The Method of Payment Decision in Australian Takeovers: An Investigation of Causes and Effects.Keywords: TAKEOVERS; METHOD OF PAYMENT; LONG-TERM PERFORMANCE; SURVIVAL BIAS; FIRM SIZE; FREE CASH FLOW. 1. Introduction In perfect capital markets, with symmetric information, no transaction costs, and no taxes, the method a firm proposes to finance its acquisitions is inconsequential. Practically however, the `method of payment' in takeover bids is usually conditioned by exactly these factors. A substantial body of literature suggests that the method of payment decision has signalling properties or transaction cost/tax implications that affect the share prices of bidding and target firms. This paper extends existing Australian literature (Bellamy & Lewin 1992; Bugeja & Walter 1995) on the shareholder wealth effects of acquisition financing. We address three questions: 1. Is there a robust empirical association between cash bids and higher abnormal return to bidder and target firms over the bid announcement period? 2. Do target firms subject to cash bids earn higher abnormal returns in an environment where capital gains tax liability is independent of method of payment? 3. Do the overseas findings that long-term post-bid performance is systematically predictable by the method of payment apply in Australia? 4. Do firm characteristics such as capital structure and free cash-flow influence the method of payment decision? Takeovers are frequent and economically significant events and so the practical importance of the questions addressed needs no emphasis. Section 2 places the questions in context by relating them to four salient findings in the relevant literature: i. most, but not all, research indicates that bidder and target firms earn higher abnormal returns over the bid announcement period when cash is offered as consideration; ii. the evidence whether tax considerations or information signalling related reasons are principally responsible for the higher returns in cash bids remains equivocal. The Australian market for corporate control facilitates a purer test of the information hypothesis because the tax liability incurred by target shareholders is independent of the method of payment; iii. the decision to make a purely cash or share offer is not independent of value-relevant characteristics of the bidding firm and its target (e.g. capital structure); and iv. the relative performance of acquiring firms in the post-bid long-term is systematically related to method of payment. This last finding presents a challenge to the efficient market hypothesis (EMH). Although this paper does not address the challenge, the evidence suggests it is causally connected to the factors that underpin similar anomalies (e.g. the long-term performance of firms that make initial public offerings). Data collection, descriptive statistics and research methods are detailed in section 3. Section 4 reports and interprets results. Conclusions and suggestions for future research comprise section 5. 2. Theory and Evidence Early research on method of payment focused on investigating the tax- and information asymmetry-based arguments for expecting the method of payment to have a significant wealth effect. The weight of research on the tax and information asymmetry arguments makes it convenient to review them first. 2.1 The Tax Explanation for Impact of Method of Payment Tax-related explanations of acquisition financing hinge on distinctions in the taxable status of takeover proceeds. For instance, the United States' Internal Revenue Code deems that acquisitions in which the stock of the acquiring company constitutes at least 50% of the total consideration have retained a `continuity of [ownership] interest'. The proceeds from these mergers are non-taxable, at least until the stock is subsequently sold. This suggests that bidding firms must offer a higher takeover premium in cash offers to compensate for the tax disadvantage to the target shareholders (Carleton, Guilkey, Harris & Stewart 1983; Wansley, Lane & Yang 1983; Brown & Ryngaert 1991).(1) 2.2 Information-Signalling Explanations for Impact of Method of Payment Notwithstanding the higher premium required in cash bids for tax reasons, share offers might not be preferred because of information asymmetry between bidder firm managers and outside investors. The premise of information asymmetry raises the proposition that managers with private information that their firm's shares are over-valued offer these shares as consideration in takeover bids. Outside investors, recognising the adverse selection problem, consequently revise their estimate of the offerer's value downwards. The target's shareholders also demand a higher premium to compensate for the `lemons' problem in share-based bids (Leland & Pyle 1977; Myers & Majluf 1984). Hansen (1987) and Fishman (1989) enrich the information-signalling model by considering the case where the target firm's managers are better informed about their firm's value. Hansen (1987) observes that allowing both bidders and targets to have private information sets up a `double lemons' problem; bidders do not offer shares when they believe their shares are undervalued and targets only accept cash when their share value (based on their private information) is less than the offer. The double lemons problem stems from both parties recognising the adverse selection bias in the other's decision. In equilibrium, there is a trade-off between medium of exchange and offer size, mediated by size of target relative to bidder and capital structure. Nonetheless, Hansen's (1987) model predicts that cash offers always send a credible signal that the bidder's shares are undervalued. In Fishman's (1989) analysis, bidding firms decide between cash and share offers on the basis of their private (and costly) information about the value of the merger. Bidders who estimate a high value make high preemptive cash bids to deter potential competing bidders. This tactic assumes competing bidders' expected payoffs are decreasing in the initial bidder's valuation of the target. However, targets with private information about their own value make offering cash risky for the bidders because of the adverse selection problem. In sum, a cash offer has the advantage of preempting potential competing bidders; the advantage of a share offer is that it induces the target to make an efficient accept/reject decision and thereby reveal its private information about expected future cash flows. Fishman's analysis predicts that an initial bidder's expected pay-off is higher if cash is offered rather than shares. The tax- and information asymmetry-based explanations of the wealth impact of method of payment are not mutually exclusive. However, their relative impact remains moot as the following review of the evidence indicates. 2.3 Evidence Relating to Target Firms In an early study, Wansley, Lane and Yang (1983) analyse returns to U.S. acquired firms involved in 102 cash offers and 87 share offers over the period January 1970 to December 1978. Over the period [-40,0] days relative to the bid announcement date, firms offered cash earned a significant average cumulative abnormal return (CAR) of 33.54% while firms offered shares earned 17.47%. Other studies report similar results. For instance, Travlos (1987) finds that U.S. target firms involved in 60 share offers over the period 1972-1982 earned a significant average CAR of 12.04% over the period [-2,0] days relative to the bid announcement date. Target firms in 100 cash bids earned 17.06% over the same event-window. An issue in assessing the information content of the medium of exchange is that the market reaction to cash and share offers might not be independent of the degree of management resistance. Entrenched target managers who derive private benefits from control are less likely to value share offers unbiasedly in the interests of their firms' shareholders and will resist offers more vigorously. In such cases, bidders offer cash to minimise target shareholders' valuation cost. The higher premium in cash offers may simply reflect that firms with entrenched managers provide greater opportunities for gain in a takeover. Huang and Walkling (1987) control for type of acquisition and for degree of managerial resistance and find that the average CAR for cash offers is 29.3% compared with 14.4% for share offers. For mixed offers, the average CAR is 23.3%. In short, Huang and Walkling's (1987) results are consistent with the proposition that cash bids are made for higher valued targets per se. The evidence reviewed so far does not indicate whether tax considerations are more important than information signalling effects because both explanations predict higher returns to target shareholders in cash bids. However, Eckbo and Langohr's (1989) analysis of, inter alia, 35 cash-based bids and 34 share-based bids made in France between 1972 and 1982 indicates that information signalling is more influential. Eckbo and Langohr find that the average (median) bid premium over the target's pre-offer price is 17.2% (19.0%) in the share-based offers, compared with 73.3% (59.0%) for the cash-based offers. Eckbo and Langohr discount a tax-based explanation for the higher bid premium to cash offer targets even though France has a capital gains tax regime that favours share offers. They do so because the average premium over the post-expiration price is virtually identical across their sample of cash and share-based offers: 22.5 and 23.7% (medians, 20.0 and 16.1) respectively. If firms making cash bids have to pay extra to compensate for the unfavourable tax treatment of cash offers the premium over the post-expiration price would be higher for cash-based bids. Davidson and Cheng (1997) report findings inconsistent with the information signalling hypotheses in their study of 219 U.S.-based takeovers made between 1981 and 1987. Davidson and Cheng agree that a tax-derived preference for share compensation entails higher bid premia in cash acquisitions. However, they argue that payment method should remain a significant return determinant after controlling for the bid premium (i.e. the difference between the pre-bid price and the price offered by the acquiring firm), if a cash offer reduces the information asymmetry costs borne by target shareholders. In line with other research, Davidson and Cheng report that cash targets receive a 39.93% average bid premium while share targets average 29.25%. However, after controlling for size of target relative to bidder, relatedness of target to bidder, undistributed target cash flows, multiple bids and bid premium, they find that method of payment is unrelated to abnormal return. They conclude that `cash offers contain no additional information, provide no additional value, and do not seem to alleviate the asymmetric information problem any more than offers made with payments of stock' (p. 477). Australian evidence on abnormal return to target firms conditioned by method of payment is limited to Bugeja and Walter (1995). For each of the 78 target firms in their sample that received a takeover bid over the period January 1981 to December 1989, Bugeja and Walter calculate the return over the period [-60,+1] days and subtract the corresponding return to the All Ordinaries Accumulation Index (AOAI). They report that the 55 target firms that received cash bids earned a statistically significant mean return (0.01 level) of 17.18%, the 11 firms that received mixed bids earned an insignificant mean return of 7.73% and the 12 firms in receipt of share bids earned a statistically significant (0.05 level) mean return of 18.33%. These results are consistent with the conclusions favoured by Davidson and Cheng (1997). However, the relatively small number of target firms in Bugeja and Walter's sample that received share bids dilutes confidence that these results are not sample-specific. 2.4 Evidence Relating to Bidding Firms Notwithstanding the mixed evidence relating to target firms, the reported returns to bidder firms have been interpreted as being supportive of the proposition that information signalling is the principal driver of the impact of method of payment. Travlos (1987), in a widely cited study, finds that U.S. acquiring firms involved in 60 common stock offers between 1972 and 1981 earned a significant average CAR of minus 1.47% over the period [-2,0] days relative to the bid announcement date. The 100 bidders in his sample that offered cash did not experience significant abnormal returns, on average. These, results are consistent with the Myers and Majluf (1984) hypothesis that share offers are interpreted as signalling that bidders believe their shares are overvalued.(2) Bellamy and Lewin (1992) obtain similar results to Travlos when reviewing abnormal returns to ASX-listed acquiring firms between 1980 and 1988. On the bid announcement day, bidders in 52 share offers earned a significantly negative average abnormal return of 2.25% while bidders in 81 cash offers earned an insignificantly positive average return of 0.03%.(3) The performance of the cash bidders improved in days [+1] and [+2] when they earned significantly positive abnormal returns of 1.3% and 0.56% respectively. In order to place the Bellamy and Lewin study in context, recall that the Australian sharemarket provides a more favourable environment than the U.S. sharemarket for investigating the information content of method of payment. This is because Australian target shareholders are fully liable for capital gains tax regardless of whether they sell their securities for cash or exchange them for acquirer firm shares. Various exceptions to the rule do exist, most significantly in relation to shares purchased prior to September 20, 1985 and not subject to transfers since that date. However, rollover relief is not available merely because a taxpayer exchanges (i.e. `rolls over') one investment for another.(4) Notwithstanding the above, there are significant caveats on the proposition that information signalling is principally responsible for the higher abnormal return to bidders making cash bids. One caveat is that not all the evidence points in this direction. For instance, Franks, Harris and Mayer (1988) find that although U.S. bidders earn significantly positive CARs in cash-bids and significantly negative CARs in share-based bids, U.K bidding firms do not exhibit significant abnormal performance regardless of method of payment. It is unlikely that Franks et al.'s results reflect the performance of an idiosyncratic sample because their 954 U.K bidders and 850 U.S. bidders comprises a large proportion of the total number of firms who made a takeover bid between 1955 and the mid-eighties in the U.K and the U.S. The Australian evidence also indicates that cash bidders do not always earn higher returns. Bugeja and Walter's (1995) investigation of ASX-listed firms involved in 78 full takeover offers over the period January 1981 to December 1989 yields results that diverge from Bellamy and Lewin's findings. Over the period [-60,+1] days relative to the bid announcement date, 55 bidders who made pure cash offers earned an average CAR of-3.36%, while 12 bidders who made pure share offers earned 4.67%. Bugeja and Walter report the difference is significant at the one per cent level. However, after repeating their analysis over the same event-window employed by Bellamy and Lewin, they find significant positive abnormal returns for cash bidders but all other results are qualitatively unchanged. One reason why cash bidders may not always earn higher returns than share bidders is that share bids yield greater flexibility to managers to exploit growth opportunities, in contrast to debt offers. Martin (1996) groups U.S. listed bidders involved in 846 acquisitions made between 1978 and 1988 into 3 financing categories: those financed solely with common shares (250 samples), those financed solely with cash, or cash plus debt (483 samples), and those financed with both shares and cash (113 samples). Martin uses the market-to-book ratio, growth in sales over the five-year period leading to the acquisition, and positive abnormal return earned over the [-250,-5] days relative to the bid announcement date as his proxies for growth opportunities. He finds that firms with higher growth opportunities are more likely to use shares to finance their acquisitions. 2.5 Potential Biases in Research Design The sensitivity of estimates of abnormal return to seemingly innocuous changes in width of the event-window (Bugeja & Walter 1995) points to the another caveat on the evidence that estimated abnormal shareholder wealth changes reflect the information signalling properties of method of payment: estimates of abnormal return may reflect biased models of expected return. Early evidence of potential biases in estimates of abnormal return is found in Franks et al. (1988) who find that acquiring firms in cash-offers typically perform better than their counter-parts in share offers in the post-merger period. The disconcerting aspect of this result is that it is inconsistent with market efficiency. It is puzzling to observe highly competitive investors who are sophisticated enough to appreciate the signalling implications of the medium of exchange taking months to fully exploit the information in their trades. Our puzzlement would, of course, vanish if errors or biases in research design were responsible for the observed drift in post-bid prices. Loughran and Vijh (1997) review the long-term performance of firms in 534(5) acquisitions between 1970 and 1989 and in doing so address the problems in measuring long-term abnormal performance identified by, inter alia, Blume and Stambaugh (1983), Roll (1983), Kothari and Warner (1997) and Barber, and Lyon (1997). They estimate abnormal return as the difference between five-year holding period returns (i.e. five-year buy-and-hold returns) of sample firms and control firms matched on size and book-to-market value. Their sample firms are split on the basis of mode of acquisition (merger or tender offer) and medium of exchange. Loughran and Vijh report that over the five-year period post-acquisition the share offer mergers earned significantly negative excess returns of 25% and the firms involved in cash tender offers earned significantly positive excess returns of 61.7%. Loughran and Vijh observe that their results `suggest that markets systematically overestimate or underestimate the efficiency gains from acquisitions. They also suggest that markets underreact to information conveyed by whether stock or cash is used to pay for acquisitions' (p. 1767). 2.6 Additional Influences on the Estimates of the Return from Takeover Loughran and Vijh (1997) identify 405 pure share offers in their initial sample of 947 firms delisted from the Center for Research in Security Prices (CRSP) tapes due to an acquisition that took place between 1970 and 1989. Of the 405 pure share offers, only eight were tender offers, 385 were mergers and 12 could not be classified unequivocally. These statistics indicate that method of payment is systematically associated with other offer or firm characteristics and they raise the possibility that the empirical association between medium of exchange and abnormal return might not be entirely connected with either information signalling or tax consequences. Galai and Masulis (1976), for instance, demonstrate that a pure share exchange can decrease the variance of equity returns due to the coinsurance effect and result in a net transfer of wealth to debtholders. This hypothesis generates the same prediction as the information signalling and tax based explanations of the medium of exchange: abnormal returns are higher in cash bids. Jensen's (1986) buttresses the above prediction with his hypothesis that cash takeover bids more closely bond the acquiring firm's managers to shareholders' interests because they reduce the potential for wasteful allocation of free cash. Jensen's bonding hypothesis also applies to firms that raise cash on the debt market to fund takeovers; managers of such firms are motivated to perform as a result of the pressure to meet the firm's debt obligations. In sum, the body of theory allied with Jensen (1986) argues that capital structure matters, i.e. capital structure changes influence investors' perceptions of the wealth consequences of the medium of exchange in takeovers. Travlos and Papaioannou (1991) examine the effect of both method of payment and capital structure changes on the equity value of acquiring firms at the initial announcement of takeover bids. They find that cash offers consistently generate higher abnormal returns than share offers and, after controlling for the method of payment, abnormal returns are affected by perceptions of changes in firms' capital structures. Although Travlos and Papaioannou's results are inconsistent with the predictions of what most would believe to be a valid theory it is worth noting that changes in capital structure around the bid announcement period are not as salient to investors as the method of payment. If, as the evidence indicates, investors take months to respond to highly salient information we may expect the reaction to low profile firm characteristics to be even slower in manifesting. 2.7 Summary of Evidence At this point we may sum the evidence as follows: a. firms involved in cash bids generally attract higher abnormal returns but the evidence on the relative impact of tax considerations and the information content of the medium of exchange remains moot; b. the positive abnormal returns to acquirers in cash bids are not as high in percentage terms as the returns earned by cash offer targets and are not robust across all countries. The Australian evidence also indicates that estimates of the abnormal returns to bidder firms are sensitive to small changes in the event-window around the bid announcement date; c. the relative performance of merged firms in the long-term over the post-acquisition period is empirically predictable by the medium of exchange; firms in cash offers earn higher returns than firms involved in share offers. This result contradicts reasonable expectations of market efficiency but is robust to biases that may have been present in early studies; and d. there is no evidence that changes in capital structure affect firms' equity values around the bid announcement period. 2.8 Implications of Findings for a Research Agenda The findings summarised above make clear that many questions remain unresolved with respect to the medium of exchange in takeovers. The facts that results are not robust across countries and that the evidence from the two Australian studies is divergent suggests that the first step in advancing the Australian literature must be to establish that ASX-listed bidders and target firms earn higher returns in cash bids. This is done in the first part of our analysis described in section 3. In the second part of our investigation, we assess the long-term performance of merged firms in the post-bid period after controlling for, inter alia, return computation biases and survival bias. If market inefficiency in the U.S. is responsible for the apparent delay in the information content of the medium of exchange in takeovers being reflected in share prices it may be that the Australian market is different. Finally, we research the influence of firm characteristics such as bidder leverage, free-cash flow (subject to growth opportunities), and size of target relative to bidder on acquiring firms' decisions to make either a share or cash offer. Although prior research does not provide strong empirical evidence to support a link the strong theoretical reasons to expect a link justify further investigation. 3. Research Method and Data 3.1 The Core Sample The bidder and target firms involved in 240 takeover bids are selected from the Securities Data Company Inc (U.S.) `SDC Platinum' Database (SDC). SDC is a commercial database that includes information on takeover bids for ASX listed firms since 1988. This information includes data on the type and value of consideration offered. The selection criteria were that each takeover bid: a. involve at least one ASX-listed firm; b. aim at majority control by an external corporate entity; c. have information available on SDC on the type and value of consideration offered; and d. offer consideration worth at least A$100,000. Takeover bids made after 31/12/96 are not included. Not all bids result in eventual acquisition of the target by the bidding firm. Table 1 and figure 1 reveal the trends in the relative popularity of each medium of payment over the period of time covered in our study. The data relate only to our experimental sample of 240 bids but they are consistent with those reported by Gresham Partners (1996) who survey the non-mining sector over the period 1990-1996. [Figure 1 ILLUSTRATION OMITTED]
Table 1
Frequency of Method of Payment by Year
Year Cash % Share Stock % Share All
(total bids) Only (cum) Only (cum) Mixed
1988 17 81 2 10 2
(21) (12) (4)
1989 9 75 1 8 2
(12) (18) (6)
1990 12 80 1 7 2
(15) (26) (9)
1991 14 50 7 25 7
(28) (35) (23)
1992 5 50 2 20 3
(10) (39) (28)
1993 18 51 10 29 7
(35) (51) (49)
1994 17 57 7 23 6
(30) (63) (64)
1995 28 70 8 20 4
(40) (82) (81)
1996 27 55 9 18 13
(49) (100) (100)
Total 147 47 46
(240)
Year % Share Cash & A Choice of Cash / Stock
(total bids) (cum) Stock Cash or Stock & Other
Securities
1988 10 1 0 1
(21) (4)
1989 17 0 2 0
(12) (9)
1990 13 2 0 0
(15) (13)
1991 25 3 1 3
(28) (28)
1992 30 1 2 0
(10) (35)
1993 20 2 2 3
(35) (50)
1994 20 2 3 1
(30) (63)
1995 10 2 2 0
(40) (72)
1996 27 4 5 4
(49) (100)
Total 17 17 12
(240)
Note: Figures in parentheses are: (i) for `Year' column: total frequency; and (ii) elsewhere: cumulative percentage. Percentages may not sum to 100 due to rounding. `Cash and stock' combinations (i.e. mixed bids) increased in relative popularity over the seven-year interval, particularly in the form of a `Cash or Stock' option. Between 1991 and 1996, `impure' bids (not 100% cash)constituted 45% of the sample. Conversely, cash bids dropped from an 81% share in 1988 to hover around the 55% mark by 1996, the decline being most noticeable between 1990 and 1991. In 1995, cash bids were responsible for virtually all of the absolute increase (10) in the number of reported bids. 3.2 Research Method and Data: Short-Interval Performance The short-run abnormal performance of our experimental sample firms is estimated each day over the period [-50,+159] trading days relative to the bid announcement date. The All Ordinaries Accumulation Index (AOAI) is used to proxy for expected return (i.e. daily abnormal return is defined as the difference between the return to firm i on day t and the natural log of the contemporaneous AOAI close-to-close price relative).(6) Returns over event-windows spanning more than one day are calculated as follows (illustrated by reference to the seven day period [-4,+2] days relative to the bid announcement date): (1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] where: [CAAR.sub.N,[-4,+2]] = cumulative average abnormal return over N firms and the interval [-4,+2]; and [AAR.sub.N,t] = average abnormal return of the N firms on day t. Single sample t-tests are used to assess significance. Tests of positive or negative mean performance imply a null of zero mean performance. A t-statistic on this basis standardises the resulting [CAAR.sub.N,(-x,+y)] by its interval cross-sectional standard error (calculated across the announcement period). Hence: (2) [Omega]([AAR.sub.N,t]) = [Omega]([AR.sub.i,t])/[square root of N] where: [Omega]([AAR.sub.N,t]) = standard error of [AAR.sub.N,t]; and [Omega]([AR.sub.i,t]) = standard deviation of [AR.sub.i,t] across the sample of N firms. (3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] where: [CAAR.sub.N,[-x,+y]] = cumulative average abnormal return over N firms and the interval (-x,+y); and [[Omega].sup.2]([AAR.sub.N,t]) = squared standard error of [AAR.sub.N,t] The resulting t-statistic is normally distributed N(0,1) with N degrees of freedom. Daily share price and volume data for our experimental samples and daily closing index values of the AOAI are obtained from the `Daily Price Series' maintained by the University of Western Australia. Share price data are available for 180 of the target firms and 176 of the bidding firms in our experimental sample of 240 takeover bids. In the interest of obtaining `cleaner' results firms are deleted from the two samples for the following reasons: a. uncertainty regarding the date of the relevant bid announcement. Three bidder/target pairs are deleted for this reason; b. infrequent trading, defined as an absence of trading for four or more consecutive trading days over the period [-10,+ 10] trading days relative to the bid announcement date.(7) Applying this filter to our sample excludes 30 bidders and 42 targets; and c. presence of potentially confounding events, identified by their proximity to the announcement date and `newsworthiness' (e.g. routine drilling reports or release of quarterly financial statement). Signal G' announcements to the ASX pertaining to the [-15,+5] interval are browsed for evidence of confounding events. The last filter removes 21 bidding and 17 target firms. Many of the exclusions relate to bidder/target pairs associated with an Australian counterbid in response to a previous bid by a non-Australian company, although simultaneous announcements are the most common cause for disqualification. A number of bidders are excluded because they made simultaneous offers as part of a planned `wave' of acquisitions. The final bidding/target samples amount to 122 and 114 firms respectively. 3.3 Research Method and Data: Long-Interval Performance When reviewing long-term performance we calculate buy-and-hold returns over the relevant event-window, in line with Loughran and Vijh (1997). Notwithstanding that buy-and-hold returns more closely approximate the returns to a feasible investment strategy and avoid the measurement bias identified by, inter alia, Blume and Stambaugh (1983) and Roll (1983), they tend to have a highly skewed distribution. Skewed distributions imply that mean measures of performance are unrepresentative. Further, Fama (1998) points out that long-term buy-and-hold returns can bias assessments of relative performance over the whole period because any early exceptional performance will be compounded over the whole period. Finally, Brown and da Silva Rosa (1998), among others, note that buy-and-hold returns preclude the use of reinvestment assumptions to account for firms with missing data towards the end of the period. This imposes a survival selection bias. In order to assess the seriousness of the concerns identified above we report long-run buy-and-hold raw returns (BHRR)(8) over various periods in table 2. All periods are expressed relative to the bid announcement month. The results show that the period [-3,+3] months accounts for almost 40% of the mean 3.25 year BHRR. However, mean results are not necessarily representative due to high positive skewness in the BHRRs distribution. In each interval, the maximum value is a large positive outlier in contrast to the (negative) minimum. Further, at least one quarter of firms with available data fail to earn positive BHRRs over any given period. Extreme kurtosis peaking at a level nearer to the median (which in each case is considerably lower than the mean) is the net result. Table 2 Buy-and-Hold Raw Returns to Acquiring Firms Over Monthly Periods Expressed Relative to the Bid Announcement Month
Variable [-3 to +3] Months [-3 to +6] Months
173 Firms 173 Firms
Mean 0.1904 0.2280
Std. Dev. 0.7904 0.8083
Kurtosis 90.7252 62.2952
Skewness 8.4230 6.5089
Minimum -0.7143 -0.9237
1st Quartile -0.0700 -0.0928
Median 0.0515 0.0934
3rd Quartile 0.2705 0.3424
Maximum 9.0000(a) 8.444(a)
Count 173 173
Variable [-3 to +36] Months [+0 to +24] Months
91 Firms 113 Firms
Mean 0.5254 0.2123
Std. Dev. 1.4443 0.8055
Kurtosis 18.8306 1.9122
Skewness 3.6037 1.3345
Minimum -0.9857 -0.9676
1st Quartile -0.2517 -0.3333
Median 0.2132 -0.0121
3rd Quartile 0.7928 0.5000
Maximum 9.6821(a) 2.9351
Count 91 113
Variable [-3 to +12] Months [-3 to +24] Months
158 Firms 116 Firms
Mean 0.2686 0.3365
Std. Dev. 0.8648 0.9676
Kurtosis 31.9658 3.5009
Skewness 4.7028 1.6531
Minimum -0.9822 -0.9572
1st Quartile -0.0672 -0.3329
Median 0.1164 0.1512
3rd Quartile 0.4469 0.5771
Maximum 7.1111(a) 3.9849(a)
Count 158 116
Variable [+1 2 to +36] Months
91 Firms
Mean 0.2551
Std. Dev. 0.8467
Kurtosis 9.7732
Skewness 2.5567
Minimum -0.9895
1st Quartile -0.1909
Median 0.1213
3rd Quartile 0.5413
Maximum 4.6615(a)
Count 91
Note: (a) = technically an outlier, being outside 3.5 standard deviations of its mean. Table 2 also reveals that the maximum number of bidding firms with available data from our initial sample of firms in 240 bids is 173 and this drops to 91 firms in the two periods that extend to three years after the bid announcement. The high attrition rate clearly raises a concern as to the generalisability of our assessments of long-term performance. Firms with unavailable share price data over the long-term are likely to exhibit a different pattern of returns to firms with available data. The fact that our experimental sample firms are highly conditioned on survival makes it difficult to define an economically meaningful estimate of their abnormal performance. For instance, assume we define the expected return to our bidding firms as the return to a set of control firms that are matched on survival with the bidding firms but otherwise randomly selected. The difference between our bidders' actual return and their expected return provides an estimate of `abnormal' performance but the `look ahead' bias implicit in matching on survival does not permit conversion of the abnormal performance into a dollar value that is attributable to the experimental event. Given the above argument, we eschew deriving single or point estimates of long-term abnormal performance and focus instead on determining if, over a given period, the performance of our sample firms is exceptional after controlling for survival. In other words, we merely test for evidence of abnormal performance but do not attempt to estimate the economic value of the abnormal performance. Brown and da Silva Rosa (1998) discuss this issue in more detail. Our resampling or bootstrapping procedure for assessing the relative performance of our experimental sample firms, described shortly, follows the protocol they adopt when reviewing the long-run performance of acquiring firms. The `bootstrapping' or `resampling' method for detecting abnormal performance in an `experimental' portfolio relies upon the construction of an empirical distribution of return results from control portfolios matched with the experimental portfolio on one or more attributes. We match on survival and on market capitalisation (given the well known systematic association between firm size and equity returns). The method we use to match on the two attributes is revealed in the steps we follow to construct 1,001 control portfolios, formed under the null hypothesis of no association between involvement in a takeover bid and share return. To facilitate exposition of the construction of the 1,001 control portfolios we describe the protocol with reference to an event-window covering [-3,+3] months relative to the bid announcement month (henceforth, all event-windows are expressed relative to the bid announcement date). The 4 steps below are repeated for each experimental sample bidder that has price data available over the period [-3,+3] months: 1. identify the set (LS) of all listed firms that survived over the period [-3,+3] months defined relative to the sample firm's bid announcement month; 2. calculate the size (i.e. price per ordinary share as at the beginning of month [-3] multiplied by number of issued shares) of each firm in set (LS) and identify the size decile of the experimental firm; 3. select a firm from the same size decile as the experimental firm, drawing randomly and sampling with replacement, and allocate it to the first of the 1,001 control portfolios; and 4. repeat step 3 one thousand times, each time moving on to the next control portfolio in the sequence of 1,001 control portfolios. The performance of the experimental sample firms are then compared against the distribution of the performance measures of the 1,001 control portfolios. An important advantage of this method is that significance tests are free of parametric assumptions since there is no reliance on formal test statistics. Our raw price data, adjusted for dividends and changes in the basis of capitalisation, are sourced from the Share Price and Price Relative (SPPR) database, supplemented in cases of missing data by the Statex Database of the ASX. SPPR is compiled by the Centre for Research in Finance (CRIF) at the Australian Graduate School of Management (AGSM). 3.4 Research Method and Data: Modelling the Decision to Offer Cash or Shares We use logistic regression to assess the influence of 5 variables on bidding firms' decision to make either a share or cash offer. The 5 variables are bidder's leverage, bidder's free-cash flow (subject to growth opportunities), bidder's pre-announcement share price standard deviation, and size of target relative to bidder. Importantly, logistic regression does not assume multivariate normality in the independent variables. This is a desirable attribute when the inputs are largely financial ratios (Deakin 1976). The general model of the logistic regression is simplest in its binomial form. When the outcome is constrained to be either 1 or 0, the estimated probability of the event (as opposed to the non-event) is modelled thus: (4) Prob (event) = 1/ 1 + [e.sup.-Zi] (5) [Z.sub.i] = [[Beta].sub.0] + [[Beta].sub.1][X.sub.1] + [[Beta].sub.2][X.sub.2] ... where: [[Beta].sub.0], [X.sub.1]/[X.sub.2]/etc = the constant, and the independent variables, of the linear function Z; and 1/(1 + [e.sup.-Zi] = the logistic (cumulative) distribution function of Prob (event), which can be proven to be unbounded in [Z.sub.i], but bounded 0 [is less than] Prob(event) [is less than] 1. The model is completed by imposing a cut-off rule for the binary decision (0 vs 1), which is generally based upon a [Z.usb.i] score cutoff of 0.5. We maintain the 0.5 cutoff in all analyses. This general model is illustrated in the first of the 4 proposed regressions in equation (6), where method of payment is modelled as a choice between cash (1) and stock (0): (6) 1n(Pcashp/1 - Pcash) = [[Beta].sub.0] + [[Beta].sub.1]BLEVTA + [[Beta].sub.2]BFCFW + [[Beta].sub.3]BBTMR + [[Beta].sub.4]RSIZEW + [[Beta].sub.5]BSTDEV where: 1n(Pcashp/1 - Pcash) = is the natural log of the `odds ratio' of bidding in cash, the result of a rearrangement of equations (4) and (5); BLEVTA = the bidder's leverage ratio, being total liabilities divided by total assets as calculated from the most recent annual report; BFCFW = the `free' cash flow (Earnings before interest, depreciation and tax--dividends paid) of the bidder as a proportion of its market value (calculated at trading day -5 relative to the announcement date); RSIZEW = natural log of the target's market capitalisation as a fraction of the bidder's market capitalisation (calculated using share prices at the time of the most recent annual report); BBTMR = bidder's book-to-market ratio (balance sheet value of bidder's non-preferred equity/bidder's market capitalisation using share prices corresponding to the annual report date). This is the `control' variable to proxy for growth/investment opportunities; and BSTDEV = the standard deviation of the bidder's share return(9) during the period [-40,-10] trading days relative to the announcement day divided by the bidder's share return as at day -5. A proxy for the uncertainty of investment opportunities. The 5 variables identified above comprise the `basic' logistic regression set for each of the four choices considered (cash or stock, cash or non-cash, non-stock or stock, cash, mixed or stock). We describe the additional variables introduced when reviewing our basic results in the next section. Running the logistic regressions requires, inter alia, the collection of financial statement information. Annual report data are obtained from (a) the AGSM Annual Report Collection (microfiche version) which contains reports up to 1993, (b) the `Connect 4' Annual Report Collection (CD Rom) Version 2.00 [C] 1996, 1997, containing reports dated from 1992 to 1997, and (c) the Australian Stock Exchange (ASX) `Datadisc' (CD Rom) Version 3.31 of July 1997 [C] 1996 by ASX Operations P/L, containing monthly financial information after 1995. As there are logistic/OLS regression variables relating to both the bidding firm and its target, it is necessary to have annual report data available with respect to both corporations before the associated `event' can enter the logistic regression sample. Eighty (80) bidder/target pairs survive this data availability criterion. 4. Results 4.1. Announcement Interval Cumulative Average Abnormal Returns (CAARs) Panels A and B of table 3 show the mean CAARs by method of payment for bidder and target firms respectively. Importantly, results are reported after deleting sample firms with outlier returns, defined as returns lying more than 3.5 standard deviations away from the relevant portfolio's mean return. Note that both cash and share bidders register insignificant negative mean returns over the 2 periods [0,+1] and [-4,+2] days relative to the bid announcement day (henceforth all periods are expressed relative to the bid announcement date unless otherwise stated). Table 3 Mean Cumulative Average Abnormal Returns (CAARs) to Bidder and Target Firms Over Various Periods of Days Relative to the Bid Announcement Day. Firms with an Outlier Return are Excluded(*)
Panel A: Bidding Firms
Cash Only Stock Only Mixed
(all)
n 66 29 25
[CAAR.sub.n[0,+1] 0.33% 1.46% 0.06%
t-stat. -0.692 -1.073 0.617
p-value 0.491 0.293 0.543
[CAAR.sub.n[-4,+2] 0.36% 2.85% 2.42%
t-stat. -0.469 -1.426 1.250
p-value 0.641 0.165 0.223
Panel B: Target Firms
Cash Only Stock Only Mixed
(all)
n 54 27 26
[CAAR.sub.n[0,+1] 6.70% 7.02% 9.25%
t-stat. 4.941 3.118 4.509
p-value <0.001 0.004 <0.001
[CAAR.sub.n[-4,+2] 10.09% 8.15% 10.42%
t-stat. 5.549 2.276 4.171
p-value <0.001 0.031 <0.001
Panel C: Wilcoxon Test
Method of Payment Cash Only Stock Only
Statistic Positive: Z Positive:
Negative (asymp. Negative
Ranks sig.) Ranks
Bidders:
Raw-Market (0,+1) 25:41 -1.900 11:19
(0.057)
Raw-Market (-4,+2) 24:42 -1.396 11:19
(0.163)
Targets:
Raw-Market (0,+1) 44:12 4.698 21:7
(0.026)
Raw-Market (-4,+2) 48:8 5.628 20:8
(<0.001)
Panel A: Bidding Firms
Cash and A Choice of Cash/Stock and
stock Cash or Stock Other Securities
n 11 10 4
[CAAR.sub.n[0,+1] 1.14% 2.45% 6.24%
t-stat. 0.600 -2.162 2.484
p-value 0.562 0.059 0.089
[CAAR.sub.n[-4,+2] 1.77% 0.33% 8.27%
t-stat. 0.574 0.120 1.737
p-value 0.579 0.907 0.181
Panel B: Target Firms
Cash and A Choice of Cash / Stock and
Stock Cash or Stock Other Securities
n 11 10 5
[CAAR.sub.n[0,+1] 10.58% 10.30% 4.25%
t-stat. 2.739 3.775 1.058
p-value 0.021 0.004 0.350
[CAAR.sub.n[-4,+2] 13.68% 14.27% 4.46%
t-stat. 3.212 4.102 -0.762
p-value 0.009 0.003 0.488
Panel C: Wilcoxon Test
Method of Payment Mixed
Statistic Z Positive: Z
(asymp. Negative (asymp.
sig.) Ranks sig.)
Bidders:
Raw-Market (0,+1) -1.224 13:13 .216
(0.221) (0.829)
Raw-Market (-4,+2) -2.047 15:11 1.206
(0.041) (0.228)
Targets:
Raw-Market (0,+1) 2.846 21:5 3.721
(0.004) (<0.001)
Raw-Market (-4,+2) 2.141 23:3 3.441
(0.032) (<0.001)
Note: p-values relate to two-tailed tests; and (*) = outliers are defined as returns lying more than plus or minus three standard deviations away from the mean sample return. Panel C of table 3 reports results from matching the experimental firms' returns with the associated daily return on the AOAI index and applying the Wilcoxon matched-pairs signed-rank test. The numbers suggest that both cash and share bidders underperform the market around the bid announcement date but the only statistically significant underperformance at the 5% level (one tailed test) is that recorded by the share bidders over the period [-4,+2] days. If cash offers signal bidder quality or bidder confidence in the positive value of the takeover this signal does not generate a net positive gain. The table 3 results for targets are consistent with many previous studies of takeovers (Jensen & Ruback 1983; Jarrell, Brickley & Netter 1988; Walter 1984; Bishop, Dodd & Officer 1986) and show target firm shareholders benefit significantly from takeover announcements. All theories on method of payment predict higher returns to targets in cash bids. Table 4 reports results from tests. To see if firms grouped according to method of payment register significantly different mean/median cumulative abnormal returns (CARs) across the periods [0,+1] and [-4,+2] days. The right-most column in table 4, which displays the results of tests of the difference in mean/median CAR to cash offer targets and share offer targets, shows an unexpectedly negative tstatistic for the [0,+1] days period. The corresponding [-4,+2] days result is positive but insignificant. However, the non-parametric Mann-Whitney test results are more strongly in the predicted direction, with the [-4,+2] days Z-statistic significant at the 5% level. Given that the target subsample is subject to high positive skewness and a high standard deviation, the non-parametric test results are probably more reliable. The results reported in the third column of table 4 show that bidders who offer cash do not earn higher returns than bidders who offer shares. Table 4 Parametric and Non-Parametric Two Sample Tests for Mean/Median Differences in Cumulative Abnormal Return (CAR) Across Portfolios Selected by Method of Payment
Bidders
CAR Interval, Cash v. Stock v. Cash v.
Statistics Mixed Mixed Stock
Panel A: Parametric (excluding outliers)
n 66:25 29:25 66:29
(0,+1)
t -0.478(a) -0.995 0.844(a)
p-value 0.636 0.324 0.405
(-4,+2)
t -1.513 -2.243 1.713
p-value 0.134 0.029 0.090
Panel B: Non-Parametric
n 66:26 30:26 66:30
(0,+1)
Mann-Whitney Z -0.694 -0.969 -0.352
p-value 0.488 0.332 0.725
(-4,+2)
Mann-Whitney Z -1.428 -2.218 -1.403
p-value 0.138 0.027 0.161
Targets
CAR Interval, Cash v. Stock v. Cash v.
Statistics Mixed Mixed Stock
Panel A: Parametric (excluding outliers)
n 54:26 28:26 54:28
(0,+1)
t -1.061 -0.795 -0.093
p-value 0.292 0.430 0.926
(-4,+2)
t -0.048 -0.578 0.704
p-value 0.962 0.566 0.484
Panel B: Non-Parametric
n 56:26 28:26 56:28
(0,+1)
Mann-Whitney Z -0.877 -1.021 0.408
p-value 0.381 0.307 0.683
(-4,+2)
Mann-Whitney Z -0.448 -1.904 1.945
p-value 0.654 0.057 0.052
Note: (a) = t-statistic calculated without assuming equal variances due to results from Levene's F-test for equal variances. The results reviewed so far are consistent with the proposition that targets subject to cash bids earn higher returns around the bid announcement period. However, if mixed-payment offers are considered an intermediate case, the results in table 3 give pause. Contrary to predictions, mixed-payment bidders earn higher returns than both cash bidders and share bidders, though only significantly so against share bidders over the period [-4,+2] days. Eckbo, Giammarino and Heinkel (1990) develop a model in which only mixed payment bids contain signalling information and synergy re-evaluation component. Blackburn, Dark and Hanson (1997) elaborate the model and develop the prediction that, in mixed payment bids, owner-controlled firms are expected to reap higher announcement period returns than managers-controlled firms. Further investigation to explain the announcement periods returns in this paper may need to take into account ownership structure. In sum, our review of the abnormal performance of bidder and target firms around the bid announcement periods mildly supports the proposition that the medium of exchange in takeover bids conveys information that is priced by the market. Cash offers are empirically associated with higher returns but the evidence is indicative only in light of the stronger results documented in other (overseas) markets. 4.2 The Long-Run Post-Bid Performance of Cash and Share Bidders Table 5 reports statistics which indicate the relative performance of bidder and target firms over the following periods of months relative to the bid announcement month: [-3,+3], [-3,+6], [-3,+12], [-3,+24], [-3,+36], [0,+24], and [+12,+36]. The first five periods listed permit observations of trends in the relative performance of the experimental sample firms as the period over which buy-and-hold returns are calculated lengthens in stages from [-3,+3] months to [-3,+36] months. This method of reviewing relative performance is vulnerable to Fama's (1998) criticism (noted earlier) that exceptional early performance is compounded in buy-and-hold returns and gives a potentially misleading indicator of relative performance over the whole event-period. Table 5 Comparative BAHR Performance of Cash and Shares Bidders--After the Removal of Outlier Returns
Consideration Cash
Time Interval Basis of Comparison Mean Median
(months), N
(-3,+3) Mean/median of experimental 10.45% 4.97%
C=101, S=41 sample
Mean/median of control 9.09% 3.71%
portfolios' means/medians
# of control portfolios with 326 299
a higher mean/median
(-3,+6) Mean/median of experimental 17.58% 10.08%
C=101, S=41 sample
Mean/median of control 13.64% 6.67%
portfolios' means/medians
# of control portfolios with 172 153
a higher mean/median
(-3,+ 12) Mean/median of experimental 22.08% 14.94%
C=92, S=37 sample
Mean/median of control
portfolios' means/medians
# of control portfolios with 321 223
a higher mean/median
(-3,+24) Mean/median of experimental 30.65% 22.63%
C=70, S=26 sample
Mean/median of control 32.25% 16.35%
portfolios' means/medians
# of control portfolios with 465 185
a higher mean/median
(-3,+36) Mean/median of experimental 39.63% 35.79%
C=57, S=18 sample
Mean/median of control 58.73% 27.96%
portfolios' means/medians
# of control portfolios with 792 234
a higher mean/median
(+0,+24) Mean/median of experimental 27.24% 23.37%
C=68, S=27 sample
Mean/median of control 34.15% 14.22%
portfolios' means/medians
# of control portfolios with 541 97
a higher mean/median
(+12,+36) Mean/median of experimental 17.14% 12.13%
C=57, S=18 sample
Mean/median of control 29.22% 12.76%
portfolios' means/medians
# of control portfolios with 835 541
a higher mean/median
Consideration Shares
Time Interval Basis of Comparison Mean Median
(months), N
(-3,+3) Mean/median of experimental 15.79% 5.91%
C=101, S=41 sample
Mean/median of control 9.50% 4.08%
portfolios' means/medians
# of control portfolios with 121 292
a higher mean/median
(-3,+6) Mean/median of experimental 16.03% 17.03%
C=101, S=41 sample
Mean/median of control 15.10% 6.87%
portfolios' means/medians
# of control portfolios with 425 49(a)
a higher mean/median
(-3,+ 12) Mean/median of experimental 14.94% 2.40%
C=92, S=37 sample
Mean/median of control 15.06%
portfolios' means/medians
# of control portfolios with 864 932
a higher mean/median
(-3,+24) Mean/median of experimental 16.28% 1.56%
C=70, S=26 sample
Mean/median of control 42.39% 16.71%
portfolios' means/medians
# of control portfolios with 859 928
a higher mean/median
(-3,+36) Mean/median of experimental 23.78% 13.57%
C=57, S=18 sample
Mean/median of control 79.67% 30.05%
portfolios' means/medians
# of control portfolios with 909 995(a)
a higher mean/median
(+0,+24) Mean/median of experimental 12.46% 21.19%
C=68, S=27 sample
Mean/median of control 32.86% 13.91%
portfolios' means/medians
# of control portfolios with 857 1001(a)
a higher mean/median
(+12,+36) Mean/median of experimental 22.84% 4.70%
C=57, S=18 sample
Mean/median of control 42.33% 16.39%
portfolios' means/medians
# of control portfolios with 782 796
a higher mean/median
Note: (a) = significant at the 5% level on the basis of pseudo p-values Fama's point is of particular concern in cases where the exceptional performance is consistently in one direction because then it is difficult to identify what proportion of measured abnormal return for the whole period is attributable to compounding effects and what proportion is due to persistence in exceptional performance. As it happens, the relative performance of our samples of cash-offer and share-offer bidders over the long-term is such that, notwithstanding Fama's valid point, we can be unequivocal in our conclusions: over the long-term post-bid period, cash bidders achieve `normal' returns but share bidders exhibit significantly negative abnormal performance. Across all seven event-windows, bidding firms that offered cash exhibit unexceptional performance in both their mean and median returns, given that under- or over-performing on the corresponding metrics of more than 950 of the 1001 control portfolios defines exceptional performance. Notwithstanding this definition of abnormal performance, it is worth noting that the median return of the cash bidders exceeds the median return of over 50% of the control portfolios across all but one event-window. Over the period [+12,+36] months the cash bidders' median return is exceeded by the median return in 541 control portfolios. The solidity of the long-term post-bid performance of the cash-bidders is made starkly apparent when contrasted with the relative performance of the bidders who offered shares. Over the period [-3,+3] months share offer bidders exhibit unexceptional performance on both mean and median return measures. The share bidders' median return over the period [-3,+6] months evidences a puzzling sharp rise in their performance as a group since just 49 of 1001 control portfolios register a higher median return. However, by the end of the period [-3,+12] months the return achieved by the share bidders is substantially reversed. In six months, the experimental portfolio falls, in terms of relative median performance, from the first decile of the control portfolio distribution to the tenth decile. While a small subset of the control portfolio maintains the level of absolute average performance up to month +36, the median buy-and-hold of the experimental sample suggests significant underperformance by share bidders over the same period.(10) While the downturn in median performance over the 3.25 years period is impressive, it is even more so at the margin. Across the period [+0,+24] months, not one of the 1001 control portfolios has a median return as low as that of the experimental sample of share bidders. Table 6 shows the relative long-term post-bid performance of just bidding firms that were successful in acquiring their target firms. Of the 68 cash bidders with data over the period [0,+24] months, 30 were successful in acquiring their targets; of the 27 share bidders with data over the same period, 16 acquired their targets. Comparison of the results over this event-window in tables 5 and 6 indicate that bid outcome does not appear to influence relative performance in the long-term post-bid period. This strengthens the argument for the view that bidders issue shares when their shares are undervalued. If poor relative performance in the postbid period were mostly due to disappointing returns from takeover then we would expect the `successful' acquirers to perform worse than firms that did not acquire their targets. Table 6 Comparative BAH Return Performance of Cash and Shares Bidders--Successful Bidders Only
Consideration Cash
Time Interval Basis of Comparison Mean Median
(months)
(-3,+3) Mean/median of experimental 11.18% 4.15%
C=51, S=23 sample
Mean/median of control 9.07% 4.48%
portfolios' means/medians
# of control portfolios with 318 559
a higher mean/median
(-3,+6) Mean/median of experimental 16.48% 11.92%
C=51, S=23 sample
Mean/median of control 14.68% 8.48%
portfolios' means/medians
# of control portfolios with 300 258
a higher mean/median
(-3,+12) Mean/median of experimental 26.76% 20.48%
C=47, S=23 sample
Mean/median of control 23.99% 14.65%
portfolios' means/medians
# of control portfolios with 299 199
a higher mean/median
(-3,+24) Mean/median of experimental 36.26% 34.29%
C=33, S=16 sample
Mean/median of control 32.75% 18.47%
portfolios' means/medians
# of control portfolios with 323 92
a higher mean/median
(-3,+36) Mean/median of experimental 51.23% 42.33%
C=25, S=10 sample
Mean/median of control 60.86% 29.90%
portfolios' means/medians
# of control portfolios with 504 251
a higher mean/median
(+0,+24) Mean/median of experimental 36.63% 33.41%
C=30, S=16 sample
Mean/median of control 27.50% 15.35%
portfolios' means/medians
# of control portfolios with 237 60
a higher mean/median
(+ 12,+36) Mean/median of experimental 25.55% 4.86%
C=25, S=10 sample
Mean/median of control 23.96% 11.79%
portfolios' means/medians
# of control portfolios with 391 752
a higher mean/median
Consideration Shares
Time Interval Basis of Comparison Mean Median
(months)
(-3,+3) Mean/median of experimental 27.99% 21.95%
C=51, S=23 sample
Mean/median of control 11.38% 5.64%
portfolios' means/medians
# of control portfolios with 10(a) 0(a)
a higher mean/median
(-3,+6) Mean/median of experimental 29.56% 30.46%
C=51, S=23 sample
Mean/median of control 19.83% 7.91%
portfolios' means/medians
# of control portfolios with 142 3(a)
a higher mean/median
(-3,+12) Mean/median of experimental 23.43% 8.58%
C=47, S=23 sample
Mean/median of control 31.86% 18.44%
portfolios' means/medians
# of control portfolios with 679 862
a higher mean/median
(-3,+24) Mean/median of experimental 15.17% 1.56%
C=33, S=16 sample
Mean/median of control 35.34% 16.04%
portfolios' means/medians
# of control portfolios with 785 888
a higher mean/median
(-3,+36) Mean/median of experimental 29.18% 1.85%
C=25, S=10 sample
Mean/median of control 55.08% 24.81%
portfolios' means/medians
# of control portfolios with 662 814
a higher mean/median
(+0,+24) Mean/median of experimental -7.13% -24.88%
C=30, S=16 sample
Mean/median of control 26.05% 11.56%
portfolios' means/medians
# of control portfolios with 973(a) 995(a)
a higher mean/median
(+ 12,+36) Mean/median of experimental 4.42% 6.28%
C=25, S=10 sample
Mean/median of control 19.59% 5.95%
portfolios' means/medians
# of control portfolios with 707 795
a higher mean/median
4.3 Influences on Method of Payment: Results From Logistic Regression Analysis This section present results of tests of the influence of five variables--bidder's leverage, bidder's free cash flow, bidder's book-to-market ratio, standard deviation of bidder's share price, and relative size of target--on the choice of payment method. The explanatory value of the five variables is assessed in four multivariate logistic regressions. Table 7 contains the results. Note the two right-most columns detail the relative probabilities of cash financing and mixed payment financing in a multinomial logistic regression where the dependent variable is constrained to one of three discrete choices--cash, mixed and shares.
Table 7
Logistic Regression Results from the Basic Variable Set
Cash = 1, Cash = 1,
Shares = 0 Non-Cash = 0
Variable Coefficient p-value Coefficient p-value
(predicted sign) (std error) (std error)
Constant -0.3181 0.809 -1.884 0.127
?(a) (1.314) (1.234)
BBTMR 0.7603 0.194 0.808 0.182
+ (0.882) (0.888)
BLEVTA 0.7608 0.345 0.446 0.405
+ (1.908) (1.850)
BFCFW 2.8020 0.145 3.578 0.078
+ (2.643) (2.528)
RSIZEW -0.5830 0.028 -0.841 0.002
- (0.304) (0.297)
BSTDEV -0.2030 0.034 -0.080 0.196
- (0.112) (0.094)
Test statistics
N 56 - 67 -
Log Likelihood -29.664 - -36.998 -
Model Chi-square 16.510 - 18.870 -
(sign.) (0.006) (0.002)
Goodness of Fit 52.176 - 66.319 -
Cox & Snell 0.255 - 0.245 -
[R.sup.2]
Prediction 71.43% - 68.66% -
Accuracy Rate
Chance 51.60% - 50.01% -
Classification
Non-Shares = 1, Prob.(Cash)/
Shares = 0 Prob.(Shares)
Variable Coefficient p-value Coefficient p-value
(predicted sign) (std error) (std error)
Constant 0.490 0.664 11.035 0.044
?(a) (1.126) (5.487)
BBTMR 0.648 0.208 0.752 0.207
+ (0.798) (0.919)
BLEVTA 1.439 0.217 1.323 0.249
+ (1.839) (1.950)
BFCFW 2.547 0.152 3.491 0.102
+ (2.482) (2.753)
RSIZEW -0.332 0.109 -11.730 0.027
- (0.269) (6.071)
BSTDEV -0.237 0.007 -0.186 0.036
- (0.096) (0.103)
Test statistics
N 67 - - -
Log Likelihood -34.912 - - -
Model Chi-square 16.364 - - -
(sign.) (0.006)
Goodness of Fit 67.180 - - -
Cox & Snell 0.217 - - -
[R.sup.2]
Prediction 73.13% - - -
Accuracy Rate
Chance 54.91% - - -
Classification
Prob.(Mixed)/
Prob.(Shares)
Variable Coefficient p-value
(predicted sign) (std error)
Constant -17.989 0.098
?(a) (10.889)
BBTMR -0.670 0.341
+ (1.643)
BLEVTA 4.519 0.099
+ (3.507)
BFCFW 1.125 0.400
+ (4.416)
RSIZEW 19.472 0.044
- (11.401)
BSTDEV -0.467 0.008
- (0.192)
Test statistics
N 67 -
Log Likelihood -51.767 -
Model Chi-square 32.139 -
(sign.) (0.004) -
Goodness of Fit - -
Cox & Snell - -
[R.sup.2]
Prediction 67.16% -
Accuracy Rate
Chance 38.74% -
Classification
Note: a = having no predicted direction, p-values correspond to a two-tailed test of significance. Generally, the test statistics in table 7 indicate the logistic models have considerable but not suspiciously high explanatory power ([R.sup.2] measures between 0.2 and 0.3). Goodness of fit, log likelihood and prediction accuracy rates are all within fair limits. Null hypotheses that the coefficients of the models are all zero are rejected in each case at the 1% level of significance by the `model' chi-square statistics. The regressions indicate that, of the 5 variables, RSIZEW and BSTDEV have the greatest contextual explanatory power. The RSIZEW variable is significant in 3 of the 4 reported regressions. Its negative sign in the Cash=1, Shares=0 regression suggests that as the target increases in size relative to the bidder, the probability that the bidder will utilise stock financing increases. This result is consistent with the Hanson (1987) model of method of payment selection under information asymmetry but there are alternative explanations. It is possible that larger acquisitions are financed in stock due to insufficient cash reserves (a `cash availability' rationale) or in response to the reduced chance of a competing bidder. Interestingly, the coefficient for the relative probability of mixed payment financing in the multinomial regression is positive, suggesting that the acquisition of relatively larger targets increase the odds of a mixed payment offer. The implied preference for mixed payment financing over cash financing is even stronger. The coefficients of the BSTDEV variable are uniformly negative and often significantly so, suggesting a strong relationship between the (proportional) standard deviation of the bidder's pre-announcement price and the probability of stock financing.(11) This finding lends support to the Myer and Majluf (1984) hypothesis that shares are offered by bidders with private information that their shares are overvalued. The coefficients for BLEVTA are insignificant in all 5 regressions. Thus there is no support for the proposition that a firm's capital structure influences the method-of-payment decision. Similarly, results for the regressions on BFCFW (bidder free cash flow) do not support the view that takeover bids are prompted by free cash-flow. In sum, our regressions support expectations that bidders with smaller proportional price standard deviation prior to the acquisition attempt, or greater size relative to the their intended target, prefer to utilise cash financing. As both of these variables relate to information asymmetry, tactical considerations seem to dominate the method of payment decision, with financial considerations perhaps being more relevant at the margin. 5. Summary, Conclusions and Suggestions for Future Research Our research is summarised in the following observations. Firstly, evidence for a method of payment effect at the announcement period is limited. Cash bidders obtain `normal' returns while share bidders obtained slightly weaker returns. The relatively strong announcement period performance of bidders offering a mixture of share and cash as consideration is inconsistent with the expectation that such bids represent an intermediate case. Notwithstanding that statistical analysis does not strongly support the hypothesis that cash bids trigger significant positive revaluations of bidding firms' shares, this may reflect the shortcomings of the relatively simple information signalling model that was tested. It is reasonable to suppose that the information content of cash bids is mediated by factors such as managerial agency costs (Blackburn, Dark & Hanson 1997). For instance, firms in which managers have unconstrained discretion over the disposition of free cash-flows may be more likely to make negative net present value (NPV) cash bids as managers seek to expand their empire. In such cases, the positive signals generated by a cash bid may be outweighed by the interpretation that a cash bid from the firm signals the presence of substantial agency costs. The severity of this kind of confounding effect is unclear but future Australian research could take it into account by controlling for ownership structure. Firms with concentrated shareholdings are less likely to have significant agency costs. The performance of bidding firms over the long-term in the post-bid period is, prima facie, consistent with view that share bids signal that bidders' shares are over-valued; the performance of cash-bidders over the long-term post-bid period is solidly unexceptional. The puzzling aspect of the evidence in relation to share bidders is that investors seemingly take a long time to act on bidding managers' (presumably inadvertent) signal that their firms' shares are over-valued. An alternative explanation is that share bids are empirically but not causally associated with long-term decline in firms' relative performance. Future research could usefully focus on investigating whether share bidders typically possess other characteristics that indicate a company is over-valued (e.g. a high price/earnings ratio relative to other firms in the bidder's industry). The argument that information signalling effects dominate other considerations in the method of payment decision gains additional support in the analysis of the variables associated with cash and share offers. The finding that the probability of a share as opposed to cash offer increases in the size of the target relative to the bidder is consistent with both a `rationed cash' explanation and with the explanation that large targets increase the costs of overpayments in cash bids. This cost is lowered when shares are offered because the target shareholders are forced into making an efficient accept/reject decision (Fishman 1989). The finding that share offers are more likely the higher the variance of the bidder's share price in the pre-bid period lends support to the Myers and Majluf (1984) hypothesis that shares are offered when bidders consider their shares are overvalued. Importantly, this conclusion is not potentially confounded by differences in tax liability associated with method of payment. In Australia, capital gains tax liability is independent of the method of payment. However, the findings are based on small samples. Assessing the robustness of the results presents an obvious avenue for future research. This paper began with the premise that the asymmetric distribution of information, costly transactions, and taxes result in the decision to offer shares or cash in takeover offers a non-trivial problem. The evidence reviewed in this paper suggests the method of payment decision has significant share valuation consequences for bidder and target companies. The analysis suggests that greater predictive power may be realised if factors such as ownership structure are taken into account i.e. if more sophisticated models that control for factors that mediate the signalling consequences of method of payment are developed. (Date of receipt of final typescript: February, 2000. Accepted by Garry Twite, Area Editor.) (1.) The impact of the higher premium on the shareholder wealth of the bidding firms may be off-set, at least in part, by an increase in depreciation tax-shields, if the market value of the target' s assets exceeds their pre-acquisition tax bases (Carleton et al. 1983). (2.) Travlos and Papaioannou (1991) observe that negative returns to bidders are consistent with overpayment by bidders. They suggest that the lack of observed losses to bidders in cash offers might be due to the positive information effect off-setting negative price reaction caused by overpayment. (3.) The total does not sum to 210 because 77 takeovers involved offers which combined cash and shares. (4.) The `Ralph Committee Report' of inquiry into business taxation in Australia, released in 1999, identifies rollover relief as a possible change to capital gains tax legislation. (5.) Although Loughran and Vijh (1997) identify an initial experimental sample of 947 firms, their principal results are based on returns to this subset of 534 firms. (6.) Two reasons motivated the use of the AOAI as the proxy for expected return. (a) Bellamy and Lewin (1992) and, Bugeja and Walter (1995) also use the AOAI as a proxy for the market portfolio when they adopt the `zero-one' market model to estimate abnormal returns. Our use of the AOAI thus enhances comparability of our results with earlier Australian findings. (b) Brown and Warner (1985) demonstrate that over short intervals the results obtained when using a daily index are reasonably robust to refinements. (7.) Bellamy and Lewin (1992) apply the same filter. (8.) The BHRR for each firm in the experimental (`sample') portfolio is calculated according to: [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] where: [R.sub.i,t] = the discrete monthly return of firm i during month t. (9.) This is the proxy for information asymmetry as used in How, Izan and Monroe (1995). (10.) The pseudo p-value is almost 0.005. (11.) The possibility that that stock bidding is associated with greater amounts of information leakage (inducing volatility) is inconsistent with the finding that the method of payment effect is most evident in the announcement period abnormal returns of share bidders. References Barber, B.M. & Lyon, J.D. 1997, `Detecting long-run abnormal stock returns: The empirical power and specification of test statistics', Journal of Financial Economics, vol. 43, pp. 341-72. Bellamy, D.E. & Lewin, W.L. 1992, `Corporate takeovers, method of payment, and bidding firms' shareholder returns: Australian evidence', Asia Pacific Journal of Management, vol. 9, no. 2, pp. 137-49. Bishop, S., Dodd, P. & Officer, R.R. 1986, Australian Takeovers: The Evidence, The Centre for Independent Studies, St Leonards, NSW. Blackburn, V.L., Dark, F.H. & Hanson, R.C. 1997, `Mergers, method of payment and returns to manager- and owner-controlled firms', Financial Review, vol. 32, no. 3, pp. 569-89. Blume, M. & Stambaugh, R.F. 1983, `Biases in computed returns: an application to the size effect', Journal of Financial Economics, vol. 12, pp. 387-404. Brown, D.T. & Ryngaert, M.D. 1991, `The mode of acquisition in takeovers: Taxes and asymmetric information', Journal of Finance, vol. 46, no. 2, pp. 653-69. Brown, P. & da Silva Rosa, R. 1998, `Research method and the long-run performance of acquiring firms', Australian Journal of Management, vol. 23, no. 1, pp. 23-38. Brown, S.J. & Warner, J.B. 1985, `Using daily stock returns: the case of event studies', Journal of Financial Economics, vol. 14, pp. 3-31. Bugeja, M. & Walter, T. 1995, `An empirical analysis of some determinants of the target shareholder premium in takeovers', Accounting and Finance, vol. 35, pp. 33-60. Carleton, W.T., Guilkey, D.K., Harris, R.S. & Stewart, J.F. 1983, `An empirical analysis of the role of the medium of exchange in mergers', Journal of Finance, vol. 38, pp. 813-26. Davidson, W.N. & Cheng, L.T.W. 1997, `Target firm returns: Does the form of payment affect abnormal returns?', Journal of Business Finance and Accounting, vol. 23, pp. 465-79. Deakin, E.B. 1976, `Distributions of financial accounting ratios: Some empirical evidence', Accounting Review, vol. 51, pp. 90-96. Eckbo, B.E., Giammarino, R.M. & Heinkel, R.L. 1990, `Asymmetric information and the medium of exchange in takeovers: Theory and tests', Review of Financial Studies, vol, 3, pp. 651-75. Eckbo, B.E. & Langohr, H. 1989, `Information disclosure, methods of payment and takeover premiums: public and private tender offers in France', Journal of Financial Economics, vol. 24, pp. 363-403. Fama, E. 1998, `Market efficiency, long-term returns, and behavioral finance', Journal of Financial Economics, vol. 49, pp. 283-306. Fishman, M. 1989, `Preemptive bidding and the role of the medium of exchange in acquisitions', Journal of Finance, vol. 44, pp. 41-57. Franks, J.R., Harris, R.S. & Mayer, C. 1988, `Means of payment in takeovers: Results for the United Kingdom and the United States', in Corporate Takeovers, Causes and Consequences, (ed.) A. Auerbach, University of Chicago Press, p. 221-58. Galai, D. & Masulis, R. 1976, `The option pricing model and the risk factor of stock', Journal of Financial Economics, vol. 3, pp. 53-81. Gresham Partners Ltd 1996, `Scrip Takeover Offers Trends and Issues', unpublished conference paper of the LCA Business Law Workshop, October 1996. How, J.C.Y., Izan, H.Y. & Monroe, G.S. 1995, `Differential information and the underpricing of initial public offerings: Australian evidence', Accounting and Finance, vol. 35, no. 1, pp. 87-107. Hansen, R.G. 1987, `A theory for the choice of exchange medium in mergers and acquisitions', Journal of Business, vol. 60, pp. 75-95. Huang, Y.S. & Walkling, R.A. 1987, `Target abnormal returns associated with acquisition announcements--payment, acquisitions form, and managerial resistance', Journal of Financial Economics, vol. 19, pp. 329-49. Jarrell, G.A., Brickley, J.A. & Netter, J.M. 1988, `The market for corporate control: The empirical evidence since 1980', Journal of Economic Perspectives, vol. 2, no. 1, pp. 49-68. Jensen, M. 1986, `Agency costs of free cash flow, corporate finance and takeovers', American Economic Review, vol. 76, pp. 323-329. Jensen, M.C. & Ruback, R.S. 1983, `The market for corporate control: The scientific evidence', Journal of Financial Economics, vol. 11, pp. 5-50. Kothari, S.P. & Warner, J.B. 1997, `Measuring long-horizon security price performance', Journal of Financial Economics, vol. 43, pp. 301-39. Leland, H. & Pyle, D. 1977, `Information asymmetries, financial structure and financial intermediaries', Journal of Finance, vol. 32, pp. 371-87. Loughran, T. & Vijh, A.M. 1997, `Do long-term shareholders benefit from corporate acquisitions?', Journal of Finance, vol. 51, no. 5, pp. 1765-90. Martin, K. J. 1996, `The method of payment in corporate acquisitions, investment opportunities, and management ownership,' Journal of Finance, vol. 51, pp. 1227-1246. Myers, S. & Majluf, N. 1984, `Corporate financing and investment decisions when firms have information that investors do not', Journal of Financial Economics, vol. 13, pp. 187-221. Roll, R. 1983, `On computing mean returns and the small firm premium', Journal of Financial Economics, vol. 12, pp. 371-86. Travlos, N.G. 1987, `Corporate takeover bids, methods of payment, and bidding firms' stock returns', Journal of Finance, vol. 42, no. 4, pp. 943-63. Travlos, N.G. & Papaioannou, G.J. 1991, `Corporate acquisitions: method of payment effects, capital structure effects, and bidding firms' stock returns', Quarterly Journal of Business, vol. 30, no. 4, pp. 3-22. Walter, T. 1984, `Australian takeovers: Capital market efficiency and shareholder risk and return', Australian Journal of Management, vol. 9, no. 1, pp. 63-118. Wansley, J.W., Lane, W.R. & Yang, H.C. 1983, `Abnormal returns to acquired firms by type of payment and method of payment', Financial Management, vol. 12, pp. 16-22. Helpful comments from an anonymous referee, Steve Easton and other participants at the 1999 Asia-Pacific Finance Association Conference organised by the School of Economics and Finance at RMIT, Melbourne are gratefully acknowledged. The authors are responsible for all errors. Raymond da Silva Rosa ([dagger]) H. Y. Izan ([double dagger]) Adam Steinbeck ([sections]) Terry Walter ([+ or -]) ([dagger]) Department of Accounting, The University of Sydney, NSW 2006. Email: raymond@econ.usyd.edu.au ([double dagger]) Division of Business, IT and Law, Murdoch University, Murdoch WA 6150 ([sections]) Analyst, Reserve Bank of Australia ([+ or -]) Department of Accounting, The University of Sydney, NSW 2006 |
|
||||||||||||||||||||
Printer friendly
Cite/link
Email
Feedback
Reader Opinion