فعالیت بازرگانی، تمرکز فروشنده و کیفیت بازار ارز

We study the relation between foreign exchange market quality and both trading activity and dealer concentration by considering two currency pairs with significant differences along both dimensions – the Euro–US dollar and Canadian dollar–US dollar. A variance ratio test reveals over-reaction in currency prices, but that this is smallest when trading activity is high and dealer concentration at its peak. A GARCH model shows that over-reaction declines as trading activity and dealer concentration increase, with the results being stronger for the Euro. Our results confirm that trading activity is an important determinant of market quality, but also point to a significant role for dealer concentration.

Market quality is a key characteristic of financial markets because it summarizes the speed and accuracy with which information is impounded in prices.1 Thus, uncovering the factors that are associated with improved market quality in financial markets is important to academics, financial market participants and regulators. Although the conventional wisdom is that market quality is highest when the level of trading activity is highest, recent work in market microstructure suggests that other factors may also influence market quality. The list of such factors includes the identity of traders (e.g. Peiers, 1997 and Sapp, 2002) and the level of competition between dealers (e.g. Huang and Masulis, 1999). These factors, however, have not received as much attention in empirical analyses of market quality. Using high frequency data for the Euro–US dollar and Canadian dollar–US dollar currency pairs, this paper investigates the effects of trading activity and dealer concentration on foreign exchange market quality.2 The new hypothesis explored in this paper is that dealer concentration (the percentage of trading involving the most active dealers) affects foreign exchange market quality. Huang and Masulis, 1999, Evans, 2002 and Evans and Lyons, 2002 suggest that greater concentration of trading in the hands of large dealers makes market prices less noisy, due to less dispersed private information and inventory effects. Therefore, at times of increased dealer concentration, currency prices should be more efficient and market quality higher. The Euro–US dollar and the Canadian dollar–US dollar are ideal cases to consider because they have very different characteristics. The Euro–US dollar (henceforth, the Euro) is the most heavily traded currency pair. It is also traded by almost all the large foreign exchange dealers and many small ones as well, so dealer concentration is low for the Euro. The Canadian dollar–US dollar (henceforth, the Canadian dollar) is significantly less heavily traded, traded predominantly by North American (and especially Canadian) dealers, and sees a much higher degree of dealer concentration3. We use these currencies to investigate if the level of trading activity is the main influence on market quality (the conventional wisdom) or whether dealer concentration also plays an important part. We study two measures of market quality. First, we consider the variance ratio statistic, which captures the efficiency with which all price relevant information is impounded in prices. In a high quality market, price changes tend to be permanent and transitory volatility low, so the ratio of a long-term return variance to an appropriately scaled short-term variance should be close to 1.0.4 This analysis is complemented by an analysis of the return dynamics for the currency pairs, the presumption being that higher market quality means more rapid and accurate incorporation of both currency-specific and common information into prices. We use a GARCH framework to study intraday returns and volatility for the Euro and Canadian dollar and explicitly relate these to trading activity and dealer concentration. Our results are based on intraday indicative quote data for the calendar year 2000. The dataset includes information on the identity of the quoting banks, allowing us to consider how the level of trading activity (as captured by aggregate quoting activity)5 and the presence of influential dealers (as measured by the percentage of quotes issued by the most active dealers) affect market quality. We proceed by calculating both aggregate quoting activity as well as the percentage of the total number of quotes submitted by the most active dealers (the reported results are based on the five largest dealers) in 15-min intervals throughout the day. We expect market quality to be directly related to aggregate quoting activity. Further, if greater participation by active dealers implies that prices are more likely to be set by better-informed traders, we expect higher dealer concentration to be associated with higher market quality.6 We start by examining the intraday (15-min) patterns in quoting activity and dealer concentration for the two currency pairs. For the Euro, quoting activity increases rapidly as the European markets open, rises further as the North American markets open, stabilizes until the European markets close, and then declines. For the Canadian dollar, quoting activity increases as the European markets open and especially as the North American markets open, peaks during the overlap in European and North American trading and remains relatively high until the close of the North American markets. These patterns suggest that trading activity for the Euro is relatively evenly spread out across the major European and North American markets (being highest when both Europe and North America are open). For the Canadian dollar, trading activity is heaviest in the North American market. Turning to the patterns in dealer concentration, we find that dealer concentration is always lower for the much more heavily and broadly traded Euro than for the Canadian dollar. For instance, the mean percentage of quotes accounted for by the top five dealers in the Euro never exceeds 45%, whereas the mean percentage of quotes submitted by the five most active dealers in the Canadian dollar typically averages 80% and is rarely less than 50%. In the case of the Euro, dealer concentration increases at the start of European trading, peaks during the heaviest European trading, declines briefly as the North American markets open, increases as North American trading moves into full swing, and then declines after the European markets close. For the Canadian dollar, dealer concentration increases during European trading but remains relatively flat during active trading in Europe and North America. Consequently, dealer concentration is not merely a function of trading activity for these two currency pairs. When we examine market quality, we find that the variance ratio for both currencies is below 1.0 for much of the trading day, suggesting some degree of over-reaction. The variance ratio for the Euro is generally farther below 1.0 than that for the Canadian dollar. Possible contributors to the larger short-term distortions in the case of the Euro include a higher rate of information arrival or dispersed inventory effects due to lower dealer concentration. Across the trading day, the variance ratio is closest to 1.0 for the Canadian dollar during North American trading, when the dollar is most actively traded and dealer concentration is highest. For the Euro, the variance ratio is closest to 1.0 when only the European markets are open. As the North American markets open – and trading activity peaks while dealer concentration declines – the variance ratio for the Euro moves further below 1.0. These patterns are not entirely consistent with the conventional wisdom that higher trading activity improves market quality, but they are consistent with the hypothesis that higher dealer concentration promotes market quality. To formally investigate the importance of quoting activity and dealer concentration, we estimate a GARCH model that captures information assimilation within and across the currency pairs. We extend prior research (e.g. Baillie and Bollerslev, 1991 and Engel et al., 1990) by including quoting activity and dealer concentration in the conditional mean and conditional variance equations. Consistent with the variance ratio results, there is negative autocorrelation in currency returns. For the Euro, the autocorrelation becomes significantly less negative as quoting activity or dealer concentration increases. The coefficients for the Canadian dollar tell a similar story although they are not significant.7 Thus, there is an improvement in the efficiency with which currency-specific information is impounded in the prices of the Euro and the Canadian dollar as either the level of quoting activity or the degree of dealer concentration increases. Cross-currency effects, which reflect the incorporation of common information, reveal some return predictability running from the Euro to the Canadian dollar but some volatility predictability in the reverse direction. These results suggest that the incorporation of currency-specific information depends on both the level of trading activity and dealer concentration, and that common information-related efficiency is somewhat higher for the more actively-traded Euro. In sum, the patterns in market quality do not mechanically follow those in aggregate trading activity, as widely believed, but also reflect patterns in dealer concentration. Thus, while we confirm the importance of the level of trading activity, the main contribution of our analysis is to highlight the role of dealer concentration in explaining variations in market quality. An additional contribution is our description of the market microstructure and market quality of a less actively traded currency, the Canadian dollar. Even though a growing portion of trading in the foreign exchange market now involves less liquid currencies such as the Canadian and Australian dollars versus the US dollar, the microstructure of these currency pairs remains largely unexplored.8 The paper develops as follows. Section 2 provides a description of the data. Section 3 describes our empirical techniques and results. Section 4 concludes.

The quality of a financial market reflects the speed and accuracy with which information is impounded in prices. In this paper, we examine foreign exchange market quality, focusing on the influence of two factors: the level of quoting activity (a proxy for overall trading activity) and dealer concentration (the share of quoting activity attributable to the leading dealers). We choose to analyze the Euro-US dollar and the Canadian dollar–US dollar currency pairs because, as we show, quoting activity and dealer concentration for these pairs differ significantly across the trading day. This allows us to pin down the effects of these two factors on market quality. Our first metric of market quality involves a variance ratio statistic, based on the ratio of the variances of longer-term and short-term returns. The variance ratio will be close to 1.0 if market quality is high and above (below) 1.0 if currency prices exhibit under- (over-) reaction. Both currencies show some degree of over-reaction, and the extent of this over-reaction varies across the trading day. Market quality for the Canadian dollar is highest (i.e. over-reaction is small) when quoting activity and dealer concentration are at their highest levels. For the Euro, market quality is highest when dealer concentration is highest and quoting activity is high, though not at its peak. Thus, market quality for both currencies appears to be highest when quoting activity is concentrated among a small number of dealers; this is at odds with the conventional wisdom which focuses on the level of trading activity alone. To formally investigate the separate roles of trading activity and dealer concentration as well as to study the efficiency of incorporation of common versus currency-specific information, we estimate a GARCH model. The results reinforce the results of the variance ratio analysis and provide additional insights into market quality. Consistent with the variance ratio results, we find an increase in the efficiency with which currency-specific information is impounded in prices both as the level of quoting activity increases and as the degree of dealer concentration increases. The results are especially striking for the Euro. Cross-currency effects show return predictability running from the Euro to the Canadian dollar but volatility predictability in the reverse direction. The GARCH analysis points to three facts about market quality. First, the incorporation of currency-specific information depends on both the level of trading activity and dealer concentration. Further, looking at common information, efficiency is somewhat higher for the more actively and globally traded Euro than the Canadian dollar. Finally, the transmission of common information in returns and in volatility is influenced by different factors. Overall, our use of foreign exchange data sheds light on the influence of trading characteristics on financial market quality. It is believed that the level of activity is a key contributor to market quality, and our analysis confirms that market quality is high when quoting activity is heavy. At the same time, however, we present evidence to suggest that market quality is high when concentration of trading among the largest dealers is high. Thus, market quality depends, not only on the level of trading activity, but also on dealer concentration. The importance of such factors as dealer concentration can only increase as trading becomes more global and dispersed across dealers and customers.