Takeover defenses and wealth effects on securityholders: The case of poison pill adoptions

Transcription

1 ELSEVIER Journal of Banking & Finance 20 (1996) Journalof BANKING & FINANCE Takeover defenses and wealth effects on securityholders: The case of poison pill adoptions Sudip Datta a,*, Mai Iskandar-Datta b ~ Department of Finance, Bentley College, 175 Forest Street, Waltham, MA 02154, USA Department of Accounting and Finance, University of Massachusetts- Dartmouth, North Dartmouth, MA 02747, USA b.. Received 7 July 1994; accepted 26 July 1995 Abstract This is the first study to examine the valuation effects of any antitakeover amendment on both bondholders and stockholders. We present new evidence documenting that, on average, there is a significant wealth loss experienced by bondholders at poison pill adoption announcement, while stockholders are unaffected. Our finding of significant bondholder losses is consistent with the proposed negative signal hypothesis. We document results which indicate that bondholders correctly anticipate the degree of leverage increase at the time of the announcement. We also show that the proportion of insider ownership is negatively related to bondholder wealth effect at announcement. This supports the notion that higher insider (manager) ownership leads to a greater alignment of manager-stockholder interests while increasing the stockholder-bondholder agency costs. Long-run analysis of leverage and performance measures reveal that pill adopting firms are not under-leveraged as compared to their industry rivals. However, supporting the negative signal hypothesis, the leverage of sample firms rises significantly after the pill adoption. Performance measures reveal that sample firms significantly underperform their industry cohorts. This result suggests that poison pill adoptions are motivated by poor managers * Corresponding author. Tel.: (1) (617) ; fax: (1) (617) Iskandar-Datta - Tel.: (1) (617) /96/$15.00 Copyright 1996 Elsevier Science B.V. All rights reserved. SSDI (95)0005 l-8

2 1232 S. Datta, M. Iskandar-Datta/Journal of Banking & Finance 20 (1996) attempting to immunize themselves from the disciplinary actions of the corporate control market. JEL classification: G34 Keywords: Corporate control; Antitakeover measures; Bondholder and stockholder wealth; Poison pills 1. Introduction During the 1980s, the proliferation of different types of takeover financing, as well as the conducive regulatory environment for corporate takeovers and restructurings, led to a substantial rise in corporate control related activities. Target managers responded by devising various takeover defense strategies. Previous researchers have focused exclusively on the valuation effect of various antitakeover measures on stockholders. The framework for their analysis has been to investigate whether such decisions benefit the stockholders (principals) supporting the stockholder-interest hypothesis or are an outcome of self-interested behavior of the managers (agents) as contended by the managerial entrenchment hypothesis. Thus far, there is no evidence in the finance literature on the wealth implication of any type of antitakeover measure on the bondholders. This paper adds a new dimension to the corporate control literature by providing evidence on the wealth effect of a poison pill adoption on bondholders. The examination of the case of poison pill plan adoptions is particularly interesting as it has received a great deal of attention in the financial literature. It is widely acknowledged that actions taken by firms can have substantially different wealth implications on different constituencies of the firm (see McDaniel, 1986; Asquith and Wizman, 1990; Lehn and Poulsen, 1991). Thus far, financial economists studied and proposed hypotheses about takeover defenses in the context of shareholder-manager agency relationship. However, a third important constituency which has been ignored by the previous studies is the bondholders. Studies by Malatesta and Walkling (1988) and Ryngaen (1988) find evidence to suggest that the negative stock price reaction to poison pill adoption is consistent with the managerial entrenchment hypothesis. In contrast, recent studies by Comment and Schwert (1995) and Brickley et al. (1994) document slightly negative but insignificant stock price reaction at poison pill announcement. In addition, Ambrose and Megginson (1992), Bhagat and Jefferis (1993) and Comment and Schwert (1995) provide evidence to indicate that poison pills have not been effective in takeover deterrence. In fact, Comment and Schwert (1995) show that such antitakeover measures are associated with higher takeover premia for target shareholders indicating a possible increase in their bargaining power against the bidders. Thus, a poison pill adoption has the potential to hurt or benefit shareholders depending on whether or not managers oppose subsequent takeover offers. We present new results which indicate that bondholders experience significant wealth loss at the announcement of poison pill adoption while the impact on

3 S. Datta, M. Iskandar-Datta/Journal of Banking & Finance 20 (1996) stockholders is insignificant. The negative impact on bondholders is in support of our contention that a poison pill adoption may signal a potential takeover threat which typically results in a substantial leverage increase. This finding is also consistent with Comment and Schwert's evidence that poison pills do not provide meaningful degree of takeover deterrence. Cross-sectional analysis reveals that the greater the increase in the firm's leverage in the post-adoption period, the larger the adverse wealth effect on bondholders. This result implies that bondholders are able to correctly anticipate the degree of leverage increase at the time of the adoption announcement. Regression results also reveal that the proportion of insider ownership is negatively related to the effect on bondholder wealth at announcement. This finding supports the notion that higher insider (manager) ownership leads to a greater alignment of manager-stockholder interests while increasing the divergence of interests between stockholders and bondholders. Finally, supporting the premise of the negative signal hypothesis, examination of the debt ratio for the whole sample indicates that firm leverage increases significantly after the pill adoption. Performance measures reveal that sample firms significantly underperform their industry cohorts. The rest of the paper proceeds as follows: Section 2 briefly describes previous research on antitakeover measures. Section 3 proposes and discusses the testable hypotheses. Section 4 details the sample selection process and the data sources. Section 5 presents the bond and stock event study methodologies. The empirical results are presented in Section 6. We conclude the paper with Section Antitakeover measures and previous research Among the various antitakeover measures that have been adopted by U.S corporations, some require the approval of the shareholders while others may be adopted independently by the incumbent managers. Takeover defenses, such as antitakeover charter amendments or dual-class recapitalizations, require the approval of the shareholders while defenses, such as poison pill adoption, litigation by the target firm management to defend against a hostile tender offer, standstill agreements and privately negotiated targeted share repurchases (greenmail), can be undertaken by the incumbent managers without any explicit shareholder approval. In general, studies have found that defensive measures adopted with shareholder approval result in an insignificant or mildly negative stock price reaction, ~ while i DeAngelo and Rice (1983) and Linn and McConnell (1983) do not find that antitakeover charter amendments adversely affect the target firm shareholders. Both studies, however, acknowledge that the adoption of such takeover defenses may signal an increased probability of the firm becoming a takeover target which may have confounded the resultant stock price effect. Jarrell and Poulsen (1987) examine stockholder wealth effects of different types of antitakeover amendments adopted during the period They report that pure supermajority amendments had no significant effect on stockholder wealth, while those with board-out provisions result in a significant percent abnormal return. In general, studies examining takeover defenses requiring shareholder approval have found that there is insignificant stockholder wealth impact resulting from adoption of such measures.

4 1234 S, Datta, M. lskandar-datta / Journal of Banking & Finance 20 (1996) those measures implemented without shareholder ratification are found to generate significant negative return for the shareholders. A common measure that falls in the latter category is the adoption of poison pills. Studies by Malatesta and Walkling (1988) and Ryngaert (1988) report significant negative abnormal stock return on the announcement of poison pill adoption. These studies interpret their findings to be in support of the managerial entrenchment hypothesis. Examining 132 firms that announced poison pill defenses between 1982 and 1986, Malatesta and Walkling (1988) report a significant negative abnormal stock return during the two-day (- l, 0) announcement period. Further, they find that managers adopting poison pills have relatively low ownership in the firm as compared to the industry average. They also report that these finns have substantially higher probabilities of being takeover targets. Similar to the Malatesta and Walkling (1988) results, Ryngaert (1988) finds that the announcement of a poison pill adoption leads to a significant wealth loss of percent for the stockholders. He partitions the sample into firms that were under a takeover threat and those that were not. Stockholders of the firms that were under a takeover threat suffered a significant wealth loss of -1.S1 percent while abnormal stock return for the other category was found to be insignificant. In contrast, Brickley et al. (1994) find support for the shareholder-interest hypothesis where the stock price response to the poison pill announcement is significantly positive for firms with boards that are dominated by outsiders and significantly negative for firms with boards that are dominated by insiders. Comment and Schwert (1995) examine poison pills, control share laws and business combination laws. Their overall evidence casts doubt on the view that antitakeover measures have been used systematically to deter takeovers and entrench incumbent management. They conclude that the demise of the corporate control market was caused by secular trends and not by the introduction of antitakeover measures. Specifically, they show that takeover premiums are higher when target firms have protection via state laws or from poison pills suggesting that the bargaining positions of targets against bidders is enhanced by these antitakeover devices. Bhagat and Jefferis (1993) examine whether classified hoard charter provisions, poison pills and fair price amendments were successful in deterring takeovers during the two years after the antitakeover measure was adopted. They find little evidence to suggest that takeover defenses are effective deterrents. Using a logit model, Ambrose and Megginson (1992) find that classified boards, fair price charter provisions, dual class recapitalizations and poison pills do not meaningfully deter takeovers. 3. Testable hypotheses Two mutually exclusive hypotheses have evolved in the literature related to the issue of adopting takeover defense measures and the resulting impact on stock-

5 S. Datta, M. Iskandar-Datta/Journal of Banking & Finance 20 (1996) holder wealth. The managerial entrenchment hypothesis is based on the potential conflicts of interest that may arise between managers who want to be insulated from the disciplinary threat of the corporate control market, and target firm shareholders who stand to gain substantially if the takeover is consummated. This hypothesis contends that, by restricting the transfer of managerial control via the adoption of the poison pill defense, target managers benefit at the expense of the stockholders. In contrast, the competing stockholder-interest hypothesis proposes that such takeover defense measures benefit the stockholders by enabling the target firm managers to extract the maximum possible bid premium. Extending the argument presented in the literature on stockholder-interest hypothesis, we propose that if the poison pill benefits the target firm in extracting the maximum possible premia from the bidder then we expect not only the stockholders, but also the bondholders to gain from such a decision, ceteris paribus. Therefore, this hypothesis may be renamed the securityholder-interest hypothesis. However, a crucial underlying assumption of this hypothesis is that the leverage of either the target or the combined entity remains unaffected after the takeover attempt. Yet takeovers are typically accompanied by a substantial increase in leverage. Various researchers have documented the impact of leverage increasing transactions on bondholders. Asquith and Wizman (1990) and Warga and Welch (1991) report that bondholders suffer abnormal losses from leveraged buyouts. Datta and Iskandar-Datta (1995) document that a leverage-increasing corporate partial acquisition has an adverse effect on bondholder wealth. Thus, in the case of poison pill adoption, any resultant higher leverage is also expected to adversely affect the bondholders. Further, Comment and Schwert (1995) document that there is a tendency for managers to adopt a shareholder rights plan shortly before impending takeover attempts. Poison pill adoptions are expected to convey a negative signal to the bondholders, based on the recent evidence by Ambrose and Megginson (1992), Bhagat and Jefferis (1993) and Comment and Schwert (1995) that such antitakeover measures are generally ineffective deterrents and the fact that takeovers are generally accompanied by an increase in leverage,. Hence, from the bondholders' perspective, this argument leads us to introduce the negative signal hypothesis. Further, it may be argued that even when finns do not encounter an immediate takeover threat, the managers may still opt to increase leverage in addition to the poison pill as a second line of defense. Finally, the managerial entrenchment hypothesis states that managers adopt the antitakeover measure in their own self-interest to maintain their tenure in office, which is potentially in conflict with the goal of firm value maximization. Although this is clearly not in the interest of the shareholders, the bondholders may gain from such an action by incumbent managers if they are successful in deterring takeovers. This gain for bondholders is likely to be due to the decreased likelihood of an immediate increase in firm leverage. Such gains may be ameliorated by losses due to managerial actions which do not maximize firm value. Given the

6 1236 S. Datta, M. Iskandar-Datta/Journal of Banking & Finance 20 (1996) arguments presented above, the predictability of the valuation effect on bondholders as a result of poison pill adoptions is not as clear and straightforward as the impact on stockholder wealth. This ambiguity concerning the impact on bondholder wealth makes it interesting and indeed worthwhile to settle the issue empirically. 4. The sample and data sources A preliminary sample of poison pills adopted between January 1985 and December 1989 was collected from Investor Responsibility Research Center's Corporate Takeover Defenses (1989) and from a text search of the firms' annual reports available on Compustat's CD Rom Disclosure database. Prior to 1985, a very small number of poison pills were adopted. Frequency of pill adoptions increased consistently after 1985 and peaked in 1988 and Our sample period is chosen to cover the years with most pill adoption activity. Ryngaert (1988) reports only 13 firms adopted poison pills from December 1982 to December The samples of both Ryngaert (1988) and Malatesta and Walkling (1988) cover the period of By extending the sample to a later period, we are able to test whether the impact of poison pills have changed over time. Of the 383 observations with an exact announcement date, 175 firms were deleted because of the absence of publicly traded straight debt while 39 observations were eliminated due to 'thin' trading. A bond is defined to be thinly traded if there are less than six trades during the 21-day event window. To be included in the sample, a bond has to trade both before and after the announcement day. An additional 78 firms were deleted due to confounding events around the poison pill announcement date. The criteria used for elimination from the sample are summarized in panel A of Table 1. The final sample contained 91 poison pill announcements. Given our research objective which constrains the sample to firms which have publicly traded debt, our final sample is composed of large firms with mean total assets of $9.55 billion at the fiscal year-end prior to the poison pill announcement as opposed to $2.86 billion for firms that do not meet our selection criteria. Panel B of Table 1 presents the distribution of poison pill adoptions in our sample by year. The highest proportion of sample firms (37.36%) are in 1986, while the lowest proportion of the sample (8.79%) falls in We collected the daily bond prices of the most frequently traded bond (one bond per firm) for 11 trading days before and 10 days after the announcement day from the Wall Street Journal (WSJ). 2 The announcement day is defined as the 2 Other recent studies examining bond prices use odd-lot transaction data similar to ours, such as Jayaraman and Shastri (1988), and Datta and Dhillon (1993), among others. Furthermore, Hand et al. (1992, p. 736) point out the advantages of using only transaction prices. They mention that large reversals are common in bond price series if transaction and bid prices are included. In addition, they assert that bid prices adjust to news announcements with a lag.

7 S. Datta, M. lskandar-datta/journal of Banking & Finance 20 (1996) I Table I Sample selection and distribution by year for a sample of 91 firms that adopted poison pills between a Panel A: Sample selection criteria Sample selection Number of firms Beginning sample with exact announcement date 383 Less: No public debt 175 Confounding event 78 Thin bond trading 39 Subtotal 292 Final sample 9 l Panel B: Sample distribution by year of poison pill adoption Year Number of firms Percent of firms a Poison pill sample was collected from Investor Responsibility Research Center Corporate Takeover Defenses (1989) and from a text search of the firms' annual reports available on Compustat's CD Rom Disclosure database. day on which the intent to adopt the poison pill appeared on the Dow Jones News Retrieval Service (DJNRS). Treasury bond prices with matching coupons and maturities as those of the sample bonds are also collected from the WSJ. To compute daily returns from bond prices, with cumulated daily coupon interest, Moody's Bond Record is used to identify the interest payment dates. Stock price data are retrieved from the CRSP NYSE/AMEX daily master tape. Finally, detailed financial information about the sample are retrieved from Compustatfiles and Moody's Industrial Manuals. Table 2 describes the bond sample characteristics based on bond rating, subordination status of the bonds and the frequency of bond trading during the event window. Examining the rating distribution of the sample bonds in panel A, we find 75 of the 91 bonds (82.4%) are investment grade with a rating of Baa or better. In addition, panel B shows that over 75 percent of the bond sample is composed of nonsubordinated issues. Finally, in panel C it is shown that nearly 77 percent of the sample bonds traded more than 14 times during the 21-day event window. We examine whether our bond selection criterion of choosing the most frequently traded bond biases our sample in favor of including higher-rated bonds.

8 1238 S. Datta, M. Iskandar-Datta/Journal of Banking & Finance 20 (1996) Table 2 Description of sample bond characteristics: rating distribution, type of bond, frequency of bond trading, and bond rating of sample bond to firm's outstanding bonds. Panel A: Rating distribution of sample bonds before adoption of poison pill Moody's Number of Percent of Rating firms firms Aaa Aa A Baa Ba B CCC Not rated Panel B: Subordination status of bond sample Type of bond Number of bonds Percent of bonds Subordinated Nonsubordinated Panel C: Frequency of bond trading in event window Frequency of trading in event window Number of bonds Percent of bonds 20 < = Trades < Trades < = < Trades < = < Trades < = < Trades < = < Trades < = Panel D: Relationship of sample bond's rating to outstanding bonds' rating Relative status of bond rating Number of bonds Percent of bonds One bond outstanding Multiple bonds outstanding: Sample bond has same rating as outstanding bonds Sample bond has higher rating than outstanding bonds Sample bond has lower rating than outstanding bonds Panel D of Table 2 documents the relation between the rating of the sample bond and that of the finn's remaining bonds. In over 20 percent of the cases, the firm has only one bond outstanding. In another 73 percent of the sample, all the finn's bonds had the same rating. Only four of the sample bonds (4.35%) had a higher rating than the remaining outstanding bonds, while in two cases (2.17%) the

9 S. Datta, M. Iskandar-Datta / Journal of Banking & Finance 20 (1996) sample bond's rating was lower than other outstanding issues. Thus, it appears that our selection criterion is not biased in favor of including higher-rated bonds only. 5. Research methods We use the mean adjusted returns methodology of Handjinicolaou and Kalay (1984) adapted for bonds to estimate bond excess returns. The corporate bonds are matched with treasury bonds according to maturity and coupon rate to adjust for shifts in the term structure of interest rates. The adjusted bond return (ABRi.~I) is calculated as follows: ABRi. d = BRi, d - TBRi, d ( 1 ) where BRi, d is the holding period bond return for bond i for day d and TBRi. d is the return over the same period for an equivalent treasury bond. The holding period return (BRi,,~) for corporate bond i for day d is calculated as follows: BR i. J = ln[ F,.J Fi,a_, ] (2) where Fi. a is the flat price for corporate bond i for day d. 3 A nineteen-day interval around the event is used to estimate the comparison and announcement period returns. The day the initial announcement of pill adoption appeared in the DJNRS is identified as day 0 in event time. The comparison period is day t-10 to day t-2 and day t+l to day t+10. The mean comparison period return (R~,cp) for bond i as derived by Handjinicolaou and Kalay (1984) is as follows: 1 ABRi, d Ri,cp (3) l, cpd - dk-, where (d k -d~_ I) is the number of trading days that elapsed between two successive trades. Since bond returns are a series of single and multiple day returns they are adjusted to yield equivalent single day returns and standardized as in the following equation: [ ABRi'a - R"c~(d~ - d~- ')] (4) BERi'a = Si~/dk - dk- 1 where BER~. a is the daily standardized excess return for bond i and Si is the 3 Flat price is calculated as follows: Fi, d = Pi,d q'(ci//180)ni where, Pi,d is the closing price for bond i on day d, C i is the semi-annual coupon payment for bond i and N i is the number of days that elapsed since the last coupon payment.

10 1240 S. Datta, M. Iskandar-Datta/Journal of Banking & Finance 20 (1996) estimated standard deviation of the comparison period returns for bond i computed as follows: 1 K ABRi,d Ri,cp~dk - dk- 1 (5) S-= 2 77~ [~ki=2 ~ ~/dk--dk I where k is the number of trading days for bond i during the event period. The standardized mean excess return (SMBER a) for the portfolio of bonds for day d is then estimated over the entire 21-day period and is given by BER i,a SMBERd= Z - - (6) i N where N is the number of bonds trading on day d. Assuming that individual standardized excess bond returns are cross-sectionally independent and normally distributed, the appropriate test statistic for any event day d can be computed as follows: z-stat = v~ * SMBERd (7) which is approximately distributed unit normal. We use the mean excess returns generated by the market model in order to determine the stock price reaction to poison pill adoption. The estimation period for the market model parameters is from day -250 to day -61. Assuming that the standardized excess returns are cross-sectionally uncorrelated, the appropriate test statistics for any event day is as follows: z-stat = V'N * SMSER a (8) where N is the number of stocks in the portfolio and SMSER a is the standardized mean stock excess return for event day d. The z-statistics for the cumulative bond and stock excess returns over various intervals from tl to t2 is computed as follows: t2 Zt z,,.,2 = ~ v/t2 _ tl + 1 (9) tl 6. Empirical evidence 6.1. Bond and stock price reactions to poison pill adoption Table 3 presents the bond price reaction to poison pill adoption announcements made by 91 sample bonds. The average standardized bond excess return on the announcement day is percent which is significant at the 1 percent level (z = -3.18). This is the largest excess return for the entire 21-day event window

11 S. Datta, M. lskandar-datta / Journal of Banking & Finance 20 (1996) Table 3 Daily bond excess returns and cumulative excess returns (CER) for a sample of 91 firms that adopted poison pills during a Event day Excess return z-statistic CER % Neg , * * * ! * * * I (/ I a The mean adjusted returns methodology adapted for bonds in Handjinicolaou and Kalay (19841 is used to estimate excess bond returns which adjusts for changes in the term structure of interest rates and multiple day returns. The comparison period, day t - 10 to day t -2 and day t + I to day t + 10, is used to estimate the comparison and announcement period returns. Day 0 is the Dow Jones News Retrieval day. * * * Significant at the 0.01 level. surrounding the announcement. The two-day (days 0 and + 1) cumulative excess bond return (CER) is percent which is also significant at the 1 percent level. The range of the two-day cumulative bond excess return spans from percent to 5.20 percent. Since 88 percent of the sample bonds reacted negatively on the announcement day we can conclude that the results are not driven by a few outliers. The non-parametric sign z-statistic is significant at the 1 percent level (z = 8.89). These results document that adoption of poison pill announcements, on average, have a significant negative impact on bondholder wealth. The discussion presented earlier in this paper suggests two opposing effects on bondholders due to poison pill adoptions. Our event study results strongly support the negative signal hypothesis which is consistent with the notion that bondholders

12 1242 S. Datta, M. Iskandar-Datta /Journal of Banking & Finance 20 (1996) Table 4 Cumulative stock excess returns (CSER) over various intervals, a Panel A: Cumulative stock excess return (CSER) over various intervals Event window CSER z-statistic Pos:Neg b Sign z-statistic - 30, : , : , 0, : , : Panel B: Three-day cumulative stock excess return for partitioned samples Subsamples Number CSER z-statistic Subject to takeover " Not subject to takeover a The market model is used to determine stock price reaction to poison pill adoption. The estimation period for the market model parameters is from day -250 to day -61. b Stock returns were not available over the announcement period for four observations. * * * Significant at the 10% level. interpret the adoption of the poison pill as a precursor to an increase in leverage. Anticipating an increase in leverage could be either due to an increase in the likelihood of a takeover or managers pursuing a broader scheme to retain corporate control (by increasing leverage as a second line of defense). Consequently, the securityholder-interest hypothesis and the managerial entrenchment hypothesis, both of which predict a positive bondholder wealth effect, are not supported by our bond results. In contrast to the bondholders, panel A of Table 4 shows that the two-day (0, + 1) and three-day (-1,0, + 1) stock excess returns are statistically insignificant. This result is similar to findings by Brickley et al. (1994) and Comment and Schwert (1995). With only 54 percent of our sample stocks experiencing negative excess returns, the non-parametric sign z-statistics are also insignificant. The range of two-day cumulative stock excess retum~ is from percent to percent. To examine whether the impact on stockholders is time period specific, we partition the sample into two groups: (a) firms that adopted the poison pill prior to 1987 and (b) those that adopted the pill during the period. Panel B of Table 4 shows that the stock excess returns are insignificant in both subperiods indicating that the results are not sensitive to the period under examination. Next, we partition the sample into (a) firms that were subject to takeover threat/rumor and (b) those that were not. We categorize firms as being under an immediate threat of a takeover or surrounded by a takeover speculation by

13 s. Datta, M. lskandar-datta/journal of Banking & Finance 20 (1996) scanning the Wall Street Journal Index and the Dow Jones News Retrieval Service for one year preceding the poison pill adoption. Of the nine firms that fell in this category, four had a takeover attempt while five firms had rumors circulating that the firm may be taken over. 4 The empirical results indicate that the stockholders of firms that were subject to a takeover experience significant losses, -2.25%, which is in support of the managerial entrenchment hypothesis. On the other hand, the mean stock excess returns of nontarget firms, 0.09%, is statistically insignificant. These stock results corroborate the findings by Ryngaert (1988) who documents that the average stock excess return is statistically significant, percent, for finns subject to takeover speculation and statistically insignificant for the nontarget firms. Since only a small portion of our total sample was under any immediate takeover threat or speculation (N = 9), our overall stock price impact is heavily influenced by poison pill adoptions undertaken by firms which are nontargets. We also calculate the total dollar gains or losses to stockholders and bondholders. The dollar stockholder (bondholder) gains/losses are calculated by multiplying the announcement period (days 0, + 1) stock (bond) excess return by the value of the common equity (total debt) at the year-end preceding the transaction, Assuming that all debtholders are impacted similarly by the adoption of the poison pill, the mean dollar losses to bondholders is -$46.86 million per firm which is significantly different from zero with a t-statistic of However, the average dollar decrease in stockholder wealth, -$4.16 million per firm, is statistically insignificant (t = -0.67). On average, total firm value decreases a statistically significant -$51.02 per firm (t = -3.12). In summary, the combined bond and stock results indicate that poison pill adoptions are value-destroying for the firm as a whole, and that the bondholders absorb the majority of the losses Bond excess returns and bond covenants Some indenture restrictions, such as constraints on additional financing, dividend payments and event risk covenants, can be critical to bondholders in protecting them against corporate control event risk and the resultant increase in leverage (see Asquith and Wizman, 1990; Lehn and Poulsen, 1991). First introduced in 1986, event-risk covenants ensure bondholder claims by giving them the right to put their bonds to the issuing firm, typically at par value, if a change of control occurs. 4 While we find that 10 percent of our sample finns were under a takeover threat or rumored to be a target, the corresponding proportions reported by Ryngaert (1988) and Comment and Schwert (1995) are 20 percent and 16.6 percent respectively. Our sample includes a somewhat smaller fraction of such firms perhaps because our sample finns are typically larger and are, thus, less likely to be takeover targets. In this regard, Mikkelson and Partch (1989) and Palepu (1986), among others, show that larger firm size reduces the likelihood of a takeover,

14 1244 S. Datta, M. Iskandar-Datta /Journal of Banking & Finance 20 (1996) Table 5 Two-day (0, + 1) bond excess returns, t-statistics and the number of events categorized by the presence or absence of various indenture restrictions, a Type of indenture Does not include Includes restrictions Bond excess return (t-statistics) N Bond excess return (t-statistics) N Dividend restriction ( ) ( ) Debt restriction ( ) ( ) Event risk restriction ( ) ( ) 87 4 a The information on bond indentures were obtained from Moody's Industrial Manuals To examine whether bondholders who are protected with covenants against the risk associated with leverage-increasing takeovers fare better than unprotected bondholders, we collected from various issues of Moody's Industrial Manual the details on various types of protective covenants, such as dividend covenants, additional debt covenants and event-risk covenants. Given our sample period and the fact that a majority of our bonds were issued before 1986, we find only four bonds with this provision. 5 Restrictions on additional debt and dividend payments were more common (15.4% and 25.3% respectively). As shown in Table 5, the bondholders protected by restrictions on debt and dividends are impacted less adversely than those without such protections. 6 However, this is not the case with event-risk protected bondholders. Given the small number of bonds with event-risk protection, no definite conclusions can be made with respect to the impact of this type of covenant on bondholders Explaining cross-sectional bond excess returns To explain bond excess returns to pill adoption announcement we estimate the following model: 5 Although Lehn and Poulsen (1991) find that the event-risk provision is prevalent in 1989 for newly issued bonds (32.1 percent), they report that only three bonds incorporated such a provision in In Table 5, the difference in excess returns between bondholder groups with and without a particular restrictive covenant is insignificant at any of the conventional levels.

15 S. Datta, M. lskandar-datta / Journal of Banking & Finance 20 (1996) Table 6 OLS cross-sectional regression explaining the announcement period bond excess returns. ~' Model: BERo. 1 = a o + al(rating)+ az(tkvr)+ a3(insider)+ a4(chgdr)e Constant RATING TKVR INSIDER CHGDR Adj. R * * ( ) ( ) ( ) ( ) [ N = 85] * * (-2.49) (0.13) (-0.90) (- 1.73) [N = 83] RATING denotes Moody's bond rating where Aaa takes a value of 6, Aa has a value of 5, etc. TKVR takes a value of 1 if the firm was under a takeover threat or speculation during the year prior to the poison pill adoption and 0 otherwise. This variable was obtained by scanning the Wall Street Journal Index for one year prior to the poison pill adoption. INSIDER represents the percent of insider ownership. This variable was collected from Compustat's CD Rom Disclosure database. CHGDR reflects the percentage change in the debt ratio from the year prior to the poison pill adoption to the year of the adoption obtain from Compustat files. (t-statistics in parentheses are computed using White's correction for heteroskedasticity). N is the number of observations used to estimate each regression. * * Significant at the 5% level. BERo,, = a o + a I.(RATING) + a2. (TKVR) + a3. (INSIDER) + a4.(chgdr ) + e (10) where BERo, I is the two-day announcement period bond excess return, RATING is a categorical variable that assumes a value of 6 if Moody's bond rating is Aaa, 5 if it is Aa etc., TKVR takes a value of 1 if the firm was under a takeover threat or speculation during the year prior to the poison pill adoption and 0 otherwise, INSIDER denotes the percent of insider ownership prior to the poison pill adoption, CHGDR is the percentage change in the firm's debt ratio (measured as book value of total debt to total assets) from the year preceding the pill adoption to the year of the adoption, and e is a random error term. Table 6 presents the regression results. The eroskedasticity correction. 7 t-statistics are computed using the White (1980) het- A higher-rated bond is expected to suffer less from the negative signal about a forthcoming increase in leverage (conveyed by the adoption of the poison pill) than a lower-rated bond. To investigate this relationship, we use the rating of the bond issue as a proxy for the riskiness of the bond. The coefficients of this variable are insignificant in both models. This result is maintained if we redefine bond rating as a dummy variable which differentiates between investment and junk grade bond categories. 7 We detected heteroskedasticity using White's general heteroskedasticity test (see Fomby et al., 1984).

16 1246 S. Datta, M. lskandar-datta/journal of Banking & Finance 20 (1996) The notion behind the managerial entrenchment hypothesis is that the managers protect their tenure in office by adopting the poison pill which makes takeover difficult, thereby depriving the target firm's securityholders from realizing their maximum value. Ryngaert (1988) reports a significant negative abnormal stock return of percent for the subsample of firms subject to a takeover speculation. To examine whether this result is cross-sectionally consistent for the bondholders, we include the TKVR variable. If firms subject to takeover threat or speculation are more likely to increase leverage, then we would expect bondholders to suffer larger losses, ceteris paribus. Although, the coefficient of the TKVR variable, -0.16, has the expected negative sign, it is statistically insignificant. This result could mean that whether or not firms are subject to a takeover threat, bondholders expect an increase in leverage at the time of the poison pill adoption announcement. Previous researchers (Malatesta and Walkling, 1988; Ryngaert, 1988) have proposed that managers with a lower proportion of insider ownership have greater incentive to adopt a poison pill thereby entrenching themselves at the cost of destroying shareholder wealth. The conjectured effect of the degree of insider ownership on bond excess returns may be opposite to that of the stockholders. This is because an increased insider (manager) ownership is expected to lead to a greater alignment of manager-stockholder interests while exacerbating the divergence of interests between stockholders and bondholders. Contrary to the predicted stockholder wealth effect, it can be argued that a higher degree of insider ownership can have an adverse effect on bondholder wealth as the probability of a takeover increases leading to a higher likelihood of a concomitant increase in leverage. Mikkelson and Partch (1989) and Song and Walkling (1990) document that the probability of a successful takeover is positively related to insider holdings. Further, some studies (Asquith and Wizman, 1990; Warga and Welch, 1991) have documented the impact of leverage increasing transactions on bondholders. Consistent with the combined implications of these studies, our regression analysis indicates that the INSIDER variable is negative and statistically significant at the 5 percent level. This finding supports the notion that higher insider ownership which reduces agency conflicts between managers and stockholders may, in fact, increase the agency costs between managers and bondholders. In the second model, in addition to RATING and TKVR variables we include the CHGDR variable which captures the percentage change in the debt ratio following the poison pill plan announcement. To the extent that investors are able to anticipate a future change in the level of debt, a negative relation between bondholder wealth change and the CHGDR variable is expected. With a coefficient of -0.55, this variable is significantly negative at the 5 percent level. The finding lends support to the proposed negative signal hypothesis which is based on investor expectation of an enhanced likelihood of a leverage increase. This finding

17 S. Datta, M. Iskandar-Datta/Journal of Banking & Finance 20 (1996) Table 7 Median book debt ratio, return on assets, and profit margin for three years centered around poison pill adoption year. Year relative to Level Industry-adjusted p-value poison pill Book debt ratio b ~ , Return on assets Profit margin - l All financial variables are obtained from Compustat files. Debt ratio is defined as book value of total debt to book value of total assets. Return on assets is calculated as net income after extraordinary items divided by total assets. Profit margin is defined as net income after extraordinary items divided by net sales. Each variable is industry-adjusted by subtracting the median value of that variable using all firms in Compustat with the same four-digit SIC code from the sample finn's value for that variable. Sample firms are excluded when calculating the industry median. Median significance levels are based on two-tailed Wilcoxon signed rank tests. b The difference between the level debt ratios of year - 1 and year 0 is statistically significant with a p-value of c The difference between the industry-adjusted debt ratios of year -1 and year 0 is statistically significant with a p-value of also suggests that bondholders are able to correctly anticipate the degree of leverage increase at the time of the poison pill adoption announcement Long-run analysis of leverage and firm performance around poison pill adoption Table 7 reports median estimates of the levels and industry-adjusted measures of the leverage and performance of sample firms for three years centered around the year of the pill adoption. We focus on the median as a measure of central tendency because the distribution of the financial ratios is skewed. Hence, the median is more reliable and more informative about the typical sample firm. In at least three respects, these financial ratios may provide insights into the various proposed hypotheses. First, the level of leverage over time should provide an indication of whether bondholder losses are prompted by changes in leverage. Second, poor performance in prior years may provide a possible motivation for the

18 1248 S. Datta, M. Iskandar-Datta/Journal of Banking & Finance 20 (1996) poison pill adoption. Third, a firm's earnings subsequent to the pill adoption may provide some indication of whether the managerial entrenchment hypothesis or the securityholder-interest hypothesis holds. Debt ratio (DR) is measured as a ratio of the book value of total debt to the book value of total assets. Two performance measures are used: annual return on assets (ROA) and profit margin (PM). ROA is defined as the ratio of net income to book value of total assets, and PM is computed as a ratio of after-tax net income to net sales. Because the sample firms' performance may be affected by economywide factors, we control for these influences by using the industry median as a benchmark to obtain industry-adjusted measures. Each variable is industry-adjusted by subtracting the median value of that variable for all firms in the Compustat database with the same four-digit SIC code. Sample firms are excluded when calculating the industry median. To test for median significance levels, we use a two-tailed Wilcoxon signed rank test. Data for all the accounting ratios are obtained from the Compustat Annual Industrial file and supplemented by information from Moody's Manuals. Three interesting observations can be made about the results relating to debt ratios. First, the sample firms were not under-leveraged before the adoption of poison pill defense. The industry-adjusted median debt ratio (0.9%) at the year-end prior to the poison pill adoption (Year -1) is significantly different from zero (p-value = 0.02). Second, we observe a significant increase in the median leverage (in levels) from Year - 1 to Year 0 (p-value = 0.02). The industry-adjusted debt ratios also show a significant increase from Year - 1 to Year 0 (p-value = 0.07). Finally, the sample firm's debt ratio remains higher than the industry in the year following the poison pill adoption. These results provide further support for the negative signal hypothesis which argues that bondholder losses at poison pill announcements are prompted by the expectation of an imminent increase in leverage. Examination of the performance measures (ROA and PM) reveals that the median pill adopting firm performs consistently worse than the industry median during and around the year of the pill adoption. Both performance measures indicate that sample firms significantly underperform their industry cohorts in each of the three years centered around the announcement year. The long-run analysis also suggests that pill adoption is motivated by poor managers attempting to protect themselves from the disciplinary actions of the corporate control market. 7. Conclusions This study illuminates a new dimension in the corporate control literature by documenting that bondholders are adversely affected by a poison pill adoption. Consistent with recent studies, we find that the wealth effect on stockholders is insignificant. Our finding of significant bondholder losses is consistent with the

19 s. Datta, M. Iskandar-Datta / Journal of Banking & Finance 20 (1996) proposed negative signal hypothesis. We also find that the stockholders of firms under a takeover speculation/threat experience a significant negative stock price reaction. Cross-sectional analysis indicates that the adverse wealth effect on the bondholders at adoption announcement is systematically related to the post-adoption increase in leverage. We also find that the proportion of insider ownership is negatively related to bondholder wealth change at pill adoption announcement. This supports the notion that higher insider (manager) ownership leads to a greater alignment of manager-stockholder interests while increasing the stockholderbondholder agency costs. Long-run analysis of leverage and performance measures reveal some interesting results. We find that pill adopting firms are not under-leveraged as compared to their industry cohorts. However, consistent with the premise of the negative signal hypothesis, the leverage of sample firms are found to rise significantly between the pre-adoption year and the year of the adoption. Compared to the industry, the debt ratios of the sample firms are significantly higher in each of the three years centered around the year of the adoption. Performance measures reveal that sample firms significantly underperform their industry cohorts. This result is in support of the argument that poison pill adoptions are motivated by poor managers to protect themselves from the disciplinary actions of the corporate control. References Ambrose, B.W. and W.L. Megginson, 1992, The role of asset structure, ownership structure, and takeover defenses in determining acquisition likelihood, Journal of Financial and Quantitative Analysis 27, Asquith, P. and T.A. Wizman, 1990, Event risk, covenants, and bondholder return in leveraged buyouts, Journal of Financial Economics 27, Bhagat, S. and R.H. Jefferis, 1993, Is defensive activity effective? Working paper (University of Colorado, Boulder, CO). Brickley, J.A., J.L. Coles and R.L. Terry, 1994, Outside directors and the adoption of poison pills. Journal of Financial Economics 35, Comment. R. and G.W. Schwert, 1995, Poison or placebo'? Evidence on the deterrent and wealth effects of modern antitakeover measures, Journal of Financial Economics, forthcoming. Datta, S. and U.S. Dhillon, 1993, Bond and stock market response to unexpected earnings announcements, Journal of Financial and Quantitative Analysis 28, Datta, S. and M.E. Iskandar-Datta, 1995, Corporate partial acquisitions, total firm valuation and the effect of financing method, Journal of Banking and Finance 19, DeAngelo H. and E.M. Rice, 1983, Antitakeover charter amendments and stockholder wealth, Journal of Financial Economics 11, Fomby, T.B., R.C. Hill and S.R. Johnson, 1984, Advanced econometric methods (Springer-Verlag, New York). Hand, J.R.M, R.W. Holthausen and R.W. Leftwich, 1992, The effect of bond rating agency announcements on bond and stock prices, Journal of Finance 47,