MAY–JUNE 2004 87

The Chinese Economy, vol. 37, no. 3, May–June 2004, pp. 87–122. © 2005 M.E. Sharpe, Inc. All rights reserved. ISSN 1097–1475 / 2005 $9.50 + 0.00.

GONGMENG CHEN, MICHAEL FIRTH, AND YU XIN

The Price-Volume Relationship in China’s Commodity Futures Markets Abstract: This study examines the relationship between returns and trading volume of four actively traded commodity futures contracts in China. Correlation analyses and Granger causality tests are used to investigate contemporaneous and lead-lag relationships between trading volume and both signed and absolute return. We find that the contemporaneous correlations between return and trading volume are not significantly different from zero, and there is no linearly significant causality following from trading volume to return or from return to trading volume. However, the contemporaneous correlations between absolute return and trading volume are significantly positive in all futures markets, and there is a significant relationship of causality following from absolute return to trading volume, which contradicts the mixture of distributions hypothesis and supports the sequential information arrival hypothesis in all of the futures markets examined except for aluminum futures. We also find a significant causality following from trading volume to absolute settlement-to-settlement return in the copper (subsample 1) futures market, but not in the copper (subsample 2) futures market. In the early 1980s, China set about reforming its stagnant economy and this involved, among other things, the adoption of market-oriented solutions to its economic reform. During the first ten years of the reform, China saw significant growth in output and some modernization of its industries. An obvious impediment to further sustainable growth was the lack of markets for financial assets and commodities. To remedy this shortcoming, China set up its first stock market and its first commodities exchange in late 1990. The ensuing decade has witnessed many developments in these markets as investors Gongmeng Chen is an associate professor in the School of Accounting and Finance, the Hong Kong Polytechnic University. Michael Firth is a professor in the School of Accounting and Finance, the Hong Kong Polytechnic University. Yu Xin is an assistant professor in the School of Business, Zhongshan University. 87

88

THE CHINESE ECONOMY

and regulators alike learn the intricacies of trading. Continuing improvements in the stock and commodities markets are vital if China is to reap the full rewards of its economic restructuring. The focus of this article is to examine the price-volume relationships of China’s commodities futures contracts. As in the developed markets of North America and Europe, commodities futures trading allows investors to hedge or speculate on futures prices and this activity can help in the price-discovery process. Return-volume studies are of interest as they may unearth dependencies that can form the basis of profitable trading strategies, and this has implications for market efficiency. A number of theories have been developed to explain the relationship between return and volume, and they have been subject to substantial empirical testing in the established markets of North America and Europe. However, the findings from these studies have been mixed and no strong consensus has emerged on issues such as whether returns lead volume or volume leads return. Even if consensus had emerged, the embryonic state of China’s commodities futures markets and its unique institutional features make foreign studies of limited relevance. Our study examines the relationship between returns and trading volume of four actively traded commodity futures contracts in China. Correlation analyses and Granger causality tests are used to investigate contemporaneous and lead-lag relationships between trading volume and both signed and absolute return. Based on the empirical results, we find that the contemporaneous correlations between return and trading volume are not significantly different from zero, and there is no linearly significant causality following from trading volume to return or from return to trading volume. However, the contemporaneous correlations between absolute return and trading volume are significantly positive in all futures markets. There is a significant relationship of causality following from absolute return to trading volume, which contradicts the mixture of distributions hypothesis and supports the sequential information arrival hypothesis in all of the futures markets examined except for aluminum futures. We also find a significant causality following from trading volume to absolute settlement-to-settlement return in the copper (subsample 1) futures market, but not in the copper (subsample 2) futures market. Theory and Literature Review This section briefly reviews the theoretical framework and the related empirical evidence on the relationships between return and volume, and between absolute return and volume from contemporaneous and causality (lead-lag) perspectives. Contemporaneous Relationships Between Return and Volume The contemporaneous relationship between return and trading volume helps reveal information about the symmetry of trading volumes in markets. Karpoff (1988) and Suominen (1996) argue that the relationship between trading volume and re-

MAY–JUNE 2004 89

turn in futures markets should not be affected by the sign of return, since the costs of taking short and long positions in futures markets are identical. Overwhelming empirical evidence exists that the contemporaneous correlation between return and volume is close to zero in futures markets (Karpoff 1987). For example, Kocagil and Shachmurove (1998) found no significant contemporaneous relationship between return and volume, thus confirming the symmetry of trading in futures markets. Lead-lag Relationship Between Return and Volume The lead-lag causality relationship between trading volume and return helps reveal the informational efficiency of futures markets. Blume, Easley, and O’Hara (1994) developed a model to investigate the informational role of volume and its applicability for technical analysis. They claimed that volume contains valuable information about the quality of information, that is, the precision of the signal of a price. Therefore, volume provides information that cannot be detected from price alone, and current trading volume can be used to predict future price movements. DeLong et al. (1990) proposed a noise-trading model, which predicts a positive feedback relationship between return and volume. The positive causality relationship running from return to volume is consistent with the positive-feedback trading strategy of noise traders who trade on the basis of past price changes, while the positive causality relationship from volume to return is consistent with the hypothesis that price change is caused by the trading strategies/actions of noise traders. The fact that past values of volume Granger-causes the actual return can be interpreted as evidence of informational inefficiency in futures markets. From an empirical perspective, many studies have found that there is a statistically significant linear causality running from the past values of return to trading volume (for example, Moosa and Al-Loughani, 1995; Kocagil and Shachmurove 1998), but past observations of trading volume do not increase the ability to forecast return in futures markets. On the other hand, Hiemstra and Jones (1994) provide evidence of significant nonlinear Granger causality from trading volume to stock returns. Following Hiemstra and Jones, Fujihara and Mougoue (1997) found a significant bidirectional nonlinear causality relationship between return and trading volume in three petroleum futures markets. Ciner (2002) found that nonlinear causality from volume to return disappears in the Tokyo commodity futures markets when the returns are adjusted for persistence in conditional volatility. Contemporaneous Relationship Between Absolute Return and Volume The basic supply and demand model1 (Crouch 1970; Clark 1973; and Westerfield 1977), the dispersion model2 (Epps and Epps 1976; Harris and Raviv 1993; and Shalen 1993), and the information asymmetry model3 (Wang 1994) are often employed to explain the positive contemporaneous relationship between volume and

90

THE CHINESE ECONOMY

absolute return. No matter which model is used, the positive contemporaneous relationship between absolute return and volume has been verified by overwhelming evidence (for example, Clark 1973; Cornell 1981; Tauchen and Pitts 1983; Grammatikos and Saunders 1986; Bessembinder and Seguin 1993; Foster 1995; Kocagil and Shachmurove 1998; and Ciner 2002). Generally, such a simultaneous relationship would imply that these markets are highly liquid, with traders being able to enter and exit as required. Lead-lag Relationship Between Absolute Return and Volume In contrast to the contemporaneous relationship, different theories predict different dynamic responses, namely, the lead-lag relationship. Specifically, the mixture of distributions hypothesis (Tauchen and Pitts 1983; and Harris 1984, 1986, and 1987) argues that price change and volume have a joint response to information due to their common distribution, which implies that trading volume and price change synchronously in response to new information. Then, the complete information equilibrium is immediately attained in a single trading round without any intermediate equilibrium. The implication is that with the mixture of distributions hypothesis, there is no information in the past absolute return that can be used in predicting future volume that is not already contained in past volume, and vice versa. In contrast, the sequential information arrival hypothesis (Copeland 1974 and 1976; Jennings, Starks, and Fellingham 1981) assumes that traders in a market receive new information in a sequential, random fashion. The information signal is observed separately by each individual, and trading occurs after each reception. When all traders have observed the information signal, a final complete equilibrium is established. Thus, there is a series of intertemporal equilibria before the final complete equilibrium is reached. This sequential information flow results in past values of trading volume having the ability to predict current absolute return and/or vice versa, which means that a causality relationship exists in both directions or either direction between absolute return and trading volume. Analyzing the relationship of causality between trading volume and price volatility (absolute return) can help us investigate speculation and its relationship with price volatility in the futures market, and it is extremely important for regulators in deciding upon the desirability of market restrictions. As Garcia, Leuthold, and Zapata (1986) have pointed out, increased trading volume may lead to increased price variability, which suggests a need for greater regulatory restrictions to curb speculative positions and daily trading activities; increased price volatility attracts more trading, which suggests that there should be less regulation of traders and their activities, and that further regulation may harm the price responsiveness of futures; trading volume and price volatility increase and decrease simultaneously, which indicates liquid and efficient markets where the response to new information is nearly instantaneous.

MAY–JUNE 2004 91

From an empirical perspective, Cornell (1981) found that the correlation between changes in price volatility and lead or lagged changes in volume was insignificant, which is consistent with the mixture of distributions hypothesis. Using causality tests, Rutledge (1979) examined four-month trading periods for 136 different futures contracts for thirteen commodities during the mid-1970s. He finds modest evidence that changes in price volatility may lead trading volume. Consistent with Smirlock and Starks (1988) and McCarthy and Najand (1993), Kocagil and Shachmurove (1998) found that price volatility Granger-causes trading volume in almost all of the futures markets examined. Herbert (1995) and Ciner (2002) found that lagged trading volume contains predictive power for current price volatility. These empirical results provide evidence against the mixture of distributions hypothesis and, instead, support the sequential information arrival hypothesis. A Review of Commodity Futures Markets in China After the abandonment of the planned economy, prices of most commodities and products were allowed to float according to supply and demand conditions. To facilitate the pricing of commodities, China began planning for formal commodities exchanges in the late 1980s, and the first one, the China Zhengzhou Grain Wholesale Market, was established in October 1990. The spot market quickly expanded to include futures transactions. Other exchanges opened shortly thereafter and the Shenzhen Metal Exchange, established in 1991, introduced the first standardized futures contract in China. The futures markets are still regarded as experimental by the Chinese authorities and so the government monitors the exchanges quite closely and has enacted various reforms that aimed to improve efficiency and reduce market manipulation. The devolution of power to cities and municipalities led to a profusion of exchanges, products, and brokerages. By the end of 1993 there were more than fifty commodity futures products, more than fifty exchanges, and in excess of 1,000 brokerages. The uncontrolled growth led to inefficient duplication of products and to cases of market manipulation and outright fraud. Faced with this situation, the Chinese government enacted regulations at the end of 1993 and the beginning of 1994 with the aim of curbing the malpractices in the system and making the markets more effective. After the regulations came into effect, the number of futures exchanges was reduced to fourteen, the exchanges became not-for-profit organizations, the number of commodities traded was reduced, restrictions were placed on investors, and a licensing system was introduced for brokers. The new laws, the establishment of clearer powers for the regulators, and the institutional changes outlined above combined to improve the credibility of the markets. While the reforms of 1993 and 1994 resulted in significant improvements in the effectiveness of commodity futures exchanges, market manipulation still persisted (Li 1999) and so the authorities launched a second round of reforms in late 1998 and early 1999. These reforms focused on enhancing market effectiveness

92

THE CHINESE ECONOMY

and efficiency. After the reforms, the number of exchanges fell to three, namely the Shanghai Futures Exchange (SHFE), the Zhengzhou Commodity Exchange (ZCE),4 and the Dalian Commodity Exchange (DCE). Commodity futures contracts were reduced to twelve, although only half of these have active markets. Higher capital requirements and tougher licensing examinations led to a reduction in the number of futures brokerage corporations to 213. The reforms expanded and clarified the legal framework for futures markets, and legal enforcement has been strengthened. Table 1 reports the detailed information of the two periods of adjustments. The reforms led to improvements in information disclosure, settlement procedures, and contract enforcement, as well as further standardization of contracts. In their zeal to dampen speculative trading, the authorities have taken actions that curtail liquidity and may inhibit the functions of futures markets. Among the actions taken by the authorities are the prohibition of using borrowed funds to finance trading, the prohibition of banks and similar institutions from involvement in futures (for example, acting as a principal, agent, and lender), and the prohibition of state-owned enterprises (SOEs) from participating in the market beyond the amount they invest in the spot market. The main investors in commodity futures are companies, SOEs, and individual investors. The relatively small lot size of contracts encourages smaller investors to participate, and this enhances liquidity at the exchanges. In many respects, the futures markets have been more successful than the spot markets (which are plagued by commercial disputes). The futures exchanges are electronic and they have borrowed the best features of exchanges in other countries. Research Design Data and Data Processing Four relatively active commodity futures—copper, aluminum, soybean, and wheat—are the subjects of the analysis, and copper futures are further investigated by dividing them into two subsamples. The sample periods are from January 4, 1999, to December 31, 2002, for aluminum and soybean futures; from January 4, 2000, to December 31, 2002, for wheat futures; from January 2, 1996, to December 31, 2002, for copper futures; from January 2, 1996, to December 31, 1998, for the first subsample of copper futures; and from January 4, 1999, to December 31, 2002, for the second subsample of copper futures. Summary statistics for China’s commodity futures markets are shown in Table 2, which shows some volatility in the turnover of different futures contracts. Natural rubber futures declined precipitously in 1998 although there is some recovery in 2002. Green bean (mung bean) futures have fallen to virtually nothing in the last few years. The four products we focus on in our study (copper, aluminum, soybean, and wheat) have had significant trading volumes since 2000, although

Table 1 China’s Commodity Futures Markets: Exchanges and Products Panel A: Commodity Futures Exchanges

Number Name

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

>50

>50c

15

14

14

14

3

3

3

3

Zhengzhou Commodity Exchange (ZCE)

Dalian Commodity Exchange (DCE) Shanghai Cereals and Oils Exchange (SCOE) Shanghai Commodity Exchange (SHCE) Shanghai Metal Exchange (SHME) Beijing Commodity Exchange (BCE) China Commodity Futures Exchange (CCFE) Chongqing Commodity Exchange (CQCE) Suzhou Commodity Exchange (SCE) Shenyang Commodity Exchange (SYCE) Shenzhen Metal Exchange (SME) etc.

Dalian Commodity Exchange (DCE) Shanghai Cereals and Oils Exchange (SCOE) Shanghai Commodity Exchange (SHCE Shanghai Metal Exchange (SHME) Beijing Commodity Exchange (BCE) China Commodity Futures Exchange (CCFE)

Zhengzhou Commodity Exchange (ZCE) Dalian Commodity Exchange (DCE) Shanghai Futures Exchange (SHFE)a

Chongqing Commodity Exchange (CQCE) Chengdu United Futures Exchange (CUFE) Guangdong United Futures Exchange (GUFE) Suzhou Commodity Exchange (SCE) Shenyang Commodity Exchange (SYCE) Shenzhen Metal Exchange (SME) Tianjin United Futures Exchange (TUFE) Changchun United Futures Exchange (CCUFE)b (continued )

MAY–JUNE 2004 93

Zhengzhou Commodity Exchange (ZCE)

94

Table 1 (continued)

Panel B: Available Commodity Futures Exchanges

Number Name

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

>50

>50c

35

35

35

35

12

12

12

12

Soybean, soybean meal, broomcorn, red bean, coffee, cocoa, green bean, beer barley, natural rubber, wheat, corn, palm oil, long-grain rice, colza meal, tung oil, aluminum, nickel, copper, tin, veneer, steel, sugar, coal, rice, colza oil, petroleum, etc.

Soybean, soybean meal, broomcorn, red bean, coffee, cocoa, green bean, beer barley, natural rubber, wheat, corn, palm oil, long-grain rice, colza meal, tung oil, aluminum, nickel, copper, tin, veneer, etc.

ZCE: wheat, green bean, red bean, and peanut kernel DCE: soybean, soybean meal, and beer barley SHFE: copper, aluminum, natural rubber, plywood, and long-grain rice

Sources: Rearrangements based on materials from China Futures Yearbook (1995); China Securities and Futures Statistical Yearbook (1998– 2002); China Securities Regulatory Commission 1999; Li 1999; Zhu 1999; and www.cfachina.org. Notes: a In April 1999 the SHFE was established after the reorganization and merging of the SCOE, SHCE, and SHME. In October 1995 Changchun United Futures Exchange (CCUFE) was closed due to irregular operations. c At the end of 1994, the actual number of actively traded commodities was much less than the reported number. b

THE CHINESE ECONOMY

China’s Commodity Futures Markets: Exchanges and Products

MAY–JUNE 2004 95

soybean and copper clearly dominate. In fact, soybean futures traded on the DCE and copper futures traded on the SHFE record the second highest volumes after the CBOT and LME, respectively. Appendix 1 gives details of the four contracts we examine. The nearby futures closing and settlement prices are selected in this study. In the meantime, to obtain an aggregate measure of trading activity in each market, volume is summed up across all outstanding contracts for each trading day. The Granger causality test requires that the related variables should be stationary. Since previous studies have found strong evidence of both linear and nonlinear time trends in raw trading volume series (Gallant, Rossi, and Tauchen 1992; Lee and Rui 2001), the raw trading volume series needs to be detrended to achieve stationarity. The following regression is conducted to detrend the raw volume series.

LOG ( ALLVOLUME ) = c + α1 TREND + α 2 TREND * TREND + ε t

(1)

where LOG(ALLVOLUME) is the natural log of the daily total trading volume (1,000 lots) for all contracts; TREND is the number of observations; and the residuals are the detrended trading volume series (DEVOLUME), which will be used in following empirical tests. Here TREND is used to model the linear time trend, and TREND*TREND is used to model the nonlinear time trend. In this study, if we find that the coefficients of TREND and/or TREND*TREND are not significant, the related terms will be excluded to obtain a new OLS regression, which will be used to filter the raw volume series. The detailed results of detrended regressions for raw trading volume are listed in Table 3.5 Generally, we find that linear and nonlinear time trends both exist in the copper futures market, and that a linear time trend exists in the aluminum and soybean futures markets. There is no linear and nonlinear time trend in the wheat futures market. Table 4 reports the results of ADF tests for the raw volume and detrended volume series. For all of the futures markets that were examined, the null hypotheses of nonstationarity are significantly rejected for detrended volume, while sometimes the null hypothesis for raw volume cannot be consistently rejected in five percent significant level. Therefore, it is appropriate to obtain stationary series through detrending the raw volume series. Table 5 presents the results of the ADF test for close-to-close and settlement-tosettlement returns, and close-to-close and settlement-to-settlement absolute returns.6 For all of the futures markets examined, the null hypotheses of nonstationarity are significantly rejected for both the return and absolute return series. Contemporaneous Correlation Test Two kinds of contemporaneous correlation, the Pearson correlation and the Spearman rank correlation, between return and trading volume, and between ab-

96

Table 2

Panel A: Yearly Turnover 1993 Yearly total turnover

770

1994 >6000

1995

1996

1997

9240.8

8411.9

6117.1

402.1 / 226.8 b 19.2 / 3.2 b 1276.8 742.0 79.5 2.3 2853.0 89.5 1274.2 na 1.9 96.2 551.8

398.2 / 271.9 b 217.2 / 10.7 b 777.7 1049.8 148.7 13.3 2487.0 1.1 81.5 na 0.4 41.9 488.8

1998

1999

2000

2001

2002

3298.6 a

2234.7

1607.3

3015.4

3948.1

464.8

424.9

503.6

651.1

918.8

8.3

38.5

73.1

199.5

319.2

30.9 675.5 na 78.6 2029.7 1.7 3.9 na na na na

28.1 642.2 – 8.7 1091.8 0.5 – – – – –

88.5 760.6 21.3 156.1 4.1 – – – – – –

5.3 1912.1 63.9 183.5 ≈0 – – – – – –

401.7 1925.5 157.7 225.2 ≈0 – – – – – –

Panel B: Turnover in Terms of Futures Products Copper

na

na

Aluminum

na

na

Natural rubber Soybean Soybean meal Wheat Green bean Long-grain rice Veneer Peanut kernel Corn Broomcorn Red bean

na na na na na na na na na na na

na na na na na na na na na na na

683.3 / 407.6 b 135.1 / 48.0 b 772.6 223.4 67.1 11.0 2738.6 na 3109.4 na 637.7 na 414.6

THE CHINESE ECONOMY

China’s Commodity Futures Markets: Yearly Turnover and Turnover in Terms of Futures Products (billion yuan)

Coffee Beer barley Palm oil Others

na na na na

na na na na

65.9 24.4 151.3 206.4

918.0 90.9 13.6 0.9

409.3 0.2 2.0 ≈0

na na na 5.2

– – – –

– – – –

– – – –

– – – –

Sources: Rearrangements based on the materials from China Futures Yearbook (1995); China Securities and Futures Statistical Yearbook (1998–2002); Zhu 1999; Futures Daily (bound edition, 1999); www.csrc.com.cn; and www.cfachina.org. Notes: a The data are incomplete here since only the turnovers of ZCE, DCE, SCOE, SHCE, and SHME are included. The turnovers of other exchanges are not included due to their closure after September 1998. b These data are the turnover from SHME.

MAY–JUNE 2004 97

98

The Results of Detrended Regressions for Raw Volume Series Copper (all-sample)

Copper (subsample 1)

Copper (subsample 2)

Aluminum (all-sample)

Soybean (all-sample)

Wheat (all-sample)

Coefficient (t-value)

Coefficient (t-value)

Coefficient (t-value)

Coefficient (t-value)

Coefficient (t-value)

Coefficient (t-value)

C

1.650699 (36.00872)***

1.307664 (16.99048)***

TREND

0.001757 (14.16433)***

0.003622 (7.510610)***

0.000803 (3.267544)***

0.003515 (10.29034)***

TREND*TREND

–3.20E-07 (–4.554471)***

–1.75E-06 (–2.766765)***

5.36E-07 (2.184909)**

Variables

R Square

0.475640

0.342660

2.670619 (51.67322)***

0.326759

–0.134376 (–1.890408)*

4.470977 (73.42605)***

3.494225 (52.25392)***

0.001349 (4.630026)***

0.000463 (1.041529)

1.43E-07 (0.416114)

3.71E-07 (1.267737)

7.81E-07 (1.260722)

0.657025

0.364695

0.108044

Notes: LOG(ALLVOLUME) is the natural log of the daily total trading volume (1,000 lots) across all contracts for the related futures product; TREND is the number of observations. * significant at the 10 percent level; ** significant at the 5 percent level; *** significant at the 1 percent level.

THE CHINESE ECONOMY

Table 3

MAY–JUNE 2004 99

Table 4 ADF Test for the Volume and Detrended Volume Series LOG(ALLVOLUME) (with constant) ADF value Copper

All-sample (1996–2002)

Subsample 1 (1996–98)

Subsample 2 (1999–2002)

DEVOLUME (with constant) ADF value

Lag 5 Lag 10 Lag 15 Lag 20 Lag 5 Lag 10 Lag 15 Lag 20 Lag 5 Lag 10 Lag 15 Lag 20

–5.945960*** –3.944368*** –3.204142** –2.823842* –4.718275*** –3.215168** –2.842379* –2.476234 –5.961207*** –4.056679*** –3.074064** –2.709030*

–10.44399*** –7.335417*** –6.024102*** –5.157085*** –6.978020*** –5.035614*** –4.525588*** –3.817800*** –9.458976*** –6.842846*** –5.228225*** –4.584215***

Aluminum

All-sample (1999–2002)

Lag 5 Lag 10 Lag 15 Lag 20

–2.950652** –2.500955 –2.280528 –2.209501

–5.325152*** –4.083281*** –3.635678*** –3.577745***

Soybean

All-sample (1999–2002)

Lag 5 Lag 10 Lag 15 Lag 20

–5.081443*** –3.438250** –2.709105* –2.172108

–7.390976*** –5.340795*** –4.457031*** –3.987804***

Wheat

All-sample (2000–2)

Lag 5 Lag 10 Lag 15 Lag 20

–5.551519*** –4.185197*** –3.738460*** –2.825217*

–5.551519*** –4.185197*** –3.738460*** –2.825217*

Notes: LOG(ALLVOLUME) is the natural log of the daily total trading volume (1,000 lots) across all contracts for the related futures product; DEVOLUME is the detrended trading volume series; DEOPIN is the detrended open interest series. For ADF test with constant, the critical values are –2.57, –2.86, and –3.44 at the 10 percent, 5 percent, and 1 percent significant levels, respectively. The null hypothesis that the series is nonstationary is rejected if the test statistic is greater than the critical value. * significant at the 10 percent level; ** significant at the 5 percent level; *** significant at the 1 percent level.

100

THE CHINESE ECONOMY

Figure 1 Null hypothesis test using F statistic Null hypotheses: β1 = β2 = . . . = βm = 0; i.e., x(y) does not cause y(x) Equation 4-2

Equation 4-3

Conclusions

Reject H0

Reject H0

Reject H0 H0 cannot be rejected H0 cannot be rejected

H0 cannot be rejected Reject H0 H0 cannot be rejected

Feedback relationship between x and y x causes y y causes x No relationship between x and y (independent)

solute return and trading volume, are calculated. Based on the above analysis, the contemporaneous correlation between return and volume should be zero due to the absence of trading cost asymmetry in futures markets, and the contemporaneous correlation between absolute return and trading volume should be significantly positive. Granger Causality Test Granger (1969) developed a methodology to examine whether changes in one series cause changes in another. If the current value of a time series, y, can be better predicted by using the past value of another time series, x, than by not doing so, and considering other relevant information including the past values of y, we may conclude that x causes y (Pindyck and Rubinfeld 1998, 243). The following two OLS regressions are used in the Granger causality test: m

yt = α0 + ∑ α i y t −i + i =1 m

xt = α0 +

∑α x i

i =1

m

∑β x i

i =1

t −i

+ εt

(2)

+ εt

(3)

m

t −i

+

∑β y i

i =1

t −i

The F statistic is calculated to test the null hypothesis. The related conclusions are listed in Figure 1. As Gujarati has pointed out (1995, 623), the Granger causality test is very sensitive to the number of lags used in the analysis. Thus, Davidson and Mackinnon (1993) have suggested using more rather than fewer lags. From a practical viewpoint, if the Granger causality test is not very sensitive to the lag length, we will have more confidence in our conclusions than if the results are very sensitive to the lag length. Therefore, in this study we report the results of the Granger causality test using three methods. The first employs the lags determined by the AIC (Akaike Information Criterion) and FPE (Final Prediction Error) critiques; 7 the

Table 5 ADF Test with Constant for the Return and Absolute Return Series

Copper

All-sample (1996–2002)

Subsample 1 (1996–98)

Subsample 2 (1999–2002)

Aluminum

Settlement-tosettlement return ADF value

Absolute close-toclose return ADF value

Absolute settlement-tosettlement return ADF value

Lag 5 Lag 10 Lag 15 Lag 20 Lag 5 Lag 10 Lag 15 Lag 20 Lag 5 Lag 10 Lag 15 Lag 20

–15.76983*** –10.89472*** –9.477387*** –8.853226*** –10.10844*** –6.832493*** –5.467781*** –5.654503*** –12.24358*** –8.682189*** –8.200146*** –6.957401***

–15.54872*** –10.97230*** –9.455237*** –8.906230*** –9.951283*** –6.937163*** –5.512255*** –5.724517*** –12.10984*** –8.661332*** –8.134053*** –6.944219***

–12.47691*** –8.690913*** –7.451506*** –6.770912*** –7.465509*** –5.396135*** –4.870229*** –4.460350*** –10.54105*** –7.255427*** –5.786084*** –5.008057***

–13.11204*** –9.126439*** –7.520050*** –7.002851*** –7.973142*** –5.936722*** –5.264226*** –4.867491*** –10.82742*** –7.233469*** –5.454933*** –4.782662***

Lag 5 Lag 10 Lag 15 Lag 20

–12.04447*** –7.872839*** –6.408500*** –5.443323***

–11.92862*** –7.838144*** –6.327634*** –5.566321***

–9.305675*** –5.762358*** –4.479372*** –3.790225***

–9.500588*** –5.709643*** –4.798337*** –3.939969***

(continued)

MAY–JUNE 2004 101

All-sample (1999–2002)

Close-to-close return ADF value

102

Table 5 (continued)

Close-to-close return ADF value

Settlement-tosettlement return ADF value

Absolute close-toclose return ADF value

Absolute settlement-tosettlement return ADF value

Soybean

All-sample (1999–2002)

Lag 5 Lag 10 Lag 15 Lag 20

–12.16677*** –8.237777*** –7.312682*** –6.068303***

–12.15435*** –8.247613*** –7.332347*** –5.955664***

–12.14458*** –8.977853*** –6.956143*** –6.328290***

–11.82785*** –8.699855*** –7.029278*** –6.557952***

Wheat

All-sample (2000–2002)

Lag 5 Lag 10 Lag 15 Lag 20

–11.12655*** –8.578416*** –7.093146*** –5.997040***

–11.04383*** –8.424210*** –7.105650*** –5.888453***

–10.38328*** –7.659530*** –6.337063*** –5.305653***

–10.75606*** –7.791635*** –6.335503*** –5.399861***

Notes: Close-to-close Return is the first difference of the LOG(NCLOSEP); Settlement-to-settlement Return is the first difference of the LOG(NSETTLEP); Absolute Close-to-close Return is the absolute value of Close-to-close Return; Absolute Settlement-to-settlement Return is the absolute value of Settlement-to-settlement Return. For ADF test with constant, the critical values are –2.57, –2.86, and –3.44 at the 10 percent, 5 percent, and 1 percent significant levels, respectively. The null hypothesis that the series is nonstationary is rejected if the test statistic is greater than the critical value. * significant at the 10 percent level; ** significant at the 5 percent level; *** significant at the 1 percent level.

THE CHINESE ECONOMY

ADF Test with Constant for the Return and Absolute Return Series

MAY–JUNE 2004 103

second employs the lags determined by the SC (Schwarz Information Criterion) critique; and the third employs the lags with 5, 10, and 15.8 By comparing the empirical results of these three methods together, a more robust and appropriate conclusion can be obtained. In this study the linear causality relationship between return and trading volume is examined in China’s commodity futures markets. If there is significant causality running from the past value of trading volume to return, this means that informational efficiency is not achieved, and technical analysis can be used to obtain abnormal profits. In the meantime, the linear causality relationship between absolute return and trading volume is examined to ascertain whether the mixture of distributions hypothesis or the sequential information arrival hypothesis is supported. If there is no significant causality from past values of absolute return to trading volume, and vice versa, the mixture of distributions hypothesis is supported. This would suggest that there is no information in past observations of absolute return that can be used to forecast trading volume, and vice versa. Further, this implies that in China’s futures markets, the dissemination of information is contemporaneous, and the final equilibrium is immediately attained. However, if there is significant causality from the past values of absolute return to trading volume and/or vice versa, the sequential information arrival hypothesis will be supported, which will suggest that there are several intermediate equilibria before the final complete informational equilibrium is reached. In addition, the test of causality between absolute return and trading volume allows us to investigate speculation in China’s commodity futures markets. On the one hand, increased trading volume that brings about increased absolute returns may suggest that speculative activity is relatively high and that imposing further regulation is necessary. On the other hand, increased absolute returns that lead to greater trading volumes may suggest a more efficient market and less regulation. Empirical Results Contemporaneous Relationship Between Return and Trading Volume The contemporaneous correlations between return and trading volume are reported in panel A of Table 6. At the 5 percent significant level, based on the Pearson and Spearman correlations no significant contemporaneous relationship exists between close-to-close return and volume for all of the futures markets examined except for copper (all-sample) futures. No significant contemporaneous relationship exists between settlement-to-settlement return and volume for copper (subsample 2), aluminum, and wheat futures; and a significant negative contemporaneous relationship exists between settlement-to-settlement return and volume for copper (all-sample), copper (subsample 1) and soybean futures in the Spearman correlation test. The above facts mean that the short mechanism works rationally and normally in the copper (subsample 2), aluminum, and wheat futures markets, but irratio-

104

Contemporaneous Correlation Between Return and Detrended Volume, and Between Absolute Return and Detrended Volume Panel A: Return vs. Detrended Volume Copper (all-sample)

Copper (subsample 1)

Close-to-close return vs. detrended volume Settlement-to-settlement return vs. detrended volume

–0.050** (0.039) –0.039 (0.110)

–0.065* (0.078) –0.066* (0.073)

Close-to-close return vs. detrended volume Settlement-to-settlement return vs. detrended volume

–0.057** (0.019) –0.058** (0.017)

–0.065* (0.076) –0.080** (0.030)

Copper (subsample 2)

Aluminum (all-sample)

Soybean (all-sample)

Wheat (all-sample)

Pearson Correlation –0.025 (0.430) 0.001 (0.968)

–0.014 (0.666) 0.000 (0.997)

–0.052 (0.110) –0.048 (0.139)

–0.036 (0.350) –0.029 (0.442)

Spearman Correlation –0.025 (0.439) –0.013 (0.678)

–0.024 (0.456) –0.009 (0.772)

–0.063* (0.052) –0.082** (0.011)

–0.059 (0.118) –0.067* (0.079)

THE CHINESE ECONOMY

Table 6

Panel B: Absolute Return vs. Detrended Volume Copper (all-sample)

Absolute close-to-close return vs. detrended volume Absolute settlement-tosettlement return vs. detrended volume Absolute close-to-close return vs. detrended volume Absolute settlement-tosettlement return vs. detrended volume

Copper (subsample 1)

Copper (subsample 2)

Aluminum (all-sample)

Soybean (all-sample)

Wheat (all-sample)

0.461*** (0.000)

0.528*** (0.000)

Pearson Correlation 0.409*** (0.000)

0.182*** (0.000)

0.146*** (0.000)

0.477*** (0.000)

0.554*** (0.000)

0.420*** (0.000)

0.174*** (0.000)

0.147*** (0.000)

0.528*** (0.000)

0.568*** (0.000)

0.152*** (0.000)

0.272*** (0.000)

0.125*** (0.001)

0.564*** (0.000)

0.604*** (0.000)

0.158*** (0.000)

0.322*** (0.000)

0.122*** (0.001)

Spearman Correlation 0.514*** (0.000) 0.547*** (0.000)

0.011 (0.777)

MAY–JUNE 2004 105

Notes: Two-tailed p values for the related correlations are reported in brackets. * significant at the 10 percent level; **significant at the 5 percent level; *** significant at the 1 percent level.

0.020 (0.605)

106

THE CHINESE ECONOMY

nally and abnormally in the copper (subsample 1) and soybean futures markets. The asymmetry of trading volume exists in the copper (subsample 1) and soybean futures markets.9 When settlement price decreases, trading volume increases significantly. The negative relationship may imply that the short mechanism works irrationally and abnormally and dominates the long mechanism. Comparing the results of two subsamples of copper futures, we find that during 1999–2002, the short mechanism of copper futures worked more rationally and normally than it did during 1996–98, which means that the asymmetry of trading volume has improved in the copper futures market during the sample period. Lead-lag Relationship Between Return and Trading Volume The results of the Grange causality test between return and trading volume are reported in Table 7 and Table 8, and are summarized in Table 11. Generally, we cannot find any consistently significant linear causality following from volume to return or from return to volume in all examined futures markets for all three methods. This fact implies that volume has no linear information content to help generate abnormal profits, and does not provide any evidence that technical analysis will be effective. Contemporaneous Relationship Between Absolute Return and Trading Volume The contemporaneous correlations between absolute return and trading volume are reported in panel B of Table 6. Based on the Pearson and Spearman correlations and at the 5 percent significant level, the contemporaneous correlation between absolute return and trading volume is significantly positive in all of the futures markets that were examined except for wheat futures in the Pearson correlation test. Generally, this empirical result is consistent with models such as the supply and demand model, the dispersion model, and the information asymmetry model. Comparing the correlations between absolute return and trading volume between two subsamples of copper futures, we find that the correlations decline during the sample period. Based on Wang’s information asymmetry model (Wang 1994), this fact implies that information asymmetry decreases and market efficiency improves during the sample period in the copper futures market because the correlation between absolute return and trading volume increases with information asymmetry. Comparing the correlations between copper and aluminum, and between soybean and wheat futures, we find that the information asymmetry of copper and soybean futures is higher than that of aluminum and wheat futures, which can be attributed to the presence of more participants with heterogeneous information.

Table 7 Linear Granger Causality Test Between Return and Detrended Volume Based on the AIC, FPE, and SC Critiques Panel A: Based on the AIC (Akaike Information Criterion) and FPE (Final Prediction Error) Critiques Close-to-close return vs. detrended volume

Settlement-to-settlement return vs. detrended volume

From volume to return

From return to volume

Lags

From volume to return

From return to volume

All-sample (1996–2002) Subsample 1 (1996–98) Subsample 2 (1999–2002)

9

0.77658

0.71172

9

1.02359

0.74387

3

0.23126

0.45165

3

0.40932

0.16574

5

1.74482

1.95862

5

1.10875

2.19396

Aluminum

All-sample (1999–2002)

9

1.68818

1.38694

9

1.94979**

1.24111

Soybean

All-sample (1999–2002)

4

1.46251

0.20551

3

2.31126

0.05960

Wheat

All-sample (2000–2)

4

1.31421

1.06880

4

1.84630

1.09706

Copper

(continued)

MAY–JUNE 2004 107

Lags

108

Panel B Based on the SC (Schwarz Information Criterion) Critique Close-to-close return vs. detrended volume

Settlement-to-settlement return vs. detrended volume

Lags

From volume to return

From return to volume

Lags

From volume to return

From return to volume

All-sample (1996–2002) Subsample 1 (1996–98) Subsample 2 (1999–2002)

2

0.11879

0.52392

2

0.67580

0.31613

1

0.29594

0.16680

1

0.04183

0.17476

1

0.15588

1.13674

1

0.01609

0.38303

Aluminum

All-sample (1999–2000)

4

1.36525

1.23794

4

0.67359

1.80331

Soybean

All-sample (1999–2002)

2

2.26118

0.42868

2

3.35798**

0.20228

Wheat

All-sample (2000–2)

1

2.93492

1.42261

1

2.41891

1.16154

Copper

Notes: F statistics are reported here; * significant at the 10 percent level (not reported here); ** significant at the 5 percent level; *** significant at the 1 percent level.

THE CHINESE ECONOMY

Table 7 (continued)

Table 8 Linear Granger Causality Test Between Return and Detrended Volume Based on the Lags 5, 10, and 15 Close-to-close return vs. detrended volume From volume to return Copper

All-sample (1996–2002) Subsample 1 (1996–98) Subsample 2 (1999–2002) All-sample (1999–2002)

Soybean

All-sample (1999–2002)

Wheat

All-sample (2000–2)

0.331393 0.730975 0.642631 0.455184 0.951632 0.787862 1.744816 1.089396 1.031048 1.093063 1.498299 1.564683 1.192698 1.103504 0.956662 1.121400 1.338079 1.275280

0.518260 0.641688 0.559328 0.449333 0.491967 1.190331 1.958619 1.586119 1.593152 1.024649 1.297273 1.690346** 0.821008 0.767453 0.754352 1.034057 0.803234 1.448009

From volume to return 0.333296 0.934696 0.856847 0.814817 1.357071 1.202592 1.108755 0.768905 0.882532 0.501442 1.705495 1.872231** 1.384029 1.199662 1.244629 1.439820 1.048911 1.044612

From return to volume 0.494593 0.682297 0.571612 0.141576 0.375215 1.111652 2.193961 1.633980 1.498138 1.457517 1.222410 1.433560 0.581047 0.676898 0.771268 1.074678 0.808346 1.462815

Notes: F statistics are reported here; * significant at the 10 percent level (not reported here); ** significant at the 5 percent level; *** significant at the 1 percent level.

MAY–JUNE 2004 109

Aluminum

Lag 5 Lag 10 Lag 15 Lag 5 Lag 10 Lag 15 Lag 5 Lag 10 Lag 15 Lag 5 Lag 10 Lag 15 Lag 5 Lag 10 Lag 15 Lag 5 Lag 10 Lag 15

From return to volume

Settlement-to-settlement return vs. detrended voume

110

Table 9

Panel A: Based on the AIC (Akaike Information Criterion) and FPE (Final Prediction Error) Critiques Absolute close-to-close return vs. detrended volume

Absolute settlement-to-settlement return vs. detrended voume

Lags

From volume to absolute return

From absolute return to volume

Lags

From volume to absolute return

From absolute return to volume

All-sample (1996–2002) Subsample 1 (1996–98) Subsample 2 (1999–2002)

15

1.41590

3.41268***

15

1.54781

2.16539***

15

1.35710

2.23250***

10

2.38515***

1.30630

15

1.62951

1.89981**

18

1.26516

1.15656

Aluminum

All-sample (1999–2002)

14

0.80317

1.29669

10

0.83407

1.05332

Soybean

All-sample (1999–2002)

15

0.88238

1.90765**

13

1.20139

1.61199

Wheat

All-sample (2000–2)

18

2.10944***

2.13255***

8

2.17910**

1.54223

Copper

THE CHINESE ECONOMY

Linear Granger Causality Test Between Absolute Return and Detrended Volume Based on the AIC, FPE, and SC Critiques

Panel B: Based on the SC (Schwarz Information Criterion) Critique

Copper

All-sample (1996–2002) Subsample 1 (1996–98) Subsample 2 (1999–2002)

8

1.59592

5.05444***

8

2.09405**

2.91796***

4

1.82997

4.64379***

4

3.31771**

2.08900

5

1.90217

3.62245***

5

2.28007**

2.68641**

Aluminum

All-sample (1999–2002)

5

1.28713

1.43727

4

1.41608

0.81084

Soybean

All-sample (1999–2002)

4

0.53205

4.25630***

5

1.61581

2.54040**

Wheat

All-sample (2000–2)

2

1.01046

4.94000***

2

1.76113

3.81457**

Notes: F statistics are reported here; * significant at the 10 percent level (not reported here); ** significant at the 5 percent level; *** significant at the 1 percent level.

MAY–JUNE 2004 111

112

Table 10

Absolute close-to-close return vs. detrended volume From volume to absolute return Copper

Aluminum

From absolute return to volume

Absolute settlement-to-settlement return vs. detrended voume From volume to absolute return

From absolute return to volume

All-sample (1996–2002)

Lag 5 Lag 10 Lag 15

2.235291** 1.439310 1.415902

6.373156*** 3.855749*** 3.412683***

3.299978*** 1.828203 1.547810

3.402339*** 2.386911*** 2.165390***

Subsample 1 (1996–98)

Lag 5 Lag 10 Lag 15

1.585168 1.573586 1.357098

4.361544*** 2.520393*** 2.232500***

2.631581** 2.385145*** 1.777910**

1.909160 1.306304 1.186552

Subsample 2 (1999–2002)

Lag 5 Lag 10 Lag 15

1.902174 1.991423** 1.629514

3.622450*** 2.439831*** 1.899812**

2.280074** 2.068806** 1.492020

2.686410** 1.698677 1.312642

All-sample (1999–2002)

Lag 5 Lag 10 Lag 15

1.287133 1.033229 0.796915

1.437273 0.775732 1.187260

0.936202 0.834067 0.758732

1.539644 1.053324 1.014161

THE CHINESE ECONOMY

Linear Granger Causality Test Between Absolute Return and Detrended Volume Based on the Lags 5, 10, and 15

Soybean

All-sample (1999–2002)

Wheat

All-sample (2000–2002)

Lag 5 Lag 10 Lag 15 Lag 5 Lag 10 Lag 15

1.047262 0.728783 0.882377 0.696865 1.395079 1.409867

3.291560*** 2.288313** 1.907646** 2.623888** 2.227784** 1.619870

1.615806 0.990241 1.143549 1.136712 1.848521** 1.451986

2.540399** 1.892346** 1.498381 1.641238 1.774947 1.472116

Notes: F statistics are reported here; * significant at the 10 percent level (not reported here); ** significant at the 5 percent level; *** significant at the 1 percent level.

MAY–JUNE 2004 113

114

Table 11

Return vs. detrended volume AIC and FPE

SC

AIC and FPE

SC

Lags 5, 10, 15

C

NO

NO

NO

AR → VOL

AR → VOL

AR → VOL

C

NO

NO

NO

AR → VOL

AR → VOL

AR → VOL

C

NO

NO

NO

AR → VOL

AR → VOL

AR → VOL

Aluminum All-sample (1999–2002)

C

NO

NO

NO

NO

NO

NO

Soybean

All-sample (1999–2002)

C

NO

NO

NO

AR → VOL

AR → VOL

AR → VOL

Wheat

All-sample (2000–2)

C

NO

NO

NO

AR → VOL

AR → VOL

AR → VOL

Copper

All-sample (1996–2002) Subsample 1 (1996–98) Subsample 2 (1999–2002)

S

NO

NO

NO

AR → VOL

AR → VOL

AR → VOL

S

NO

NO

NO

VOL → AR

VOL → AR

VOL → AR

S

NO

NO

NO

NO

AR → VOL

VOL → AR

Copper

All-sample (1996–2002) Subsample 1 (1996–98) Subsample 2 (1999–2002)

Lags 5, 10, 15

Absolute return vs. detrended volume

THE CHINESE ECONOMY

Comments on the Empirical Results of Tables 7, 8, 9, and 10

Aluminum All-sample (1999–2002)

S

VOL → R

NO

NO

NO

NO

NO

Soybean

All-sample (1999–2002)

S

NO

VOL → R

NO

NO

AR → VOL

AR → VOL

Wheat

All-sample (2000–2)

S

NO

NO

NO

VOL → AR

AR → VOL

NO

Notes: Five percent significant level is employed in this table; C means close-to-close; S means settlement-to-settlement; VOL means detrended volume; R means return; AR means absolute return; NO means no causality relationship; → means the direction of the related causality.

MAY–JUNE 2004 115

116

THE CHINESE ECONOMY

Lead-lag Relationship Between Absolute Return and Trading Volume The results of the Grange causality test between absolute return and trading volume are reported in Table 9 and Table 10, and are summarized in Table 11. Generally, for close-to-close absolute return, consistently significant causality followed for all three methods from absolute return to trading volume in all of the futures markets that were examined except for aluminum futures; the reverse causality relationship does not consistently hold for these markets; and there is no significant relationship of causality between absolute return and trading volume in the aluminum futures market. The evidence of a significant causality from absolute return to trading volume contradicts the mixture of distributions hypothesis and supports the sequential information arrival hypothesis. Thus, in China’s copper, soybean, and wheat futures markets, new information is absorbed sequentially, and the intermediate informational equilibrium is reached before the final equilibrium is found; while in China’s aluminum futures market, the mixture of distributions hypothesis is supported, which means that the dissemination of information is contemporaneous, and the final equilibrium is immediately reached in the aluminum futures market. For absolute settlement-to-settlement returns, we cannot find consistent results among the three methods in the copper (subsample 2), soybean, and wheat futures markets, which may be due to the special characteristics of settlement price.10 We also find the same empirical results as in close-to-close absolute return in the aluminum and copper (all-sample) futures markets. One interesting finding relates to the copper (subsample 1) futures market. There is a consistently significant causality following from trading volume to absolute settlement-to-settlement return in this market, which is definitely different from the results based on close-to-close absolute return. This relationship of causality from volume to absolute settlement-to-settlement return holds only for the first subsample (1996–98), and not for the second subsample (1999–2002). The fact that increased trading volume causes increased absolute return may suggest that speculative activity was relatively high during the first subsample period. Then, because of the effectiveness of the second more powerful adjustment by the authorities, this relationship of causality disappeared and speculation was reduced during the second subsample period. Conclusions Based on the empirical results, the basic conclusions of this article are as follows: (1) Generally, the contemporaneous correlations between return and trading volume are not significantly different from zero, due to the absence of trading cost asymmetry and the effectiveness of short mechanisms in China’s commodity futures markets, except for the Spearman correlations between settlement-to-settlement return and trading volume in the copper (subsample

MAY–JUNE 2004 117

1) and soybean futures markets. For the copper futures market, the asymmetry of trading volume improved during the sample period. (2) There is no linearly significant causality following from trading volume to return or from return to trading volume, so there is no indication that technical analysis can be used to earn superior returns. (3) The contemporaneous correlations between absolute return and trading volume are significantly positive in all futures markets, and the information asymmetry decreases during the sample period for copper futures. (4) The evidence of a significant relationship of causality following from absolute return to trading volume contradicts the mixture of distributions hypothesis and supports the sequential information arrival hypothesis in all of the futures markets examined except for aluminum futures. Thus, in China’s copper, soybean, and wheat futures markets, new information is absorbed sequentially and the intermediate informational equilibrium is reached before the final equilibrium is found. In China’s aluminum futures market, however, the mixture of distributions hypothesis is supported, the dissemination of information is contemporaneous, and the final equilibrium is immediately reached. (5) There is significant causality following from trading volume to absolute settlement-to-settlement return in the copper (subsample 1) futures market, but not in the copper (subsample 2) futures market. This may suggest that the speculative activity is relatively high during the first subsample period. Speculation then declined during the second subsample period because of the effectiveness of the second powerful adjustment. (6) Generally, market efficiency improved significantly during the second subsample period in the copper futures market. One interesting question is: Why does the aluminum futures market (mixture of distributions hypothesis is supported) seem more efficient than the copper futures market (sequential information arrival hypothesis is supported) in the context that copper futures are more active than aluminum futures and copper has been considered one of successful futures products by the professionals in China? Conducting this investigation needs the behavioral finance interpretation and more detailed microstructure information about the two futures markets, and we leave it for future studies. This study examines the relationship of causality relying on the restrictive assumption of linearity. However, the issue of testing for a relationship of nonlinear causality between two time series has recently received a considerable amount of attention in the literature (for example, Hiemstra and Jones 1994; Fujihara and Mougoue 1997; Ciner 2002). Therefore, nonlinear causality can be investigated in a future study, since the univariate time series of futures prices and trading volumes are likely to be generated by nonlinear processes.

118

THE CHINESE ECONOMY

Notes 1. Starting from an initial position of equilibrium, a price change occurs due to the change in demand. The related adjustment induces transactions to react to the change in demand until a new equilibrium is reached. Thus, trading volume increases as price changes, regardless of the direction of the changes. 2. Trading volume is positively related to the extent to which traders disagree when they revise their price expectations, and an increase in the extent to which traders disagree is associated with a larger absolute price change. 3. Investors are heterogeneous in their information and private investment opportunities. As the asymmetry of information increases, uninformed investors require a higher discount in price when they buy contracts from informed investors to cover the risk of trading against private information. Therefore, trading volume is always positively related to absolute return, and the correlation increases with the level of information asymmetry. 4. For a review of the development of the ZCE see Williams et al. (1998). 5. It should be noted that if the coefficients of TREND and/or TREND*TREND are not significant, they will not be included in the detrending procedure. 6. Close-to-close Return is the first difference of the LOG(NCLOSEP); Settlement-tosettlement Return is the first difference of the LOG(NSETTLEP); Absolute Close-to-close Return is the absolute value of Close-to-close Return; Absolute Settlement-to-settlement Return is the absolute value of Settlement-to-settlement Return. 7. In this study, the AIC and FPE critiques have the same lags for all of the futures markets examined. 8. For this method, a relationship of causality exists only if at least two of three F values with 5, 10, 15 lags are significant at the 5 percent level. 9. This effect may also exist in wheat futures, since the Spearman correlation between settlement-to-settlement return and volume are marginally and negatively significant in the wheat futures market. 10. Settlement price is determined based on the weighted volume for the whole day, and much noisier trading may be reflected in the settlement price series. However, the closing price is determined by the last transaction; thus, the information on that day will have been sufficiently absorbed in the closing price series.

References Bessembinder, H., and P.J. Seguin. 1993. “Price Volatility, Trading Volume and Market Depth: Evidence from Futures Markets.” Journal of Financial and Quantitative Analysis 28: 21–39. Blume, L., D. Easley, and M. O’Hara. 1994. “Market Statistics and Technical Analysis: The Role of Volume.” Journal of Finance 49: 153–81. China Securities Regulatory Commission. 1999. The Examination Guide for Qualification as Futures Practitioners. Beijing: China Finance Press (in Chinese). Ciner, C. 2002. “Information Content of Volume: An Investigation of Tokyo Commodity Futures Markets.” Pacific-Basin Finance Journal 10: 201–15. Clark, P. 1973. “A Subordinated Stochastic Process Model with Finite Variance for Speculative Prices.” Econometrica 41: 135–55. Copeland, T.E. 1974. “A Model of Asset Trading Under the Assumption of Sequential Information Arrival.” Ph.D. dissertation, University of Pennsylvania. ———. 1976. “A Model of Asset Trading Under the Assumption of Sequential Information Arrival.” Journal of Finance 31: 1149–68.

MAY–JUNE 2004 119

Cornell, B. 1981. “The Relationship Between Volume and Price Variability in Futures Markets.” Journal of Futures Markets 1: 303–16. Crouch, R. 1970. “A Nonlinear Test of the Random-Walk Hypothesis.” American Economic Review 60: 199–202. Davidson, R., and J.G. Mackinnon. 1993. Estimation and Inference in Economics. Oxford: Oxford University Press. DeLong, J., A. Shleifer, L. Summers, and B. Waldmann. 1990. “Positive Feedback, Investment Strategies and Destabilizing Rational Speculation.” Journal of Finance 45: 379–95. Epps, T.W., and M.L. Epps. 1976. “The Stochastic Dependence of Security Price Changes and Transaction Volumes: Implications for the Mixtures-of-Distribution Hypothesis.” Econometrica 44: 305–21. Foster, A.J. 1995. “Volume-Volatility Relationships for Crude Oil Futures Markets.” Journal of Futures Markets 15: 929–51. Fujihara, R.A., and M. Mougoue. 1997. “An Examination of Linear and Nonlinear Causal Relationships Between Price Variability and Volume in Petroleum Futures Markets.” Journal of Futures Markets 17: 385–416. Gallant, A.R., P.E. Rossi, and G. Tauchen. 1992. “Stock Prices and Volume.” Review of Financial Studies 5: 199–242. Garcia, P., R. Leuthold, and H. Zapata. 1986. “Lead-lag Relationships Between Trading Volume and Price Variability: New Evidence.” Journal of Futures Markets 6: 1–10. Grammatikos, T., and A. Saunders. 1986. “Futures Price Variability: A Test of Maturity and Volume Effects.” Journal of Business 59: 319–30. Granger, C.W.J. 1969. “Investigating Causal Relations by Econometric Models and Cross-Spectral Methods.” Econometrica 37: 424–38. Gujarati, D.N. 1995. Basic Econometrics (international edition). Singapore: McGrawHill Inc. Harris, L. 1984. “The Joint Distribution of Speculation Prices and of Daily Trading Volume.” Working Paper No. 34-84. Department of Finance and Business Economics, University of Southern California, Los Angeles. ————. 1986. “Cross-Security Tests of Mixture of Distributions Hypothesis.” Journal of Financial and Quantitative Analysis 21: 39–46. ————. 1987. “Transaction Data Tests of the Mixture of Distributions Hypothesis.” Journal of Financial and Quantitative Analysis 22: 127–41. Harris, M., and A. Raviv. 1993. “Differences of Opinion Make a Horse Race.” Review of Financial Studies 6: 473–506. Herbert, J.H. 1995. “Trading Volume, Maturity and Natural Gas Futures Price Volatility.” Energy Economics 17: 293–99. Hiemstra, C., and J. Jones. 1994. “Testing for Linear and Nonlinear Granger Causality in the Stock Price-Volume Relation.” Journal of Finance 49: 1639–64. Jennings, R.H., L.T. Starks, and J.C. Fellingham. 1981. “An Equilibrium Model of Asset Trading with Sequential Information Arrival.” Journal of Finance 36: 143–61. Karpoff, J.M. 1987. “The Relation Between Price Changes and Trading Volume: A Survey.” Journal of Financial and Quantitative Analysis 22: 109–26. ————. 1988. “Costly Short Sales and the Correlation of Returns with Volume.” Journal of Financial Research 11: 173–88. Kocagil, A.E., and Y. Shachmurove. 1998. “Return-Volume Dynamics in Futures Markets.” Journal of Futures Markets 18: 399–426. Lee, B.S., and O.M. Rui. 2001. “Empirical Identification of Non-Informational Trades Using Trading Volume Data.” Review of Quantitative Finance and Accounting 17: 327–50. Li, Q. 1999. Theory and Practice in China’s Futures Markets. Beijing: China Finance and Economic Press (in Chinese).

120

THE CHINESE ECONOMY

McCarthy, J., and M. Najand. 1993. “State Space Modeling of Price and Volume Dependence: Evidence from Currency Futures.” Journal of Futures Markets 13: 335–44. Moosa, A.I., and N.E. Al-Loughani. 1995. “Testing the Price-Volume Relation in Emerging Asian Stock Markets.” Journal of Asian Economics 6: 407–22. Pindyck, R.S., and D.L. Rubinfeld. 1998. Econometric Models and Economic Forecasts (international edition). Singapore: McGraw-Hill Companies. Rutledge, D.J.S. 1979. “Trading Volume and Price Variability: New Evidence on the Price Effects of Speculation.” In International Futures Trading Seminar, 160–74. Chicago: Chicago Board of Trade. Shalen, C.T. 1993. “Volume, Volatility, and the Dispersion of Beliefs.” Review of Financial Studies 6: 405–34. Smirlock, M., and L. Starks. 1988. “An Empirical Analysis of the Stock Price-Volume Relationship.” Journal of Financial Research 8: 31–41. Suominen, M. 1996. “Trading Volume and Information Revelation in Stock Markets.” Working paper, Department of Economics, University of Pennsylvania. Tauchen, G., and M. Pitts. 1983. “The Price Variability-Volume Relationship on Speculative Markets.” Econometrica 59: 371–96. Wang, J. 1994. “A Model of Competitive Stock Trading Volume.” Journal of Political Economy 102: 127–68. Westerfield, R. 1977. “The Distribution of Common Stock Price Changes: An Application of Transactions Time and Subordinated Stochastic Models.” Journal of Financial and Quantitative Analysis 12: 743–65. Williams, J., A. Peck, A. Park, and S. Rozelle. 1998. “The Emergence of a Futures Market: Mungbeans on the China Zhengzhou Commodity Exchange.” Journal of Futures Markets 18: 427–48. Zhu, G.H. 1999. The Research for China’s Futures Markets. Beijing: China Commerce Press (in Chinese).

To order reprints, call 1-800-352-2210; outside the United States, call 717-632-3535.

Appendix 1 Detailed Contract Information on Copper, Aluminum, Soybean, and Wheat Futures Products Copper

Aluminum

SHFE

SHFE

Trading unit

5 tonnes/lot

5 tonnes/lot

Quotation unit

RMB/ton

RMB/ton

Tick size

RMB10/ton

RMB10/ton

Daily price limit

3 percent above or below the previous day’s settlement price

3 percent above or below the previous day’s settlement price

Contract months

January, February, March, April, May, June, July, August, September, October, November, December

January, February, March, April, May, June, July, August, September, October, November, December

Trading hours

9:00–11:30 A.M., 1:30–3:00 P.M., Monday to Friday

9:00–11:30 A.M., 1:30–3:00 P.M., Monday to Friday

Last trading day

15th of the delivery month (postponed in the case of legal holidays)

15th of the delivery month (postponed in the case of legal holidays)

Delivery period / last delivery day

16th–20th of the delivery month (postponed in the case of legal holidays)

16th–20th of the delivery month (postponed in the case of legal holidays)

Transaction margin

5 percent of the contract value

5 percent of the contract value

Transaction fee

Less than 0.02 percent of the trading value (including risk reserve payment)

Less than 0.02 percent of the trading value (including risk reserve payment) (continued )

MAY–JUNE 2004 121

Exchange

122

Appendix 1 (continued)

Soybean

Wheat

Exchange

DCE

ZCE

Trading unit

10 tonnes/lot

10 tonnes/lot

Quotation unit

RMB/ton

RMB/ton

Tick size

RMB1/ton

RMB1/ton

Daily price limit

3 percent above or below the previous day’s settlement price

3 percent above or below the previous day’s settlement price

Contract months

January, March, May, July, September, November

January, March, May, July, September, November

Trading hours

9:00–11:30 A.M., 1:30–3:00 P.M., Monday to Friday

9:00–11:30 A.M., 1:30–3:00 P.M., Monday to Friday

Last trading day

10th trading day of the delivery month (postponed in the case of legal holidays)

7th last trading day of the delivery month (postponed in the case of legal holidays)

Delivery period / last delivery day

7th day after last trading day (postponed in the case of legal holidays)

From the first trading day to the last trading day during the delivery month (postponed in the case of legal holidays)

Transaction margin

5 percent of the contract value

5 percent of the contract value

Transaction fee

RMB4/contract

RMB2/contract (including risk reserve payment)

Sources: China Securities and Futures Statistical Yearbook (1998–2002); www.shfe.com.cn; www.dce.com.cn; and www.czce.com.cn.

THE CHINESE ECONOMY

Detailed Contract Information on Copper, Aluminum, Soybean, and Wheat Futures Products