Defence versus Offence: Disclosure and Media in Takeovers

This paper analyzes a target firm’s decision to voluntarily disclose information during a takeover event and the effect of such disclosures on the outcome of the takeover. In the model the acquirer may also run a media campaign. The model predicts that a voluntary disclosure of positive information by the target decreases the likelihood that the takeover succeeds. The empirical analysis confirms this prediction by showing that positive earnings forecasts by target firms during takeover events increase the probability of takeover failure. Overall, it is shown that information dissemination through voluntary disclosures by target firms is an important factor affecting takeover outcomes.


Introduction
Some takeovers may be classified as win-lose games played between the target firm and the acquirer. This is especially true for hostile takeovers. Various tactics are intended to lure the shareholders to their own sides. They do so because target shareholders determine the outcome of a takeover by voting. One of the frequently employed tactics by the acquirer is the financial media. Buehlmaier (2011) both theoretically and empirically shows that financial press coverage about the acquirer can predict takeover outcomes. In particular, positive media content about the acquirer gives rise to takeover success. On the other hand, one of the popular defence weapons employed by the target is voluntary disclosures in the form of earnings/profit forecasts. Target firms send these profit forecasts to shareholders with other takeover documents. The aim may be to show that the shares are worth more than the bid price or to illustrate that the current management is better at running the company than the potential acquirer. Gray et al. (1991) give evidence from the UK that more forecasts are voluntarily disclosed during hostile takeovers. Similarly, Sudarsanam (1994) finds that targets disclose a profit forecast in 45% of hostile and competing bids in UK. There are also some well-known examples of this phenomenon from the US, like Avon Products Inc. forecasting good news after a hostile takeover bid by Amway in 1989 which finally failed (Ruland et al., 1990). Thus, an interesting question is whether profit/earnings forecast disclosures, which seem to be preferred by target firms as a defence instrument, really have an effect on the takeover outcomes. This paper analyzes a target firm's decision to disclose financial information during a takeover event and the effect of these disclosures on the takeover outcome. In the model the acquirer may also run a media campaign to affect the shareholders' perception. The target firm may make informative or uninformative disclosures. The so-called informative disclosure in the model may correspond to quantitative and realistic management earnings forecasts by the target firms. This type of disclosures is an effective tool to communicate the real type of the target to shareholders. In contrast, an uninformative disclosure does not give much information. One may consider it as not disclosing or providing qualitative and 'soft' statements that are not easy to interpret. This is consistent with the idea of no disclosure does not necessarily imply keeping 'bad news'. Hence, uninformative disclosure provides noisier information about the real type of the target compared to the first alternative.
The economic intuition behind the model is as follows. Target shareholders are not always perfectly aware of the company's real worth and also do not know whether a takeover leads to value creation or destruction. Shareholders are going to approve a takeover if they only believe that in expectation the target firm value with the takeover is higher than the value without the takeover. Hence, voluntary disclosures of the target firm and financial news about the acquirer alleviate the information asymmetry problem of the shareholders. The target management chooses its disclosure policy by taking into account its interest to show the firm value under its control as valuable as possible either to maintain its private benefits of control or to increase the offer price. Thus, a high-value target has the incentive to release as much information as possible about its type. The high-value target thus makes an informative disclosure to distinguish itself from the low type. However, the low-value target that does not want to be identified but cannot provide completely false information, prefers uninformative disclosures in order to create a noisy signal and confuse shareholders. Shareholders still pay attention to voluntary disclosures since they know that the high-value target makes an informative disclosure that contains rather precise information. Shareholders also follow the financial press about the acquirer because the good and the bad type of acquirers play different media strategies. In particular, the good type runs a media campaign to separate itself from the bad type, while the bad type does not have any incentives to mimic the good type when the media campaign involves high costs.
We argue that both voluntary disclosures by the target and the financial news about the acquirer play a role in takeovers. In particular, the success probability of a takeover decreases with a positive disclosure signal and increases with a positive media signal. That is the case because shareholders know that a positive disclosure signal occurs due to the informative disclosure of the high-value target, and a positive media signal comes from the media campaign of the good type of acquirer. This consideration immediately yields the following empirical prediction: while disclosures that include positive news about the target decreases the likelihood of takeover success, disclosures with positive information in the financial media about the acquirer increases this likelihood.
This main prediction of the model is confirmed by an empirical investigation. We use the complementary log-log model specification and the 'rare events procedure' as a robustness check since the takeover outcome is a binary variable with an uneven distribution: less failed takeovers compared to the successful ones. We attain the disclosure variable by identifying the target firms in the sample that disclose positive news in the form of increasing management earnings forecasts for future years during the takeover event. The results confirm that the availability of positive news through management earnings forecasts has a significant and negative effect on the likelihood of takeover success. Including disclosure variable as an explanatory variable leads to also higher goodness of fit. Finally, the analysis also confirms that positive media content about the acquirer improves takeover success supporting Buehlmaier (2013).
The paper shows the impact of information dissemination on the likelihood of takeover success along with deal and firm characteristics that the previous literature has identified. Differently than Buehlmaier (2013), it considers the possibility of information dissemination by the target firm through voluntary disclosures. As it is with the acquirer, the information dissemination tool of the target firm also affects the takeover outcome. This paper demonstrates that the effect of voluntary disclosures in the form of earnings forecasts is not only limited to an increase in offer prices as it is shown in Brennan (1999), but it also reduces the takeover success probability. It is also observed that the voluntary disclosure practices of target firms during a takeover event seem to be in accordance with the general pattern of voluntary disclosures made in routine situations: while 'good news' is given by quantitative forecasts, target firms are likely to share 'bad news' with qualitative statements. Last but not least, the paper is also closely related to the emerging literature that analyzes the link between financial markets and corporate takeovers. Edmans, Goldstein, and Jiang (2011) show that a non-fundamental undervaluation of the target firm creates a profit opportunity for acquirers and so triggers takeovers. Our paper supports this view by showing that voluntary disclosures in the form of management earnings forecasts can be an effective way to alleviate such undervaluations for target firms. Our finding is also consistent with Safieddine get a noisy signal s ∈ 0,1 on the occurrence of a media campaign through the press. The precision of the signal is given by, δ P s 0|m 0 P s 1|m 1 1 2.
In addition, shareholders receive a noisy signal s d ∈ h, l about the type of the target based on its disclosure policy. The probability that the shareholders observe the correct signal about the type of the target is higher if the target makes an informative disclosure. In particular, where 0 ϵ 1 2. Hence, the precision of the signal increases with an informative disclosure. The reader may think of an informative disclosure as a quantitative accounting statement like a management's earnings forecast by which the target tries to convey its type. Alternatively, the target may provide no statements at all or may assert some vague qualitative statements about its future prospects.
Shareholders then update their priors by using all recently available information revealed by both signals. Accordingly, their posterior beliefs about the acquirer and the target after observing the signal realizations are denoted by, respectively, β 1 P τ G|s 1 and β 0 P τ G|s 0 , α h P t H|s d h and α l P t H|s d l .
The purpose of noisy signals is to have a more realistic model. There is always the chance that target shareholders misinterpret what they read in the voluntary disclosure documents. But the misinterpretation risk gets smaller with the informativeness of the target´s disclosure. A similar comment is applied to the media signal. Target shareholders may think that the media content that they read is due to a media campaign when no such media campaign was initiated (and vice versa).
At the final stage of the model, shareholders decide whether or not to tender their shares in a simultaneous-move game. Each shareholder owns one share. The acquirer needs to obtain at least k shares to get control of the company. Otherwise, the takeover fails. Conditional on having observed the noisy disclosure signal s d , the shareholders expect the target's share price to be worth p s d , where under the current management. Moreover, having also observed the media signal s, they expect the share price after a successful takeover to be where s ∈ 0,1 , s d ∈ h,l and P .,.|s d ,s are the joint posterior probabilities about the types of the acquirer and the target. Accordingly, if a shareholder does not tender and the takeover goes through, her expected payoff is p s,s d p s d , whereas if she tenders and the takeover goes through her expected payoff is b p s d . If the takeover fails, each shareholder obtains a zero payoff. The acquirer obtains non-monetary private benefits of control z by taking over the target. Even the bad type-B acquirer has an incentive to take over the target even though it will destroy value since in expectation it benefits: The final payoff of the acquirer's management is then where j is the number of shares tendered by the shareholders and 1 is the indicator function. On the other hand, the target management loses its private benefits of control y in case the takeover succeeds.

Model Solution
Our first lemma determines how target shareholders utilize the new information conveyed through both the target's disclosure policy and the acquirer's media campaign to update their beliefs to, β s =P(τ=G|s) and α s d =P (t=H|s d    takeover would fail and they would obtain a zero payoff. (iv) Last but not least, suppose that , . Then, there are two different equilibria. Either all shareholders tender or no shareholder tenders. From the equilibrium with no shareholder tendering, noone has an incentive to deviate because if she instead tendered, the takeover would still fail and the payoff would stay the same. Similarly, noone has an incentive to deviate from the equilibrium where all shareholders tender. If a shareholder were to deviate and not tender, the takeover would still succeed but she would obtain a lower payoff. The last equilibrium with each shareholder tendering shares a similar logic with the well-known bank run equilibrium of Diamond and Dybvig (1983). However, the equilibrium with no shareholder tendering is more sensible in a takeover context, especially considering the fact that shareholders obtain negative payoff if they all tender. The paper concentrates on the type of equilibrium with no shareholder tendering.
Given the four different scenarios, one may deduce that target shareholders tender k shares if , or if , . It is clear that the acquirer has no incentives to bid less than since doing so would lead to failure of the takeover. If the acquirer bids more than , the takeover succeeds. But the acquirer can do better by lowering the bid to . Lemma 3 below shows formally that the acquirer's utility is a decreasing function of the bid price for .
Lemma 3. The acquirer bids optimally * in equilibrium.
Lemma 3 implies that shareholders care only whether the expected posterior price after a successful takeover, p s,s d , exceeds the expected price under the current management, . In other words, shareholders tender k shares and the takeover succeeds if , . On the other hand, no shareholder tenders and the takeover fails if , . This result is intuitive: Shareholders compare the expected values of the target with and without the takeover. They approve the takeover only if they feel certain that in expectation the target's value with the takeover is higher than its value without the takeover.
Another implication of Lemma 3 is that target shareholders tender at most k shares. This means that the acquirer's final payoff is at most , by recalling equation 2 . The model would be interesting only if the cost of the media campaign is lower than the maximal expected amount that the acquirer gains from the takeover. Otherwise there would be no incentives for any type of the acquirer to run a media campaign. In this respect, suppose that

Equilibria
This section determines the optimal disclosure and media campaign decisions by the target and the acquirer. Define , and also the lower threshold for the cost of the media campaign, where c c since p t,B p t,G .
Proof. It is clear that β l,1 β h,1 when α 0 and α 1. For the values of 0 α 1 there exists a positive difference between β l,1 and β h,1 , which follows from δ ∈ 1 2,1 , ∈ 0,1 2 , p t,B p t p t,G and for all values of 0 α 1 given that all previous conditions are true.
One may interpret these four different β thresholds presented above as the different levels of shareholders' prior www.ccsenet.org/ijef International Journal of Economics and Finance Vol. 7, No. 3;2015 belief or, in other words, their optimism (pessimism) about the acquirer. Shareholders are most pessimistic about the acquirer in the region β ∈ 0, β l,1 because they are quite certain that they face a bad type. On the other hand, shareholders are most optimistic about the acquirer if β ∈ β h,0 ,1 . In the middle range β ∈ β l,1 , β h,0 , shareholder's uncertainty about the acquirer's type is relatively high and while their pessimism increases moving in the direction of β l,1 , their optimism increases in the direction of β h,1 . The interpretation of the thresholds for the cost of the media campaign is as follows: For the existence of a separating media equilibrium by the acquirer in general (Note 1), the cost of the media campaign should lie in an intermediate range (c ∈ c, c ). A media campaign should be expensive enough that the bad B type of the acquirer is not able to afford it. It does not pay off to spend money for the expensive media campaign, since the costs of a media campaign together with the destruction in the target's value by the bad B type surpass the private benefits of control. On the other hand, the cost of a media campaign should not be too high, so that it does not even pay off for the good G type of acquirer.
The next theorem states the conditions and the characteristics of a separating equilibrium both by the target and the acquirer. Under this separating equilibrium, the information dissemination tools of both the acquirer and the target have an effect on the takeover outcome. In other words, both voluntary disclosures by the target and financial news about the acquirer play an important role. The high-value target H strictly prefers to make an informative disclosure if shareholders are relatively uncertain about the acquirer's type, i.e. β ∈ β l,1 , β h,0 . In contrast, the low-value target L chooses noisy or non-informative disclosure to confuse shareholders. However, shareholders still pay attention to voluntary disclosures since they know that the high-value target H makes an informative disclosure. If an informative disclosure involved (high) costs and became unaffordable for the high-value H target, then voluntary disclosures would have no effect on the takeover outcome. Yet it is plausible to assume that releasing a voluntary disclosure, containing useful information about the company, has a negligible cost. Moreover, now due to the informational effects of voluntary disclosures by the target, the media campaign should cost less so that the good G type of the acquirer still finds running a media campaign worth paying for (see below c 1 c ). To say it differently, now the cost of the media campaign should be less for the existence of a separating media equilibrium by the acquirer since the voluntary disclosure of the target firm is also informative, which decreases the marginal benefit of a media campaign. The underlying reason is that shareholders get informed about the target firm by looking at the disclosure signal and move already to influence the takeover result. The important take-away of this theorem is that while positive media content about the acquirer may entail takeover success, positive disclosure about the target may decrease its likelihood.
• For β ∈ β l,1 , β h,1 the takeover succeeds only after s d l,s 1 and fails otherwise.
• For β ∈ β h,1 , β l,0 the takeover succeeds after s d l,s 1 , s d h,s 1 and fails after s d l,s 0 , s d h,s 0 .
• For β ∈ β l,0 , β h,0 the takeover fails only after s d h,s 0 and succeeds otherwise. First, evaluate the separating equilibrium, * , * 1,0 and * , * 1,0 . Shareholder uncertainty about the acquirer is high β ∈ β l,1 , β h,0 . The information that the shareholders obtain through voluntary disclosures is especially important because they face great uncertainty about the acquirer's type. If the shareholders were instead quite certain that the acquirer was a good type G (bad type B) of acquirer, they approved (prevented) the takeover already and the voluntary disclosures were pointless. The high-value target H thus prefers an informative disclosure to separate itself from the low-value target L, which opts for the non-informative disclosure to create noise and to confuse shareholders. Additionally, the acquirers play a separating media equilibrium now in a smaller region of the cost of a media campaign, which has moved towards the left due to the information dissemination through voluntary disclosures by the target. The bad B type of the acquirer has no incentives to mimic the good G type also in this region (c c c 1 ), as it has been before. As a result, shareholders learn from both disclosure and media signals. Shareholders believe that they deal with a high-value target H after observing a positive disclosure signal s d h and a good type of acquirer after observing a positive media signal 1. Shareholders' beliefs affect the outcome in return. That is why a positive disclosure signal causes takeover failure since acquisition destroys value for the high-value target . On the other hand, a positive media signal triggers takeover success since being taken over by a good type G of acquirer increases value.
Consider now the case when shareholders are pretty sure about the type of acquirer ( ∉ , , , Vol. 7, No. 3;2015 outcome together with the financial press coverage about the acquirer. In particular, a voluntary disclosure containing positive prospects ( ) about the target decreases the likelihood that the takeover succeeds. Empirically, target firms may release management's earnings forecasts for the upcoming periods in their voluntary disclosure documents. Those target firms that announce increasing earnings forecasts compared to the previous year's earnings are the ones providing positive prospects. On the other hand, those firms that announce decreasing earnings forecasts (not expected in equilibrium) and/or qualitatively negative forward-looking statements about the next periods are the ones providing negative news ( ). The likelihood that the takeover fails (succeeds), is expected to be higher for those firms with positive (negative) news. On the other hand, it is not very uncommon that some target firms do not provide either earnings forecasts or qualitative statements. In the following empirical analysis, those targets with no disclosure are classified together with the firms announcing negative news. By doing this, the empirical analysis focuses on determining the sole effect of positive disclosures on the likelihood of takeover success.

Data and Empirical Analysis
Takeover data consists of takeover attempts with announcement dates between January 1, 2000 and December 31, 2006 from the SDC Platinum Mergers & Acquisitions database as in Buehlmaier (2011). Deal and firm characteristics of targets and acquirers (Note 2) are also from SDC and they are standard control variables used in the literature. The prediction that positive news about the acquirer in the financial press leads to takeover success is kept and further tested in the analysis by including the media variable of Buehlmaier (2011). Data on voluntary disclosures required to test the main prediction of the current paper, which is positive prospects about the target decreases the takeover success probability, are obtained from Dow Jones Factiva and the SEC EDGAR Database. The details and the summary statistics of deal and firm characteristics as well as the details regarding how the voluntary disclosures data are obtained could be found in Orhun (2013).
The general binary outcome model is suitable in order to check the main empirical prediction of the model since takeover outcome is a binary variable: 1 for completed takeovers and 0 for failed ones. In particular, we choose complementary log-log model specification. This specification is preferred if binary outcome data is unevenly distributed: the positive (or negative) outcome is rare. The number of failed takeovers is much smaller compared to the number of successful ones: 28 failed vs. 286 successful takeovers. (Orhun, 2013) Complementary log-log models capture this effect by being asymmetric around zero.
The function is the complementary log-log link function, is a 1 vector of control variables and is the 3 1 parameter vector to be estimated. is the media variable of Buehlmaier (2011). is the disclosure variable, which is 1 for target firms that provide increasing earnings forecasts ("positive disclosure") and 0 for target firms with "no disclosure" and "negative disclosure" as it is explained in Orhun (2013). The parameter shows the effect of positive (news) disclosure by the target firm in the form of management's earnings forecasts on takeover outcome. If the empirical prediction of the model is valid, is negative and significant and also the sample average of the marginal effect is negative.

Results
The model (6) is estimated with maximum likelihood. The control variables include deal characteristics and firm characteristics of target firms and acquirers. In addition, prevtakeovers, which is the number of successful takeovers in the previous 100 days for each takeover attempt in the sample, is also included. The latter variable is necessary to account for the effect of merger waves and macroeconomic factors. Table 1 presents the estimation results of the complementary log-log specification together with the results of the classic probit model to highlight the difference. The table includes the coefficients of the explanatory variables that remain to be statistically significant after a standard stepwise regression procedure. In this procedure, all explanatory variables are included at the start. But at each new step the least significant explanatory variable is dropped and then the model is re-estimated until all explanatory variables are significant. The results of the two different models are very similar. The significance of some variables slightly changes and averages of the sample marginal effects of most variables increase in absolute value with complementary log lof model. McFadden's also increases by 1%. The coefficient of the disclosure variable is negative and is statistically significant at 1% level. This result is supportive of the model's prediction. Last but not least, the marginal effect of disclosure (providing "positive disclosure") changes from -0.08 to -0.09 with complementary log-log model. This implies an increase of one unit of disclosure yields a (larger) decrease of approximately 0.09 units in the probability that the takeover succeeds. The coefficient of media is positive and significant. This means that positive media content about the acquirer increases the probability that the takeover succeeds. The coefficients of all other www.ccsenet.org/ijef International Journal of Economics and Finance Vol. 7, No. 3;2015 remaining variables are as expected (Note 3). The results of the complementary log-log model confirm the main prediction of the model even when the uneven distribution of the takeover data is taken into consideration. One other alternative to the complementary log-log model specification if one deals with such 'rare events' data, is proposed by political scientists Zeng (2001a, 2001b). Political scientists, who generally deal with very unevenly distributed binary dependent variables, with dozens to thousands of times fewer ones (events such as wars, coups, etc.) than zeros (non-events), have become aware that logit and probit models underestimate the probability of an event with very few observations. In their situation with very rare ones (events), 1 will be systematically underestimated. This result occurs due to classification errors: the ability to accurately find a 'cutting point' to distinguish zeros 0| from ones 1| is biased in the direction of favoring zeros at the expense of ones. This result arises naturally since the model has better information about the distribution of zeros than ones and so is better at classifying zeros than ones. To summarize, the probability of the outcome with very few observations is underestimated. This means that in our situation 0 will be underestimated since there are fewer zeros (failures) than ones (successes). This problem especially affects the constant term , which turns out to be biased although the rest of the estimates of ... are consistent. Accordingly, Zeng (2001a, 2001b) outline an alternative procedure to cope with these issues. Their strategy is to select on the dependent variable by collecting all those very few observations available for which 0 (the 'cases') and a random selection of observations for which 1 (the 'controls') (Note 4). In econometrics, this method is called choice-based or case control sampling. Their suggestion is not to collect more than 2-5 times as many ones ('frequent outcome') as zeros ('rare outcome'). In order not to lose many observations, here the upper boundary will be preferred, meaning that 5 times as many ones (successes) as zeros (failures) will be randomly selected. This translates into selecting a nearly 50% random sample of ones (successes). After this step, the procedure of Zeng (2001a, 2001b) prescribes to run a logit analysis on www.ccsenet.org/ijef International Journal of Economics and Finance Vol. 7, No. 3;2015 the new sample and then correct for the bias. The easiest way to correct for the bias of is called prior correction. In this regard, the following prior-corrected estimate is consistent for the constant term : where is the estimate that results from the logistic regression on the newly selected sample, is the fraction of zeros ('rare outcome') in the original data and ȳ is the fraction of zeros ('rare outcome') in the new sample. As stated before, all usual estimates for ... are statistically consistent with case control sampling. The results of the logit regression after case control sampling and prior correction for are presented in Table 2 together with the results of the original logit regression for comparison purposes. The first observation is that the estimated intercept and its significance change considerably as expected. The significance of all other variables also slightly changes and averages of the sample marginal effects of all variables increase significantly in absolute value. As with the previous specifications while the coefficient of disclosure is significant and negative, the coefficient of media is significant and positive. The 'rare events' analysis outlined by Zeng (2001a, 2001b) is quite useful when positive (or negative) outcome is very rare. However, it should be kept in mind that this procedure was originally designed and works best for very large data sets.

Conclusion
This paper tries to find out the role of voluntary disclosures by target firms during a takeover event on the likelihood of takeover success. It approaches to this issue both from theoretical and empirical angles. Target shareholders who determine the outcome of the takeover are not always perfectly aware of the company's real worth. The target firm may provide informative or uninformative disclosures in order to affect the shareholders' www.ccsenet.org/ijef International Journal of Economics and Finance Vol. 7, No. 3;2015 approval decision. In this situation, the high-value target has strong incentives to distinguish itself with an informative disclosure. On the other hand, the low-value target prefers an uninformative disclosure because this increases the chances that it stays unidentified. Yet voluntary disclosures do have an effect on the shareholders' approval decision due to the following consideration: if shareholders observe a positive disclosure signal, they are less likely to tender because they believe that this signal is due to the informative disclosure of the high-value target. In addition, shareholders pay attention to the financial press about the acquirer because the good and the bad type of acquirers may play a separating equilibrium. The prediction of the model is tested empirically with complementary log-log model and rare events procedure. Our findings confirm that positive earnings forecasts by target firms decrease the probability of takeover success once after the uneven distribution of the takeover outcome data is taken into account. This result implies that voluntary disclosures by a target firm in the form of earnings forecasts during a takeover event convey useful information for shareholders. In this regard, shareholders pay attention to these disclosures by target firms to decide for the outcome of the takeover since such disclosures reduce the information asymmetry problem.