Maximizing Returns, Minimizing Max Draw Down

Transcription

1 RISK MANAGEMENT CREATES VALUE Maximizing Returns, Minimizing Max Draw Down For EDHEC Hedge Funds Days 10-Dec.-08

2 Agenda > Does managing Extreme Risks in Alternative Investment make sense? Will Hedge Funds still be there tomorrow? Can quant models work when the market melts down? > After the storm, which new deal for the survivors? Risk transparency: the real commitment. Risk process: to manage diversification in any circumstances... and the fastest road to all outstanding opportunities

3 What sacrifice must we make to the gods of the market? > Complaining about the market is useless Nate Beeler, The Washington Examiner, Oct 17, 2008 > Rationally, current events are not unpredictable, nor unbelievable > Elsewhere, we fall into irrationality: people want a sacrifice

4 HEDGE FUNDS? > They did not have a great performance during the crisis, even when adjusting the leverage to make it comparable with other investments in term of tail risks Jan-07 Feb-07 Leveraged strategies from janvier-yy up to octobre-yy targeting a worst case per month of 0.4% HFRI FOF Composite (1) Gold SP500 Tbond 10 Y Oil Equity World High Yield Mar-07 Apr-07 May-07 Jun-07 Jul-07 Aug-07 Sep-07 Oct-07 Nov-07 Dec-07 Jan-08 Feb-08 (1) Source: Hedge Fund Research Inc. HFR Inc. 30 Sep Mar-08 Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Oct-08

5 HEDGE FUNDS? > But in the medium and the long terms, they remain by far the best risk-adjusted return on investment! Oct-98 Apr-99 Leveraged strategies from October-98 up to October-08 targeting a worst case per month of 4.0% HFRI FOF Composite (1) Gold SP500 Tbond 10 Y Oil Equity World High Yield Oct-99 Apr-00 Oct-00 Apr-01 Oct-01 Apr-02 Oct-02 Apr-03 Oct-03 Apr-04 Oct-04 (1) Source: Hedge Fund Research Inc. HFR Inc. 30 Sep Apr-05 Oct-05 Apr-06 Oct-06 Apr-07 Oct-07 Apr-08 Oct-08

6 HEDGE FUNDS? > A significant proportion of them had a POSITIVE performance in September/October > With the highest proportion of winners within Macro (HFR category, it includes CTA) 100% 90% 80% 70% 60% 50% 40% Winner, positive returns Follower, less than 11% Loser, more than 11% 30% 20% 0% Equity Hedge Event-Driven Fund of Funds Macro Relative Value

7 Will Hedge Funds still be there? > Despite disappointing performance during the crisis, Hedge Funds lost far less than Equity > In the long term, they have outperformed the best asset class during the period by a factor of 3 > And we had a large dispersion of performance in September/October, except for Fund of Funds > Fund of Funds performance raised the issue of Investors approach in term risk diversification

8 Are quant models working? Bad quants don t > Past Returns are not a very good predictor of future returns 20% 6% 6% 9% > To Invest only on the 20% top performers prior the crisis (right bubble on the chart) leads to a high proportion of top loser during the crisis Sept Oct 08 Perf 0% 16% 12% 7% - -20% 14% 21% -30% -40% -0.5% 0.0% 0.5% 1.0% 1.5% 2.0% Average Montly performance prior to the crisis

9 Even sophisticated quants can failed completely > More advanced performance based quant also failed > Enclose modified VaR (to account for fat tails) used as a predictor > To invest only on low Risk fund (left bubbles) lead to high proportion on losers, and only few winners Sept Oct 08 Perf 20% 14% 4% 3% 0% 13% 13% - 4% 14% -20% 26% -30% -40% 0% 2% 4% 6% 8% 12% 14% 16% 18% Modified VaR

10 What Failed is LOOKING BACKWARD! > Qualitative management of diversification also failed see FoF performance > But LOOKING FORWARD models work! > PROOF:

11 PROOF n 1: CTA & Macro Funds performances in September/October 100% 90% 80% 70% 60% 50% 40% 30% The one who are using Looking forward models Winner, positive returns Follower, less than 11% Loser, more than 11% 20% 0% Equity Hedge Event-Driven Fund of Funds Macro Relative Value

12 PROOF n. 2: A good Risk Model accurately predicted the impact of Sept / Oct > For all the 3,147 funds that reported their performance in October to HFR and are alive for at least 2 years > We compare actual and predicted performance based on our FOFIX models - according from data in August, > And obtain a very good match with a perfect prediction, with a very small proportion of errors! 20% 15% Error of Type 2: Predicted Loosed, Actual Winner 5% 7% 8% 5% Actual Performances 0% -5% - -15% -20% 16% 7% 14% 5% Error of Type 1: Predicted winner, Actual Loser -25% 36% 1% -30% -35% -40% -30% -20% - 0% 20% 30% 40% Predictions

13 Why good quant (nonlinear factor model) succeeded, while bad quant (performance based) failed? The Time Bomb effect 15% > Here we compare the Max Drawn down prior to the crisis with the Factor VaR (ie the tail betas) prior to the crisis 5% 0% 11% 5% 4% > A high ratio means that there are a lot of hidden risk: predicted risk is much higher than apparent risk. Low ratio means that apparent risk is smaller than actual risk. Sept Oct 08 Perf -5% - -15% 9% 11% 15% > Most of the big losers are among the one who hide their risks (high losers) while most of the winners are among the one who exhibit their risks -20% 4% 15% 25% -25% -30% 0.0% 50.0% 100.0% 150.0% 200.0% 250.0% 300.0% 350.0% 400.0% 450.0% Time Bomb Ratio (Factor Var/Max Draw Dow n)

14 Pure Return-Based Analysis Miss Hidden Risks janv-03 avr-03 > Credit driven fund: juil-03 oct-03 janv-04 avr-04 juil-04 oct-04 janv-05 avr-05 juil-05 oct-05 janv-06 avr-06 juil-06 Long AAA bonds, Short T-bonds, duration 10Y Sharpe = 1.3 Annualised Volatility = 2.4% Annualised return = 6.5% VaR 99 = 0.9% (1.3 sigma) Peak to valley = 1.1% Skew = +0.6 Excess Kurtosis = 0.2 No apparent risk: low vol, low peak to valley, positive skew oct-06 janv-07 avr-07 juil-07 oct-07 janv-08

15 Pure Return-Based Analysis Miss Hidden Risk janv-03 avr-03 > Credit driven fund: juil-03 oct-03 janv-04 avr-04 juil-04 oct-04 janv-05 avr-05 juil-05 oct-05 janv-06 Long AAA bonds, Short T-bonds, duration 10Y Ex post: huge loss, and very high risk Sharpe = Annualised Volatility = 3.4% Annualised return = 2.6% VaR 99 = 3.5% (3.5 sigma) Peak to valley = 12.2% Skew = -1.0 Excess Kurtosis = 3.0 avr-06 juil-06 oct-06 janv-07 avr-07 juil-07 oct-07 janv-08

16 Factor Analysis Doesn t Miss Hidden Risks > Credit driven fund vs. AAA spread over T-Bonds This fund was just surfing the good wave during the analysis period Will it last?

17 If It Is Nonlinear S&P500 Quarterly Returns vs Credit Spread BAA % 20% 15% S&P500 5% 0% Nonlinear Regression -5% - -15% -20% Credit Spread BAA

18 What Failed? What Works? > What failed: Looking backward: pure performance based data crunching Purely qualitative management of diversification Belief in Santa Claus, i.e. that excess return + liquidity + risk free is possible > What works for a New Deal Risk Transparency which relies on stable ex ante risk measures, i.e. measures encompassing change in market regime Extreme Risk diversification, based on nonlinear factor model Combine Quant and qualitative approaches

19 Risk Transparency Required Stability 20% 14% 0% 3% 4% Sept Oct 08 Perf - -20% 4% 13% 14% 13% -30% 26% -40% 0% 2% 4% 6% 8% 12% 14% 16% 18% Modifie d VaR > Beyond the fact that it does not handle time bombs properly > the other issue with sophisticated performance based models are their instability >..leading to a very high proportion of false warning (error of type 2, the blue bubble), and inability to spot the potential winners

20 Risk Transparency Required Stability 20% 11% 4% 5% 0% Sept Oct 08 Perf - -20% 3% 9% 15% 11% 31% -30% -40% 0% 5% 15% 20% 25% Factor VaR (Worst Beta--, Beta++) > While nonlinear factor based models are much more stable > minimizing false warning (5% instead of 14%) and helping spotting the winner (11% instead of 3%)

21 Stability to Manage EXTREME RISK DIVERSIFICATION > One tends to focus on business as usual diversification, while only extreme risk diversification really matters 25% 20% 15% S&P500 Quarterly Returns vs Credit Spread BAA > Managing extreme risk diversification requires paying attention to 2 phenomena: Change in correlation Time bomb issue, ie exceptional events S&P500 5% 0% -5% - -15% Nonlinear Regression -20% Credit Spread BAA > Only Nonlinear Factor modeling, using long-term factor history, can catch these phenomena

22 X Risk Diversification: Good vs. Bad Quant Rules Passive Risk Based Allocation 14%Funds predicted as Winner by FOFiX 20% Low Extreme Betas 9.50% 11.60% 20% Low Corning Fisher VaR 3.35% 20% Low Vo 3.60% 0.00% 2.00% 4.00% 6.00% 8.00% 10.00% 12.00% 14.00% Overperformance vs benchmark Looking backward bring small benefits, but looking forward very tangible ones

23 Active Risk Budgeting: Extreme Risk Diversification Benchmark (Equal Risk Allocation) Active extreme risk diversification From a random sample of 50 funds, run FOFIX portfolio builder in May 05 lead to a portfolio of 15 funds, to be implemented in July May 07: the portfolio builder is run again, to be implemented in July May 08: portfolio builder processed again Jul-05 Sep-05 In rally times, extreme risk diversified portfolio underperform the benchmark Nov-05 Jan-06 Mar-06 May-06 Jul-06 Sep-06 Nov-06 Jan-07 The cost of hedging risk in rally time is pay back xx times in shaky times Mar-07 May-07 Jul-07 Sep-07 Nov-07 Jan-08 Mar-08 May-08 Jul-08 Sep-08

24 LOOKING FORWARD, BOTH QUALITATIVELY AND QUANTITATIVELY > Are you going to eliminate all the funds that performed badly in September/October??? > Or are you going to look FORWARD > Example of Merger Arbitrage: They structurally become correlated with markets that are similar to the September/October markets In other words, they behaved like a short put out of money Qualitative analysis could tell that in the coming years a number of opportunities for Merger Arbitrage will arise Quantitative analysis helps discriminate between true alpha and theta i.e. to compare apples to apples Quantitative analysis helps to find ways to hedge short gamma

25 CONCLUSION > A tremendous amount of opportunities for the survivors > If they gain true transparency on their risks, going far beyond displaying Value-at-Risk > If Investors & Fund of Funds efficiently manage extreme risk diversification, using efficient nonlinear factor models, going back to the true value proposition of Alternative Investments > And stop looking backward!