ارتباط اطلاعاتی در سراسر مناطق تجاری: شواهدی از بازار ارز خارجی

Using a new high-frequency data set from Electronic Broking Services (EBS), this paper examines informational linkages in the euro–dollar and dollar–yen exchange rates across five trading regions: Asia Pacific, the Asia–Europe overlap, Europe, the Europe–America overlap, and America. Information is proxied by exchange rate return, direction of return, volatility, trading activity, and order flow. We find that informational linkages are statistically significant at both own-region and inter-region levels, but own-region spillovers dominate in economic significance, especially for volatility and trading activity. In addition, order flow spillovers from the Europe–America overlap trading region are the most important source of spillovers to other trading regions for both currency pairs.

Foreign exchange is traded around-the-clock in major global financial centers, including Tokyo, London, and New York. With this market structure, exchange rates can incorporate new information instantly. Thus, we expect information that is incorporated into exchange rates to pass on from one financial center to the next. However, there are only a few studies that examine informational linkages across major financial centers, and they use only a short sample period which makes the results hard to generalize (e.g., Engle et al., 1990 and Baillie and Bollerslev, 1990). In addition, previous studies use only exchange rate volatility as a proxy for information. This paper examines informational linkages in the global interdealer foreign exchange market across major financial centers. We use a unique data set of prices, trading volume, and order flow in spot euro–dollar and dollar–yen trading from Electronic Broking Services (EBS) covering the period from January 1999 to February 2004. This EBS data set has two important advantages. First, the EBS data set consists of transactable quotes, as opposed to indicative quotes from Reuters used in previous studies. The EBS data set also contains actual trading volume and order flow that were previously not available. Second, EBS has become the major trading platform for the two most traded currency pairs, the yen and the euro, making the results based on this data set a true representation of the behavior of global interdealer foreign exchange markets. As informational linkages can manifest themselves by the transmission of exchange rate return, direction of return, volatility, trading activity, and order flow, we investigate these financial variable spillovers across three major trading areas (centered in Tokyo, London, and New York) for two currency pairs (euro–dollar and dollar–yen).2 We are interested in the extent to which changes in exchange rates, volatility, trading activity, and order flow in one financial center influence those in the other financial centers. Specifically, this paper explores and makes contributions to several issues as follows. First, this paper uses regional trading volume from the EBS data set to classify different trading regions in the global foreign exchange market for the euro–dollar and dollar–yen currency pairs. Trading volume is a more accurate measure of trading activity than quote frequency used in previous studies (e.g., Melvin and Yin, 2000 and Melvin and Melvin, 2003). EBS records regional trading volume by transactions that occur between pairs of the three physical locations of the EBS computer centers: Tokyo, London, and New York. These computer centers are linked together in real time, and EBS classifies the origin of each trade by the physical location of the computer center that serves each region.3 Based on our assessment, we define five distinct trading regions: Asia Pacific, the Asia–Europe overlap, Europe, the Europe–America overlap, and America.4 Second, we explore the spillovers of exchange rate return, direction of exchange rate return, return volatility, trading activity, and order flow across different trading regions for the two currency pairs. We test two hypotheses first studied in Engle et al. (1990) for exchange rate volatility: the meteor shower hypothesis and the heat wave hypothesis. The meteor shower hypothesis suggests that there are information spillovers across trading regions. For example in the case of volatility, high (low) volatility in one region today tends to be followed by high (low) volatility in the next trading region. The heat wave hypothesis suggests that information persists within trading regions. For example, high (low) volatility in a region today tends to be followed by high (low) volatility in the same region on the following day. To our knowledge, this paper is the first to study a comprehensive set of variables that proxy for information in the foreign exchange market. The use of order flow to study information linkages provides us a clean way to measure information transmission across trading regions. Evans and Lyons (2002) note that order flow can be viewed as a proxy for information about economic fundamentals. Since informational linkages usually manifest themselves by both volatility and trading activity, we use trading activity as a robustness check for the information transmission dynamics. We use both trading volume and the number of transactions to measure trading activity. Previous studies have attempted to proxy for actual transaction volume by using futures exchange volume data (e.g., Chaboud and LeBaron, 2001) or the frequency of indicative quotes on Reuter's data screens (e.g., Melvin and Yin, 2000), but these measures have been shown to be poor proxies. We also repeat the test with the number of transactions. Jones et al. (1994) find that in the case of equity markets the key information in trading volume is captured by the number of transactions. Third, we use high-frequency data to measure daily volatility, i.e., realized volatility (Andersen et al., 2001), and revisit the issue of volatility clustering – the meteor shower and heat wave hypotheses. Engle et al. (1990) find support for the meteor shower effect, while Baillie and Bollerslev, 1990 and Hogan and Melvin, 1994, and Melvin and Melvin (2003) report evidence for the heat wave effect. These conflicting results may reflect differences in their sample period, mostly very short, and sampling frequency. This paper contributes to this research area by using high-frequency intraday data with a longer and more recent sample period to test the heat wave and meteor shower hypotheses. In addition, our measure of volatility allows us to treat volatility as observable as opposed to latent, which makes our results independent of statistical models for volatility (e.g., GARCH and stochastic volatility models). We estimate a regional model of each proxy for information (i.e., return, direction of return, return volatility, trading activity, and order flow) by allowing each region's information proxy to depend on its own past value and other regions' past values, similar to Melvin and Melvin (2003) in the case of volatility. The model is estimated for each information proxy one at a time. Our main findings can be summarized as follows. First, we find statistically significant evidence of volatility, trading activity, and order flow spillovers across trading regions (i.e., the meteor shower effect), but weak statistical evidence of return and direction of return spillovers across trading regions. This finding suggests that there are information linkages across trading regions, as shown by spillovers in volatility, trading activity, and order flow, and that the foreign exchange market is efficient in processing new information, as shown by the lack of spillovers in return and direction of return. This interpretation is consistent with the finding in Andersen et al. (2003) that exchange rates respond instantaneously to macroeconomic news surprises, but that exchange rate volatilities increase for a longer period of time before they decrease to their normal levels (i.e., 5 min for returns versus 1 h for volatilities). Second, the economic significance of volatility and trading activity spillovers across regions (i.e., the meteor shower effect) is much smaller than that of volatility persistence within regions (i.e., the heat wave effect). This finding is consistent with the idea that information flow is correlated within a region and/or that the behavior of traders within a region is a driving force for volatility and trading activity clustering. Third, spillovers in order flow from the Europe–America overlap and America to other trading regions are most important in euro–dollar, whereas spillovers from Asia, the Asia–Europe overlap, and the Europe–America overlap are most important in dollar–yen. The importance of the Europe–America overlap for both currency pairs reflects the fact that London is the most important trading center for foreign exchange and that major U.S. economic releases occur in this overlap.5 This finding also highlights the importance of country-specific news for the two exchange rates. The remainder of the paper is organized as follows. Section 2 describes data sources. Section 3 provides details on trading region classification. Sections 4 and 5 present empirical models and empirical results, respectively. Section 6 concludes the paper.

This study examines informational linkages in the foreign exchange market. Our results are based on a unique data set from EBS, which represents most of the global interdealer trading activity in the euro–dollar and the dollar–yen exchange rates, contains transactable quotes, and contains a longer and more recent sample period than data sets used by previous studies (e.g., Engle et al., 1990, Baillie and Bollerslev, 1990, Hogan and Melvin, 1994 and Melvin and Melvin, 2003). These advantages of our data should make our results a true representation of the behavior of the global interdealer spot exchange rate. Unlike most previous studies which only study volatility spillovers, this paper studies spillovers in return, direction of return, volatility, trading activity, and order flow across trading regions in the foreign exchange market. Based on our regional models, we find statistically significant evidence of volatility, trading activity, and order flow spillovers across trading regions (i.e., the meteor shower effect), but weak statistical evidence of return and direction of return spillovers across trading regions. We interpret the results as evidence for informational linkages across different trading regions. However, the economic significance of informational linkages is smaller than that of spillover within regions (i.e., the heat wave effect), especially for volatility and trading activity. This finding is consistent with the idea that information flow is correlated within a region and/or that the behavior of traders within a region is a driving force for volatility and trading activity clustering. In addition, we find that information originating from the Europe–America overlap trading region, as measured by order flow, is the most important source of spillovers to other trading regions for both euro–dollar and dollar–yen exchange rates, pointing to the importance of London as a financial center for foreign exchange trading and the importance of major U.S. economic data releases. Information spillovers from America are also important for euro–dollar, whereas spillovers from Asia and the Asia–Europe overlap are also important for dollar–yen. This finding also highlights the importance of country-specific news for the two exchange rates.