Extreme events are an important source of financial risk, but they present special challenges in quantitative forecasting.  In this paper we describe an empirical approach to forecasting extreme risk and evaluate its accuracy out-of-sample on a range of factor-based strategies and pair trades.  Our results show that for a large majority of strategies, our model is more consistent with market behavior than a conditional Normal model.

Factor models are standards in investment management. For decades, Barra factor models have provided valuable risk forecasts and inputs for the portfolio construction process. Most uses of factor models have targeted longer horizons of months or years. However, we demonstrate in this paper that factor models can also provide accurate risk forecasts for shorter horizons of one to ten days. Furthermore, factor models have the advantage of explaining risk sources and providing consistency in...

Systematic model bias has been implicated in the global recession that began in 2007, and this bias can be traced back to assumptions about the normality of data. Nonetheless, the normal distribution continues to play a foundational role in quantitative finance. One reason for this is that the normal often emerges, without prompting, as the distribution of sums or averages of large collections of random variables. Precise statements of this miracle are known as Central Limit Theorems, and...

We show that for a diverse collection of 74 US equity portfolios, risk forecasts based on an extreme value theory model greatly outperform a conditional normal model with a 23-day halflife.