محتوای اطلاعات حجم: بررسی بازارهای آتی کالای توکیو

This study examines the relationship between volume and price changes for Tokyo commodity futures contracts by focusing on the predictive power of volume. The findings indicate a positive simultaneous relation between volume and absolute returns. The relation is not entirely contemporaneous since lagged volume contains predictive power for absolute returns. However, linear and nonlinear causality tests show that volume does not forecast returns. The results are qualitatively the same for contracts traded with different methods.

The nature of price–volume relationship in asset markets has long been a subject of financial research. Many papers document a positive contemporaneous relation between volume and absolute value of returns in both futures and equity markets. This is usually explained as a result of the same variable, the flow of information, directing changes in prices and volume as in the mixture of distributions hypothesis (MDH) of Clark (1973). Gallant et al. (1992) conducted an extensive analysis of stock price–volume relation for the U.S. equity market and concluded that more can be learned by studying prices jointly with volume than by examining them alone. This paper investigates the relationship between price changes and trading volume for gold, platinum and rubber futures contracts traded on the Tokyo Commodity Exchange (TOCOM). The main research question of the study is whether volume contains information useful for predicting future price movements. The information content of volume is examined for the direction and magnitude of price changes, i.e. for returns and absolute value of returns. The theoretical motivation for the study is provided by Blume et al. (1994), who argue that volume conveys information to the market that cannot be deduced from price alone. Specifically, their model suggests that volume contains information about the precision of price's signal and, therefore, current trading volume could improve forecast of price movements. Their analysis is consistent with the use of technical analysis in financial markets. An investigation of the TOCOM is of interest for two main reasons. First, the evidence on price–volume relation in futures markets is mainly from U.S. futures exchanges. Results from an international market can be useful for comparison. Second and more importantly, the TOCOM provides an opportunity to examine whether market structure affects price–volume relation in futures markets since two different methods of trading are used on the TOCOM. The first is systems trading, which is continuous trading similar to the methods used by U.S. futures exchanges. Gold and platinum futures contracts on the TOCOM are traded with the systems method. The second is Itayose trading, an auction-like periodic call market trading, under which all orders are treated as having arrived at the same time. The Itayose system closely resembles a classical Walrasian auction, where recontracting is allowed at provisional futures prices until an equilibrium (market consensus) price is determined. Rubber futures contract is traded using the Itayose method. Currently, no theory exists to link market structure to information content of trading volume. However, it can be argued that the predictive power of volume for future price movements should be greater for continuous trading systems. On a continuous trading market, trades occur one at a time and prices are not market consensus. The price in any transaction will differ depending on a number of factors, whether the trade is large or small being one of them. Hence, as Schwartz (2000) argues, price discovery will be more accurate on periodic call markets since trades are executed only at market consensus prices.1Webb (1995) similarly notes that prices produced by periodic call markets will be less noisy compared to those produced by continuous trading systems. These arguments suggest that prices will be more revealing on a periodic call market and, hence, trading volume as an additional statistic should be less informative. In fact, consistent with this argument, trading volume emerges as a useful statistic in the Blume et al. (1994) model only because prices are noisy and traders cannot obtain the full information signal from price alone. The empirical analysis begins with an examination of the relationship between daily trading volume and absolute value of returns. It is important to account for a simultaneous relationship between the variables in modeling. A structural model, which treats volume as an endogenous variable and controls for the simultaneity bias, is constructed and estimated using an instrumental variable (IV) estimator as a GMM estimator. The findings indicate a positive contemporaneous relationship between volume and absolute value of returns, supporting the MDH and consistent with prior studies from U.S. futures markets. Moreover, it is documented that past volume contains information to predict absolute value of returns, consistent with the arguments of Blume et al. (1994). The results are similar for all three of the contracts. The second part of the empirical analysis examines the dynamic relationship between volume and returns. A finding suggesting that volume can be used to predict returns could cast doubt on market efficiency.2 However, linear Granger causality tests, conducted within the context of vector autoregression (VAR) models, do not suggest that volume contains information to forecast returns. Hiemstra and Jones (1994) and Fujihara and Mougoue (1997) show that, for equity and futures markets, respectively, volume can have nonlinear predictive power. These studies use the modified Baek and Brock (1992) test, a nonparametric test designed to detect relations that cannot be captured by conventional linear causality tests, to examine nonlinear Granger causality dynamics. Indeed, an application of the modified Baek and Brock (1992) test suggests that bidirectional nonlinear Granger causality exists between volume and returns. However, this nonlinear Granger causality from volume to returns can be due to simple volatility effects. The MDH implies that a latent-variable representing daily information flow to the market affects trading volume and price contemporaneously. If lagged volume captures persistence in daily information flow, spurious causality may be detected. To investigate this possibility, returns are adjusted for conditional variance using an ARCH-type model and nonlinear causality tests are reconducted. Consistent with the implications of the MDH, nonlinear causality from volume to returns disappears after this adjustment. The remainder of the paper is organized as follows: The next section provides a brief review of prior work. Econometric approach of the study is discussed in the third section, while the fourth section presents the data set. The fifth section provides the empirical findings and the final section contains the concluding remarks of the study.