Idiosyncratic Volatility Strategies in Commodity

Transcription

1 1

2 Idiosyncratic Volatility Strategies in Commodity Futures resmarkets Joëlle Miffre Professor of Finance, EDHEC Business School Member of EDHEC Risk Institute Joint work with Adrian Fernandez Perez (ULPGC) and Ana Maria Fuertes (Cass Business School)

3 Idiosyncratic Volatility Strategies in Commodity Futures Markets Single sort idiosyncratic volatility strategy Portfolio construction Benchmark definition Performance evaluation Triple sort strategy combining idiosyncratic volatility, momentum and term structure signals Can the idiosyncratic volatility portfolios be used for risk diversification and inflation hedging? Robustness checks Transaction costs Liquidity 3

4 Portfolio constructionction Active strategies based on idiosyncratic i volatility (hereafter, IV) have been shown to be profitable in equity markets, where stocks with high past levels of idiosyncratic volatility tend to exhibit low future average returns (Ang, Hodrick, Xing and Zhang, 2006, 2009) Measuring the idiosyncratic volatility of stocks is straightforward: Standard ddeviation of the residuals from the 3 factor model dlof Fama and French (1993) and the 4 factor model of Carhart (1997) R St ( RMt R ft ) + β2smbt + β HMLt ε St R ft = 0 + β1 3 β + Normal excess return defined by 3-factor model Time t idiosyncratic return IV S = σ ( ε ) St 4

5 Portfolio construction These benchmarks cannot be readily dl applied to commodities Erb and Harvey (2006): Low correlation between commodities and traditional risk factors S&P GSCI is sub optimal We use hdi hedging pressure benchmarks of Basu and Miffre (2012) based on positions of hedgers and speculators R Ct = 0 1 ( Backwardationt Contangot ) ε Ct β + β + Normal excess return Idiosyncratic return IV = σ ε C ( ) Ct 5

6 Hedging pressure benchmarks Futures price F t, T > FT, T F t, T < FT, T Contango Net long hedgers / Net short speculators Backwardation Net short hedgers / Net long speculators OI ( LongHedgers) HPHedgers = OI ( Long Hedgers + Short Hedgers ) OI F T, ( LongSpec) T HPSpec = OI ( Long + Short ) g Spec Spec t Maturity = T Time 6

7 Hedging pressure benchmarks Ranking period (R): 4, 13, 26 and 52 weeks Holding period (H): 4, 13, 26 and 52 weeks HP Hedgers HP Speculators Up to 27 commodities Low HP Hedg (50%) Net short hedgers High HP Spec (40%) Net long speculators Low HP Spec (60%) High HP Spec (60%) Buy Backwardation (20%) Ignore remaining 60% High HP Hedg (50%) Net long hedgers Low HP Spec (40%) Net short speculators Sell Contango (20%) 7

8 Idiosyncratic volatility portfolios Ranking period (R) over which IV is modelled: 52 weeks Holding period (H): 4, 13, 26 and 52 weeks Up to 27 commodities Low IV (20%) High IV (20%) Buy low IV Sell high IV Ignore remaining 60% R Ct = 0 1 ( Backwt Contangot ) ε Ct = σ ( ε ) β + β + IVC Ct 8

9 Portfolio construction Strategies implemented on a cross section of 27 (agricultural, l energy, livestock and metal) commodity futures using front contracts Sample: Sept 1992 March 2011 The constituents are equally weighted The long and the short portfolios are fully collateralized li Since we have 16 hedging pressure (HP) benchmarks (that come from the permutation of 4 ranking periods and 4 holding periods), we end up with 16 IV strategies For the HP benchmarks and the IV portfolios, we present the performance ofa portfolio that equally weights these 16 portfolios 9

10 Evolution of $1 invested in long only and longshort benchmarks or in IV strategy 10

11 Mean excess return, Standard deviation and Sharpe ratio Annualized mean excess returns Annualized SD Sharpe ratio Panel A: Idiosyncratic volatility strategy Best strategy, R = 4, H = 4 weeks 6.18% (2.69) 9.62% Worse strategy, R = 52, H = 4 weeks 4.67% (1.96) 9.85% Average performance 5.49% 9.56% Panel B: Hedging pressure benchmark Best strategy, R = 13, H = 26 weeks 7.18% (3.06) 9.82% Worse strategy, R = 26, H = 52 weeks 2.60% (1.19) 9.13% Average performance 4.54% 9.68% Panel C: Long only benchmarks S&P GSCI 4.28% (0.81) 84.48% Equally weighted portfolio 0.64% (0.21) 22.73% R is the ranking period of the HP based benchmark H is the holding period of the HP benchmark used to model IV and the holding period of the IV portfolio t statistic in parentheses 11

12 Abnormal performance (Alpha) of IV portfolios relative to HP benchmarks R = Ranking period of HP benchmark H = Holding period of HP benchmark used to model IV and holding period of IV portfolio 12

13 Idiosyncratic Volatility Strategies in Commodity Futures Markets Single sort idiosyncratic volatility strategy Portfolio construction Benchmark definition Performance evaluation Triple sort strategy combining idiosyncratic volatility, momentum and term structure signals Can the idiosyncratic volatility portfolios be used for risk diversification and inflation hedging? Robustness checks 13

14 IVversus momentum and TS strategies Two strategies have been shown to work well in commodity futures markets Momentum strategy that buys past winners and sells past losers (Erb and Harvey, 2006; Miffre and Rallis, 2007) Term structure (TS) strategy that buys commodities with high roll returns and sell commodities with low roll returns (Erb and Harvey, 2006; Gorton and Rouwenhorst, 2006) Combining momentum and TS signals in a double sort generates an even better performance (Fuertes, Miffre and Rallis, 2010) To which extend does our new IV strategy overlap with the double sort strategy based onmomentum andts? 14

15 IVversus momentum and TS strategies Average correlation between double sort portfolios based on momentum and term structure and IV portfolios is merely at 0.11 Low IV portfolios buy commodities that are not necessarily past winners with high average roll returns High IV portfolios sell commodities that are not necessarily past losers with low average roll returns Canperformance be further enhanced by combining the signals in a triple sort? 15

16 Portfolio construction Ranking period (R) overwhichiv is modelled: 52 weeks R of momentum and TS signals: 4, 13, 26 or 52 weeks Holding period (H): 4, 13, 26, 52 weeks IV Momentum TS Up to 27 commodities Low IV (80%) High IV (80%) High Mom (50%) Low Mom High Mom Low Mom (50%) High Roll (50%) Low Roll High Roll Buy (20%) Ignore remaining 60% Low Roll Sell (50%) (20%)

17 Evolution of $1 invested in long short IV strategies 17

18 Mean excess return, Standard deviation and Sharpe ratio We present summary statistics i for portfolios that equally weight ihall 16 combinations of ranking and holding periods for single sort IV, double sort Mom TS, triple sort that combines IV, Mom and TS signals Annualized mean excess returns Annualized SD Sharpe ratio Single sort IV 5.48% (2.45) 9.36% Double sort Mom TS 6.57% (3.46) 7.71% 71% Triple sort Mom TS IV 6.84% (3.91) 7.31% Triple sort Mom IV TS 7.11% (4.05) 7.34% Triple sort IV Mom TS 7.15% (3.80) 7.64% t statistics in parentheses 18

19 Alphas of single and triple sort IV portfolios relative to HP benchmarks 19

20 Performance of triple sort strategy depends on weighting assigned to IV signal 20

21 Idiosyncratic Volatility Strategies in Commodity Futures Markets Single sort idiosyncratic volatility strategy Portfolio construction Benchmark definition Performance evaluation Triple sort strategy combining idiosyncratic volatility, momentum and term structure signals Can the idiosyncratic volatility portfolios be used for risk diversification and inflation hedging? Robustness checks Transaction costs Liquidity 21

22 CanIVportfolios be used for risk diversification? Relative to long-only, l long-short IV portfolios act as worse diversifiers ifi of fixed income risk and better diversifiers of equity risk 22

23 CanIVportfolios be used to hedge inflation? Relative to long-only, l long-short IV portfolios fail to hedge inflation shocks 23

24 Idiosyncratic Volatility Strategies in Commodity Futures Markets Single sort idiosyncratic volatility strategy Portfolio construction Benchmark definition Performance evaluation Triple sort strategy combining idiosyncratic volatility, momentum and term structure signals Can the idiosyncratic volatility portfolios be used for risk diversification and inflation hedging? Robustness checks Transaction costs Liquidity 24

25 Superior performance is not wiped out by transaction costs The results presented here overstate performance by an amount equal to the cost of trading Trading costs should not be overstated Transaction costs: Futures: % 0004% to 0.033% 033% (Locke and Venkatesh, 1997) Large capitalization stocks: 0.5% Small capitalization stocks: 2.3% Not much problem of liquidity No restrictions on short selling Small cross section of commodity futures on which the strategy is implemented (up to 27 contracts) Infrequent rebalancing 25

26 Superior performance is not wiped out by transaction costs 26

27 Superior performance is not merely a compensation for a lack of liquidity To address concerns over lack of liquidity, we redeploy the IV strategy, systematically excluding the 10% of commodities with the lowest average open interest over the R weeks preceding portfolio formation Single sort IV Triple sort IV Panel A: Mean excess return Including all 27 commodities 5.49% 7.00% Excluding 10% least liquid commodities 3.50% 5.67% Panel B: Annualized alpha Including all 27 commodities 4.62% 5.59% Excluding 10% least liquid commodities 3.07% 4.44% 27

28 Conclusions We present new active strategies in commodity futures markets based on idiosyncratic volatility (IV) Buying commodities i with ihlow IV and selling commodities i with ih high IV generate an average alpha of 4.62% a year Combining i the IV signal with momentum and term structure t signals enhances average alpha further, to 5.59% a year The IV portfolios can be used to diversify if equity risk ikbut not to hedge inflation The strategy t is cheap to implement, profitable net of transaction costs and does not merely pick illiquid contracts 28