ABSTRACT

Periods of turbulence are often characterized by observed consecutive drops in prices, called “drawdowns”. In such cases, static, one-period measures of risk insufficiently describe downside risk. In this paper, we assess risk by formally modeling these random strings of negative returns. We use long-tailed distributions from extreme value theory to model the severity and duration of the drawdowns. This leads to the concept of drawdown-at-risk (DAR) and conditional DAR. The proposed distributions adjust properly to extreme values previously found to be outliers. The paper illustrates the importance of these concepts with stock market indices.