Globalization and the value of US listing: Revisiting Canadian evidence

Transcription

1 Journal of Banking & Finance 27 (2003) Globalization and the value of US listing: Revisiting Canadian evidence Usha R. Mittoo Asper School of Business, University of Manitoba, 181 Freedman Crescent, Winnipeg, Man., Canada R3T 5V4 Abstract As capital markets become integrated, the value of foreign listing and, consequently, the number of foreign listings should be expected to decline. The dramatic surge in foreign listings on US stock exchanges in recent years suggests that motivations and value of listing may also be changing as markets become increasingly globalized. We explore this issue by comparing both short- and long-run valuation effects of Canadian listings in the US in pre- and post periods. We document that the positive price and liquidity effects for Canadian stocks surrounding US listing have declined over time. The analysis of long-run performance, however, shows that Canadian firms list in the US after a strong market performance but underperform Canadian market indexes by 13 30% over the three years after the listing in both pre- and post-1990 periods. Further, the determinants of long-run performance appear to be significantly different from that in the short-run. Our evidence suggests that valuation effects of US listing may be driven by several factors, including liquidity and industry factors, that vary cross-sectionally and over time. Ó 2003 Elsevier B.V. All rights reserved. JEL classification: F30; G15; G30 Keywords: Foreign listing; Segmentation; Liquidity; Short-run effects; Long-run performance 1. Introduction The geography of cross-border listings has changed significantly in the last decade. While US exchanges have seen a dramatic increase in foreign listings, most other exchanges have experienced a significant decline in the 1990s. For example, the number of foreign listings on the major US stock exchanges surged from less address: umittoo@ms.umanitoba.ca (U.R. Mittoo) /$ - see front matter Ó 2003 Elsevier B.V. All rights reserved. doi: /s (03)

2 1630 U.R. Mittoo / Journal of Banking & Finance 27 (2003) than 200 in 1990 to over 750 in 2000, an increase of about 400 percent while that on the European exchanges declined from 757 in 1991 to 627 in 1997, a decline of over 20 percent. 1 What makes US exchanges such attractive places for listing? This question has generated considerable debate among practitioners and academics. Traditionally, firms listed in the US primarily to access the larger US capital markets and to enhance stock liquidity. These motivations have become less important in recent years as barriers to international capital flows have fallen and an increasing number of institutional investors are buying foreign shares on their home exchanges. Thus, many doubt if the US listing adds any shareholder value. Others argue that the motivations as well as the benefits of US listing may have also changed as product markets become increasingly globalized. A US listing enhances visibility and profile of a company and signals to the market that it has become a global player. Survey evidence by Bancel and Mittoo (2001) and Fanto and Karmel (1997) suggests that recent US listings appear to be driven primarily by business rather than financial considerations. Pagano et al. (2001) also document that firms appear to follow their industry peers in selecting a foreign exchange of listing. Our study contributes to this debate by providing some insights into whether the motivations as well as the value of US listing have changed through time. We examine this issue by comparing the US listing effects for Canadian listings in the pre-1990 and post-1990 periods. The Canadian context provides a unique sample to conduct this experiment for several reasons. First, in contrast to most non-us firms, Canadian firms have a long tradition of listing on the US markets. Until 1990, Canadian firms comprised about half of total foreign listings in the US and these firms listed primarily to raise cheap US capital and to increase stock liquidity. These reasons have become less important in the 1990s as Canadian markets have become increasingly integrated with US markets and Canadian stock exchanges have become more competitive with their US counterparts. Thus, we should expect that the value of a US listing for Canadian firms and, consequently, the number of Canadian listings in the US should decline over time. However, the trend has been just the opposite as a record number of Canadian firms have chosen to interlist in the US in 1990s. For example, the number of Canadian firms interlisting on the New York Stock Exchange (NYSE) increased from only 28 in 1990 to 70 in 2000, an increase of 150 percent. Whether these listings enhance shareholder value is an important issue. Second, a vast majority of the pre Canadian listings in the US are resource stocks that form a significant component of the Canadian economy. The post-1990 listings, in contrast, are dominated by non-resource stocks spanning a wide range of industries including high technology, communications, transportation, and finance sectors. This dichotomy of resource versus non-resource stocks provides a natural partition to examine the role of industry factors on the valuation effects of US listing. Finally, several previous studies have examined the impact of US listing on risk, return and liquidity in pre-1990 Canadian listings which provide useful benchmarks to measure changes in the listing 1 See Pagano et al. (2002).

3 U.R. Mittoo / Journal of Banking & Finance 27 (2003) effects over time. Most of these studies, however, have concentrated on the shortterm effects of US listing. Recent literature suggests that long-run effects of an event can be significantly different from those in the short-term. In this paper, we also examine changes in long-term performance of US listing through time by comparing performance of pre-1990 and post-1990 Canadian listings. Our empirical analysis shows some interesting findings. We document significant differences in the short- and long-run performance of Canadian listings in the US. In the short-run, we document a decline in the positive impact of US listing on stock return and trading volume over time. In the long-run, Canadian listings underperform Canadian market indexes in both pre- and post-1990 periods. A typical Canadian stock listed in the US underperforms Canadian market indexes by percent during three years subsequent to US listing. Our results also suggest that the determinants of long-run performance are different from those in the short-run and that value of US listing is driven by factors that vary cross-sectionally and over time. Our study should be of interest to corporate managers, stock exchanges and academics. Whether US listing adds value to a firm is relevant to managers in their foreign listing decision. Further, Canadian stock exchanges have faced fierce competition from US exchanges in the last decade and have undertaken several measures to meet this challenge. To what extent, these measures have been successful in maintaining their market share in cross-listed stocks should be of interest not only to Canadian exchanges but also to other stock exchanges around the world. The rest of the paper is organized as follows. Section 2 presents literature review and testable hypotheses. Section 3 discusses the sample and methodology. Sections 4 and 5 present univariate and multivariate test results respectively. In Section 6, main findings and conclusions are presented. 2. Literature review and hypotheses Theoretically, the major financial benefits of foreign listing occur when national markets of different countries are segmented. Several theoretical models predict that listing a firmõs securities on foreign exchanges dismantles some of the investment barriers that induce market segmentation and results in a lower cost of capital and a higher stock price for the firm. 2 A large number of empirical studies have also examined this hypothesis with mixed results. While most studies document a decline in risk and expected return surrounding foreign listing consistent with market segmentation, some results are puzzling. For example, while firms from most geographical regions including Canada experience a decline in the estimated cost of capital after listing in the US, the opposite is observed for European (non-uk) firms. 3 Further, there is significant cross-sectional variation in the impact of foreign listing even 2 See for example, Stapelton and Subrahmanyam (1977), Errunza and Losq (1985), and Alexander et al. (1987). 3 Karolyi (1998) provides a comprehensive survey of international listing literature. The difference is most pronounced between UK and non-uk European firms; UK firms experience a decline of about 2.92% whereas non-uk European firms experience an increase of 0.49% in their cost of capital.

4 1632 U.R. Mittoo / Journal of Banking & Finance 27 (2003) among firms within a country. This evidence is consistent with findings of managerial surveys that motivations and net benefits of foreign listing vary widely across firms. Recent managerial surveys suggest that motivations of foreign listing may also have changed over time. Traditionally firms listed overseas primarily for financial reasons such as raising cheap capital and increasing liquidity but strategic reasons such as enhancing visibility or gaining entry into foreign markets are becoming increasingly important in recent years. 4 However, untangling these cross-sectional and time varying effects in a multicountry setting is a challenging task for researchers. We investigate these issues in a single country context. In this section, we develop testable hypotheses pertaining to the motivations and value of US listing over time for Canadian listings in the US. We focus on three major effects of US listing based on theoretical and empirical literature: capital market segmentation, stock liquidity, and industry-specific factors Capital market segmentation Although Canada and US are the worldõs largest trading partners, some barriers to capital flows (e.g., preferential tax treatment of dividends from Canadian firms and limits on foreign content in investments of pension funds) do exist between the two countries. Evidence also suggests that Canadian markets were segmented from the US markets until early 1980s but have become increasingly integrated as many regulatory barriers have fallen in recent years. 5 A major change occurred in July 1991 when Canada and the US implemented the Multijurisdictional Disclosure System (MJDS) that allowed eligible Canadian firms to list and issue securities in the US using the disclosure requirements of their home country. More importantly, the Canadian issuers did not have to be interlisted in the US to qualify under the MJDS. The MJDS was considered as a major first step in integrating the Canadian and US capital markets. Doukas and Switzer (2000) examine the impact of the MJDS on the Canadian market integration for Canadian listings in the period but do not find support for this hypothesis. A plausible explanation is that their sample period contained mostly the early stage of the MJDS implementation when it was available only to a few large Canadian firms. The major impact of the MJDS occurred after 1994 when its availability was expanded to a much larger number of Canadian firms and other regulatory changes were also implemented. 6 The number of Canadian 4 See for example, Bancel and Mittoo (2001), Fanto and Karmel (1997), and Mittoo (1992a) for survey literature and Foerster and Karolyi (1999), Miller (1999), Doidge et al. (2001), and Reese and Weisbach (2002) for tests of different hypotheses pertaining to the value of US listing. These include enhanced liquidity and investor recognition, access to larger and cheaper capital markets, better alignment of controlling and non-controlling shareholdersõ interests, and legal bonding hypotheses. 5 Jorion and Schwartz (1986), Mittoo (1992b), and Booth (1987) provide evidence of segmentation in Canadian markets in early 1980s. 6 These include harmonization of Canadian and US disclosure requirements and relaxation of reporting requirements by the US Securities and Exchange commission (SEC) for foreign firms issuing cross-border securities or listing in the US. In addition, the Canadian pension funds were allowed to gradually increase the foreign investment content from 10 percent of their book value to 30 percent during the 1990s.

5 firms issuing cross-border securities as well as listing in the US increased dramatically after This increased integration of the Canadian and US capital markets is likely to lower the benefits of US listing for Canadian firms. 7 We examine the following hypothesis: H1: The value of US listing for Canadian firms should decline as Canadian and US markets become increasingly integrated over time, all else equal Stock liquidity U.R. Mittoo / Journal of Banking & Finance 27 (2003) Theory suggests that listing on larger and more liquid stock exchanges enhances stock liquidity and lowers expected stock return. For example, in Amihud and MendelsonÕs (1986) theoretical model, expected stock returns are an increasing and concave function of stock liquidity measured by stockõs bid ask spread. Chowdhry and Nanda (1991) develop a model of multimarket trading which predicts that the trading of a security on multiple markets results in lower trading costs for that security because of increased competition among market makers. Canadian stock exchanges are generally smaller and less liquid than their US counterparts. Fowler et al. (1980) document that non-trading was a major problem on Canadian exchanges in the 1970s while Tinic and West (1974) report that bid ask spreads on the Toronto Stock Exchange (TSE), the largest Canadian stock exchange, were much higher compared to that on either the NYSE or even the over-the-counter market. 8 Managerial surveys show that most Canadian firms interlist on US exchanges to enhance stock liquidity. Evidence also supports that firms experience substantial gains in trading volume and significant declines in bid ask spreads after US listing. More importantly, liquidity gains are also major determinants of the positive impact on stock price surrounding US listing (see Karolyi, 1998). In the last decade, the TSE has adopted several measures in response to the increasing competition from US stock exchanges including extended hours of trading, increased transparency, and lower trading costs. A major change occurred on April 15, 1996 when the TSE switched to decimal pricing for stocks trading at $5 per share or more, and implemented a minimum quotation increment reductions (MQIR). Several studies document significant reductions in the quoted and effective spreads for the TSE stocks in the post-decimalization period. For example, Ahn et al. (1998) report that spreads decreased by 27% on the TSE but remained unchanged on the NYSE/American Stock Exchange (AMEX) for TSE stocks interlisted on the NYSE/AMEX while spreads decreased by 16% on the TSE and by 8% on the National Securities Dealers Automatic Quotation System (NASDAQ) for TSE 7 It is likely that the MJDS would have also reduced the cost of US listing for Canadian firms. However, Houston and Jones (1999) in a survey of Canadian firms report that the MJDS had little impact on listing costs for Canadian firms. 8 For example, only 37.5 percent of stocks listed on the TSE during traded at least once every month.

6 1634 U.R. Mittoo / Journal of Banking & Finance 27 (2003) stocks interlisted on the NASDAQ. 9 We would expect that liquidity gains of US listing for TSE stocks should decline as a result of these competitive measures by the TSE. Formally stated: H2: The liquidity gains for Canadian listings in the US should decline over time as Canadian exchanges become more competitive with US exchanges, all else equal Industry-specific factors In a recent survey, Fanto and Karmel (1997) report that industry-specific factors are cited as the main reasons by many non-us firms for listing in the US. These factors are likely to vary widely across firms and industries. For some firms, foreign listing may be part of their corporate strategy to expand foreign operations and to signal that they have become global players. Such firms may seek enhanced visibility and corporate profile as primary benefits from the US listing. Other firms may seek foreign listing because their business is not well understood on their home exchanges and listing on foreign exchanges, where their industry peers are also listed, could provide better valuation for their securities. For example, it is often argued in the financial press that high technology firms are likely to get better valuations by listing on NASDAQ. Finally, for some firms, following their industry peers on foreign exchanges may not be a matter of choice but a question of survival. Pagano et al. (2001) argue that cross-listing behaviour is likely to be affected by informational cascades and firms may imitate their industry peers because they may perceive that failing to do so might put their company at a competitive disadvantage in the industry. Such a strategy may, however, involve significant costs for some firms. Thus, whether the net benefits of foreign listing based on the industry-specific factors are positive or negative is unclear. Resource stocks comprise a significant proportion of the Canadian capital market and evidence also suggests that resource stocks are priced in a relatively integrated market compared to their non-resource peers. 10 Until 1990, Canadian listings in the US were also dominated by resource stocks but this trend has reversed as a majority of US listings in the 1990s are comprised of non-resource stocks. This dichotomy between resource and non-resource stocks allows us to test whether US listings are driven primarily by market segmentation or by industry 9 See for example, Chung et al. (1996), and Ahn et al. (1998). Both the NYSE and NASDAQ also reacted competitively and reduced their MQIR from 1/8th to 1/16th for their listed issues with a bid price of $10. Kryzanowski and Zhang (2002) show that the TSE price advantage changed significantly after both TSEÕs MQIR reduction and that of its US competitors. 10 Booth (1987) shows that preferential dividend tax effect for Canadian firms induces segmentation in Canadian market and since resource stocks generally pay less dividends, their stock is priced in a relatively integrated North American market compared to their non-resource peers. Foerster and Karolyi (1993) also confirm that resource stocks have a significantly lower price impact surrounding US listing than their non-resource counterparts while there is no difference in liquidity effects between the two in their pre-1990 sample.

7 U.R. Mittoo / Journal of Banking & Finance 27 (2003) factors. Under market segmentation hypothesis, the difference in the value of US listing between resource and non-resource stocks should decrease over time as Canadian markets become increasingly integrated with US markets. On the other hand, if industry factors are the major determinant of the value of US listing, this difference is ambiguous and could be positive, negative, or insignificant. We test the following hypothesis: H3: As Canadian and US capital markets become increasingly integrated over time, the difference in the value of US listing between the resource and non-resource Canadian stocks should decline, all else equal. 3. Data and methodology 3.1. Data Our pre-1990 and post-1990 samples consist of Canadian firms that interlisted in the US during and periods respectively. The initial sample is identified from lists provided by the NYSE, AMEX, and NASDAQ. The final sample consists of stocks that are listed on the TSE, listed for the first time in the US, and have daily stock return data available in the Canadian Financial Markets Research Centre (CFMRC) database. The post-1990 sample consists of 108 Canadian stocks, almost twice that in the pre-1990 sample (56), consistent with the accelerated pace of US listings in recent years. Foerster and Karolyi (FK;1993,1998) use a sample of 53 Canadian firms listed in the US during the period to examine short-run effects of Canadian listings in the US. Since our pre-1990 sample overlaps significantly with their pre-1990 sample, we benchmark our post-1990 results with their pre-1990 results for short-run analysis. For long-term analysis, we use our pre-1990 and post-1990 samples for comparisons. Table 1 presents the frequency distribution of the post-1990 sample by year, by the US exchange of listing, and by industry. Panel A shows that a preponderance of listings (about 70 percent) are clustered during the period that witnessed lowering of many regulatory barriers between the two countries. There are also some hot years of US listing; about half of NYSE listings occurred in 1996 and about 40 percent of NASDAQ listings are in the period. Both NYSE/AMEX and NASDAQ listed stocks are well represented. NYSE listings comprise about 30 percent of the post-1990 sample, almost three times higher than that in the pre-1990 sample, reflecting results of aggressive marketing efforts by the NYSE in attracting foreign listings. Table 1 (Panel B) compares industry characteristics of post-1990 listings with those of a pre-1990 sample analyzed in Foerster and Karolyi (FK;1993,1998). The differences between the two are striking. The pre-1990 sample is dominated by the Resource sector (62 percent) consisting of metals and minerals, gold and silver, oil and gas, and paper and forest products; some TSE 300 sectors have no US listing. In contrast, the post-1990 sample is a well-diversified group of stocks, representing

8 1636 U.R. Mittoo / Journal of Banking & Finance 27 (2003) Table 1 Sample distribution by year, by US exchange of listing and by industry Year NYSE AMEX NASDAQ Total Panel A: Listing frequency and location of listing on US exchanges (post-1990 sample) Total Industry Post-1990 sample Pre-1990 (FK, 1993) sample Number Percent TSE 300 weight (%) Number Percent TSE 300 weight (%) Panel B: Comparison of pre-1990 and post-1990 Canadian listings in the US by TSE industry group versus TSE 300 group weight Metals and minerals Gold and silver Oil and gas Paper and forest products Consumer products Industrial products Real estate and construction Transport and environment services Pipelines Utilities Communications and media Merchandising Financial services Conglomerates Total The post-1990 sample consists of 108 Canadian stocks that are listed on Toronto Stock Exchange and interlisted on US exchanges between 1991 and of the 14 industry groups that comprise the TSE 300 index. 11 The concentration of US listings, however, varies substantially across sectors. For example, Resource and Industrial Product sectors comprise about 32 percent and 30 percent of the 11 The TSE 300 Index is a Canadian market index comprising of 300 firms representing different industrial sectors and selected on the basis of liquidity and trading volume. The industrial composition is based on the TSE 300 index as on December 31, 1998.

9 post-1990 sample respectively while each of these comprises less than 20 percent of the TSE 300 index. The proportion of US listings in Communications and Media stocks is also much larger (10 percent) relative to their weight in the TSE 300 (4 percent). Utilities and Financial Services sectors, on the other hand, have a much higher weight in the TSE 300 index (35 percent) but few US listings (about 7 percent). The clustering of the listings in certain sectors suggests that industry factors may play an important role in the listing decision of Canadian firms Methodology U.R. Mittoo / Journal of Banking & Finance 27 (2003) Univariate tests: Short-term analysis Impact of US listing on risk and return. To analyze the impact of US listing on risk and return of stocks, we use an event study framework similar to that in FK (1993) and compare our post-1990 results with their pre-1990 results. 12 The Ôevent windowõ is defined as a seven-day window surrounding the listing date (three days before and three days after the listing date) and pre- and post-listing periods comprise 100 days prior and 100 days after the event window respectively. The market portfolio is proxied by the CFMRC value-weighted index, and daily abnormal returns for each stock are computed as stock returns minus the market returns on that day (excess-of-market returns). Risk of a stock is proxied using standard deviation of returns and the market risk (beta) Impact of US listing on stock liquidity. The changes in monthly trading volume are computed from three-months prior to six-months after the US listing. The pre-listing volume consists of volume on the TSE only whereas the post-listing volume consists of volume on both the TSE and US exchanges. We use two measures of volume similar to that in FK (1993): the number of shares traded (unadjusted trading volume), and the ratio of the stock trading volume to the TSE trading volume (TSE adjusted trading volume). To examine the effect of listing on trading costs, we follow FK (1998). The daily closing bid and ask quotes, trading costs, transaction prices, trade sizes and volumes for each day are measured in a 60-day window surrounding the US listing (from day t ¼ 30 relative to the listing day t ¼ 0 to day t ¼þ29) and these observations are averaged across securities on an equal-weighted basis. The trading cost is proxied using both spread premium (SP%) and liquidity premium (LP%) as follows: SP% ¼ðASK BIDÞ=MIDPOINT; ð1þ LP% ¼þ½PRICE MIDPOINTŠ=MIDPOINT; ð2þ where MIDPOINT is the midpoint of the ask and bid prices and PRICE is the actual transaction price. The impact on trading costs is measured in a cross-sectional regression framework after controlling for changes in price, volume, and trade size associated with US listings. 12 See Brown and Warner (1985) for detailed methodology.

10 1638 U.R. Mittoo / Journal of Banking & Finance 27 (2003) Univariate tests: Long-term analysis The long-term analysis is undertaken from 12 months prior to 36 months after US listing. 13 As a starting point, we benchmark the long-term performance of our sample stocks relative to the value-weighted (VW) Canadian market index. However, benchmarking with the VW index is problematic in the Canadian case because the US interlisted Canadian stocks are generally much larger than their domestic-listed counterparts and form a significant component of this index. To address this problem, we also benchmark performance with an equally weighted (EW) Canadian market index that reflects more the performance of the domestic-listed Canadian stocks. To minimize the impact of infrequent trading in estimation, we restrict the analysis to only those stocks that trade in the month of listing and in at least in 75 percent of the months in our analysis period (month )12 to month +36). A total of 87 stocks in the post-1990 sample and 41 stocks in the pre-1990 sample met these criteria. Monthly abnormal returns for each stock are calculated in excess of the market index returns. Average holding period returns in different subperiods are calculated as total geometric returns and as simple cumulative returns over consecutive months. Recent literature on long-term performance shows that significant biases can arise when benchmarking solely with indices. 14 To ensure robustness of our results, we also use two additional benchmarks. First, for each sample firm, we use an industry matched domestic listed TSE 300 index firm as a benchmark. Second, we also examine the operating performance of our sample firms from 12 months prior to 36 months after US listing Multivariate analysis: Short-term and long-term analysis A limitation of the univariate analysis is that it does not account for changes in risk that are also likely to accompany the US listing. Evidence suggests that after US listing, Canadian stockõs sensitivity to the US market factors is likely to increase while its sensitivity to the Canadian market factors is likely to decrease. 15 We conduct multivariate tests for both short- and long-run effects of US listing using a two-step procedure. In the first step, the abnormal returns are computed after adjusting for the market risk in a two-factor international asset pricing model (IAPM) using both Canadian and US market risk factors. In the second step, cross-sectional regressions of these abnormal returns are run on variables commonly employed in the foreign listing literature to gain some insights into the major determinants of short- and long-run performance. This methodology is discussed more fully in Section To our knowledge, Alexander et al. (1988) is the only previous study that has examined the impact of US listing on long-term performance of Canadian stocks in a sample of 13 Canadian listings in the period. 14 See for example, Barber and Lyon (1997) and Lyon et al. (1999). 15 For a survey of literature on changes in risk exposures around US listings, see Karolyi (1998). Typically, foreign companies listing on the US experience a decline in their home market beta and an increase in US market beta. These changes vary across firms from different geographical regions and are generally less significant for Canadian companies.

11 4. Empirical results 4.1. Short-term analysis U.R. Mittoo / Journal of Banking & Finance 27 (2003) Effect of the US listing on risk and return Table 2 provides results in the post-1990 sample and compares with those reported by FK (1993) in their pre-1990 sample. Panel A shows that mean abnormal return in the seven-day event window surrounding US listing has declined from 1.97% in pre sample to 0.68% in the post-1990 sample. This decline of over 60 percent appears consistent with the market segmentation hypothesis. However, a comparison of cumulative mean abnormal returns (CARs) from day )100 to day +100 shows an Table 2 Short-term analysis: Impact of US listing on risk and return of stocks; comparison of pre-1990 and post Canadian listings in the US 100 days pre-listing: )103 to )4 Event window 7 days ()3 to+3) 100 Days post-listing: +4 to +103 Pre post difference t-test Pre post W -test (signed rank) Panel A: Overall sample Post-1990 (N ¼ 108) Cumulative average 4.85% 0.68% )2.60% abnormal return (CAR) Average of betas Pre-1990 (FK, 1993) (N ¼ 49) Cumulative average 9.35% 1.97% )9.71% abnormal return (CAR) Average of betas Panel B: NYSE and AMEX versus NASDAQ (in Bold) Post-1990 samples NYSE and AMEX (N ¼ 37) versus NASDAQ (N ¼ 57) Cumulative average 2.87 (6.13)% )0.20 (1.245)% )0.06 ()0.42)% 1.78 (1.01) 1.94 (0.85) abnormal return (CAR) Average of betas 0.61 (1.20) 0.81 (0.93) )1.68 (2.21) )1.52 (1.86) Pre-1990 (FK, 1993) samples NYSE and AMEX (N ¼ 11) versus NASDAQ (N ¼ 38) Cumulative average 5.06 (10.59)% 4.20 (1.32)% )14.37 ()8.36)% 1.46 (1.99) 1.87 (1.95) abnormal return (CAR) Average of betas 1.46 (1.16) 1.53 (0.99) )0.19 (0.61) )0.27 (1.07) Panel C: RESOURCE versus NON-RESOURCE (in Bold) Post-1990 samples RESOURCE (28) versus NON-RESOURCE (N ¼ 66) Cumulative average 0.5 (6.7)% )0.60 (1.19)% )0.95 ()3.3)% 0.17 (2.05) 0.34 (1.87) abnormal return (CAR) Average of betas 1.04 (0.91) 0.96 (0.85) 0.54 (0.52) 0.62 (0.08) Pre-1990 (FK, 1993) samples RESOURCE (29) versus NON-RESOURCE (N ¼ 20) Cumulative average 1.99 (20.01)% 0.89 (3.54)% )7.31 ()13.19)% 0.92 (2.62) 0.59 (2.80) abnormal return (CAR) Average of betas 1.17 (1.31) 0.75 (1.64) 1.45 ()0.89) 2.00 ()1.05) The cumulative average abnormal returns are returns in excess of the market returns proxied by Canadian valueweighted market index. The parametric t-statistics (Column 4) and non-parametric Wilcoxon Signed Rank test statistics (Column 5) test that there is no difference in risk and return of listed stocks in pre- and post- listing periods.,, denote significance at less than 0.01, 0.05, 0.10 levels respectively.

12 1640 U.R. Mittoo / Journal of Banking & Finance 27 (2003) opposite pattern. The mean net abnormal return for the post-1990 listings in this window is 2.9 percent (derived by adding 4.85% from days )103 to )4, plus 0.68% from days )3 to +3, and minus 2.60% from days +4 to +103; see Table 2, panel A), almost twice than in pre-1990 listings (1.61 percent). This evidence is contrary to the market segmentation hypothesis. This difference cannot be attributed to the differences in average betas because the decline in betas is very similar in both pre- and post-1990 samples and is not statistically significant in either case. 16 The subsample analysis (Panels B and C) also exhibits substantial variation across stocks and over time. The pre-1990 NYSE/AMEX stocks have a much higher stock price impact (4.2 percent) compared to their NASDAQ counterparts (1.32 percent) in the seven-day window whereas the reverse holds for the post-1990 stocks. Panel C also reveals that positive abnormal returns surrounding US listings are driven primarily by non-resource stocks in both pre- and post-1990 listings. Moreover, there is no evidence that the difference in US listing effects between resource and nonresource stocks has declined over time as predicted in hypothesis 3 under market segmentation. These results suggest that effects of US listing may be driven by factors that may also be varying over time Effect on stock liquidity Effect on trading volume. Table 3 benchmarks changes in trading volume in post-1990 sample with those reported in FK (1993) in their pre-1990 sample. These comparisons show that trading volume gains for Canadian listings have declined significantly over time. For a typical pre-1990 stock, the monthly trading volume jumps from less than 1 million shares prior to listing to about 2 million shares after listing, an increase of over 120 percent. This increase is less than half (60 percent) in a typical post-1990 listed stock. The difference is even more severe when the TSE adjusted trading volume is used for comparison. The average increase in trading volume in the post-1990 stocks is only 40 percent in this case, almost one third of that in the pre-1990 listed stocks (112 percent). 17 The TSE trading volume in the interlisted stocks also shows a similar pattern of decline over time. Prior to 1990, the TSE trading volume in the US interlisted stocks increases by about 16 20% after US listing. This increase is much lower for the post-1990 stocks; the TSE adjusted trading volume in the interlisted stocks actually declines by )3.3 percent after US listing in post-1990 sample. Fig. 1 compares the cross-sectional variation in the trading volume gains in the post-1990 stocks with that reported in Mittoo (1997) in a pre-1990 sample. 18 These 16 We also examine the difference in average standard deviations in pre- and post-listing periods. These results are qualitatively similar to that of average beta, although the significance levels vary. For example, the average standard deviation in the post-1990 sample declines from 2.74 percent in the pre-listing period to 2.57 percent in the post-listing period. 17 The results are qualitatively similar using the monthly stock turnover as a proxy for trading volume. These results are not provided to conserve space. 18 The pre-1990 sample consists of 55 stocks and the change in average trading volume is measured from one year prior to one year after US listing.

13 Table 3 Short-term analysis: Effect on trading volume; comparison of pre-1990 and post-1990 Canadian listings in the US Month Average volume Volume as % of TSE Post-1990 Pre-1990 (FK, 1993) Post-1990 Pre-1990 (FK, 1993) Pre-listing volume on TSE )3 2,213, , % 0.193% )2 2,030, , % 0.203% )1 2,483, , % 0.219% Average ()3to)1) 2,242, , % 0.205% TSE only TSE and US TSE only TSE and US TSE only TSE and US TSE only TSE and US Post-listing volume on TSE only and on both TSE and US exchanges +1 2,434,509 3,754,670 1,523,045 2,848, % 0.245% 0.360% 0.673% +2 2,944,928 4,256,861 1,001,140 1,775, % 0.247% 0.227% 0.403% +3 2,798,051 4,057, ,883 1,386, % 0.243% 0.178% 0.308% +4 2,350,693 3,236,773 1,029,008 1,744, % 0.192% 0.231% 0.392% +5 2,309,407 3,180,325 1,044,262 1,989, % 0.179% 0.226% 0.431% +6 2,085,083 3,156, ,713 1,804, % 0.182% 0.203% 0.396% Average (+1 to +3) 2,726,962 4,035,279 1,108,356 2,003, % 0.245% 0.255% 0.461% Average (+4 to +6) 2,248,394 3,190, ,994 1,846, % 0.184% 0.220% 0.406% Average (+1 to +6) 2,486,565 3,586,871 1,054,175 1,924, % 0.213% 0.238% 0.434% % change from pre-listing ()3 to )1) to post-listing (+1 to +6) 10.88% 59.95% 19.69% % )3.29% 40.13% 16.10% % U.R. Mittoo / Journal of Banking & Finance 27 (2003)

14 1642 U.R. Mittoo / Journal of Banking & Finance 27 (2003) Fig. 1. Effects of US listing on total trading volume: Cross-sectional variation pre-1990 versus post-1990 samples. graphs show that the average effect may be misleading because it conceals a large cross-sectional variation across stocks and is strongly influenced by a few outliers. There are a few big winners and losers of trading volume gains after the US listing. While about one third of stocks are big winners experiencing increases of over 150 percent, about one third of stocks also suffer declines in trading volume after US listing. This pattern of significant cross-sectional variation is observed in all stocks, although the proportion of big winners is higher and that of the big losers is lower among stocks listed prior to A similar pattern is observed for trading volume on the TSE (Fig. 2). The subsample analysis by industry and location of US listing (not presented to conserve space) also supports a similar variation across stocks and over time Impact on the trading costs. Table 4 provides summary statistics of trading costs and trading volume in the pre-listing (days )30 to )1) and post-listing (days 0 to +29) periods. The post-1990 sample statistics are compared with those reported in 19 There may be double counting problem for the NASDAQ listed stocks, see for example, Heidle and Huang (2001). This is, however, not a major concern in our study since we focus on changes in volume over time.

15 U.R. Mittoo / Journal of Banking & Finance 27 (2003) Fig. 2. Effects of US listing on TSE trading volume: Cross-sectional variation pre-1990 versus post-1990 samples. FK (1998) in their pre-1990 sample. This table shows that average trading costs decline after US listing in both pre- and post-1990 listings, although the magnitude of decline varies depending on whether the LP% or SP% is used as a proxy. There are also changes in volumes, trade size, and other variables after US listing that are likely to influence the trading costs and need to be controlled for any meaningful comparison of trading costs. To account for these changes, we run cross-sectional regressions of trading costs on changes in volumes, trade size, and average price per share as follows: LP% it ¼ a 0 þ b 0 VOL it þ c 0 PR it þ d 0 TRD it þ D t ða 1 þ b 1 VOL it þ c 1 PR it þ d 1 TRD it ÞþE it ; ð3þ SP% it ¼ a 0 þ b 0 VOL it þ c 0 PR it þ d 0 TRD it þ D t ða 1 þ b 1 VOL it þ c 1 PR it þ d 1 TRD it ÞþE it ; ð4þ where VOL it is the trading volume for stock i on day t, PR it is the average transaction price for stock i on day t, TRD it is the average trade size for stock i on day t, and D t is a dummy variable that equals one for post-listing days (days 0 to +29). Both b and c coefficients are expected to have negative coefficients because trading costs are typically inversely related to the trading volume and stock price. The

16 1644 U.R. Mittoo / Journal of Banking & Finance 27 (2003) Table 4 Short-term analysis: Impact on stock liquidity; comparison of pre-1990 and post-1990 Canadian listings in the US Variable Pre-listing period (days )30 to )1) Mean Standard deviation Post-listing period (days 0 to +29) Mean Standard deviation % Change pre to post listing Post-1990 sample (N ¼ 108) LP% )5.51 SP% )1.46 Volume per day Average trade size Pre-1990 sample (FK, 1998) (N ¼ 52) LP% )1.67 SP% )3.75 Volume per day )4.55 Average trade size )5.28 Liquidity premium percentage (LP%) is the absolute difference between the trade price and midpoint of standing bid ask quotes, as a fraction of the midpoint. Spread percentage (SP%) is the difference between the ask and bid quotes, as a fraction of the midpoint. The number of quotes and trades represents the average number of quotes or trades per day, and the average volume represents the average number of shares traded per day. coefficient of d is expected to be positive because large trades are likely to be undertaken mostly by informed traders and therefore, market makers are likely to charge higher spreads for such trades. The coefficients multiplied with D t represent the post-listing incremental changes in each variable relative to the pre-listing variables, conditional on the listing not yet having occurred. Table 5 presents the regression results in the post-1990 sample. The results reported in FK (1998) in their pre-1990 sample are also provided for easy comparison. For the pre-1990 sample, there is a significant reduction in trading costs after controlling for determinants of the bid ask spreads, irrespective of whether LP% or SP% is used for analysis. The declines in LP% and SP% are )7% ()0.11 relative to 1.54) and )11% ()0.45 relative to 4.18) respectively, and are statistically significant. In contrast, while the signs of both a 1 coefficients (for LP% and SP%) are negative in the post-1990 sample, neither is statistically significant. This evidence supports that the impact of US listing on trading cost has declined through time. To examine the impact of multimarket competition predicted in Chowdhry and Nanda (1991), we also run separate regressions in subsamples partitioned by volume increasing and volume decreasing firms similar to that in FK (1998). These results are presented in Panel B and are very similar for both pre- and post-1990 listings. Only volume increasing firms experience a reduction in trading costs in both samples. In fact, a 1 coefficients are positive for volume decreasing firms, suggesting that trading costs actually increase for these firms after US listing.

17 Table 5 Short-term analysis: Impact of US listing on trading costs; regression tests of differences in spread and liquidity premiums around US listing Model a 0 b 0 c 0 d 0 a 1 b 1 c 1 d 1 R 2 Panel A: All US listings: Pre-1990 versus post-1990 samples Post-1990 sample LP% ) ) ) ) ) (42.540) ()8.013) ()16.429) (3.305) ()0.469) (2.287) ()0.148) ()3.07) SP% ) ) ) ) (50.202) ()9.998) ()19.765) (3.158) ()1.371) (2.920) ()0.396) ()2.907) Pre-1990 sample (FK, 1998) LP% ) ) ) ) (36.2) ()8.62) ()13.3) (0.67) ()1.79) (2.07) ()0.23) (0.82) SP% ) ) ) (38.8) ()9.52) ()13.5) (1.86) ()2.98) (2.56) (0.95) (0.87) Subsample analysis: Panel B: Volume-increasing versus volume-decreasing firms Post-1990 sample Volume-increasing firms (N ¼ 44) LP% ) ) ) ) (31.523) ()6.381) ()11.690) (0.813) ()2.782) (1.836) (1.095) ()0.190) SP% ) ) ) ) (35.415) ()7.267) ()13.447) (0.889) ()3.256) (1.695) (1.779) ()0.133) Volume-decreasing firms (N ¼ 50) LP% ) ) ) ) ) (29.143) ()5.293) ()11.549) (3.802) (1.810) ()1.523) ()0.310) ()2.288) SP% ) ) ) ) (35.581) ()7.584) ()14.256) (3.920) (0.793) ()0.931) (0.106) ()2.06) Pre-1990 (FK, 1998) sample Volume-increasing firms (N ¼ 22) LP% ) ) ) ) (20.4) ()5.19) ()13.4) (1.24) ()2.71) (3.43) (1.15) ()0.34) U.R. Mittoo / Journal of Banking & Finance 27 (2003)

18 Table 5 (continued) Model a 0 b 0 c 0 d 0 a 1 b 1 c 1 d 1 R 2 SP% ) ) ) ) (23.4) ()6.38) ()14.2) (2.50) ()3.55) (4.29) (1.71) ()1.13) Volume-decreasing firms (N ¼ 30) LP% ) ) ) ) ) (29.9) ()6.41) ()9.69) ()1.81) (2.16) (0.85) ()2.37) ()1.09) SP% ) ) ) ) ) (32.8) ()8.10) ()18.7) ()1.14) (1.09) ()0.89) ()1.74) (0.57) Panel C: TSE volume Winners versus Losers firms Post-1990 sample TSE volume Winners firms (N ¼ 46) LP% ) ) ) ) (25.950) ()5.878) ()9.383) (1.725) (0.989) (0.720) ()0.405) ()1.646) SP% ) ) ) ) (36.031) ()8.892) ()13.319) (1.727) ()0.913) (1.858) (0.392) ()1.624) TSE volume Losers firms (N ¼ 46) LP% ) ) ) ) ) (25.921) ()4.864) ()10.521) (3.584) ()0.205) (1.993) ()0.322) ()2.091) SP% ) ) ) ) ) (28.801) ()5.952) ()11.599) (3.519) ()0.080) (2.509) ()0.330) ()2.055) Pre-1990 (FK, 1998) sample TSE volume Winners firms (N ¼ 16) LP% ) ) ) ) (29.3) ()6.56) ()15.4) ()0.64) (0.98) (0.89) ()2.00) (1.08) SP% ) ) ) (32.2) ()7.54) ()17.5) (0.77) (0.15) (0.36) ()1.51) (1.25) 1646 U.R. Mittoo / Journal of Banking & Finance 27 (2003) TSE volume Losers firms (N ¼ 16) LP% ) ) ) ) (18.4) ()6.44) ()12.1) (1.41) ()3.41) (4.82) (1.03) ()1.02)

19 SP% ) ) ) ) (21.1) ()7.98) ()12.9) (2.02) ()4.78) (5.71) (1.72) ()1.03) Comparison of pre-1990 versus post-1990 Canadian listings in the US. Liquidity premium (LP%) is defined as the absolute difference between the trade price and midpoint of standing bid ask quotes as a fraction of the midpoint, and spread premium (SP%) is defined as the difference between the ask and bid quotes as a fraction of the midpoint. To test for differences between the pre-listing (day )30 to )1) and post-listing (days 0 to +29) trading costs, we estimate the following regressions: LP% it ¼ a 0 þ b 0 VOL it þ c 0 PR it þ d 0 TRD it þ D it ða 1 þ b 1 VOL it þ c 1 PR it þ d 1 TRD it Þþe it SP% it ¼ a 0 þ b 0 VOL it þ c 0 PR it þ d 0 TRD it þ D t ða 1 þ b 1 VOL it þ c 1 PR it þ d 1 TRD it Þþe it where VOL it is the total trading volume for stock i on day t,pr it is the average transaction price for stock i on day t,trd it is the average trade size for stock i on day t, and D t is a dummy variable that equals one for days 0 to +29. Both b and d coefficients are multiplied by T -statistics are reported in parentheses below each coefficient. R 2 is the coefficient of determination adjusted for degree of freedom.,, denote significance at the 0.01, 0.05, and 0.1 levels respectively. U.R. Mittoo / Journal of Banking & Finance 27 (2003)

20 1648 U.R. Mittoo / Journal of Banking & Finance 27 (2003) We also compare the effect on trading cost for the TSE volume increasing (winners) and volume decreasing (losers) firms. These results are strikingly different in the pre- and post-1990 samples (Panel C). For the pre-1990 sample, the TSE losers also experience a significant decline in trading costs while there is no effect on the TSE winners. In comparison, there is no significant impact on either the TSE winner or TSE loser stocks in post-1990 listings. FK (1998) interpret the pre evidence of a decline in trading costs for the TSE losers as consistent with the prediction in Chowdhry and Nanda (1991) of increased market competition among market makers. They argue that because of substantial competition from US market makers, TSE market makers had to respond to TSE volume losers but they did not have to respond to TSE volume winners. Following their logic, TSE market makers ought to have responded to TSE losers in post-1990 sample as well. However, a lack of such response by TSE market makers to TSE losers in post-1990 sample suggests that the TSE may have become more competitive with US exchanges in the 1990s. This interpretation is also supported by the observed decline of about 25 percent in average effective spreads on the TSE prior to US listing, from 1.12 percent in the pre-1990 sample to 0.85 percent in the post-1990 sample (see Table 4). Additional subsample analysis (not presented for brevity) reveals no significant differences between the NYSE/AMEX and NASDAQ listed stocks in both pre and post-1990 samples but some differences in resource versus non-resource stocks. The non-resource stocks experience significant decline in trading cost compared to the resource stocks in the post-1990 sample while there is no difference between the two in the pre-90 sample (FK, 1998). In summary, the short-term analysis shows that the positive impact of US listing on both trading volume and trading costs has declined over time. However, there are significant differences in the cross-section and over time which suggest that determinants of the US listing effect may be influenced by many factors that may also be time varying Long-term performance Market adjusted return performance Table 6 presents cumulative abnormal returns (CAR) employing both Canadian VW and EW indexes and Fig. 3 plots CARs from 12 months prior to 36 months after US listing for both pre- and post-1990 samples. Table 6 also presents analysis of subsamples partitioned by US exchange of listing and by resource and non-resource firms. Three observations are noteworthy from this analysis. First, firms generally list in the US after a strong market performance but experience a steady decline in performance after listing. Surprisingly, these results hold for both pre- and post-1990 listings. In fact, post-listing decline is even more severe for stocks listed prior to These stocks not only lose all abnormal returns earned prior to listing but end up with negative cumulative abnormal returns ranging from )6 percent to )15 percent three years after listing. Second, performance is generally lower when the EW rather