Seasoned Offerings, Imitation Costs, and the Underpricing of Initial Public Offerings

THE JOliRNAI. OF FINANCE • VOL. XLtV, NO, 2 • JUNE 1989 Seasoned Offerings, Imitation Costs, and the Underpricing of Initial Public Offerings IVO WELCH* ABSTRACT This paper presents a signalling mode! in which high-quality firms underprice at the initial public offering (IPO) in order to obtain a higher price at a seasoned offering. The main assumptions are that low-quality firms must invest in imitation expenses to appear to be high-quality firms, and that with some probability this imitation is discovered between offerings. Underpricing by bigh-quality firms at the IPO can then add sufficient signalling costs to these imitation expenses to induce low-quality firms to reveal their quality voluntarily. The model is consistent with several documented empirical regular- ities and offers new testable implications. In addition, the paper provides empirical evidence that many firm.s raise substantial amounts of additional equity capital in the years after their IPO. EVER SINCE IBBOTSON (1975) FIRST rigorously documented the large underpricing of initial public offerings (IPOs), it has puzzled researchers. Rock (1986) has offered an equilibrium model for this phenomenon in which uninformed investors face a winner's curse when they submit an order for IPO shares. Since informed investors withdraw from the market when the issue is priced above its value, uninformed investors are more likely to receive a full allocation of shares if the offering is overpriced and a rationed allocation if it is not. Firms are forced to underprice their IPOs in order to compensate uninformed investors for this adverse selection. Beatty and Ritter (1986) extend the model to show that the value of information and, thus, both the bias against uninformed investors and the necessary underpricing are higher for issues for which there is greater uncertainty about their value. One problem with the winner's curse explanation of underpricing is that it could be easily avoided. For example, underwriters could reduce the adverse selection problem by offering IPOs only in pools or by agreeing to withdraw an issue or compensate uninformed investors if demand from informed investors is not forthcoming (see also Ritter (1987)). Alternatively, venture capitalists could provide the expertise and capital funding to reduce or avoid IPO underpricing. Yet Barry, Muscarella, Peavy, and Vetsuypens (1988) find that venture-capital- backed IPOs are even more underpriced than non-venture-capital-backed issues. * School of Business, University of Chicago. I am grateful for the support and encouragement of Randolph Beatty, Sandra Chamberlain, Jennifer Dropkin, Robert Gertner, Milton Harris, Mark Lang, Albert Madansky, Peter Pashigian, John Persons, Jay Sankaran, Robert Vishny, the partici- pants of the Workshop of Economics and Econometrics at the University of Chicago, Rene Stulz (the editor), and Jay Ritter (the referee). I would also like to thank Jay Ritter and Hyuk Choe for permitting me access to their data bases. All remaining errors are my responsibility alone. 421 Administrator 高亮 422 The Journal of Finance Ibbotson, in listing possible reasons for the underpricing of IPOs, stated that the prevalent explanation on Wall Street is that issuers may want to " 'leave a good taste in investors' mouths' so that future underwritings from the same issuer could be sold at attractive prices."' This paper formalizes this argument in a two-period signalling model in which firms are rational actors with superior information in a perfectly competitive capital market. The main assumptions are that low-quality firm owners must incur imitation costs to appear to be high- quality firms, and that nature may nevertheless reveal the firm's true quality after the IPO but before a seasoned offering (SO). Consequently, low-quality issuers face a tradeoff. They can invest in imitation activity but face the possible loss of some of this investment if they are discovered, or they can reveal their quality and forego the higher price they could have received at the IPO and the SO had their imitation not been discovered. This paper shows that the additional costs of underpricing can induce low-quality firms to voluntarily reveal their quality when real imitation costs alone are not sufficient. There are three distinctive features of this model. First, the information asymmetry is due to the firm owner knowing more about the firms' value than investors. In Rock's model, the asymmetry is between informed investors on one hand and uninformed firm owners and investors on the other. Second, in my model it is high-quality firms whose quality is not otherwise known by the market that underprice. Third, this model implies that high-quality firms value under- pricing as a signalling device. Therefore, firms have no incentive to avoid underpricing. In Rock's model, firms reluctantly underprice only to keep unin- formed investors in the market. Three recent papers also model IPO underpricing as a signal from better- informed firm owners to less informed investors. In Nanda (1989), projects with higher expected return are also less risky. Since (unidentified) riskier firms would prefer to issue debt over equity because of limited liability, high-quality firms can use (underpriced) equity contracts as a signalling mechanism to communicate their quality. In Grinblatt and Hwang (1989) a generalization of Leiand and Pyle (1977) is presented in wbich a continuum of risk-averse firm types signal two characteristics, the mean and the risk of their projects. To do so, firms employ a second signal, underpricing, in addition to Leiand and Pyle's original signal, the fraction ofthe firm retained by the firm owner. In Allen and Faulhaber (1987), high-quality IPO firms offer to trade off a lower IPO price against a more favorable interpretation of future high dividends. Low-quality firms are more reluctant to offer this exchange, since they are less likely to experience high future cash flows and therefore are less likely to pay high future dividends. Consequently, investors can rationally interpret future dividends more favorably for firms that underprice at their IPO. The latter two models are particularly closely related to this model insofar as, in all three, IPO underpricing results in higher proceeds for high-quality firms in future selling activit}'. In contrast, however, this model focuses on financial markets' ability to observe firms' past and current real activities at the IPO. I assume that there are direct costs that low-quality firms must incur in order to ' See Ibbotson (1975, p. 264). Underpricing of Initial Public Offerings 423 imitate observable operations by high-quality firms. Therefore, this model offers comparative statics quite different from the above models. The remainder of this paper is organized as follows: Section I provides the details of the model, including a description of the assumptions, the sequence of actions and events, and the definition of equilibrium. Section II analyzes the pricing and operating choices of issuers in different equilibria. Section III presents the empirical implications of the model. Section IV provides preliminary empir- ical evidence consistent with a central implication of the model: some firms should issue only a portion of their equity in the IPO and the remainder in SOs. I find that nearly a third of the approximately one thousand IPO firms during 1977-1982 had issued SOs by 1987. Section V contains concluding remarks. I. The Model A. Setting Consider an economy in which risk-neutral individuals own one of two types of firms, high-quality firms, H, and low-quality firms, L. The utility of each firm owner depends only on the sum of the issuing proceeds from an initial public offering (IPO) and a single seasoned offering (SO). The risk-neutral, perfectly competitive market would pay V '^ for a high-quality firm and V'' for a low-quality firm (V'- < V")- However, investors cannot directly verify the quality of an individual firm; they know only the aggregate proportion of high-quality and low- quality firms, h and 1— h, respectively, h can therefore be interpreted as the market's prior that a firm is high-quality. Both high-quality and low-quality owners know their firm's true value, but high-quality owners cannot credibly communicate their knowledge through simply announcing their quality. To receive V" (at least at the SO), high-quality firms may be forced to adopt a mechanism that speaks louder than words. Fortunately for the high-quality owners in this model, nature reveals the true firm value with probability r between the IPO and SO. This represents credible outside information that becomes available randomly. For example, an oil foun- tain may emerge from the drilling site in time or a disgruntled employee may reveal his employer's type. I shall sometimes refer to this type of information disclosure as detection or revelation. Further, assume that it is costly for a low-quality firm to imitate a high-quality firm. There are publicly observable costs that a high-quality owner always finds optimal to incur but that are not optimal for low-quality owners. For example, a high-quality oil company may order a pipeline, while a low-quality oil company would rather not. I assume that these activities (sometimes referred to as operations)^ are efficient for high-quality firms in the sense that, even in a world of perfect information, high-quality firms must perform them to be worth V"; and that the imitation of these operations, with cost C, is of no value to low- quality firms other than to imitate the observable actions or attributes of high- ^ A firm that does not operate may, of course, conduct business other than that necessary to appear to be of high quality. 424 The Journal of Finance Table I Market Value of Firms Conditional on Investors' Information Set and the Firm's Operations (o) A known high-quality firm (H) is worth V" if it operates (o = O), V'- if it does not operate (a = NO). A known low-quality firm (L) is worth V' if it does not operate. If it does operate, it expends costs (C) and is thus worth only V'- - C. If the market cannot distinguish between these firms, it is willing to pay the expected value of an unidentified firm. This is just the value of high-quality firms and low-quality firms weighted by their proportion in the market, ft and 1 - h, respectively. Known Known tf Firm L Firm Unknown Firm Firm operates (o = O) V" V'- - C hV" + (1 - h){V'- - C) Firm does not operate (o = NO) V" V' V"' quality firms. For simplicity, I further assume that high-quality firms that do not operate are worth V'', and that there are no default considerations (V'- - C> 0). Table I summarizes the market values of firms.'' Next, assume that firms are so wealth constrained that they must raise the capital necessary to fund their operations, and that C covers the full outlays necessary for a firm to operate.-* (Recall that only high-quality firms eventually recoup this outlay.) I further assume that firms can raise their minimum certain value after beginning operations iV'- - C) from alternative sources, but that these cannot completely finance the firm's operations, that is V^ - C < C.^ Without loss of generality, the minimum funding and maximum borrowing constraints can be combined into a minimum proceeds constraint where a^P, denotes the proceeds of the IPO, and B the amount a firm borrows from alternative sources. ' Because V" and V"' are defined to be the perfect information values or high-quality firms and low-quality firms, C is subtracted from a low-quality firm's value only if the low-quality firm attempts to imitate a high-quality firm. To illustrate, consider land that can be used for either farming or oil exploration. If there is exploration and oil is found, the value of the land is V . If there is no exploration, the value of the land is that to the farmer, V: These wouid be the full-information values. However, when a low-quality imitator explores oil-empty land, he or she (a) incurs exploration expenses and (b) foregoes farming rent. Therefore, such land would not be worth V'% but only V^ - C. Note also that with limited stockholder liability, it is unclear who would hear possible default costs if I allowed V' < C. In the strictest sense, one may therefore interpret tbis model as applying only to firms with V" > C. ••This assumption was introduced by Leland and Pyle (1977) and has often been adopted in the IPO literature (e.g., see Rock (1986) and Allen and Faulhaber (1987)). ^ Firms that can finance their operations by borrowing may not have to approach the IPO market for a while and thus may be able to—and, as shall be discussed in section HE, may prefer to— advertise sufficiently early to signal quality (for a subsequent IPO) without Securities Exchange Commission (SEC) restrictions. It is also common for IPO firms to engage in several rounds of venture capital financing prior to going public. Firms communicate some proprietary information to venture capitalists, whose investments part.ially certify that the issuer is a high-quality firm. Underpricing of Initial Public Offerings 425 B. Sequence of Actions and Events The game played by managers and investors is now described in more detail, and the sequence of actions and events is specified. The players are the two types of firms and the market. The market is both passive and efficient. Investors purchase an entire offering if and only if, given their current information, its expected value weakly exceeds its price. Firm owners are active players. They maximize the expected sum of proceeds from two offerings, one in each of two stages. In stage 1, the firm concurrently issues its IPO and begins operations, denoted by 0 (o E {O, NO}, where O [NO] stands for operation [no operation]). The market observes o and can then purchase the offering, involving the proportion ai of the firm for price Pi. Firm owners in turn can use the proceeds of this offer to fund operations. Thus, the market uses information from the firm's action triple (rt,, Puo) e ([0, 1] x R^ x \0, NOj) in its decision to purchase the IPO. After the IPO, nature reveals to all players the firm's quality with probability r. Let RH (RL) be the state (denoted by R) in which the firm is revealed to be high quality {low quality), and NR the state where quality is not revealed. I assume that this information is not noisy—// (L | RH) = // (H | RL) = 0, where /- stands for probability—and that the revelation probability is nonzero and inde- pendent of any other variables, in particular, of firm quality: //{R}i\H) = / . (RL|L) = r, and/-.{NR|//) = / . (NR|L) = 1 - r. In stage 2, the market observes another action by the firm, the seasoned offering (SO), analogously denoted (a^, P>) G ([0, 1] X R*). Hence the market can use the information set {(ai, Pi,o), R, («2> Pj) I to judge firm quality in stage 2. The game ends with stage 2. Therefore each firm owner chooses a2 to be 1 — «!, and P2 to be the highest price the market is willing to pay. Market beliefs about firm quality are in turn determined by known past events, first periods' actions by the firm {ai, Pi, 0) and the state of detection R. Consequently, only strategic signalling behavior in stage 1 needs to be considered, ia-,, P>) contains no additional information about firm quality to the market, and \iai. Pi, o), R\ is a sufficient statistic for market beliefs. C. Definition of Equilibrium I first introduce the notation used to define and describe equilibria. Let ai [(ai, Pi, 0) I Q] denote the probability of a firm of type Q £ \H, L\ to choose action («!, Pi, 0) in stage 1, and let cT^iia-z, P^)! Q, (aj. Pi, o), /?l denote the probability of a firm of type Q to choose action («;., P;.) in stage 2 (after [a] having engaged in o operations, [b] offered «] ofthe firm for P] at the IPO, and [c] experienced detection R). The manager's optimal strategy depends on the market's reaction to the specified strategy. Let //i[Q\iai, Pi, 0)] denote the market's belief in stage 1 that the firm is type Q (after having observed [a] the firm's o operations and [b] the firm's IPO of «, of the firm for I\), and let /.o[Q|(a,, Pi, 0), R, (a.. P.)] denote the market's belief in stage 2 that the firm is type Q (after having observed [a] the firm's 0 operations, [b] the firm's IPO of fraction «] of the firm for Pi, [c] the state of detection R, and [d] the firm's SO of a^ of the firm for Pz)- Since 426 The Journal of Finance [d] contains no information about firm quality, /.^[Q \ (ai, Pi, o), R, {a^, P2}] is henceforth abbreviated to /z-^lQ \ («i, Pi, 0), R] to reduce clutter. The link between market beliefs and firm strategies is the equilibrium defini- tion that specifies the way in which each is rational given the other. I employ the concept of Bayesian equilibrium. An equilibrium is a set of strategies ai, a-^, and of (posterior) market beliefs ^1, /^ a, such that (a) at each stage both high- quality and low-quality owners' choice of stragegy—given their own information, future (wealth-maximizing) strategies, and market beliefs—is wealth maximizing, and (b) at each stage the market's posterior beliefs about firm quality are consistent with Bayes' rule and the specified and observed strategies of firms. Thus, for an equilibrium to exist, there must be a set of market beliefs that are self-fulfilling and against which firm's actions are optimal. In this model, this implies that the set of equilibria is defined against market beliefs that consider a firm that acts in an out-of-equilibrium fashion to be of low quality (with value y- ~ C ii 0 = O, V'^ if 0 = NO)—unless, of course, nature reveals this firm to be of high quality. However, as I proceed, I shall apply some of Cho and Kreps' (1987) equilibrium refinements. Since these constrain rational market beliefs such that investors do not consider out-of-equilibrium firms to be of low quality in certain situations, high-quality firms can have an opportunity to profitably deviate. This allows some economically implausible Bayesian equilibria to be eliminated. For example, consider an equilibrium in which high-quality firms are strictly worse off and low-quality firms are strictly better off than in a second equilibrium. Cho and Kreps (1987) argue that the first equilibrium is untenable, because a high-quality firm could adopt the action of the second equilibrium, and make a speech to the market that its (out-of-equilibrium) action should not be interpreted to imply low quality. This is because only a high-quality firm could have an incentive to be compensated according to the second equilibrium. Therefore, all high-quality firms would pursue this out-of-equilibrium strategy, and the original equilibrium is no longer tenable. II. Equilibria of the Issuing Game In this section, three pure equilibria of the issuing game are presented.*' The first equilibrium is a pooling equilibrium in which investors cannot distinguish at the IPO between low- and high-quality firms since both raise enough capital at their IPO to operate and do operate. The second is a first-best separating equilibrium in which all high-quality firms operate but do not underprice, and low-quality firms voluntarily reveal their identify. The third is a signalling equilibrium in *• The choice of presented equilibria is not exhaustive, and I do not even discuss the full set of these equilibria. Instead I present and contrast only some economically plausible equilibria. In particular, 1 omit an equilibrium where neither firm operates and the market believes every firm to be of value V'\ since it is pareto dominated by some other equilibrium over all possible parameter values. Further, I omit a mixed equilibrium in which some bigh-quality firms pool and others underprice. This equilibrium is not included in the text because an argume