Changes in systematic risk following global equity issuance

Transcription

1 Journal of Banking & Finance 24 (2000) 1491± Changes in systematic risk following global equity issuance Latha Ramchand *, Pricha Sethapakdi Department of Finance, College of Business Administration, University of Houston, Houston, TX , USA Received 13 March 1999; accepted 16 July 1999 Abstract This paper examines changes in systematic risk following global equity issues by US rms. Models of market segmentation show that if international capital markets are not fully integrated and demand curves for securities are downward sloping, rms issue equity at higher prices by issuing in multiple markets compared to issuance on a single domestic market. This would imply a reduction in the rmõs cost of capital and an increase in rm value. Using a sample of global equity o ers during 1986±1993, we nd that US rms that issue equity abroad experience a decline in systematic risk subsequent to issuance. After controlling for size, volume, and leverage e ects, we nd that this decline in systematic risk is larger in magnitude for global compared to a control sample of domestic equity issues. The larger the proportion of the o er sold abroad and the larger the increase in trading volume, the bigger the decline in systematic risk. Using a two-factor global risk model we nd that while rms issuing equity abroad experience a decline in the domestic component of systematic risk, the foreign component increases. Overall, however, the net e ect is a decline in the cost of capital. Ó 2000 Elsevier Science B.V. All rights reserved. JEL classi cation: F30; G15 Keywords: Global; Equity; Systematic; Risk; Capital * Corresponding author. Tel.: ; fax: address: ramchand@jetson.uh.edu (L. Ramchand) /00/$ - see front matter Ó 2000 Elsevier Science B.V. All rights reserved. PII: S (99)

2 1492 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± Introduction The internationalization of security markets has enabled rms worldwide to seek and obtain alternative sources of capital. While US rms have been listing on foreign markets since the 1960s, a more recent phenomenon involves global equity o erings by US rms. 1 A global equity o er involves the simultaneous sale of equity by a US rm on the US and one or more foreign markets. The development of this market can be traced to the late 1980s. While there were a couple of global equity issues in 1983, the real origins of this market can be traced to 1985 when there were a total of 12 issues. The total amount of global equity raised was roughly $600 million. By 1993 both the total number of issues as well as the total volume of global equity issued by US rms had increased nearly sevenfold ± a total of 99 issues were made that raised close to $4 billion in equity. The proportion of the o er raised abroad is on average 21%. Increasingly, larger equity issues are being o ered for sale on multiple markets. As Chaplinsky and Ramchand (1998) document, while 60% of o ers in excess of $100 million were o ered in multiple markets in the period 1989±1991, this proportion had increased to 70% for 1992±1995. The trend seems to be for US rms desiring to issue large amounts of equity to raise capital on multiple markets. The motivations to issue global equity are similar to those found in the literature on cross-border listings: 2 to enhance the liquidity of the rmõs shares by increasing heterogeneity of the investor base and to reduce the rmõs cost of capital. Such o ers are believed to increase demand and hence share price and market value. With downward sloping demand curves for the rm's shares, global equity o ers lead to a rightward shift of the demand curve. This reduces the price pressure e ects when equity is raised leading to a higher share price and hence market value relative to o ers that are sold exclusively on domestic markets. Chaplinsky and Ramchand (1998) document that the negative price reaction at announcement associated with equity o ers is signi cantly lower for global compared to domestic o ers by 0.8%. Global equity issuance can also be undertaken with the strategic objective of gaining a foothold in a foreign market so as to link the rmõs product and capital markets. Investment bankers and rms also cite reduction in stock price volatility as another reason for doing a global as opposed to a domestic o er. Stock price volatility is driven by systematic as well as unsystematic risk. Placing shares in the hands of foreign investors, it is believed could make these securities less sensitive to domestic systematic risk. On the other hand, this could increase the rmõs exposure to foreign market shocks including changes 1 See Chaplinsky and Ramchand (1998). 2 See Stapleton and Subramanyam (1977), Alexander et al. (1987) and Karolyi (1998).

3 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± in exchange rates and foreign interest rates. Changes in risk are important not merely for theoretical reasons but also since they a ect the required rate of return on equity and hence the rm's cost of capital, which is of concern to managers. This paper examines changes in systematic risk subsequent to the issue of equity by US rms in multiple markets (global o ers). We examine a sample of global o ers from 1986±1993 and document changes in stock price volatility and systematic risk after the issue. Based on changes in risk we then document the changes in cost of capital, if any, for rms that issue on foreign markets. Speci cally we examine changes in stock price volatility and systematic (beta) risk with respect to the domestic (US) market for US rms that issue equity abroad. These changes are measured around a 250 day trading window both before and after the event as well as using a 48 month window before and after the event. We then compare these changes to changes in domestic beta risk for a control sample of rms issuing equity on domestic markets only. Following this, we examine if global equity issuance is also accompanied by changes in exposure to foreign market risk. Using the results on changes in systematic risk, both domestic and foreign, we examine changes in the cost of capital in the context of a two-factor model. Changes in risk ought to a ect investorsõ rates of return. Using methodology similar to that employed in Ritter (1991) and Foerster and Karolyi (1998a,b) we examine the pattern of abnormal returns of global issuers the year after the o er relative to returns for rms that raise equity on domestic markets. Our results indicate that beta estimates decline subsequent to equity issuance, consistent with the decline in nancial leverage. 3 We also nd that after controlling for changes in trading volume and issue characteristics, rms issuing abroad experience a greater decline in beta compared to rms issuing equity on domestic markets only. We con rm the robustness of our results using monthly returns as well as by using a pooled cross-section/time series regression. We also nd that the larger the proportion of the issue raised abroad, greater is the decline in systematic risk subsequent to the o er. Using a two-factor model with the domestic and a foreign index as the two factors, we nd that for global issues, while the domestic component of systematic risk (domestic beta) declines, the systematic risk with respect to the foreign market (foreign beta) increases, subsequent to the issue. The net e ect is a decline in the 3 This is consistent with the results in Denis and Kadlec (1994). Denis and Kadlec (1994) demonstrate how changes in trading activity associated with the issuance of equity can lead to biased estimates of systematic risk. They demonstrate how OLS regressions can result in increases in beta estimates following equity issuance since these estimates do not correct for changes in the frequency of trading. They further show that when these changes are accounted for using the Scholes and Williams (1977) correction procedure, consistent with a leverage e ect, beta declines following equity issues.

4 1494 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491±1513 overall cost of capital although the decrease is not signi cant. Finally there is some evidence to suggest that global issues experience a signi cantly higher one-year return compared to domestic issues. The remainder of this paper is organized as follows: Section 2 examines the theoretical motivations for global equity issuance and its impact on risk. Section 3 describes our data. Section 4 reports the empirical results on changes in stock price volatility and trading volume and systematic risk for global issues and compares them to a control sample of domestic issues. This section also examines the implications of changes in systematic risk on the cost of capital using a two-factor model as in Karolyi (1998). Section 5 gives our conclusions. 2. Global issuance and changes in risk If international capital markets are not completely integrated and demand curves are downward sloping, a global equity issue can, by shifting the demand curve to the right, mitigate the downward price pressure on prices when new equity is issued. This would imply a higher price for the shares and hence a reduction in the cost of capital to the rm. Market segmentation could result from various factors ranging from legal investment barriers to taxes and transactions costs. Models of market segmentation suggest that any activity by the rm that lowers the cost of segmentation can enhance rm value. 4 A global issue could also result in a lowering of the transactions and information costs associated with the purchase of these securities by foreign investors. Parsons and Raviv (1985) and Benveniste and Spindt (1989) point out that the marketing e orts accompanying a rm-commitment o er can potentially increase the o er price for an issue. While foreign investors can purchase these securities on a US exchange even in the absence of a global o er, the selling activities associated with a global o er can reduce the information costs faced by foreign investors. The introduction of a new market competing for order ow could also reduce trading costs. 5 Finally, to the extent that global issues a ord foreign investors tax bene ts relative to purchases of securities directly on a US exchange, the return on these securities could be lower. All these factors imply that if capital markets are segmented, securities o ered on global markets will command a higher price and hence lower the rm's cost of capital. Segmentation of markets could also lead to changes in systematic risk when rms access foreign capital markets. Previous studies on the e ect of cross- 4 See Black (1974), Stapelton and Subramanyam (1977), Stulz (1981), Errunza and Losq (1985), Alexander et al. (1987, 1988) and Stulz and Wasserfallen (1995) for models of market segmentation. 5 See Chowdhury and Nanda (1991) and Domowitz et al. (1997, 1998).

5 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± border listings nd that subsequent to the listing, rms experience changes in the domestic market beta. 6 These studies also nd that such listings lead to changes in systematic risk while keeping overall variances unchanged. 7 For cross-listed stocks, changes in systematic risk result from changes in exposure to both the domestic as well as the foreign market. To the extent that the domestic and foreign markets are not completely integrated, foreign listings increase the exposure of the rmõs returns to foreign market uctuations as well as changes in exchange rates. For instance, Howe and Madura (1990) nd that for US rms listing in Europe or Japan, domestic betas measured with respect to the S&P 500 index drops while the beta with respect to the foreign market increases. 8 Further, Varela and Lee (1993) nd that US rms listing in London experience a decline in the cost of capital to the tune of 240 basis points subsequent to listing. As a complement to this literature, Foerster and Karolyi (1999) nd that foreign rms listing in the US experience a decline in their local market betas while global market risk is unchanged. 9 Jayaraman et al. (1993) report that the listing of American Depositary Receipts results in a permanent increase in risk and return of the underlying securities (see also Miller, 1999). Generalizing these ndings on cross-listings to equity issues would suggest that equity issues by US rms on a foreign market could also result in changes in systematic risk. It should be noted, however, that equity issuance on a foreign market does not automatically imply that the shares are listed on the foreign market. In fact only a small proportion of the rms in our sample (22%) also list their shares abroad. At the same time rms need not always issue equity on markets where they are also listed. Listing could have a di erent impact on trading volume and trading patterns when compared to issuance. For instance, rms that issue equity but do not have a foreign listing could experience a decline in trading volume if foreign investors do not trade as frequently, either because their objectives are di erent or because they face signi cant transactions costs. Foerster and Karolyi (1998a) nd that post-issue abnormal returns of foreign rms issuing equity in the US are positively related to the magnitude of trading volume shifted to the US market. For US rms issuing abroad, the same need not be true to the extent that the proportion of trading volume shifted abroad may not be as high as that of foreign rms issuing in the US. This would imply a smaller increase in the systematic risk with respect to the foreign index. 6 See Karolyi (1998) and Stulz (1998) for an exhaustive review of this literature. 7 See Howe and Kelm (1987), Lee (1991), Torabzadeh et al. (1992), Damodaran et al. (1993), Varela and Lee (1993), Lau et al. (1994), Rothman (1995), Foerster and Karolyi (1998a,b). 8 Also see Torabzadeh et al. (1992), and Damodaran et al. (1993). 9 Karolyi (1998) uses a multi-factor model to study the impact of changes in the components of systematic risk on the rmõs cost of capital and nds that non-us rms listing in the US experience a decline in the cost of capital averaging 114 basis points.

6 1496 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491±1513 To the extent that the foreign tranche of the equity o er subsequently results in a `` ow-back'' to the domestic market, these e ects will be reduced resulting in risk and return remaining unchanged. The change in systematic risk is important since it impacts on the required rate of return and hence the rm's cost of capital. Whether or not systematic risk and hence the required rate of return declines, is an empirical issue that we examine below. 3. Data Our sample consists of global equity o ers made by US rms during 1986± By global o er we refer to the simultaneous sale of equity by a US rm on the US and one or more foreign markets. This de nition is largely an artifact of the data we use in which all issues involve a foreign as well as a domestic tranche. The domestic tranche varies anywhere from 11% to 50% with an average of 21%. While the prospectuses state the amount of the o er to be sold in the foreign tranche, there is almost always no mention of the exact foreign market where the issue will be sold. The prospectuses merely state that a certain percentage of the issue will be sold outside the US. All o ers in this sample are seasoned equity o ers (SEOs). Since all the issues in this sample involve simultaneous issues in the domestic as well as one more foreign markets, these o ers do have to meet SEC requirements and have to be o ered for sale at the same price on all markets. The tax implications on these securities are no di erent from those relevant for US holdings by foreign investors in general and are detailed in the prospectus. Furthermore, the institutional details of the marketing process are similar to those of a domestic o er except for the fact that the underwriting syndicate comprises international underwriters who, in general, are the international a liates of the domestic book manager. Also, as explained in Chaplinsky and Ramchand (1998), these o ers are almost always marketed at least initially to foreign institutional investors. We restrict our attention to rm commitment o ers made by industrial rms, all of which involve the sale of primary shares. 11 By primary we refer to new shares as opposed to the sale of existing shares. We also eliminate from our sample equity o erings connected with closed-end investment funds. Our sample includes 309 issues by 262 rms of which there are 190 NYSE/AMEX 10 This sample is a subset of the global equity issues used in Chaplinsky and Ramchand (1998) obtained from the New Issues Database of the Securities Data Company (SDC). 11 We exclude equity o erings by utilities and nancials as well as those by closed-end investment funds.

7 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± and 72 NASDAQ rms. 12 For all of our o ers we require stock return data from the Center for Research in Security Prices (CRSP) as well as company speci c information from the COMPUSTAT tapes. Since we examine changes in systematic risk using daily returns as well as monthly returns we require data on daily returns around a 250 day window both before and after the event as well as monthly returns around a 48 month window before and after the event. As a result of this our nal sample reduces to 147 issues of which 122 are NYSE/AMEX rms and 25 are NASDAQ. 13 Our control sample of domestic equity issues is obtained from the Disclosure 33 CD database. The control sample is limited to primary equity issues that are sold in the US market only. To ensure proper control we also eliminate from this group closed-end investment funds and OTC rms. The control sample is constructed in the following manner: Step 1: For every global issue, we choose a domestic issue that is closest in rm size to the global issue. This is done using the rmõs market value of equity (MVEQ) prior to the issue. Step 2: We then choose from among the subset of the domestic issues that are rm size matched, that issue which is closest in terms of the issue date to the global issue. 14 Since trading volume changes induced by equity issuance could be di erent across NYSE/AMEX versus NASDAQ rms, throughout the paper we separate the two groups. 15 Table 1 reports issue characteristics of NYSE/AMEX and NASDAQ rms separately. From Panel A, the average size of global 12 There are 4 OTC rms that are eliminated from this sample. 13 Our analysis of systematic (beta) risk is based on daily returns over a 250 day period both prior to and after the event. In addition, we estimate monthly betas using a 48 month horizon prior to and after the event. If the rm does not have at least 50 observations of daily returns both prior to and after the event or if it does not have at least 30 months of data prior to and after the event, it is eliminated from the sample. The binding constraint is the requirement of 30 monthly returns before and after the event, which eliminates rms that made initial public o erings within a 2 yr period prior to our o er date (the median time between IPO and the seasoned o er for rms in our sample is 29 months). This reduces the sample size to 174 issues. The remaining loss is due to lack of nancial data on COMPUSTAT. All results on global issues in Tables 1±5 are based on this nal sample of 147 rms. 14 We did analyze the results using two other ways : One, matching by asset size prior to the issue and then by issue date and two, by matching using book-to-market ratios prior to the o er and then by the issue date. The results using these methods are qualitatively identical to those reported in the paper. 15 Size, both issue size and rm size, would be one reason to separate the NYSE/AMEX from the NASDAQ rms. Di erences in size are often used as proxies for risk. Hence, the di erences reported across these two categories could be associated with risk di erences. It could also be due to di erences in liquidity across the two markets leading to di erential changes in systematic risk after equity issue. Bessembinder and Kaufman (1997) show that trading costs on the NYSE measured as realized bid±ask spreads are lower than those on the NASDAQ by a factor of two to three.

8 1498 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491±1513 Table 1 Variable NYSE/AMEX NASDAQ Panel A: ± Issuer characteristics of global issues ± 1986±1993 a Issue size ($ millions) $250 $67 # Shares issued (millions) % Issued abroad 21% 22% # Shares issued abroad (millions) Size/MVEQ MVEQ ($ millions) $1823 $322 CAR 1; Change in D/E N Panel B ± Number of issues by year of issue Year NYSE/AMEX NASDAQ Total b a Size is the issue size in millions of dollars. # Shares Issued is the number of shares o ered for sale in millions. % Issued Abroad is the proportion of the issue sold abroad. # Shares Issued Abroad is the number of shares in millions o ered for sale in the foreign markets. Size/MVEQ is the size of the issue relative to the market value of the issue at the time of the o er. MVEQ is the market value of the rm's equity at the time of the o er in millions of dollars. CAR ( 1,+1) is the cumulative abnormal return calculated around a 3 day window relative to the announcement date of the o er using market adjusted returns and change in D/E is the change in the debt to equity ratio after the issuance of equity. N denotes the number of observations in each group. b Total denotes the total number of issues for a given year for the 147 global issues in our nal sample. * Signi cant at 1% level. equity o ers by NYSE/AMEX rms is $250 million, of which, $53 million or roughly 21% is raised abroad. While the number of global issues has been increasing over time (Panel B), the average proportion of the foreign equity tranche (Panel A) has remained more or less constant at 20%. On average global issues result in an increase in the number of shares by about 8 million for NYSE/AMEX and 3 million for NASDAQ rms (Panel A). The number of shares sold abroad is roughly 20% of the total number of shares issued and ranges from about 1.53 million for NYSE/AMEX rms to about 0.6 million for NASDAQ rms (Panel A). Both in terms of the average size and the

9 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± number of shares issued, global issues are much larger than domestic issues. 16 On average, global issue sizes are about 19% of the existing market value of the rm for NYSE/AMEX rms and about 35% for NASDAQ rms. The average global issuer is also typically a large rm as evidenced by the MVEQ in Panel A of Table 1. The mean market capitalization of the global issuer in the NYSE/ AMEX group is almost $2 billion. 17 Panel A of Table 1 also reports the average cumulative abnormal return CAR 1; 1 calculated using a 3 day window relative to the announcement date for global issues. The mean CAR 1; 1 is 2.3% for NYSE/AMEX and 2.6% for NASDAQ issuers both of which are signi cant at 1%. Chaplinsky and Ramchand (1998) report that after controlling for rm size and risk, the CAR 1; 1 is less adverse by 0.8% for global issues compared to domestic issues sold exclusively on the domestic market. The last row of Panel A is the change in the debt to equity ratio after the issuance of equity. With the issue of new equity, the debt to equity ratio declines by 0.29 for NYSE/AMEX rms and declines by 0.03 for NASDAQ rms. 4. Empirical results 4.1. Changes in stock price volatility and trading volume We document the changes in volatility accompanying global issuance by comparing pre and post issue stock price volatility. The market microstructure literature relates changes in stock price variance to changes in volume and trading activity. 18 Hence we do the same comparison for trading volume. 19 We also compare these changes to the control sample of domestic issues matched by market value and date of the o er. These results are reported in Table 2. Panel A of Table 2 reports the pre and post o er standard deviation of the stock price for NYSE/AMEX rms. This is calculated around a 250-day window relative to the listing and o er dates. More precisely the pre-o er standard deviation is the standard deviation of the rmõs stock price measured over 250 days starting 300 days before the announcement date and ending The average size of a comparable NYSE/AMEX domestic issue is $117 million and the MVEQ is $1.6 billion. For NASDAQ issues the issue and rm size of a domestic issue are $44 and $320 million, respectively. 17 Some of the more well-known rms in this group are Time Warner, Occidental Petroleum, Westinghouse Electric, Union Carbide, Union Texas Petroleum, Sears Roebuck, Motorola and Inland Steel to name a few. 18 See Forster and Viswanathan (1993). 19 See also Christie and Huang (1993) and Foerster and Karolyi (1998a).

10 1500 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491±1513 Table 2 Changes in stock price volatility and trading volume: Comparison between global and domestic o ers ± 1986±1993 a Variable Global o ers Domestic o ers Di erence Pre-o er Posto er di erence period o er di erence groups Ratio/ Pre-o er Post- Ratio/ between period period (p-value) period (p-value) (p-value) Panel A ± NYSE/AMEX rms Standard (0.01) deviation (0.31) (0.40) Trading volume % % (<0.01) (<0.01) (<0.01) Panel B ± NASDAQ rms Standard (0.31) deviation (0.46) (0.41) Trading % % (0.83) volume (<0.01) (<0.01) a The Pre-o er period is a 250-day window that starts 300 days before the announcement and ends 50 days before the announcement. The post-o er period is a 250-day window starting 50 days after the o er date and ending 300 days after the o er date. The rst row in columns 2, 3, 5 and 6 are means and the second row reports medians. The column labelled ÔRatio/Di erence (p-value)õ reports the ratio of pre to post-o er standard deviations and the p-value for a di erence in standard deviations using an F-test. For trading volume, the value in this column is the di erence between the pre and the post o er period and the p-value (in parantheses) of a di erence of means t-test between the pre and post o er values. The last column titled ÔDi erence between the groupsõ is the p-value associated with a di erence of means t-test across the two groups viz. Global and domestic o ers. days prior to the announcement date. Similarly the post o er period starts 50 days after the o er date and ends 300 days after the o er date. Trading volume is measured as the number of shares traded daily as a proportion of the total number of shares outstanding as of that day. The results in Panel A suggest that NYSE/AMEX rms making a global o er experience a decline in stock price volatility of 8.7% ( ˆ (0:021 0:023)/0.023) subsequent to the issue. This contrasts with an increase in volatility of 4% for rms issuing equity on domestic markets only. Furthermore, the di erence across the two groups is signi cant at the 1% level. Trading volume increases for both groups although the increase is signi cantly larger for global issues compared to domestic issues. 20 In the case of NASDAQ rms reported in Panel B, global as well as domestic issues result in a decline in stock price volatility and increase in trading volume of similar magnitudes. In other words, NASDAQ rms which 20 It should be noted that these comparisons do not control for rm size, number of shares and other rm speci c characteristics.

11 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± are on average smaller in size, experience similar changes in liquidity that arise from a larger equity base, no matter where they issue Changes in systematic risk Next we estimate changes in systematic (beta) risk using the same 250 day window before the announcement date and after the o er date. We calculate beta by regressing daily stock returns on the CRSP equal-weighted index. To account for estimation biases resulting from frictions in the trading process, we follow Denis and Kadlec (1994) and estimate betas using a symmetric lead/lag of 1, 2, 5, 10 and 15 days but only report results using OLS betas and betas using a lead/lag of 2 and 10 days. 21 This is done using the Scholes±Williams (1977) method regressing returns on leading, contemporaneous and lagged market returns (see also Fowler and Rorke, 1983). The results for the global as well as the control sample of domestic equity issues are reported in Table 3. Consistent with Denis and Kadlec (1994), as the number of leads and lags is increased, estimates of systematic risk decline for both domestic and global o ers. Using a symmetric 2-day lead/lag structure, the mean beta for NYSE/ AMEX rms declines from 1.28 in the pre-o er period to 1.23 post-o er (medians decline from 0.81 to 0.76) while average betas using a 10 day lead/lag structure change from 0.96 to 0.77, a decline of 0.19 that is signi cant at the 10% level. 22 The longer lead/lag structure is meant to reduce the problems associated with infrequent trading that could characterize OLS betas. As a further check, the last row has betas estimated using monthly returns for a 48- month window starting one month before and ending 48 months before the announcement month. The post-o er betas using monthly returns are similarly calculated using monthly returns starting one month after and ending 48 months after the o er month. Similar to the 10-day lead/lag betas, the monthly betas for global issues decline from 0.95 to 0.82, a decline that is signi cant at the 10% level. Monthly betas for domestic issues also decline although the decline is not signi cant. Furthermore, the p-values for the Wilcoxon signed rank test imply that the decline in betas is signi cantly lower for the global compared to the domestic issues at the 5% level. For NASDAQ rms in Panel B of Table 3, beta estimates increase for global o ers and decline for domestic o ers. 23 In general, the changes are not signi cant. In addition, the p values for a di erence of means t test across the two groups and for a Wilcoxon 21 The results using betas computed with 1, 5 and 15 day lead/lags are qualitatively identical. 22 p-values reported are associated with a di erence of means t-test and a non-parameteric Wilcoxon signed rank test. 23 Foerster and Karolyi (1999) nd that foreign rms listing in the US experience a decline in their domestic component of risk whereas Jayarman et al. (1993) report a permanent increase in risk after the listing of ADRs on the US market.

12 1502 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491±1513 Table 3 Changes in systematic risk for domestic and global equity o erings ± 1986±1993 a Beta Global o ers Domestic o ers Di erence Pre-o er Post-o er Change Pre-o er Post-o er Change p-value Panel A ± NYSE and AMEX rms OLS L L Monthly Panel B ± NASDAQ rms OLS L <0.01 L Monthly a Beta is calculated by regressing the daily stock return on the CRSP equal weighted index return for a 250 day window. Pre-o er betas are calculated using the 250 window starting 50 days before and ending 300 days before the announcement day. Post-o er betas are calculated using the 250 day window starting 50 days after and ending 300 days after the o er day. OLS is the ordinary least squares regression and LN is the beta calculated using the Scholes and Williams (1977) correction using N symmetric leads and lags for the market index. Monthly refers to betas calculated using monthly returns over a 48 month interval. Pre-o er monthly betas are calculated starting one month prior to the announcement month and ending 48 months prior to the announcement month. Post-o er monthly betas are calculated starting one month after and ending 48 months after the o er month. The rst value in each cell is the mean for that variable while the second is the median. Change is the di erence between pre and post-o er betas (means and medians). The last column titled ``Di erence p-value'' contains the p value for di erence of means t test across the domestic and global sub-groups ( rst row) and for the Z statistic associated with a Wilcoxon signed rank test (second row). * Signi cance level 5%. ** Signi cance level 10%. signed rank test suggest that the di erences across the two groups is not signi cant. 24 Our results suggest that in most cases, when estimation biases caused by trading frequencies are corrected for, equity issuances by NYSE/AMEX rms 24 The NASDAQ rms have higher betas on average re ecting the higher risk and hence higher cost of capital for these rms. This is consistent with the overall results in the earlier tables suggesting that these rms are smaller rms (Table 1) with more variable stock prices (greater volatility as in Table 2).

13 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± are followed by a decline in a rmõs equity beta. This is consistent with a leverage e ect. From Hamada (1972): b e ˆ b a b a b d D=E ; 1 ) Db e ˆ b a D D=E ; 2 where b e, b a, and b d represent the systematic risk of the rmõs equity, assets, and debt, respectively. D represents change and D/E is the rmõs debt to equity ratio. The issuance of new equity by reducing (D/E) results in a lower beta. To the extent that global issues result in a bigger decline in the debt to equity ratio, the decline in systematic risk can be attributed to a leverage e ect. From Table 1, the decline in the debt to equity ratio is 0.29 for NYSE/ AMEX rms making global equity o ers. On the other hand, the debt to equity ratio declines by 0.12 for our control sample of domestic issues. This di erence is signi cant at the 5% level. The implication of Eq. (2) is that the decline in beta should be greater for global issues. In fact, using Eq. (2) above, for NYSE/AMEX rms, the predicted change in beta (PCB), is on average, 0.12 for global issues and 0.06 for the domestic control sample. For NASDAQ rms, the PCB is 0.07 for global and 0.03 for domestic issuers. These univariate comparisons between global and domestic issues, however, do not control for di erences in issue size, both absolute and relative, changes in volume and other characteristics across the two groups. Hence we regress the actual change in beta on the predicted change as well as on variables intended to control for size, leverage e ects and volume. We rst analyze these e ects for global and domestic issues separately. The coe cient of the PCB is used to determine if the actual change in beta is larger or smaller than the predicted change. Furthermore the di erence in magnitude of the coe cient across the two groups indicates the relative di erences between the two groups. Table 4 reports these regressions separately for the NYSE/AMEX and NASDAQ rms, respectively. The dependent variable in speci cations (1) and (2) is the change in beta estimates using a 10 day lead and lag structure. 25 The independent variables used are SIZE, which is the issue size in millions of dollars, S/M which is the issue size relative to MVEQ at the time of issuance, BTM which is the book to market ratio used in various studies of returns (see Fama and French, 1995), PCB, the PCB using Eq. (2) above and VC which measures the percentage change in volume subsequent to issuance. The PCB is 25 The results are similar when the changes in betas are estimated using 5 day lead lags.

14 1504 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491±1513 Table 4 Cross-sectional regressions of changes in beta following equity issuance 1986±1993 a Independent variable (1) Change in 10 day lead/lag beta global issues (2) Change in 10 day lead/lag beta domestic issues (3) Change in monthly beta global issues (4) Change in monthly beta domestic issues (5) Change in monthly beta domestic and global issues (6) Change in monthly beta domestic and global issues Panel A ± NYSE/AMEX rms Constant (0.15) (0.24) (0.11) (0.07) (0.01) (0.01) PCB (<0.01) (0.09) (0.01) (0.01) (<0.01) (<0.01) PCB GD (0.07) PCB %IA (0.08) SIZE 2.3E E E E E E 11 (0.16) (0.05) (0.25) (0.01) (0.23) S/M (0.07) (0.29) (0.05) (0.46) (0.11) (0.14) BTM (0.10) (0.37) (0.28) (<0.01) (0.04) (0.04) VC (0.15) (0.26) (0.25) (0.39) (0.37) (0.19) VC GD (0.41) VC %IA (0.20) R N Panel B ± NASDAQ rms Constant (0.40) (0.31) (0.30) (0.24) (0.27) (0.28)

15 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± PCB (0.26) (0.11) (0.38) (0.26) (0.03) (0.02) PCB GD (0.07) PCB %IA (0.04) SIZE 8.33E E E E 9 2.3E E 9 (0.44) (0.04) (0.49) (0.25) (0.10) (0.10) S/M (0.25) (0.07) (<0.01) (0.09) (0.13) (0.14) BTM (0.30) (0.24) (0.44) (0.28) (0.31) (0.33) VC (0.30) (0.28) (0.03) (0.21) (0.47) (0.48) VC GD (0.25) VC %IA (0.22) R N a PCB is the predicted change in beta from Hamada (1972). PCB GD is an interactive variable equal to GD times PCB where GD is a dummy variable ˆ 1 for a global issue and 0 otherwise. PCB %IA is an interactive variable equal to PCB times the proportion of the o er issued abroad %IA. SIZE is the issue size in millions of dollars. S/M is the ratio of issue size to market value of equity one year prior to the o er. BTM is the book value to market value of equity calculated one year prior to the o er. VC is the change in trading volume calculated using a 250 day window before the announcement date and after the o er date where volume is measured as the number of shares traded daily as a proportion of the total number of shares outstanding on that day. VC GD is an interactive variables equal to GD times VC. VC %IA is an interactive variable equal to VC times the proportion of the o er issued abroad %IA; R 2 is the adjusted R squared of the regression and N denotes the number of observations in each speci- cation. Heteroscedasticity consistent p-values are in parantheses.

16 1506 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491±1513 meant to control for leverage e ects as explained above. We also check the robustness of these results using monthly beta changes as the dependent variable in speci cations (3) and (4). From Panel A for NYSE/AMEX rms making a global issue, relative issue size is positively while book to market is negatively related to the change in beta. The coe cient of the change in volume is not signi cant although the positive sign indicates that rms experiencing a larger increase in volume experience an increase (smaller decline) in beta. The coe cient of the PCB, PCB is signi cant and greater than one in magnitude for global issues suggesting that the actual change in beta is larger than the predicted change for global issues. For instance the magnitude of the coe cient of PCB suggests that for global issues, the change in beta should be 0:16 ˆ 1:349 0:12. This contrasts with the ndings for the domestic issues in (2) where the coe cient of PCB is less than one in magnitude implying that the change in beta is lower than the predicted change. Using the coe cients again, the change in beta for domestic issues is (using the regression coe cients) 0:05 ˆ 0:787 0:060. The same pattern is observed when changes in monthly betas are used to estimate changes in systematic risk and used as the dependent variable in the regression (speci cations (3) and (4)). The coe cient of PCB for the global issues ˆ 1:196 compared to for domestic issues. We also check the robustness of these results using a pooled regression of the global and domestic issues. These results using monthly beta changes are reported in speci cations (5) and (6). PCB GD is an interactive term, where GD is a global dummy ˆ 1 if the issue is a global issue and 0 otherwise. The coef- cient of this variable captures the di erential e ect between global and domestic issues of the PCB. The coe cient of PCB is while that of the interactive term is and both these are signi cant at 10%. Along with the earlier reported result that the average PCB for global issues is 0.12 while that for domestic issues is 0.06 these results imply that global issues experience a decline in beta that is greater in magnitude than the predicted change (as given by leverage changes). In addition, the magnitude of the decline is signi cantly larger for global compared to domestic issues. In the last column (6) we replace the global dummy with the proportion of the issue sold abroad which is zero for all domestic issues and ranges from a minimum of 11% to a maximum of 50% for NYSE/AMEX rms. The interactive term, PCB%IA is positive and signi cant suggesting that the larger the magnitude of the issue sold abroad the greater is the decline in beta. We also include another interactive term VC GD which represents the di erential e ect of a change in volume for global compared to domestic issuers. Our choice of this variable is motivated by the ndings in Foerster and Karolyi (1998a) who nd that foreign rms listing ADRs in the US market are able to increase returns if they can also shift a greater proportion of trading volume to the US market. This variable, VC%IA is our best proxy for the increase in trading volume attributable to the foreign tranche of the

17 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± o er. 26 Since both VC and %IA are positive, this implies that the greater the proportion of the issue sold abroad, the larger is the decline in beta. 27 Panel B replicates these results for NASDAQ rms. The coe cients of PCB in speci cations (1)±(4) are not signi cant for the global as well as the domestic issues taken separately. One reason for this could be the smaller sample size. The pooled regressions, however, recon rm the ndings for the NYSE rms that global issues experience a larger decline in beta compared to their domestic counterparts. The coe cient of PCB is however negative and signi cant while that of the interactive term PCB GD is positive and signi cant. This suggests that the change in beta for domestic issuers is positive while that for global issuers is negative and the di erence is signi cant. 28 Taken together these ndings suggest that global issues are associated with bigger declines (or smaller increases) in systematic risk after equity is issued, ceteris paribus. Also, the decline is greater even after controlling for leverage (Hamada, 1972), volume, size and time e ects. In addition, the larger the foreign tranche of the o er, the greater is the decline in systematic risk compared to the predicted decline and greater is the shift in trading volume to a foreign market, the greater is the reduction in domestic systematic risk We recognize that in the absence of a knowledge of the foreign market and the actual volume in that market, this measure is a crude proxy for the change in volume driven by the foreign sale of equity. As mentioned above, the actual market of issuance for these issues is not known (we looked through several prospectuses) and in the absence of this information, this is the best proxy we have. We tried other measures such as the raw turnover after the issue times %IA but the variable is not signi cant. Besides, to the extent that only a small proportion of the rms in our sample that issue equity abroad also list their shares abroad, this variable may not represent actual foreign trading volume. 27 While not reported here, qualitatively identical results obtain if our control sample is constructed by rst matching by date of the o er and then by either MVEQ or by book to MVEQ ratios. 28 Similar conclusions obtain when alternative speci cations using other variables such as the actual volume prior to the o er are used in place of the volume change. The coe cient of the interactive term is always positive and signi cant for both the NYSE/AMEX and NASDAQ rms. 29 We also examined (not reported) changes in beta in a more dynamic context where the betas are calculated using a pooled cross-section/time series regression of monthly returns. To do this we stack for each rm, the rm as well as the market return for the entire period under analysis i.e.for each rm we stack the 96 month returns (48 months before and 48 months after the event). The data for all rms in each group (NYSE/AMEX and NASDAQ) is then pooled and we run the following market model regression: R ˆ a b R m C T R m e t ; R ˆ R 1 ; R 2 ; R 3 ;...; R N ) where R N is a column vector that includes 96 monthly returns for each rm. Similarly R m is the market return corresponding to each of the 96 months for each rm while b is the estimate of systematic risk. T is a time index whose value is increasing in event time. More precisely T ˆ 1 for the 48th month before the o er and increases to T ˆ 96 for the 48th month after the o er. We do this for global and domestic issues separately as well as pooled. In all cases, the coe cient c is negative and larger in magnitude for global issues implying that the decline in beta is greater for this group.

18 1508 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± Changes in the cost of capital The evidence presented above suggests that rms choosing to raise equity capital in multiple markets experience a decline in systematic risk with respect to the domestic component. Even if systematic risk with respect to the domestic (US) market declines, for global issues, systematic risk with respect to a foreign index could change post issuance. We investigate such changes using a two-factor model with the domestic (US) and a foreign index as the two factors. Ideally the two factors ought to be the US index and the index of the country in which these shares are issued. To our knowledge, the exact market of issuance of these securities is not publicly available. 30 Hence, we use a broad based foreign index, the Europe, Asia and Far East (EAFE) index that represents most of the major equity markets except the US market. In order to avoid problems related to non-synchronous trading, we use monthly data and the following two-factor model to estimate the two components of systematic risk: E R i ˆR f b i;us E R US R f Š b i;eafe E R EAFE R f Š; 3 where R i is the return on stock i, R f is the risk free rate (30 day US T-Bill rate), R US and R EAFE are the returns on the US market index (CRSP equal-weighted) and the EAFE index, respectively. E denotes the expectation operator. The results of this estimation for global issues are presented in Table 5 for NYSE/ AMEX and NASDAQ rms. For NYSE/AMEX rms, the average beta with respect to the US index declines from 0.94 before the o er to 0.75 after the o er and the decline is signi cant at the 5% level. The median beta declines from 0.66 to On the other hand, the beta estimates with respect to the foreign index, the EAFE index, increase from 0.05 to 0.24 on average for NYSE/ AMEX rms an increase of 0.19 which is also signi cant at the 1% level. For NASDAQ rms, systematic risk with respect to the US index declines from 1.29 to 1.12 and that with respect to the EAFE increases from 0.04 to Neither of these changes is however, statistically signi cant. Using the pre and post-event estimates of systematic risk we next compute for each rm, the cost of capital before and after the issue using (3) above We examined various sources including the prospectuses of these issues. While the prospectus states the proportion of the o er (the number of shares) that is to be issued outside the US, no mention is made of the foreign market. 31 Although these betas are estimated using monthly returns as in Table 3, the estimates di er from those in Table 3. The di erence arises because the estimation in Table 3 uses a one-factor whereas in Table 5 uses a two-factor model. 32 We follow the procedure used in Karolyi (1998) to estimate the cost of capital using a twofactor model.

19 L. Ramchand, P. Sethapakdi / Journal of Banking & Finance 24 (2000) 1491± Table 5 Changes in systematic risk and cost of capital using a two-factor model for global issues ± 1986± 1993 a NYSE/AMEX Change NASDAQ Change Pre-o er Post-o er Pre-o er Post-o er b US b EAFE Cost of capital (% monthly) N a The model used here to compute the cost of capital is: E R i ˆR f b i;us E R US R f Š b i;eafe E R EAFE R f Š Eq: 3 in text where E(R i ) is the cost of capital of the ith rm, E(R US ) is the average return on the CRSP equal weighted index for the time period 1970±1996 and E(R EAFE ) is the average return on the Europe, Asia and Far East (EAFE) index for the time period 1970±1996 and R f is the average rate on the 3 month Treasury bill for the period 1970±1996. b i;us and b i;eafe are the estimates of systematic risk with respect to the US and the EAFE monthly indices. The numbers in the rst row of each cell are means and the second row is the median. Change is the di erence between the pre and the post-o er period values of the variable. indicates signi cance at the 5% level and indicates signi cance at the 10% level. These are calculated using a t-test for di erence of means ( rst row of the cell) and a Wilcoxon signed rank test (second row of the cell). N denotes the number of observations used to calculate the cost of capital in each group. The risk free rate R f is the 30-day US. Treasury bill rate for the period 1970± 1996 which is 6.74% annualized. The returns on the US and the EAFE index used in (3) are geometric average rates for the period 1970±1996. The average monthly rate of return for the CRSP equal-weighted index is % and that for the EAFE (monthly) index is %. The average monthly cost of capital for NYSE/AMEX rms in the pre-event period is 1.20% while that in the post event period is 1.18%, a decline (of 27 basis points when annualized), which is not, however, signi cant. The median cost of capital changes from 1.17% to 0.87%. For NASDAQ rms the cost of capital declines from 1.43% to 1.38% and the median cost of capital from 1.67% to 1.31%. 33 Overall these results suggest that global equity issues by US rms are accompanied by a decline in stock price volatility and systematic risk with respect to the domestic (US) index. The decline in systematic risk is greater for global compared to domestic issues. Furthermore, while systematic risk declines rises with respect to the domestic index, that with respect to a foreign index 33 Karolyi (1998) nds that the cost of capital declines on average by 1.01% for Canadian rms and 2.92% for UK rms listing ADRs in the US.