Ownership Concentration, Asymmetric Information, and the. Positive Announcement Effect of New Equity Placements

Transcription

1 Ownership Concentration, Asymmetric Information, and the Positive Announcement Effect of New Equity Placements Xueping WU * Department of Economics and Finance, City University of Hong Kong Tat Chee Avenue 83, Kowloon, Hong Kong Tel: (852) ; Fax: (852) efxpwu@cityu.edu.hk Zheng WANG Tel: (852) ; Fax: (852) efzwang@cityu.edu.hk This version: March 15, 2002 * Corresponding author. Both authors are from the Department of Economics and Finance of City University of Hong Kong. We wish to thank Espen Eckbo, Ned Elton, Lawrence Khoo and Sheridan Titman for helpful discussions and comments on an earlier version of the draft. WU gratefully acknowledges financial support from the City University of Hong Kong (DAG project No ).

2 Ownership Concentration, Asymmetric Information, and the Positive Announcement Effect of New Equity Placements Abstract The existing literature has shown that new equity financing can have a positive announcement effect (especially in private placements). The two widely cited explanations are Wruck's (1989) ownership concentration changes hypothesis and Hertzel and Smith's (1993) asymmetric information mitigation hypothesis. However, these two hypotheses are premised on a debatable idea that private investors play an active, value-enhancing role through new equity purchases (as challenged by Barclay, Holderness and Sheehan 2001). Further questioning the validity of these hypotheses, this paper shows that both the hypotheses are based on results obtained from using an unusual measure of announcement return that biases the related tests in favor of these hypotheses. Using a standard return measure and a unique Hong Kong data set that can isolate the factors that underlie the two hypotheses, we find no evidence for these two hypotheses. A more interesting and important finding is that positive announcement returns of equity placements (including some non-private placements) reflect an endogenously determined information effect, consistent with a generalized Myers-Majluf model. Key Words: Equity Placement, Announcement Effect, Ownership Structure, Asymmetric Information, Investment Opportunities JEL Classification Code: G14, G32

3 2 Ownership Concentration, Asymmetric Information, and the Positive Announcement Effect of Seasoned Equity Placements 1. Introduction The existing finance literature has shown that new equity financing can have a positive announcement effect (especially in private placements). A great deal of theoretical and empirical analysis has been devoted to understanding the phenomenon. It has been a tradition to view private placements differently from ordinary placements. In ordinary placements, both existing and new investors are deemed to be passive and the adverse-selection effect of equity financing becomes a well-known result (Myers and Majluf 1984). In contrast, in private placements, new equity sales may enable investors participating in new equity purchases (or private investors) to play an active, organizational or informational role. Two hypotheses, changes in ownership concentration (Wruck 1989) and mitigation of asymmetric information (Hertzel and Smith, 1993), have become well-cited explanations for the positive announcement effect of private placements. Both the hypotheses are premised on the idea that the active role of private investors can be value enhancing. Wruck (1989) finds that the announcement returns are related to changes in ownership concentration, consistent with the non-linear relationship between firm value and ownership structure suggested by Morck, Shleifer and Vishny (1988). Wruck argues that new equity blocks placed cause substantially changes in ownership concentration, thus creating a monitoring effect or a management entrenchment effect. Using the same definition for announcement returns as introduced by Wruck (1989), Hertzel and Smith (1993) find that the announcement returns are related to placement price discounts and new block sizes (or fraction placed). They argue that price discounts reflect information costs that the new investors

4 3 incur in private placement. As a result, the gain of firm value should be related to fraction placed (a proxy for the amount of valuation uncertainties to be mitigated). They hypothesize that in private placement new blockholders play a role mitigating asymmetric information, or creating a certification effect. In this paper, we start with showing that both the above hypotheses are based on results obtained from using an unusual measure of announcement return and are questionable. The announcement return measure concerned (introduced by Wruck and also adopted by Hertzel and Smith) is the sum of a standard announcement return measure plus an additional term that includes the exclusive compensations to new blockholders. 1 This introduces a problem: the measure they use obviously has a built-in relationship to ownership concentration changes (as in Wruck 1989) and more directly to fraction placed (as in Hertzel and Smith 1993), two explanatory variables used in the empirical tests of the above two hypotheses, respectively. As a result, their tests are biased in favor of their respective hypotheses. 1 Wruck (1989) claims that this new return measure is originated from Bradley and Wakeman (1983) who conduct an event study of stock repurchases. The return measure is defined as the change in firm value due to the valuation effect of a corporate event divided by the pre-announcement firm value. But Wruck seems to ignore a key difference between new equity issues and stock repurchases. Unlike in stock repurchases, there are usually new investors created in new equity issues. While this return measure means correctly the return to old shareholders (both participating and non-participating) in Bradley and Wakeman (1983), it is neither an appropriate measure of the return to old shareholders nor the return to both old and new shareholders in new equity issues. Besides, Bradley and Wakeman only use the return measure that Wruck refers to in the wealth transfer analysis. After all, in their crosssectional analysis of announcement effects, Bradley and Wakeman follow the literature to employ a standard measure of announcement return to non-participating shareholders.

5 4 Once it is recognized that private investors may not necessarily have a value-enhancing role, it will become more convincing that there is no monitoring or certification effect in private placements. This seems to be the case. Barclay, Holderness and Sheehan (2001) have already cast doubt on the above two hypotheses. Using U.S. data, they find that new blockholders are usually not active in association with private placement (a situation deemed more likely to occur with non-private placements). Thus, the positive announcement effect of seasoned equity placements remains largely unexplained. The main purpose of the paper is twofold. First, we add further evidence to the literature that the monitoring or certification effect is absent in private placements. Second, we shed new light on the determinants of the positive announcement effect of new equity placements that include some nonprivate placements. To these ends, we make use of a unique data set from Hong Kong. The major form of seasoned equity offerings (SEOs) in Hong Kong is placements. Placements, both private and non-private (or ordinary), are typically firm commitment underwritten offers through investment banks or brokerage firms to new investors, who have no connections to corporate insiders of the issuers. In addition, the concentration of controlling ownership in Hong Kong is notably high, with an average controlling ownership of about 40 percent. In our sample, new equity placed is on average more than 10 percent of existing shares outstanding. As a result, a new equity issue to non-existing shareholders causes a substantial change in the concentration of controlling ownership in private as well as ordinary placements. These unique characteristics of the Hong Kong data provide us an opportunity to examine any potential relationship between announcement effects of new equity issues and ownership concentration changes, not only in private placements but also in non-private placements. The direct comparison between our private and non-private placements can help isolate and test the monitoring effect. Unlike

6 5 ordinary placements, private placements usually generate new substantial (block) shareholders. If new blockholders are active investors, this creates a monitoring effect (Wruck 1989). However, we will show that ordinary placements (where the monitoring effect is apparently absent) can also have, on average, a positive announcement effect. This suggests that factor(s) other than the monitoring effect is causing the positive announcement effect. Our results are rich and interesting. We leave an overview to the conclusion of this paper. For the moment, two main points should be highlighted. First, we find positive announcement effects in both private and non-private placements in Hong Kong. Interestingly, we do not find significant difference in announcement effects between these two samples. A positive announcement effect in ordinary placement is inconsistent with the Myers and Majluf (1984) model. However, there is a literature that allows for positive announcement effects in ordinary equity placement (see Ambarish, John and Williams, 1987; Cooney and Kalay, 1993; Wu and Wang, 2001). 2 Second, using an announcement return measure common in the event studies literature, we find no evidence of either a monitoring or a certification effect in private placement. Instead, we find that positive announcement returns in both private and non-private placements simply reflect an endogenously determined information effect, consistent with the generalized Myers-Majluf framework of Wu and Wang (2001). An endogenous-determined information effect is purely related to information asymmetries regarding assets-in-place and new investments. These information asymmetries create an effect conditional on the news to the market that contains no new information except about the equity 2 For example, Cooney and Kalay (1993) and Wu and Wang (2001) suggest that the Myers-Majluf model can be generalized to predict positive as well as negative announcement effects of seasoned equity issues. See the literature review in Section 2.

7 6 placement itself. This effect is fundamentally different from an exogenously driven information effect: for example, the certification effect that requires an additional assumption that the active, informational involvement of new investors is value enhancing. To the extent that an endogenously determined information effect produces positive announcement returns in SEOs, the understanding of a generalized Myers-Majluf framework is important. The Myers-Majluf model has been generalized to be consistent with both positive and negative announcement effects (Cooney and Kalay, 1993; Wu and Wang, 2001). By introducing private benefits of control into managers/controlling shareholders' wealth objective function, Wu and Wang (2001) show that it is incentive compatible for managers to undertake negative net present value (NPV) projects. Unlike the original Myers-Majluf model, the generalized Myers-Majluf model allows the market to assign positive probabilities to both future underinvestment and future overinvestment. As a result, the pre-announcement equilibrium stock price is lower than that predicted by the original Myers- Majluf model. Consequently, SEO announcement effects are likely to be positive if asymmetric information arises more from new investments than from assets-in-place (see Cooney and Kalay, 1993; Wu and Wang, 2001). In this situation, the adverse-selection effect can be overwhelmed by good news about new investments (however, this never occurs in the original Myers-Majluf model). The remainder of this paper is organized as follows. Section 2 surveys the literature. Section 3 describes the data and offers sample statistics. Section 4 discusses the validity of the announcement return measure that Wruck (1989) introduces into the literature on private placements. Section 5 presents the empirical evidence concerning SEO announcement effects using Hong Kong data. Section 6 examines the determinants of announcement effects. Section 7 concludes the paper. 2. The Positive Announcement Effect of SEOs: Hypotheses and Theories in the Literature In this section, we review major models and hypotheses concerning the positive announcement

8 7 effect of seasoned equity offerings (SEOs), along with related previous empirical studies. In the U.S. market, the positive announcement effect, found largely in the case of private placements, has been explained in the literature by two widely cited hypotheses: changes in ownership concentration and mitigation of asymmetric information (Section 2.1). However, there are also other theories, such as those that directly generalize the adverse-selection model of Myers and Majluf (1984), that can predict positive announcement effects of SEOs (Section 2.2). 2.1 Hypotheses to Explain Positive Announcement Effect in Private Placement The positive announcement effect of SEOs has been found especially in association with private equity issues, where issuers are mostly small. The evidence is also international (e.g., Wruck, 1989 and Hertzel and Smith, 1993, for the U.S.; Kato and Schallheim, 1993 and Kang and Stulz, 1996, for Japan; Cronqvist and Nilsson, 2000, for Sweden). Since private placements are highly concentrated, and are deemed to involve active new investors, it seems to be understandable that their positive announcement effect may not fit into the adverse-selection framework of Myers and Majluf (1984). The existing literature offers two widely cited hypotheses. First, Wruck (1989) argues that private equity sales are usually highly concentrated. Therefore, insiders ownership structure is significantly changed after the placement. By introducing and using a non-standard measure of announcement return, Wruck finds a significant relationship between the announcement returns and ownership concentration changes. In particular, the relationship shows that private sales that cause managers to be more aligned with the interests of shareholders add value. Conversely, private sales that cause management entrenchment or less interest-alignment with shareholders destroy value. Thus, new private equity sales induce a nonlinear relationship between firm value and ownership, in line with the Morck, Schleifer and Vishny (1988) hypothesis. From this evidence, Wruck concludes that the monitoring/entrenchment effect can explain the positive announcement effect of private equity placements. Hertzel and Smith (1993) provide a second alternative hypothesis. They argue that private equity

9 8 placements are a solution to the Myers-Majluf adverse selection problem. They emphasize that private investors are not organizational active but instead are informational active. Private equity sales tend to reduce asymmetric information because managers can better communicate with a small number of private investors. It follows that placement discounts must reflect compensations for information costs borne by private investors. Using the same definition for announcement returns as introduced by Wruck (1989), Hertzel and Smith find a significant relationship between the announcement returns and new block sizes (or fraction placed). The larger the new investment (or fraction placed), the more difficult it is for the private investors to value the uncertain investment, as they argue. Thus, they interpret that the above relationship is consistent with a certification effect. The certification comes from the value-enhancing, informational involvement of private investors, which is deemed to cause a positive shift in the market s assessment of the firm value of the issuers. Both the above two hypotheses are premised on private investors' active involvement through new equity purchases. However, Barclay, Holderness and Sheehan (2001) have recently challenged this premise directly. Using a sample of the U.S. seasoned equity issues much larger than used in previous studies, Barclay et al. find that private investors are actually not active. Thus, it is unlikely that private investors fulfill a value-enhancing, organizational or informational role: as such, this finding casts doubt on the monitoring and certification hypotheses. 2.2 Can Announcement Effects also be Positive in Non-Private Equity Issues? Suppose investors behave passively in both private and non-private equity issues. There will be little difference in the role that new investors play through the two forms of placements. However, unlike in private equity issues, the announcement effects in non-private equity issues have been found to be negative on average (especially in the U.S. market, see the surveys by Smith, 1986, and Eckbo and Masulis, 1995). There are two major theories in the literature that explains the negative market reaction to non-private equity issues: the underinvestment model of Myers and Majluf (1984) and the

10 9 overinvesment-driven agency model of Stulz (1990) and Jung, Kim and Stulz (1996) (detailed in Section 2.2.1). However, these models do not adequately explain all phenomena observed. There are some theories that are consistent with both positive and negative announcement effects. The multiple-signal model of Ambarish, John and Williams (1987) allows the possibility of positive announcement effects in new equity issues, and links positive (negative) announcement effects to overinvestments (underinvestments) of the issuers (Section ) Also, direct extensions of the Myers-Majluf framework can generate models with both positive and negative announcement effects (Section 2.2.3) Models that Explain Negative Announcement Effects There are two major models that explain negative SEO announcement effects. The first is the underinvestment model of Myers and Majluf (1984). This model, built on the adverse-selection problem of Akerlof (1970), produces two well-known predictions. First, firms may skip positive NPV projects if they are forced to issue new equity for financing investment hence creating a problem of underinvestment. Second, the issuing firms stock prices always drop at the announcement of new issues. In their model, Myers and Majluf make three major assumptions, that managers act in the interest of existing (passive) shareholders, that managers accept only non-negative NPV projects, and that managers know more than outside investors about both the value of assets-in-place and the NPV of investment opportunities of their firms. Myers and Majluf argue that a new issue must signal an overvaluation of a firm s assets-in-place, because the managers would forego (positive) new investments if they knew that the current stock price of their firm was undervalued. The Myers-Majluf model does not imply that there is no good news signaled by new issues. But, in the Myers-Majluf logic, since the market has always expected good investment opportunities, the

11 10 decision to issue equity mainly signals an overvaluation of the firm s assets-in-place bad news that overwhelms the good news that managers are able to undertake (already much expected) good projects. 3 The literature has extended the Myers-Majluf model to predict that new issues cause smaller drops in stock prices when there are smaller information asymmetries. Korajczyk, Lucas and McDonald (1991) and Dierkens (1991) have found that smaller information asymmetries relieve negative announcement effects. 4 Choe, Masulis and Nanda (1993) have also found that the tide of new issues has to do with business cycles. They show that announcement effects are significantly related to market conditions, since adverse selection effects and investment opportunities naturally vary with market conditions. Apart from the Myers and Majluf (1984) underinvestment model, there is an alternative explanation for negative SEO announcement effects. Stulz (1990) develops an agency model largely on the market's concern about overinvestment. Jung, Kim and Stulz (1996) subsequently apply this model to SEOs. In contrast to Myers and Majluf s underinvestment argument, the agency model suggests that the negative announcement effect of new issues signals that managers are likely to take on bad projects. It 3 It has been observed that there are significantly negative stock price responses to the announcement of SEOs, particularly in the U.S. market (e.g., Asquith and Mullins, 1986; Masulis and Korwar, 1986, among others). In survey papers, both Smith (1986) and Eckbo and Masulis (1995) document an average abnormal return of about 3.0 percent for U.S. industrial firms. This stylized evidence is consistent with the prediction by Myers and Majluf (1984). 4 Since debt issues involve less adverse-selection than equity issues do, the findings that debt issues have virtually no impact on stock prices are interpreted as being consistent with the Myers-Majluf model (e.g., Dann and Mikkelson, 1984; Eckbo, 1986; Mikkelson and Parch, 1986). This also gives rise to Myers (1984) pecking order model of financing, due to the difference in information costs of various forms of capital in the capital markets.

12 11 has been recognized in the previous literature that the agency problem is an important concern for investors. Managers have a potential desire to invest in self-interested projects (for example, corporate empire building) rather than to maximize shareholder wealth. Thus, managerial discretion may give rise to an agency problem that leads to overinvestment (Jensen, 1986). Firms with free cash flows potentially allow managers to waste shareholders money a situation more likely to occur in mature firms. Under asymmetric information, when firms are forced to raise external funds for their new projects, it is difficult for managers to convince the market that their projects are worthy of investment and are not a waste of money (Stulz, 1990). Stulz argues that debt issues are more appropriate than equity issues for firms with free cash flows because debt plays a disciplinary role on managerial discretion (see also Jensen and Meckling, 1976; Jensen, 1986). Thus, the agency model predicts that the announcement effects of new equity issues are negative when the issuing firms do not show much growth. One might argue that firms with poor investment opportunities could also be financially distressed and are simply not mature. Financially distressed firms are less likely to have nimble managers. And the market is more likely to believe that less nimble managers will take on more negative NPV projects. Jung, Kim and Stulz (1996) find evidence in support of the overinvestment driven agency model. They show that firms with poor investment opportunities that otherwise should issue debt experience an extremely significant drop in stock prices in response to their new equity issues. In contrast, they find no adverse shocks in stock prices for issuing firms with more valuable investments Positive Announcement Effect and The Multiple-Signal Model Is it possible for new equity issues to signal overwhelmingly good news? As noted above, the standard Myers-Majluf model rules out this possibility. The overinvestment driven agency model concentrates mainly on the adverse impact of the agency problem in the announcement effects of new equity issues. Nevertheless, such a possibility has received formal academic attentions since the development of the

13 12 multiple-signal model. Building on the work of Miller and Rock (1985) and John and Williams (1985), the multiplesignal model of Ambarish, John and Williams (1987) considers an efficient signaling equilibrium with both dividends and new issues (investment). Their model predicts that paying dividends increases stock prices. In addition, the announcement returns of new equity issues, given the dividend policy, are negative if asymmetric information arises more from assets-in-place than from investment opportunities, and are positive otherwise. Since there is an underlying coordination between the dividend and financing policies, this model implies that there is a positive relationship between the announcement effects of new issues and the dividend policy. However, empirical studies using U.S. firms have found little support for such a coordinated relationship (Loderer and Mauer, 1992). The most provocative prediction of the Ambarish-John-Williams model is the positive relationship between announcement effects of new equity issues (which are likely to be positive) and investment opportunities. Previous studies have found a significantly positive cross-sectional relationship between announcement effects and firms investment opportunities. But the search for significantly positive announcement effects is, on the whole, disappointing, even for high growth firms (see Pilotte, 1992; Denis, 1994) Positive Announcement Effect and The Generalized Myers-Majluf Framework The Myers-Majluf prediction that issuers stock prices always drop when new equity issues are announced hinges crucially on one key assumption that managers only take on positive NPV projects (as articulated by Cooney and Kalay, 1993). Cooney and Kalay argue that if managers can take on negative NPV projects as well as positive ones, issuers stock prices may in some cases rise when new equity issues are announced. The intuition is that, in the Myers-Majluf model, the anticipation of always-

14 13 positive NPV projects has been rationally reflected in the firm s pre-announcement stock prices, limiting the information content in the positive news about new projects. In contrast, if managers are expected to accept negative NPV projects as well, the pre-announcement stock prices, compared to the Myers and Majluf case, are adjusted downward accordingly. Thus, the announcement of new issues may signal overwhelmingly positive news about new investments. This is likely to happen when the asymmetric information stems largely from investment opportunities (this is similar to the prediction of the Ambarish-John-Williams model). Cooney and Kalay, however, offer no strong rationale to explain why managers may be willing to take on negative NPV projects. Myers and Majluf (1984) forcefully reject the idea that managers are willing to take on bad projects, by arguing that managers would, ex post, put the raised capital into zero- NPV investments (such as tradable securities) instead. Thus, investors could rationally expect that all projects that managers undertake have non-negative NPVs. In short, the analysis in Cooney and Kalay (1993) does not seem to be subgame perfect. 5 To solve this incentive compatibility problem, Wu and Wang (2001) introduce private benefits of control into the Myers-Majluf framework. When managers/controlling shareholders can derive private benefits from new projects, they are willing to take on negative NPV projects if the private benefits overwhelm the detrimental effects of the negative NPV investments on the manager/controlling 5 In addition, Cooney and Kalay still assume that managers maximize existing shareholders value as Myers and Majluf (1984) do. Thus, the trade-off between the existing shareholders benefits and costs regarding managers undertaking bad projects is not defined. Consequently, because they consider all bad projects available, the market's concern about overinvestment, which is incorporated in the pre-announcement prices, tends to be overstated in their analysis.

15 14 shareholders equity holdings. The Wu and Wang (2001) model reconciles the Myers-Myers underivestment model and the overinvestment-driven agency model, and it also allows new equity issues to have positive announcement effects. Like the Ambarish-John-Williams model, the model of Wu and Wang (2001) shows that the positive announcement returns are more likely to be related to relative uncertainties about investment opportunities rather than to the expected growth per se. But the key difference between the two models is that the model of Wu and Wang (2001) does not directly attribute the negative announcement effects of new issues to underinvestment and the positive effects to overinvestment. The novel prediction of the model is that the positive relationship between the announcement effects and investment opportunities (or the expected NPV of a new project) is not monotonic. In particular, when the expected growth, given asymmetric information, increases to a high level, the announcement effect decreases. The intuition for this is that, if a new investment is so good that the new issue is almost inevitable, there will be little information content in the new issue. Thus, it may be difficult to find positive announcement effects for issuers with very high growth opportunities, something that the empirical literature has shown. The measurement of the relative uncertainties about investment opportunities is critical to empirical studies. Wu and Wang (2001) propose that firm size is a reasonable proxy for uncertainties about growth, relative to those about assets-in-place, because the smaller an issuer s firm size, the more asymmetric information may come from growth rather than from assets-in-place. Using this assumption, they come up with a testable prediction: the announcement effects are negatively related to firm size, with smaller issuers being more likely to enjoy positive announcement effects of new equity issues. Kang and Stulz (1996) come across this very relationship when they report a small (0.5 percent) but significant average announcement return with a sample of 185 Japanese non-private equity issues. However, they find it puzzling. Traditionally, it has been perceived in the literature that small firms are

16 15 particularly fraught with asymmetric information. Therefore, the announcement of new issues should cause a larger drop in small firms stock prices. Likewise, Fama and French (2002) view it perplexing when they find that less-levered small-growth firms favor new equity issues. According to Myers (1984) pecking order model, this should be the last resort in financing. This puzzle can be resolved according to the Wu and Wang (2001) model. First, (outside) debt is unlikely for small firms that usually have little collateral value and limited reputations. Second, if asymmetric information stems more from positive investment opportunities than from assets-in-place as is likely for some small firms, asymmetric information may not necessarily be disadvantageous in new equity issues. 6 In this case, the adverse-selection effect may be overwhelmed by the good news about new investment. The possibility for such a situation, however, is absent from the original framework of Myers and Majluf (1984). If the positive announcement effect of private placement does not come from the active, organizational or informational role of new blockholders, it may well arise from the relative asymmetric information about investment opportunities. This is an endogenously determined information effect, since private equity issuers are usually small. By "endogenously determined" we mean that the new 6 But this does not mean that the new equity issues by small firms would become rampant. First, if issuers are expected to have bad projects, announcement effects are likely to be negative (see the simulation results in Wu and Wang, 2001). Second, if a very bad future investment is the inside information, a new equity issue can eventually cause an adverse effect on managers/controlling shareholders equity ownership, thus making their private benefits from the new investment not to pay. Third, the advantage of a small growth firm in subsequent new equity issues will endogenously diminish as the small firm gets bigger after the current new issue. It is worth mentioning that the above arguments are independent to the repeated-game argument by Ambarish, John and Williams (1987) to prevent cheating.

17 16 information conveyed by the announcement of a new issue is only that a new issue occurs. Model parameters are kept constant from before to after the event. The equilibrium stock prices are adjusted according to this new information in a way predetermined by the model. In contrast, hypotheses such as the certification effect reflect an "exogenously determined" information effect. Such a hypothesis requires, for example, an additional assumption that new investors play an active, informational role that is value enhancing. This active role favorably revises the market assessment of the issuers' firm value. The generalized Myers-Majluf framework hence has an important application. Private placements can be treated the same way as ordinary equity issues in this framework. Since private and non-private issues are pooled together, we will need to unify the measure of announcement returns in finding out the determinants of the announcement effects. As mentioned above, the literature on private placement has been using a non-standard measure, first introduced by Wruck (1989). In the next section, we argue that the Wruck measure is not interpretable, and that no exception should be made for private placements, regarding the measure of announcement returns. Put differently, the standard measure of announcement returns should be used. 3. The Validity of Wruck s (1989) Measure of Announcement Returns In this section, we question the validity of Wruck s (1989) measure of announcement returns when used in a cross-sectional regression analysis of new equity placements (Section 3.1). We then show that this measure causes a bias in favor of hypotheses in both Wruck (1989) and Hertzel and Smith (1993) (Section 3.2).

18 Announcement Returns to Whom? Wruck (1989) introduces a new measure of announcement returns into the literature about private placements. 7 This measure departs from the standard measure of announcement returns commonly used for ordinary seasoned equity offerings and other event studies. Let P 0 and P 1 be the pre- and post-announcement stock price, P x the placement price, N the shares outstanding before the placement (old shares), and N the new shares placed. The equilibrium price after the announcement is: P = ( P N + P N + NPV ) /( N + N ), (1) 1 0 x where NPV is the market assessment on the change in firm value due to the valuation effect of the placement. Apparently, NPV is shared by both the old and new shareholders. From (1), we simply have: NPV = P ( N + N ) ( P N + P N ) 1 = ( P P )N + ( P P ) N, x x (2) where the first term in the second equation of (2) is the gain to old shareholders and the second term is the benefit to new shareholders. Wruck (1989) defines AR NPV, or NPV divided by P 0 N, as the measure of the return to old shareholders due to the new information. Let AR =(P 1 -P 0 )/P 0, the standard measure of the return to old shareholders, and d=(p 0 -P x )/P 0, the (positive) discount that new shareholders exclusively enjoy in the placement. According to (2), Wruck s definition is actually: 7 Wruck does use the standard measure to document announcement effects of private placements but switches to her new measure when analyzing the cross-sections of the announcement returns.

19 18 AR NPV ( P1 P0 )N + ( P1 P0 + P0 Px ) N = P N N = AR + ( AR + d ). N 0 (3) It might be debatable whether the definition of (3) makes sense but it is questionable enough when we call it, as Wruck does, the return to old shareholders while the return apparently includes the exclusive compensations to new shareholders in the placement. At least, one would be confused under the definition of (3) when asking the question: Which objective, old shareholders wealth alone or old and new shareholders combined wealth, is on the mind of the managers? Wruck (1989) claims that AR NPV originates from the one used in the analysis of targeted shares repurchases in Bradley and Wakeman (1983) (see also footnote 1). Indeed, the definition of (3) is exactly the same as the one (also labeled (3)) in Bradley and Wakeman (1983) if one uses unified definitions (signs) of equity changes and discount/premium in both new issues and shares repurchases. But Wruck seems to ignore the fact that, unlike in the new equity placement, the shares outstanding before the announcement of repurchases involve both non-participating and participating shareholders. Both Bradley and Wakeman (1993) and Wruck (1989) divide the total informational effect (NPV) by the preannouncement firm value. Yet, Bradley and Wakeman correctly call it the information-induced percentage change in the value of the assets of the repurchasing firm. This illustrates that the measure is neither the return to old shareholders nor the return to both old and new shareholders in new equity offerings. In addition, Bradley and Wakeman never use the definition of (3) in any cross-sectional regression; their use of definition (3) is purely for the analysis of the wealth transfer between, and the information effect shared by, participating and non-participating shareholders in shares repurchases. Instead, Bradley and Wakeman use the (dollar) return to non-participating shareholders, W, as the preannouncement equity value of non-participating shareholders multiplied by the standard measure of the

20 19 announcement return, AR, in their cross-sectional regression (5). In short, the use of AR NPV introduced by Wruck (1989) in the literature about private placements appears strange. 3.2 The Built-in Relationship between AR NPV and Ownership Structure Variables At this point, it should be obvious that AR NPV contains an additional term compared to the standard measure, AR, as shown in (3). This additional term embeds shares dilution, N/N, a near-perfectly correlated variant of fraction placed, N/(N+ N), and placement discount, d. Let Insider Shares be corporate controlling insiders number of shares outstanding and Conown the controlling ownership in percentage, Insider shares/n, before the placement. Assuming that the insiders and existing shareholders do not participate in private placements and the new shareholders (actively) join the insiders after the placement, the ownership structure changes, Ownership, in Wruck (1989) is simply Ownership = Insider Shares + N N + N Insider Shares N = N N + N N N + N Insider Shares N (4) = Fraction Placed (1 - Conown). Wruck (1989) finds a significant relationship between AR NPV and Ownership, support for the ownership concentration hypothesis. Hertzel and Smith (1993) document a significant relationship between AR NPV and Fraction Placed, support for the asymmetric information mitigation hypothesis. We show below that their analysis is biased in favor of their hypotheses. There is a built-in relationship of AR NPV with Ownership or even more directly with Fraction Placed. Assuming that Fraction Placed is not correlated with the standard measure, AR, nor correlated with the placement discount, d, in any way, and substituting the near-perfect correlated share dilution, N/N, with Fraction Placed, the covariance between AR NPV and Fraction Placed, according to (3) is:

21 Cov{ ARNPV,Fraction Placed } = Cov{ AR + Fraction Placed (AR + d), Fraction Placed} 20 = Cov{Fraction Placed (AR + d), Fraction Placed} = Cov{Fraction Placed (m + e), Fraction Placed} (5) = m Var{Fraction Placed}, where m is the mean of AR+d and e is the residual that is not correlated with any moments of Fraction Placed, as we assumed in the beginning. In private placements, both AR and d are usually positive on average. Thus, the slope estimate in a regression of AR NPV on Fraction Placed equals a positive number, m, even when there is no relationship between the standard measure of announcement returns, AR, and Fraction Placed. A similar bias can be shown when one runs a regression of AR NPV on Ownership because Ownership in (4) is also closely related to Fraction Placed by definition. Because of this property, the R-squares in both the regression results of Wruck (1989) and Hertzel and Smith (1993) are unusually high, around 50 percent. To illustrate numerically that the relation of AR NPV to Fraction Placed or to Ownership arises mainly from the components of AR NPV that are, by definition, related to Fraction Placed and Ownership, we do a similar analysis using a sample of Hong Kong private placements. It should be mentioned that, if the private investors are not active, the new shares should not be included into the ownership concentration. In this case, the change in the existing controlling ownership is negative due to the new issue: Conown= Ownership-Fraction(Placed). Thus, Conown is an alternative measure of ownership concentration change when new investors do not play any active role in the issuing firms. As shown in Table 1, each of the measures that is relevant to the ownership concentration changes (Fraction Placed, Ownership and Conown), even though evidently not correlated with AR (the standard measure of announcement returns), is significantly correlated with AR NPV. As we have already pointed out, this significant correlation actually comes from the significant correlation with the additional term in AR NPV : Dilution*(AR+d). This occurs through a significant relation to Dilution ( N/N)

22 21 or Discount. Note that the correlation coefficient between Conown and Dilution is negative, -0.89, because Conown is always negative in new issues to outside new investors. In short, there is a permanent bias in the relationship between AR NPV and variables relevant to ownership concentration changes. Thus, the use of AR NPV by the literature to study the announcement effect or the firm valuation effect of private placements should be reconsidered. In summary, it is appropriate to use the standard return measure of AR to study private as well as non-private placements. Applying the standard return measure to the U.S. data to replicate the tests of Wruck (1989) and Hertzel and Smith (1993) does not seem promising, though. Barclay, Holderness and Sheean (2001) have already used the U.S. evidence to challenge the very notion that new shareholders in private placements play an active, organizational and informational role. We turn instead to more international data for possible explanations of the positive announcement effect of new equity placements. 4. Data The SEO data on announcement date, issue price, and issue amount are collected from The Securities Journal published monthly by the Hong Kong Stock Exchange from 1989 to Only local nonfinancial and non-utility firms listed on the Stock Exchange of Hong Kong (SEHK) are considered. The SEOs included in our sample are either private or non-private (ordinary) equity placements and are purely new equity sales to outside or independent investors that have no connections to corporate insiders (required by the local listing rules), the main form of seasoned equity issues in Hong Kong. The new issues are firm commitment underwritten usually through investment banks and brokerage firms. In the private placements, the new equity is sold to a limited number of identifiable outside investors (also labeled in the data source as private placing ) while, in the non-private placements, both the

23 22 number and identity of investors are unknown to the public. New issues to corporate insiders or parent firms, rights-issues, and issues of warrants and convertibles are excluded from the sample. The controlling ownership data are collected from various volumes of Hong Kong companies handbooks and guides. 8 Controlling ownership includes the holdings of the largest shareholder and, where applicable, also the holdings of the family members of the largest shareholder. Finally, stock returns and financial statement data where available are retrieved from the Pacific-Basin Capital Markets (PACAP) databases. Table 2 reports that there were a total of 409 SEOs of listed Hong Kong firms during the period from 1989 to As shown in Panel A, there are 99 private placements and 310 ordinary placements in the sample and the latter dominate the former in every year. During this nine-year period, the number of SEOs in Hong Kong fell to 13 issues, the lowest, in 1995, but peaked twice, with 73 issues in 1993 and 75 issues in This phenomenon of waves of SEOs is consistent with evidence from the U.S. observed by Choe, Masulis and Nanda (1993). Panel B of Table 2 shows the dollar amount placed. For the SEOs in our sample, HK$ 36,517 million (equivalent to US$ 4,681 million) were placed, for which private placements accounted for HK$ 10,677 million while ordinary placements for HK$ 25,840 million. Table 3 reports the issue and firm characteristics of the SEOs. As shown in Panel A, the average issuing amount, Offersize, is HK$ million for private placements while it is HK$ million for ordinary placements. It seems that the private placements, on average, raised more money than did the 8 The list of the sources is Companies Handbook, 1988, published by The Stock Exchange of Hong Kong Ltd; Corporate International s Company Handbook: Hong Kong, 1992, 1993, 1994, Published by Corporate International; Thornton guide to Hong Kong Companies, 1995, 1996, published by Edinburgh Financial Pub. (Asia).

24 23 ordinary placements. In terms of shares placed as a percentage of shares outstanding at the end of the previous month of the issuance announcement, Dilution, the private placements, on average, also create more new shares than do the ordinary placements (18.69 versus percent). SEOs are often placed at a discount in percentage, Discount, measured by one minus the offer price over the closing price two event days before the announcement of an issue. The private placements have, on average, a slightly deeper discount than do the ordinary placements (5.75 versus 4.20 percent). The median discount rates are actually similar (5.56 versus 5.78 percent). Thus, we do not observe, on average, a deeper discount in offer prices of private placements in Hong Kong. A bigger discount in private placements is an important way to compensate private investors for information signaling (certification) as hypothesized by Herzel and Smith (1993). Panel B of Table 3 reports firm characteristics of the SEOs. The average firm size by market value (market equity plus book debt at the end of the previous fiscal year), MV, is much smaller for firms with private placements (equal to HK$ 1,571.2 million) than for firms with ordinary placements (equal to HK$ 4,032.6 million). However, their median firm sizes are similar (HK$ million versus HK$ million), as also confirmed by Wilcoxon ranks sum test. The fact that the median firm sizes are smaller than the average ones indicates that smaller issuers dominate the Hong Kong market. Another observation is that the firm size variation is much smaller among private placements than among ordinary placements (HK$ 6,634 million versus HK$ 16,334 million by standard deviation). This means that firms with private placements cluster more as small firms. As also shown in Panel B of Table 3, like the average controlling ownership, Conown, (39.0 versus 45.0 percent), the median controlling ownership of the firms with private placements is

25 24 significantly lower than that of the firms with ordinary placements (40.0 versus 45.5 percent). 9 Given the substantial dilution in new equity placement documented above, the high ownership concentration in Hong Kong would render both private and non-private new equity placements to cause substantial changes in ownership concentration. Panel B of Table 3 further shows, firms with private placements tend to have less growth prospects in terms of market value over book value, MV/BV, and lower returns on equity, ROE, than do firms with ordinary placements. But other variables such as leverage and dividend yield, DivYd, do not show much a difference between the two types of issuers. In sum, compared with firms with ordinary placements, firms with private placements have significantly lower concentrations of controlling ownership and less growth prospects. Also important, private issuers clusters more as small firms. In Section 6, we will test whether these firm characteristics are important to explain the cross-sections of announcement returns. For the moment, in the next section, we first document the average announcement effects, which are one of the important concerns discussed in the literature. 5. Announcement Effects of New Equity Placements in Hong Kong We essentially follow the event study method as in Asquith and Mullins (1986) and Kang and Stulz (1996). Daily stock returns with reinvested dividend are used. Each year, we form ten equally weighted portfolios among all listed Hong Kong stocks in the PACAP database, ranked according to the Scholes- 9 For a number of 466 non-utility and non-financial Hong Kong listed firms as the whole population in 1996, a typical controlling ownership is higher (the mean and median values are 49.3 and 50.6 percent, respectively).

26 25 Williams (1977) beta estimates of individual stocks computed against the PACAP equally weighted market portfolio for Hong Kong. The decile portfolio to which an issuing firm s stock belongs is thus its control portfolio. On a particular day, the abnormal return of the issuing firm is defined as the return of the firm in excess of its expected return estimate or the return of the control portfolio. At any event-time cross-section, t, an average abnormal return, AAR t, is taken across the issuing firms. To gauge the significance of the average abnormal return on each event day, we calculate, from the estimation event time period, 89 to 11, the standard deviation of the event-time-series of the average abnormal returns accounting for the Newey-West (1987) auto-correlation adjustment up to the fourth lag. For a multi-day announcement window, [event day t 1 to t 2 ], a cumulative average abnormal return, CAR[t 1,t 2 ], is defined as the sum of the time-series of AAR t within the event window. And its standard deviation is the standard deviation for the one-day AAR t multiplied by the square root of the number of event days in the event window. Table 4 shows, for individual event days from 15 up to +15, the daily average abnormal returns, AAR, cumulative average abnormal returns, CAR, and their t-values for the private and ordinary equity placing firms in the period from 1989 to We highlight the CARs of two-day [-1, 0] and three-day [- 1, 0, +1] event windows, accordingly, at the bottom of Table 4. The announcement effects of the seasoned equity offerings (SEOs) in Hong Kong, regardless of private or non-private placements, turn out to be, on average, significantly positive. The two-day and three-day average CARs are 1.97 (tvalue=5.01) and 3.51 (t-value=7.28) for private placements and 1.90 (t-value=7.76) and 3.08 percent (tvalue=10.29) for ordinary placements. Interestingly, the positive announcement effects are statistically indifferent between the private and ordinary placements in Hong Kong (p-values >0.21). Moreover, while the findings of positive announcement effects for private placements are consistent with the empirical literature (e.g., Wruck, 1989), the findings for ordinary placements are in sharp contrast to findings from the U.S. market (see the empirical literature survey by Eckbo and Masulis, 1995).