Event

QWAFAFEW Denver - May 19th, 5:30PM - Dan diBartolomeo

The Choice of Model Factors under Multiple Definitions of Risk with Dan diBartolomeo, Northfield

Risk assessment in asset management is intrinsically multi-dimensional. Investors may be concerned in differing degrees with a multitude of risk measures such as tracking error, active risk, absolute volatility, VaR, CVaR, and first passage risk (drawdowns). We will begin with a review of the relevant literature such as Roll (1992) and Wilcox (1994) and Kritzman and Rich (2002). The presentation will then examine how the different definitions of risk might influence our choice of which factors include in both alpha and risk models. If I hold a benchmark index fund and add a low beta stock to it, tracking error will increase but absolute risk decreases. Some factors such as balance sheet gearing (debt/equity or similar ratio) would cause comparable behavior. If my portfolio has an average gearing level greater than X, this will increase tracking error and increase absolute risk, while portfolio gearing below X will increase tracking error but must reduce absolute risk (a company with no debt cannot go bankrupt). However, for many factors (e.g. momentum, book/price) there is no immediate intuition as to which way factor bets normally contribute to absolute risk, potentially making factors bets harder to interpret. The presentation will conclude with a numerical simulation that illustrates the multi-period problem for long-term investors. For the normal IID (random walk) almost all risk measures are a scalar of a standard deviation around a mean which itself is unknown. In the long run, we demonstrate that uncertainty of the mean is at least as important as volatility around the mean. We show that it is often helpful to take bets on alpha factors that have zero factor return on average if the factor returns are negatively correlated with the market and hence decrease the absolute volatility and tail risk of the portfolio.