An empirical examination of white knight corporate takeovers: synergy and overbidding.

The purpose of this study is to investigate the characteristics of takeover contests which involve a "white knight," i.e., a friendly bidding firm actively sought by a target which is resisting acquisition by a hostile bidding firm. In view of existing evidence that participation in an acquisition does not, in general, confer substantial benefits on bidding firm shareholders,(1) the white knight's motivation for attempting to outbid the hostile rival is unclear. Previous studies suggest that due to irrationality, hubris, miscalculation, the "winner's curse," or managerial objectives other than stockholder wealth maximization, some bidding firms may often simply overbid for their respective targets (Boebel and Harris |5~, Giammarino and Heinkel |9~, Morck, Shleifer, and Vishny |18~, Roll |21~, Ruback |22~, Shleifer and Vishny |25~, and Varaiya and Ferris |27~). The likelihood of overbidding is increased by the white knight's position in the sequential bidding process: by definition, white knights are participants in competitive acquisitions involving more than one bidder, and the white knight enters the contest subsequent to the first bidder's attempt. Prior research has shown that competition has a negative effect on bidding firm shareholder gains, and this effect is more pronounced for subsequent bidders than for firms which make the first bid (Bradley, Desai, and Kim |7~). Similarly, Banerjee and Owers |3~ investigate the auction properties of white knight acquisitions, and find evidence that white knights, as subsequent bidders, experience losses upon entry into a bidding contest. Clearly, under the hypothesis that white knights systematically overbid, becoming a white knight does not serve the interests of the bidder's shareholders. Rather, the beneficiaries of a white knight acquisition are target shareholders, who receive inflated acquisition premia, and target managers, who would likely lose their jobs if the firm was taken over by a hostile bidder.(2) The hypothesis that white knights differentially and systematically overbid for their targets has the testable implication that white knight acquisitions generate greater gains to target shareholders and greater losses to bidding firm shareholders than do non-white-knight acquisitions.

Another view of white knights is advanced by Shleifer and Vishny |24~, who present a model in which acquisition by a white knight represents the optimal synergistic combination of target and bidding firm assets (termed the "maximum synergy hypothesis" in this study). The key element of this model is that the target firm has private information as to how to achieve the highest valuation of its assets, and this information is shared with the white knight, but withheld from other bidders. The most obvious testable implication of the maximum synergy hypothesis is that combined target and bidding firm shareholder wealth effects (i.e., synergies) are predicted to be greater in white knight acquisitions than in non-white-knight acquisitions. Shleifer and Vishny's model does not specify how the optimal synergistic gains are split between

shareholders of the target and the white knight, nor does the model preclude overbidding.

This study explores the validity of the maximum synergy and overbidding hypotheses of white knight corporate acquisitions. The empirical tests compare white knight bids with two control groups: (i) hostile bids (to which target management has voiced an explicit objection); and (ii) friendly non-white-knight bids (which have the cooperation of target management, but which are not white knights). The major empirical results are summarized below: (i) White knights experience significant abnormal losses upon announcing entry into a bidding contest. The losses to white knights exceed announcement period losses to friendly non-white-knight bidders, and in some cases also exceed losses to hostile bidders. White knights appear to account entirely for the previously reported finding that shares of subsequent bidders experience significantly greater losses than those of first bidders (Bradley, Desai, and Kim |7~ and Boebel and Harris |5~).

(ii) Targets acquired by white knights experience higher abnormal gains around the acquisition announcement than targets acquired by friendly non-white knights and about the same abnormal gains as targets acquired by hostile bidders. However, once bidder competition is controlled, white knight status does not significantly affect target gains. Targets acquired by white knights receive about the same gains as targets acquired by non-white knights in multiple-bidder acquisitions.

(iii) White knight acquisitions generate positive synergies, on average, as do acquisitions by friendly non-white-knight and hostile bidders. Synergies created in white knight acquisitions significantly exceed those created in friendly non-white-knight acquisitions, and are insignificantly different from those created in hostile acquisitions. However, the difference in synergies between white knight and friendly acquisitions stems from bidder competition. White knight acquisitions generate about the same synergies as non-white-knight acquisitions involving bidder competition.

The results outlined above support the overbidding hypothesis. Although white knight acquisitions exhibit significantly positive average synergies, overbidding apparently leads to the observed distribution of gains between the shareholders of the target and the white knight.(3) The evidence for the maximum synergy hypothesis is less persuasive. First, acquisition by a white knight does not result in significantly different synergy than acquisition by a hostile bidder. Second, the synergies generated in white knight acquisitions are insignificantly different from those produced in non-white-knight acquisitions once the effects of bidder competition are controlled. The major contribution of this paper is to show that white knights represent an ex ante identifiable segment of bidding firms which systematically experience greater abnormal losses upon entry into an acquisition contest than their non-white-knight counterparts. Indeed, 45% of the white knights examined in this study themselves became targets, restructured, or adopted anti-takeover defenses during the five years subsequent to becoming a white knight. This is consistent with Mitchell and Lehn's |17~ conclusion that one important motive for acquisition is to discipline managers of firms which have previously made value-reducing acquisitions. Furthermore, the finding that subsequent bidders suffer from the "winner's curse" (Bradley, Desai, and Kim |7~ and Boebel and Harris |5~) is apparently entirely attributable to white knights. These results have serious welfare implications for the shareholders of the white knight, and bring into question the efficiency of managerial decisions to become a white knight.

The remainder of this study is structured as follows: Section I describes the acquisition data used in empirical tests. Section II describes the methodology used to measure the abnormal target and bidder stock price response to an acquisition announcement and aggregate acquisition synergies. Section III presents the empirical results. Section IV contains concluding remarks.

I. Data

Acquisition attempts are classified into three target management response categories: (i) white knight, (ii) friendly non-white knight, and (iii) hostile, using information from entries in the Wall Street Journal Index as well as the corresponding articles in the Wall Street Journal. The operational definition of "white knight" used in this study is a friendly bidder which launches an acquisition attempt subsequent to, and in competition with, a hostile bidder. A bidder is classified as hostile if at any time during the event there is specific evidence of hostility and resistance by target management. For example, target management may reject a bid outright, file a lawsuit against the bidder, or appeal to a regulatory agency such as the SEC, FTC or Justice Department for assistance in fending off the bidder. A bidder will be classified as hostile if target management initially resists and then subsequently cooperates; a bidder will also be classified as hostile if target management is initially friendly and then later resists. In order to be classified as "friendly," there must be explicit evidence throughout the event that target management consents to the bidder's offer, e.g., the language of the Wall Street Journal articles must include "approval" or "agreement" by target management. Thus, a bidder is classified as a "white knight" if it meets the criteria for the "friendly" classification and the target has already received an attempt from another bidder which meets the "hostile" criteria. A bidder is classified as a "friendly non-white knight" if it meets the criteria for the "friendly" classification and there are no existing bids which meet the "hostile" criteria. The data used in this study were identified primarily from acquisition attempts involving the 890 NYSE and AMEX targets which were delisted during the period 1974-1984 due to being acquired, and an additional 13 acquisitions which were identified from the MERC tender offer database (version ending in September 1980) which meet the criteria for inclusion outlined below. Relevant details of each acquisition, including the date of the first announcement that the target is the object of a takeover attempt, were obtained from the Wall Street Journal Index. The Wall Street Journal Index was checked for prior acquisition attempts on each target firm, in order to be sure that each announcement date did not simply capture a response to some previous bid. This study uses a scheme similar to that described in Bradley, Desai, and Kim |6~, |7~ to bundle multiple bids for the same target into "events." An event begins when the Wall Street Journal first mentions that a particular firm is the target of an acquisition. In the case of a single bid, the event ends on the day that bid is resolved.(4) If there are other bids outstanding on that day, or forthcoming within 40 trading days, the event is extended through the resolution date of the last bid which meets these criteria. Two samples, one composed of bidders and one of targets, were constructed for this study. The target sample consists of firms successfully acquired during the period 1974-1984. The bidder sample consists of firms, successful and unsuccessful, which bid upon the targets. Sample firms had to meet the following requirements for inclusion in this study:

(i) Sample firms must be listed on the NYSE or the AMEX so that the stock data may be obtained from the CRSP Daily Master and Daily Returns files.

(ii) For a target to be included, its ultimately successful acquirer must own less than 50% of target stock at the time of the first acquisition announcement, must announce an intention to increase its ownership in target stock to at least 80%, and must successfully increase its ownership to at least 80%. Firms entering the bidder sample must also own less than 50% of target stock at the time of the first acquisition announcement and must announce an intention to increase ownership in target stock to at least 80%. These criteria ensure that the targets and bidders analyzed in this study are involved in events in which the bidder seeks substantial control of target assets. Smaller stock purchase transactions, in which the bidder's intentions are sometimes unclear, are excluded.(5)

(iii) Acquisitions must be structured as mergers or tender offers; going-private transactions are excluded.

(iv) Sufficient information must be available from the Wall Street Journal Index and from Wall Street Journal articles to code the takeover attempts involved in each acquisition event as "hostile," "friendly non-white knight," or "white knight" using the criteria described above.

(v) Each target and bidding firm must have traded at least 100 of the 160 trading days beginning 200 trading days before and ending 41 trading days before the Wall Street Journal announcement date of the acquisition. This requirement ensures that sufficient returns data exist to estimate market model parameters. (vi) Each target must have traded on the Wall Street Journal announcement date of the last revised or competing bid for that target. Each bidder must have traded on the Wall Street Journal announcement date of the last revised bid by that bidder for its respective target. This requirement facilitates computation of the cumulative abnormal stock price response for both targets and bidders. (vii) There must be no confounding acquisition activity during the 200 trading days preceding the first announcement in the event. The 200-day period subsumes the pre-event market model estimation period.

Application of these criteria yielded 376 targets and 334 bidding firms. The numbers of bidders and targets differ primarily because: (i) some CRSP-listed targets are sought by non-CRSP-listed bidders; and (ii) sometimes more than one bidding firm bid on a particular target. Exhibit 1 shows the sample distribution by target management response for the target and bidder samples, respectively. The 376 targets are classified by the target management response to the ultimately successful bid for that target, and include 279 targets acquired by friendly non-white-knight bidders, 51 targets acquired by hostile bidders, and 46 targets acquired by white knights. The 334 bidders include 211 non-white knights making friendly attempts, 73 making hostile attempts, and 50 white knights. Exhibit 1 further segments the data by acquisition method; note that most of the white knight and hostile bids are tender offers, while most of the friendly non-white-knight bids are mergers. In addition, Exhibit 1 shows the frequencies of single-bidder versus multiple-bidder events across target management response groups. For both the target and bidder samples, hostile bids are associated with a greater frequency of competing bidders than are friendly non-white-knight bids (by definition 100% of white knight bids involve competition).

By construction, the target sample includes only successfully acquired targets. The bidder sample contains both successful and unsuccessful CRSP-listed bidders making attempts on successfully acquired CRSP-listed targets. The bottom rows of Exhibit 1 describe the frequency of bidding firm success across target management TABULAR DATA OMITTED response categories. As expected, resistance by target management greatly increases the likelihood of bid failure: friendly non-white-knight bidders are associated with a higher frequency of success (97%) than hostile bidders (42%). Becoming a white knight in no way guarantees success: 10 (20%) of the white knights in the bidder sample were unsuccessful.(6)

II. Measuring Shareholder Wealth Effects Around the Announcement

Several of the tests presented in this study rely on measures of target and bidding firm shareholder wealth effects due to the acquisition. Daily stock returns are assumed to be distributed multivariate normal, so that the market model may be used to extract abnormal bidder or target returns due to the announcement of an acquisition:

|u.sub.jt~ = |R.sub.jt~ - (|Alpha~ + ||Beta~.sub.j~|R.sub.mt~), (1)

where

|R.sub.jt~ = Continuously compounded rate of return for firm j on day t.

|R.sub.mt~ = Continuously compounded rate of return for CRSP value-weighted market index on day t.

||Alpha~.sub.j~, ||Beta~.sub.j~ = Scholes-Williams estimates of regression parameters for firm j estimated over the 160-day period beginning 200 trading days and ending 41 trading days prior to announcement of the acquisition event in the Wall Street Journal.

|u.sub.jt~ = Market model prediction error for firm j, day t.

The announcement is defined as the date on which the acquisition is first disclosed in the Wall Street Journal. Event windows for cumulating abnormal bidder and target returns begin prior to this date to capture both pre-announcement leakage of acquisition news and possible disclosure of the acquisition via the Dow Jones Broad Tape the day before its Wall Street Journal announcement. In measuring the effect of the announcement on bidding firm share price, it is important to choose a relatively short event window in which to observe abnormal returns, because any abnormal returns to bidding firm shares are likely to be small relative to their standard error. Consistent with previous work by Asquith, Bruner, and Mullins |2~, this study isolates the acquisition announcement effect on bidding firm shareholder wealth by using a bidding firm event window which begins at day -1 preceding the acquisition announcement in the Wall Street Journal.(7) To account for the additional information contained in any revised bids made by the bidding firm, the event window is extended through the announcement date of the last bid made by that bidder for its respective target. Thus, the variable BCPE represents cumulative market model prediction errors from day -1 preceding announcement of the first bid by a bidding firm, through the announcement of that bidder's last revision: |BCPE.sub.j~ = |summation of~ |u.sub.jt~ where t = b1 - 1 to b2, (2)

where

|Sigma~(|BCPE.sub.j~) = ||Sigma~.sub.j~ |square root of |L.sub.j~~. (2a)

||Sigma~.sub.j~ = Estimation period residual standard deviation from market model regression for bidding firm j.

b1 = Date of first public announcement of acquisition attempt by bidder j for a particular target.

b2 = Date of announcement of last bid by bidder j for its particular target.

|L.sub.j~ = Number of trading days over which the up accumulated to compute |BCPE.sub.j~.

For bidding firms making only one bid for their respective targets, BCPE will comprise two days of prediction errors; for bidding firms making more than one bid, BCPE will comprise more than two days of prediction errors. Since BCPE is not i.i.d. across bidding firms, significance tests involving this variable are based on a cumulative standardized prediction error (SBCPE) using the methodology described in Patell |20~:

|SBCPE.sub.j~ = 1/|square root of |L.sub.j~~ |summation of~ |u.sub.jt~/||Sigma~.sub.j~|square root of |C.sub.jt~~ where t = b1 - 1 to b2, (3)

where

|Mathematical Expression Omitted~.

|T.sub.j~ = Number of days in the bidding firm's parameter estimation period.

Past studies have reported that the target share price response to the acquisition disclosure begins well before announcement in the Wall Street Journal (see Jensen and Ruback |14~ for a survey). In order to capture as much of the informational impact of the acquisition announcement on target share price as possible, the target event window begins at day -40.(8) Again, in order to incorporate the additional information contained in revised and competing bids for a particular target, target cumulative prediction errors are extended through the announcement of the last bid (by any bidder) for that particular target:

|TCPE.sub.k~ = |summation of~ |u.sub.kt~ where t = t1-40 to t2, (4)

where

|Sigma~(|TCPE.sub.k~) = ||Sigma~.sub.k~|square root of |L.sub.k~~. (4a)

||Sigma~.sub.k~ = Estimation period residual standard deviation from market model regression for target k.

t1 = Date of first public announcement that target k is the object of a takeover attempt.

t2 = Date of announcement of last revised or competing bid for target k.

|L.sub.k~ = Number of trading days over which the |u.sub.kt~ are cumulated to compute |TCPE.sub.k~.

The cumulative prediction error measure for targets, represented by TCPE, will comprise 41 days of returns for targets receiving only one bid, and more than 41 days for targets receiving revised and/or competing bids. As with bidders, significance tests of TCPE are based on a cumulative standardized prediction error (STCPE):

|STCPE.sub.k~ = 1/|square root of |L.sub.k~~ |summation of~ |u.sub.kt~/||Sigma~.sub.k~|square root of |C.sub.kt~~ where t = t1-40 to t2. (5) where

|Mathematical Expression Omitted~

|T.sub.k~ = Number of days in the target firm's parameter estimation period.

Announcement period abnormal gains to target and bidding firm shareholders are also measured in (nominal) dollars to gauge the comparative economic significance of friendly non-white-knight, hostile and white knight acquisitions. Dollar gains are computed by multiplying the announcement period cumulative prediction error by the aggregate market value of the firm's common stock measured one day preceding the beginning of the cumulation period. For targets, the value of common stock is observed at day -41; for bidders, at day -2.(9)

$BCPE = |BSIZE.sub.-2~ x BCPE, (6)

STCPE = |TSIZE.sub.-41~ x TCPE, (7)

where

$BCPE, $TCPE = Abnormal dollar gains to bidding firm and target firm shares, respectively, during the announcement period.

|BSIZE.sub.-2~, |TSIZE.sub.-41~ = Aggregate market value of bidder and target common stock, measured one day preceding the beginning of the cumulation periods for computing BCPE and TCPE, respectively.

Aggregate dollar synergies are computed by summing the target and bidding firm dollar gains for matched pairs of targets and bidders involved in ultimately successful acquisitions. Gains on target shares already held by the bidding firm are excluded from the dollar synergy computation, since these gains will be reflected in the bidder's dollar gain:

$SYNERGY = $BCPE + $TCPE x (1 - HELD), (8)

where

HELD = Proportion of target common stock held by the bidding firm prior to the announcement.

Synergies are also expressed as a proportion of the combined pre-announcement market value of target and bidding firm common stock. The value of target shares already held by the bidder is omitted from the preannouncement market value of target stock, since this amount will be reflected in the pre-announcement market value of bidding firm shares:

SYNERGY = $SYNERGY / |BSIZE.sub.-2~ + |TSIZE.sub.-41~ x (1 - HELD). (9)

III. Empirical Tests of Overbidding and Maximum Synergy

A. Abnormal Returns to Bidder and Target Common Stock Around the Announcement

This section presents univariate tests comparing announcement period bidder and target abnormal returns and dollar gains across white knight, hostile, and friendly non-white-knight bids. Multivariate regressions are also presented which control for bid characteristics in addition to target management response to the bid. Since white knight status is correlated with other bid characteristics which have been shown to influence abnormal gains to TABULAR DATA OMITTED bidder and target shareholders, e.g., bidder competition and method of acquisition, the regression analysis will help to establish whether observed white knight status effects on bidder and target abnormal returns are independent of the effects of other bid characteristics.(10) 1. Univariate Results

Exhibits 2 and 3 contain summary statistics for acquisition announcement period abnormal returns to bidders and targets (BCPE and TCPE, respectively), segmented by target management response. Exhibit 2 shows that in the full sample, the mean bidding firm experienced a statistically significant loss of 2.2% due to the acquisition announcement, although values range from -71% to +27%. Segmenting by target management response into white knights, hostile bidding firms and friendly non-white knights reveals differences in BCPE across these groups. The largest losses accrue to white knights, which, on average, experience a -3.8% return during the acquisition announcement period, compared with -2.1% for friendly non-white knights and -1.5% for hostile bidders. White knight losses are also largest when announcement period abnormal returns are expressed in dollars: shareholders of the median white knight lost $13,366,800, compared with losses of $3,811,300 for hostile bidders, and $4,521,530 for friendly non-white knights. T-tests and Wilcoxon rank sum tests presented in Panel B show that white knight losses, expressed both as cumulative returns and in dollars, significantly exceed those experienced by their friendly non-white-knight and hostile counterparts.(11) Announcement period abnormal returns are not significantly different across hostile and friendly non-white-knight bidders. Exhibit 3 shows that targets of white knights experience greater announcement period gains than targets of friendly non-white knights and about the same gains as targets acquired in hostile bids. The average target of a successful white knight acquisition earned a 45.9% abnormal return, compared with 45.1% for targets acquired by hostile bidders and 33.8% for targets acquired by friendly non-white knights. As with the evidence for bidders from Exhibit 2, a similar pattern is seen when target gains are expressed in dollars. The median announcement period gain to shareholders of targets of white knights is $41,257,670, compared with $35,913,440 to targets of hostile bidders and $12,756,970 to targets of friendly non-white knights. Significance tests of differences in target gains, both in cumulative return and dollar terms, are presented in Panel B. The difference in target shareholder gains is significant in the comparison of white knight versus friendly non-white-knight acquisitions, but not in the comparison of white knight versus hostile. The Wilcoxon rank sum Z-statistics indicate that target gains are significantly greater in hostile than in friendly non-white-knight acquisitions. The univariate evidence presented in Exhibits 2 and 3 suggests overbidding by white knights: white knight bids are associated with larger losses to bidding firm shareholders and larger gains to target shareholders than hostile and friendly bids, and these differences are statistically significant in most of the comparisons involving white knights (with the exception of the comparison between target shareholder gains in white knight versus hostile acquisitions). These conclusions are robust to measuring announcement period abnormal wealth effects in dollars. Indeed, the largest bidding firm dollar loss ($864 million) and the largest target firm dollar gain ($3.4 billion) are both attributable to the 1981 acquisition of Conoco by white knight Du Pont. The next section uses linear regression models in order to establish whether the evidence of relatively large bidder losses and target gains observed among white knight acquisitions persists after controlling for other bid characteristics. 2. Regression Results

In this section, a regression model is presented in which the dependent variable of interest (BCPE or TCPE) is compared across target management response (hostile, white knight, friendly non-white knight) holding constant the method of acquisition (merger versus tender offer) and the acquisition payment method (100% cash, 100% common stock, mixed cash-plus-stock).(12) Bids involving payment methods other than 100% cash, 100% common stock or mixed cash-plus-stock (e.g., debt securities, combinations of cash or equity with debt) were eliminated from the regression tests in order to effectively control for payment method. Target regressions also include a dummy variable to distinguish events which involve more than one bidder from those which involve only one bidder. A slightly different distinction is made in the bidding firm regressions: a dummy variable is included to distinguish subsequent bids from first bids.(13) Since the bidding firm sample includes both successful and unsuccessful bidders, bidding firm regressions include a dummy variable for success.(14) TABULAR DATA OMITTED

a. Bidder Regressions

In order to ascertain the marginal effect of white knight status on acquisition period abnormal returns to bidding firm shareholders, holding constant the effects of other bid characteristics, the following regression is estimated: BCPE = ||Alpha~.sub.0~ + ||Alpha~.sub.1~WK + ||Alpha~.sub.2~HOST + ||Alpha~.sub.3~CASH + ||Alpha~.sub.4~ MIX + ||Alpha~.sub.5~SUBSEQ + ||Alpha~.sub.6~TNDOFR + ||Alpha~.sub.7~SUCCESS, (10)

where

WK = 1 if the bidder is a white knight; equals 0 otherwise.

HOST = 1 if the bidder is hostile; equals 0 otherwise.

CASH = 1 if the payment offered is 100% cash; equals 0 otherwise.

MIX = 1 if the payment offered is a combination of cash and stock; equals 0 otherwise.

SUBSEQ = 1 if the bidding firm enters the contest subsequent to the first bidder; equals 0 if the bidding firm makes the first bid for its particular target.

TNDOFR = 1 if the bid is a tender offer; equals 0 otherwise.(15)

SUCCESS = 1 if the bidder is ultimately successful; equals 0 otherwise.

This regression is estimated using WLS, in order to correct for possible heteroskedasticity arising from cross-sectional differences in market model residual standard errors and in the number of days over which prediction errors are cumulated.(16) The results of estimating various versions of this model are shown in Exhibit 4. Regression (1) of Panel A includes the payment method (CASH, MIX) and target management response (WK, HOST) variables as regressors. The coefficient on the cash (CASH) dummy variable is significantly positive, consistent with the findings of previous studies, and the coefficient on the white knight dummy variable (WK) is significantly negative. Relative to the intercept, which picks up friendly non-white-knight bids involving payment via stock, white knight bidders realize significantly negative announcement period returns. White knights also realize lower announcement period returns than do hostile bidders, and the F-statistic shown in the first row of Panel C shows that this difference is at least marginally significant (F = 3.0083, p = 0.0840).

Regression (2) adds dummy variables for subsequent bids (SUBSEQ), tender offers (TNDOFR) and ultimate bid success (SUCCESS); none of the coefficients on these variables is statistically significant.(17) The coefficient on WK in Regression (2) continues to be significantly negative, and its point estimate is not greatly changed from the coefficient estimated in Regression (1). However, equality of the coefficients on WK and HOST can no longer be rejected according to the F-statistic in Row (2) of Panel C.

The persistence of the significant negative coefficient on the white knight dummy variable in Regression (2) suggests that white knights experience greater losses than other types of subsequent bidders. Further evidence of this is shown in Regression (3) of Panel B, in which BCPE of subsequent bids only is regressed on WK, HOST, CASH and MIX. Even among subsequent bids, the coefficient on the white knight dummy is negative, statistically significant and of about the same magnitude as in the full sample Regressions (1) and (2). Again, the F-statistic shown in Row (3) of Panel C indicates that we can not reject equality of BCPE across hostile and white knight bidders.

The regressions confirm the univariate results from Exhibit 2: white knight losses in the announcement period significantly exceed the losses experienced by their friendly non-white-knight counterparts. White knight losses also exceed losses to hostile bidders, although this difference is statistically significant only in Regression (1) of Exhibit 4. Controlling for each bidder's position in the sequential bidding process in Regressions (2) and (3) reveals that white knights lose more than non-white-knight subsequent bidders. In fact, white knights seem to account entirely for the winner's curse among subsequent bidders, since non-white knights do not exhibit differential returns across first and subsequent bidders. When white knights are deleted from the sample, the mean BCPE for first and subsequent bidders is -1.8% and -2.4%, respectively, and the difference between these is not statistically significant. Thus, the well-known result that subsequent bidders experience greater announcement period losses than first bidders (Bradley, Desai, and Kim |7~ and Boebel and Harris |5~) is apparently attributable to white knights. b. Target Regressions

Similar to the model of bidding firm announcement period returns discussed above, the following model is estimated in order to examine the effects of white knight TABULAR DATA OMITTED status on announcement period abnormal returns to target shareholders:

TCPE = ||Alpha~.sub.0~ + ||Alpha~.sub.1~WK + ||Alpha~.sub.2~HOST + ||Alpha~.sub.3~CASH + ||Alpha~.sub.4~MIX + ||Alpha~.sub.5~MULT + ||Alpha~.sub.6~TNDOFR, (11)

where

MULT = 1 if more than one bidding firm competed for the target; equals 0 otherwise.

WK, HOST, CASH, MIX and TNDOFR are defined in Section III.A.2.a., and are coded such that they represent the characteristics of the ultimately successful bid for the target.

The target model is slightly different from the bidder regression model. The target regression includes a dummy variable for multiple bidders (MULT) rather than the first/subsequent distinction made in the bidder regressions. The results of WLS estimation of various versions of this regression are shown in Exhibit 5. WLS weights for the target regressions are computed in the same way as for bidders in Exhibit 4. Regression (1) controls for method of payment (CASH, MIX) and target management response (WK, HOST), with the intercept picking up the effect of friendly non-white-knight stock bids. The coefficients of the white knight dummy, the cash dummy and the mixed cash-plus-stock dummy are all positive and significantly different from zero. Relative to targets acquired in friendly non-white-knight acquisitions, targets of ultimately successful white knights earn significantly positive abnormal returns around the acquisition announcement. However, the F-statistic in Row (1) of Panel C indicates that we can not reject equality of target gains across white knight and hostile acquisitions.

Regression (2) adds the dummy variable for multiple bids (MULT), which has a significantly positive coefficient. With the addition of this variable, the coefficient on TABULAR DATA OMITTED the white knight dummy becomes insignificant, suggesting that the superior abnormal returns earned by targets of successful white knights reported in Regression (1) are attributable to the effects of bidder competition. Regression (3) adds a dummy variable for acquisition method (TNDOFR), which has a significantly positive coefficient. Addition of TNDOFR to the model does not alter the conclusion from Regression (2) that white knight effects on TCPE are due to bidder competition. Again, the F-statistics in Rows (2) and (3) of Panel C indicate that we can not reject equality of the coefficients of the white knight and hostile dummy variables. Similar to the bidding firm tests, single-bidder events are omitted from the regression reported in Panel B in order to further explore the connection between white knight effects on target announcement period abnormal returns and bidder competition. Again, the coefficient of the white knight dummy in this regression is insignificantly different from zero, and the F-statistic presented in Row (4) of Panel C indicates that we can not reject equality of the coefficients of WK and HOST. These findings are consistent with the notion that targets of white knights earn about the same announcement period abnormal returns as targets of non-white-knight multiple bidders.(18)

The results suggest that white knight acquisitions generally lead to greater bidding firm shareholder losses (particularly, in comparisons of white knight versus friendly non-white-knight bidders) and greater target shareholder gains (due to competition) than do non-white-knight acquisitions. Bidder competition plays a key role in the determination of wealth effects experienced by white knights and their targets. White knight bidder losses apparently represent a unique white knight effect that is not found among other subsequent bidders as a group, while gains to targets of white knights seem to stem from bidder competition in general.

B. Overbidding Revisited

The overbidding hypothesis presumes a wealth transfer from the shareholders of the white knight to the shareholders of the target. If one assumes that the assets of the bidding firm are, on average, correctly priced, the empirical results shown in Exhibits 2 through 5, taken together, provide strong support for the overbidding hypothesis. White knight bids generate the largest announcement period bidder losses and target gains observed among the three target management response groups. However, white knight losses at the announcement could reflect overbidding for target assets, a downward revaluation of the white knight's assets in place, or both. In this section, the relative sizes of the target and bidder are examined to gain insight into whether white knight losses reflect revaluations of existing assets or overbidding. If white knight acquisitions involve predominantly small targets, then it would be unlikely that white knight losses are driven by overbidding. In such cases, a large bidding firm price effect at the announcement must reflect factors other than the acquisition itself.

Relative size is computed as the ratio of target to bidding firm common stock market values, observed one day preceding the beginning of their respective cumulation periods for the measurement of abnormal returns (i.e., firm size is measured on day -2 for bidders and day -41 for targets). The value of target shares already held by the acquiring firm is omitted from the computation of the pre-announcement market value of target common stock.

Median relative sizes for matched pairs of targets and bidders in ultimately successful acquisitions are presented in Exhibit 6. The median relative size is 0.179 for white knights, 0.112 for friendly non-white knights, and 0.367 for hostile bidders. Thus, white knight acquisitions do not comprise the smallest acquires when compared to hostile and friendly non-white knights, and so overbidding can not be ruled out on this basis as at least a partial explanation for the differential white knight losses documented in this study. However, revaluation of existing assets may also provide a partial explanation for white knight losses. A WLS regression of announcement period abnormal bidder returns (BCPE) among white knights on the relative size measure (results not shown) yielded an insignificant relation between these two variables, indicating that white knight losses are not sensitive to the relative size of the target. Similar results were obtained using an alternative procedure in which the data were ranked on relative size, the middle third deleted, and BCPE was averaged among observations in the smallest third and the largest third, separately. White knight losses at the announcement are similar in magnitude for both groups, approximately three to tour percent, and are statistically significant at better than the five percent level, although not significantly different from each other. Despite the strength of the evidence concerning overbidding, other factors probably drive white knight losses in the smallest size group. It seems likely that both overbidding and revaluation of existing assets are present among white knights. C. Analysis of Aggregate Synergies

Exhibit 6 contains summary statistics for proportional and dollar synergies (SYNERGY and $SYNERGY, respectively) for matched pairs of bidders and targets involved in ultimately successful acquisitions. For the full sample, the median dollar synergy is $13,879,490 (mean = $46,305,760), which corresponds to a median value of 3.1% when measured as a proportion of the summed pre-announcement target and bidding firm common stock market values (mean = 4.4%).(19) Median dollar synergies are significantly positive for all three management response groups, suggesting that even if overbidding is present, target gains do not come entirely at the expense of bidding firm shareholders. Significance tests of differences in synergies across the three groups are shown in Panel B of Exhibit 6. Acquisitions by both hostile bidders and white knights generate significantly greater dollar and proportional synergies than acquisitions by friendly non-white knights. Although the Wilcoxon rank sum Z-statistics indicate that the difference in synergies across white knight and hostile acquisitions is not significant, hostile TABULAR DATA OMITTED acquisitions generate the largest synergies of the three groups, contradicting the maximum synergy hypothesis.

Similar to the regressions reported earlier in this study, the measure of total synergies as a proportion of pre-acquisition summed target and bidder equity values (SYNERGY) is regressed on variables reflecting target management response, payment method and bidder competition:

SYNERGY = ||Alpha~.sub.0~ + ||Alpha~.sub.1~WK + ||Alpha~.sub.2~HOST + ||Alpha~.sub.3~CASH + ||Alpha~.sub.4~MIX + ||Alpha~.sub.5~MULT + ||Alpha~.sub.6~TNDOFR, (12)

where WK, HOST, CASH, MIX, MULT and TNDOFR are defined as in Sections III.A.2.a. and III.A.2.b., and are coded such that they represent the characteristics of the ultimately successful bid for the target.

TABULAR DATA OMITTED

The WLS regression results for the dependent variable SYNERGY are shown in Exhibit 7.(20) In Regression (1), the coefficients on WK, HOST, CASH and MIX are positive and statistically significant, indicating that each of these factors increases synergies over and above the level observed for friendly non-white-knight stock bids, which are picked up by the intercept. In Regressions (2) and (3), which contain the multiple-bidder dummy MULT and in Regression (4), which is run on multiple-bidder events only, the significance of the white knight dummy variable disappears. Thus, synergies generated in white knight acquisitions ate about the same, on average, as in non-white-knight acquisitions involving multiple bidders.(21) F-tests are unable to reject equality of the coefficients on WK and HOST in any of the regressions presented in this exhibit.

The addition of the tender offer dummy variable in Regression (3) eliminates the significance of the hostile dummy variable. Closer inspection of the data revealed that of the 31 successful hostile bids shown in the bidding firm sample in Exhibit 1, only six were not tender offers. The TNDOFR and HOST variables in Regression (3) probably proxy for the same underlying effect.

The evidence presented in this exhibit indicates that the univariate result from Exhibit 6 of greater synergies in white knight than friendly non-white-knight acquisitions is driven by bidder competition. Acquisition by a white knight does not generate significant additional synergies over and above those associated with bidder competition in general. Although white knight acquisitions generate positive average synergies, the regressions presented in Exhibit 7 do not support the maximum synergy hypothesis.

IV. Conclusion

This study empirically examines the characteristics of white knight acquisitions in comparison with hostile and friendly non-white-knight acquisitions. The sample examined in this study, which includes the bid histories surrounding ultimately successful acquisitions consummated during the 1974-1984 period, is characterized by positive synergies, on average, and this observation is true for all of the segmentations of the data on the basis of target management response to the bid. White knight acquisitions are not a zero-sum game, as suggested by Roll's |21~ hubris hypothesis. Since white knight synergies are positive, benefits to target shareholders do not come entirely at the expense of bidding firm shareholders.

However, the empirical results are strongly supportive of overbidding by white knights (which is differentially severe relative to non-white-knight bidders). White knight losses represent a unique effect of white knight status on bidding firm shareholder wealth, which is not found among other subsequent bidders as a group. The results show that previous findings by Bradley, Desai, and Kim |7~ and Boebel and Harris |5~ of greater announcement period losses to subsequent bidders than first bidders are entirely attributable to white knights. Bradley, Desai, and Kim surmise that their result is driven by white knights, and their reasoning is confirmed by this study.

The prediction of the maximum synergy hypothesis that white knight acquisitions generate greater synergies than non-white-knight acquisitions does not receive strong support. First, there are no significant additional synergies associated with acquisition by a white knight once bidder competition is controlled. Second, although Shleifer and Vishny |24~ predict that targets are hostile to lower-synergy bidders and friendly to higher-synergy bidders (i.e., white knights), the evidence shows that white knight synergies are insignificantly different from synergies generated in hostile acquisitions. These results suggest that many bidding firms which become white knights take on acquisitions which are not in their shareholders' best interests. Roll |21~ suggests that the benefits to bidding firm shareholders of the firm's involvement in an acquisition are elusive, and perhaps nonexistent. Future work must address the white knight's motivation for participating in acquisitions by examining the performance of the post-acquisition firm. The distinction between the 45% of the white knight sample which became targets, restructured or implemented takeover defenses during the five years after becoming white knights and the 55% which experienced none of these events provides a starting point for this post-acquisition analysis. Towards this end, it would also be useful to document the frequencies of financial distress, executive turnover and divestitures of target assets by the post-acquisition firm. 1 Jensen and Ruback |14~ conclude that shareholders of bidding firms neither gain nor lose on average, earning a normal rate of return on their investment in the target. Jarrell, Brickley, and Netter |12~ conclude that at least for acquisitions via tender offer, gains to bidding firms are small, with the bulk of acquisition synergies accruing to target shareholders. They also conclude that shareholders of bidding firms which defeat a competing bidder are as likely to lose as to gain from the transaction. Studies by Huang and Walkling |10~, Travlos |26~, Asquith, Bruner, and Mullins |2~, and Franks, Harris, and Mayer |8~ show that the returns to bidding firm shareholders are significantly negative if the acquisition is equity-financed, and either zero or slightly positive if the acquisition is cash-financed. Jarrell and Poulsen |13~ show that gains to tender offer bidders have declined over the period from 1962 to 1985, with significant bidder gains observed for 1960s tender offers and insignificant losses observed for those from the 1980s.

2 Indeed, Bhagat, Shleifer, and Vishny |4~ analyze hostile takeovers during the period 1984-1986, and find that white knight acquisitions lead to fewer layoffs (blue collar and white collar) than successful hostile acquisitions or cases in which the target remained independent.

3 One caveat must be considered before attributing the results described above to overbidding. An acquisition offer may reveal information to the market about the true value of bidding firm assets as well as revealing information about the target and/or any synergies which would be created by the proposed combination of the two firms. The negative abnormal stock price response observed for white knights upon entering a bidding contest could be due to overbidding, or a negative revision in the value of the white knight's assets, or both. This issue is addressed in Section III.B; the results indicate that overbidding can not be ruled out as at least a partial explanation for the findings of this study. 4 For successful mergers, the resolution date is generally the day on which target shareholders approved acquisition by the bidder. For successful tender offers, the resolution date is the day on which the bidder was announced to have obtained at least 80% of target shares outstanding. For unsuccessful attempts, the resolution date is the date on which the bidding firm withdraws from trying to acquire the target, or some other bidding firm successfully acquires the target.

5 Control of target assets is conservatively defined here as controlling 80% or more of target common stock because many firms now have provisions in their corporate charters requiring supermajority assent for a merger to be approved. 6 Six of the 10 unsuccessful white knights were outbid and one withdrew its bid voluntarily. For the remaining three observations, information from the Wall Street Journal is inconclusive as to whether the white knights were outbid. 7 Prior studies have suggested that when the bidder has an acquisition program, part of the value of any of its acquisitions is anticipated and capitalized at the outset of the program. Schipper and Thompson |23~ find that announcements of acquisition programs are associated with positive and significant abnormal returns. Malatesta and Thompson |16~ find that acquisition announcements by firms with ongoing programs are partially anticipated. However, Malatesta and Thompson conclude that investors do not perfectly anticipate the timing of announcements, and past acquisitions impart little information about future acquisitions. Asquith, Bruner, and Mullins |1~ find that the announcement period abnormal returns to bidder shareholders are of approximately equal magnitude for each of the first four mergers in an acquisition program. Asquith, Bruner, and Mullins' evidence suggests that the bidder shareholder wealth effect measure used in this study (BCPE) probably does not seriously understate the share price response at the announcement for bids which are part of an ongoing acquisition program. In addition, numerous prior studies have computed bidding firm share price responses to the acquisition announcement which do not adjust for partial anticipation and acquisition programs (e.g., Asquith, Bruner, and Mullins |2~, Bradley, Desai, and Kim |6~, |7~, Huang and Walkling |10~, and Travlos |26~).

8 The target results reported in this study are robust to using shorter event windows beginning at day -15 and day -5 relative to the first acquisition announcement.

9 There were 15 bidding firm observations for which the day -2 share price was missing. For these firms, dollar gains are computed by multiplying BCPE by the market value of the bidder's common stock at day -3.

10 Bradley, Desai, and Kim |6~, |7~, and Franks, Harris, and Mayer |8~ document the effects of acquisition method and bidder competition on abnormal gains to bidder and target shareholders. The correlation between competition, acquisition method and white knight status is apparent from Exhibit 1.

11 T-tests for differences in $BCPE, $TCPE and $SYNERGY across target management response groups are not presented, since these dollar gains variables are not normally distributed. Instead, the nonparametric Wilcoxon rank sum test is used to compare these variables across white knight, hostile and friendly non-white-knight classifications.

12 Data on the method of payment are from the Wall Street Journal and the Commerce Clearing House Capital Changes Reporter.

13 All of the bidding firm tests were replicated using a dummy to distinguish single-bidder from multiple-bidder events, and the results were virtually identical to those reported here using the first/subsequent dummy.

14 The time period of this study (1974-1984) avoids the need to control for time series shifts in acquisition gains due to the passage of the Williams Act in 1968 (Jarrell and Bradley |11~). However, previous studies have found time shifts in target and bidder gains during the sample period. Nathan and O'Keefe |19~ find that merger and tender offer premia experienced an upward shift in 1974, and Bradley, Desai, and Kim |7~ allow target and bidder abnormal returns to shift in the 1981-1984 period. Inspection of mean standardized target and bidder gains, segmented by announcement year (not reported), reveals that indeed, the mean standardized abnormal bidder gain is significantly smaller for bids announced in the 1981-1984 period than in the earlier years. Also the mean standardized target gain is unusually high for bids announced in 1974, consistent with Nathan and O'Keefe, although the data do not indicate a shill in standardized target gains in the 1981-1984 period. The test results presented in Exhibits 4 and 5 have been replicated including a dummy variable for 1981 to 1984 in the bidder regressions and a dummy for 1974 in the target regressions, with virtually no effect on the other coefficients. Thus, the regressions presented in this study do not include variables to control for time period. 15 In a multiple bid event, the characteristics of the revised bid may differ from the first bid, in terms of payment method or acquisition method. In such cases, the variables CASH, MIX and TNDOFR are classified according to the characteristics of last bid made by the bidder for its target.

16 In the WLS estimation, each row j of the data matrix is multiplied by:

1/||Sigma~.sub.j~|square root of |L.sub.j~~,

where

||Sigma~.sub.j~ = Estimation period residual standard deviation for bidding firm j.

|L.sub.j~ = Number of trading days in bidding firm j's cumulation of BCPE.

Computation of the F-statistics testing the null hypothesis that the coefficients of WK and HOST are equal (shown in Panels C of Exhibits 4, 5 and 7) is described in Johnston |15, pp. 198, 199~.

17 Ultimate bid success is not a significant determinant of BCPE, presumably because the window over which BCPE is computed is short and does not impound information about the bid outcome. When the regressions from Panel A, Exhibit 4 are run on ultimately successful bids only, the results are similar to those reported in the paper for the full sample of bidders.

18 In contrast to the bidder results from Exhibit 4, white knights do not account entirely for the effects of bidder competition on target gains. When white knights are deleted from the sample, a single-/multiple-bidder effect on TCPE persists (results not reported).

19 These mean values are smaller than the mean dollar synergy of $117,110,000 and proportional synergy of 7.43% reported by Bradley, Desai, and Kim |7~. Bradley, Desai, and Kim use a somewhat different methodology and a longer time period than this study, and they confine their attention to acquisitions via tender offer.

20 Sources of heteroskedasticity affecting SYNERGY include factors affecting the variance of TCPE and BCPE: the number of days over which prediction errors are cumulated for each target and bidder, and the estimation period residual standard error of each target and bidder. Thus, WLS is used to estimate the regression for SYNERGY, with weights equal to:

|Mathematical Expression Omitted~,

where ||Sigma~.sub.j~, ||Sigma~.sub.k~ equal estimation period residual standard errors for bidder and target, respectively, and |L.sub.j~, |L.sub.k~ equal the number of days in cumulation for BCPE and TCPE, respectively.

Using these weights assumes that the covariance between TCPE and BCPE is zero, which is unlikely. However, it is probable that these WLS weights adjust for a significant portion of the heteroskedasticity in the variable SYNERGY.

21 White knights do not account entirely for previous findings that bidder competition is associated with higher synergies (Bradley, Desai, and Kim |7~ and Boebel and Harris |5~). When white knights are deleted from the sample, a synergy difference between single- and multiple-bidder non-white-knight acquisitions persists (results not reported).

References

1. P. Asquith, R.F. Bruner, and D.W. Mullins, Jr., "The Gains to Bidding Firms from Merger," Journal of Financial Economics (April 1983), pp. 121-139.

2. P. Asquith, R.F. Bruner, and D.W. Mullins, Jr., "Merger Returns and the Form of Financing," Unpublished Working Paper, Harvard University, 1987.

3. A. Banerjee and J.E. Owers, "Wealth Reduction in White Knight Bids," Financial Management (Autumn 1992), pp. 48-57.

4. S. Bhagat, A. Shleifer and R.W. Vishny, "Hostile Takeovers in the 1980s: The Return to Specialization," in Brookings Papers on Economic Activity: Microeconomics, M.N. Baily and C. Winston (eds.), Washington, D.C., Brookings Institution, 1990, pp. 1-84.

5. R.B. Boebel and R.S. Harris, "Competition in the Bidding Market and the Shareholder Wealth of Target and Acquiring Firms," Unpublished Working Paper, University of Virginia, 1989.

6. M. Bradley, A. Desai, and E.H. Kim, "The Rationale Behind Interfirm Tender Offers: Information or Synergy?," Journal of Financial Economics (April 1983), pp. 183-206.

7. M. Bradley, A. Desai, and E.H. Kim, "Synergistic Gains from Corporate Acquisitions and Their Division Between the Stockholders of Target and Acquiring Firms," Journal of Financial Economics (May 1988), pp. 3-40.

8. J.R. Franks, R.S. Harris, and C. Mayer, "Means of Payment in Takeovers: Results for the United Kingdom and the United States," in Corporate Takeovers: Causes and Consequences, A.J. Auerbach (ed.), Chicago, University of Chicago Press, 1988.

9. R.M. Giammarino and R.L. Heinkel, "A Model of Dynamic Takeover Behavior," Journal of Finance (June 1986), pp. 465-480.

10. Y. Huang and R.A. Walkling, "Target Abnormal Returns Associated with Acquisition Announcements: Payment, Acquisition Form, and Managerial Resistance," Journal of Financial Economics (December 1987), pp. 329-349.

11. G.A. Jarrell and M. Bradley, "The Economic Effects of Federal and State Regulations of Cash Tender Offers," Journal of Law and Economics (October 1980), pp. 371-407.

12. G.A. Jarrell, J.A. Brickley, and J.M. Netter, "The Market for Corporate Control: The Empirical Evidence Since 1980," Journal of Economic Perspectives (Winter 1988), pp. 49-68.

13. G.A. Jarrell and A.B. Poulsen, "The Returns to Acquiring Firms in Tender Offers: Evidence from Three Decades," Financial Management (Autumn 1989), pp. 12-19.

14. M.C. Jensen and R. Rubark, "The Market for Corporate Control: The Scientific Evidence," Journal of Financial Economics (April 1983), pp. 5-50.

15. J. Johnston, Econometric Methods, New York, McGraw-Hill, 1972.

16. P.H. Malatesta and R. Thompson, "Partially Anticipated Events: A Model for Stock Price Reactions with an Application to Corporate Acquisitions," Journal of Financial Economic's (June 1985), pp. 237-250.

17. M.L. Mitchell and K. Lehn, "Do Bad Bidders Become Good Targets?," Journal of Political Economy (April 1990), pp. 372-398.

18. R. Morck, A. Shleifer, and R.W. Vishny, "Do Managerial Objectives Drive Bad Acquisitions," Journal of Finance (March 1990), pp. 31-48.

19. K.S. Nathan and T.B. O'Keefe, "The Rise in Takeover Premiums: An Exploratory Study," Journal of Financial Economics (June 1989), pp. 101-119.

20. J.M. Patell, "Corporate Forecasts of Earnings Per Share and Stock Price Behavior: Empirical Tests," Journal of Accounting Research (Autumn 1976), pp. 246-276.

21. R. Roll, "The Hubris Hypothesis of Corporate Takeovers," Journal of Business (April 1986), pp. 197-216.

22. R. Ruback, "Assessing Competition in the Market for Corporate Acquisitions," Journal of Financial Economics (April 1983), pp. 141-153.

23. K. Schipper and R. Thompson, "Evidence on the Capitalized Value of Merger Activity for Acquiring Firms," Journal of Financial Economics (April 1983), pp. 85-119.

24. A. Shleifer and R.W. Vishny, "Greenmail, White Knights, and Shareholders' Interest," Rand Journal of Economics (Autumn 1986), pp. 293-309.

25. A. Shleifer and R.W. Vishny, "Value Maximization and the Acquisition Process," Journal of Economic Perspectives (Winter 1988), pp. 7-20.

26. N.G. Travlos, "Corporate Takeover Bids, Methods of Payment and Stockholders' Returns," Journal of Finance (September 1987), pp. 943-963.

27. N.P. Varaiya and K.R. Ferris, "Overpaying in Corporate Takeovers: The Winner's Curse," Financial Analysts' Journal (May-June 1987), pp. 64-70.

Cathy M. Niden is an Assistant Professor of Finance, College of Business Administration, University of Notre Dame, Notre Dame, Indiana.