فعالیت های بازرگانی و قیمت معکوس در بازارهای آتی

We use the standard contrarian portfolio approach to examine short-horizon return predictability in 24 US futures markets. We find strong evidence of weekly return reversals, similar to the findings from equity market studies. When interacting between past returns and lagged changes in trading activity (volume and/or open interest), we find that the profits to contrarian portfolio strategies are, on average, positively associated with lagged changes in trading volume, but negatively related to lagged changes in open interest. We also show that futures return predictability is more pronounced if interacting between past returns and lagged changes in both volume and open interest. Our results suggest that futures market overreaction exists, and both past prices and trading activity contain useful information about future market movements. These findings have implications for futures market efficiency and are useful for futures market participants, particularly commodity pool operators.

A large body of finance literature shows that past prices contain useful information about future market movements in equity markets. For example, Lehmann (1990) and Lo and MacKinlay (1990) find that a short-term contrarian portfolio strategy of buying past losers and selling past winners generates significant profits. Lehmann also shows that the contrarian profits remain significant after corrections for plausible transaction costs. The evidence on short-horizon return predictability is consistent with the overreaction hypothesis, namely, traders over-adjust their posterior beliefs to news by more than what is warranted by fundamentals (Lehmann, 1990), although market microstructure biases like lead–lag and bid–ask spread effects may explain away some of the contrarian profits (Lo and MacKinlay, 1990; Conrad et al., 1997). Both academics and practitioners have also been interested in the role played by trading volume in predicting future market movements. Blume et al. (1994) present a model in which traders can profit from using volume information in addition to historical price information to forecast future price changes. Recent empirical finance research has shown that the relation between trading volume and expected stock returns is in general negative, although the interpretation of the negative relation has been controversial. The traditional liquidity premium hypothesis suggests that trading volume is a measure of liquidity, and high (low) volume assets should command lower (higher) returns, on average (Amihud and Mendelson, 1986; Brennan and Subrahmanyam, 1996; Brennan et al., 1998).1 Hence, a negative relation between volume and expected returns is consistent with the notion of liquidity premiums. Several behavioral finance studies contend that individuals are overconfident about their ability to evaluate securities, in the sense that they overestimate precision of private information signals, resulting in overreaction to private information and causing asset prices to temporarily swing away from the fundamental value (DeBondt and Thaler, 1985; Odean, 1998; Gervais and Odean, 2001). Overconfidence and overreaction themselves imply a large volume of trading and are thus positively related to the magnitude of price reversals.2 Therefore, the investor irrationality-induced market inefficiency gives rise to a negative relation between volume and expected returns (DeBondt and Thaler, 1985; De Long et al., 1990; Odean, 1998; Statman and Thorley, 1999).3 Although past prices and trading activity are closely watched by futures market participants, the informational role of past prices and trading activity in broader futures markets has not been well studied. Evidence of predictable futures returns has profound implications for market efficiency as well as practical trading strategies, because arbitrage strategies are more readily available than in equity markets due to the low cost and high liquidity of futures trading. This paper adds to the literature by providing evidence on short-horizon (1–8 weeks) return predictability for a sample of 24 actively traded US futures contracts over the July 1983–June 2000 interval. The primary focus of this study is on the following two issues: first, following a standard contrarian portfolio approach similar to that of Lehmann (1990), we examine whether consistent contrarian profits in futures markets exists; and second, we investigate futures return predictability by interacting between past returns and lagged changes in trading activity (volume and/or open interest). An important feature of futures markets is that open interest provides an additional measure of trading activity. Open interest represents total long or short positions of all traders in a market and has been shown to covary with price changes. Bessembinder and Seguin (1993) document a positive (negative) relation between price volatility and volume (open interest) and show that the effect of volume on volatility depends on whether volume generates changes in open interest (p. 38). These authors argue that open interest represents uninformed trading by hedgers or hedging activity and is thus an important determinant of market depth.4 Other things being equal, a market becomes deeper as open interest increases, leading to smaller price volatility. However, there has been little evidence on the relation between open interest and inter-temporal return predictability. We find strong evidence of futures return reversals over the 1-week horizon. A contrarian strategy of buying past losers and selling past winners gives rise to an average return of 0.31% per week (16.12% per annum). This return is both statistically and economically significant given the high leverage of futures trading and high liquidity of futures markets and is robust to a one-way transaction cost (commissions and one-half the bid–ask spread) of up to 0.16%. We also find a significant relation between lagged changes in trading activity (volume and open interest) and contrarian profits, but the relation differs for the two measures of trading activity: The relation between changes in volume and contrarian profits is positive, whereas there is a negative relation between changes in open interest and contrarian profits. As a result, when implemented in high-volume (low-open interest) contracts, a contrarian strategy is more profitable than that in low-volume (high-open interest) contracts. For example, the average return from a contrarian strategy implemented in high-volume contracts is 0.41% per week (t=3.07), whereas it is only 0.04% per week (t=0.28) for a contrarian strategy implemented in low-volume contracts. 5 For low-open interest contracts, a contrarian strategy generates an average return of 0.33% per week (t=2.56), whereas the mean return for a contrarian strategy implemented in high-open interest contracts is only 0.10% per week (t=0.72). A contrarain strategy is most profitable if it is implemented in high-volume together with low-open interest contracts. Under this circumstance, the mean return for the 1-week contrarian strategy is as high as 0.57% per week (t=3.86) or 29.64% per annum. Further examination of return predictability for a formation/holding horizon spanning from 2 to 8 weeks shows no evidence of contrarian profits. This result holds true even if a contrarian strategy is conditional on lagged changes in trading activity (volume and open interest). The evidence of significant contrarian profits over the 1-week horizon is in line with the findings in equity markets (Lehmann, 1990; Lo and MacKinlay, 1990; Mei and Gao, 1995). Market condition and investors' perception of risk should remain relatively unchanged over a short interval; issues like lead–lag, bid–ask spread, and nonsynchronous trading effects are trivial in futures markets (see also Table 1), and therefore the result of consistent contrarian profits points toward the market overreaction. The finding of futures market inefficiency is surprising given that futures markets are very close to a textbook model of competition and arbitrage conditions are most satisfied.6 Nevertheless, our result tends to be in line with the previous futures market literature, for example, Ma et al. (1989), Ma et al. (1990), and Park et al. (1997). These authors find evidence of overreaction in several futures markets. Table 1. Summary statistics for futures returns, trading volume, and open interest Mean St. Dev. t-value Autocorrelations ρ1 ρ2 ρ3 ρ4 Panel A: Summary statistics for weekly futures returns Financials Eurodollar 0.005 0.210 0.62 0.002 −0.034 −0.007 −0.098* T-notes 0.039 0.979 1.17 −0.031* 0.041* 0.022 0.030 T-bills 0.004 0.203 0.61 −0.040* −0.048* −0.027 0.038 S&P 500 0.269 2.145 3.76 −0.032* −0.007 −0.014 −0.042 NYSE 0.236 2.087 3.39 −0.027* 0.001 −0.017 −0.040* Currencies British pound 0.010 1.560 0.18 −0.020 0.000 −0.033 0.012 Deutsche mark 0.036 1.576 0.68 −0.035* 0.051* −0.033 0.017 Japanese yen 0.107 1.609 1.95 −0.002 0.011 0.007 0.002* Swiss franc 0.044 1.708 0.76 −0.002 0.065* 0.007 0.020 Agriculturals Corn 0.005 3.170 0.05 −0.001 0.039 −0.019 0.024 Cotton 0.037 4.089 0.31 −0.002 0.015 −0.023* 0.069* Feeder cattle 0.052 1.876 0.82 −0.011 −0.036* 0.015 0.019 Live cattle 0.036 2.194 0.44 −0.069* −0.003 0.019 −0.003 Soybeans 0.020 2.939 0.20 −0.038* 0.027 −0.022 −0.023 Soybean oil 0.036 3.338 0.32 −0.065* 0.075* −0.035 0.039 World sugar 0.131 5.527 0.70 0.002 0.070* −0.034 0.058 Wheat 0.027 3.292 0.24 −0.026* −0.033* −0.025 −0.012 Commodities Cocoa −0.027 3.915 −0.20 −0.059* −0.023 −0.008 0.017 Coffee 0.106 5.471 0.56 −0.031* −0.027 −0.075* −0.014 Crude oil 0.120 4.886 0.73 −0.073* 0.067* −0.045* −0.010 Gold −0.024 1.885 −0.37 −0.092* −0.041* 0.074* 0.093* Heating oil 0.118 4.920 0.72 0.007 −0.012 −0.060 0.032 Platinum 0.076 3.056 0.73 −0.097* −0.010 0.010 0.058* Silver −0.040 3.385 −0.35 −0.032* −0.035* −0.059* 0.084* Market 0.009 0.949 0.42 −0.059* 0.001 −0.068* 0.027 Volume Open interest Change in trading activity Volume Open interest Mean St. Dev. Max Min Mean St. Dev. Max Min Mean St. Dev. Autocorrelations Mean St. Dev. Autocorrelations ρ1 ρ2 ρ1 ρ1 ρ1 ρ2 ρ1 ρ1 Panel B: Summary statistics for trading activity Financials Eurodollar 229.66 181.29 901.05 2.99 1,417.36 1,124.37 13,274.76 29.59 0.001 0.014 −0.03 −0.15 −0.06 0.07 0.002 0.012 −0.48 −0.00 −0.00 −0.01 T-notes 56.83 51.52 336.06 1.58 206.70 176.56 694.36 9.03 0.002 0.015 −0.19 −0.02 −0.03 −0.05 0.003 0.002 0.42 0.09 −0.07 −0.10 T-bills 5.30 4.62 30.80 0.01 28.48 16.56 61.34 0.12 −0.003 0.017 −0.39 −0.03 −0.01 −0.05 −0.002 0.003 0.07 −0.18 −0.12 −0.08 S&P 500 67.87 31.87 271.23 9.74 300.02 117.19 576.81 44.03 0.000 0.011 −0.02 −0.19 −0.12 −0.04 0.001 0.002 −0.13 −0.04 −0.04 −0.06 NYSE 5.91 3.57 17.53 0.12 6.96 10.97 19.89 1.13 −0.002 0.008 −0.15 −0.23 −0.04 0.00 −0.002 0.005 0.15 −0.21 −0.13 −0.08 Currencies British pound 11.28 4.81 35.34 1.66 35.64 13.05 76.78 12.65 −0.000 0.014 −0.28 −0.15 −0.11 0.03 0.000 0.005 0.16 −0.23 −0.11 −0.02 Deutsche mark 28.98 14.74 80.83 0.01 67.25 34.49 175.53 0.64 −0.000 0.011 −0.27 −0.13 −0.11 0.01 −0.000 0.005 0.21 −0.22 −0.10 −0.07 Japanese yen 21.63 10.02 83.97 1.51 62.66 27.14 154.49 13.23 0.000 0.014 −0.24 −0.22 0-.06 0.05 0.000 0.004 0.16 −0.23 −0.12 −0.10 Swiss franc 16.77 8.68 46.38 4.50 39.73 14.02 103.26 16.96 −0.000 0.010 −0.28 −0.12 −0.12 −0.02 0.000 0.004 −0.33 −0.13 −0.00 0.02 Agriculturals Corn 51.79 21.73 154.79 6.95 237.31 122.36 940.46 43.69 0.001 0.010 −0.34 0.01 −0.10 −0.08 0.002 0.006 −0.28 −0.23 0.23 −0.01 Cotton 6.95 4.14 78.21 0.03 6.96 10.97 289.36 0.27 0.001 0.015 −0.33 −0.17 0.12 −0.04 0.001 0.004 0.21 0.06 0.08 0.04 Feeder cattle 1.99 0.96 30.78 0.00 12.71 5.49 140.07 0.04 −0.000 0.011 −0.30 −0.12 −0.02 0.06 0.001 0.003 0.07 0.10 −0.09 −0.07 Live cattle 14.52 5.07 30.55 0.01 71.12 28.10 135.12 0.02 −0.000 0.009 −0.34 −0.11 −0.00 −0.00 0.001 0.002 −0.13 −0.12 0.10 −0.07 Soybeans 41.08 17.43 14.17 1.83 118.07 43.35 213.22 12.53 −0.000 0.011 −0.35 −0.11 0.05 −0.05 0.001 0.003 0.07 0.00 0.03 0.06 Soybean oil 17.70 5.96 54.84 3.36 81.34 29.22 168.68 28.26 0.001 0.010 −0.32 −0.05 −0.09 −0.04 0.001 0.002 −0.07 −0.14 0.12 −0.10 World sugar 23.00 16.96 54.56 0.02 113.25 54.87 216.95 0.64 0.002 0.014 −0.33 −0.08 0.01 −0.08 0.002 0.003 0.34 −0.02 0.01 0.01 Wheat 19.39 6.20 150.84 2.68 78.16 74.58 553.79 17.38 −0.000 0.012 −0.24 −0.16 0.11 −0.09 0.001 0.007 0.058 −0.04 −0.13 −0.34 Commodities Cocoa 6.08 3.30 18.73 0.02 55.09 28.79 115.79 3.39 0.001 0.014 −0.35 −0.11 0.04 −0.06 0.002 0.002 0.27 −0.05 −0.10 −0.73 Coffee 6.50 3.81 51.79 0.01 30.72 15.76 236.39 0.20 0.001 0.013 −0.37 −0.06 −0.04 0.01 0.002 0.002 −0.01 −0.35 −0.11 0.22 Crude oil 78.21 43.72 230.35 0.51 289.57 157.53 641.07 5.24 0.002 0.010 −0.22 −0.32 −0.16 0.27 0.002 0.002 −0.12 0.09 −0.04 0.01 Gold 30.78 13.94 108.69 2.00 140.25 41.39 232.36 12.19 0.001 0.015 −0.32 −0.10 0.03 −0.12 0.001 0.003 −0.06 −0.18 0.01 −0.06 Heating oil 25.12 12.01 59.99 3.47 100.43 51.84 445.45 15.21 0.001 0.009 −0.37 −0.09 −0.04 0.05 0.001 0.005 −0.42 0.00 0.01 0.03 Platinum 3.28 1.91 18.86 0.00 17.48 5.02 87.15 0.00 0.000 0.016 −0.36 −0.06 −0.02 0.01 −0.000 0.003 −0.02 −0.03 −0.06 −0.02 Silver 17.49 7.40 57.74 3.46 86.97 20.26 261.63 26.88 0.000 0.015 −0.28 −0.13 −0.41 −0.06 0.000 0.004 −0.34 −0.00 −0.04 −0.08 The return is measured as weekly (Wednesday–Wednesday) return, in percentage. The market denotes the equal-weighted average returns of all the futures contracts. Trading volume and open interest are the weekly average of daily (Thursday–Wednesday) volume and open interest respectively, in 1000 contracts. ρj is the jth-order autocorrelation coefficient of return or change in trading activity for individual contracts. The change in trading volume (open interest) is the weekly change in average daily trading volume (open interest) that is detrended using its sample mean. * Indicates significance at the 0.05 level. Table options The result on the volume–return relation is also similar to that in equity markets. However, it is unlikely for the liquidity premium hypothesis to explain the informational role of volume in our highly liquid futures markets. Therefore, changes in trading volume are likely to capture overconfidence and overreaction and is thus positively associated with the magnitude of return reversals (DeBondt and Thaler, 1985; Odean, 1998; Statman and Thorley, 1999; Lee and Swaminathan, 2000; Gervais and Odean, 2001). Although there is no straightforward explanation for the relation between open interest and expected returns in the literature, our result tends to support Bessembinder and Seguin's (1993) argument that open interest represents uninformed trading by hedgers and is a proxy for market depth. Therefore, changes in volume conditional on an increase in open interest are, on average, associated with weaker price reversals than those conditional on a decrease in open interest. This is because an increase in market depth counterbalances the effect of overreaction on prices. In other words, speculators tend to move prices to a less extent in a deeper market than in a market with reduced depth. Therefore, the findings of the relation between trading activity (volume and open interest) and contrarian profits provides further support for the overreaction hypothesis. The remainder of this paper is organized as follows. Section 2 describes the data and the contrarian portfolio strategy used throughout the paper. The empirical evidence is presented in Section 3. Section 4 contains brief conclusions of the paper.

We employ the standard contrarian portfolio approach in equity market research to examine short-horizon return predictability in broader US futures markets. We find strong evidence of price reversals over the 1-week horizon, similar to that from equity market studies. The 1-week contrarain strategy remains profitable after corrections for plausible transaction costs. We also find that return predictability is substantially improved by interacting between past returns and lagged changes in volume and/or open interest. However, volume and open interest tend to 1356 C. Wang, M. Yu / Journal of Banking & Finance 28 (2004) 1337–1361 have the opposite effect on contrarian profits. There is a positive relation between lagged changes in volume and contrarian profits, whereas the relation between lagged changes in open interest and contrarian profits is negative. Furthermore, a contrarian strategy implemented in a portfolio of contracts that experienced an increase in volume together with a decrease in open interest generates the most pronounced profits over the short horizon. It is reasonable to assume that investors
perception of risk is not significantly changed over a short interval. Bid–ask bounce, lead–lag effect, and nonsynchronous trading problems do not appear to play a role in explaining the contrarian profits in our highly liquid futures markets. Therefore, our result of consistent contrarian profits points toward the direction of market inefficiency, that is, the market overreaction. This result is particularly interesting because futures trading is very close to the textbook model of perfect competition, and a majority of market participants are professional players. 17 The result on the relation between trading volume and contrarian profits is similar to that in equity market studies (Brennan et al., 1998; Hameed and Ting, 2000) and in general consistent with the implications of behavioral finance theories (DeBondt and Thaler, 1985; Odean, 1998; Gervais and Odean, 2001). This paper provides novel evidence on the negative relation between lagged changes in open interest and contrarian profits. This result appears to be consistent with the hypothesis that open interest is a proxy for market depth: A decrease (increase) in open interest is related to reduced (increased) market depth, and, consequently, the effect of overreaction on futures prices is larger (smaller), as are the contrarian profits. Therefore, our finding on the relation between trading activity and contrarian profits provides further support for the overreaction hypothesis. Although an overreaction phenomenon has been seen as the foundation of many existing regulatory measures such as daily price limits and ‘‘circuit breaker’’ systems in futures and stock markets, this paper is the first study that systematically examines overreaction in broader futures markets. This study also contributes to the literature by showing that lagged trading volume and open interest in addition to past prices correlate with future price changes in futures markets. Therefore, our findings have implications for market efficiency and futures market regulation and are also important for futures traders, commodity pool operators in particular