Risk Convection in Commodities Futures Markets

1. Risk Convectionin Commodities Futures Markets Ing-Haw ChengUniversity of Michigan Andrei Kirilenko CFTC Wei Xiong Princeton University This presentation and the views presented here represent only our views and do not necessarily represent the views of the Commission, Commissioners or staff of the Commodities Futures Trading Commission.

2. Motivation • Do financial intermediaries amplify shocks? • Common to observe increased correlations between different asset markets during market stresses. • Due to increased underlying correlation, increased volatility, or both. • Is the discount rate explanation a complete story? • Cochrane, 2011: “Ideally, one should tie price or discount-rate variation to central items in the models, such as the balance sheets of leveraged intermediaries, data on who is actually active in segmented markets, and so forth.” Cheng, Kirilenko and Xiong 2011

3. Commodities open interest 1. Collapse in open interest just as uncertainty spiked. 2. Concurrent with significant price drops. 3. Consistent across many markets- not substitution. Cheng, Kirilenko and Xiong 2011

4. Motivation Cheng, Kirilenko and Xiong 2011

5. Research question 1. Did distressed financial institutions exacerbate the effects of outside shocks on commodities futures markets? • Outside the initial center of the financial crisis. • Comprehensive position data at a daily frequency. 2. What were the commercial hedgers doing? • Reduced hedges as the VIX rose (open interest fell). Why? • Implications for the real economy and corporate hedging. Cheng, Kirilenko and Xiong 2011

6. Our paper Idea: Examine whether risk was systematically re-allocated among market participants as uncertainty spiked. • Discount rate story may be incomplete. • Focusing on the VIX allows us to contrast theories of risk allocation. • Deeper implications for corporate hedging behavior. Strategy: Looking at price patterns can only get you so far. Position quantities and responses to outside market movements can yield more insight into asset market dynamics. (Who was selling?) Key theme: We think of financials as providing liquidity, and commercial hedgers as using futures markets to offload risk. Did this reverse? Cheng, Kirilenko and Xiong 2011

7. Results: a preview Our main findings are as follows. • After the crisis, financials sold in response to increases in the VIX, while commercial hedgers bought. • We observe particularly high sensitivities among certain types of financials, “commodities index traders,” whose CDS spreads were high. • Long commercial hedgers did not reduce their hedges, suggesting this was not driven by a desire to reduce hedging. Punchline: There was a flow of risk away from distressed financials towards commercial hedgers after the crisis and an amplifying role for financial traders, or a risk convection. Cheng, Kirilenko and Xiong 2011

8. What is convection? Flow of moisture Cheng, Kirilenko and Xiong 2011

9. Flow of moisture Cheng, Kirilenko and Xiong 2011

10. Who operates in commodities markets? • Commercial Hedgers: • Offload price risk to speculators for a risk premium • Canonical hedging pressure theory (Keynes, Hicks) • Typically short in most markets • Hedge Funds: • Long short mixed dynamic strategies, • Leverage and exposure to funding risk • How do shocks affect the market? • Is it a common shock? • Are some groups offloading risk? • Which direction does the risk flow? • CITs: • A new breed of investors treatingcommodities as an asset class • Exposure to equities throughother portfolio holdings Cheng, Kirilenko and Xiong 2011

11. Theories and predictions Suppose the VIX spikes up. Who trades and why? Standard view: representative agent theory Makes little prediction on systematic re-allocation of risk in response to shocks. While massive discount rate shocks surely affected prices during the crisis, indications of systematic re-allocation of risk suggest that this may be an incomplete story. Cheng, Kirilenko and Xiong 2011

12. Theories and predictions Another standard view: hedging pressure theory Commercials (farmers) consume liquidity to hedge future price risk (Keynes and Hicks; Hirshleifer 1988), with financials taking the other side. Motivated by (among other theories): • Underinvestment costs (Froot, Scharfstein and Stein 1993) • Minimizing financial distress (Smith and Stulz 1985) When the VIX spikes, hedgers may want to increase their hedges. Long and short hedgers should have opposite reactions to the VIX; this would be true even if hedgers wanted to reduce their hedges. Did they? Cheng, Kirilenko and Xiong 2011

13. Theories and predictions Distressed financials hypothesis Emphasizes the that during the crisis, financials such as CITs and hedge funds became distressed. • Distressed selling of commodity futures at distressed prices, consuming liquidity • Naturally, commercial hedgers must take the other side: re-allocation of risk back towards commercial hedgers Cheng, Kirilenko and Xiong 2011

14. Empirical strategy Exploit the cross-section of traders and trader groups and the differential predictions of who should be selling. • Look at which groups responded to the VIX before and after the crisis. • Examine financials at the micro level. Did distressed financials sell? • Examine hedgers. Can theories of hedging explain the pattern of position changes and prices? • Examine the medium/long-run responses of trading. Was there a persistent re-allocation of risk? Cheng, Kirilenko and Xiong 2011

15. Data CFTC’s Large Trader Reporting System (LTRS), 2000-2011. • Provides detailed daily data on traders’ long and short positions on individual futures contracts. • Traders with positions in excess of a reportable level are reported to the CFTC by clearing members. • Generally 70-90% of the open interest. • Basis of weekly “Commitment of Traders” public reports. Use this data to jointly look at reactions of all groups to the shock. • Less likely to miss the effect of any one group because of excessive focus on other groups. • Allows us to construct finer categorizations of traders than in the publicly available versions of the data. • Other data from Bloomberg, FRB. Cheng, Kirilenko and Xiong 2011

16. Table 1: Commodities Cheng, Kirilenko and Xiong 2011

17. Trader classifications The LTRS contains information filed by traders as to their purposes of trade. • Basis of classification in public COT reports. • We extend this by combining information about their trading behavior in the previous year. Commodity index traders • Traders with who invested in 8 or more commodities in the previous year. • Were long on average in the commodities in which they had exposure. • Intersect this with the CFTC CIT classification, constructed through interviews with specific market participants. Cheng, Kirilenko and Xiong 2011

18. Trader classifications Hedge funds • Traders registered as commodity pool operators (CPOs), commodity trading advisors (CTA), or managed money. Commercial hedgers • Traders of types “Dealer/Merchant,” “Agricultural,” “Manufacturer,” “Producer,” or Livestock feeder/slaughterer. Others • Many traders may fall outside of our strict classification scheme. Leave the behavior of these traders as an empirical question. One trader may have multiple classifications. Because we are interested in the time series properties of position responses, we separate these out. Cheng, Kirilenko and Xiong 2011

19. Trader classifications Other Reporting Traders Non-Reporting Traders Commodity Index Traders CIT-HF CIT-Hedger Triple Hedge Funds / Managed Money Commercial Hedgers Hedger- HF Cheng, Kirilenko and Xiong 2011

20. Market participation in 2010 Cheng, Kirilenko and Xiong 2011

21. Evolution of market participation Cheng, Kirilenko and Xiong 2011

22. Evolution of market participation Cheng, Kirilenko and Xiong 2011

23. Evolution of market participation Cheng, Kirilenko and Xiong 2011

24. Evolution of market participation Cheng, Kirilenko and Xiong 2011

25. Evolution of market participation Cheng, Kirilenko and Xiong 2011

26. Evolution of market participation Cheng, Kirilenko and Xiong 2011

27. Evolution of market participation Cheng, Kirilenko and Xiong 2011

28. Evolution of market participation Cheng, Kirilenko and Xiong 2011

29. Evolution of market participation Cheng, Kirilenko and Xiong 2011

30. Evolution of market participation Cheng, Kirilenko and Xiong 2011

31. Evolution of market participation Cheng, Kirilenko and Xiong 2011

32. Table 2: Trader characteristics Cheng, Kirilenko and Xiong 2011

33. Commodity exposures Cheng, Kirilenko and Xiong 2011

34. Basic exercise Cheng, Kirilenko and Xiong 2011

35. Table 4: Price correlation Cheng, Kirilenko and Xiong 2011

36. Table 5: Position response Cheng, Kirilenko and Xiong 2011

37. Table 5: Position response Cheng, Kirilenko and Xiong 2011

38. Table 6: Distress of CITs Cheng, Kirilenko and Xiong 2011

39. Table 6: Distress of CITs • Consistent with distressed financial institution hypothesis that vulnerable financial institutions are selling. Suggests convection was towards commercials. • Instead of exploiting relative ranking of CDS spreads, could have interacted absolute level – same. • Effect could be due to selling of own proprietary positions co-mingled with the account. Or clients may withdraw their investment as the institution is under distress. Cheng, Kirilenko and Xiong 2011

40. Table 7: Hedging pressure • Alternative story: hedgers wanted to reduce hedges. Convection was towards financials. • Commercials would want to reduce their hedges when the VIX rises if… • Commodity price volatility was dropping. (It wasn’t.) • Risk of financial distress was declining (Smith and Stulz 1985) – unlikely. • Cost of external financing declined (FSS 1993) – unlikely. • Investment opportunity set declined (FSS 1993) – possibly. Although, this is open to interpretation, since hedgers are short; they would make money when price falls and need less cash. • Suggests a test: classify long hedgers and short hedgers and look for differential response (reductions in hedges). Cheng, Kirilenko and Xiong 2011

41. Table 7: Hedging pressure Cheng, Kirilenko and Xiong 2011

42. Table 7: Hedging pressure Several theories would actually suggest that hedgers would want to increase their hedges as the VIX increased, rather than decrease. • Short hedgers would have been making money as prices fell; why reduce the futures position? Expectations for demand could have fallen, but data on production in corn and wheat suggest quantities did not decline very much through the crisis. Story of hedging must fit all these facts. Cheng, Kirilenko and Xiong 2011

43. World corn production (USDA data) Cheng, Kirilenko and Xiong 2011

44. Table 8: Active and inactive hedgers Alternative story: financials were exploiting an informational advantage. Did hedgers just miss the boat? Classify hedgers as “active” and “inactive” based on median weekly position change in previous year. • Take top 10% in the previous year’s cross-section and classify as active for this year. Compare sensitivity of aggregate position changes across these two groups. Cheng, Kirilenko and Xiong 2011

45. Table 8: Active and inactive hedgers Cheng, Kirilenko and Xiong 2011

46. Table 8: Active and inactive hedgers Cheng, Kirilenko and Xiong 2011

47. Robustness: Informational Advantages? One alternative is that perhaps the VIX is bad news, and hedgers react slowly. This would allowing CIT and hedge funds to exploit them by selling and then buying back. Predicts position reversal in response to the VIX. Cheng, Kirilenko and Xiong 2011

48. Table 9: Reversals Cheng, Kirilenko and Xiong 2011

49. Table 9: Reversals Cheng, Kirilenko and Xiong 2011

50. Table 9: Reversals Cheng, Kirilenko and Xiong 2011