The cost of capital effects of overseas listings:

Market sequencing and selection

Sergei Sarkissian Michael J. Schill McGill University University of Virginia

April 14, 2003 ___________________________ * Sarkissian is from the McGill University Faculty of Management, Montreal, QC H3A1G5, Canada. Schill is from the Darden Graduate School of Business Administration, University of Virginia, Charlottesville, VA 22906, USA. Sarkissian may be reached at: (514) 398-4876, -3876 (fax), or sergei.sarkissian@mcgill.ca. Schill may be reached at: (434) 243-7736, -7676 (fax), or schill@virginia.edu. We thank Joshua Coval, Protiti Dastidar, Vihang Errunza, Erik Lie, Vijay Singal, Wanda Wallace and participants of workshops at College of William & Mary, McGill University, Virginia Tech, and the FMA-2002 meeting for comments and suggestions. Sarkissian acknowledges financial support from FCAR and IFM2.

1

The cost of capital effects of overseas listings:

Market sequencing and selection

ABSTRACT Using a broad sample of 1298 overseas listings spanning most world markets, we examine the long-run times-series and cross-sectional behavior of returns for firms choosing to list in foreign markets. Over an extended event window of ten years before and after the listing, we identify short-term and long-term patterns in firm returns that vary with different home and host markets for securities, listing sequencing, and firm characteristics. Controlling for the temporary return effects associated with exchange listing, we find that the cost of capital gains to overseas listings are more modest than those reported in earlier studies: the average cost of capital decline is 2.5 percentage points for the entire sample of cross-listings and 5.9 percentage points for non-U.S. stocks listed in the U.S. We find that contrary to the conventional notion of international diversification in segmented markets, firms that list across markets with low return correlation experience relatively little to negative improvements in cost of capital. Rather, our evidence suggests that firms obtain the most substantial improvements in cost of capital by listing in markets with better shareholder protection. JEL classification: G15; G32

2

1. Introduction

This paper considers the choice of security trading venue for firms in a cross-border market

context. In a frictionless world market, firms are indifferent to the choice of security trading

venue. However, cross-border barriers to investments (Black, 1974; Solnik, 1974; Stulz, 1981;

Errunza and Losq, 1985) and information flow (Merton, 1987), as well as market differences in

liquidity (Tinic and West, 1974; Foerster and Karolyi, 1998; Domowitz et al., 1998; Werner

and Kleidon, 1996), disclosure requirements (Biddle and Suadagaran, 1992; Fuerst, 1998;

Huddart et al., 1999), and shareholder expropriation risk (La Porta et al. 1997, 1998; Reese and

Weisbach, 2002; Doidge et al., 2003), make the selection of the listing venue an important

corporate decision. Studies such as Foerster and Karolyi (1999) and Errunza and Miller (2000)

suggest that non-US firms listing their equity in the United States experience dramatic declines

in their cost of capital. They attribute this effect to a reduction of the super risk premium

warranted of stocks in segmented and disadvantaged markets. Errunza and Miller (2000), for

example, find that foreign firms achieve on average an 11 percentage points reduction in the

cost of equity capital. Panel A of Table 1 reviews the major studies that compute the cost of

capital gains for cross-listed firms. Using at most three years of return data before and after the

listing, nearly all of the studies find positive abnormal returns prior to the listing followed by

negative abnormal returns after the listing.1

In this paper we raise several concerns with this literature. Our first concern is that

current estimates of the reduction in firm cost of capital are more appropriately attributable to

temporary abnormal effects around the listing than to equilibrium performance as asserted. A

growing literature documents substantial long-run abnormal performance before and after

major corporate events, such as domestic equity listing and equity issuance. Sanger and

McConnell (1986) McConnell and Sanger (1987) and Dharan and Ikenberry (1995) report

abnormal returns around domestic equity listing, while Asquith and Mullins (1986), Ritter

(1991), and Loughran and Ritter (1995) around equity offering (see Panel B of Table 1).

Foerster and Karolyi (2000) observe long-run abnormal returns prior to and following global

equity offerings. Although there is some debate about whether the abnormal performance is

consistent with explicit market timing, there is substantial evidence that the valuation effects

during the periods surrounding major corporate events are unusual. Therefore, firms with

1 See Karolyi (1998) and Stulz (1999) for a comprehensive review of the literature on overseas listings.

3

overseas listings may experience valuation effects similar to those observed among firms that

list their shares on a new domestic exchange or issue new equity. If returns surrounding an

event such as a foreign market listing are abnormal, studies that use the return data from this

period are likely to experience substantial bias in their estimate of equilibrium effects such as in

measuring changes in the cost of capital.

Our second concern with the literature is that work to date on the cost of capital effects

of overseas listings is almost exclusively based on the U.S. host market experience.2 Given that

the U.S. market is substantially larger than all other markets at this time and that foreign

listings on U.S. markets represent less than a third of all foreign listings (Sarkissian and Schill,

2003), it is reasonable to question the representativeness of this sub sample.

Our third concern with the literature is that almost no work on overseas listings

explicitly controls for the sequence of foreign market listings followed by a firm.3 Studies that

only look at the cost of capital effects of listings on one market, such as the U.S. market, treat

all such listings the same. In such a context the 1997 ADR listings on the NYSE of French

telecommunication companies Alcatel and France Telecom would be considered equivalent

despite the fact that at the time Alcatel stock was already listed on six other foreign markets

while France Telecom was only listed in Paris. If there are diminishing gains to foreign listing,

then ceteris paribus one would expect the most profound response to be associated with a

firm’s first overseas listing. Since the motives and gains for subsequent foreign listings may be

very different from initial listings we expect that it is important to control for a firm’s listing

sequence. The impact of foreign market selection and listing sequencing on firm performance

has not been yet appreciated in the overseas listing literature.

Finally, current research on overseas listing motives focuses primarily on evaluating

individually the various hypotheses proposed for listing. The conventional motives include

diversification gains, information benefits, and market-wide improvements in liquidity,

disclosure, and shareholder protection laws. We are concerned with which hypotheses can best

2 Studies that examine non-US listings include Howe and Madura (1990), Lau et al. (1994), Pagano et al. (2002). 3 The only paper that we are aware of which attempts to disentangle the effects of listing sequencing is that of Howe and Kelm (1987). The authors examine listings of U.S. firms on four overseas markets. As a result of small and non-representative sample, they fail to find any substantial sequencing effects. More importantly, they find no positive abnormal returns before listings or negative abnormal returns after listings, thus contradicting many earlier and later studies. Chaplinsky and Ramchand (2000) test for sequence effects in the announcement returns associated with global equity offerings. They find a modest and statistically insignificant decline in the announcement returns with subsequent offerings.

4

explain the cost of capital gains of listing. A test that “horse races” the predominant

hypotheses seems appropriate.

This paper addresses these four concerns. We accomplish our goal using a

comprehensive, hand-collected sample of firms with overseas listings from all geographic

regions that have overseas listing by the end of 1998. It consists of 816 firms from 35 countries

with 1298 listings placed in 24 foreign host markets. Thus, we are able to examine the

performance of firms originating not only from or listed in the U.S. but also from and in other

markets. We compute residual returns based on various market model specifications using a

wide window comprised of 120 months prior to the listing and 120 months after the listing.

This extended window allows us to better differentiate abnormal return event effects from

equilibrium returns. Since about 41% of our dataset includes firms with more than one

overseas listing, we are able to evaluate the effects of listing sequence on returns. Our broad

sample provides for stronger tests with respect to the cross-section of market and firm

characteristics. In particular, it helps us identify which standard hypotheses best explain how

foreign listings affect the costs of capital.

Our tests results allow us to make several observations. First, we document that

overseas listing is associated with a fundamental change in return behavior. Firms with

overseas listings experience a substantial and sustained pre-listing run-up and a profound post-

listing decline in their returns that takes place over the five years before and after the listing,

respectively. Second, we detect a modest but significant cost of capital reduction during the

five to ten year period following the listing relative to the equivalent pre-listing period. Third,

we show that return patterns vary widely with different home and host markets for securities,

listing sequencing, and firm characteristics. We estimate that the average cost of capital

decline is 2.5% for the entire sample of cross-listings and it is 5.9% for non-U.S. stocks listed

in the U.S. The first listing is associated with very large temporary abnormal returns around

the listing; however, we do not find statistical support for diminishing cost of capital gains to

overseas listing. The gains to initial listings are statistically no different than those of

subsequent listings by the same firm. We observe that firms producing tradable output

experience greater reduction in cost of capital, implying that firms that are relatively more

familiar to foreign investors before the listing benefit more by the listing. Finally, contrary to

the conventional notion that cross-listed firms achieve cost of capital gains through

diversification, our results show that these gains are primarily related to the better investor

5

protection environment of the foreign market. In fact, firms that list across markets with low

return correlation experience relatively little to negative improvements in their cost of capital.

We do not test any hypotheses regarding the source of the particularly strong returns

during the five years before the listing and the particularly weak returns during the five years

after the listing. This is not the object of this paper. There is a large literature on the hazards

associated with measuring long-run abnormal performance. Recognizing this difficulty, our

tests are simply meant to identify changes in firm cost of capital. We make no claims regarding

pricing efficiency issues in this market.

The rest of the paper is organized as follows. Section 2 describes our sample of

overseas listings, return data, as well as methodological issues. Section 3 is devoted to the

examination of market timing and cost of capital effects across different home and host markets

for listings and listing sequencing. Section 4 reviews the perceived motives for issuing shares

in foreign markets and analyzes the relation between the valuation effects of subsequent

overseas listings and market-wide characteristics. Section 5 reports the valuation effects of

cross-listings with respect to firm characteristics. Section 6 concludes.

2. Data and Methodology

2.1. Listing and return data

We survey the world stock exchanges to identify the universe of foreign listed equities at the

end of 1998. In constructing our sample, we excluded investment funds or trusts as well as

firms from countries with unreliable or limited financial data and corporate tax havens, such as

the Bermuda, Cayman Islands, Jersey, and others. We use stock exchange, financial institution,

public sources, or in some cases the firms themselves to identify listing dates. We restrict our

sample to those firms for which we also have return data from CRSP or Datastream within 120

months of the listing date. This selection procedure gives us 1298 foreign listings from 816

firms. The sample is broad and balanced with no country contributing more than 20 percent of

the listing firms. We know of no other study that includes a more comprehensive sample of

overseas listings and returns.

Panel A of Table 2 gives the information on the country-to-country distribution of

overseas listings. There are 35 countries that place shares overseas and 24 countries that host

6

shares from overseas. The largest numbers of foreign listings come from the U.S., Canada, and

Japan – 352, 148, and 148, respectively, although the majority of Canadian listings are in the

U.S. Some countries, such as Luxembourg and Switzerland, have very few firms traded

overseas, yet they host a substantial number of listings from other countries.

Our sample of cross-listings contains firms with both only one overseas listing and

multiple (two and more) listings on foreign exchanges. The occurrence of single and multiple

listings varies widely across calendar periods as well as home and host countries. It is

important to note that the listing sequencing that we present below may not necessarily

coincide with the actual listing chronology. Our sample is not immune from the survivorship

bias. Some firms may have placed their shares in foreign markets but then delisted them before

1998. These listings are not represented in our sample. The sequencing documented in our

study presents the effects of a relative sequence of overseas listings not an absolute sequence.

Moreover, we do not require that the first foreign listing is preceded by a domestic listing.

Some firms in our sample list on foreign exchanges before listing in their home market. This

phenomenon is particularly strong among Canadian and Israeli firms.

Panel B of Table 2 shows the distribution of overseas listings based on the listing

number across four calendar periods: before 1970, 1970-1979, 1980-1989, and 1990-1998. We

provide the frequency of the first, second, third, fourth, fifth, and the sixth and higher number

listings. The listing number is determined based on the actual available listing dates as of 1998.

For instance, if the return data for some firm with multiple listings is available only starting

with the second listing, the second listing is considered as a second listing in our dataset and no

first listing appears in the sample.

Since the latest period in our sample coincides with a high cross-listing activity

throughout the world, and due to survivorship bias, it is not unexpected to find that the largest

number of listings (661 out of 1298) is placed on foreign exchanges in the 1990s. Our sample

has 120 listings with the listing dates before 1970. More important is the break-up of the

listings by listing number. There are 764 first listings and 534 second and higher number

listings. The sample contains 60 listings that are the sixth or higher listing numbers for their

respective firms.

Panels C and D of Table 2 show the distribution of listings across home and host

markets for different listing numbers. Panel C presents the distribution of listings across 35

domicile countries. Most of developed markets tend to be dominated by firms that maintain

7

multiple listings. The only emerging market with significant number of multiple listings is

South Africa. There are ten countries with firms that have placed their shares in six or more

foreign markets. The firm with most overseas listings is German firm Bayer with 11 listings.

Both German firm Volkswagen and Phillip Morris from the U.S. have 10 listings.

Panel D of Table 2 shows the frequency distribution of foreign listings across 25 host

markets. Some host countries appear to be particularly attractive to firms making higher-order

listings. For instance, Italy maintains only two foreign listings in our sample – the 11th listing

of Bayer and the 10th listing of Volkswagen. Other countries that host a large proportion of

multiple listings relative to the number of first listings are Austria, Belgium, France, Germany,

Japan, Spain and Switzerland. Luxembourg, on the other hand, with more listings than, say,

France, has very few second or higher listings relative to the large base of first listings. This is

not unexpected since many smaller firms and firms from emerging countries often list in

Luxembourg before they become eligible for trading at larger world exchanges.

Finally, Panel E of Table 2 shows the median lag length in calendar months between

two consecutive listings. We provide summary statistics across the whole sample of listings as

well as the three sub-samples: US listings abroad, overseas listings in the U.S., and overseas

listings issued outside the U.S. Across all firms, the median lag between first and second

listing, second and third, and third and fourth are 33, 27, and 24 months, respectively. The

pattern suggests that firms with multiple foreign listings list in additional foreign markets every

two to three years on average.

Table 3 presents the distribution of sample monthly return observations by respective

home countries. Individual security returns of US firms are from CRSP; non-US firms are

from Datastream. All non-US firm returns are converted to US dollars using the corresponding

monthly exchange rates from Datastream. The excess returns are constructed by subtracting

the return on the 30-day US Treasury Bill. For each country, the table shows the number of

monthly observations available, the number of firm-listing months, as well as the respective

number for a sample that excludes the overlapping returns for firms that list in multiple

markets. It also reports the first and the last observation dates. There are many countries,

mainly from emerging markets, for which the return series are not available for the entire 241-

relative-month window. The smallest windows are for Polish and Portuguese firms with only

43 and 41 calendar months of observations, respectively. For Peru, Portugal, and Spain, we

have only one firm generating the monthly return series. Since the CRSP return data is

8

available well before the Datastream data, all returns between 1926 and 1972 are for US

companies. The last return date is December 2001 for all countries but Denmark and Spain, for

which it is 1997-05 and 1999-05, respectively. Given our ten-year post-listing window, this

implies that last listings from those two countries occurred in 1987-05 and 1989-05,

respectively.

2.2. Methodological issues

To estimate the impact of foreign listings on the cost of capital, we employ a two-stage

procedure. In the first stage, we compute a times series of residual returns for each firm listing

for all available months around the event window.4 We compute residual returns based on

three common market risk models namely,

Model 1: itLtiLiit erbar ++= , (1)

Model 2: itFtiFLtiLiit erbrbar +++= , (2)

Model 3: itWtiWLtiLiit erbrbar +++= , (3)

where itr , Ltr , Ftr , and Wtr are the excess returns on firm i, on the home market index, on the

host market index, and on the world market index in US dollars at time t, respectively. The

market index return for the U.S. is the CRSP value-weighted market return; the returns for the

world market portfolio and individual countries are the returns on the corresponding market

indexes from Datastream. 5 Our objective is to calculate residual returns, not abnormal returns

as in a typical event study. Consequently, we restrict the coefficients in these regressions to be

4 Due to difficulties with obtaining the announcement dates, we analyze performance of stock returns only around listing dates. Most of announcement dates occur within close to listing dates. In this respect, Forester and Karolyi (1999) report that the median difference between the announcement and listing dates of foreign firms in the U.S. is 44 days while the proportion of these firms that had announcement dates 100 or more days before listing is negligibly small. Since our analysis covers an extended period before and after the listing date, any effect from the listing announcement will be reflected in the stock return pattern immediately before the listing. 5 It is important to note that any model-based computation of residual returns is likely to be misspecified around listing dates. It is unclear if domestic market betas represent the entire risk of firms that are about to be cross-listed in a foreign market. Errunza and Miller (2000) point out, for example, that when firms start listing abroad, some market-wide liberalization processes can also be simultaneously taking place. This concern may be especially relevant for firms with multiple listings since they are among the largest and well-known companies in their respective countries.

9

constant over the full event window. We impose this restriction because we want to embed any

cost of capital changes into the residual term that we use in the second stage estimation. If we

allow the coefficients to absorb the cost of capital changes, our second stage tests have no

power. We recognize that this restriction assumes that the only systematic changes in cost of

capital over the extended event window are due to the foreign listing. We provide some tests of

this important assumption in Section 3.1. Alternatively, we could follow Foerster and Karolyi

(1999) and others and first estimate market betas in models (1-3) over the earliest pre-listing

period and then use these estimates in computing residual returns in the next periods. We do

not take this path because it substantially shortens our observation period.

In the second stage, we regress the residual returns on various market, sequence, and

firm characteristics to identify the magnitude and cause of any changes in firm returns. In

testing the impact of these characteristics, we use the following general regression setup,

omitting the time and firm subscripts,

( ) ( ) ( )( ) ( ) ( ) ( ){ }( ) ( ) ( ) ( ){ }( )

1 1 1 3

2 2 2 2

3 3 3 3

4 4

60, 1 , 59 60, 120

60, 1 , 59 60, 120

60, 1 , 59 60, 120

60,

e D t t D t t D t t

D Home M D t t D t t D t t

D Host N D t t D t t D t t

D First Listing D t t

α φ θ δ

α φ θ δ

α φ θ δ

α φ

= + − − + + + + +

+ = + − − + + + + +

+ = + − − + + + + +

+ + −( ) ( ) ( ){ }( ) ( ) ( ) ( ){ }( ) ( )

4 4

5 5 5 5

1 910

1 , 59 60, 120

60, 1 , 59 60, 120

1926 ... 2001

D t t D t t

D Firm X D t t D t t D t t

D Feb D Dec u

θ δ

α φ θ δ

η τ η τ

− + + + + +

+ = + − − + + + + +

+ = + + = +

, (4)

where e is the series of residual returns from models (1, 2, or 3), D(t-J,t-K) is a dummy variable

that is equal to 1 if the return occurs between event-time months J and K, D(Home=M)

indicates whether the firm originates from country M, D(Host=N) indicates whether the listing

is placed in market N, D(First Listing) indicates whether the listing is the earliest listing in our

sample for that firm, D(Firm=X) indicates whether the listing firm has characteristic X. These

variables enter the regression model separately and through the interaction with each other.

D(τ=Z) indicates whether the return occurs in calendar month Z. We include 910 monthly

calendar time dummy variables because of possible colinearity across observations in any

common calendar month.

10

3. Firm Return Performance around Listing

3.1. Aggregate performance

Figure 1 shows the annualized moving-average of gross and residual returns over the +/- 120-

month window around the listing date for all the listings in our sample. The residual returns are

computed based on the local-market model (Model 1). To annualize the monthly return series

we average the monthly returns across firms in event time and then cumulate the aggregate

returns over months t-12 to t+11, where t is the listing month, and divide by two.

Both gross and residual returns show a similar pattern composed of four periods. In the

first period, ending about five years before the listing date, the residual returns fluctuate around

the zero return mark. In the second period, within approximately five years before the listing,

there is a profound increase in both gross and residual returns. On average, the annualized

gross return rises from about 10% to 20%. This effect is consistent with the documented

increases in returns in other long-run performance studies (see Table 1). The third period

covers the interval from the time of listing to about five years after the listing. During this

period, we notice a profound and steep decline in returns. At its minimum, the annualized

gross return is about 7%. However, after the initial large decline, returns increase and generally

level off during the last period between five and ten years after the listing. Note that the

residual return series suggests that the level of returns in the fourth period is markedly lower

than that in the first period. Although this decline in the range of two-three percent is clearly

visible, it is much smaller than the change in returns between periods 2 and 3.

To examine the above patterns in statistical terms, we regress monthly residual returns

from the three market models on dummy variables associated with the four relative periods

suggested by Figure 1. We report our results in Table 4.6 In Panel A we use an extreme skip

period in our cost of capital tests; we break periods 1 and 2 at month -60 and break periods 3

and 4 at month +59. The first regression in Panel A compares the difference between the

average residual return of periods 2, 3, and 4 to that of the first period (t-120 to t-60). The

regression coefficient indicates that in period 2 the average monthly return is 0.42% per month

(5.0% a year) higher than in the preceding period. This increase in returns is followed by an

almost equally strong decline during the first 60 months after the listing. The average monthly

residual return in period 3 is lower by 0.36% per month (4.3% a year) than during period 1. In

11

the last period (starting 60 months after the listing), the average return is smaller in magnitude

but still negative and statistically significant relative to period 1. The coefficient on the period

4 dummy is -0.21% per month or -2.5% per year. We interpret this estimate as the average cost

of capital reduction for a cross-listed firm in our sample.

In the second regression in Panel A, we shift the reference period from period 1 to

period 2. Our interest is to replicate the tests of other studies that measure the effect of

overseas listing on the cost of capital using a smaller event window. The results indicate that

the prelisting increase in residual returns is a distinct phenomenon. The residual returns during

this period are statistically higher than during all other valuation periods. If we presume that

the difference in returns between period 2 and period 3 measures listing-induced change in cost

of capital, the magnitude of the change is much greater as indicated by the coefficient on the

period 3 dummy variable of -0.79% or -9.5% annually. Note that at least some of the prelisting

run-up in prices is fully in line with the notion that in the anticipation for a firm to become less

segmented from the world market, the firm’s higher prices reflect lower expected returns and

the decrease in the future cost of capital. However, a substantial and sustained

underperformance of cross-listed stocks after the listing that lasts three to five years is also

consistent with a “window of opportunity” explanation of the returns in periods 2 and 3. The

returns within three to five years of an overseas listing may reflect both timing effects as well

as reductions in the cost of capital. The inability to disentangle these two effects motivates cost

of capital tests that use an extended skip period.

In regressions 3 and 4 of Panel A, we repeat the regression 1 specification with other

market benchmark adjustments. The inferences are similar with the different models. The only

difference we note is a marginal decrease in the magnitude and significance of the period 4

coefficient, suggesting that the cost of capital may be even less than the annual decline of 2.5%.

We recognize that skipping ten years of data may appear an overly aggressive solution to the

measurement problem. In Panel B we reduce the skip period to six years by extending period 1

to month t-37 and period 4 to month t-36. This specification causes a dramatic increase in the

magnitude of the estimated cost of capital gain. The period 4 coefficient increases in absolute

value by more than 75% from -0.21% to -0.37% for the Model 1 regressions. In Panel C we

increase the skip period to 12 years by reducing period 1 to month t-85 and period 4 to month t-

84. With the longer skip period, the coefficient on the period 4 dummy across the three models 6 We do not report the calendar time dummy coefficients for the sake of convenience.

12

is not materially different from the respective coefficients in Panel A. The standard errors,

however, increase substantially.

Based on the results from the various event windows and market adjustment models, we

conclude that the 120-month skip period is the best specification for reducing the probability of

capturing abnormal return behavior and minimizing the reduction in power caused by the

reduction in data. For the results presented in the remainder of this paper, we report only tests

done with the local market model (Model 1) and the 120-month skip period. We have

conducted the tests using the other specifications as well. Although the choice of the market

model makes little difference, the results are somewhat sensitive to the choice of the skip

period length.

As a final specification test, we test the changes in market risk over the event window.

In evaluating the time variation in market risk, our concern is not for changes in risk between

period 2 and 3, since any systematic changes in market risk over that period are most likely due

to the effect of the foreign listing. The more important question is whether our sample

experiences systematic changes in market risk over either the pre-event period or the post-event

period. If this is the case, our long skip period may result in a comparison of returns for assets

that are in fact fundamentally very different. To test this, we perform the following regression:

( ) ( ) ( )( )( ) ( ) ( )( ) itLt

PiLt

POSTi

POSTi

LtPiLt

PREi

PREiit

ettttDrbttDrbttDa

ttttDrbttDrbttDar

+++++++++

+−−−−+−−+−−=

120,60120,120,120,

1,601,1201,1201,1204

2

. (5)

Due to the test specification, the coefficients in this regression are the same as if the sample

was divided into pre-listing and post-listing sub-samples. In this regression 2Pib measures the

change in the home market risk between periods 1 and 2 and 4Pib measures the change in the

home market risk between periods 3 and 4. If these coefficients are on average different from

zero, we must reject the conjecture that the market risk remains unchanged over the pre-event

or post-event periods. The mean coefficients from these regressions and their associated t-

statistics are reported in Panel D of Table 4. In both cases the test coefficient is insignificantly

different from zero. The estimate for 2Pib is -0.0211 with a t-statistic of -1.15 and the estimate

for 4Pib is -0.0055 with a t-statistic of -0.23. We conclude that the 120-month skip period

13

specification does not suffer from problems due to systematic changes in market risk. The test

provides some interesting additional information about return behavior of firms in our sample.

The intercept is on average extremely high for the pre-event period (1.14% per month or 13.7%

annually) consistent with a pre-event run-up. Although the intercept is much smaller in

magnitude for the post-event period, the intercept is also positive on average. We also observe

that the point estimate of the home market beta increases slightly from 0.95 in the pre-listing

period to 1.00 in the post-listing period.

3.2. Performance across markets and listing sequence

In this section we examine the differences in return patterns around the listing by home and

host markets, as well as by listing order sequence. We begin by highlighting differences in

return patterns for three subsets of listings: the listings of US firms abroad, the listings of non-

US firms in the U.S., and the listings of non-US firms on markets outside the U.S. Figure 2

depicts the annualized residual returns for three subsets of listings and the corresponding

observation frequencies (shown in the lower part of the figure). The number of firm-listing-

month observations is always greater than 50. Since the CRSP return data is available for a

much longer time series, the number of observations for the U.S. firm sub-sample is less

affected by the extended event window. The return plot shows that the general pattern in

returns around an overseas listing that we described earlier is also present at the disaggregate

level, although relative magnitudes vary across markets. Foreign firms that list in the U.S.

experience the largest pre-listing increase in residual return reaching about 25% per year. In

comparison, the maximum level of annualized residual returns for foreign listings outside the

U.S. is less than 10% per year. The post-listing decline in returns is again markedly steeper for

listings in the U.S. than for the other two groups of listings.

Second, we examine the impact of the listing sequence. Studies that only look at the

cost of capital effects of listings on one market, such as the U.S. market, consider all such

listings the same. If there are diminishing gains to foreign listing, then ceteris paribus one

would expect the most profound response to be associated with a firm’s first overseas listing.

Since the motives and gains for subsequent foreign listings may be very different from initial

listings we expect that it is important to control for a firm’s listing sequence. We emphasize

again that our sample is not immune from the survivorship bias and so the sequencing

documented in our study presents the effects of a relative sequence of overseas listings not an

14

absolute sequence. Figure 3 shows the annualized residual returns of cross-listed firms sorted

by listing sequencing, namely, for the first, second, third, and the fourth-and-greater listing

group. The latter group incorporates all listing of order four and above and the corresponding

observation frequencies (see the lower part of the figure). The most interesting feature of the

return plot is the profound difference in the valuation effects between the first listing and the

subsequent listings. The first listing is associated with the largest run-up in residual returns

reaching over 12% annually prior to the listing. The post-listing decline is equally profound:

the residual returns approach -9% per year in the two-three years following the listing. The

second and the third listings are associated with a much smaller run-up in the stock prices prior

to the listing and decrease after the listing date. The returns associated with higher-order firm

listings (Fourth+) show hardly any pattern at all.

To test the statistical significance of the return patterns we follow the analysis of Table

4 but add additional market variables as provided in Equation (4) using host market model

(Model 1) and the 120-month skip period. Table 5 presents these results. In Panel A, we show

the results of two regressions. The first regression provides the information on the relative

performance around the listing of US firms listed overseas and non-US firms listed in the U.S.

Consistent with Figure 2, we find that the pre-listing run-up is particularly acute for firms

listing in the U.S. with monthly residual period 2 returns of 0.62% greater than those listed on

other markets. Foreign listings of US firms are associated with particularly poor period 2

monthly returns of 0.37% lower than in other markets. We find no statistical cross-sectional

differences in period 3 returns. Regression 1 also provides evidence on the cross-sectional

differences in cost of capital gains. We find that the statistically significant reduction in the

period 4 returns is unique to the US-hosted firms with coefficient of -0.50% per month or -

6.0% per year. The overall period 4 coefficient is an insignificant -0.05%.

To examine whether this finding is specific to the U.S. or is more generally a large-

market effect, we perform a similar regression (Regression 2) but substitute a G5 country

(France, Germany, Japan, U.K., and the U.S.) dummy, for the U.S. dummy in Regression 1.

The estimation results provide a remarkably different picture. The overall coefficients are now

similar to those for the full sample of data as in Table 4. As a group, the G5 countries do not

appear to provide significant cost of capital gains to foreign firms listed on their exchanges.

The interactive G5-period-4 coefficient of -0.01% is insignificantly different from zero. The

U.S. effect does not appear at first glance to be a large-market effect.

15

The U.S. has attracted the largest number of new overseas listings in the 1990s,

including a substantial number of first-time listings. To investigate the potential impact of

listing sequencing on valuation differences across markets, we include a dummy variable that

captures the first-listing effect. These results are presented in Panel B of Table 5. The first

regression shows that the period 2 run-up is particularly important and possibly unique to first

listings. The first listing period 2 coefficient is 0.48% higher than that for the subsequent

listings. We also find that the first listing is associated with substantially greater post-listing

abnormal performance. The first listing period 3 coefficient is 0.32% lower than that for the

subsequent listings. The first listing does not appear to be associated with statistically

significant differences in cost of capital effects. The first listing period 4 coefficient is -0.17

with a t-statistic of -1.07. In the second regression of Panel B, we observe that the U.S. host

market survives a first listing control variable. The period 4 return reduction is -0.47

percentage points per month over non-US-hosted listings. We also note that even after

controlling for the first listing, the period 2 increase in residual returns is still highly significant

for the US-hosted listings. Finally, in the third regression of the panel, we again observe that

even after controlling for the first listing, the G5 countries as a group do not provide any

specific cost of capital advantages to foreign firms cross-listed on their exchanges nor do

domestic G5 firms experience any unique benefits from being listed abroad.7

We have found that the U.S. market is unique among the largest economies in its impact

on the reduction in the cost of capital of cross-listed firms. In Table 6 we test whether the U.S.

effect is common among other large foreign listing markets. The table shows the results of a

single regression in which we incorporate country specific dummy variables for the nine largest

host markets for listings (markets hosting more than 30 listings). In this regression we also

control for the first listing. In addition to the G5 group of countries, the large foreign listing

hosts are all European countries, namely, Belgium, Luxembourg, the Netherlands, and

Switzerland. Possibly due to the large number of coefficients being estimated in this

7 Note that the first listing sample in period 4 contains a larger proportion of US firms than in periods 2 and 3. This is because many first listings from outside the U.S. have occurred during the 1990’s, resulting in fewer observations over the five-to-ten year period after the listing (see lower plots of Figures 2 and 3). Therefore, we have also estimated the cost of capital advantages of the first listing separately for US and non-US firms. For US firms the cost of capital gains associated with the first listing are insignificant and practically indistinguishable from higher order listings. For non-US firms, we observe a substantial economic significance of the first listing: the reduction in the cost of capital is 0.57% per month (6.8% a year). However, due to the smaller sample size, we are unable to support this finding in statistical terms. These results are available on request.

16

regression, most of the slope estimates in the table are insignificant. We do find that besides

the U.S., firms listing in Belgium, which hosts 59 listings, also appear to generate significant

reductions cost of capital. Although not significant, the large period 4 point estimates

associated with foreign listings in France, Switzerland, and the U.K., -46%, -0.47%, and -

0.39% per month, respectively, are economically sizeable. Although again insignificant, the

period 4 point estimate for Luxembourg would indicate a particularly poor cost of capital gain

for firms that list in that country. We do not find an obvious pattern in the countries that is

consistent with the estimated cost of capital effects reported in Table 6. We investigate more

closely the effect of market characteristics on cost of capital gains in the next section.

Finally, we recall from Panel E of Table 2 that many firms place subsequent listings

within a relatively short time period (about three years) of earlier listings. This implies, to a

certain degree, an overlap in returns with the earlier listings. Our concern is that the t-statistics

presented are upwardly biased because we are assuming observation independence when in fact

the same firm returns may be highly overlapping. To decouple the listing sequencing effect

from other effects, we exclude all sample listings that occur within five years of an earlier

listing. This procedure reduces the overall sample by about 50%. In unreported results using

this non-overlapping sample we find that the results are similar.

4. Return Performance and Relative Market Characteristics

We observe large cross-sectional variation in the cost of capital gains across various home and

host markets. In this section we investigate the inherent strengths and weaknesses of world

markets. Our goal is to run a horse race among conventional motives for issuing listings

overseas to determine which hypotheses best explain the cost of capital gains to overseas

listing.

4.1. Motives for overseas listings

In a frictionless world market, firms are indifferent to the choice of security trading venue.

However, cross-border barriers to investment capital and information flow, as well as market

differences in liquidity, disclosure requirements, and shareholder expropriation risk make the

listing venue decision relevant. We discuss each of these hypotheses in turn.

17

Cross-border barriers to investment capital. Based on the international asset pricing models of

Black (1974), Solnik (1974), Stulz (1981), a large number of papers argue that international

investing provides large diversification gains. If investors are restricted for investing overseas,

firms gain from the premium that investors are willing to pay for restricted foreign securities,

Errunza and Losq (1985). With cross-border trading friction, overseas investors are most

interested in capital from markets that will minimize their portfolio risk. Consequently, firms

may capture the premium that foreign investors are willing to pay for equity from markets with

little correlation to their home market.8 Other things equal, firms should realize the greatest

cost of capital gain by selecting those host markets whose returns are least correlated with the

home market.

Cross-border barriers to information flow. Merton (1987) argues that incomplete information

generates a firm-specific risk premium in the cost of capital for lesser-known firms. Firms can

reduce this cost by increasing investor awareness and thus expanding the investor base. A

number of studies provide empirical support for Merton's proposition. Baker et al. (2002),

Lang et al. (2003), Ahearne et al. (2001) find that foreign firm listings in the U.S. are associated

with greater foreign market media and analyst exposure. Foerester and Karolyi (1999) find that

returns behavior of their ADR sample of firms within one year of listing is consistent with

Merton’s proposition. For our sample one prediction from this literature is that a firm’s

reduction in the cost of capital is positively correlated with the relative capitalization of the

home and host markets.

Market differences in liquidity. Overseas listings may be placed abroad in a search for greater

liquidity. Many papers find that foreign firms listed in the U.S. achieve significantly lower bid-

ask spread (e.g., Tinic and West, 1974; Domowitz et al., 1998; Foerster and Karolyi, 1998;

Werner and Kleidon, 1996). Chowdhry and Nanda (1991) document that liquidity plays an

important role in the ability of markets to attract trading volume. If increased liquidity is

important in the listing decision, we expect that those foreign markets that provide the greatest

improvement in liquidity will generate the largest gains in cost of capital.

8 See Sarkissian and Schill (2003) for a more formal development of this assertion.

18

Market differences in disclosure. Biddle and Suadagaran (1992) find that more difficult

disclosure standards in a given country hinder the access of foreign firms to that market. Fuerst

(1998) and Huddart et al. (1999), on the other hand, argue that firm managers prefer listing

stock in the markets with higher accounting standards. The existence of these two opposing

views motivate the examination of the relation between cost of capital effects of overseas

listings and the relative disclosure standards in the home and host markets.

Market differences in shareholder expropriation risk. Firms may seek to list abroad in order to

reduce investor expropriation risk. La Porta et al. (1997, 1998) find that countries with poor

investor protection system have low financing activity. Reese and Weisbach (2001), Doidge et

al., 2003, Lins et al. (2001), and Doidge (2003) show that improving shareholder protection

may explain why some non-US firms cross-list in the United States. In our framework, we

expect that firms list in markets with lower shareholder expropriation risk should experience

the greater cost of capital gains.

4.2. Return performance and home-host market characteristics

Table 7 provides aggregate economic, equity market, exchange listing standards, and investor

protection measure for the 35 countries in our sample.9 Our measure of domestic market size is

the total equity market capitalization (CAP). Our liquidity measure is the same as in

Domowitz, Glen and Madhavan (2002) – the Elkins/McSherry estimates of average one-way

trading cost for pension funds, investment managers and brokerage houses. The next two

columns report the accounting standards present in each country, based either on the

International Accounting Standards (IAS) or on US GAAP standards. The last column gives

the investor protection measure for each market from La Porta et al. (1997). This measure

accounts for the character of legal rules as well as the quality of law enforcement in the

respective countries. We obtain proxy measures of the relative characteristics of the sample

markets for each pair of home and host countries in our sample. We also proxy the

diversification potential between each pair of countries by the correlation of their respective

equity market returns.

To summarize our proxy variables, we examine the variation in relative market

characteristics along the average listing sequence. We regress each of the relative home-host

19

market characteristics on three dummy variables associated with the first, second, and third

listing. In this regression each listing is treated as an independent observation. Our results are

reported in Table 8. The intercept of the individual regressions shows the average importance

of these characteristics in explaining market choice. We find that on average firms list in

smaller markets with greater shareholder protection and disclosure requirements. The

coefficient on the first-listing dummy shows that first listings are unique. The results suggest

that first listings tend to occur in markets that are larger, more liquid, and require greater

disclosure, but maintain weaker shareholder protection status. There is no indication that first

listings are placed in the markets that are the least correlated with the home market.

We know turn to our cross-sectional test of cost of capital gains. In this test we run the

following regression of residual returns on the relative characteristics of the home and host

markets,

( ) ( ) ( )( ) ( ) ( ) ( ){ }( ) ( ) ( ) ( ){ }( )

1 1 1 3

2 2 2 2

3 3 3 3

60, 1 , 59 60, 120

60, 1 , 59 60, 120

60, 1 , 59 60, 120

SIZE

e D t t D t t D t t

D First Listing D t t D t t D t t

D CORRELATION D t t D t t D t t

D MARKET

α φ θ δ

α φ θ δ

α φ θ δ

= + − − + + + + +

+ + − − + + + + +

+ + − − + + + + +

+ ( ) ( ) ( ){ }( ) ( ) ( ) ( ){ }( ) ( ) ( ) ( ){ }( ) ( ) ( )

4 4 4 4

5 5 5 5

6 6 6 6

7 7 7 7

60, 1 , 59 60, 120

60, 1 , 59 60, 120

60, 1 , 59 60, 120

60, 1 , 59 6

D t t D t t D t t

D LIQUIDITY D t t D t t D t t

D DISCLOSURE D t t D t t D t t

D LAW D t t D t t D t

α φ θ δ

α φ θ δ

α φ θ δ

α φ θ δ

+ − − + + + + +

+ + − − + + + + +

+ + − − + + + + +

+ + − − + + + +( ){ }( ) ( )1 910

0, 120

1926 ... 2001

t

D Feb D Dec uη τ η τ

+

+ = + + = +

. (6)

Panel A of Table 9 presents the results. Again the focus of attention is on the period 4

coefficients in the last column, which correspond to the extended-term cost of capital gains.

We observe that all of the interactive period 4 coefficients have the hypothesized sign expect

for that of CORRELATION. However, only the coefficient on the investor protection variable

(LAW) is statistically significant. The LAW coefficient estimate suggests that those firms that

list in markets with greater shareholder protection achieve cost of capital gains of -0.39

percentage points larger than those that list in other markets. The coefficient of -0.39 has a

9 This table is a reduced version of a similar table in Sarkissian and Schill (1999),

20

very significant t-statistic of -3.07, and it is equivalent to a 4.7% decrease in the annualized cost

of capital.

One concern with the Panel A results is that the market characteristics may vary over

time. If, for example, disclosure requirements have changed substantially over the past 20

years, foreign listings placed in the 1970s may have been motivated by a very different set of

relative characteristics. Although we would like to improve the timeliness of our right-hand

side variables to match the market characteristics at the event time, data availability prohibits

such variable collection. However, we can test the importance of this concern by trimming our

sample to only those listings that occur within a reasonable time of the market characteristic

estimation. In Panel B we report the results for a re-estimation of Equation (6) based on

listings that occur in the 1990s. As before, we observe that the enhanced investor protection of

the host market plays the key role in the reduction of the cost of capital of cross-listed firm.

While the statistical significance of the LAW variable is smaller due to the reduced sample, its

economic significance is twice as large as that for the entire sample. The correlation variable

again has an opposite to the intuition sign but now is significant. This result implies that firms

that list in markets that are weakly correlated with the home market, which presumably can

provide higher diversification gains, experience increase rather than decrease in their cost of

capital.

The results suggest that the investor protection argument provides the best explanation

of foreign listing returns. Moreover, we reject the barriers to investment explanation since

investors do not appear to be willing to pay a premium for capital from markets with weak

correlation to their own. These results are consistent with those of Sarkissian and Schill (2003)

who argue that firms are more successful in listing in familiar and therefore empirically more

correlated markets.

5. Return Performance and Firm Characteristics

Kang and Stulz (1997) and Sarkissian and Schill (2003) argue that overseas listings reflect

rather than reduce the information barriers that lead to investor home bias in international

investment. Kang and Stulz find that foreign ownership increases with firm size and exports.

Sarkissian and Schill suggest that the wider international familiarity associated with large firms

and those that produce tradable output allow these firms greater flexibility in their choice of

21

foreign listing market. If such firm characteristics affect foreign ownership and foreign market

choice, it is likely that such characteristics also affect the associated cost of capital gain. We

test the cross-sectional affects of firm size and output tradability on our return measures to see

if larger firms and those producing tradable output generate larger cost of capital reductions.

We classify firms into tradable and non-tradable by their industry following Sarkissian

and Schill (2003). There are eight tradable industries, chemicals, consumer goods (including

food), electronics (including electrical equipment and computers), manufacturing, healthcare

(i.e., pharmaceuticals, drugs, biotechnology), mining, oil and gas, and paper (including forestry

and printing); and eight non-tradable industries, construction, financials (including banks and

insurance), leisure (i.e., hotels and leisure equipment), retail (including distributors), support

services, telecommunications (including media), transportation, and utilities. In our sample

65% of the listings are from firms we classify as tradable. We construct a dummy variable that

takes the value of one for tradable firms.

We classify firms based on their size as follows. First we obtain the U.S. dollar-

denominated market capitalization from CRSP or Datastream at the end of the calendar month

preceding the listing date. We adjust each size value to the end of 1998 using the U.S. market

index. We classify firms as large if their capitalization exceeds the sample median value for all

other firms within their respective G5 category. Since market size and firm size are correlated,

we use separate market-size breakpoints so that the size groupings include an equal number of

G5 and non-G5 firms. We construct a dummy variable that takes the value of one for large

firms.

We repeat the Table 4 experiment by adding the two firm characteristics to our

regression models and present the test results in Table 10. We present the test results on output

tradability in Regression 1. We find that tradable firms underperform the average firm in

period 3 by -0.56% per month (or 6.7% in annual terms). The period 4 interactive coefficient is

negative but statistically insignificant. In the second regression model, we perform a similar

exercise with firm size. The estimation shows that large firms exhibit significantly smaller

abnormal return behavior around the listing. The period 2 residual returns of large firms

relative to those of an average firm are lower by 0.53% per month. Moreover, the period 3

returns for the large firms are significantly larger than those of an average firm, by 0.49% per

month. Such return patterns of large firms are consistent with the notion that they are more

fairly priced in the domestic market than smaller firms and therefore exhibit less market timing

22

effects around their listing on foreign exchanges. The period 4 cost of capital reduction among

large firms is insignificantly different from that of small firms. The fact that large firms do not

show additional cost of capital advantages from overseas listings in the long run may be

explained by relatively diversified nature of these firms that pushes their cost of capital, even

before listing on a foreign exchange, to the level more common among integrated with the

world market firms.

In the third regression of Table 10 we examine the importance of the two firm-specific

characteristics while controlling for the major host and home markets for listings and listing

sequencing. The overall results are qualitatively very similar to those based on the first two

regressions. The most important quantitative difference is that tradable firms now show not

only economically but also statistically significant period 4 cost of capital gains at the 10%

level relative to those for an average firm. Firms classified as producing tradable output

experience a 0.30% per month (3.6% annually) greater reduction in the cost of capital than the

non-tradable control group. This result is consistent with the investor familiarity arguments of

Kang and Stulz (1997) and Sarkissian and Schill (2003).

6. Conclusions

Many papers have attempted to measure the valuation effects of overseas listings. The main

conclusion of these studies is a major reduction in the cost of capital among cross-listed firms.

In spite of many similarities, there has been no convincing indication that the valuation of firms

issuing foreign listings could resemble the performance of domestic firms moving their stock to

a new exchange or issuing new equity. It is well known that these firms show substantial

market-timing effects: they usually have large pre-listing or pre-issuing returns, but, after the

event, they underperform the control group of companies over a substantial period of time. The

inability to detect market timing in overseas listing decisions can be explained by the fact that

most of related studies in this area consider limited time period around the listing date and very

restricted set of host and home markets for listings. In addition, they do not account for the fact

that many global listings represent not the first but rather the subsequent listings of firms in

foreign markets.

23

This paper provides evidence of the presence of both market timing and cost of capital

effects in the performance of firms around their overseas listing placements. We use a

comprehensive sample of global listings across all geographic regions and analyze the

valuation effects over 241-month window around the listing while controlling for the listing

sequencing. We find that overseas listings lead to a substantial pre-listing run-up and a

profound post-listing decline in firm returns that take place on average five years before and

after the listing, respectively. We also detect significant cost of capital reduction five to ten

years following the listing relative to the equivalent pre-listing period. All the valuation

patterns vary widely with listing sequencing, different home and host markets for securities,

and firm characteristics. The first listing is associated with very large abnormal returns around

the listing, while the U.S., Belgium and several other markets seem to provide firms listed on

their exchanges with large cost of capital gains. We observe that firms producing tradable

output experience substantial cost of capital reduction, which implies that firms that are

relatively more familiar to foreign investors before the listing benefit more after the listing.

Large firms, on the other side, show neither significant abnormal returns around the listing nor

significant cost of capital reduction in the long-run, which implies that they are priced more

fairly in the domestic market before listing than small, less known and less diversified firms.

Contrary to the conventional notion of international diversification in segmented markets, firms

that list across markets with low return correlation experience relatively little to negative gains

in their cost of capital. We find that the foreign listings in markets with better shareholder

protection generate the most substantial improvements in the cost of capital.

References:

Alexander, G., C. Eun, and S. Janakiramanan, 1988, International listings and stock returns: Some empirical evidence, Journal of Financial and Quantitative Analysis 23, 135-151.

Asquith, P. and D.W. Mullins, 1986, Equity issues and offering dilution, Journal of Financial Economics 15, 61-89.

Baker, H.K., J.R. Nofsinger, and D.G. Weaver, 2002, International cross-listing and visibility, Journal of Financial and Quantitative Analysis 37, 495-521.

Black, F., 1974, International capital market equilibrium with investment barriers, Journal of Financial Economics 1, 337-352.

24

Chaplinsky, S. and L. Ramchand, 2000, The impact of global equity offerings, Journal of Finance 55, 2767-2789.

Chowdhry, B. and V. Nanda, 1991, Multi-market trading and market liquidity, Review of Financial Studies 4, 483-511.

Doidge, C., G.A. Karolyi, and R. Stulz, 2003, Why are foreign firms listed in the U.S. worth more?, Journal of Financial Economics, forthcoming.

Doidge, C., 2003, U.S. cross-listings and the private benefits of control: Evidence from dual class firms, Journal of Financial Economics, forthcoming.

Domowitz, I., J. Glen, and A. Madhavan, 1998, International cross-listing and order flow migration: Evidence from an emerging market, Journal of Finance 53, 2001-2027.

Domowitz, I., J. Glen, and A. Madhavan, 2002, Liquidity, volatility, and equity trading costs across countries and over time, forthcoming in Journal of International Finance.

Errunza, V. and E. Losq, 1985, International asset pricing under mild segmentation: Theory and tests, Journal of Finance 40, 105-124.

Errunza, V. and D. Miller, 2000, Market segmentation and the cost of capital in international equity markets, Journal of Financial and Quantitative Analysis, 35, 577-600.

Foerster, S. and G.A. Karolyi, 1993, International Listings and Stock Price Reactions: The Case of Canada and the U.S., Journal of International Business Studies 24, 763-784.

Foerster, S. and G.A. Karolyi, 1999, The effects of market segmentation and investor recognition on asset prices: Evidence from foreign stocks listing in the United States, Journal of Finance 54, 981-1013.

Fuerst, O., 1998, A theoretical analysis of the investor protection regulations: Argument for global listing of stocks, Working paper, Yale University.

Howe, J. and K. Kelm, 1987, The stock price impact of overseas listings, Financial Management, Autumn, 51-56.

Howe, J. and J. Madura, 1990, The impact of international listings on risk: Implications for capital market integration, Journal of Banking and Finance 14, 1133-1142.

Huddart, S., J. Hughes, and M. Brunnermeier, 1999, Disclosure requirements and stock exchange listing choice in an international context, Journal of Accounting & Economics 26, 237-269.

Jayaraman, N., K. Shastri, and K. Tandon, 1993, The impact of international cross-listings on risk and return: Evidence from American Depository Receipts, Journal of Banking and Finance 17, 91-103.

Kadlec G.B. and J.J. McConnell, 1994, The effect of market segmentation and illiquidity on Asset Prices: Evidence from exchange listings, Journal of Finance 49, 611-636.

Karolyi, G.A., 1998, What happens to stocks that list shares abroad? A survey of evidence and its managerial implications (New York University Salomon Bros. Center Monograph, Vol. 7, no. 1, New York, N.Y.).

La Porta, R., F. Lopez-De-Silanes, A. Shleifer, and R.W. Vishny, 1997, Legal determinants of external finance, Journal of Finance 52, 1131-1150.

25

La Porta, R., F. Lopez-De-Silanes, A. Shleifer, and R.W. Vishny, 1998, Law and finance, Journal of Political Economy 106, 1113-1155.

Lang M., K. Lins, and D. Miller, 2002, ADRs, analysts, and accuracy: Does cross listing in the U.S. improve a firm's information environment and increase market value?, Journal of Accounting Research, forthcoming.

Lins, K., D. Strickland, and M. Zenner, 2000, Do non-U.S. firms issue stock on U.S. equity markets to relax capital constraints?, Working paper, University of North Carolina.

Lau, S., D. Diltz, and V. Apilado, 1994, Valuation effects of international stock exchange listings, Journal of Banking and Finance, 18, 743-755.

Loughran T. and J. Ritter, 1995, The new issues puzzle, Journal of Finance 50, 23-51.

Loughran, T., J. Ritter, and K. Rydqvist, 1994, Initial public offerings: International insights, Pacific-Basin Finance Journal 2,165-199.

Marsh, P., 1982, The choice between equity and debt: An empirical study, Journal of Finance 37, 121-44.

Merton, R., 1987, Presidential address: A simple model of capital market equilibrium with incomplete information, Journal of Finance 42, 483-510.

Miller, D., 1999, The market reaction to international cross-listings: Evidence from depositary receipts, Journal of Financial Economics 51, 103-123.

Pagano, M., A.A. Roell, and J. Zechner, 2002, The geography of equity listing: Why do European companies list abroad?, forthcoming in Journal of Finance.

Ritter, J., 1991, The long-run performance of Initial Public Offerings, Journal of Finance 46, 3-27.

Sanger, G.C. and J.J. McConnell, 1986, Stock exchange listings, firm value, and security market efficiency: The impact of NASDAQ, Journal of Financial and Quantitative Analysis 21, 1-25.

Sarkissian S. and M.J. Schill, 2003, The Overseas Listing Decision: New Evidence of Proximity Preference, Review of Financial Studies, Forthcoming.

Solnik, B., 1974, An equilibrium model of the international capital market, Journal of Economic Theory 18, 365-378.

Stulz, R., 1981, On the effects of barriers to international asset pricing, Journal of Finance 25, 783-794.

Stulz, R., 1999, Globalization of equity markets and the cost of capital, Working paper, Ohio State University.

Tinic, S. and R. West, 1974, Marketability of common stocks in Canada and the USA: A comparison of agent versus dealer dominated markets, Journal of Finance 29, 729-746.

Werner, I. and A. Kleidon, 1996, U.S. and U.K. trading of British cross-listed stocks: An intraday analysis of market integration, Review of Financial Studies 9, 619-664.

Tab

le 1

R

epre

sent

ativ

e st

udie

s on

valu

atio

n ef

fect

s of o

vers

eas l

istin

g, n

ew e

xcha

nge

listin

g, a

nd n

ew e

quity

issu

ing

Th

e ta

ble

sum

mar

izes

the

resu

lts o

f rep

rese

ntat

ive

stud

ies

on fi

rm p

erfo

rman

ce a

roun

d ne

w li

stin

g an

d eq

uity

issu

ance

dat

es in

chr

onol

ogic

al o

rder

. Pa

nel A

sho

ws

stud

ies

on v

alua

tion

effe

cts

of o

vers

eas

listin

g, P

anel

B –

on

valu

atio

n ef

fect

s of

new

exc

hang

e lis

ting

and

equi

ty is

suan

ce.

The

first

col

umn

indi

cate

s th

e pu

blis

hed

stud

y. T

he n

ext t

wo

colu

mns

in P

anel

A c

orre

spon

d to

the

fore

ign

and

hom

e m

arke

ts fo

r lis

tings

, res

pect

ivel

y. T

he sa

me

colu

mns

in P

anel

B c

orre

spon

d to

the

new

an

d ol

d m

arke

t/exc

hang

e or

equ

ity ty

pe, r

espe

ctiv

ely.

Acr

oss

both

pan

els,

the

four

th, f

ifth,

and

six

th c

olum

ns in

dica

te th

e tim

e pe

riod

cons

ider

ed in

a g

iven

stu

dy,

num

ber o

f lis

tings

or e

quity

issu

ance

obs

erva

tions

, and

the

even

t win

dow

, res

pect

ivel

y. T

he se

vent

h co

lum

n, c

alle

d th

e “S

KIP

Per

iod”

show

s the

tim

e pe

riod

arou

nd

the

listin

g or

equ

ity is

suan

ce th

at is

exc

lude

d fr

om th

e pr

e- a

nd p

ost-l

istin

g pe

rfor

man

ce c

ompa

rison

. Th

e la

st tw

o co

lum

ns s

how

the

mar

ket-a

djus

ted

perf

orm

ance

of

firm

s be

fore

and

afte

r th

e lis

ting

or e

quity

iss

uanc

e da

te, r

espe

ctiv

ely.

Th

e pe

rfor

man

ce i

s gi

ven

in t

erm

s of

cum

ulat

ive

abno

rmal

ret

urns

(C

AR

) an

d it

is

annu

aliz

ed fo

r all

the

data

in th

e or

igin

al st

udie

s tha

t are

shor

ter o

r lon

ger t

han

one

year

.

Pane

l A: S

tudi

es o

n va

luat

ion

effe

cts o

f ove

rsea

s lis

ting

Mar

ket

Pe

rfor

man

ce (C

AR

, %)

Ref

eren

ce

Fo

reig

n

Hom

e

Tim

e Pe

riod

Num

ber o

f Ev

ents

Ev

ent

Win

dow

Sk

ip

Per

iod

Pr

e-lis

ting

Po

st-li

stin

g

Ale

xand

er e

t al.,

88

USA

G

loba

l 19

69-1

982

34

+/-3

6 m

onth

s N

one

10.6

-1

7.5

Foer

ster

-Kar

olyi

, 93

USA

C

anad

a 19

76-1

992

56

+/-1

2 m

onth

s O

ne w

eek

25.4

-2

6.3

Jaya

ram

an e

t al.,

93

USA

G

loba

l 19

83-1

988

95

+/-1

mon

th

Non

e 15

.1

-11.

1

Lau

et a

l., 9

4 G

loba

l U

S 19

62-1

990

346

+6 m

onth

s /

-1 w

eek

Non

e 0

-7.9

Foer

ster

-Kar

olyi

, 99*

U

SA

Glo

bal

1976

-199

2 15

3 +/

-12

mon

ths

One

wee

k 17

.0

-12.

0

Mill

er, 9

9 U

SA

Glo

bal

1985

-199

5 18

3 +/

-1 m

onth

N

one

0 -3

4.7

Erru

nza-

Mill

er, 0

0 U

SA

Glo

bal

1985

-199

4 12

6 +/

-36

mon

ths

12 m

onth

s 10

.3

-1.1

Foer

ster

-Kar

olyi

, 00

USA

G

loba

l 19

82-1

996

333

+36

mon

ths

-12

mon

ths

One

mon

th

18.0

-1

1.7

* Th

e pe

rfor

man

ce d

ata

is o

btai

ned

from

a fi

gure

.

28

Tab

le 1

(con

tinue

d)

Pa

nel B

: Stu

dies

on

valu

atio

n ef

fect

s of n

ew e

xcha

nge

listin

g an

d ne

w e

quity

issu

ing

Mar

ket/E

quity

Perf

orm

ance

(CA

R, %

) R

efer

ence

New

Old

Tim

e Pe

riod

Num

ber o

f Ev

ents

Ev

ent

Win

dow

Sk

ip

Per

iod

Pr

e-lis

ting

Po

st-li

stin

g

New

exc

hang

e lis

ting

Sang

er-M

cCon

nell,

86

NY

SE

OTC

1 19

66-1

970

153

+/-1

2 m

onth

s N

one

23.2

-2

.7

N

YSE

O

TC2

1971

-197

7 16

6 +/

-12

mon

ths

Non

e 19

.8

-3.3

McC

onne

ll-Sa

nger

, 87

NY

SE

OR

A

1926

-198

2 24

82

+12

mon

ths

Non

e N

/A

-3.6

Dha

ran-

Iken

berr

y, 9

5 N

YSE

, ASE

A

SE, N

ASD

19

62-1

990

2889

+3

6 m

onth

s N

one

N/A

-4

.0

New

equ

ity is

suin

g

Asq

uith

-Mul

lins,

86*

SEO

U

S: P

ublic

19

63-1

981

531

+22

mon

ths /

-2

6 m

onth

s N

one

14.2

-4

.2

Ritt

er, 9

1 IP

O

US:

Priv

ate

1975

-198

4 15

26

+36

mon

ths

Non

e N

/A

-8.9

Loug

hran

-Ritt

er, 9

5 IP

O

US:

Priv

ate

1970

-199

0 47

53

+60

mon

ths

Non

e N

/A

-6.7

SE

O

US:

Pub

lic

1970

-199

0 37

02

+60

mon

ths

Non

e N

/A

-8.3

* A

squi

th a

nd M

ullin

s (1

986)

com

pute

sha

re p

rice

reac

tion

rela

tive

to th

e an

noun

cem

ent d

ate

of th

e eq

uity

issu

ance

. It

is a

ssum

ed th

at o

n av

erag

e th

e lis

ting

date

fo

llow

s the

ann

ounc

emen

t dat

e in

two

mon

ths.

OTC

1 –

pre-

Nas

daq

over

-the-

coun

ter m

arke

t O

TC2 –

post

-Nas

daq

over

-the-

coun

ter m

arke

t O

RA

– O

TC, r

egio

nal e

xcha

nges

, and

ASE

N

ASD

– N

asda

q

Table 2 Distributional characteristics of overseas listing

The table provides the distributional characteristics of overseas listings as of 1998 that have return data available between 120 months before and 120 months after the listing. The total sample is comprised of 816 firms with 1298 overseas listings. Panel A provides the country-to-country frequency distribution of multiple cross listings as of 1998. The other panels present distribution of overseas listings based on the listing number. Panels B-D show frequency of listings across different calendar periods and home and host countries. Firms with six or more overseas listings are combined into one group (6+). Panel E shows the median duration between subsequent listings.

Panel A: Country-to-country distribution of listings Host country Home country A

ustra

lia

Aus

tria

Bel

gium

C

anad

a D

enm

ark

Fran

ce

Ger

man

y H

. Kon

g Ir

elan

d Ita

ly

Japa

n Lu

xem

. M

alay

sia

Net

her.

N Z

eala

nd

Nor

way

Pe

ru

Sing

apor

e S.

Afr

ica

Spai

n Sw

eden

Sw

itz.

UK

U

SA

Tota

l

Australia 3 1 3 1 27 2 7 18 62Austria 1 4 5Belgium 1 1 2 5 1 10Brazil 3 14 17Canada 2 3 3 2 1 1 6 4 126 148Chile 8 8Czech Rep. 4 4Denmark 1 2 1 4Finland 1 2 3 2 2 10France 4 1 6 1 3 2 3 6 18 44Germany 11 4 7 2 9 5 6 1 2 1 18 8 9 83Hong Kong 1 1 6 1 2 11India 42 13 55Ireland 17 5 22Italy 3 4 1 6 14Japan 1 4 1 21 41 15 15 3 9 18 20 148Luxem. 1 1 2Mexico 4 4Nether. 3 7 3 14 1 3 1 1 10 8 14 65N. Zealand 13 3 16Norway 1 1 1 2 1 5 5 16Peru 1 1Philippines 5 1 1 7Poland 1 4 5Portugal 1 1Singapore 1 2 3S. Africa 5 2 2 3 1 8 6 27Spain 1 1Sweden 1 1 3 1 2 3 4 7 22Switz. 1 1 1 6 2 1 1 4 17Taiwan 14 1 7 1 23Thailand 1 1Turkey 1 3 4UK 3 4 3 5 5 1 6 4 1 1 6 1 4 42 86USA 6 2 26 6 32 38 22 1 57 1 1 5 64 81 352

30

Table 2 (continued)

Panel B: Calendar period frequency distribution of foreign listings Listing number 1 2 3 4 5 6+ Total

Pre-1970 62 35 15 6 2 120 1970-1979 81 31 24 16 9 8 169 1980-1989 162 81 44 28 18 15 348 1990-1998 459 75 41 28 21 37 661 Total 764 222 124 78 50 60 1298

Panel C: Home country frequency distribution of foreign listings Number of listings 1 2 3 4 5 6+ Total

Australia 42 13 6 1 62 Austria 5 5 Belgium 4 4 1 1 10 Brazil 17 17 Canada 127 8 5 4 3 1 148 Chile 8 8 Czech Rep. 4 4 Denmark 2 1 1 4 Finland 5 2 1 1 1 10 France 23 9 4 3 2 3 44 Germany 27 15 7 7 8 19 83 Hong Kong 10 1 11 India 53 2 55 Ireland 18 4 22 Italy 9 3 2 14 Japan 73 31 19 13 7 5 148 Luxembourg 1 1 2 Mexico 4 4 Netherlands 21 14 10 5 5 10 65 New Zealand 15 1 16 Norway 7 4 1 1 1 2 16 Peru 1 1 Philippines 7 7 Poland 5 5 Portugal 1 1 Singapore 3 3 South Africa 18 3 4 2 27 Spain 1 1 Sweden 7 3 4 4 2 2 22 Switzerland 10 2 1 1 1 2 17 Taiwan 22 1 23 Thailand 1 1 Turkey 4 4 UK 57 17 6 4 1 1 86 USA 152 84 51 31 19 15 352

31

Table 2 (continued)

Panel D: Host country frequency distribution of foreign listings

Number of listings 1 2 3 4 5 6+ Total

Australia 20 4 1 25 Austria 5 5 1 3 2 2 18 Belgium 22 14 15 5 1 2 59 Canada 18 1 2 1 2 24 Denmark 1 1 France 29 20 14 14 1 3 81 Germany 44 27 20 12 13 15 131 Hong Kong 1 1 Ireland 5 1 6 Italy 2 2 Japan 9 16 2 3 7 7 44 Luxembourg 88 5 1 4 2 100 Malaysia 1 1 Netherlands 53 17 8 8 3 5 94 New Zealand 21 2 4 1 28 Norway 2 2 Peru 1 1 Singapore 14 4 2 1 1 22 South Africa 1 1 Spain 1 2 3 Sweden 7 2 1 1 1 2 14 Switzerland 38 38 21 9 5 6 117 UK 127 37 18 12 6 4 204 USA 260 27 17 5 6 4 319

Panel E: Lag length in months between consecutive listings

All listings Listings

from the U.S. Listings

in the U.S. Non-U.S.

listings

Obs Median Obs Median Obs Median Obs Median

1-to-2 178 33 76 103 19 12 83 29 2-to-3 114 27 51 48 14 21 49 20 3-to-4 71 24 30 42 4 111 37 15 4-to-5 44 71 19 96 5 105 20 25 5-to-6+ 51 29 15 35 4 0 39 41

32

Table 3 Distribution of return data across markets

The table shows the distribution of sample return data within the period of +/-120 months around listing months by respective home countries. Number of months refers to the total number of unique calendar months in the sample of returns. Number of firm-listing-months is the total number of return observations used in the study. Number of firm-months excludes the overlapping returns for firms that list in multiple markets. The first and last month are the first and last calendar month, respectively, in the series of unique months.

Number of months

Number of firm-listing-months

Number of firm- months

First month Last month

Australia 346 10286 6914 3/1973 12/2001 Austria 194 639 639 11/1985 12/2001 Belgium 291 1515 1058 3/1973 12/2001 Brazil 81 684 684 9/1994 12/2001 Canada 346 19063 17117 3/1973 12/2001 Chile 148 912 912 9/1989 12/2001 Czech Rep. 90 304 304 5/1994 12/2001 Denmark 291 647 304 3/1973 5/1997 Finland 164 1059 500 5/1988 12/2001 France 346 4859 3480 3/1973 12/2001 Germany 346 12285 5926 3/1973 12/2001 Hong Kong 264 1807 1628 7/1980 12/2001 India 143 7581 7439 2/1990 12/2001 Ireland 346 2970 2499 3/1973 12/2001 Italy 280 1880 1372 3/1973 12/2001 Japan 346 26713 16832 3/1973 12/2001 Luxembourg 96 163 163 1/1994 12/2001 Mexico 151 539 539 6/1989 12/2001 Netherlands 346 8685 4888 3/1973 12/2001 New Zealand 167 1757 1638 2/1988 12/2001 Norway 263 2049 1101 2/1980 12/2001 Peru 95 95 95 2/1994 12/2001 Philippines 157 775 775 12/1988 12/2001 Poland 43 155 155 6/1998 12/2001 Portugal 41 41 41 8/1998 12/2001 Singapore 175 333 333 6/1987 12/2001 South Africa 346 3736 3007 3/1973 12/2001 Spain 119 119 119 7/1989 5/1999 Sweden 238 2970 1226 3/1982 12/2001 Switzerland 346 2885 1872 3/1973 12/2001 Taiwan 163 2598 2598 6/1988 12/2001 Thailand 86 86 86 7/1994 12/2001 Turkey 141 337 337 4/1990 12/2001 UK 346 13742 10848 3/1973 12/2001 USA 912 71652 46904 1/1926 12/2001

33

Table 4 Residual returns for overseas listings across event time periods

This table shows the results of a times-series cross-section regression of monthly residual returns for firms listing in a foreign market. The dependent variable, residual return, is computed based on home market model (Model 1), a 2 factor home and world market model (Model 2), and a 2 factor home and host market model (Model 3) where all returns are in US dollars less the return on the 30-day Treasury Bill. The market index for the U.S. is the CRSP value-weighted market return; the market returns for other countries are the returns on the corresponding market indexes from Datastream. D(*) refers to a corresponding dummy variable, such that D(t-120,t-60) is equal to 1 for observations 120 to 60 months prior to the month of the overseas listing date and equal to 0 otherwise. For each regression an array of monthly calendar time dummy variables from February 1926 to December 2001 is also included but the coefficient estimates are not reported. T-statistics are reported in parentheses. Panel D reports the mean beta coefficients from a regression of event firm returns on the home market index return. Panel A. Return behavior (Skip period: t-60, t+59)

Intercept D(t-120,t-61) D(t-60,t-1) D(t,t+59) D(t+60,t+120)

(1) Model 1 -0.3641 (-0.72)

0.4207 a (5.11)

-0.3665 a (-4.46)

-0.2076 b (-2.41)

(2) Model 1 0.0566 (0.11)

-0.4207 a (-5.11)

-0.7872 a (-10.94)

-0.6283 a (-8.16)

(3) Model 2 -0.4540 (-0.84)

0.4313 a (4.27)

-0.3230 a (-3.24)

-0.2035 c (-1.95)

(4) Model 3 1.0327 a (2.95)

0.4555 a (4.80)

-0.3400 a (-3.60)

-0.1663 c (-1.67)

Panel B. Return behavior (Skip period: t-36, t+35)

Intercept D(t-120,t-37) D(t-36,t-1) D(t,t+35) D(t+36,t+120)

(1) Model 1 -0.2505 (-0.50)

0.3836 a (4.57)

-0.4488 a (-5.51)

-0.3735 a (-5.18)

(2) Model 1 0.0133 (0.26)

-0.3836 a (-4.57)

-0.8324 a (-9.36)

-0.7571 a (-9.38)

(3) Model 2 -0.3603 (-0.67)

0.4620 a (4.55)

-0.3712 a (-3.85)

-0.3364 a (-3.86)

(4) Model 3 1.1285 a (3.25)

0.4395 a (4.63)

-0.4298 a (-4.70)

-0.3338 a (-4.07)

a, b, and c denote statistical significance at the 1%, 5%, and 10% levels, respectively.

34

Table 4 (continued) Residual returns for overseas listings across event time periods

Panel C. Return behavior (Skip period: t-84, t+83)

Intercept D(t-120,t-85) D(t-84,t-1) D(t,t+83) D(t+84,t+120)

(1) Model 1 -0.3299 (-0.65)

0.3069 a (3.24)

-0.3610 a (-3.76)

-0.1769 (-1.59)

(2) Model 1 -0.0230 (0.05)

-0.3069 a (-3.24)

-0.6679 a (-10.44)

-0.4838 a (-5.62)

(3) Model 2 -0.4071 (-0.75)

0.2996 b (2.54)

-0.3373 a (-2.84)

-0.2068 (-1.52)

(4) Model 3 1.0819 a (3.04)

0.3170 a (2.85)

-0.3364 a (-2.99)

-0.1954 (-1.51)

Panel D. Changes in market risk (Model 1)

Intercept RL RL * D(t-60,t-1) RL * D(t+60,t+120)

D(t-120,t-1) * Column variable

1.1356 a (7.31)

0.9489 a (26.20)

-0.0211 (-1.15)

D(t,t+120) * Column variable

0.2041 a (4.07)

1.0001 a (73.83)

-0.0055 (-0.23)

a, b, and c denote statistical significance at the 1%, 5%, and 10% levels, respectively.

35

Table 5 Residual returns for overseas listings across event time periods, markets, and listing sequence

This table shows the results of a times-series cross-section regression of monthly residual returns for firms listing in a foreign market. The dependent variable, residual return, is computed based on the local CAPM,

itLtiLiit erbar ++= , where all returns are in US dollars less the return on the 30-day Treasury Bill. D(*) refers to a corresponding dummy variable, such that D(t-120,t-60) is equal to 1 for observations 120 to 60 months prior to the month of the overseas listing date and equal to 0 otherwise. D(Home=G5) and D(Host=G5) indicate whether the listing firm is from a G5 country or was in a G5 market, respectively. G5 countries are France, Germany, Japan, the U.K., and the U.S. D(First Listing) indicates whether the listing was the earliest calendar time listing in our database. For each regression an array of monthly calendar time dummy variables from February 1926 to December 2001 is also included but the coefficient estimates are not reported. T-statistics are reported in parentheses.

Panel A: Valuation effects across major listing markets

Intercept D(t-60,t-1) D(t,t+59) D(t+60,t+120)

Regression 1

Column variables -0.3298 (-0.65)

0.4542 a (3.66)

-0.3767 a (-3.10)

-0.0475 (-0.37)

D(Home=US) * Column variables

0.0562 (0.40)

-0.3752 b (-2.12)

0.0574 (0.33)

-0.1631 (-0.92)

D(Host=US) * Column variables

-0.0420 (-0.23)

0.6237 a (2.71)

0.1033 (0.48)

-0.4976 b (-2.13)

Regression 2

Column variables

-0.2816 (-0.55)

0.9708 a (6.23)

-0.5383 a (-3.52)

-0.3550 b (-2.14)

D(Home=G5) * Column variables

-0.0479 (-0.31)

-0.4798 b (-2.47)

0.2063 (1.12)

0.2356 (1.22)

D(Host=G5) * Column variables

0.0469 (0.33)

-0.4229 b (-2.28)

0.1120 (0.64)

-0.0145 (-0.08)

36

Table 5 (continued)

Panel B: Valuation effects across major listing markets and listing sequence

Intercept D(t-60,t-1) D(t,t+59) D(t+60,t+120)

Regression 1

Column variables -0.3696 (-0.73)

0.1729 (1.56)

-0.1895 c (-1.72)

-0.1226 (-1.07)

D(First Listing) * Column variables

0.0750 (0.61)

0.4841 a (3.03)

-0.3154 b (-2.06)

-0.1706 (-1.07)

Regression 2

Column variables -0.3609 (-0.70)

0.2755 c (1.89)

-0.1877 (-1.32)

0.0076 (0.05)

D(Home=US) * Column variables

0.0694 (0.49)

-0.3252 c (-1.83)

0.0090 (0.05)

-0.1710 (-0.96)

D(Host=US) * Column variables

-0.0644 (-0.35)

0.5540 b (2.37)

0.1840 (0.84)

-0.4702 b (-1.98)

D(First Listing) * Column variables

0.0985 (0.79)

0.3337 b (2.03)

-0.3490 b (-2.21)

-0.1228 (-0.75)

Regression 3

Column variables

-0.3249 (-0.62)

0.7636 a (4.03)

-0.3353 c (-1.81)

-0.2565 (-1.31)

D(Home=G5) * Column variables

-0.0387 (-0.25)

-0.4176 b (-2.13)

0.1561 (0.84)

0.2176 (1.12)

D(Host=G5) * Column variables

0.0626 (0.43)

-0.3840 b (-2.05)

0.0694 (0.39)

-0.0327 (-0.18)

D(First Listing) * Column variables

0.0888 (0.71)

0.2816 c (1.71)

-0.2751 c (-1.74)

-0.1545 (-0.95)

a, b, and c denote statistical significance at the 1%, 5%, and 10% levels, respectively.

37

Table 6 Residual returns for overseas listings across event time periods for largest host markets

This table shows the results of a times-series cross-section regression of monthly residual returns for firms listing in a foreign market. The dependent variable, residual return, is computed based on the local CAPM,

itLtiLiit erbar ++= , where all returns are in US dollars less the return on the 30-day Treasury Bill. D(*) refers to a corresponding dummy variable, such that D(t-120,t-60) is equal to 1 for observations 120 to 60 months prior to the month of the overseas listing date and equal to 0 otherwise. In the regression an array of monthly calendar time dummy variables from February 1926 to December 2001 is also included but the coefficient estimates are not reported. T-statistics are reported in parentheses.

Host market Intercept D(t-60,t-1) D(t,t+59) D(t+60,t+120)

Column variables -0.4316 (-0.80)

0.4134 (1.58)

-0.4304(-1.71)

c 0.1596 (0.60)

D(First Listing) * Column variables

0.0915 (0.71)

0.2990 (1.75)

c -0.2612(-1.59)

-0.1751 (-1.03)

D(Belgium)* Column variables

0.3476 (1.03)

-0.2359 (-0.54)

-0.2545(-0.61)

-0.8002 (-1.87)

c

D(France)* Column variables

0.0940 (0.32)

-0.3784 (-0.98)

0.2799(0.76)

-0.4655 (-1.23)

D(Germany)* Column variables

-0.0441 (-0.18)

-0.5308 (-1.64)

0.4657(1.50)

0.1567 (0.47)

D(Japan)* Column variables

-0.1732 (-0.54)

-0.0204 (-0.05)

0.5629(1.34)

-0.1389 (-0.32)

D(Luxembourg)* Column variables

0.0538 (0.12)

0.0496 (0.10)

-0.1766(-0.35)

0.2059 (0.39)

D(Netherlands)* Column variables

-0.1181 (-0.43)

-0.2825 (-0.77)

0.5324(1.54)

0.0414 (0.12)

D(Switzerland)* Column variables

0.1335 (0.52)

-0.3607 (-1.06)

0.1064(0.32)

-0.4700 (-1.39)

D(United Kingdom)* Column variables

0.1340 (0.55)

-0.4266 (-1.34)

0.1663(0.55)

-0.3971 (-1.25)

D(United States)* Column variables

-0.0752 (-0.31)

0.4393 (1.40)

0.3749(1.27)

-0.5553 (-1.74)

c

a, b, and c denote statistical significance at the 1%, 5%, and 10% levels, respectively.

38

Table 7 Market characteristics

This table, which is a reduced version of Table 6 in Sarkissian and Schill (2003), provides aggregate economic, equity market and exchange listing standards data for our sample of 35 countries that have issued listings abroad. Market capitalization is from the International Federation of Stock Exchanges (FIBV). Liquidity is based on the Elkins/McSherry estimates of one-way transactions costs in basis points. The listing standards for IAS GAAP are acceptable if the country allows foreign firms to list with IAS GAAP-based financial statements. The listing standards for US GAAP are acceptable if the country allows foreign firms to list with US GAAP-based financial statements. The investor protection measure is based on La Porta et al. (1997).

Listing standards

Market cap ($B)

Liquidity

IAS GAAP

US GAAP

Investor protection

Australia 329 54.7 Yes No 10.00Austria 36 43.8 Yes Yes 10.00Belgium 246 35.0 Yes No 10.00Brazil 161 58.0 No No 6.32Canada 543 52.4 No No 10.00Chile 52 84.3 No No 7.02Czech Rep. 10 143.7 No No n.a.Denmark 99 40.7 Yes Yes 10.00Finland 155 43.4 Yes No 10.00France 992 29.5 Yes No 8.98Germany 1,094 37.7 Yes No 9.23Hong Kong 344 59.8 No No 8.22India 93 71.6 Yes No 4.17Ireland 67 130.7 No No 7.80Italy 570 34.8 Yes No 8.33Japan 2,440 41.3 Yes No 8.98Luxembourg 38 63.8 Yes Yes 10.00Mexico 92 61.7 Yes No 5.35Netherlands 603 42.2 Yes Yes 10.00New Zealand 25 47.2 Yes Yes 10.00Norway 46 44.6 Yes Yes 10.00Peru 10 95.8 Yes No 2.50Philippines 35 112.7 No No 2.73Poland 21 n.a. No No n.a.Portugal 63 62.7 Yes No 8.68Singapore 97 77.5 Yes No 8.57South Africa 151 81.6 Yes No 4.42Spain 402 41.9 No No 7.80Sweden 279 35.8 No No 10.00Switzerland 689 38.5 Yes Yes 10.00Taiwan 261 74.6 No No 8.82Thailand 34 89.1 Yes Yes 3.45Turkey 34 64.6 Yes Yes 5.18UK 2,373 54.5 Yes No 8.57USA 12,926 38.1 No Yes 10.00

39

Table 8 Listing sequencing and market selection

This table shows regression results various host and home market characteristics on the dummies representing the sequencing order of the listing. CORRELATION is the correlation coefficient between the host and home equity market indices over the period 1990 to 1998. MARKET SIZE is the US dollar market capitalization of the host market divided by the capitalization of the home market. LIQUIDITY is the Elkins/McShery measure of round-trip transaction costs of the host market divided by the cost measure of the home market. STRINGENCY is a dummy that takes the value of one if host market has more stringent disclosure requirements than that of the home market. LAW is a dummy variable that takes the value one if the legal protection variable from La Porta et al. (1997) is greater that that of the home market. T-statistics are reported in parentheses.

Market properties Intercept First listing Second listing Third listing

CORRELATION

0.4670 a (33.96)

0.0149 (0.98)

0.0082 (0.45)

0.0180 (0.85)

MARKET SIZE

-0.9418 a (-5.17)

1.4699 a (7.22)

0.2047 (0.82)

0.0823 (0.28)

LIQUIDITY

0.0344 (1.56)

-0.1465 a (-5.93)

-0.0545 c (-1.81)

-0.0837 b (-2.39)

STRINGENCY

0.4162 a (11.39)

0.1499 a (3.68)

0.1043 b (2.10)

0.1160 b (2.01)

LAW 0.7541 a (21.9)

-0.1234 a (-3.17)

-0.0411 (-0.88)

-0.0261 (-0.48)

a, b, and c denote statistical significance at the 1%, 5%, and 10% levels, respectively.

40

Table 9 Residual returns for overseas listings across event time periods and market characteristics

This table shows the results of a times-series cross-section regression of monthly residual returns for firms listing in a foreign market. The dependent variable, residual return, is computed based on the local CAPM,

itLtiLiit erbar ++= , where all returns are in US dollars less the return on the 30-day Treasury Bill. D(*) refers to a corresponding dummy variable, such that D(t-60,t-1) is equal to 1 for observations 60 to 1 month prior to the month of the overseas listing date and equal to 0 otherwise. CORRELATION is the correlation between the host and home equity markets. MARKET SIZE is the relative size of the host market equity capitalization divided by the market equity capitalization of the home market. LIQUIDITY is the ratio of the host country liquidity to that in the home country as measured by Elkins/McShery. STRINGENCY is a dummy, which is set to 1 if the host country has more stringent disclosure requirements than the home country. LAW is a dummy, which is set to 1 if the host country has better investor protection laws than the home country. For each regression an array of monthly calendar time dummy variables from February 1926 to December 2001 is also included but the coefficient estimates are not reported. T-statistics are reported in parentheses.

Panel A: Full sample

Intercept D(t-60,t-1) D(t,t+59) D(t+60,t+120)

Column variables

-0.5594 (-1.01)

-0.3620 (-1.63)

0.0373 (0.18)

0.4795 c (1.95)

D(First Listing) * Column variables

0.0877 (0.71)

0.3016 c (1.80)

-0.2757 c (-1.72)

-0.1820 (-1.09)

CORRELATION * Column variables

0.6253 c (1.69)

-0.1438 (-0.45)

-0.4741 (-1.22)

MARKET SIZE * Column variables

0.0515 c (1.86)

0.0359 (1.53)

-0.0370 (-1.39)

LIQUIDITY * Column variables

-0.7837 a (-3.41)

0.4037 b (2.12)

0.3152 (1.50)

DISCLOSURE * Column variables

0.1670 (1.46)

-0.0649 (-0.64)

-0.1836 (-1.64)

LAW * Column variables

0.2900 b (2.16)

-0.0843 (-0.74)

-0.3912 a (-3.07)

a, b, and c denote statistical significance at the 1%, 5%, and 10% levels, respectively.

41

Table 9 (continued)

Panel B: 1990s sample

Intercept D(t-60,t-1) D(t,t+59) D(t+60,t+120)

Column variables

0.0221 (0.13)

-0.7033 c (-1.73)

0.0361 (0.09)

1.6216 a (2.66)

D(First Listing) * Column variables

0.0871 (0.34)

0.8067 b (2.29)

-0.5812 c (-1.71)

-0.3002 (-0.72)

CORRELATION * Column variables

0.4204 (0.63)

0.2119 (0.36)

-1.9450 b (-2.04)

MARKET SIZE * Column variables

-0.0128 (-0.23)

0.1157 b (2.41)

-0.0996 (-1.35)

LIQUIDITY * Column variables

-0.91015 c (-1.88)

0.7941 b (1.99)

0.4055 (0.68)

DISCLOSURE * Column variables

0.3372 (1.42)

-0.1058 (-0.50)

-0.3887 (-1.25)

LAW * Column variables

0.4011 (1.54)

-0.1598 (-0.70)

-0.6214 c (-1.95)

42

Table 10 Residual returns for overseas listings across event time periods and firm characteristics

This table shows the results of a times-series cross-section regression of monthly residual returns for firms listing in a foreign market. The dependent variable, residual return, is computed based on the local CAPM,

itLtiLiit erbar ++= , where all returns are in US dollars less the return on the 30-day Treasury Bill. D(*) refers to a corresponding dummy variable, such that D(t-60,t-1) is equal to 1 for observations 60 to 1 month prior to the month of the overseas listing date and equal to 0 otherwise. Firm tradability is defined on an industry classification of the exportability of the firm’s output. Small firms are those whose market capitalization is below the monthly median for all overseas listing firms within the sample period and within the same G5 classification. In the regression an array of monthly calendar time dummy variables from February 1926 to December 2001 is also included but the coefficient estimates are not reported. The t-statistics are reported in parentheses.

Intercept D(t-60,t-1) D(t,t+59) D(t+60,t+120)

Regression 1

Column variables -0.5573 (-1.09)

0.4814 a (3.37)

0.0104 (0.08)

-0.0195 (-0.13)

D(Firm=Tradable) * Column variables

0.2581 b (1.99)

-0.0879 (-0.52)

-0.5581 a (-3.47)

-0.2694 (-1.58)

Regression 2

Intercept * Column variables

-0.3015 (-0.59)

0.1282 (1.06)

-0.1414 (-1.22)

-0.1715 (-1.45)

D(Firm=Large) * Column variables

0.0738 (0.61)

-0.5333 a (-3.34)

0.4938 a (3.25)

0.1294 (0.81)

Regression 3

Column variables -0.4608 (-0.86)

0.5631 b (2.37)

0.2110 (0.92)

-0.0465 (-0.19)

D(Firm=Tradable) * Column variables

0.2756 b (2.07)

-0.0188 (-0.11)

-0.5957 a (-3.64)

-0.2978 c (-1.71)

D(Firm=Large) * Column variables

0.1014 (0.76)

-0.3722 b (-2.12)

0.4588 a (2.77)

0.1249 (0.71)

D(Home=G5) * Column variables

-0.0528 (-0.34)

-0.4606 b (-2.31)

0.3023 (1.61)

0.2747 (1.39)

D(Host=G5) * Column variables

0.0105 (0.07)

-0.2460 (-1.25)

-0.0546 (-0.29)

-0.0571 (-0.30)

D(First Listing) * Column variables

0.1487 (1.15)

0.1754 (1.02)

-0.2280 (-1.39)

-0.1560 (-0.92)

a, b, and c denote statistical significance at the 1%, 5%, and 10% levels, respectively.

43

-10%

0%

10%

20%

30%

-120 -60 0 60 120

Months relative to listing

Annu

aliz

ed re

turn

Gross returnResidual return

Period 1 Period 2 Period 3 Period 4

Figure 1. Annualized moving-average returns for overseas listings. Monthly gross and residual dollar-denominated returns are averaged across firms and then annualized by cumulating the returns over months t-12 to t+11 and dividing by two. The residual returns are computed for each firm based on the home market risk model (Model 1). The lot also shows the four distinct periods in the dynamics of firm returns around the listing.

44

-20%

-10%

0%

10%

20%

30%

-120 -60 0 60 120

Months relative to listing

Annu

aliz

ed re

sidu

al re

turn

.Host: USHome: USHome/Host: Non-US

0

200

400

600

800

-120 -60 0 60 120Months relative to listing

Freq

uenc

y of

obs

erva

tions

Host: USHome: USHome/Host: Non-US

Figure 2. Annualized moving-average returns for overseas listings by home and host markets. Monthly residual dollar-denominated returns are averaged across firms and then annualized by cumulating the returns over months t-12 to t+11 and dividing by two. The residual returns are computed for each firm based on the home market risk model (Model 1). The number of monthly observations for the respective sub-samples is presented in the lower plot.

45

-10%

0%

10%

20%

30%

-60 0 60 120

Months relative to listing

Annu

aliz

ed re

sidu

al re

turn

.FirstSecondThirdFourth+

0

200

400

600

800

-60 0 60 120Months relative to listing

Freq

uenc

y of

obs

erva

tions

FirstSecondThirdFourth+

Figure 3. Annualized moving-average returns for overseas listings by listing sequence. Monthly residual dollar-denominated returns are averaged across firms and then annualized by cumulating the returns over months t-12 to t+11 and dividing by two. The residual returns are computed for each firm based on the home market risk model (Model 1). The first second, and third listing refer to the calendar order of listing date in our data set. The fourth and above listings are pooled into one group. The number of monthly observations for the respective sub-samples is presented in the lower plot.