محل برگزاری تجارت بین المللی و رقابت جهت جریان

This paper tests a generalized version of the investor clientele hypothesis of Amihud and Mendelson [J. Finan. Econ. 17 (1986) 223]. This international trade-venue clientele effect hypothesis is supported for the Canadian cross-listed firms undifferentiated and differentiated by US trade venue, except for TSE shares cross-listed on NASDAQ. The hypothesized relationship between relative holding periods (measured using shares outstanding and share float) and effective half-spreads changes after TSE decimalization, and differs if the cross-listed shares have options traded on them. The empirical findings suggest that the TSE lost (won) executed order flow relative to the AMEX and to NYSE for shares with (without) options traded on them.

Amihud and Mendelson (1986) theoretically demonstrate that shares with higher bid–ask spreads are held for longer periods than shares with smaller spreads, all else held equal. This theoretical inference is supported by the finding of Atkins and Dyl (1997) that holding periods are positively associated with the size of the quoted bid–ask spread for NASDAQ and NYSE stocks. A generalized Amihud and Mendelson hypothesis, referred to herein as an international trade-venue clientele effect, is the focus of this paper which empirically examines this generalized hypothesis for a sample of Canadian firms cross-listed on the Toronto Stock Exchange (TSE) and various US trade venues. 1 This generalized hypothesis states that, in equilibrium, same-firm shares with increasing relative spreads across trade venues are held in portfolios with the same or expected increasing relative holding periods, and vice versa for decreasing relative spreads. Thus, a decrease in relative trade costs as measured by the effective spreads for the same-firm shares in favor of the US versus say the Toronto trade venue is expected to result in a decrease in the relative holding period or increase in the relative volume turnover on the US trade venue relative to that on the TSE. The existence of an international trade-venue clientele effect implies a possible win-lose situation for the market shares for order-flow execution across trade venues. A smaller ratio of the holding periods between the US trade venue and the TSE for the same-firm cross-listed shares implies that trading volume or share turnover has increased on the US trade venue relative to that on the TSE. In turn, this change normally is associated with a higher decrease or lower increase in effective spreads on the US trade venue relative to the lower decrease or higher increase in effective spreads on the TSE. If the synchronized increase (decrease) in the ratio of holding periods for the shares cross-listed on the US trade venue and the TSE is associated with a synchronized increase (decrease) in effective spreads for the same-firm shares on the US trade venue and the TSE, no international trade-venue clientele effect occurs. This would suggest that markets are becoming highly competitive, and that the differentials in trade costs and holding periods across trade venues will tend to disappear. In turn, this would lead to the full integration of the Canadian and US stock markets.2 Related to the generalized hypothesis is the issue of the impact of the TSE decimalization on order-flow execution and trade-venue clientele behavior. Since an objective of the TSE decimalization, which occurred on April 15, 1996, was to implement a minimum quotation increment reduction (MQIR) in order to increase TSE trading volumes relative to that on competing trade venues, this event is expected to have an impact on the relative market shares of trade venues competing for the order flow of our sample of Canadian cross-listed shares. A third issue is the impact of options trading on internationally listed same-firm shares. Various studies reviewed by Damodaran and Subrahmanyam, 1992 and Coughenour and Shastri, 1999 find that option trading decreases information asymmetry and enhances market efficiency for the underlying shares. In turn, all else held equal, the decrease in information asymmetry should lower bid–ask spreads. Lower bid–ask spreads should themselves be associated with shorter holding periods. The above discussion raises three questions that are addressed in this paper: (1) Does an “international trade-venue clientele effect” exist for the same-firm shares cross-listed on the TSE and various US trade venues? (2) What is the impact of the TSE decimalization on order-flow execution and trade-venue clientele behavior? (3) Does the international trade-venue clientele effect differ depending upon whether the internationally listed same-firm shares have options traded on them? In addressing these questions, this paper not only tests whether an international trade-venue clientele effect exists for Canadian firms cross-listed on the TSE undifferentiated by US trade venue but also differentiated by one of the three major US listing venues (NYSE, AMEX or NASDAQ). The primary reason for examining TSE shares cross-listed on the three major US listing venues on a differentiated basis is based on well-documented studies that report that the trading behavior of a stock not only changes for both domestic and international trade-venue migrations or cross-listings but the nature of the change depends upon the pair of trade venues examined. Specifically, the quoted spread declines, return volatility declines, and share turnover increases when a stock moves from the AMEX or NASDAQ to the NYSE,3 or when shares cross-list internationally into a more liquid market, such as non-US companies listing in the US.4 A secondary reason for our choice of firm groupings examined herein is that the trading behavior of a stock may be due to its special characteristics, its membership in a specific economic sector or market index or the characteristics of the trading mechanism where the stock trades. For example, shares that are listed on the NYSE/AMEX usually belong to traditional sectors compared to those listed on the NASDAQ, which usually are members of the high-tech sector. The largest and primarily resource-based Canadian firms tend to cross-list on the NYSE, while the smaller and primarily high-tech Canadian firms tend to cross-list on the NASDAQ. In addition, the different trading mechanisms at the NYSE/AMEX and NASDAQ may affect the trading behavior of firms in the same sector but trading on distinct trade venues differently (see Sanger and McConnell, 1986 and Affleck-Graves et al., 1994). Our empirical findings support the international trade-venue clientele effect hypothesis. The first major finding is that the ratio of effective spreads for international trade-venue pairings is positively associated with the ratio of their corresponding holding periods when the entire sample of TSE cross-listed firms is not differentiated by US trade venue. When the TSE cross-listed firms are grouped or differentiated by US trade venue, the international trade-venue clientele effect occurs for the TSE shares cross-listed on the AMEX and on the NYSE (the latter for shares with and without options traded on them), and primarily for the period before TSE decimalization that occurred on April 15, 1996. No evidence is found that the international trade-venue clientele effect occurs for the TSE shares cross-listed on the NASDAQ. The second major finding is that, for most time lags considered, the ratio of holding periods Granger causes the ratio of effective spreads for the TSE shares cross-listed on the AMEX, NYSE for shares with and without options traded on them, and NASDAQ. This supports the notion that TSE decimalization, whose objective was to reduce bid–ask spreads, was a reaction to the decreasing market share of the TSE for the executed order flow for cross-listed shares. MQIRs implemented by the AMEX, NYSE and NASDAQ after the TSE MQIR may have contributed to the decrease in relative executed order flow. However, if measured using effective spreads as the sole performance metric,5 these events did not consistently have the desired effect of decreasing trade costs for the cross-listed shares across trade venues, since effective spreads on average increased for the cross-listed shares on the AMEX and on the NASDAQ post-TSE decimalization. The third major finding is that the TSE shares cross-listed on the NYSE with options traded on them have smaller effective half-spreads and are associated with shorter holding periods than their non-optioned counterparts. This is consistent with the theoretical and empirical literature. Furthermore, the international trade-venue clientele effect appears to be enhanced during the post-TSE decimalization period only for the TSE shares cross-listed on the NYSE with options traded on them. In contrast, the international clientele effect weakens during the post-TSE decimalization period for the shares cross-listed on the AMEX and on the NYSE for shares without options traded on them. The remainder of this paper is organized as follows. The hypotheses to be tested are presented in Section 2. The sample and the data are described in Section 3. The general test methodology is introduced in Section 4. The empirical results are reported and analyzed in Section 5. Section 6 concludes the paper.

The main purpose of this paper was to test whether or not the empirical evidence supports the generalized version of the Amihud and Mendelson (1986) clientele hypothesis. Their hypothesis states that shares with higher bid–ask spreads are held for longer periods than shares with smaller spreads, all else held equal. When generalized to deal with same-firm share trades across multiple trade venues located in different countries, the new international trade-venue clientele effect hypothesis predicts that, in equilibrium, shares with increasing relative spreads across trade venues are held in portfolios with the same or expected increasing relative holding periods, and vice versa for decreasing relative spreads, all else held constant. The findings reported herein for Canadian cross-listed share trades executed on the US trade venues for the overall sample of firms undifferentiated by US trade venue are consistent with the international trade-venue clientele effect hypothesis (our first hypothesis). When the sample is differentiated by US trade venue, the clientele effect continues to hold for the Canadian cross-listed share trades executed on the TSE, AMEX and NYSE, and not NASDAQ. The relative effective half-spreads between the Canadian cross-listed share trades executed on these US trade venues and the same-firm share trades executed on the TSE are positively associated with the same or longer relative holding periods, particularly prior to TSE decimalization that occurred on April 15, 1996. Changes in effective half-spreads appear to be the only variable that is consistently associated with changes in holding periods for the various trade-venue categories of cross-listed share trades examined herein. The relative holding periods for the share trades executed on the US versus TSE trade venue are decreasing, and are associated with decreasing relative effective half-spreads for the TSE shares cross-listed on the AMEX and on the NYSE for shares with options traded on them. This implies that the TSE has consistently lost executed order-flow share to the AMEX and to the NYSE (the later for shares with options traded on them), and that this is due to increasing relative effective spreads for Canadian cross-listed shares. TSE decimalization did not obtain the desired outcome of avoiding the future loss of trading volume by the TSE to the US trade venues by reducing the MQI for its cross-listed shares. For the TSE shares cross-listed on the NYSE without options traded on them, the positive relationship between holding periods and effective half-spreads reverses post-TSE decimalization. The disappearance of the international trade-venue clientele effect post-TSE decimalization implies that the TSE won executed order-flow share from the NYSE for share trades of TSE shares cross-listed on NYSE without options traded on them. These results suggest that the existence of traded options on the underlying cross-listed shares may have important implications, as predicted by our third hypothesis, for the relative strength of the various trade venues competing for the same order-flow execution. TSE decimalization increased the inter-venue competition for order-flow execution. This competition was intensified by the MQIRs implemented on the AMEX and NASDAQ approximately one year later. Counter to the prediction of our second hypothesis, these events seem to affect the magnitude, direction and significance of the relationship between holding periods and effective half-spreads differently for the share trades executed on the TSE and the same-firm share trades executed on each US trade venue. In turn, they may have weakened the international trade-venue clientele effect by reducing the materiality of effective half-spread differences between trade venues. Nevertheless, it is difficult to conclude whether or not the loss, gain or maintenance of TSE competitiveness in terms of trading volumes and effective half-spreads would have been more pronounced without the first mover action of the TSE to decimalization.