Volume or Order Flow? : Which contains more information in really traded Yen/Dollar Foreign Exchange Market?

Masayuki Susai (Nagasaki University)

Empirical Results

Table 7 shows the estimation results of Base Model. As we confirmed in Table 5 and 6 that t distribution should be employed, we estimate GARCH (1,1) with t-distribution model here.

Discussions

The analyses up to here are showing the strong impact of trading volume on foreign exchange rate volatility without any exception regarding the representation of volatility and data set. From these implications, we confirm that volume can be the good proxy for information arrival. Two methods are employed to test whether volume can be the proxy of information arrival or not.

We construct foreign exchange rate volatility directly from raw data first. With this volatility of foreign exchange rate, we estimate the impact of volume directly. Following the Clark (1973) and Bauwend, et al, (2006), trading volume and log difference of trading volume were introduced into the model. Beside volume data, we also use two order flow indicators as the candidate of the proxy. Because many researches are pointing out that order flow is playing a important role in explaining asset price volatility.

Differ from the method Bauwend, et al. (2006) used, we employed GARCH with tdistribution and drew a implication from the parameter of volume related data and other variables in mean equation. As for the volume related data, the parameters of volume and log difference of volume data were estimated significantly. These significant parameters strongly suggested that volume can be the proxy of information arrival from the Mixture of Distribution Hypothesis.

In the next step, we introduce volume data into conditional variance equation and to check the impact of volume on GARCH effect. Lamoureux, et al. (1990) suggested that if volume can play a significant role that Clark (1973) pointed out, then GARCH terms fades away when volume data is introduced into conditional variance equation. GARCH with t-distribution model is also employed to check the volume impact. In estimating the model, we control the price pressure arising from order flow.

All the results are showing that trading volume can reduce the level of GARCH term parameters. Our results are not as clear as Lamoureux, et al. (1990) found, but we can confirm that trading volume works as the substitute of GARCH terms even though we control the price pressure. At the same time, we may overestimate the trading volume impact without any concern about the control variables of price pressure to foreign exchange rate.

With these two methods, we check the impact of order flow on volatility of foreign exchange rate to explore the information content in it. To figure out the information content in order flow data, we compare the impact of volume and order flow on the volatility. As the order flow data, we construct the difference of the number of bid and ask as OF, and the order direction based net volume as OFVOL.

As mentioned above, trading volume is playing a significant role in all analyses. Only order direction based net volume shows significant impact on the volatility in the first method. Two order flow variables were estimated significantly in conditional variance equation in GARCH model, but those do not have any impact on GARCH effect. From these results, the impact of order flow data on volatility of foreign exchange rate might be limited. In terms of the information content, trading volume has richer content than any other order flow variables. Among the order flow variables, not the direction of transaction but the volume might be important.

Search
Information

Empirical Study on Asian Financial Markets

Edited by Masayuki Susai Hiromasa Okada

本書は、2006年12月に長崎大学において開催された国際カンファレンス等で報告された東アジアの金融市場を対象とした金融および会計学における実証研究の成果をまとめたものである。