April 2015 Momentum Strategies Across Asset Classes Risk Factor Approach to Trend Following Quantitative and Derivatives Strategy Marko Kolanovic, PhDAC (Global) [email protected] Zhen Wei, PhD (Asia) [email protected] See page 183 for analyst certification and important disclosures, including non-US analyst disclosures. Marko Kolanovic Systematic Cross-Asset Strategy (1-212) 272-1438 15 April 2015 [email protected] Zhen Wei, CFA (852) 2800-7749 [email protected] 2 Marko Kolanovic Systematic Cross-Asset Strategy (1-212) 272-1438 15 April 2015 [email protected] Zhen Wei, CFA (852) 2800-7749 [email protected] April 2015 Dear Investor, In the current environment of low yields and high valuations, interest in Risk Premia strategies continues to grow. In our primer on Systematic Strategies, we developed a framework for Risk Factor investing across assets, and in the subsequent report we provided a detailed overview of Equity Risk Premia Strategies. In this report we focus on Momentum Strategies Across Assets. Momentum Strategies tend to have positive performance in rising markets and can also outperform traditional assets during market corrections. Low or negative correlation during market corrections has been an attractive feature of investing in Momentum Strategies and CTA funds. Momentum Strategies are not without risks – they can suffer during market turning points such as sharp market recoveries and can also underperform in mean reverting markets. As a result, Momentum Strategies tend to have higher Sharpe ratios than traditional assets, but also higher tail risk and negative skewness – a common feature of Risk Premia strategies. In the first two chapters of the report, we designed and studied prototype Momentum factors across assets. As the virtually unlimited number of possible implementations may confound an investor, we first provide a framework for designing and testing Momentum Strategies. We have examined single asset and multi asset strategies, Absolute and Relative Momentum, various Momentum filters, lookback windows, rebalancing frequencies and investment horizons. The third chapter of this report is dedicated to risk management techniques such as stop-loss, mean reversion signals and diversification across assets and signals. In the last chapter we analyze Seasonality strategies, where we treat Seasonality as Momentum of non- consecutive asset returns. Marko Kolanovic, PhD Global Head of Quantitative and Derivatives Strategy J.P. Morgan Securities LLC 3 Marko Kolanovic Systematic Cross-Asset Strategy (1-212) 272-1438 15 April 2015 [email protected] Zhen Wei, CFA (852) 2800-7749 [email protected] 4 Marko Kolanovic Systematic Cross-Asset Strategy (1-212) 272-1438 15 April 2015 [email protected] Zhen Wei, CFA (852) 2800-7749 [email protected] Contents Introduction to Momentum Investing ................................... 7 Overview of Trend-Following Strategies .................................................................. 9 Risks of Momentum Strategies ............................................................................... 12 CTAs and Momentum Strategies ............................................................................ 14 Prototype Momentum Factors ............................................. 17 Absolute Momentum Prototypes ............................................................................. 20 Relative Momentum Strategies ............................................................................... 25 Correlation of Momentum Strategies ...................................................................... 30 Selection of Trend Signal ........................................................................................ 36 Investment Horizon, Rebalance Frequency and Transaction Costs......................... 42 Dynamically Rebalanced Signals ............................................................................ 46 CTA Exposure to Prototype Momentum Factors .................................................... 48 Risk Management and Portfolio Construction ................... 53 Stop-Loss and Volatility Signals ............................................................................. 56 Incorporating Value/Reversion Factors ................................................................... 62 Risk Adjusted Momentum ....................................................................................... 66 Long Only Momentum ............................................................................................ 69 Risk Methods: J.P. Morgan Mozaic and Efficiente ................................................. 71 Diversified Trend-Following Strategies .................................................................. 73 Other Potential Enhancements ................................................................................ 79 Seasonality............................................................................ 83 Introduction to Seasonality ...................................................................................... 85 Quantifying Seasonality Across Assets ................................................................... 89 Prototype Seasonality Risk Factors ......................................................................... 98 Seasonality and Momentum Factor Portfolios ...................................................... 102 Appendices ......................................................................... 108 J.P. Morgan Investable Momentum Indices .......................................................... 110 J.P. Morgan Research on Momentum Strategies ................................................... 113 An Introduction to Commodity Trading Advisors (CTAs) ................................... 124 Mathematics of Trend Filtering Methods .............................................................. 136 CTA Exposure to Fung and Hsieh Factors ............................................................ 139 Rebalancing and Investment Horizons .................................................................. 140 Transaction Cost Analysis ..................................................................................... 146 Performance of Alternative Trend Signals ............................................................ 150 Stop-Loss Trigger and Short-Term Reversion Sensitivities .................................. 153 Macro and Market Regimes .................................................................................. 161 Performance-Risk Analytics .................................................................................. 163 Review of Portfolio Construction Methods ........................................................... 166 Academic References ............................................................................................ 171 Glossary ................................................................................................................. 177 Contacts and Disclaimers ...................................................................................... 181 5 Marko Kolanovic Systematic Cross-Asset Strategy (1-212) 272-1438 15 April 2015 [email protected] Zhen Wei, CFA (852) 2800-7749 [email protected] 6 Marko Kolanovic Systematic Cross-Asset Strategy (1-212) 272-1438 15 April 2015 [email protected] Zhen Wei, CFA (852) 2800-7749 [email protected] Introduction to Momentum Investing 7 Marko Kolanovic Systematic Cross-Asset Strategy (1-212) 272-1438 15 April 2015 [email protected] Zhen Wei, CFA (852) 2800-7749 [email protected] 8 Marko Kolanovic Systematic Cross-Asset Strategy (1-212) 272-1438 15 April 2015 [email protected] Zhen Wei, CFA (852) 2800-7749 [email protected] Overview of Trend-Following Strategies The existence of a Price Momentum effect implies that the price of an asset exhibits trends as opposed to being randomly distributed. A trending price means simply that an asset that recently appreciated is more likely to continue moving higher, and an asset that recently declined is more likely to continue moving lower. The existence of Momentum effects would violate the efficient markets hypothesis (which states that past prices cannot predict future performance) and enable Momentum traders to consistently outperform the broad market. The concept of Momentum trading could be traced back at least to the 18th century, and well elaborated Trend-Following strategies have been known for close to 100 years1. A host of academic literature describes and gives potential explanations for the Momentum effect. They include increased loading of high Momentum assets to systematic risk2, inefficiencies in investor behavior (over/under-reaction to news or investor herding)3, macroeconomic supply and demand frictions, positive feedback loops between risk assets and economic growth (e.g. strong equity markets can create wealth effects that boost consumer spending, and in turn corporate earnings and equities), and even in the market microstructure. Interested readers can refer to our primer report on Cross Asset Systematic Strategies and Equity Risk Premia Strategies for more discussions and analysis of Momentum premium across different asset classes Momentum in Equities is a well researched topic. One of the early papers to document equity Momentum was published by Jegadeesh, and Titman (1993). Examples of various equity Momentum improvements can be found in our report Investment Strategies no 89: Equity Momentum. The existence of Momentum in Commodities is well known to CTA practitioners over the past 30 years. A detailed review of commodity Momentum strategies can be found in our report: Investment Strategies No. 25: Momentum in Commodities as well as in reports by Erb and Harvey (2006), Miffre and Rallis (2007) and Fuertes, Miffre, and Rallis (2010). Momentum effects have also been documented in the Fixed Income space. For instance, our report on Momentum in German government bonds (Investment Strategies No. 27: Euro Fixed Income Momentum Strategy) demonstrates a strong Momentum signal with a 2-3 week time horizon. The existence of fixed income Momentum across global bond markets was also shown in the work of Asness, Moskowitz, Pedersen (2011). The Momentum effect in Currency Markets was tested and demonstrated for example in the research of Okunev and White (2003), Burnside, Eichenbaum, and Rebelo (2011) and Menkhoff, Sarno, Schmeling, and Schrimpf (2012). Interested readers can refer to the Appendix on page 171 for a list of academic literature on the Momentum risk premium. In our previous report on Cross Asset Systematic Strategies, we examined Momentum effects across assets and showed the performance of simple Momentum strategies over the past 40 years. Our prototype Momentum factors were monthly rebalanced portfolios that went long assets with the highest, and sold short assets with the lowest 12-month price returns.4 1 David Ricardo (1772-1823) quoted “Cut short your losses; let your profits run on”. Further, see William Dunnigan’s ‘Trading With the Trend’ in 1934, and Richard Donchian’s ‘Trend-Following Methods in Commodity Price Analysis’ in the Commodity Yearbook of 1957. 2 Liu and Zhang (2008) provide evidence that high Momentum stocks have excess exposure to macro growth risk. 3 See, for example, Barberis, Sheleifer and Vishny (1998), Hong, Lim and Stein (2000) and Dasgupta, Prat and Verardo (2011) 4 These non-tradable prototype Momentum Risk Factors are constructed as follows (they are constructed to illustrate long-term risk properties and we later consider more liquid contracts in tradable versions of prototype Momentum Risk Factors in Chapters 2-4): Equity Momentum Factor: Excess return of a long position in three equity indices with the highest past 12-month returns and a short position in the three equity indices with lowest past 12-month returns (monthly rebalanced). Our index universe of country equity benchmarks was: Australia, Canada, France, Germany, HK, Italy, Japan, Netherlands, Spain, Sweden, Switzerland, the UK, and the US. Bond Momentum Factor: Excess return of a long position in the three 10-year government bonds with the highest past 12-month returns and a short position in the three 10-year government bonds with lowest past 12-month returns (monthly rebalanced). The universe was comprised of government bonds from Australia, Belgium, Canada, Germany, Denmark, Japan, Sweden, the UK, and the US. Currency Momentum Factor: Excess return of a long position in the three G10 currencies with the highest past 12-month returns and a short position in the three G10 currencies with lowest past 12-month returns (monthly rebalanced). Commodity Momentum Factor: Excess return of a long position in three commodity futures with the highest 12-month returns and a short position in three futures with lowest 12-month returns. The commodity futures universe: Brent and WTI oil, Heating Oil, Gasoil, Gasoline, Natural Gas, Gold, Silver, Cocoa, Coffee, Cotton, Feeder Cattle, Wheat, Lean Hogs, Live Cattle, Soybeans, Sugar, and Wheat. 9 Marko Kolanovic Systematic Cross-Asset Strategy (1-212) 272-1438 15 April 2015 [email protected] Zhen Wei, CFA (852) 2800-7749 [email protected] Even these simple prototype models have delivered positive long-run risk premia across major asset classes as shown in Figure 1. The summary of performance statistics in Table 1 shows that Momentum strategies in various asset classes delivered Sharpe ratios in a 0.2 to 0.6 r ange. With the benefit of cross asset diversification, an Equal Marginal Volatility (EMV) weighted portfolio of prototype Momentum Risk Factors generated a Sharpe ratio of 0.78 during 1972-2014. 5 These solid Sharpe ratios were at least in part a compensation for the negative skewness and elevated tail risk (positive excess kurtosis) Momentum strategies delivered. Figure 1: Performance of Prototype Momentum Risk Factors by Asset Table 1: Performance and Risk Statistics for Prototype Momentum Class Factors 1200 Equity Bond Equity Bond Curncy Comdty EMV Currency Commodity 1000 EMV Portfolio Excess Ret (%) 4.6 3.5 1.7 7.3 4.1 STDev (%) 17.7 6.2 9.0 15.1 5.3 800 MaxDD (%) -37.5 -19.1 -27.9 -33.3 -13.2 600 MaxDDur (yrs) 16.4 6.0 15.2 5.7 6.0 t-Statistic 2.2 3.9 1.5 3.6 5.2 400 Sharpe Ratio 0.26 0.57 0.18 0.49 0.78 200 Hit Rate (%) 51.7 63.4 56.8 56.2 63.8 Skewness 0.51 -0.10 -0.33 0.01 -0.32 0 Kurtosis 6.17 3.67 1.87 1.45 1.13 72 75 78 81 84 87 90 93 96 99 02 05 08 11 14 Source: J.P. Morgan Quantitative and Derivatives Strategy. Source: J.P. Morgan Quantitative and Derivatives Strategy. * For comparison purpose, Equity, Bond, Currency and Commodity prototype Momentum strategies are scaled to have an ex-post volatility of 10% per annum in the chart. Similar to our tests, Ooi, Moscowitz, and Pedersen (2011) documented Momentum effects in global equity index, currency, commodity and bond futures markets since the 1970s. Based on extended datasets, Hurst, Ooi, and Pedersen (2012) validated significant Momentum effects across assets since 1903, while Lemperiere, Derenble, Seager, Potters and Bouchaud (2014) did a similar exercise for Equity index and commodity markets since 1800. We have even reviewed 800- year backtests reported in a book ‘Trend-Following with Managed Futures’ by A. Greyserman and K. Kaminski that suggests a long-term Sharpe ratio of 1.16 for trend following strategies. However, we do caution against relying on very long backtests (e.g. more than ~50 years) as they often underestimate risks and transaction costs and hence may produce misleading results (see risk section below). In this report we further expand our analysis of Momentum effects and develop a framework for systematic Trend- Following strategies. We also address the question of risks embedded in Momentum factors by quantifying these risks and analyzing Momentum strategies in the context of Risk Factor investing. An investor looking to design a Momentum strategy may find the large number of possible implementations overwhelming. For instance, one needs to choose the asset universe, type of Momentum signal, signal lookback window, rebalance frequency, and investment horizon, even before considering various risk management methods such as stop loss or volatility targeting. A large number of possible parameters in a Momentum strategy may naturally raise concerns of over-fitting and in-sample choice of parameters. In this report, we define a framework for designing a Momentum strategy and test a broad range of implementations. Figure 2 below summarizes various elements of a Trend-Following strategy we address in the rest of the report. First, one needs to decide if the strategy will be trading instruments from one or multiple asset classes (i.e. specify the asset 5 At each month-end portfolio rebalancing, each Momentum Risk Factor is weighted inversely with respect to its past 24-month volatility. See Chapter 3 of our report on Cross Asset Systematic Strategies for more details on portfolio allocation methodologies and their applications. Also see the Appendix on page 166 for a brief summary of these methods. 10