+ The Stock Market Price of Commodity Risk November 2013 Martijn Boons, Nova School of Business and Economics Frans de Roon, Tilburg University, CentER Marta Szymanowska, Rotterdam School of Management + 2 Motivation Commodity Futures Modernization Act (CFMA) Dramatic change in size and composition of futures markets 16 TOI in 33 commodities 14 12 10 Energy Agriculture 8 6 4 2 0 Metals & Fibers Livestock & Meats EW Average + 3 Motivation CFMA: break point in the behavior of (institutional) investors Pre-CFMA commodity exposure position limits in futures markets commodity-related equity, physical commodities (Lewis, 2007) Post-CFMA commodity exposure commodity index investment (CII) by institutions from 6% of total open interest (< 10$ bln) in 1998 to 40% (> 200$ bln) in 2009 + 4 Our goal We want to understand commodity prices as a source of risk price of this risk in the stock and commodity futures markets impact of CFMA / changing investment behavior This will allow us to shed light on a link between stock and commodity futures markets “financialization” of commodities stock market strategies to hedge or speculate on commodity prices + 5 Our Approach A model with investors exposed to commodity price risk Testable implications in the spirit of Hirshleifer (1988,1989), Bessembinder and Lemmon (2002) Study the effect of position limits on demand and prices Sort stocks on commodity beta Sort commodity futures on stock market risk Main empirical findings 1. 2. Commodity risk is priced in stock market in the opposite way before and after CFMA Stock market risk is priced in the commodity futures market post-CFMA + 6 The model Agents Commodity Producers (business exposed to commodity price risk 𝑟𝑆,𝑡+1 and trade futures contract 𝑟𝐹,𝑡+1 ) Specialized Speculators (e.g. CTA's, trade futures contract) Investors Position limit (pre-CFMA): invest in stocks (𝑟𝑟,𝑡+1 ) only No limit (post-CFMA): invest in both stocks and futures contract Standard, two-date, mean-variance framework Investors are exposed to commodity price risk: inflation, state variable Today: available futures contract is a perfect hedge (𝜌𝑆𝐹 = 1) + 7 The Investor’s problem Excess portfolio return 𝑟𝑝,𝑡+1 = 𝑤 ′ 𝑟𝑡+1 + 𝜑𝑟𝑆,𝑡+1 , such that 1. With limit (w = 𝑤 ′ 𝑟 0 ′ ) max w r 2. γI w'r μ r μ S ( w'r Σ rr w r 2 w'r Σ rS 2 SS ) 2 Without limit (𝑤 = 𝑤 ′ 𝑟 𝑤𝐹 ′ ) γI max w w' μ μ S ( w' Σw 2 w' Σ S 2 SS ) 2 Optimal portfolios (1) with and (2) without limit 1 1. w r γ I Σ rr1μ r Σ rr1Σ rS 1 w r γ I Σ rr1μ r Σ rr1Σ rS w F,spec 0 K 2. , w F,spec wF 1 where w F,spec γ I ee1a using rF,t 1 a b' rr ,t 1 et 1 + 8 Expected stock returns with commodity price risk With limits E (ri ,t 1 ) γ I im γ I iS Cross-hedging demand implies a negative (positive) risk premium when φ < 0 (φ > 0) and high commodity prices are bad (good) news Without limits E (ri ,t 1 ) γ I im γ I w F,spec iS Risk premium determined by speculative investment in commodities If zero CAPM! + 9 Risk premiums in the futures market 1. With limit: Producers and Speculators only E (rF ,t 1 ) P P S γ P FF (1 ) i N i / γ i , i P, S 2. Without position limits: stock market risk is priced due to presence of Investors P γ P (1 ) Iγ I I γ I E (rF ,t 1 ) FF FT with P S I P S I FF i N i / γ i , i P, S , I and I I ee FT cov(rF ,t 1 , rT ,t 1 ) where w T rr1 r + 10 Data and method: stock market All CRSP stocks, French’s 48 industry portfolios OIW index of 33 commodities (from CRB and FII) Sorts on rolling 60 month commodity beta Robust: EW index, S&P-GSCI index Variation across commodity sectors Mean and risk-adjusted returns (CAPM, FF3M and FFCM) of High minus Low (HLCB) portfolios Pre- versus Post-CFMA: split around December 2003 Robust Different break points Different rebalancing Fama-MacBeth cross-sectional estimates Between/within industry sort Controlling for inflation + 11 Stock market: pre-CFMA Stocks 48 Industries 15.00 15.00 10.00 10.00 FFCM 5.00 CAPM Low -10.00 FFCM FF3M 0.00 -5.00 5.00 2 Means 3 4 High FF3M 0.00 Low -5.00 -10.00 CAPM 2 3 4 High Means + 12 Stock market: post-CFMA 48 Industries Stocks 15.00 15.00 10.00 10.00 FFCM 5.00 FF3M CAPM 0.00 Low -5.00 5.00 2 0.00 Means 3 4 High -5.00 -10.00 -10.00 FFCM FF3M CAPM Means + 13 Means and FFCM alphas Pre-CFMA Size quintile Means One-way Size quintile One-way OIW OIW OIW OIW OIW EW OIW OIW OIW OIW OIW EW S 3 B Stocks 48 Ind. Stocks S 3 B Stocks 48 Ind. Stocks H 5.88 3.55 2.33 1.91 5.00 4.45 12.13 15.29 15.10* 14.85* 14.57 11.93 4 8.88* 6.90* 7.04* 6.58* 8.23* 5.77 12.02 9.97 4.78 5.64 5.97 7.33 3 10.56* 9.44* 6.32* 7.04* 7.84* 8.25* 11.07 8.58 2.08 3.58 6.62 5.16 2 10.55* 11.32* 9.24* 9.53* 10.07* 8.81* 9.25 7.91 3.08 3.87 6.47 5.07 L 8.93* 13.03* 10.01* 10.02* 9.72* 9.33* 1.88 1.98 3.25 2.77 2.35 3.24 -3.04 -9.47* -7.68* -8.11* -4.72* -4.88 12.08* 12.22* 8.69 H -1.73 -6.12* -5.52* -6.67* -4.75* -3.52 1.65 6.81 11.30* 9.82* 8.60* 6.23 4 0.69 -3.23* -0.97 -1.73 -0.92 0.40 2.40 2.46 1.67 1.33 -0.82 1.76 3 2.41 0.43 -0.61 -0.13 -1.99 0.76 1.60 1.66 -1.83 -0.93 1.08 1.16 2 2.82 3.48* 3.22* 3.33* 2.13 1.08 0.77 1.53 -0.47 -0.19 1.23 1.18 L 2.75 5.59* 5.88* 4.99* 2.12 2.77* -6.66* -4.67* 0.36 -1.08 -2.01 -0.09 -6.87* -6.30* 8.31* 11.48* 10.94* 10.90* 10.60* 6.32 HLCB FFCM Post-CFMA HLCB -4.48* -11.71* -11.39* -11.66* * Indicates significance at the 5%-level 10.25* 13.31* 11.85* + The reversal in the commodity risk premium I Recall: E (ri ,t 1 ) γ I im γ I iS vs E (ri ,t 1 ) γ I im γ I wF , spec iS 𝑎 Reversal obtains when (1) 𝜑 < 0 and (2) 𝛾𝐼 𝑤𝐹,𝑠𝑝𝑒𝑐 = > 0. 𝜎𝑒𝑒 Plausible: 1. Negative exposure to commodity price risk for Investors from Inflation: commodity prices are most volatile components State-variable risk: Energy and Metals prices predict negative stock returns (e.g., Driesprong et al. (JFE, 2008), Jacobsen et al. (2013)) Results driven by commodities from Energy and Metals sectors 14 + 15 The reversal in the commodity risk premium II A positive speculative investment in commodity futures 𝑎 ( > 0) obtains when 2. 𝜎𝑒𝑒 Hedging pressure from Producers is sufficiently large, i.e., the group of Producers is relatively large and risk averse (“normal backwardation”) Indeed, we find that commercial hedger’s short positions are sufficient to cover non-commercial speculators long positions Cheng et al. (2011): hedgers short positions increase in lockstep with CIT’s long positions Consistent with diversification benefits in Gorton and Rouwenhorst (FAJ, 2006) and Erb and Harvey (FAJ, 2006) + 16 Hedgers versus Speculators + 17 Commodity futures risk premiums With and without limit: a “classic“ hedging pressure effect In both sub-periods, sorting on hedging pressure works Without limits, stock market risk is priced in the futures market Using that T=M+H, sort commodities on beta with respect to the MKT and HLCB portfolio High stock market beta commodities outperform ONLY postCFMA, as predicted! Pre-CFMA Post-CFMA HLCB HLCB MKT Low High MKT Low High Low 1.36% 7.43% Low -2.23% 8.09% High -2.25% 3.83% High 9.06% 9.16% HH-LL 2.48% HH-LL 11.38% t(HH-LL) (0.55) t(HH-LL) (1.77) + 18 Conclusion Focus on the structural break in investor’s behavior Study a model with Investors exposed to commodity price risk Analyze the effect of position limits related to CFMA We find Commodity risk is priced in stock market in the opposite way pre- versus post-CFMA Stock market risk is priced in the commodity futures market postCFMA Consistent with Investors seeking commodity exposure in the stock market pre-CFMA and subsequently in the commodity futures markets Stocks to hedge or speculate on commodity prices + 19 Within-industry sort “Out-of-sample” test: spreads exist when using only withinindustry variation in commodity beta Hedge, while keeping industry exposure constant Within-industry H 1980-2003 (Pre-CFMA) 2004-2010 (Post-CFMA) Industries sorted on commodity beta Industries sorted on commodity beta 4 3 2 L -6.13* -4.17 -3.34 -4.72 Means HLCB -3.39 FFCM HLCB -6.92* -7.58* -4.37 Average -4.35* -4.86* -9.01* -6.55* H 4 3 2 L Average 13.64* 11.01* 5.38 19.05* 9.37 11.69* 13.92* 9.76 14.58* 5.48 9.18* 2.17 + 20 Industry composition of High and Low portfolio 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 1980-1990 1991-2000 2001-2010 0.25 0.20 0.15 0.10 0.05 1980-1990 0.00 1991-2000 2001-2010