when options are present in the portfolio.
Our approach is more akin to a statistician observing large quantities of data. It calculates the probabilities of extreme events happening and factors those probabilities into the calculations. For example: it knows that correlations and volatility increase in periods of large oil declines (as an example), resulting in large price oscillations for securities.
Stress tests is one of the most important tools in the investor quiver. It aims to determine how a portfolio will respond to extreme events and which securities (or group of securities) will benefit or detract on those scenarios.