When do high stock returns trigger equity issues?

Transcription

1 When do high stock returns trigger equity issues? Aydoğan Altı University of Texas at Austin Johan Sulaeman University of Texas at Austin November 2007 Preliminary. Any comments and suggestions are welcome.

2 When do high stock returns trigger equity issues? Abstract One of the most prominent stylized facts in corporate finance is that equity issues tend to follow periods of high stock returns. We document that firms exhibit such timing behavior only in response to high returns that coincide with strong institutional investor demand for their stock. When not accompanied by institutional purchases, stock price increases have little impact on the likelihood of equity issuance. The results suggest that potential issuers pay close attention to who stands behind their market valuations.

3 Introduction It is well-known that equity issues tend to follow periods of unusually high stock returns. A firm s recent stock return is in fact a better predictor of its equity issuance behavior than most other factors that are relevant for financial policy. These observations have lead many researchers to argue that firms equity issue decisions are largely affected by market timing considerations. Indeed, equity market timing is often described as the practice of issuing shares following a substantial runup in the stock price. In this paper, we take a closer look at the timing of equity issues and find that issuers do not respond to stock returns per se. High stock returns trigger equity issues only when they coincide with strong demand from institutional investors. When not accompanied by institutional purchases, high stock returns have little impact on the likelihood of equity issuance. Thus, issuers seem to care about who stands behind their stock prices as well as how high those prices are. We argue that the sensitivity of equity issues to institutional investor demand is a manifestation of issuers concerns about the sustainability of their stock prices. When firms are privately informed about their intrinsic values, the equity issuance decision is subject to an adverse selection problem; consequently, the market greets issue announcements as bad news (Myers and Majluf (1984)). In practice, both the stock price prior to the announcement and the market reaction following it are determined as outcomes of trading activity among investors with varying qualities of information. In their effort to avoid negative stock price reactions, firms are likely to time their equity issues to coincide with periods during which they expect well-informed investors to be active traders in their stock. We identify institutional investors as the more likely candidates for being well-informed, and recent institutional investor demand as the signal potential issuers monitor. 1 Our empirical analysis concerns the timing and outcomes of seasoned equity offerings (SEOs). 1 See Section I for a more detailed discussion of the informational mechanisms through which institutional investor demand may affect firms equity issuance decisions. 1

4 We highlight two main sets of results. First and as already discussed above, we find that high stock returns trigger equity issues only when accompanied by strong institutional investor demand. In particular, the issuance decision is highly sensitive to the number of institutional investors who establish new holdings in the stock. To give a sense for magnitudes, the unconditional per-quarter probability that a firm announces an SEO is 1.27% in our sample. When the previous quarter s stock return is in the top quintile but the number of institutions initiating new holdings in the stock is in the bottom quintile, the SEO announcement probability is only 0.74%. However, when both the stock return and the number of new holdings are in their respective top quintiles, the SEO announcement probability jumps to 4.85%. A closer look at these periods reveals an increased frequency of large holding initiations by institutions, which is indicative of informed trading activity. Interestingly, firms do not appear to condition on the holding changes of their existing institutional shareholders in deciding whether to issue equity. It is only the demand from new institutional shareholders that matters for the issuance decision. Second, we analyze how the market s response to the equity issuance decision depends on recent institutional investor demand for the issuer s stock. We find that high-demand issuers are able to sustain their stock prices at pre-announcement levels. As is well known, SEO announcements generate negative stock price reactions on average. This initial price reaction is negative in our sample as well, and similar for high- and low-demand issuers. However, while stock prices continue to decline for low-demand issuers until the offer date, they fully rebound from the initial negative reaction for high-demand issuers. In other words, high-demand issuers are able to complete their offers at their pre-announcement stock price levels on average. We interpret this finding as an indication of trading by informed institutions following the announcement. The results discussed above concern institutional investor demand prior to the issuance decision. We also examine whether institutional investors interest in the offer reveals substantial valuation-relevant information to other market participants. We find that issuers with high offerperiod institutional demand substantially outperform those with low demand in the immediate post-offer period. This indicates that market participants pay close attention to the resolution of uncertainty regarding institutions participation in the offer. 2

5 The final part of our analysis focuses on the long-run stock returns of issuers. Previous research has documented that SEOs underperform in the long run. Whether such underperformance is due to mispricing is a source of ongoing debate in the literature. The results discussed above show that firms strongly time their equity issues to coincide with high institutional investor demand. Furthermore, high-demand issuers exhibit substantially better stock return performance than low-demand issuers around and following the SEO. Perhaps firms view periods of high institutional demand for their shares as windows of opportunity in which they can sell overvalued equity. If so, one might expect high-demand issuers to underperform low-demand issuers in the long run. We find that this is not the case. While we confirm the finding in previous studies that SEOs underperform in general, there is no evidence of stronger underperformance for issuers with high institutional demand. In particular, the short-run gains of high-demand issuer that we discuss above are not reversed in the long run. Taken together, our results point to the difficulty of successfully timing equity issues. Firms would clearly like to time the market and issue at price peaks, but investors are likely to recognize this timing motive and price the stock of issuers accordingly. Our results show that firms are concerned about how investors will react to an equity issue even after periods of exceptionally high stock returns. Potential issuers require not only high but also sustainable stock prices. The remainder of the paper is organized as follows. Section I discusses the motivation and hypotheses. Section II describes the empirical setup. Section III presents the results. Section IV concludes. I. The timing of equity issues A. Background and motivation A large body of empirical work in corporate finance analyzes timing of equity issues. Research in this area focuses on three sets of questions. First and of particular importance for the current study, several papers document that firms tend to issue equity when their market values are high 3

6 relative to book or past market values. 2 This tendency is often interpreted as direct evidence of market timing attempts by issuers. The second approach is to detect market timing ex post by examining long-run stock returns of issuers. While the basic finding in this regard is that issuers exhibit low returns, there is an ongoing debate in the literature about whether these low returns result from successful timing. 3 Finally, a number of recent studies analyze the effects of market timing on firms capital structures and suggest that these effects may be quite persistent. 4 The literature described above defines equity market timing as the practice of issuing shares when equity is overvalued. The idea is that firm insiders, having an informational advantage over outsiders, recognize and exploit the so-called windows of opportunity, periods in which market valuations deviate substantially from fundamentals. There is an obvious difficulty with this idea: firms would naturally want to exploit their informational advantage, but why would investors be systematically fooled? The market reaction to the issuance announcement should eliminate any predictable overvaluation and ensure that the newly-issued shares are fairly priced on average a basic prediction of theories of financing under asymmetric information (e.g., Myers and Majluf (1984)). However, proponents of the market timing view argue that the announcement effects observed in practice are too small and exhibit too little time variation to have a meaningful impact on firms timing considerations. 5 More generally, equity market timing is usually portrayed as a primarily behavioral phenomenon whereby firm insiders believe, rightly or wrongly, that they can exploit mispricing successfully. The tendency of equity issues to follow high stock returns is often viewed through this behavioral lens. Other stylized facts about equity issues, however, are more difficult to reconcile 2 Empirical evidence can be found in Taggart (1977), Marsh (1982), Jalilvand and Harris (1984), Asquith and Mullins (1986), Rajan and Zingales (1995), Jung, Kim, and Stulz (1996), Pagano, Panetta, and Zingales (1998), and Hovakimian, Opler, and Titman (2001). Also, survey evidence in Graham and Harvey (2001) indicates that recent stock price performance is one of the most important factors affecting the equity issuance decision. 3 Low post-issue returns are first documented by Ritter (1991) in the context of IPOs and Lougran and Ritter (1995) and Spiess and Affleck-Graves (1995) for SEOs. Market timing interpretation of these findings has been countered by alternative explanations such as benchmark misspecification (Brav, Geczy, and Gompers (2000), Eckbo, Masulis, and Norli (2000)) and changing risk characteristics (Carlson, Fisher, and Giammarino (2006)). 4 See, for example, Baker and Wurgler (2002) and Huang and Ritter (2007). As with long-run underperformance, the persistence results are contested in the recent literature along a number of dimensions. 5 For example, Loughran and Ritter (1995) write...does it make sense that a firm would wait years to issue equity just to save 10 cents on a $25 issue? Our focus is on whether the company can sell at an offer price of $28.80 rather than $20, not whether it will save 10 cents. 4

7 with the market timing view. First, equity issues are quite rare: as Eckbo, Masulis, and Norli (2007) document, only about one-quarter of publicly traded firms ever undertake an SEO following their IPOs. If firms could easily issue equity without triggering large price declines, they would presumably take advantage of overvaluations more frequently. 6 Second, when firms do issue equity, they attempt to place the newly-issued shares primarily with institutional investors. Given their sophistication relative to retail investors, it is difficult to see why institutions would constitute the more likely targets of market timing attempts. The above observations motivate our broad point. We argue that equity market timing is at best a difficult task that requires more than detecting overvaluation of stock. Concerns about receiving a negative market reaction make firms reluctant to issue equity for the most part, even (or perhaps especially) when their stock is favorably valued. Equity issues, which are rare, take place only when issuers feel sufficiently confident that their stock prices will hold their ground. In short, potential issuers require their stock prices to be not only high but also sustainable. B. Institutional investor demand and the equity issuance decision Our specific focus on institutional investors stems from two premises. First, institutional investors are more likely to possess high-quality information on a firm s intrinsic value than do other types of investors (e.g., individuals). This is not to say that most or even some institutions are always well-informed about a given stock. Rather, we argue that when high-quality information does exist among investors (which may be the case only occasionally), it is more likely to be in the hands of a number of institutions. Second, trading by these well-informed institutions causes their information (or at least a substantial part of it) to be revealed to other market participants and hence incorporated into the stock price. The general idea that trading conveys information is standard. In the context of institutional investors, we emphasize a particular asymmetry caused by the presence of short-selling constraints. 7 While high institutional 6 There should be plenty instances of overvaluation (as well as undervaluation) from the perspective of informationally-advantaged firm insiders. 7 Short-selling constraints are prevalent for various types of institutional investors. For example, Almazan et al. (2004) document that a vast majority of mutual funds are subject to either explicit restrictions or implicit policies against short selling. 5

8 demand reveals positive information on the part of purchasing investors, low demand may obtain either when institutions possess negative information (and cannot/do not short sell) or when they possess no material information. Thus, high realizations of institutional investor demand convey more information than do low realizations. B.1 Information content of institutional investor demand: An example Before discussing the implications of the above premises for the equity issuance decision, it is useful to illustrate their impact on stock prices using a simple theoretical example. Consider a setting with three dates (0, 1, 2), no discounting, and risk-neutral investors. There is one firm whose intrinsic value v is distributed uniform in the interval [0, 1]. The firm pays out v as a liquidating dividend at date 2. For the discussion that follows this example, it is useful to think of v as known by firm insiders. At date 0, two informational events occur. First, all market participants observe a noisy signal of firm value, s = v + ɛ, where ɛ is distributed standard normal. The public signal s represents a summary of various valuation-relevant news (e.g., earnings announcements). Second, a group of investors that we refer to as institutions potentially receive high-quality information about firm value. 8 Specifically, with probability q institutions observe v with certainty, whereas with probability 1 q they receive no private information. Institutions hold no shares of the stock at date 0. The stock is traded once at date 1 at market price P. If institutions know that v P, they buy d = N shares. 9 If institutions know that v < P, or if they are uninformed, they buy d = 0 shares. Notice that there is a binding short-selling constraint in the former case. We use the rational expectations equilibrium concept to characterize the stock price. Formally, P equals the expected value of v conditional on all publicly available information, including information reflected by P itself. Informally, one can think of a competitive market maker 8 We emphasize that institutions do not represent all institutional investors; rather, they are a select group of investors that are potentially well-informed about the firm s intrinsic value. 9 One can think of N as the number of informed institutions, each buying one share. 6

9 setting the stock price given the public signal s and institutions demand d. Consider first the two extreme cases. When q = 0, institutions are never informed. In this case, the stock price is based only on the public signal s: P = E(v s) = s + φ(s) φ(s 1) Φ(s) Φ(s 1), (1) where φ and Φ denote the standard normal density and cumulative distribution functions, respectively. When q = 1, on the other hand, institutions are always informed. In this case, the equilibrium price fully reflects institutions information; that is, P = v. Intuitively and using the informal analogy again, the market maker starts from the valuation implied by the public signal, E(v s), and revises the price upward (downward) until demand from institutions subsides (appears). The main case of interest is q (0, 1). The stock price in this case is given by v P = E(v s) (s) if institutions are informed and v E(v s) (s) otherwise, (2) where ( ) is a positive-valued function. 10 The intuition for (2) is as follows. If the condition in the first line of (2) is satisfied, institutions information is revealed by their demand, as in the case with q = 1. If that condition is not satisfied, then d = 0. In response to lack of demand, the market maker lowers the stock price below the valuation implied by the public signal, E(v s). But unlike in the case with q = 1, the market maker stops revising the stock price downward at some point, which is given by E(v s) (s). This is because d = 0 may obtain either when institutions are informed but v is relatively low, or when institutions are uninformed. Since the market maker assigns positive probability to the latter possibility, she calculates a strictly positive expected value for v even in the absence of institutional demand. 10 Formally, the stock price P in the second line of (2) is the unique solution of the following equation: It follows that (s) E(v s) P is positive. 1 (E(v s) P ) φ(s x)dx = q P (P x)φ(s x)dx. x=0 1 q x=0 7

10 Notice that the stock price is determined by two factors, public signal s and institutions demand d. In fact, any stock price P (0, 1) can obtain in two ways: 1. Supported by demand: The public signal s is such that P E(v s) (s), institutions are informed, and v = P. 2. Supported by news: The public signal s is such that P = E(v s) (s), and either institutions are uninformed or they are informed but v < P. The difference between the two cases is the uncertainty about v. In the first case, institutions demand fully reveals v = P. In the second case, P is the best estimate of v given all available public information, but there is non-trivial estimation error. 11 B.2 Implications for the equity issuance decision Consider a firm that contemplates issuing equity. Keeping the firm s current stock price constant, how does recent institutional investor demand affect the firm s issuance decision? The above example offers one answer to this question. Strong institutional demand reveals the presence of well-informed institutions standing behind the firm s current stock price. This reduces the valuation uncertainty faced by other investors (individual investors, other (less-informed) institutions, market makers, etc.), making it easier for the firm to issue equity without triggering a big decline in its stock price. Essentially, strong institutional demand serves the role of third party certification regarding the firm s market valuation and thus alleviates the adverse selection problem. The above discussion describes situations where the stock price already reflects substantial private information (e.g., end of date 1 in the example). In practice, trading activity is likely to convey information at a gradual pace, with institutional order flow slowly building up. Thus, 11 The stark difference between the two cases is clearly due to the simple structure of our example. The more general takeaway is that the short-selling constraint causes an asymmetry in the information content of institutional investor demand. 8

11 firms with pressing capital or liquidity needs may have to make their issuance decisions before institutional demand fully unfolds and the stock price settles. Such firms are likely to respond to increased demand relatively more quickly. Following an initial spike in institutional demand, firms are in a good position to predict whether strong demand will continue or fade. This is partly because firms have inside information about their own intrinsic values, which helps them better evaluate the information content of trading. More importantly, anecdotal evidence suggests that potential issuers utilize their investment bankers communication channels to receive privileged information about pending institutional investor demand. These observations suggest a second rationale for a positive institutional demand-equity issuance relationship. An initial spike in institutional demand alerts a potential issuer to the likely presence of well-informed institutions. If the firm anticipates or knows that strong demand will continue, it will be more confident that the stock price will hold its ground following the offer announcement. In other words, the firm may count on strong institutional demand during the offer period (i.e., between the announcement and the offer) to alleviate the adverse selection problem. 12 II. Empirical setup A. Data and sample construction The initial sample of potential SEO announcers consists of all firm-quarter observations in the intersection of CRSP, COMPUSTAT and CDA/Spectrum Institutional (13F) databases for 84 quarters from 1985 : 1 to 2005 : 4. We restrict the sample to exclude financial firms (SIC codes ), utilities (SIC codes ), and firms that have issued equity within the previous two quarters. In most of our analysis, we also exclude firms that had less than 5% institutional 12 The discussion here is from the viewpoint of a firm with high intrinsic value, for which strong demand is justified. Of course, it is also possible that the firm s intrinsic value is in fact low, that the initial spike in institutional investor demand was just noise, and that a continuation is unlikely. If this is the case, the firm would expect to see its stock price decline following the offer announcement. The important point is that such a firm will still be willing to issue, since the price decline in this case represents a correction of overvaluation. Put differently, a resolution of the informational asymmetry should not reduce the willingness of low intrinsic-value firms to issue. 9

12 ownership in the previous four quarters. Our final sample of potential SEO announcers consists of 242, 601 firm-quarter observations. We use data from Securities Data Corporation (SDC) to identify SEO filing announcement and offer dates. The initial SEO announcements sample consists of all SEC registration filings by firms that are in our final sample of potential SEO announcers. We restrict the sample to exclude spinoffs, unit offers, and rights offerings. Our final SEO sample consists of 3, 133 announcements and 2, 820 completed offers. We use the CDA/Spectrum database to calculate the institutional ownership and demand variables. CDA/Spectrum reports quarterly snapshots of institutional investors portfolios extracted from 13F reports filed with the SEC. 13 The types of institutions covered by this database are banks, insurance companies, investment companies, independent investment advisors, and pension funds. As in previous studies that use this data, we approximate institutions quarterly trades by calculating the difference in their holdings between two consecutive reports. Throughout the analysis we use stock returns that are characteristic-adjusted with respect to size and book-to-market (B/M) benchmarks. To construct the benchmark portfolios used in quarter t, firms are first sorted into five size portfolios based on their market equity at the end of quarter t 2 using NYSE size quintile breakpoints. Within each of these size portfolios, firms are then sorted into five B/M portfolios based on their book equity reported in quarter t 2 divided by market equity at the end of t The characteristic-adjusted stock return is defined as the raw stock return minus the equal-weighted average return of the benchmark portfolio to which the stock is assigned. 13 Institutions are allowed to omit reporting small positions, defined as those that meet both of the following two criteria: (i) the institution holds less than 10,000 shares of a given issuer; (ii) the aggregate fair market value of the holdings in the same issuer is less than $200, We calculate benchmark portfolios with a two-quarter lag to address the concern that issuers may have high returns prior to their SEO announcements for idiosyncratic reasons and hence may not share the characteristics of the firms in the resulting higher size or lower B/M portfolios. The results are very similar when a one-quarter lag is used instead. 10

13 B. Definitions and summary statistics of institutional demand variables Institutional investor demand for a given stock can be measured in several different ways. In this regard, we make two particular choices. First, we use variables that reflect changes in (as opposed to levels of) institutional holdings. While clearly informative about institutions views, level variables are affected by other factors that create noise for our purposes. For example, one would expect certain types of stocks (e.g., larger, more liquid) to have higher levels of institutional ownership. Change variables track demand shifts more closely; they also provide more meaningful comparisons to the effects of stock returns. Second, our demand variables are count-based (e.g., the number of institutions purchasing the stock). The alternative would be to aggregate trades across institutions (e.g., change in institutional ownership). While the two approaches provide qualitatively similar results, the count-based approach appears to capture the information content of institutional demand more effectively. 15 Specifically, we construct two institutional demand variables for each stock i-quarter t pair. Our main variable is new holdings, defined as the number of institutions that initiate a holding in stock i in quarter t. A second variable that we use in some parts of the analysis is change in existing holdings, defined as the number of institutions that increase their holdings (excluding position initiations) minus the number of institutions that decrease their holdings (including position terminations) in stock i in quarter t. Both variables are normalized by the number of institutions holding the stock at the beginning of quarter t. In calculating the institutional demand variables, we restrict attention to firms that have more than 5% institutional ownership during the past four quarters. Panel A of Table I reports distribution statistics of the institutional demand variables for three different samples: the unconditional sample of all firm-quarters (A.1), firm-quarters that immediately precede SEO announcements (A.2), and firm-quarters in which offers take place (A.3). Relative to its unconditional distribution, the variable new holdings is unusually high in 15 For instance, Sias, Starks, and Titman (2006) document that stock returns are more strongly correlated with contemporaneous changes in the number of institutional shareholders than contemporaneous changes in institutional ownership. 11

14 the quarter prior to an SEO announcement. Notice that this is not the case for change in existing holdings; this variable s pre-seo announcement distribution is very similar to its unconditional distribution. These findings provide a first glimpse at the impact of institutional demand on the issuance decision, which we analyze more formally in the next section. As one would expect, new holdings tends to be very high in the offer quarter (A.3); this simply reflects the fact that the placement of newly-issued shares creates new institutional shareholders. Perhaps more surprising is the offer-quarter trading behavior of existing institutional shareholders. Relative to a typical quarter (A.1), these investors appear to be more inclined to reduce their holdings in the offer quarter. Panels B through D of Table I report various correlations of the institutional demand variables. The reported statistics are calculated as time-series averages of quarterly cross-sectional correlations. The highlights from these panels are as follows: - Both institutional demand variables exhibit moderate positive autocorrelation. There is very little correlation between the two demand variables. - New holdings is positively correlated with market-to-book ratio and negatively correlated with firm age and firm size. However, these correlations are quite small in magnitude. - New holdings is positively correlated with past stock returns. This is consistent with the finding in previous studies that institutional investors tend to engage in positive-feedback trading (see, for example, Grinblatt, Titman, and Wermers (1995)). New holdings is positively correlated also with the contemporaneous (i.e., same quarter) stock return, and this correlation is stronger than those with past returns. This is indicative of price impact of institutional trades. Interestingly, change in existing holdings has almost no relationship to past and contemporaneous returns. Overall, the institutional demand variables appear to exhibit substantial variation that is independent of their own past values, firm characteristics, and stock returns. 12

15 III. Results A. The equity issuance decision In this section, we present our main results that relate the likelihood of equity issuance to stock returns and institutional investor demand. We analyze the equity issuance decision at quarterly frequency and measure it by whether the firm makes an SEO announcement during the quarter. Thus, the quantity of interest is the probability that a firm-quarter observation is associated with an SEO announcement. For ease of interpretation and to better illustrate non-linear relationships, we report SEO announcement probabilities for groups of firms sorted based on the quintiles of explanatory variables. The results are quantitatively similar when we resort to regression analyses instead. It is well-known that firms are more likely to issue equity following periods of high stock returns. We start by verifying this stylized fact in our sample. In each quarter, we sort firms into quintiles based on previous quarter s characteristic-adjusted stock return. We then calculate the percentage of firms within each quintile that announce an SEO during the quarter. Panel A of Table II reports the time-series averages of these percentages and the t-statistic of the difference between the highest and the lowest return quintiles. 16 As expected, stock returns have a strong impact on the equity issuance decision. Firms in the highest return quintile are ten times more likely to announce SEOs than firms in the lowest return quintile. We also report the results separately for sub-samples of firms with more and less than 5% institutional ownership. Notice that firms with low institutional ownership are less likely to issue equity in general. Of more interest is the sub-sample of firms with more than 5% institutional ownership, which constitutes the main sample for the rest of the analysis in the paper. 17 Here, we continue to observe the strong relationship between stock returns and the equity issuance decision. Of the 3, 133 SEO announcements in this sample, 1, 477, or 47%, are 16 The t-statistics in this panel, as well as all others in Tables II and III, are based on time-series standard errors that are calculated with a four-quarter lag Newey-West correction. 17 Recall that we calculate the institutional demand variables only for firms with more than 5% institutional ownership during the past four quarters. 13

16 by firms that are in the highest stock return quintile (the numbers of SEO announcements are not reported in Table II). Next, we analyze the likelihood of an SEO announcement as a function of both stock returns and institutional investor demand. In each quarter, firms are sorted independently into quintiles based on previous quarter s characteristic-adjusted stock return and institutional investor demand. Within each return-demand group, we calculate the percentage of firms that announce an SEO during the quarter. The results are reported in Panels B and C of Table II, where the institutional demand variables are new holdings and change in existing holdings, respectively. Panel B presents our main result, namely, that equity issues tend to follow stock price runups that coincide with strong institutional investor demand. When the stock return is in its highest quintile but new holdings is in its lowest quintile, the probability that the firm announces an SEO is only 0.74%. This is quite low relative to the unconditional announcement probability, which is 1.27% in our sample. When both the stock return and new holdings are in their highest quintiles, however, the announcement probability jumps up to 4.85%. To give a sense for the number of announcements, of the 1, 477 announcements in the highest return quintile, 959 are by firms that are also in the highest demand quintile, whereas only 43 are in the lowest demand quintile. Notice that the two sorting variables have a strong non-linear interaction effect on the issuance decision. Moving from an adjacent group to the highest return-highest demand group increases the SEO announcement probability by about two percentage points. In Panel C we repeat the same exercise by sorting firms based on the stock return and change in existing holdings. The picture that emerges here is quite different than in Panel B. The stock return is now the only significant variable in predicting SEOs. Within any return quintile, moving from the lowest to the highest demand group causes at best a very small increase in the announcement probability. Simply put, change in existing holdings does not appear to affect the equity issuance decision. Before discussing the interpretation of the results in Table II in greater detail, we first present two refinements to our tests. We report the results of these refinements only for new holdings, 14

17 since this is the demand variable that has a significant impact on the equity issuance decision. Applying the same tests to change in existing holdings reconfirms the conclusion that this demand variable does not affect the issuance decision. The first refinement concerns the benchmarking of the probabilities reported in Table II. Our analysis so far does not control for various other factors that are known to affect the equity issuance decision. These factors may account for part of the variation in announcement probabilities that we report in Table II. To address this concern, we adjust the announcement probabilities relative to their estimated benchmarks. Specifically, we first estimate the following panel Probit regression for all firm-quarter observations in our sample: Pr(SEO announcement) = F (M/B, firm size, firm age, IPO dummy, profitability, leverage, volatility, turnover, institutional ownership, quarter fixed effects). (3) Regression (3) estimates the probability that a firm will announce an SEO in quarter t + 1 as a function of various firm and stock characteristics measured as of the beginning of quarter t. 18,19 The firm characteristics that appear in the first line of (3) are clearly important determinants of financial policy. The stock characteristics in the second line of (3) are included to control for factors that relate to the liquidity of the stock, which is likely to affect the equity issuance decision as well. For each firm-quarter observation in the sample, we set the benchmark SEO announcement probability as the predicted value from (3). We then replicate the analysis in Panel B of Table II by subtracting the average benchmark announcement probability in each 18 We lag the explanatory variables by one quarter because we are interested in the effects of the stock return and institutional demand in quarter t on the announcement probability in quarter t + 1. Thus, we need to set the benchmark for this probability as of the beginning of quarter t. 19 The variable definitions are as follows. Market-to-book ratio M/B is market equity plus book assets minus book equity divided by book assets. Firm size is the logarithm of book assets in 2005 dollars. Firm age is the logarithm of the number of quarters passed since the first appearance of the firm in CRSP. The IPO dummy equals one if the firm has been public for less than two years (we take the first appearance of the firm in CRSP as the IPO date). Profitability is measured by EBITDA over book assets. Leverage is book debt divided by book assets. Volatility is the standard deviation of daily stock returns measured over one quarter. Turnover is quarterly trading volume divided by shares outstanding. Institutional ownership is the fraction of shares outstanding owned by institutional investors in our dataset. The accounting variables used in M/B, firm size, profitability, and leverage are from the most recent annual financial reports filed by the firm. Since trading volume is reported differently for NASDAQ stocks, we include a separate turnover coefficient for NASDAQ stocks in (3). 15

18 return-demand group from the realized announcement probability. The results are reported in Panel A of Table III. The main takeaway from Table II that the two sorting variables have a strong interaction effect on the issuance decision is robust to the benchmark adjustment. As before, the announcement probability is much higher in the highest return-highest demand group relative to other return-demand groups. Importantly, since the probabilities in Table III are benchmark-adjusted, they can be evaluated based on their absolute values as well and not just relative to each other. In this regard, notice that high stock returns positively affect the issuance decision only when institutional demand is also high. Firms with high returns but low demand are in fact slightly less likely to issue relative to what one would expect based on their characteristics. The second refinement concerns the sources of variation in new holdings. As indicated in Section II.B, new holdings is positively correlated with past stock returns, suggesting the use of positive-feedback trading strategies by some institutions. This raises the possibility that the impact of new holdings on the equity issuance decision is spurious and driven by stock returns prior to the most recent quarter. Another concern is that new holdings is correlated with several firm and stock characteristics. While the benchmark adjustment in Panel A of Table III partially deals with this issue, the adjustment does not control for potential interactions between the stock return and the characteristics. We address these concerns by isolating the surprise component of new holdings. First, we estimate the following linear Tobit regression in the panel of all firm-quarters: 20 New holdings = F (past stock return, M/B, firm size, firm age, IPO dummy, (4) volatility, turnover, institutional ownership, quarter fixed effects). In (4), past stock return refers to the six-month characteristic-adjusted stock return prior to the quarter in which new holdings is measured. Other variables are defined as in (3). Next, we 20 The specification is Tobit because new holdings is bounded below by zero. The results are similar with an OLS specification. 16

19 calculate a new demand variable called residual new holdings, which is defined as the difference between the realized value of new holdings and its predicted value from (4). 21 Finally, we replicate the analysis in Panel A of Table III by sorting firms based on residual new holdings. The results of this exercise, reported in Panel B of Table III, confirm the qualitative conclusions that we reach based on new holdings. Essentially, high stock returns positively affect the issuance decision only when they are accompanied by sufficiently high demand from new institutional investors. 22 We now turn to the interpretation of the results. Our basic finding is that equity issues are most likely following periods in which both the stock return and institutional investor demand are high. What causes the increase in institutional demand in such periods? One possibility is that the increase is at least in part due to purchases by well-informed institutions, as we hypothesize. Another possibility is that we are simply picking up institutions response to samequarter high stock returns. Perhaps institutions herd into purchasing the stock in some periods, and they are more likely to do so when the stock return is high. The equity issues that follow may then represent opportunistic behavior by firms that anticipate continued herd demand to support the stock price and absorb the newly issued shares. In other words, firms may simply be timing irrational swings in institutional demand. The difference between the information and the herding hypotheses is a nuanced one. At a very general level, one can argue that the only way to distinguish between the two is to analyze subsequent returns, as any theory involving information can be recast in terms of perceived (irrational) information. 23 Furthermore, even if the herding story explains firms issuance behavior, it is still interesting to see that firms do not respond to high stock returns indiscriminately. When the stock return is high but institutional demand is low, firms are averse to issuing equity, suggesting that they are concerned about the market reaction. Nevertheless, the data may provide some clues regarding the difference between informed 21 The predicted value calculations account for the truncation in the Tobit error term 22 In unreported analysis, we conducted another robustness test by defining new holdings demand quintiles based on firms rankings relative to other firms in their size and book-to-market groups (i.e., not relative to all other firms). The results are quantitatively similar to those reported in Tables II and III. 23 See Section 3.D for an analysis of long-run stock returns following SEOs. 17

20 trading and herding. If institutions simply respond to high returns by purchasing the stock en masse, one might expect most of these purchases to be small. On the other hand, institutional trades that convey information and move the stock price are likely to be (or perhaps need to be) large relative to shares outstanding. Accordingly, we compare the share of small versus large purchases across different return-demand groups analyzed in Table II. Since we are primarily interested in the effect of new holdings, the purchases in question are holding initiations. Specifically, we define a holding initiation to be large (small) if the number of shares purchased divided by the number of shares outstanding is in the top (bottom) 25% of the distribution of the same variable across all the comparable initiations. The set of comparable initiations is defined as those that take place in the same quarter for firms that are in the same size decile as the firm in question. Table IV reports the results. The numbers reported in Panel A are the average values of new holdings in each return-demand group. Notice that there is a substantial spike in the highest return-highest demand group (41.06 holding initiations per 100 existing institutional shareholders). We are interested in whether this spike represents informed trading or herding. In Panel B, we report the average numbers of large and small holding initiations (both normalized by beginning-of-quarter number of institutional shareholders). If the size of initiations were random, one would expect large and small initiations to each account for 25% of the numbers reported in Panel A. We find that this is not the case. Consider the highest return quintile, which is of most interest. When demand is low, more of that demand comes in the form of small initiations. This picture is reversed as demand increases. For example, of the initiations in the highest return-highest demand group, (about 31%) are large initiations, whereas 7.55 (about 18%) are small initiations. Importantly, there is no discernible increase in the number of small initiations relative to other groups within the highest demand quintile. However, the number of large initiations increases significantly within the highest demand group as the return increases. While not conclusive, the evidence is in line with the hypothesis that high return-high demand periods exhibit an increased frequency of informed trades. There is no evidence that a wave of small purchases dominate these periods. 18

21 We find that equity issues are highly sensitive to new holdings but not sensitive at all to change in existing holdings. What accounts for this stark difference? One explanation is that new holdings is a less noisy indicator of the valuation-relevant information institutional investors possess. Changes in existing holdings are driven to a greater extent by fund flows and diversification considerations. The fact that change in existing holdings exhibits almost zero correlation with the contemporaneous stock return (Panel D of Table I) is consistent with this view. A second explanation unrelated to information is that it is difficult or costly to place the newlyissued shares with the existing institutional shareholders. Purchases by existing holders tend to be fewer and smaller in size relative to new holdings. Persuading existing holders to absorb a substantial amount of shares may thus require large price concessions (e.g., as compensation for holding less diversified portfolios). Accordingly, potential issuers may pay relatively less attention to the trading demand of their existing institutional shareholders. To summarize, we find that equity issues tend to follow periods in which both the stock return and institutional investor demand as measured by new holdings are high. Such periods exhibit an increased frequency of large purchases by institutional investors. The equity issuance decision does not appear to be affected by the trading patterns of existing institutional shareholders. B. Market reaction to the issuance decision By announcing an SEO, a firm indicates its intention to issue equity. The offer itself typically takes place several weeks after the announcement (in our sample, the median number of trading days between the announcement and the offer is 22). At the time they announce their SEOs, firms are primarily concerned about the price at which they will place the new shares. Accordingly, we now analyze stock price changes during the SEO registration period (i.e., the period between the announcement and the offer). In particular, we ask whether the stock return following an announcement in quarter t can be predicted based on institutional investor demand in quarter t 1. The registration period varies in length across SEOs and can be quite long in some cases. We use two different approaches to deal with this problem. First, we analyze 60 trading- 19

22 day post-announcement returns for all SEO announcements (i.e., including those that are not followed by an offer within 60 trading days). This approach provides a clear representation of the relationship between pre-announcement institutional investor demand on post-announcement returns. However, some of the observations correspond to delayed or withdrawn offers. As a second approach, we focus on SEOs that are completed within 60 trading days following the announcement. The 60-day cutoff alleviates the concern that the analysis is dominated by SEOs that stay in registration for extremely long periods. 24 Panel A of Table V reports the results. Firms that announce SEOs are sorted into five demand groups based on their new holdings quintile assignment in the quarter prior to the announcement (i.e., as in Section III.A). Since there are few SEOs with very low pre-announcement demand, Table V reports the results by combining the lowest three demand groups into one. In the first two columns, the sample includes all SEO announcements. The first column reports the SEO announcement effect, defined as the characteristic-adjusted stock return in the ( 1, +1) threeday window around the announcement. The second column reports the characteristic-adjusted stock return in the (+2, +60) window following the announcement. In the third column the sample is restricted to announcements that are followed by an offer within 60 trading days; this column reports the characteristic-adjusted stock return in the (+2,offer) window following the announcement. The last column shows the fraction of announcements that are followed by an offer within 60 trading days. As in previous studies, the initial market reaction to the issuance decision as measured by the three-day SEO announcement effect is negative on average in our sample. More importantly, this initial market reaction does not vary significantly across announcements preceded by high versus low institutional investor demand. In contrast, stock returns following the announcement exhibit significant differences. Firms with low pre-announcement institutional demand perform close to their return benchmarks during the 60 trading days following the announcement About 80% of announced SEOs are completed within 60 trading days. The choice of 60 trading days (about one quarter) is admittedly arbitrary; however, the results are robust to changing this threshold. 25 The return reported in Table V for the low-demand group, 0.83%, is not significantly different from zero (the t-statistic, which is not reported in the table, is 1.24). 20