International Sources of Risk: Evidence from Cross-Border Mergers Richard A. Brealey* Ian A. Cooper** Evi Kaplanis*** London Business School First Version: August 1998 This Revision 18/8/98 * London Business School and Bank of England, Threadneedle Street, London EC2R 8AH ** (corresponding author) London Business School, Sussex Place, Regent’s Park, London NW1 4SA, England, +44-171-262-5050, [email protected] *** London Business School 1 1. Introduction This paper studies the stochastic structure of international equity returns. In particular, we are interested in the fact that some studies have found that share prices of international companies may move more closely with their market of listing than one would expect from the degree of international diversification of their operations (Jacquillat and Solnik (1978), Froot and Dabora (1996)). Other studies find, however, that the degree of international diversification of operations does have a significant effect on the international factor structure of equity returns (Agmon and Lessard (1977), Yang, Wansley and Lane (1985)). Further evidence that the covariance structure of returns is affected by the country of trading is provided by U.S. closed-end funds. Bonser-Neal et al (1990) and Bailey and Lim (1992) have documented a relationship between the returns on closed-end country funds and the market returns of the country in which the funds are traded. Bodurtha, Kim and Lee (1995) also report a positive correlation between changes in the premium over net asset value of country funds traded in the U.S. and the returns on a U.S. market index. Similar results are reported by Bekaert and Urias (1996) and Hardouvelis, La Porta and Wizman (1993). Evidence that the location of listing has an influence is also apparent for firms with foreign listings. Several studies of foreign listings have looked at the effect on the variance of returns, but the only analysis to look at the covariance is Urias (1994).1 Urias examines the issuance of ADRs by seven Chilean and six Venezuelan companies. In the case of the Chilean ADRs the stock’s exposure to the U.S. factor increases in six cases and its exposure to the Chile factor decreases in five cases. In the case of the Venezuelan ADRs the exposure to the U.S. factor declines in all cases and the exposure to the Venezuelan factor declines in four out of six cases. Our analysis is similar in motivation to that of Urias. Rather than testing directly whether country factors in stock returns are related to the country of operations or listing, we focus on cases where there is a discrete shift in domicile as a result of a cross-border merger or acquisition.2 We compare the covariance with country factors of the combined firm before and after merger. We hypothesise that the effect of the merger is to bring about a shift in the stockholder clientele and that this will cause the stock to move more closely with the local market than the combination of the two stocks before merger. 2. Cross-border Mergers and Returns Campbell and Mei (1993) show that the covariance of an asset’s returns with the returns on the market portfolio can be broken into three components associated with 1 A number of studies have estimated the wealth effects of foreign listings, with inconclusive results. See, for example, Lee (1991), Howe and Kelm (1987), Lau, Diltz and Apilado (1994), Alexander, Eun and Janakiraman (1988), Jayaraman, Shastri and Tandon (1993), and Sundaram and Logue (1996). 2 Henceforth we use the terms ‘merger’ and ‘acquisition’ interchangeably. 2 changing expectations of future cash flows, real interest rates and future excess returns. They use this decomposition to identify the components of domestic betas. Their empirical estimates lead them to conclude that the covariances of expected excess-returns with the market are typically much larger in absolute value than the covariance of cash flows with the market. Thus, it appears that a high proportion of the covariance of returns is due to common changes in discount rates rather than to the comovement of the cash flows. The Campbell-Mei finding concerns a single equity market. Its implications for the comovement of international asset prices are unclear. For instance, the covariance of the return to an individual equity with a foreign stock index will depend on the extent to which shocks to cash flows, real interest rates, and expected excess returns are correlated between countries. A theoretical model of an international covariance decomposition along the lines of Campbell and Mei requires, therefore, a model of the way that expected returns are related internationally. One extreme would be that international equities are fully integrated and behave essentially as a single market. In this case, we might expect that the component of an equity return that is caused by a shock to its expected excess returns will be independent of the market in which it is listed. It is well known, however, that equity markets show some signs of being not fully integrated. In particular, investor portfolios are not broadly diversified across different national markets, but instead are subject to significant home bias (Cooper and Kaplanis (1986) and (1994), French and Poterba (1991)). So international equity markets may be, at least partially, segmented from each other. Such segmentation may result in expected return factors in different markets that are specific to the market and do not represent global factors. For instance, suppose that investors in one country become more risk averse (perhaps as a result of a reduction in wealth) and demand a higher premium for holding risky assets. This will result in a common decline in the price of locally traded stocks and, depending on the portfolio shifts that are forced on investors in other countries, a smaller adjustment in the prices of internationally traded stocks and of the other countries’ locally traded stocks. Thus changes in risk aversion that are country specific will induce common price adjustments in locally traded stocks that are not highly correlated with changes in expected returns on other markets. If expected excess returns depend thus on the firm’s domicile and Campbell and Mei are correct in their suggestion that a large part of stock comovement is due to shifts in the expected excess return, then country factors in stock returns may depend as much on the firm’s domicile as on the countries in which the firm does business. Consider now the effect of changes in the domicile of a company brought about by an acquisition. The impact of the fact that the merger is cross-border on the value of the target company will be unpredictable. If international capital markets are not fully integrated, there will be an initial adjustment in both portfolio holdings of this stock and its value. This effect could either raise or lower the value of the acquired firm (Black (l974), Cooper and Kaplanis (1997)). If the acquisition is of a company whose shares are traded in a small capital market (‘local market’) by a company from a large capital market (‘international market’), there will, on average, be an initial fall in the 3 required return, and a commensurate rise in value. Conversely, if an internationally traded stock becomes only locally traded, there will be an initial fall in price. Where the stock’s status is unchanged (for example, a locally traded stock remains only locally traded), price may either rise or fall. In contrast, the impact of the cross-border nature of the merger on the stochastic structure of returns is more predictable. If markets are integrated, then the only effect should be through the impact on the operations of the merging firms. On the other hand, if the markets are segmented, then we would expect country-specific variations in the expected excess return to cause the stock of the combined firm to comove more closely with the market of the combined firm’s domicile and less with the market of the target firm. 3. Cross-border Mergers A number of studies have examined cross-border mergers, but their focus has been on the consequent wealth changes rather than on the impact on the covariance structure. Kang (1993) examined 102 Japanese acquisitions of U.S. firms between 1975 and 1988 and concluded that they created significant wealth gains for both bidders and targets. Conn and Connell (1990) examined a sample of cross-border mergers involving U.S. and U.K. firms 1971-1980 and reported abnormal returns of about 20 percent for the target firms and about 2 percent for the acquirers. Several other studies have compared the gains to the target companies from foreign and domestic mergers. For example, Harris and Ravenscraft (1991) examined a sample of 1114 U.S. domestic mergers and 159 cross-border mergers between 1970 and 1987. They concluded that foreign acquisitions of U.S. firms resulted in a 13.4 percent higher abnormal return for the target firm. Swenson (1993) in a study of 477 domestic acquisitions and 226 foreign acquisitions found that foreign acquisitions provided a 10.8 percent higher abnormal return than the domestic equivalents. Dewenter (1995) compared 90 foreign acquisitions of U.S. firms in the chemical and retail industries 1979-1989 with 294 domestic acquisitions. In both cases the mean abnormal return to the target shareholders was about 20 percent, but there was no significant difference between the two samples. Doukas (1995) and Doukas and Travlos (1988) have examined returns to bidders in cross-border acquisitions. The former looked at 463 international acquisitions by U.S. firms between 1975 and 1989 and found small positive abnormal returns to the bidder, while the latter found no evidence of abnormal returns. Finally, Fatemi and Furtado (1988) looked at a sample of 117 U.S. bidders 1974-1979 and found negligible abnormal returns. 4. Covariances If a cross-border merger simply involves a combination of the two firm’s cash flows and the rates at which these flows are discounted are unchanged, then the covariance structure of the stock returns should be unaffected. In other words, the sensitivity of the returns of the merged firm’s stock to any factors, including national market returns, should be the same as that of a portfolio of the stocks of the two separate firms. 4 There are two principal reasons that the factor loadings may change as a result of the merger. First, cross-border mergers may affect the aggregate cash flow stream to the equity holders. For example, the acquirer may divert production to its home country or to the country of the target; there may be synergies that are realized by the combined entity; the merger may be accompanied by a change in the currency of the firm’s debt. Such moves could be expected to induce a change in the exposure of the stock of the combined firm to country factors, though the sign of the changes in the stock’s exposure to these factors is ambiguous. Second, if international equity markets are segmented and there are country factors in expected returns, then the merger will result in a shift in investor clienteles and thereby unambiguously cause the stock of the merged firm to move more closely with the returns of the acquirer’s national market. 5. Hypotheses The month of the merger is defined in terms of its announcement month Ta and the completion month Tc. All variables are ordered in time. For economy in notation we omit the time subscript. We define variables as follows: RA returns to acquirer RT returns to target RM post-merger returns to the pooled acquirer and target We also define the derived variable: RS = A RSA + (1-A )RST where A is the market value at Tc of the acquirer’s stock as a proportion of the aggregate value of the acquirer and target. Thus the variable RS is the return to the combination of the acquirer and the target before the acquisition occurs. The superscript ‘S’ refers to the fact that the variable refers to the period when the two companies are separate. We define the market environment by a set of variables: RMA returns to the market of the acquirer RMT returns to the market of the target RW returns to the world market We again use the superscripts ‘S’ and ‘M’ to denote whether the returns refer to the period before the merger (‘S’) or after the merger(‘M’). 5 We define F as a vector of factors and as a vector of factor loadings. Prior to the merger the factor loadings are given by the regression: RS = S.F + eS The return variable in this regression is the combination of the returns to the two separate firms prior to the merger. After the merger the loadings are given by the post-merger regression: RM = M.F + eM Here the return variable is for the merged firm. The hypotheses we wish to test concern the factor loadings, . Hypothesis 0: Neutrality The first proposition that we wish to test is that the stochastic structure of returns is unaffected by the acquisition. This corresponds to the view that the acquisition does no more than pool the stochastic properties of the two separate companies, without altering them in any way: H0 : M = S If neutrality is rejected, there is a variety of alternate hypotheses that could be tested. Some concern the factor loadings with respect to currency variables. For instance, we could test whether hedging lowers the exposure with respect to the currency of the target firm’s home currency. The focus of this paper is, however, on the factor loadings with respect to the home stock markets of the acquirer and the target. The alternate hypotheses discussed below concentrate on this dimension of the impact of the merger on risk. Hypothesis 1: Operational Effects The most obvious alternative to the neutrality hypothesis is that the operational risks of the target, the acquirer, or both are affected by the acquisition. If the future operating cash flows of either the target or the acquirer are expected to change, then the stochastic behaviour of the present value of these flows will change. Some possible sources of this impact are: The acquirer is expected to make efficiency gains in the acquiree The acquisition signals the intent of the acquirer to expand overseas There are synergies to the acquisition that reduce common fixed costs The acquisition signals increased growth opportunities in general for the acquirer 6 If the gains are efficiency gains to the target, or if the acquisition signals the intent of the acquirer to make more acquisitions, then we would expect the beta in the target country to rise and the beta with respect to the acquirer market to fall: H1A(increased foreign activity): TM TS and AM AS If there are reduced common costs to the merged firm, then we would expect the betas with respect to both the acquirer’s and the target’s home markets to fall as operating leverage is reduced: H1B(cost cutting): TM TS and AM AS If the acquisition signals increased growth opportunities in general for the acquirer, then it should increase the betas with respect to both markets: H1C(increased growth opportunities): TM TS and AM AS Hypothesis 2: Purchaser Domicile Matters While operating effects are the most obvious source of change in betas, the literature on location of trade effects means that there is another hypothesis. This is that the result of the merger will be that the beta with respect to the market of the acquirer will rise and the beta with respect to the target will fall because the trading of the merged entity will be in the acquirer’s home market: H2: TM TS; AM AS An operating effect that might also generate this pattern of betas would be that the merger signals for the acquirer a substitution of costs in the target market for costs in its home market beyond that included in the purchase of the target. If costs are positively related to the market in which they are generated, then the betas would behave as in H2. We note that the ‘purchaser domicile’ effect may simply be due to this type of operating effect and discuss below how to distinguish between them. Table 1 summarises the four alternate hypotheses in terms of their implications for the betas of the pooled company relative to the markets of the acquirer and the target. The null hypothesis of neutrality implies no change in the betas. Table 1: Implications of the hypotheses for the change in betas. Hypothesis H1A: Inc. foreign activity H1B: Cost cutting H1C: Increased growth H2:Purchaser domicile Beta with respect to Beta with respect to target’s home market acquirer’s home market UP DOWN UP DOWN DOWN DOWN UP UP 7 6. Data IFR Securities Data provided us with a record of all cross-border mergers involving two public companies between 1987 and 1992.3 This amounted to 570 mergers. For each merger we required the country of domicile of the acquirer and target We also required the first announcement date of the merger, the effective date and the proportion of the target owned immediately before the announcement and after the effective date. We used Datastream to extract monthly returns for the two firms prior to the merger and for the combined firm afterwards. Datastream also provided the market capitalisation of their equity on the merger announcement date. There were a number of cases of missing data for target companies and as a result our final sample size consisted of 74 mergers. For each month we also collected from Datastream monthly data for the national stock market index of both the acquirer and target and for the exchange rate between the currencies of the acquirer and the target. For each merger we also compute a “world index” return series, which is an equally weighted average of the returns on N national market indexes. The world index excludes the national market indexes of the acquiring and target firms. The first merger announcement in our sample occurred in February 1987 and the last occurred in October 1992. As Table 2 shows, there is a broad distribution of mergers by year. Table 2 Distribution of mergers by year of announcement No. of mergers Year Working sample 1987 6 1988 22 1989 20 1990 12 1991 8 1992 6 Total 74 The acquiring firms come from a total of 15 countries and the targets come from 10. As Table 3 shows, Anglo-Saxon countries account for a preponderance of the target firms and a somewhat smaller proportion of the acquirers. Other countries such as Japan, Switzerland and Sweden feature heavily as the homes of acquirers but not of targets. 3 The record of cross-border mergers is not fully complete for 1987 and 1992. 8 Table 3 The countries of origin for the acquiring and target firms (number of firms) USA Canada UK Ireland Scandinavia Switzerland France Other Europe Japan Australasia South Africa Acquirers Targets 9 8 32 6 18 5 4 5 6 6 7 6 1 26 0 1 2 0 4 0 3 0 The target firms were much smaller than the acquirers and the mean value of the acquired stock as a proportion of the stock of the acquirer was only 15.3 percent. We discuss below the implications of this for methodology. Monthly returns for each stock were calculated as ln(Pjt + Djt) - ln(Pjt-1), where Pjt is the stock price at the end of month t and Djt is the dividend paid in month t, expressed in terms of the currency of the acquiring firm.4 National and world market returns were also expressed in the currency of the acquiring firm. 7. Methodology We wish to compare the covariance structure of stocks before the merger with the structure after merger. We measure prior returns relative to the merger announcement date Ta. Subsequent returns are measured relative to completion date, Tc. The mean time between the announcement and completion dates is 92 days. Our null hypothesis is that the factor loadings of the combined firm on the national market indexes are unaffected by the merger. The alternative hypothesis which is consistent with segmented capital markets is that the loading on the national market of the acquirer increases while that on the market of the target declines. To test whether the factor loadings change, we form for each merger a capitalizationweighted portfolio of the two stocks and measure its returns RSAT for the months 4 The choice of currency is largely arbitrary. Our results are essentially unaffected if returns are stated in the local currency. 9 January 1983 to Ta - 1.5 The portfolio weights are the market capitalizations at Ta 1of the acquirer’s stock and of the shares in the target firm that are acquired. We wish to compare the factor loadings of this portfolio with the loadings of the merged firm. A simple way to test the effect of each merger on the factor loadings is to undertake an OLS regression of the postmerger returns on the returns of the two national markets, where the variables are ordered by time: RMAT = M0 + MA RMA + MT RMT + eM (1) A similar regression could then be undertaken for the returns of the combination of the two firms before the merger and a comparison made of the estimated regression coefficients. One problem with these regressions is that the returns on different national markets are collinear. We can reduce the problem of colinearity in the premerger period and potentially reduce the estimate errors by imposing the prior that before the merger the returns on the two firms load only on their national markets with a coefficient of unity. In this case the null hypothesis is that the coefficients after the merger, MA and MT, are simply equal to the relative market capitalizations of the two firms. We, therefore, need to estimate equation (1) only for the postmerger period (Ta to Ta+61) and compare the estimated coefficients with their expected values of MA = a, and MT = 1 - a. Our first set of tests are based on equation (1). Since many of the firms that engage in cross-border mergers are multinationals with exposure to a number of national markets, equation (1) is liable to suffer from an omitted variables problem. Therefore, it is helpful to include in our regression the return (rW) on an equally weighted portfolio of the national markets other than those of the acquirer and target. We refer to this portfolio as the “world” portfolio. As noted above, the returns on different national markets are correlated. To reduce the problem of colinearity, we first regress the returns on the acquirer’s national market and the target’s national market on the world market returns, and we calculate the residuals (RMA* and RMT*) from these regressions. We take these residuals as measures of the national market factors, purged of any world market influence. Thus in our second set of tests we estimate the loadings of the combined company on the country and world factors for the premerger months (January 1983 to Ta-1) by the following expanded OLS regression: RSAT = S0 + SA RMA* + ST RMT* + SW RW + eS (2) Similarly, we estimate the loadings of the merged company on the country and world factors by the second OLS regression: RMAT = M0 + MA RMA* + MT RMT* + MW RW + eM (3) 5 For N firms data is available for at least 24 months but less than 60 months. In these cases we used the maximum available data. 10 Under the null hypothesis the loadings on the national market indexes are the same before and after the merger. Under the alternative hypothesis that the domicile of the purchaser matters, we expect the first inequality to show that the beta against the target market falls after the acquisition (MTST) and the second inequality to show that the beta against the acquirer’s market rises (MASA). Where the target is very small relative to the acquirer, the likely change in the regression coefficient is likely to be negligible relative to the estimate error. For this reason, tests based on an unweighted average of the estimated coefficients are liable to be inefficient. After the merger we can observe only the betas of the combined firms, but we would like to be able to break out the betas of the target component. The segmented market hypothesis predicts that the regression coefficients of the acquiring firms are unaffected by the merger. We assume that this is so and estimate the beta of the target component in the merged firm as T = (ATM - aAS)/(1 - a), where the prefix denotes that the regression coefficient refers to the acquirer, the target, or the combined firm. This allows us to derive betas for the target component of the merged firm both against the target’s index and the acquirer’s index. We also compute the error variance of the estimated target betas as V(T) = [V(ATM) + a2 V(AS)]/[1 a]2. Finally, we compute the differences between the premerger target betas and the postmerger target betas and the standard error of these differences. We calculate a weighted mean of these differences, where the weights are the inverse of the standard errors. 8. Results 8.1 Tests on averages For our first test we estimate equation (1) for the post-merger period and compare the estimated coefficients against the relative market capitalizations of the two firms. The results are summarised in Table 4. Table 4 Difference between coefficients from a regression of the returns from the combined firm on the two national market indexes (equation (1)) and the relative market capitalizations of the two firms. Mean difference MT - a -.04 (se = .06) MA - (1 - a) +.12 (se = .05) MT - a MA - (1 - a) % -ve at .05 sig. 12 % +ve at .05 sig. 19 Since the mean value of the target as a proportion of the combined firm was a = .15, under the null hypothesis the expected mean values of MT and MA are respectively .15 and .85. The actual mean of the 74 estimated values for MT is .11, while the mean for MA is .97. These estimates are respectively .62 and 2.35 standard errors from the expected values under the null. In 53 percent of the cases the regression coefficient on the target index was lower than the target’s relative market capitalisation and in 12 percent of the cases it was significantly lower at the .05 11 significance level. In 59 percent of the cases the coefficient on the acquirer’s index was higher than acquirer’s relative market capitalisation and in 19 percent of the cases it was significantly higher. In our second set of tests we estimate equations (2) and (3) and examine the changes in the loadings on the national market factors. We first employ a non-parametric test in which we look at the estimated coefficients on the two home markets of the merging firms and we count the incidence of the inequalities: MTST and MASA Under the null hypothesis we expect each of the four possibilities to have a frequency of one quarter. The results are shown in Table 5. In 39 percent of the cases the regression coefficient of the combined firms’ returns on the market of the acquirer increases while the coefficient on the market of the target declines. The null hypothesis that the observations are equally distributed across the four cells can be rejected at the 5 percent significance level. The segmented market hypothesis implies that there should be an excess of observations in the bottom left-hand cell. On a one-tailed test we can reject the hypothesis that the proportion of observations in this cell is less than or equal to 25 percent at the 1 percent significance level. Table 5 Percentage of firms with a specified change in the regression coefficients on the national market returns (equations (2) and (3)) TST MTST M N = 74 MASA 24 39 H0: 25% of obs. in each cell H0: 25% of obs. in bottom-left cell MASA 16 20 2 = 8.9, df = 3 2 = 7.9, df = 1 In a number of cases the value of the target company is very small relative to the acquiring firm. In this case the changes in the estimated regression coefficients are likely to consist largely of noise. To reduce the weight placed on these small mergers, we arbitrarily weight each occurrence by the relative size of the target and again summed the number of observations in each cell. Table 6 shows that the conclusions are even stronger even though the effective sample size is somewhat reduced. Table 6 Percentage of firms (value weighted) with a specified change in the regression coefficients on the national market returns (equations (2) and (3)) MTST MTST MASA 15 58 H0: 25% of obs. in each cell MASA 9 18 2 = 45, df = 3 12 H0: 25% of obs. in bottom-left cell 2 = 43, df = 1 A few of the mergers involve the purchase of relatively small fractions of the target’s shares. Table 7 shows the results of excluding those mergers where less than 50% of the target’s shares are acquired in the merger transaction. Again, the conclusions are largely unaffected. Table 7 Percentage of firms with a specified change in the regression coefficients on the national market returns (equations (2) and (3)) excluding acquisitions of 50%. MASA 26 42 MTST MTST N = 65 H0: 25% of obs. in each cell H0: 25% of obs. in bottom-left cell MASA 14 18 2 =12, df = 3 2 =10, df = 1 Our second parametric test involves estimating both pre- and postmerger the regression coefficients of the combined firm on a world index and the national indexes (purged of any correlation with the world index (equations 2 and 3). The mean betas against the acquirer’s and target’s markets are shown in Table 8 for both the pre-merger and post-merger periods. Since the target firms are smaller than the acquirers, the returns of the combined firm have a relatively low loading on the target firm’s national market. As predicted, after the merger the mean coefficient of the combined firm on the target market declines, while that on the acquirer’s market rises. The change is both statistically and economically significant. Table 8 Coefficients of the returns from the combined firm on the returns of the world market and of the two national market factors (equations (2) and (3)) MT -ST Mean coefficients ST MT SA MA .21 .09 .76 .93 MA -SA Mean % -ve at Mean change .05 sig. change -.12 15 +.17 (se = .07) (se = .06) % +ve at .05 sig. 16 In our final test using sample averages we assume that the regression coefficients of the acquiring firm are unaffected by the merger and use this assumption to break out the beta of the target component of the merged firm. The results are summarised in Table 9. The estimates for the implied target are weighted by precision, as some have very large standard errors. The loadings on each of the three indexes post-merger are similar to those of the acquirer pre-merger. This implies that the target component has become less sensitive to changes in the target’s market index and more sensitive to changes in the acquirer’s index. Indeed, the combined beta with respect to the 13 target’s market is insignificantly different from zero after the merger, even though it was significant before the merger. Table 9 Mean estimated coefficients of the returns from the separate firms on the returns of the world market and of the two national market factors T .12 (.05) .87 (.08) .21 (.05) .09 (.06) -.18 (.15) Regressor: Acquirer pre-merger Target pre-merger Combined pre-merger Combined post-merger Implied target post-merger* A .86 (.04) .26 (.09) .76 (.04) .93 (.05) .97 (.16) ( ) = se’s W .80 (.05) 1.02 (.05) .81 (.05) .78 (.04) .79 * Post-merger coefficients for acquirer assumed unchanged 8.2 The effect of merger type If the cause of the change in betas is operating effects, we might expect mergers within industries to exhibit different effects to cross-industry mergers. Our sample contains 35 mergers within the same industry (defined by 2-digit SIC code) and 39 mergers across industries. Table 10 shows the result of splitting the sample into these two groups. Table 10 Percentage of firms with a specified change in the regression coefficients on the national market returns. Same industry first, cross industry second. MTST MTST MASA 23,26 37,41 MASA 17,15 23,18 N = 35,39 Table 11 Change in betas for mergers within industries and between industries. (t-values in parentheses) Same industry bT2-bT1 bA2-bA1 Different industry bT2-bT1 bA2-bA1 -0.097 (0.784) 0.185 (2.098) -0.136 (1.63) 0.158 (1.82) The results in Table 10 show that there is no difference between the two groups in the effect of the merger on betas. This is confirmed by the changes in betas shown in 14 Table 11. For both types of merger, the beta with respect to the target country falls and the beta with respect to the acquirer’s country rises. The sizes of the effects are the same for the two types of merger. 8.3 Cross-sectional relationships In an attempt to gain more insight into the structure of the changes in risk, we now estimate cross-sectional relationships in the impact of cross-border merger on risk. We cannot observe the separate betas of the merging companies once they have merged. So separating the changes in betas between impacts on the acquirer, the target, or synergy gains is difficult. Initially, therefore, we estimate the standard predictive beta regression on the pooled betas: Mj = constant + coefficient * Sj + error (j = T,A,W) The coefficients of this regression tell us how t predict the betas of the merged company from the pooled betas of the pre-merger companies. Table 12 shows the results of this regression. The first row shows that there is no persistence of the beta with respect to the target market for the merged company. Furthermore, the constant term with respect to the target market is insignificantly different from zero. The second row shows that there is also little persistence in the beta with respect to the acquirer’s market. So the merger appears to reduce the amount of beta predictability that we would normally expect to see. The constant in the regression with respect to the acquirer’s market is, however, large. So the best prediction of the merged beta with respect to the acquirer’s market is largely independent of the pre-merger betas. The regression of the betas with respect to the world index shows more normal behaviour. There is persistence of beta from premerger to post-merger, although the coefficient is not as high as one would normally see. Table 12: Cross-sectional regressions of post-merger betas on pooled pre-merger betas. Index constant SE coeff. SE R2 obs. T A W 0.09 0.84 0.52 0.07 0.12 0.09 0.04 0.12 0.31 0.14 0.14 0.10 0.00 0.01 0.12 74 74 74 The cross-sectional regressions of pooled betas suggest that the optimal forecast of the post-merger betas with respect to the target and acquirer markets are approximately zero and one respectively. If these effects are primarily driven by changes in operations, we would expect the size of the change in the beta to be related to features of the particular merger such as the size of the synergy, the size of the target, and the difference between target and acquirer premerger betas. To test this, we develop in the Appendix a model of the relationship between premerger and post merger betas. The model is based on three assumptions: 15 A1: There is no change in the betas of the acquirer’s existing business A2: The betas of the target company’s business are affected by the merger. The size of the effect is . So the target company’s betas move towards the acquirer’s betas by the factor . A3: The betas of the synergy are weighted averages of the betas of the acquirer and the target. The weight of the acquirer is . If is zero, the synergy beta is the value-weighted beta of the companies. If is one, the synergy beta is the beta of the acquirer. So measures the extent to which the synergy beta is more heavily influenced by the beta of the acquirer. We estimate the parameters by taking the residuals from the cross-sectional regression of the pooled betas and regressing them on two variables that reflect the impact of the acquirer on the target’s beta and the synergy beta: e ji constant j X 1 ji j X 2 ji u ji where: eji is the residual from the cross-sectional beta regression X1 ji Ti ( Aji Tji ) for firm i X 2 ji Si ( jiA (ai jiA (1 ai ) jiT )) for firm i. uji is an error term The variable X1 is the proportion of the value of the merged entity represented by the target multiplied by the difference between the target beta and the acquirer beta. X2 is the proportion of the value of the merged entity represented by synergy multiplied by the difference between the acquirer’s beta and the pooled beta of the two companies. j then measure the degree to which the target beta and The coefficients j and synergy beta with respect to index j are pulled towards the acquirer’s beta. The correlations between X1 and X2 are -0.38 for betas with respect to the target market and 0.33 for betas with respect to the acquirer’s market, so they are measuring different effects. Table 13 shows the result of running this cross-sectional regression. In Panel A the results for the betas with respect to the target’s market are given. None of the coefficients, including the constant, are significant. This suggests that the best conditional forecast of the beta with respect to the target’s market is that the beta after the merger is still approximately zero, even once the particular features of the merger are allowed for. In Panel B the results for the betas with respect to the acquirer’s market are given. Once again, all the coefficients in this cross-sectional regressions are insignificant, implying that there is no relationship between the change in betas and the characteristics of the merger. So the impact of the merger on betas does not appear to have the characteristics that one might expect if it were the result of operating changes. 16 Table 13: Cross-sectional estimation of the Risk effect. Panel A: With respect to target market Coefficient 0.004 0.022 -1.143 Constant SE 0.063 0.266 1.285 2 R 2 = 0.01; R = - 0.01; Observations = 74 Panel B: With respect to acquirer market Coefficient -0.011 0.194 -2.626 Constant SE 0.060 0.367 2.090 2 R 2 = 0.02; R - 0.01; Observations = 74 9. Abnormal Announcement Returns Most previous studies of the wealth effects of cross-border mergers have been limited by data availability to acquisitions by, or of, U.S. companies. Thus, while our main focus in this paper is on the covariance structure of the merging firms, our sample also allows us to provide some further evidence on the abnormal returns in the months surrounding the announcement date of a cross-border merger. These abnormal returns are shown in Table 14. The abnormal returns for the target and acquirer are broadly consistent with other studies of both domestic and cross-border mergers. The target shareholders realise high percentage returns, while the acquirer receives a statistically insignificant percentage gain. While the target is typically much smaller, even in money terms the target appears to realise almost six times the gain of the acquirer. Table 14 Mean cumulative abnormal returns* over a 5-day period centered on the merger announcement date Target Acquirer Combined Merger gain as % of target Abnormal return % +24.5 (4.0) +.9 (2.0) +3.6 (2.0) +26.4 (8.2) ( ) = se’s * Defined as return in excess of the national index return 17 Although studies of merger gains commonly report simple means of abnormal returns to acquirer and target, it seems highly likely that the impact of an acquisition on the acquirer is a function of the size of the acquisition. In this case averaging the gains to the acquirer across large and negligibly small acquisitions will be sample-dependent and misleading. We therefore calculate the aggregate money abnormal return to the target and acquirer and express this as a proportion of the value of the target. There appears to be little or no relation between this proportionate gain from merger and the merger’s relative size (Spearman’s rank correlation is .06) However, if the acquirer is relatively large, the sampling error in measuring the acquirer’s abnormal return will show up as a large error in our measure of the proportionate gain. To estimate the standard error of the proportionate gain we divide our sample into 4 size groups and calculate the standard error of the mean proportionate gain for each size group. The standard error declines monotically from 31.8 percent for the smallest acquisitions to .09 percent for the largest acquisitions. To calculate the grand mean of the proportionate gains, we weight the means for the different size buckets by the inverse of their error variances. Thus we estimate that on average cross-border acquisitions result in a total gain amounting to 26.5 percent of the value of the acquired firm. 10. Summary and Conclusions In this paper we have examined the effect of cross-border mergers on the covariance structure of the stock returns. The principal finding is that the effect of the merger is to cause the returns to load more heavily on the market of the acquiring firm and less heavily on the market of the target. While our sample is relatively small, the effect is sufficiently strong that the result is highly significant. It is also robust to changes in the way that we measure the influence of the national market factors. There are two reasons that one might expect a cross-border merger to change the covariance structure. One is that the market expects changes in the structure of the underlying cash flows. Since, for example, production could be shifted to or from the target’s country, the effect of operational changes on the covariance structure is ambiguous. Moreover, we might expect operational effects to be more marked in the case of within-industry mergers and we might expect the change in the loadings on the two market factors to be a function of the loadings for the two firms before the merger. The data do not support either conjecture. The alternative explanation, which we believe is more consistent with the data, is that the covariance structure is a function of the country in which the firm’s stock is traded and thus the change in domicile that results from the merger also changes the covariances. The consequence is that the structure after the merger is effectively independent of the loadings of the target company on the national market factors. This apparent importance of firm domicile suggests (a) that the comovement between stocks may be driven as much by common changes in the discount rate as by common influences on the cash flows, and (b) that international capital markets are segmented so that discount rates in different countries are imperfectly linked. 18 References Adler, M. and Dumas, B., “International portfolio choice and corporation finance: a synthesis,” Journal of Finance, 38, 925-984 (1983). Agmon, T. and Lessard, D.R., “Investor Recognition of Corporate International Diversification,” Journal of Finance, 32: 1049-1055 (Sep 1977). Alexander, G.J., Eun, C.S. and Janakiramanan, S., “International Listings and Stock Returns: Some Empirical Evidence,” Journal of Financial and Quantitative Analysis, 23: 135-152 (June 1988). Bailey, W. and Lim, J., “Evaluating the Diversification Benefits of the New Country Funds,” Journal of Portfolio Management, 74-80 (Spring 1992). Bekaert, G. and Urias, M.S., “Diversification, Integration and Emerging Market Closed-End Funds,” Journal of Finance, 51: 835 - 869 (Jul 1996). Black, F., “International Capital Market Equilibrium with Investment Barriers,” Journal of Financial Economics, 1: 337- 352 (1974). Bodurtha, J.N., Kim, D. and Lee, C.M.C., “Closed-end Country Funds and U.S. Market Sentiment,” Review of Financial Studies, 8: 879-918 (1995). Bonser-Neal, C., Brauer, G., Neal, R., and Wheatley, S., “International Investment Restrictions and Closed-End Country Fund Prices,” Journal of Finance, 45: 523-548 (1990). Campbell, J.Y. and Mei, J., “Where Do Betas Come From? Asset Price Dynamics and the Sources of Systematic Risk,” Review of Financial Studies, 6: 567-592 (1993). Conn, R.L. and Connell, F., “International Mergers: Returns to U.S. and British Firms,” Journal of Business Finance and Accounting, 17: 689-711 (Winter 1990). Cooper, I.A. and Kaplanis, E., “Costs of crossborder investment and international equity market equilibrium,” in Jeremy Edwards (ed.), Recent advances in corporate finance, Cambridge University Press, Cambridge (1986). Cooper, I.A. and Kaplanis, E., “Home bias in equity portfolios, inflation hedging, and international capital market equilibrium,” Review of Financial Studies, 7: 45-60 (1994). Cooper, I.A. and Kaplanis, E., “Partially segmented international capital markets and international capital budgeting,” working paper, London Business School, (1997). Dewenter, K.L., “Does the Market React Differently to Domestic and Foreign Takeover Announcements? Evidence from the U.S. Chemical and Retail Industries,” Journal of Financial Economics, 37: 421-441 (1995). 19 Doukas, J., “Overinvestment, Tobin’s q and Gains for Foreign Acquisitions,” Journal of Banking and Finance, 19: 1285-1303 (1995). Doukas, J. and Travlos, N., “The Effect of Corporate Multinationalism on Shareholder Wealth: Evidence from International Acquisitions,” Journal of Finance, 43: 1161-1175 (1988). Eun, C.S. and Janakiramanan, S., “A Model of International Asset Pricing with a Constraint on the Foreign Equity Ownership,” Journal of Finance, 41: 897-914 (Sep 1986). Fatemi, A.M. and Furtado, E.P.H., “An Empirical Investigation of the Wealth Effects of Foreign Acquisitions,” in Khoury, S.J. and Ghosh, A. (eds.), Recent Developments in International banking and Finance, Vol 2, Lexington Books, Lexington, Mass., 1988. French, K. and Poterba, J., “Investor diversification and international equity markets,” American Economic Review 81: 222-226 (1991). Froot, K.A. and Dabora, E., “How are Stock Prices Affected by the Location of Trade?,” unpublished paper, Harvard University, 1996. Hardouvelis, G., La Porta R. and Wizman, T., “What Moves the Discount on Country Equity Funds,” NBER working paper No. 4571, 1993. Harris, R.S. and Ravenscraft, D., “The Role of Acquisitions in Foreign Direct Investment: Evidence from the U.S. Stock Market,” Journal of Finance, 46: 825-844 (1991). Howe, J.S. and Kelm, K., “The Stock Price Impacts of Overseas Listings,” Financial Management, 16: 51-56 (Autumn 1987). Jacquillat, B. and Solnik, B.H., “Multinationals are Poor Tools for Diversification,” Journal of Portfolio Management, 8-12 (Winter 1978). Jayaraman N., Shastri K. and Tandon K., “The Impact of International Cross Listings on Risk and Return: The Evidence from American Depository Receipts,” Journal of Banking and Finance, 2: 139-154 (1992). Kang, J-K., “The International Market for Corporate Control: Mergers and Acquisitions of U.S. Firms by Japanese Firms,” Journal of Financial Economics, 34: 345-371 (1993). Lau, S.T., Diltz, J.D. and Apilado, V.P., “Valuation Effects of International Stock Exchange Listings,” Journal of Banking and Finance,18: 743-755 (1994). 20 Lee, I., “The Impact of Overseas Listings on Shareholder Wealth: The Case of the London and Toronto Stock Exchanges,” Journal of Business Finance and Accounting, 18: 583-592 (1991). Sundaram, A.K. and Logue, D.E., “Valuation Effects of Foreign Company Listings on U.S. Exchanges,” Journal of International Business Studies, 27: 67-88 (First Quarter 1996). Swenson, D., “Foreign Mergers and Acquisitions in the United States,” in K.A. Froot (ed.), Foreign Direct Investment, University of Chicago Press, Chicago, Ill., 1993. Urias, M.S., “The Impact of Security Cross-Listing on the Cost of Capital in Emerging Markets,” unpublished paper, Stanford University, November 1994. Yang, H.C., Wansley, J.W. and Lane, W.R., “Stock Market Recognition of Multinationality of a Firm and International Events,” Journal of Business Finance and Accounting, 263-274 (1985). 21 Appendix: Cross-sectional estimation of the risk effect. In this section we present a model of the effect of cross-border merger on risk. We use the following notation: VA VT VM VS Value of acquirer ex merger prospects Value of target ex merger prospects Post-merger value (adjusted for market movements) Value of synergy The synergy value is defined by: VS VM VT VA We define the value proportions of the merged firm by: A VA / VM ; T VT / VM ; S VS / VM and the value proportions of the unmerged firm by: a = VA/(VA+VT) The betas we observe are: Before the merger: Aj beta of VA with respect to market j Tj beta of VT with respect to market j Market j can be the world portfolio, the acquirer’s market or the target’s market. After the merger: M j betas of merged value with respect to markets j. We do not observe the separate betas of the acquirer, target and synergy after the merger. These are denoted by: Aj , Tj , Sj , To generate parametric predictions about the behaviour of the post-merger betas of the merged firm, we need a model about its constituent betas. We assume the following: A1: There is no change in the betas of the acquirer’s existing business 22 Aj Aj A2: The betas of the target company’s business are affected by the merger. The size of the effect is . So the target company’s betas move towards the acquirer’s betas by the factor : Tj (1 ) Tj Aj Tj Aj Tj So measures the extent to which the post-merger beta of the target is moved towards the acquirer beta. A3: The betas of the synergy are weighted averages of the betas of the acquirer and the target: Sj (1 )[(1 a) jT a jA ] jA (1 )(1 a) jT (1 )a jA jA If is zero, the synergy beta is the value-weighted beta of the companies. If is one, the synergy beta is the beta of the acquirer. So measures the extent to which the synergy beta is more heavily influenced by the beta of the acquirer. The betas we observe after the merger are those for the merged entity: M j A Aj T Tj S Sj Substituting the expressions for Aj , Tj , Sj gives: in the expression for M j ( (a jM A S A j A j T T j A j S (a jA (1 a ) jT ) T ( jA jT ) (1 a ) jT )) The first term in this expression is the beta for the merged entity if the merger has no impact on operations (the ‘neutral’ beta). The second and third terms measure the regression of the target beta to the acquirer and the third is the impact on the beta of synergy. The terms inside the square brackets are observable. We estimate the parameters by a two-stage procedure. First, to allow for the fact that the betas are estimated with error, so we cannot directly observe the ‘neutral’ beta, we run a standard cross-sectional predictive beta regression: jM constant + j jiM e ji 23 where: M ji is the observed merged beta for firm i against index j jiM Ai jiA Ti jiT Si (ai jiA (1 ai ) jiT ) for firm i eji is an error term We interpret the predictive part of this equation as our estimate of the ‘neutral’ beta. We then estimate j and j by running the regression: e ji constant j X 1 ji j X 2 ji u ji where: X1 ji Ti ( Aji Tji ) for firm i X 2 ji Si ( jiA (ai jiA (1 ai ) jiT )) for firm i. uji is an error term j then measure the degree to which the target beta and The coefficients j and synergy beta with respect to index j are pulled towards the acquirer’s beta. 24 Other studies of the wealth effect of cross-border mergers include Parhizgari, A.M. and De Boyrie, M.E. (1995). [**Check Chang et al (1993) and Diwan et al (1993)] [**Urias preliminary. Cite only if published] Unfortunately, while we can keep the weights of the portfolio constant before the merger, we cannot do so afterwards, for, once the merger has been completed, the relative values of the two components are unobservable but not constant. If, for example, there was a systematic increase in the value of firms located in the acquirer’s country relative to that of firms in the target’s country, then the loading of the merged firm’s stock on the acquirer’s local market would be expected to increase. To control for such effects, we multiplied the pre-merger market capitalization of each firms’ stock by the cumulative returns of its national market from Ta - 1 to Tc + 61. We then recalculated the weights a and (1 - a) using these estimates of the market capitalization of the two components at Tc + 61. In practice the choice of weights made a negligible difference to our findings and we report results only for the weights calculated from the premerger values. Brewer, H.L., “Investor Benefits from Corporate International Diversification,” Journal of Financial and Quantitative Analysis, 16: 113-126 (Mar 1981). ** not referred to Chang, E., Eun, C. and Kodolny, R., “International Diversification through ClosedEnd Country Funds,” working paper, University of Maryland, 1993. Diwan, I., Errunza, V. and Senbet, L., “National Index Funds: Perspectives,” working paper, World Bank, 1993. Empirical Fatemi, A.M., “Shareholders’ Benefits from Corporate International Diversification,” Journal of Finance, 39: 1325-1344 (Dec 1984). Kaplanis, E. and Schaefer, S.M., “! Loderer, C. and Jacobs, A., “The Nestle Crash,” Journal of Financial Economics, 37: 315-339 (1995). Morck, R. and Yeung, B., “Internationalization: An Event Study Test,” Journal of International Economics, 33: 41-56 (1992). **not referred to Parhizgari, A.M. and De Boyrie, M.E., “Return to Shareholders of U.S. targeted Companies Acquired by Foreign Corporations,” Applied Financial Economics, 5: 265-272 (1995). **not referred to Senschak, A. and Beedles, W.L., “Is Indirect Diversification Desirable,” Journal of Portfolio Management, 49-57 (1980). **not referred to 25 Servaes, H. and Zenner, M., “Taxes and the Returns to Foreign Acquisitions in the United States,” Financial Management, 23: 42-56 (1994). **not referred to Solnik, B., Bourcelle, C. and Le Fur, Y., “International Market Correlation and Volatility,” Financial Analysts Journal, 52: 17-34 (September/October 1996). 6intmerg 26
© Copyright 2025 Paperzz