The differences in information content of the four trader classes

國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

78

Model 3 puts futures returns and futures transactions into the same regression equation. Except for a slightly lower level of significance for each coefficient, there is no fundamental difference in the conclusion. From an information-based perspective, these results indicate that the futures market leads the spot market and that the leading relationship is mainly present in the futures returns, rather than the futures volume.

This conclusion is similar to that found in CCF’s analysis of the option market.

The alternative methodology is the VAR model, which is represented in equations (2) to (4). According to the model’s results,⁵ the lag TX returns have significantly positive effects on the TAIEX returns, but the lag TAIEX returns have no significant effects on the TX returns. This indicates a one-way leading relationship between the TX returns and the TAIEX returns. The outcome is the same as that found in the general regression. In addition, none of the lags of TX net open buy have a significant effect on the TAIEX returns. Thus, by merely observing the overall TX net open buy, I am unable to perceive the information effect on the spot market.

4.2. The differences in information content of the four trader classes

In the previous analysis, I found that the futures returns do contain information on the spot returns, but I could not observe similar information content arising from futures net open buy. According to CCF, because informed traders are less aggressive in submitting orders (e.g., they only submit limit orders instead of market orders), they are able to influence the quoted price but not drive the trades. Alternatively, another plausible reason may be that the leading relationship of trading activity can only be observed on the transactions conducted by specific traders. To verify this, in the following analyses the futures traders are divided into four classes: foreign institutional traders, futures dealers, domestic institutional traders, and individual

5 The results are reported in Appendix A, Table A2.

‧

I first study the relationship between the spot returns and the net open buy of the four trader classes in the futures market. Equation (1) is revised as follows:

3 3

where , represents foreign institutional traders, futures dealers, domestic institutional traders, and individual traders, respectively, and NBj,t represents the net open buy of trader class j. The regression results are presented in Table 3.4.

Panel A shows the results of four regressions based on the net open buy of each trader class. In Panel B, the net open buy of all trader classes have been put into one regression.

[Insert Table 3.4]

The results in Panel A confirm my predictions. The lag one coefficient (0.0054) of foreign institutional traders is significantly positive but the coefficients of leads one and two (0.0091 and 0.0044) are also significantly positive. These results imply that there is a feedback relationship between the TAIEX returns and the net open buy of foreign institutional traders. The lag-two coefficient of futures dealers and the lag-one coefficient of individual traders are significantly positive. However, the magnitude of the positive coefficients almost be canceled out by the negative coefficients. From an information-based viewpoint, only the transactions of foreign institutional traders carry information on the spot market; the net open buys of futures dealers, domestic institutional traders, and individual traders have no directionally predictive power for the TAIEX returns.

Panel B shows a similar outcome to Panel A. The lead-one, lead-two, and lag-one net open buy of foreign institutional traders are positive and significant (coefficient

, , , and jFI D DI I

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

80

0.0051, 0.004 and 0.0047, respectively), which indicates a feedback relationship between the TAIEX returns and the net open buy of foreign institutional traders. The lagged net open buys of the other three trader classes are either negative or insignificant. Only the coefficient (0.005) of lag-one net open buy of individual traders is significantly positive. However, the magnitude of the positive coefficient is canceled out by the negative coefficients. The evidence again proves that the net open buy of futures dealers, domestic institutional traders, and individual traders have no directionally predictive power for the spot returns.

While Chou and Wang (2009) found that foreign institutional traders and futures dealers seem to be better informed about the TX price movements than the others in Taiwan futures market, our results provide evidence that only foreign institutional traders seem to be better informed about the spot price movements. Our finding is more consistent with that of Chang et al. (2009). How could domestic traders have less information about their own market? One possible explanation suggested by Huang and Shiu (2009) and Chou and Wang (2009) is that foreign institutional traders may have superior technological, financial, or human expertise, experience, or resources. Furthermore, these advantages are more evident when domestic traders are from emerging markets. Another possible reason mentioned by EOS is that hedging demand may also be an important motivation for trading futures. For example, futures dealers may trade exchange-traded funds (ETFs), construct stock portfolios, or trade TAIEX options based on information and hedge them in futures market. Especially, some futures dealers have qualified as market makers in the TAIEX option market, their hedging demand may be stronger. As noted by Fahlenbrach and Sandas (2003), the cheapest way to hedge delta risk of index options is to use the nearby index futures. In this situation, the trading activity of futures dealers may appear to be inversely related to the spot returns. This may partly explain why the coefficient on

‧

the lag-one net open buy of futures dealers is strongly negative in Table 3.4.

In sum, our results show that the leading relationship of futures trading activity can only be observed from a specific trader class and this class is the foreign institutional traders.⁶ However, within the sample period, the foreign institutional traders’ open-buy and open-sell volume is only 7.76% of the overall TX’s open-buy and open-sell volume. Because the proportion is very low, the results are easily dominated by the trading activities of the other trader classes if overall volume is used.

That’s the reason why we observe that the futures returns informationally lead the spot returns but the overall futures trading activity informationally lags the spot returns as shown in Table 3.3.

4.2.2 The futures market

In addition to analyzing cross-market information, this article also briefly discusses price-quantity relationships solely within the futures market. The regression model is similar to equation (5); the dependent variable becomes TX returns, and the lead, contemporaneous, and lag spot returns are controlled:

3 3

The regression results are shown in Table 3.5. Panel A reports the results of four regressions based on the net open buy of each trader class. Panel B presents the result of putting the net open buy of all trader classes into one regression.

[Insert Table 3.5]

According to Table 3.5, there exists a feedback relationship between the spot returns and TX returns because the lag-one, contemporaneous, and lead-one

6 In the literature of options markets, Chang et al. (2009) used daily data to examine the one-way relation between the option market and spot market and found similar results to ours.

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

82

coefficients on the spot returns are all significantly positive. The lag-three coefficient of individual traders is positively significant at the 10% level. However, the magnitude of the positive coefficient is smaller than the negative one. None of the lag net open buys of the other trader types shows significantly positive. This implies that all of the trading activities of the four trader types have no predictive power for TX returns.

If effect, if I do not control the spot returns, the results of Panel A becomes that the foreign institutional traders’ ( jFI) TX net open buy leads the index futures returns. The lag-two and lag-three coefficients are significantly positive (coefficients 0.0093 and 0.009, and the t-statistics are 2.46 and 2.64, respectively). The lead coefficients are all negative. These results suggest that the net open buy of foreign institutional traders informationally leads the TX returns. The lagged net open buys of the other three classes of traders appear to be significantly negatively correlated with the TX returns, and their lead net open buys appear to be significantly positively correlated with the TX returns. Therefore, the trading activities of futures dealers, domestic institutional traders, and individual traders all informationally lag behind the TX returns. These results imply that the foreign institutional traders are the only class whose trading activity have information content relating to the futures returns.

4.2.3 Robustness check

As mentioned previously, information may not be the only factor that pushes traders to trade in futures market. Traders may trade futures contracts based on other purposes such as hedging or arbitrage. To mitigate the noise from non-information-based trading, I drop the transactions from arbitrager and re-run equations (5) and (6). That is if a trader trade finance sector index futures (TF) or/and electronic sector index futures (TE) and TX at the same day, it may be an

‧

arbitrage-based transaction and I exclude it from the sample. The regression results are shown in Appendix A, Tables A3 and A4.

After purging the effect of arbitrage, the new regression results show no conclusive change. The patterns reported in Tables A3 and A4 are quite similar to those in Tables 3.4 and 3.5, which confirm the robustness of my conclusions.