Risk management for hedge funds AQF 2005 Nicolas Papageorgiou - PowerPoint PPT Presentation

Risk management for hedge funds AQF 2005 Nicolas Papageorgiou

Outline • VaR and drawbacks • Nonlinearities • Credit risk and liquidity

Can risk management be a source of alpha • Consider fund with E(r)= 10% and volatility=70% – Not very interesting • Consider risk management strategy that eliminates returns lower than -20% • Truncation of distribution will double expected returns and reduce volatility • But what is the cost? • According to BS – 15% of the portfolio • Risk management can add value and hence be a source of alpha

Why not VaR • Cannot capture the spectrum of risks that hedge funds cover • VaR is a purely statistical measure with no economic interpretation – It is a static snapshot of marginal distribution of a portfolio’s profit-and- loss.It does no capture • Liquidity risk • Event risk • Credit risk • Factor exposures • Time varying risks due to dynamic trading strategies that may be systematically keyed to market conditions (contrarian, short volatility, credit spread strategies • VaR difficult to estimate for rare events • - rare events.doc • VaR is an unconditional measure of risk

Dynamic Risk-Analytics • Can static measures capture the risk of dynamic strategies • Example of CDP

CDP returns (92-99) S&P500 CDP Monthly mean 1.4% 3.7% Std.dev 3.6% 5.8% Min -8.9% -18.3% Max 14.0% 27.0% Annual Sharpe 0.98 1.94 # negative months 36/96 6/96 Correlation w SP500 100% 59.9% 367.1% 2721.3%

What is CDP’s strategy • Short OTM S&P500 put options on each monthly expiration date for maturities < 3 months and with strikes approx. 7% OTM. • Would you be willing to pay high fees for this strategy? • Managers can easily use this strategy to increase returns…

Dynamic strategies • Managers can also replicate option like payoffs by using leverage on momentum based strategies – Voir CTAs and lookback straddles

Non-linearities • Phase locking behavior – Uncorrelated actions suddenly become synchronized • How do we capture this « phase locking » behavior • Phase locking behavior.doc

Non-linearities • Let us assume p=0.0001 • Generally expected return (and volatility) will not depend on Z. • What if volatility of Z is much greater than those of idiosyncratic risk and market factor? • Then Z will dominate expected returns when I=1. • correlation.doc

Non-linearities • How about calculating corelation using unconditional variances? • Unconditional correlation.doc

Non-linearities • Assymetric sensitivity to the S&P500. • Different beta for up market and down markets

Liquidity and credit risk • Autocorrelation as an indication of liquidity exposure – The more efficient, the less predictable…

Other readings • Risk and portfolio decisions involving hedge funds, Agarwal and Naik 2004.