Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Market Returns without Downside Risk Or The Difference Between Beta and the Equity Premium Max R. Arai, Ph.D., CFA Head of Portfolio Construction and Trading Acadian Asset Management, Inc. Q3 2007 Abstract Reproducing results from Haugen and Clarke et al, I find that minimum variance equity portfolios in the US and globally exhibit returns comparable to the returns of capitalization weighted portfolios. Following the standard textbook description of CAPM but using empirical data, one finds that the frontier corresponding to market clearing prices has been essentially flat; capitalization weighted portfolios (CWP) have added risk with little or no incremental return. In contrast, historical manager alpha forecasts lead to an efficient frontier dominating the market frontier. This demonstrates the inefficiency of the capitalization weighted portfolios. To gain insight into the inefficiency, I analyze the implied alpha of the CWP using a performance attribution system, and find that the style and stock bets in CWP have no efficacy. Using quintile portfolios to test these implied alphas confirms the lack of efficacy. I interpret these findings to demonstrate that “cheap beta” does not efficiently realize the equity premium. Asset allocation frameworks frequently assume that capitalization weighted portfolios are effective proxies for the equity premium. The use of minimum variance portfolios as a more efficient proxy can markedly increase the target allocation to equities, and improve the returns of the overall plan. Minimum variance portfolios represent a passive investment. Active alternatives further strengthen the case for portfolios managed relative to total risk. Various applications of leverage further improve on the expected returns. I conclude with a discussion of the possible limitations of minimum variance portfolios, and specifically discuss trading and capacity limitations for these strategies. Due to the presence of investors with dramatically varying levels of assets, these effects are likely to be persistent. Outline • Exploring a client request – Return characteristics of systematic risk factors – CAPM and Minimum Variance Portfolios: theory, history and simulations • • • • • Asset Allocation Active management relative to a cash benchmark Capacity limitations, persistence Why this is timely Concluding remarks Exploring a client request • A client requests a portfolio seeking equity marketlike returns with limited downside risk • Possible approaches: – Call options (downside risk is the premium) – Market Neutral (low absolute risk, high returns with good manager forecasts) – Long only equity but control (negative) fat tails • Consider long only fully invested equity – Client seeks to market the product globally Some initial ideas to limit downside risk • Possible ways to improve portfolio skewness – Take advantage of factor return skewness, esp. country and risk factors – Large emerging market exposure • Portfolio total risk scales the width of the probability distribution of returns – Low total volatility • Incorporate alpha – To pay transaction costs – To compensate for downside Controlling Fat Tails • Moments of a probability distribution ( µ n = E ( X − µ )n ) • Dimensionless moments ⎛⎛ X − µ ⎞ n ⎞ µˆ n = E ⎜⎜⎜ ⎟ ⎟⎟ ⎝⎝ σ ⎠ ⎠ – Standard deviation σ sets scale of moments – Skewness proportional to σ3 – Kurtosis proportional to σ4 • To control fat tails, can we control σ? • Recall Chebyshev: P(|X-µ|>kσ) < k-2 Non-parametric Factor returns - US model • Measure using performance attribution for unit exposure assets over entire model history (198611-200706) • Factor skewness of both signs >0: Growth, Earnings Yield, Volatility… <0: Value, Size, Momentum,… • All factors have excess kurtosis Factor returns – US model Descriptive Statistics for UB model risk factors 8.00 7.00 6.00 5.00 stdev 4.00 mean skewness 3.00 kurtosis 2.00 1.00 0.00 Source: underlying data from MSCI Barra UB_YLD UB_VOL UB_VAL UB_TRA UB_SNL UB_SIZ UB_NEU UB_MOM UB_LEV UB_GRO UB_EYL UB_EVR UB_CUR -1.00 Factor returns – US model QQ plots for Ub model risk factor payoffs 6.00 UB_CUR 4.00 UB_EVR UB_EYL 2.00 UB_GRO UB_LEV -4.00 -3.00 -2.00 -1.00 0.00 0.00 -2.00 UB_MOM 1.00 2.00 3.00 4.00 UB_NEU UB_SIZ UB_SNL -4.00 UB_TRA UB_VAL -6.00 UB_VOL UB_YLD -8.00 Source: underlying data from MSCI Barra Factor returns - Global model • Measure using performance attribution for unit exposure assets over entire model history (198802-200705) • Factor skewness of both signs >0: Growth, Earnings Yield, Volatility… <0: Value, Size, Momentum,… • All factors have excess kurtosis MS__SIZ MS__VAL MS__AEROSP MS__AIRTRN MS__APPLNC MS__AUTO MS__BANKS MS__BEVTOB MS__BLDG MS__BRDCST MS__BUSSVC MS__CHEM MS__CONSTR MS__DATAPR MS__ELCTEQ MS__ELECMP MS__ENERGY MS__ENRGEQ MS__FINANC MS__FOODHS MS__GOLD MS__HEALTH MS__INSURE MS__LEISUR MS__MACHIN MS__MATER MS__MRCHND MS__MULTI MS__NFMET MS__PAPER MS__REALES MS__RECCON MS__RRTRAN MS__SCS MS__SHPTRN MS__SPR MS__STEEL MS__TELECM 0.00 -5.00 MS__SIZ MS__VAL MS__AEROSP MS__AIRTRN MS__APPLNC MS__AUTO MS__BANKS MS__BEVTOB MS__BLDG MS__BRDCST MS__BUSSVC MS__CHEM MS__CONSTR MS__DATAPR MS__ELCTEQ MS__ELECMP MS__ENERGY MS__ENRGEQ MS__FINANC MS__FOODHS MS__GOLD MS__HEALTH MS__INSURE MS__LEISUR MS__MACHIN MS__MATER MS__MRCHND MS__MULTI MS__NFMET MS__PAPER MS__REALES MS__RECCON MS__RRTRAN MS__SCS MS__SHPTRN MS__SPR MS__STEEL MS__TELECM MS__TEXTL MS__UTIL MS__AUSMKT MS__AUTMKT MS__BELMKT MS__CANMKT MS__DENMKT MS__FINMKT MS__FRAMKT MS__GERMKT MS__GREMKT MS__HKGMKT MS__IREMKT MS__ITAMKT MS__JPNMKT MS__NETMKT MS__NORMKT MS__NZEMKT MS__PORMKT MS__SINMKT MS__SPAMKT MS__SWEMKT MS__SWIMKT MS__UKIMKT MS__USAMKT MS__ARGMKT MS__BHRMKT MS__BRAMKT MS__CHIMKT MS__CHNMKT MS__COLMKT MS__CZEMKT MS__EGYMKT MS__HUNMKT MS__IDNMKT MS__INDCMP MS__ISRMKT MS__JORMKT MS__KORMKT MS__MALMKT MS__MEXMKT MS__MORMKT MS__OMNMKT MS__PAKMKT MS__PERMKT MS__PHIMKT MS__POLMKT MS__RUSMKT MS__SAFMKT MS__SLVMKT MS__SRIMKT MS__TAIMKT MS__THAMKT MS__TURMKT Factor Returns – Global Model 25.00 250 20.00 200 15.00 150 mean 10.00 100 5.00 9.00 8.00 7.00 6.00 5.00 4.00 3.00 2.00 1.00 0.00 -1.00 -2.00 100 50 0 Source: underlying data from MSCI Barra stdev skewness excess kurtosis 50 count 0 250 200 mean 150 stdev skewness excess kurtosis count Hard to Predict Higher Moments • Stationarity of factor returns suspect – Are any country returns actually thin tailed? – Time dependent coverage • Stock returns often have non–stationary distributions - true even at the index level Higher Moments for Some Sample Indexes Skewness - 3 year rolling window monthly returns 1.5 1 0.5 -1 -1.5 Source: underlying data from MSCI Barra, JP Morgan, Acadian simulated portfolios 6/30/07 4/30/07 2/28/07 12/30/06 8/30/06 10/30/06 6/30/06 4/30/06 2/28/06 12/30/05 8/30/05 10/30/05 6/30/05 4/30/05 2/28/05 12/30/04 8/30/04 -0.5 10/30/04 6/30/04 4/30/04 2/29/04 12/30/03 8/30/03 10/30/03 6/30/03 4/30/03 2/28/03 12/30/02 8/30/02 10/30/02 6/30/02 4/30/02 2/28/02 12/30/01 8/30/01 10/30/01 6/30/01 4/30/01 2/28/01 12/30/00 8/30/00 10/30/00 6/30/00 0 Active Absolute JPM-1-3Yr Bond Index JPM-10yr Bond Index Passive Absolute MSCI Australia MSCI Europe MSCI Japan MSCI World MSCI US MSCI EAFE Small MSCI EMF MSCI GCC Higher Moments for Some Sample Indexes Kurtosis - 3 year rolling window monthly returns 3 2 1 -1 -2 -3 -4 Source: underlying data from MSCI Barra, JP Morgan, Acadian simulated portfolios 6/30/07 4/30/07 2/28/07 12/30/06 8/30/06 10/30/06 6/30/06 4/30/06 2/28/06 12/30/05 8/30/05 10/30/05 6/30/05 4/30/05 2/28/05 12/30/04 8/30/04 10/30/04 6/30/04 4/30/04 2/29/04 12/30/03 8/30/03 10/30/03 6/30/03 4/30/03 2/28/03 12/30/02 8/30/02 10/30/02 6/30/02 4/30/02 2/28/02 12/30/01 8/30/01 10/30/01 6/30/01 4/30/01 2/28/01 12/30/00 8/30/00 10/30/00 6/30/00 0 Active Absolute JPM-1-3Yr Bond Index JPM-10yr Bond Index Passive Absolute MSCI Australia MSCI Europe MSCI Japan MSCI World MSCI US MSCI EAFE Small MSCI EMF MSCI GCC Review of CAPM Return Capital market line Risk Free Rate Maximum Sharpe Ratio portfolio Minimum variance portfolio Risk Minimum Variance Portfolios in the context of CAPM • CAPM makes heroic assumptions (equilibrium, fully informed investors who agree, no taxes, borrowing and lending at the same risk-free rate, …) • Under CAPM, the Maximum Sharpe ratio portfolio is the market portfolio. • CAPM has an efficient frontier (relative to the risk free asset) • Assuming CAPM, minimum variance portfolio has lower return and Sharpe ratio than the market portfolio Empirical cash relative frontier Return Capital market line Maximum Sharpe Ratio portfolio Cap weighted portfolio Risk Free Rate Minimum variance portfolio Risk Source: illustration of Acadian simulations Market consensus efficient frontier Minimum Variance Portfolios (MVP) • Clarke et al. “Minimum-Variance Portfolios in the U.S. Equity Market” , Journal of Portfolio Management, Fall 2006 • Haugen and Baker “The Efficient Market Inefficiency Of Capitalization- Weighted Stock Portfolios”, Journal of Portfolio Management Spring 1991 • Acadian simulations confirm Clarke's findings for US, Australian and global portfolios. Global results illustrate the benefits of currency hedging for long term results. Performance attribution for cap weighted portfolios • Use minimum variance portfolio as the benchmark • Neither style nor asset selection added value • The benchmark to assess the cap weighted portfolio (CWP) and active manager portfolios should be the minimum variance portfolio. • Manager frontier and market consensus frontier can differ, and can reveal manager skill (using CWP can obscure manager skill). Example performance attribution ATTRIBUTION REPORT Annualized Contributions To Total Return Managed vs minimum variance MSCI World vs minimum variance Managed vs MSCI World Source Contribution Risk Contribution Risk Contribution Risk of Return (% Return) (% Std Dev) (% Return) (% Std Dev) (% Return) (% Std Dev) 1 Risk Free 3.59 N/A 3.59 N/A 3.59 N/A 2 Total Benchmark 8.83 8.80 8.83 8.80 7.07 14.41 3 Country Selection -0.52 1.94 -0.90 1.60 0.40 2.48 4 Currency Selection 0.60 1.21 0.69 1.39 -0.10 1.21 5 Cash-Equity Policy -0.07 0.06 -0.09 0.11 0.00 0.00 6 Asset Allocation [3+4+5] 0.01 2.12 -0.30 2.06 0.31 2.71 7 Local Market Timing 0.53 1.84 1.70 8.14 -2.02 7.70 8 Risk Indices 1.42 1.21 -1.29 2.10 2.60 2.74 9 Industries 0.25 1.20 -0.09 1.70 0.54 2.01 10 Asset Selection 2.69 2.26 -1.72 1.87 5.15 1.93 11 Within Market [7+8+9+10] 4.88 3.30 -1.40 9.35 6.27 9.53 12 Trading 0.00 0.40 N/A N/A 0.00 0.40 -1.24 N/A N/A N/A -1.23 N/A 3.62 3.42 -1.76 9.41 5.38 9.45 12.45 10.15 7.07 14.41 12.45 10.15 13 Transaction Cost 14 Total Active [6+11+12+13] 15 Total Managed [2+14] Source: Acadian simulated performance Efficacy of implied alpha • Extract implied alpha for a cap weighted benchmark • Use implied alphas to construct quintile portfolios to measure spreads • The market consensus offers negligible forecast value (no free lunch). Beta ≠ Equity Premium • Minimum variance portfolio is fully invested and captures the equity premium • Beta relative to the cap weighted portfolio adds uncompensated risk • Better market proxy (missing assets)? – Capitalization would have to be enormous to change the result Simplified Asset Allocation Framework • Select some investible proxies for asset classes of interest – JPMorgan 1-3 year bond index – JPMorgan 10 year bond index – MSCI indexes (World, AU, JP, US, Europe, EAFE Small cap, EMF) – Passive and active absolute return portfolios Simplified Asset Allocation Framework (2) • Assume long only allocations • Use historical covariance • Optimize allocations using historical returns and forecast returns – Historical data provides a reference to place the covariance data in context – Forecast data seeks to avoid the rear view mirror Asset Allocation Methods • Compute covariance from monthly returns (in Excel) • Use Solver to optimize mean variance returns at varying levels of risk aversion – – – – U = ∑ R w − λ∑ w Γ w U is utility wk is weight of kth asset Rk is return of kth asset Γkk’ is covariance of kth and k’th assets k k k k kk' k,k' k' JPM-1-3Yr Bond Index JPM-10yr Bond Index MSCI EMF MSCI Japan MSCI Australia MSCI US MSCI Europe MSCI EAFE Small MSCI World Passive Absolute Active Absolute 100% 71% -32% -16% -18% -30% -30% -24% -32% -16% -19% 71% 100% -23% -13% -12% -22% -21% -7% -23% -6% -6% Historical Returns Historical Risk Forecast Returns 4.5% 1.7% 5% 7.7% 16.9% 2.9% 11.3% 6.2% 9.2% 13.7% 6.3% 8.7% 21.9% 20.2% 18.4% 15.8% 16.7% 15.8% 15.0% 6% 14% 6% 10% 8% 9% 14% 9% -32% -23% 100% 55% 73% 70% 70% 77% 77% 58% 55% -16% -13% 55% 100% 60% 47% 45% 73% 61% 65% 61% Source: underlying data from MSCI Barra, JP Morgan, Acadian simulated portfolios -18% -12% 73% 60% 100% 64% 67% 75% 74% 71% 66% -30% -22% 70% 47% 64% 100% 81% 58% 96% 70% 70% -30% -21% 70% 45% 67% 81% 100% 68% 92% 75% 70% -24% -7% 77% 73% 75% 58% 68% 100% 71% 77% 72% -32% -23% 77% 61% 74% 96% 92% 71% 100% 80% 77% -16% -6% 58% 65% 71% 70% 75% 77% 80% 100% 95% Active Absolute Passive Absolute MSCI World MSCI EAFE Small MSCI Europe MSCI US MSCI Australia MSCI Japan MSCI EMF JPM-10yr Bond Index JPM-1-3Yr Bond Index Historical Correlations -19% -6% 55% 61% 66% 70% 70% 72% 77% 95% 100% 6.4% 11.4% 9.4% 11.0% 9% 11% Allocation Results using Forecast Returns Three Asset Control Study Four Asset Allocation: Two Bonds, MSCI World, Passive Absolute Strategy Dominance of Active Absolute Strategy 100% 100% 100% 80% 80% 80% 60% 60% 60% MSCI World JPM-10yr Bond Index JPM-1-3Yr Bond Index 40% 20% MSCI World Passive Absolute JPM-10yr Bond Index JPM-1-3Yr Bond Index 40% 20% 0% 20% 0% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 0% 1 -20% MSCI World Passive Absolute Active Absolute JPM-10yr Bond Index JPM-1-3Yr Bond Index 40% 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 1 -20% 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 -20% Three Asset Control Study Four Asset Allocation: Two Bonds, MSCI World, Passive Absolute Strategy Dominance of Active Absolute Strategy 12.00% 120 12.00% 120 12.00% 120 10.00% 100 10.00% 100 10.00% 100 8.00% 8.00% 80 6.00% 60 4.00% 2.00% 0.00% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 expected return expected risk utility lambda 80 expected return expected risk utility lambda 8.00% 80 6.00% 60 6.00% 60 40 4.00% 40 4.00% 40 20 2.00% 20 2.00% 20 0 0.00% 0 0.00% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Source: underlying data from MSCI Barra, JP Morgan, Acadian simulated portfolios 25 26 27 28 29 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 expected return expected risk utility lambda Allocation Results using Historical Returns Three Asset Control Study Four Asset Allocation: Two Bonds, MSCI World, Passive Absolute Strategy Five Asset Allocation: Two Bonds, MSCI World, Active and Passive Absolute Strategy 100% 100% 100% 80% 80% 80% 60% 60% MSCI World JPM-10yr Bond Index JPM-1-3Yr Bond Index 40% 20% 60% MSCI World Passive Absolute JPM-10yr Bond Index JPM-1-3Yr Bond Index 40% 20% 0% 20% 0% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 0% 1 -20% MSCI World Passive Absolute Active Absolute JPM-10yr Bond Index JPM-1-3Yr Bond Index 40% 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 1 -20% 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 -20% Three Asset Control Study Four Asset Allocation: Two Bonds, MSCI World, Passive Absolute Strategy Five Asset Allocation: Two Bonds, MSCI World, Active and Passive Absolute Strategy 12.00% 120 12.00% 120 12.00% 120 10.00% 100 10.00% 100 10.00% 100 8.00% 80 6.00% 60 4.00% 2.00% 0.00% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 8.00% 80 6.00% 60 40 4.00% 20 2.00% 0 0.00% expected return expected risk utility lambda 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Source: underlying data from MSCI Barra, JP Morgan, Acadian simulated portfolios 22 23 24 25 26 27 28 29 8.00% 80 6.00% 60 40 4.00% 40 20 2.00% 20 0 0.00% expected return expected risk utility lambda 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 expected return expected risk utility lambda Impact of Minimum Variance Portfolios on Asset Allocation • The addition of minimum variance portfolios to the standard asset allocation framework increases the equity contribution to both risk and return • Domination of the cap weighted equity proxy by absolute return portfolios highlights the difference between beta and the equity premium • Material improvements in overall plan returns possible Active Management Relative to a Cash Benchmark • Describe capacity limitations • Describe turnover • Describe the noise trader argument for likely persistence of the effects • Describe how the persistence of beta versus equity premium relates to long term market disequilibrium Limitations of minimum variance portfolios • Capacity limitations As assets become large, cap weights are the only investment available • High tracking error when viewed from a conventional viewpoint • Relatively high turnover (typically over 10% per month) Capacity considerations • Underlying experiments need to be completed Persistence of capitalization weighted benchmark portfolios • Could be due to noise trading – Advances in Behavioral Finance, Thaler Ch.2 – Cap weighting easy to implement – Huge percentage of market assets managed to cap weighted benchmark – Material risk for arbitrageurs Role of shorting in active management relative to a cash benchmark Annualized Contributions To Total Return Source of Return US strategy Russell 1000 Russell 3000 Minimum Variance long only 60bp limit long only 100bp limit 130/30 100bp limit 1 Risk Free 3.53 3.53 3.53 3.53 3.53 3.53 2 Total Benchmark 7.14 7.14 7.14 7.14 7.14 7.14 3 Expected Active -0.02 0.00 -3.47 -2.94 -3.10 -3.85 4 Market Timing -0.05 0.00 2.22 2.29 2.37 2.44 5 Risk Indices -0.32 0.00 3.55 4.01 4.00 5.31 6 Industries 0.00 0.00 0.21 -0.29 -0.31 -0.35 7 Asset Selection 0.30 0.00 0.83 2.01 2.67 3.77 8 Trading N/A N/A N/A N/A N/A N/A 9 Transaction Cost N/A N/A N/A N/A N/A N/A 10 Total Exceptional Active[4+...+9] -0.07 0.00 6.81 8.02 8.73 11.17 11 Total Active [3+10] -0.09 0.00 3.34 5.09 5.63 7.32 7.05 7.14 10.48 12.23 12.77 14.45 12 Total Managed [2+11] Source: underlying data from MSCI Barra, JP Morgan, Acadian simulated portfolios Why is this timely • 20 years ago, even US transactions costs were high, ~100bp • Commercial risk models available since 1980’s, but supporting tools for portfolio construction and trading both expensive and weak • Today, direct market access available in most developed markets • The bloom is off the rose for cap weighted benchmarks; active investing well established Outlook and conclusions • Absolute return portfolios may be exiting the pioneering phase and moving into rapid growth • Increased attention to liability driven investment (LDI) and stability of pension plan funding status may lead to demand for more efficient investments Legal Disclaimer • The presentation material was prepared by Acadian Asset Management, Inc. with information it believes to be reliable and contains Acadian’s confidential and proprietary information. The views expressed in the presentation and presentation material are those of Acadian and are subject to change with market conditions. This presentation is for informational purposes only and should not be construed as investment advice, or an offer to sell or a solicitation of an offer to buy any security. • The presentation material has not been updated since it was published and may not reflect the current views of the author(s) or recent market activity. Market conditions are subject to change. Historical economic and performance information is not indicative of future results. • This document may not be reproduced or disseminated in whole or part without the prior written consent of Acadian Asset Management, Inc. © Acadian Asset Management, Inc. 2007. All rights reserved.