Market Reaction, Revised Proceeds, and the Classification of Seasoned Equity Offerings *

Transcription

1 Market Reaction, Revised Proceeds, and the Classification of Seasoned Equity Offerings * Abstract SEO registrations galvanize information gathering about issuing firms. Issuers also provide additional information in the registration period. We posit that market reaction to such new information influences issuers decision about their final offer size. The offer size relative to the amount filed initially further reveals the quality of the issuing firm and helps predict subsequent firm performance. Improved offerings, whose offer size exceeds the amount registered originally, record significantly positive price reaction on the offer date and do not underperform post-issuance. The converse is true for their complement: Regular offers experience significantly negative reaction on the offer date and underperform their benchmark following issuance. JEL Classification: G14, G32, G39 Key Words: SEOs, Market Timing, Market Reaction, Revised Proceeds, Registration Period, Offer-date returns

2 Market Reaction, Revised Proceeds, and the Classification of Seasoned Equity Offerings 1. Introduction Seasoned equity offerings (SEOs) typically follow a stock price run-up, and their announcements evoke a significantly negative price reaction (Asquith and Mullins, 1986; Masulis and Korwar, 1986; among others). While these stylized facts are well-established, there is considerable debate about the motivation underlying SEOs. For example, the market timing hypothesis argues that managers exploit windows of opportunity to sell overvalued stock (e.g., Myers and Majluf, 1984; Baker and Wurgler, 2000). Also, Loughran and Ritter (1995, 1997) report that SEOs exhibit poor performance in the long run. In contrast, an investment story suggests that the equity issuance is motivated by pressing financing needs (e.g., Zhang, 2005; Carlson, Fisher, and Giammarino, 2006; DeAngelo, DeAngelo, and Stulz, 2010). 1 To address the abovementioned debate, we propose a Revised Proceeds Hypothesis based on an important, previously unexplored element of the SEO process. Specifically, we refer to the informational exchange between the investment banker/issuer and prospective buy-side investors during the SEO s registration period. 2 We posit that this information exchange and the attendant market reaction influence the manager s final offer-size decision. We further hypothesize that the offer-size decision has important information content about the quality and the post-issuance performance of the issuer. In line with the Revised Proceeds Hypothesis, we construct an ex ante measure 3 which classifies an observation as Improved if the offering proceeds exceed the amount filed initially at registration and the complement as Regular SEOs. This parsimonious ex ante classification based on the Revised Proceeds Hypothesis successfully separates market timers from 1 A more comprehensive review of the literature is contained in the following section. 2 We define the registration period as the interim between the filing date (i.e., the SEO announcement date) and the offer date. Issuers and their investment banks conduct the offer s road show during the registration period. 3 It is an ex ante measure relative to the issuers post-issue performance, as this measure classifies SEOs as Improved vs. Regular as soon as the offer size is finalized on the offer date. 1

3 the rest. Our findings, discussed below, highlight the importance of making the distinction between the two types of SEOs. We find that Improved issuers experience significant stock price appreciation whereas Regular offers encounter negative returns, during their respective registration periods. The difference between the two is highly significant. Further, on the offer date, the stock price reaction for Improved SEOs is positive and significant while it is significantly negative for Regular SEOs. 4 Prior research has not made the distinction between the two types of SEOs and reports an insignificant stock price reaction on the offer date. Indeed, if we do not distinguish between Improved and Regular SEOs, we also find, for our overall sample, an insignificant offer-date price reaction. These results reconcile the aforementioned debate in the literature and indicate that the SEO population is not homogenous with respect to the motivations underlying their issuance. With respect to long-term performance, Regular offers significantly underperform their benchmark in the post-issuance period while Improved offers do not. The difference in post-issue returns between the two offer types is about 0.6% per month, or more than 7% per annum. In addition, Regular SEOs experience deteriorating fundamentals and analysts revise earnings forecasts downward in the post-issuance period. Improved issuers, on the contrary, make relatively larger investments following their SEOs and attract more institutional investors. 5 These results are consistent with the Revised Proceeds Hypothesis and indicate that the final offer size relative to the amount filed initially at announcement has significant valuation implication on the offer date and beyond. 6 Could our results be affected by mimicry? It may appear at first glance that a Regular issuer 4 This result is robust to controlling for the offer price discount which we compute more accurately than has been done in prior literature. Our measure is described in detail in the data section (Section 3). 5 If indeed institutional investors have superior stock selection skills, this result would suggest that Improved issuers are of higher quality relative to Regular issuers. 6 The increased offer size relative to the amount originally filed could arise from a price change during the registration period and/or managerial response to the registration-period market reaction. We investigate the marginal effect of the information released in the managerial response on the offer date with respect to the final offer size using a two-stage approach in Section 4. 2

4 might easily mimic an Improved offering by low-balling the initial registration amount and subsequently increasing the offer size. We believe that such mimicking is unlikely. Mimicking an Improved issuer requires the final offer size to exceed the originally registered amount. Since the final offer-size decision has to be arrived at jointly by the SEO issuers and their investment bankers, false signaling would imply that the bankers are complicit in this misrepresentation. Conversations with investment bankers indicate that they loathe permitting any practices that may hurt their buyside clients. This is because buy-side clients are repeat business customers, who also direct sizable market-trades through these banks. It is highly unlikely that investment banks would jeopardize this lucrative repeat business by permitting false signaling. What if we were to misclassify issuers while testing the Revised Proceeds Hypothesis? A misclassification would occur if issuers with positive prospects were mistakenly identified as Regular and/or overvalued issuers were classified as Improved. To the extent that there is misclassification, the empirical evidence will be inconsistent with the predictions of the Revised Proceeds Hypothesis. In particular, if overvalued issuers were to be misclassified as Improved, then such market-timing issuers would not have genuine projects to invest the increased SEO proceeds into nor would they be favored by the smart-money or institutional investors. This, in turn, would lead to lower institutional demand and evidence of deteriorating firm performance in the post-issuance period. Consequently, for these misclassified Improved SEOs, evidence of (a) pronounced post-issuance underperformance, (b) lower investments, and (c) lower institutional demand, or a combination thereof, would negate our Revised Proceeds Hypothesis. Likewise, if issuers with good prospects were misclassified as Regular offers, then we would expect to see improved performance for such offers in the post-seo period, contrary to the Revised Proceeds Hypothesis. Further, a misclassification of either type of offer would reduce the statistical significance of their differences, making it less likely to separate the two types. Our finding that there are significant differences between the two types of SEOs suggests that issuers of Regular SEOs are more likely to be market 3

5 timers while the Improved offers tend to be made for bona fide reasons. The rest of the paper is organized as follows. Section 2 discusses the related literature and our Revised Proceeds Hypothesis. Section 3 describes the sample and summary statistics. In Section 4, we present announcement, registration-period, and offer-date market reactions for both types of SEOs. Section 5 discusses long-run results including long-run stock returns, operating performance, and analyst forecasts and revisions. An examination of changes in the issuers investments in the post-seo period is also included in Section 5. Section 6 contains robustness checks and examinations of alternative hypotheses. Our conclusions are contained in Section Literature review and Revised Proceeds Hypothesis 2.1. Literature review Managerial surveys (Graham and Harvey, 2001) reveal, consistent with the adverse selection argument, that managers are concerned about the timing of their equity issues. In a similar vein, Myers and Majluf (1984) suggest that the SEO announcement itself reveals to the market that issuers stock is overvalued. Lucas and McDonald (1990) argue that undervalued firms wait for a stock price run-up before issuing additional equity, whereas overvalued firms issue equity immediately upon the arrival of a value enhancing project. 7 In either case, the market infers that the issuer is overvalued, as documented by a negative announcement period stock price reaction. Furthermore, Loughran and Ritter (1995, 1997) document long-run underperformance in stock returns and accounting earnings subsequent to the SEO. More recently, Baker and Wurgler (2000, 2002) argue that equity offerings tend to be made when the market as a whole is overvalued and predict negative long-run stock performance following equity issues. These findings are consistent with the notion that managers 7 In Lucas and McDonald (1990), the arrival of value-enhancing projects is independent of the firm s price history and the firm needs the proceeds from the SEO to invest in the projects. The overvalued firms issue equity immediately upon the arrival of the positive NPV project whereas the undervalued firms wait for the undervaluation to vanish before issuing additional equity. 4

6 exploit windows of opportunity to sell overvalued equity. On the other hand, DeAngelo, DeAngelo, and Stulz (2010), Butler, Cornaggia, Grullon, and Weston (2011), among others, recognize that there are alternative investment-based explanations. For example, Zhang (2005) and Carlson, Fisher and Giammarino (2006) argue that the exercise of profitable growth options may produce a similar abovementioned pattern in stock returns. Thus, there is an on-going debate as to which of these competing explanations better explains firms issuance decisions and it remains an unresolved question whether the firm decides to issue additional equity to strategically time the sale of overvalued stock or to finance genuine corporate capital requirements. 8 A recent strand of literature reports results supportive of both sides of the debate. Notably, Kim and Weisbach (2008) examine the use of the funds raised through stock offerings in a comprehensive sample of more than thirteen thousand international SEOs, and note that both sets of reasons discussed above could motivate equity offerings. They conclude that, Like IPOs, one motivation for SEOs appears to be raising capital for Investments while other SEO firms may be issuing equity when their stock price is high, even if the capital raised in the offering is not required for financing investments. (page 283). While the Kim and Weisbach (2008) results support both motivations for SEOs, their objective is not on ex ante identification of market-timers nor is it to examine how their performance differs from other issuers Revised Proceeds Hypothesis The filing of the SEO with the Securities and Exchange Commission (SEC) marks the beginning of a process that culminates in the stock issuance on the offer date. 9 We posit that prior to the filing of the offer, managers of the issuing firm and investors in the market each possess 8 There is another set of papers that offers explanations quite distinct from those related to either side of the aforementioned debate. Dittmar and Thakor (2007) argue that SEOs are made when the level of agreement (on firms future investment plans) is high between managers and stockholders and that the pre-seo stock-price run-up is evidence of growing consensus between the two. DeAngelo, DeAngelo, and Stulz (2010) find that SEOs are employed to overcome liquidity issues and not for market timing purposes. 9 As noted previously, the registration period is the interim between the SEO filing date and the offer date. 5

7 incomplete and partially overlapping information sets regarding the issuer s future prospects and demand for the firm s shares. The filing of the offer with the SEC is an information disclosure event, and triggers attention on the firm. Investors, alerted by the filing of the offer, seek new information about the firm s prospects. Enriched by (a) the information in the registration statement, and (b) information gathered from the SEO s roadshow conducted during the registration period, investors reassess and update their prior beliefs about the issuer s implicit value. Issuing firm managers observe the market s reaction to the proposed issuance, and accordingly decide the size of the offer. 10,11 Managers final dollar offer-size decision is thus the culmination of a dynamic process of information exchange in the registration period. Such interaction between the firm and the market plays a pivotal role in shaping outcomes on, and beyond, the offer date. We posit that the final offer size relative to the amount initially filed further discerns the quality of the issuer, and helps predict the subsequent outcome of the SEO. Within this framework, we construct a new empirical measure which compares the final dollar offering size against the amount filed originally at registration with the SEC. We define SEOs as Improved if their offering proceeds exceed the amount filed at registration and the complement as Regular SEOs. This measure and the related SEO classification incorporate two factors. First, it captures the market s updated perception of the firm s prospects as reflected by the change in the stock price during the registration period. Second, it incorporates the attendant final dollar offer size. It should be emphasized that the final dollar offer size is a response jointly arrived at by the SEO issuers and their investment bankers. The final offer size, and consequently, the revision from the 10 Similar informational exchange and associated strategic behavior has been proposed elsewhere in the literature. For example, in the context of IPOs, Jegadeesh, Weinstein, and Welch (1993) argue that the market is better informed than the issuer and that positive post-ipo returns indicate to the issuer that they have underestimated the marginal return to the project being funded. They suggest that the issuer uses this information and increases the scale of the project by raising additional capital. Allen (1993) also argues that even though the firm may be better informed than any single investor, in the aggregate the market may be better informed than the firm and concludes that checking by investors ensures that resources are allocated to viable firms. 11 The theoretical framework by Dow and Gorton (1997) and Subrahmanyam and Titman (1999) and the empirical evidence from Chen, Goldstein, and Jiang (2007), support the idea that managers glean information about their firm s prospects from market determined stock prices. 6

8 filed amount, is revealed only on the offer date. Improved SEOs, that receive the market s approval as captured by a positive registration period return, and/or the investment banker/manager s increased offer-size decision, are more likely good quality firms seeking to fund bona fide capital needs. The reverse is hypothesized for Regular SEOs. The negative reception from the market during the registration period indicates its perception of a strong likelihood that these offers are designed to capture stock overvaluation. Despite the negative market reaction during the registration period (road-show), opportunistic managers continue with the SEO as long as the stock price remains above their private valuation. 12 Thus, the decision to go through with the offering reveals that the stock is still overvalued Predictions of the Revised Proceeds Hypothesis The classification of offers into Improved and Regular SEOs is at the heart of the Revised Proceeds Hypothesis and its associated predictions. This classification is free of look-ahead bias and permits a parsimonious identification of issuers who are more likely to be opportunistic and time their equity offers versus those that issue equity for bona fide reasons. As we discuss below, our measure and the associated SEO classification predict cross-sectional differences for SEOs Offer-date returns First, for Improved issuers, the market s prior beliefs that these offers are primarily made to fund bona fide capital needs would be reflected in the positive registration period returns. Despite the anticipation that these will be improved offers, the market s beliefs are confirmed only when the offer size is finalized in managerial revisions on the offer date. Accordingly, the Revised Proceeds 12 If the stock price drops below the managers private valuation of their stock, they will withdraw the SEO, rather than persist with an undervalued stock sale. A withdrawn SEO thus indicates that the price may have been driven too low and that the stock is now undervalued. This potentially explains the positive market reaction to a withdrawn offer (see Mikkelson and Partch, 1988). 7

9 Hypothesis predicts a positive offer-date price reaction for Improved offers (net of the market reaction during the registration period). Further, to the extent that the revision from the amount filed to the final offer size reflects the information about issuers future prospects, the greater the upward revision in offer size, the more positive the predicted stock returns on the offer date. Next, for Regular issuers, the negative registration-period returns indicate that the market believes that the issuer s stock is overvalued. The decision to go through with the offer, rather than withdrawing it, is in itself a negative signal. On the offer date, when it is revealed that the offer-size is below the dollar amount filed at registration, it further serves to confirm the market s priors regarding overvaluation of the issuer. Therefore, the Revised Proceeds Hypothesis predicts negative offer-date returns for Regular issuers Post-issuance performance If indeed Regular issuers aim to exploit mispricing and they are still overvalued at their offer date, they are likely to underperform their benchmarks following the SEO. Therefore, the Revised Proceeds Hypothesis predicts that Regular SEO firms will exhibit poor post-issue performance and experience deterioration in operating fundamentals after the offering. On the other hand, Improved issuers seeking to raise equity to fund genuine corporate capital needs are likely to perform better than Regular offers in the post-issuance period. The hypothesis also suggests that managers who learn from the information revealed by the market s reaction during the registration period will incorporate it into their corporate investment decisions (e.g., Allen, 1993; Jegadeesh, Weinstein, and Welch, 1993; Chen, Goldstein, and Jiang, 2007). In other words, Improved issuers raise funds to satisfy investment needs. Therefore, the Revised Proceeds Hypothesis predicts that Improved issuers will invest relatively more than their Regular counterparts. The positive registration period returns can be an outcome of increased demand from 8

10 institutional investors in response to the issuer s road-show. 13 It is also possible that the issuer s offersize decision, revealed on the offer date, stimulates demand among institutional investors for the issuers shares. Accordingly, we predict that Improved SEOs generate higher institutional demand for their equity issues around the offer date relative to Regular SEOs. 3. Data 3.1. Sample selection Thomson Reuters Securities Data Corporation (SDC) New Issues database is first used to generate the initial sample of fully-marketed SEOs over the period. Since we examine post-seo performance and investment activities up to two years following the issuance, the sample period is from 1997 to The sample starts in 1997, the beginning of the SEC EDGAR era, when original filings with the SEC became publicly available on its website. We hand-check original documents filed by each issuer on EDGAR to ensure that the screens, described below, have been accurately applied. The initial screen excludes offers below $25 million. We also exclude offers by financial firms, pure secondary offers, rights offers, dual-class offers, unit offerings, shelf registration (Rule 415) offerings, and Rule 429 filings under which shares registered (but unissued) in a previous filing are added to the shares filed in the current offer. We then remove all F-1, F-3, and other foreign issuers filings. Finally, we drop accelerated offers (Gao and Ritter, 2010), whose registration period (from filing to public offering) is less than 6 days. Throughout the paper, we define an Improved SEO as an offer whose total dollar amount of proceeds has been increased from its original filing amount. The 13 Chemmanur, He and Hu (2010) and Gibson, Safieddine and Sonti (2008) find that institutional investors possess private information about SEOs, and trade in the same direction as their information. We posit that their private information is confirmed and augmented by the information gleaned from the issuer s road-show. 9

11 remaining offers are classified as Regular SEOs. 14 Table 1 reports the chronological distribution of the 815 SEOs in our final sample. [Insert Table 1 here] 3.2. Firm characteristics Table 2 presents the mean and median of firm characteristics for the full sample and the subsamples of Improved versus Regular SEOs. For the full sample, the average total shares offered are 4.49 million shares, about 22% of total shares outstanding a day prior to the offer (event day -1). The average proceeds are $138 million, relative to the $136 million filed initially. The average firm size prior to the offer is $0.913 billion. The average book-to-market ratio is 0.20, suggesting that the sample firms have relatively high growth opportunities. [Insert Table 2 here] The comparison between the two types of SEOs shows that Improved SEOs generally have a higher relative offer size (measured as the ratio of total shares offered to the existing outstanding shares), receive more proceeds, and have a higher share price and a lower average book-to-market ratio prior to the issuance. The lower book-to-market ratio for the Improved SEO firms suggests that they exhibit higher market valuation and higher growth prospects. 15 Consistent with the Revised Proceeds Hypothesis, Improved offers start smaller at filing but end up significantly larger by the offer date. Specifically, Improved SEOs increase their average offer proceeds to $ million from the $ million amount filed whereas Regular issuers drop from an average $ million at filing to $ million by the offer date. Interestingly, the increase/decrease in the dollar amount from filing to offer is accompanied by a similar 14 Among the Regular SEOs, there are only two issues whose total dollar value of final proceeds is the same as the original amount filed. Our results are robust to dropping these two issues. 15 Besides the variables discussed thus far, there are several measures shown in Table 2 where there is no significant difference between Improved and Regular SEOs. These are: Fraction of Primary shares, Underwriter reputation, Length of registration period, Age, Return Volatility, Leverage, Operating Performance, and Investments. 10

12 increase/decrease in the number of shares filed versus offered for the Improved and Regular issuers, respectively. The Improved SEOs increase the number of shares from an average of 4.14 million shares at filing to million shares offered. The Regular SEOs reduce the number of shares from 4.75 million shares filed to an average million shares offered. These results indicate that the increase/decrease in the final offering proceeds, and thus our Improved/Regular classification, captures not only stock price appreciation/decline during the registration period but also managerial revisions to increase/decrease the number of shares offered in the SEO process. Among other data variables, we hand collect the offer-date, last reported trading price for the issuer s common stock, and the SEO offer price from the original prospectus filed with SEC. This is an important issue because it affords us accuracy in variable measurement. First, prior literature employs spikes in trading volume and/or press release dates (when available) to identify offer dates. In comparison, we identify the precise offer date from the prospectus. Second, prior literature typically uses the pre-offer day closing price as the benchmark to estimate the offer-price discount, an important determinant of offer-date returns. Instead, we use the last reported trading price to compute the offer price discount. The last reported trading price is used by the issuer and its investment banks in the offer s pricing meeting as the basis against which to set the offer price. Hence, its use provides a more accurate measurement of the offer-price discount. In our sample, the mean offer-price discount is 3%. However, there is a significant difference between the averages for Improved (2.65%) and Regular (3.36%) offers. 4. Stock price response 4.1. Market reactions surrounding SEO announcement and offer dates Table 3 reports the stock price response prior to, and surrounding the offer date. 16 Consistent 16 As mentioned previously, we hand collect offer dates from the original prospectuses on the SEC s web site and do not rely on trading volume spikes or other means used in prior literature. 11

13 with prior studies, we find that both Improved and Regular SEOs experience a significant run-up in stock price over event days (-46, -1) relative to their announcement date (see Panel A). The difference in the pre-filing stock price run-up is statistically indistinguishable between the two types of offerings. Following this price run-up when the offer is filed, the market reacts negatively with an average abnormal return of -2.26% for Improved SEOs and -3.55% for Regular SEOs (see Panel B). This result, consistent with prior literature, suggests that at least initially, investors view all SEO announcements negatively. [Insert Table 3 here] However, as investors and managers learn from each other with additional information exchange during the registration period, the market reacts differently to the two types of SEOs. Specifically, for Regular SEOs, investors continue to react negatively during the registration period (Panel C) and on the offer date (Panel D). In contrast, for Improved SEOs, we document a positive 14.45% abnormal price run-up (which corresponds to an average daily abnormal return of 0.74%) during the registration period (see Panel C) and a positive 1.99% offer-date abnormal return (see Panel D). The difference in average daily return between the Improved and Regular SEOs is over 1% per day during the registration period. In the event window (0, +1) where the offer date is event day 0, the mean abnormal return difference is 3.68%. These differences are both economically meaningful and statistically significant. These results are consistent with our Revised Proceeds Hypothesis. They suggest that Regular SEOs are perceived as overvalued at announcement and continue to be regarded as such upon further scrutiny of the information revealed during the firm s road show. Conversely, for Improved offers, the substantial price run-up during the registration period suggests that the market reassesses the firm s prospects and reacts favorably. The follow-up questions are: What causes the market to react positively to Improved SEOs on their offer date despite the increased offer size and higher price 12

14 pressure? 17 And what causes the market to react even more negatively to Regular SEOs on the offer date? Jegadeesh, Weinstein, and Welch (1993) argue that the market is better informed than the issuer and hence a high return implies that the issuer has underestimated the marginal return to the project. Consistent with their argument, our results indicate that the encouraging registration-period market reaction influences issuers, and managers respond by increasing the offer size. In turn, this managerial response is viewed favorably by the market on the offer date, perhaps as confirmation that the issuing firms can exploit economies of scale in their new investment activities. Thus, the increased offer size and the attendant positive price reaction on the offer date for Improved SEOs are supportive of our Revised Proceeds Hypothesis. In contrast, for Regular SEOs, despite an unenthusiastic market reaction during the registration period, if the firm persists with the offering, it further serves to confirm the market s suspicions that the stock is still overpriced. This argument is consistent with the negative offer-date return observed for Regular SEOs SEO classification and offer-date returns In this section, we check whether the price reaction during the registration period is associated with the Improved status of the SEO. We also examine whether the Improved status of the SEO provides additional information, on the offer date, to the market controlling for the registrationperiod price change. We first run Logit regressions to identify the determinants of the indicator variable, Improved dummy, which equals one if the total value of the SEO s proceeds is greater than the amount filed originally (i.e., an Improved SEO), and is zero otherwise (i.e., a Regular SEO). The results are reported in Panel A of Table Prior literature documents the price pressure effect, suggesting that the larger the relative size of SEOs, the greater should be the downward pressure on stock prices on the offer date. 18 In unreported results, we also include pre-issue operating performance and investment variables in the model. The results are statistically the same. These results are not reported in this section, due to missing values in these variables for more than 40 observations. 13

15 Model 1, Panel A of Table 4 is the most general model. Our main variable of interest is the Registration period average daily abnormal return (RP average daily abnormal return). 19 We find that the RP average daily abnormal return is positively associated with Improved SEOs and is highly significant. This result is robust to inclusion of several control variables and specifications as seen in Models 2 through 4. In particular, when the RP average daily abnormal return is included as the only independent variable (Model 3), the model manifests a pseudo R-square comparable in magnitude to other models. In Model 4, even when we include the change in institutional ownership by assuming that managers have the ability to predict future institutional demand, the RP average daily abnormal return remains significant in the model and the R-square does not improve appreciably. Therefore, the Improved status of an SEO is highly associated with the market reaction in the registration period. As suggested in Panel A of Table 4, the higher the registration-period return, the more likely the offer will be an Improved SEO. [Insert Table 4 here] Next, we examine if the managerial offer-size decision (which results in our SEO classification, i.e., Improved vs. Regular SEOs) conveys additional information to the market and affects the offerdate return. In other words, we test how the market reacts to the Improved SEO decision as reflected in the offer-date return. We run cross-sectional regressions of offer-date returns and the results are reported in Panel B of Table 4. In Model 1, we include the Improved dummy, together with other control variables, and find that its coefficient is positive and significant This suggests that the issuer s offer-date return 19 We employ the Registration period average daily abnormal return because it standardizes the returns over registration periods of different duration across the sample. 20 Prior literature has shown that the offer-price discount (OPD) is inversely related to the offer-date price reaction as investment bankers use OPD as a signal to reveal the issuing firm s quality to their buy-side clients. Therefore, to empirically test our Revised Proceeds Hypothesis, we test whether the effect of the Improved variables is present after controlling for the OPD. Consistent with Altinkilic and Hansen (2003), and Krishnan, Ergungor, Laux, Singh and Zebedee (2010), we also find that the offer date return is significantly inversely related to the OPD. In addition, we find that the offer date return s relation to the Improved status of the SEO is robust to the inclusion of the OPD. 21 We also include a battery of other control variables in the models of Table 4, Panel B, including relative offer 14

16 is highly correlated with its Improved status. However, this association does not necessarily indicate that the market reacts positively to additional information acquired through managerial decisions of an Improved SEO as revealed on the offer date. It is conceivable that the impact of the Improved dummy on the offer-date returns might derive from the registration-period return. Specifically, Panel A of Table 4 shows a significant correlation between the probability of an Improved dummy and the registration period return. Accordingly, the effect of the Improved dummy in Model 1 could simply be a manifestation of the registration-period return. To investigate whether the managerial decision has an effect on the offer-date return, independent of the registration-period return, we perform the following two tests. First, in Model 2 of Panel B, we replace the Improved dummy variable with the registration period return (RP average daily abnormal return). We find that the offer-date return is insignificantly related to the registration-period return, which is an interesting result. Specifically, while the Improved dummy is significantly related to both the registration period return and the offer date return, the latter two are unrelated to one another. 22 This supports the view that the Improved status of the offering is related to information revealed during the registration period. This also implies that on the offer date, it is the managerial response to finalize the Improved status that reveals additional information (beyond what is captured by the registration-period return) to which the market reacts. These results are consistent with the Revised Proceeds Hypothesis. Next, we use a two-stage process to separate the information contained in the registration-period return from the additional information revealed by the manager s revised-proceeds decision on the size, pre-offer market capitalization, pre-offer book-to-market ratio, the reciprocal of the stock price, and stock exchange dummy, etc. These variables are included to control for well-known factors in the SEO-related literature (see, for example, Altinkilic and Hansen, 2003; Corwin, 2003; Mola and Loughran, 2004; Butler, Grullon, and Weston, 2005; Kim and Park, 2005; Lee and Masulis, 2009). 22 To verify these relationships, we perform parametric and nonparametric correlation analyses between Improved dummy, RP average daily abnormal return, and the offer date return. We find robust evidence that the Improved dummy is indeed significantly correlated with both the RP average daily abnormal return and the offer date return. On the other hand, both parametric and nonparametric tests reveal an insignificant correlation between the other two variables. 15

17 offering date. The first stage of the process employs the Logit regression of Model 3 in Panel A of Table 4. This model is used to estimate the expected component of the indicator variable, Improved dummy, based on the registration-period return as the explanatory variable. The unexpected component (i.e., Unexpected Improved dummy) is then computed as the standardized Pearson residual from the Logit model. Thus, the Unexpected Improved dummy captures the additional information revealed by the managerial revised-proceeds decision on the offer date that is not embedded in the registration-period return. The second stage Models 3 through 5 in Panel B of Table 4 employ the Unexpected Improved dummy and the registration-period return (RP average daily abnormal return) in conjunction with other control variables to explore cross-sectional variation in the offer-date returns. With the exception of Model 4, the RP average daily abnormal return is insignificant. More importantly, the Unexpected Improved dummy is positive and significant in all three models. Interestingly, the significance of the Unexpected Improved dummy matches that of the offer-price discount (OPD) implying that the information effects of each are of comparable importance. These results show a robust relation between the offer-date return and the manager s revised-proceeds decision. 23 They suggest, consistent with the Revised Proceeds Hypothesis, that both the market and managers possess incomplete and partially overlapping information sets. The manager learns from the market s reaction during the registration period and revises the offer proceeds accordingly. In turn, the market reassesses the issuing firm s future prospects based on the manager s offer-size revision Although the market reacts positively to stock issuance where the offer size has increased from the original filing, we also find a negative relation between the offer-date stock return and Relative offer size. Presumably, price pressure causes a decline in stock price for sizable equity offerings due to the market s limitations in absorbing an additional supply of shares, as noted in Corwin (2003). Nevertheless, the market responds positively to the increase in offer size for Improved SEOs indicating that the information effect dominates the price pressure related effect for these offerings. 24 In Model 5, we include variables that proxy for institutional holdings over the SEO quarter, which are not observable on the offer date. Despite endowing the market with clairvoyance and the ability to forecast institutional demand perfectly, the Unexpected Improved degree continues to be the dominant variable in predicting offer-date returns. It indicates that our results extend beyond the effect described in Gibson, Safieddine, and Sonti (2004), and Chemmanur, He, and Hu (2009) wherein they show that institutional investors are prescient when it comes to 16

18 The abovementioned result naturally begs the question whether the extent by which the offer is improved is related to the magnitude of the offer-date return. Consequently, we create a new variable, Improved degree, defined as the total dollar proceeds divided by total dollar amount filed originally, minus one. Unlike the binary indicator variable, Improved dummy, the new variable, Improved degree, is continuous; positive for Improved SEOs and non-positive (negative or zero) for Regular offers. Our empirical analysis proceeds along the same lines as that for the Improved dummy. First, in Panel A of Table 5, we present results from regressing Improved degree on the registration-period return to investigate whether the extent to which the revised offer size (as reflected in Improved degree) is predictable based on the stock returns during the registration period. Next, we take the regression residual of the Improved degree that is orthogonal to the registration-period return (i.e., Unexpected Improved degree) and employ it in the second stage (Panel B of Table 5) to explore its relation with the offer-date return. Consistent with our earlier findings in Table 4, Panel A of Table 5 shows that Improved degree is positively related to the RP average daily abnormal return. These results support our contention that the registration-period return is an important determinant of the offer-proceeds revision. In Panel B of Table 5, we use the Unexpected Improved degree to capture the information in Improved degree that is not embedded in the registration-period return and to test its impact on the offer-date return. The coefficient of Unexpected Improved degree is positive and significant, and is independent of the registration-period return after controlling for other factors, including the offer-price discount, OPD. The RP average daily abnormal return is insignificantly related to the offer date return in all models of Panel B, Table 5. [Insert Table 5 here] investing in better performing SEOs. A more comprehensive discussion related to institutional investor demand is contained in Section 6. 17

19 Overall, the results from Tables 4 and 5 show that both Improved dummy and the extent to which the offer proceeds are revised from the initial filing (i.e., the Improved degree) are significantly and positively related to the offer-date returns. While the registration-period return is a key determinant of an SEO s improved status, the managerial response to revise the final offer size contains new information and is significantly associated with the offer-date returns. 5. Long-run stock return performance The Revised Proceeds Hypothesis predicts differential post-seo performance predicated on the status of the SEO as an Improved or Regular offering. In this section, we test this prediction. Specifically, we ask: do Regular SEO firms exhibit inferior performance? Do Improved SEO issuers invest more and perform better than Regular firms? To this end, we examine long-run post-offer stock returns. We employ two methods to track long-run stock returns of SEOs. The first approach employs calendar-time portfolio regressions. This approach is appealing because the time-series variation of portfolio returns accurately captures the effect of correlation across event stocks (Fama, 1998). The cross-sectional dependence, commonly observed in the popular buy-and-hold abnormal return (BHAR) measure, which leads to poorly specified test statistics (Fama 1998; Lyon, Barber, and Tsai 1999; Brav 2000), is less of a concern under this technique. Further, in running calendar-time portfolio regressions we use monthly returns that can greatly mitigate the skewness problem in the BHAR method (Fama, 1998; Mitchell and Stafford, 2000). From February 1997 to December 2006, we form a portfolio of firms that have conducted SEOs in the past two years. 25 We regress monthly portfolio returns against the Fama-French (1993) three factors as follows: We choose two years to examine post-issue stock performance to be consistent with the horizon over which we 18

20 R p, t Rf, t ( Rm, t Rf, t ) ssmb hhml e (1) t t t where Rp is the SEO portfolio return, Rf is the risk-free rate, Rm is the market portfolio return, SMB is the small-firm portfolio return minus the big-firm portfolio return, HML is the high book-to-market portfolio return minus the low book-to-market portfolio return. The abnormal returns are measured by regression intercept alphas in equation (1). Since SEOs generally follow a price run-up (Asquith and Mullins, 1986; Lucas and McDonald, 1990), we also perform the Carhart (1997) four-factor regression by adding the momentum factor to equation (1) as shown below: R p, t Rf, t ( Rm, t Rf, t ) ssmb hhml mwml e (2) t t t t where WML is the past winner portfolio return minus the past loser portfolio return. 27 We apply two portfolio formation schemes, namely equal-weighted (EW) and log valueweighted (LW) approaches, to compute the calendar-time portfolio returns. Previous studies argue that the abnormal performance (if any) of corporate events occurs in small stocks only (Fama 1998; Brav, Geczy, and Gompers 2000). A value-weighted approach is an alternative that can mitigate this small firm effect. However, the results based on a value-weighted approach can be dominated by just a few extremely large firms in the sample. Therefore, we use a log value-weighted approach (by taking the natural logarithm of market capitalization of sample firms as the weight to compute portfolio returns) to reduce the potential impact of including very large issuers. Table 6 shows the two-year abnormal stock returns based on calendar-time portfolio regressions. Panels A and B report the long-run abnormal returns for Improved and Regular SEOs, respectively. We find that Improved SEOs do not exhibit any significant long-run abnormal returns. study analyst forecasts. We also examine the post-issue stock performance based on a three-year holding horizon and the results are qualitatively similar. 26 We drop calendar months with less than 3 stocks in the portfolio to avoid the possibility that a few months with only one or two issuers bias our results. However, our results hold when imposing different minimum numbers of stocks in a calendar month. An alternate solution is to run weighted least squares (WLS) regressions where the weight is the number of observations in each month (see Ikenberry and Ramnath, 2002). In WLS, we can retain all observations but assign a greater (lesser) weight to calendar months with more (fewer) observations. Our findings are robust to this empirical test. The WLS results are not tabulated but available upon request. 27 Risk factors are from Professor Ken French s website 19

21 In other words, their returns are commensurate with their risk factors. Conversely, Regular SEOs generate significantly negative long-run returns regardless of the factor model and weighting scheme we use. In Panel C, we compare Improved versus Regular SEOs. The differences in alpha estimates between the two SEO groups are economically meaningful, ranging from 0.60% to 0.65% per month, and statistically significant. [Insert Table 6 here] This asymmetric pattern in the long-run returns between the Improved and Regular issuers is possibly because good news travels early while negative information is held back. For Improved offers, good news is revealed during the road show and at the offer date. Since the impact of this new information is captured in the stock returns through the offer date, no long-run abnormal returns would occur. Indeed, their long-run stock performance matches their benchmark. On the other hand, Regular issuers have the incentive to postpone the disclosure of negative information. Therefore, when the additional negative information is subsequently released, there is a further negative reaction by the market resulting in long-run underperformance. 28 Additionally, we also examine the post-issue stock returns using Fama-MacBeth (1973) monthly cross-sectional regressions the results of which appear in Table In Model 1, the Improved SEOs outperform Regular SEOs by 0.87% per month (t-stat of 2.56), controlling for firm size and book-to-market ratio. The return difference between the Improved and Regular groups is consistent with the magnitude we previously reported in Table 6. Next, we test if the Improved 28 Titman, Wei, and Xie, 2006 find a strong negative relation between corporate investments and future stock returns. Accordingly, we experiment with using a four-factor model by including the investment factor into the Fama-French model. Our results are robust to this modification. 29 Similar to the calendar-time portfolio regressions, the Fama-MacBeth procedure employs monthly returns, which mitigates the skewness problems in buy-and-hold returns. More importantly, we can use firms market capitalization and book-to-market ratio, rather than factor realizations, to control for the size and book-to-market effects in stock returns. Daniel and Titman (1997) argue that firm characteristics (such as size and B/M) better explain crosssectional stock returns than factor loadings. Hence, this approach may capture abnormal returns more accurately. For sample firms that do not have a positive book-to-market ratio, we include a book-to-market indicator variable. We drop four months with smaller than 10 observations in the regression to avoid measurement errors in estimating the regression coefficients. However, our results are similar when imposing different minimum numbers of stocks in a calendar month. 20

22 degree variable is associated with post-issuance stock returns. Model 2 shows that Improved degree is significantly positively related to post-issue returns. For example, for a 10% increase in final proceeds over the original filing amount, the long-run return would increase by 15.3 basis points per month. The results in Models 1 and 2 suggest that Regular SEOs significantly underperform Improved SEOs in the post-issuance period. Also, as implied by our Revised Proceeds Hypothesis, the more firms revise their offer size upward (downward) relative to original filing amount, the better (worse) their post-issuance stock returns Robustness tests A body of research finds that institutional investors avoid investing in offers that decline subsequently. Gibson, Safieddine, and Sonti (2004) find that issuers who experience an increase in institutional ownership around their offer outperform SEO firms whose institutional ownership declines. Chemmanur, He, and Hu (2009) confirm and extend these findings. They document that institutional investors possess private information about SEOs and are able to identify and obtain more allocations in SEOs with better long-run stock returns. 31 Institutional investors are considered to be sophisticated players and better informed relative to other market participants, and their trades have informational value. 32 Therefore, to examine the market demand for additional share issuance, we investigate and compare the institutional demand for both the Improved and Regular SEOs In unreported results, we also run event-time cross-sectional regressions by regressing post-issuance two-year buy-and-hold returns on Improved dummy (and control variables) and find a significantly positive relationship for the Improved dummy. The Improved degree variable is also positively related to long-run returns in similar tests. 31 The dataset used by Chemmanur, He and Hu (2009) is proprietary and unique. In general, it is not possible to determine aggregate institutional ownership until the quarterly ownership reports (used by Gibson, Safieddine, and Sonti 2004) become publicly available after the SEO. Thus, the institutional ownership data provides an ex post measure of the SEO firms performance. 32 For example, see Yan and Zhang (2009). 33 One possibility is that during the registration period, managers and investment bankers reveal new information to market the offer and institutional investors react to this new information. Alternatively, institutional investors are pleasantly surprised during the road shows, by firms intent to undertake their preferred activities through SEOs, and in response, institutional investors reveal their unanticipated high demand for the offer. Based on the enthusiastic feedback from the market, managers and investment bankers decide to increase the offer size, resulting in Improved SEOs. 21