Product Market Advertising and Initial Public Offerings: Theory and Empirical Evidence

Current Version: May 2005 For helpful comments or discussions, we thank Sonia Falconieri, Gang Hu, Blake LeBaron, Holger Muller, Debarshi Nandy, Tom Noe, Imants Paeglis, Jay Ritter, Susan Shu, Narayanan Subramanian, and James Weston, as well as participants at the 2004 Financial Intermediation Research Society Conference at Capri, and seminar participants at Boston College, Brandeis University, and Tilburg University. We alone are responsible for any errors or omissions.

Product Market Advertising and Initial Public Offerings: Theory and Empirical

Evidence

Abstract

We develop a theoretical model of the interaction between a firm’s product market advertising and itscorporate financing decisions. We consider a firm which faces asymmetric information in both the product andfinancial markets, and which needs to raise external financing to fund its growth opportunity (new project). Anyproduct market advertising undertaken by the firm is visible to the financial market as well. We show that, inequilibrium, firms may use a combination of product market advertising, IPO underpricing, and underfinancing(raising a smaller amount of external capital than the full information optimum) to convey its true productquality and the intrinsic value of its projects to consumers and investors. Further, we characterize the conditionsunder which firms will use only advertising and underfinancing as signals, and those in which they will use acombination of IPO underpricing, advertising, and underfinancing. Our model has several predictions for IPOunderpricing and product market advertising, two of which are as follows. First, firms will choose a higher level ofproduct market advertising when they are planning to issue new equity or other information-sensitive securities,compared to situations where they have no immediate plans to do so. Second, product market advertising andIPO underpricing are substitutes for a firm going public. We present empirical evidence consistent with thesetwo predictions of our model.

1 Introduction

The role of the underpricing of initial public offerings (IPOs) in signaling firm insiders’ private information

to the equity market has been extensively analyzed (see, e.g., Allen and Faulhaber (1989) or Welch (1989)).

However, recently, some authors have questioned whether underpricing is the most efficient way to signal firm

value, and have raised the possibility that it may be more efficient for firms to use other signals around new

equity issues. For example, Ritter and Welch (2002) comment in their recent review of IPOs: “On theoretical

grounds, however, it is unclear why underpricing is a more efficient signal than, say, ... advertising.” Advertising

is a particularly interesting signaling alternative to underpricing, since there is some anecdotal evidence that,

in practice, some managers may indeed attempt to convey their firm’s intrinsic value to the financial market

making use of product market advertising (particularly in the context of an upcoming IPO). Consider, for

example, the following quote from an article in the Wall Street Journal (“In Web Firms’ Ad Blitz, an Eye on

Wall Street,” August 19, 1999): “As they plaster ads everywhere consumers might turn, companies are hoping

to catch investors’ eyes too...businesses are often as interested in selling stock as in selling products; a high

voltage advertising spree could serve as a critical prelude to an initial public offering.”1 Of course, whether

these and similar anecdotes reflect the special situation, during a special time period, of only a few companies

(e.g., internet firms going public during the bubble period), or whether they reflect the general situation of firms

making equity issues, is an empirical question which has been unanswered so far in the literature.

The objective of this paper is to explore, both theoretically and empirically, how the extent of product

market advertising undertaken by a firm may affect (and be affected by) the prospect of its upcoming IPO.

We develop our theoretical analysis in a setting where insiders of a private firm, with information about its

intrinsic value superior to outsiders, raise external financing to fund its positive net present value project. We

address several related questions in the above setting. First, how will a firm choose the extent of its product

market advertising in a setting where this advertising is visible to the financial market as well as the product

market? Second, will the equilibrium level of advertising chosen by a firm be different in situations where

1 Another article, in the Boston Globe (“Internet Firms Bet on Super Bowl Ads to Reach Investors, Consumers,” January23, 2000), deals with television advertising during the Super Bowl. We quote: “Hoping to impress Wall Street as well as fans,advertisers with names such as Pets.com, Lifeminders.com, and Ourbeginning.com will pony up at least $2.2 million for each 30second slot on next Sunday’s football game. Skip Pile, of Pile and Co., a Boston firm that helps companies hire ad agencies, offersthe following appraisal: ‘A Super Bowl ad can legitimize a brand among multiple constituencies: the company’s employees, venturecapitalists, the investment community, and lastly, the target audience of football fans.”’

1

the firm plans to make a public offering of new equity (either an IPO or a seasoned equity offering (SEO))

compared to a situation where it has no immediate plans to make such an offering of equity or other financial

assets?2 Third, of three alternative signals easily available to firm insiders, namely, underpricing, advertising,

or “underfinancing” (raising a smaller amount of equity than the optimal in a full information setting), under

what conditions will each signal be employed (either individually, or in combinations with the other signals)?3

In particular, are advertising and underpricing substitutes for a firm in the context of its IPO? After addressing

the above questions theoretically, we go on to provide some evidence regarding two of the implications of our

theory. In particular, we empirically document how firms may alter the extent of their advertising in the context

of an upcoming IPO, and how the extent of advertising relates to the extent of underpricing in the IPO.

In our theoretical model, a private firm has an existing product, an ongoing project, and a growth opportunity

(new project). Firm insiders have private information not only about the quality of the firm’s products, but also

about the true value of its projects: in other words, the firm faces asymmetric information in both the product

and financial markets. While the firm has some internal capital available, this capital is not adequate to cover

both the investment required in its ongoing project and to fund its growth opportunity. The firm therefore needs

to raise external financing for investment by making an IPO. We assume that any product market advertising

undertaken by the firm is visible to financial market investors as well.

In the above setting, product market advertising can be thought of as playing two different roles. The first

role played by advertising is that of signaling quality to the product market, thereby allowing consumers to

price the firm’s products correctly in equilibrium.4 In our setting, however, product market advertising plays

a second role: that of signaling the true value of a firm’s projects to potential stock market investors, thus

allowing them to price the firm’s equity correctly in equilibrium. Since a firm’s product quality and the value

of its projects may not be perfectly correlated, an outsider (consumer or potential investor) who knows only

the true quality of a firm’s existing product cannot perfectly infer the true value of its projects; conversely, an

2 We will present much of our analysis in the context of a private firm raising external capital by making an IPO of equity.However, our analysis goes through with minor modifications for the case of a publicly traded firm making a seasoned issue ofequity or other information-sensitive financial assets. Our model therefore has implications for these situations as well.

3 If a firm underfinances, it is forced to underinvest in its growth opportunity. Further, even though outsiders may not be ableto directly observe the investment level chosen by a firm, they can infer its investment level by observing the amount of externalcapital raised by it. Thus, underfinancing and underinvestment are equivalent signals in our setting.

4 This product market role of advertising in our model is similar to the role played by advertising in the industrial organizationliterature, which has long argued that product market advertising plays an important role in conveying information about productquality to consumers (see, e.g., Nelson (1974), Kihlstrom and Riordan (1984), or Milgrom and Roberts (1986)).

2

outsider who knows only the true value of its projects cannot infer existing product quality perfectly. However,

the firm need not use product market advertising alone as a signal, either to the product market or to the

financial market: in a setting where the firm interacts with the equity market as well as the product market, it

can also signal by underpricing equity in its IPO, or by underfinancing.

In equilibrium, the firm uses the least-cost combination of the above three signals to convey its product

quality as well as project value to outsiders. The equilibrium choice of signaling mix by the firm depends on

the extent of asymmetric information facing the firm and the internal capital available to it. Consider first the

case where the extent of asymmetric information facing the firm is relatively small. In this case, firms with

superior quality products and higher intrinsic value projects (which we refer to as higher type firms) will use

underfinancing alone as a signal, since, by itself, underfinancing is a less costly signal for the higher type firm

to use than either advertising or underpricing. While underfinancing requires the higher type firm to scale back

its investment in its growth opportunity, the cost of this underinvestment is partially offset by the reduced

dilution in insiders’ equity holdings (which results from its raising a smaller amount of external financing).5

Therefore, if the extent of underfinancing required to deter mimicking by lower type firms (those with either

inferior products, lower intrinsic value projects, or both) is small enough that the higher type firm has to sacrifice

investment only in the less productive range of its growth opportunity, then it can be shown that the firm will

use only underfinancing as a signal.

Consider now the case where the extent of asymmetric information facing the firm is more severe, so that, if

the firm were to use only underfinancing as a signal, the reduction in investment required would be significant

enough that the firm would have to sacrifice investment in the higher productivity range of its growth opportu-

nity. In this case, the cost of signaling by underfinancing alone becomes prohibitive, and the higher type firm

can lower its aggregate signaling cost by adding either advertising, underpricing, or both, to its signaling mix.

The firm’s choice between advertising and underpricing as the signal to add to underfinancing depends upon the

internal capital available to it prior to the equity issue. In order to advertise, the higher type firm needs to cut

back on investment in its ongoing project, thereby reducing firm value and diluting insiders’ equity holdings.

If the internal capital available to the firm is large enough that only investment in the low productivity range

5 By dilution, we refer to the fact that when a firm sells equity at a lower price, insiders have to give up a greater share of thefirm’s equity to new investors in return for external financing. If the firm raises only a smaller amount of external financing, thisdilution in insiders’ equity holdings will be smaller.

3

of the ongoing project has to be sacrificed to fund the required amount of advertising, then it can be shown

that the firm will add only advertising to the signaling mix. If, however, the internal capital available to the

firm is smaller, so that the firm has to sacrifice investment even in the high productivity range of its ongoing

project, then minimizing the signaling cost involves the firm adding both advertising and IPO underpricing to

the signaling mix, thus using all three signals to convey its true type to outsiders.6

Our model has several implications for product market advertising in the context of equity issues, as well for

IPO underpricing. First, our model predicts that firms will choose a higher level of product market advertising

when they are planning to issue new equity compared to situations where they have no immediate plans to sell

new equity or other information-sensitive securities. Second, it predicts that, in the context of an IPO, product

market advertising and IPO underpricing are substitutes: the greater the extent of product market advertising,

the lower the extent of underpricing. Further, the extent of advertising and underpricing undertaken by the firm

depends upon how much internal capital is available to it prior to going public: as the firm is more financially

constrained, it cuts back on the amount of advertising, while increasing the extent of underpricing. Third, it

predicts that, when they plan to issue new equity or other information-sensitive securities, firms may advertise

even in the absence of asymmetric information in the product market. Finally, our model implies that, in

many situations, a combination of IPO underpricing and advertising (as well as underfinancing) may provide

a lower cost signal to the product market compared to a situation where the firm is constrained (for some

reason) to signal using advertising alone. In particular, it implies that firms which are financially constrained

will signal to the product and financial markets using a combination of advertising and underpricing (as well as

underfinancing).

Our paper is the first in the literature to demonstrate theoretically how product market advertising can

serve as a signal to the financial market in the context of new equity issues. It is also the first to study the

interaction between product market advertising and underpricing in the context of IPOs, characterizing the

6 The relative cost of underfinancing, advertising, and underpricing as signals depends upon how expensive it is for the highertype firm to undertake each activity, relative to the cost to lower type firms undertaking the same activity to a similar extent(i.e., their cost of mimicking). When the extent of asymmetric information facing the firm is relatively small, underfinancing willbe used alone as a signal, since the cost of a given extent of underfinancing is lower for a higher type firm compared to the samecost for a lower type firm which seeks to mimic it. In contrast, if a higher type firm funds advertising by cutting back only oninvestment in the low productivity range of its ongoing project, then the cost of advertising for the higher type firm will be similarin magnitude to the cost for a lower type firm which seeks to mimic it. Thus, advertising alone will be a costlier signal comparedto underfinancing in our setting. Finally, for underpricing alone to serve as a signal, the higher type firm has to price equity in itsIPO below the intrinsic value of the equity of the lower type firm (since, otherwise, the lower type would not incur any cost frommimicking). Given this, the cost to a higher type firm of diluting insiders’ equity holdings arising from underpricing will always begreater than the cost to a lower type firm which seeks to mimic the higher type firm. This means that underpricing alone will be acostlier signal than either underfinancing or advertising.

4

conditions under which various combinations of advertising and underpricing will be employed by firms to

signal their private information in equilibrium. Thus, we provide insight into the ongoing debate in the IPO

literature (mentioned earlier) regarding the efficiency of using underpricing as a signal. Our theoretical analysis

demonstrates that it is indeed efficient for firms which are financially constrained to include both underpricing

and advertising in their signaling mix.7 Our paper makes an additional contribution by demonstrating how

IPO underpricing can allow the firm to signal quality more effectively to the product market by reducing the

amount of advertising required.

We provide evidence consistent with two of the predictions of our model making use of advertising (and other

product market) data as well as the financial market data of a sample of firms going public (IPOs). Consistent

with the first implication of our model, we find that firms indeed increase their product market advertising

in the year of their IPO relative to a benchmark year (when they had no immediate plans to make an equity

offering). Further, we find that, in the five year span around the IPO year (i.e., the IPO year, and the two years

before and the two years after the IPO year), the peak advertising level is reached in the IPO year.8 Second,

we find that product market advertising and IPO underpricing are indeed substitutes: controlling for various

other variables, we find that the extent of underpricing is smaller as the extent of product market advertising

is greater. Our evidence indicates that managers do indeed take into account the effect of their product market

advertising on investors in their firms’ equity when making their advertising and corporate financing decisions.

This research is related to several strands in the theoretical corporate finance literature. Apart from the

theoretical literature (discussed earlier) demonstrating the use of underpricing to signal insiders’ private in-

formation at the time of an IPO, this paper is perhaps most closely related to the theoretical literature on

the interaction between the product and financial markets in the context of IPOs: see, e.g., Bhattacharya and

Ritter (1983), who point out that one cost of going public arises from the need to release information in the

firm’s IPO prospectus (since this information may be used adversely by competitors in the product market);

7 The existing literature (e.g., Allen and Faulhaber (1989) and Welch (1989)) has demonstrated that underpricing can signalinsiders’ private information about firm value to the financial market. In this literature, IPO underpricing works as a signal onlybecause insiders price equity in the IPO in anticipation of a second round of financing subsequent to the IPO and the possibilityof true firm value being revealed exogenously between the two rounds of financing. In contrast to the above literature, here IPOunderpricing serves as a signal in the context of a one-shot equity offering.

8 This prediction of our model applies not only to initial public offerings, but also to seasoned equity offerings (SEOs). Inregressions not reported in the paper, we also provide evidence consistent with this prediction in the context of SEOs. We find thatthe amount of product market advertising of a firm in the year of its SEO is greater than the advertising amount in a benchmarkyear two years prior to the SEO. The results from these regressions are available upon request. Note to the referee: While theSEO results are not to be published, we have attached them to the paper as a separate Appendix (Appendix B)for the purpose of refereeing.

5

and Maksimovic and Pichler (2001)), who study how the possibility of such an information release to product

market competitors may affect the timing of a firm’s going public decision.9 Stoughton, Wong, and Zechner

(2001) argue that the decision of a firm to go public may serve to signal high quality to the product market.10

None of the above papers, however, address the role of product market advertising in IPOs (or in other equity

issues); neither do they study how product market advertising interacts with underpricing in the context of

IPOs.

To the extent that we also empirically study underpricing in IPOs, our paper is also related to the large

empirical literature on IPO underpricing (see Ritter and Welch (2002) for a review). The two empirical papers

most closely related to ours, however, are Grullon, Kanatas, andWeston (2004) and Demers and Lewellen (2003).

The former paper documents that firms with a greater level of product market advertising have a significantly

larger number of both individual and institutional investors in their equity, lower bid-ask spreads (indicating a

smaller amount of adverse selection in the market for these stocks), smaller price impacts, and greater market

depth. In contrast to the above paper, which focuses on the behavior of stock market investors in response

to increased product market advertising by a firm, our empirical analysis focuses on how firms choose their

advertising levels in the context of a new equity issue. However, by demonstrating the role played by product

market advertising in conveying firm insiders’ private information to equity market investors, our theoretical

analysis is able to explain the reduction in adverse selection associated with a greater level product market

advertising documented by Grullon, Kanatas, and Weston (2004). Demers and Lewellen (2003) document that,

for a sample of internet firms, IPO underpricing is associated with a post-IPO increase in their website traffic.

Unlike their paper, where the focus is on how internet firms’ IPOs affect their product market demand, our

empirical analysis focuses on how the prospect of a firm’s IPO affects its choice of advertising level, and how its

choice of product market advertising level affects underpricing in its IPO. However, our theoretical analysis has

implications for the empirical findings of Demers and Lewellen (2003) as well (see implication 5 in section 4).

The rest of this paper is organized as follows. In section 2, we describe the essential features of our model,

and in section 3, characterize its equilibria. In section 4, we describe some of the testable predictions of our

9 The broader theoretical literature on the going public decision (e.g., Chemmanur and Fulghieri (1999)), and the broadertheoretical literature on IPO underpricing (e.g., Chemmanur (1993)) are also indirectly related to this paper.

10 Our paper is also indirectly related to the literature on product and financial market interactions outside the context of equityissues: see, e.g., Gertner, Gibbons, and Scharfstein (1988) analyze an informed firm’s choice of financial structure when the financingcontract is observed not only by the capital market but also by a competing firm.

6

model. In section 5, we provide evidence consistent with two of these predictions. We conclude in section 6.

The proofs of all propositions are confined to the appendix.

2 The Model

The model has four dates (time 0, 1, 2, and 3). Consider an entrepreneur owning a private firm, which has

available to it an amount of internal capital W . At time 0, the firm has an existing positive net present value

project which it plans to fund using (part of) this internal capital W .11 The firm has an existing product,

which it plans to sell in the product market in two rounds of sales, at time 1 and time 2, respectively.12

A new positive net present value project (growth opportunity) becomes available to the firm at time 1. In

order to fund this new project, the firm has the opportunity to go public at time 1. If it chooses to go public,

outsiders come to know this decision right away. The firm chooses the price of the equity to be issued in the

IPO, as well as the amount of the external capital C to be raised in the IPO, at this time. If the firm goes

public at time 1, it invests the external capital raised, C, in addition to any amount left over from its internal

capital W , in the new project. If, however, the firm does not go public at time 1, it invests only the left-over

capital in the new project at time 1.

At time 0, the firm chooses the amount of investment to be made in the existing project, as well as the

amount of product market advertising to undertake, A. We assume that any advertising the firm undertakes in

the product market is observable not only by consumers, but also by potential investors in the financial market.

All cash flows to the firm are realized at time 3 including the cash flows from the two rounds of sales

of the firm’s product and that from the firms’ investments in its two projects. We assume that all agents

(entrepreneurs, consumers, and investors) are risk-neutral, and that the risk-free rate of return is zero. The

sequence of events is depicted in figure 1.

11 Our assumption of an existing project merely serves to capture the notion that the firm has other ongoing business activitiesfor which it also requires funding. In other words, the assumption of an existing project allows us to endogenize the fact thatthe firm has an opportunity cost of internal funding, with this opportunity cost increasing as the amount of internal financingavailable is smaller. All our results go through even in the absence of an existing project if we make the above assumption aboutthe opportunity cost of the firm’s internal financing exogenously.

12 We assume that issuing equity is the only source of external financing, so that firms cannot fund their projects by issuing debtor other financial assets.

7

1 2 3

• Firm determines the amount of advertising to be undertaken.

• Firm invests in the existing project, using its internal capital.

• Firm goes public in the stock market: it determines the amount of capital raised and the pricing of equity in the IPO.

• Firm implements its investment in the new project using the external capital raised from the IPO and the internal capital left over from time 0.

• First round of sales of the firm’s product.

Second round of sales of the firm’s product.

• All cash flows are realized including the cash flows from the two rounds of sales of the firm’s product, and that from the firm’s investments in its two projects.

• All asymmetric information in the product and financial markets is resolved.

t = 0

Interim signal of product quality is received.

Figure 1: Sequence of Events

2.1 Information Structure in the Product and Financial Markets

Both the product and financial markets are characterized by asymmetric information. In particular, we

assume that there are products of two quality levels in the product market: Superior (S) and Poor (P ), with the

quality of superior products being higher than that of poor products. We also assume that there are two kinds

of projects: Good (G) and Bad (B). For any given level of investment, good projects yield higher cash flows

than bad projects (i.e., the NPV of good projects is greater than that of bad projects at any given investment

level). While firm insiders know the true quality of the firm’s product as well as the net present value of its

projects, outsiders (be they consumers in the product market or investors in the financial market) observe only

the prior probability distributions over product and project types.

In order to capture the above asymmetric information in the simplest possible manner, we assume that there

are three types of firms: H (high), M (medium), or L (low). We denote firm type by k, k ∈ {H,M,L}. We

assume that a type H firm always has superior quality products as well as good projects; a type M firm has

poor quality products but good projects; a type L firm has poor quality products and bad projects.13 Figure

2 summarizes our definition of firm type as a function of product quality and project value.

While firm insiders observe the type of their own firm, outsiders observe only the prior probability distribution

13 Thus, we assume here that the product quality and project value are correlated, but not perfectly so. Clearly, assuming thatboth product quality and project value are perfectly correlated is somewhat less interesting, since, in that case, if product qualityis revealed, project value is completely known as well. Note that our results are not driven by the correlation between productquality and project value. At the expense of some additional modeling complexity, it can be shown (using a four type model) thatour results hold even when product quality and project value are completely uncorrelated.

8

Bad (B)Poor (P)Low (L)

Good (G)Poor (P)Medium (M)Good (G)Superior (S)High (H)

Project ValueProduct QualityFirm Type (k)

Figure 2: Firm Types

across firm types at time 0: outsiders believe that the firm is of type H with probability γ1; of type M with

probability γ2; and of type L with probability γ3; γ1 + γ2 + γ3 = 1. At time 3, all cash flows are realized so

that the asymmetric information between firm insiders and outsiders is resolved completely.

2.2 The Firm’s Product and the Product Market

We assume that, in the absence of asymmetric information, consumers value superior quality products at

RS , and poor quality products at RP , RS > RP . We assume that the demand for the firm’s products is perfectly

elastic, so that, in a full information setting, RS and RP will be the prices prevailing in the product market for

superior and poor quality products, respectively. Each kind of product will be sold in the product market for

two rounds, i.e., at time 1 and time 2. Between time 1 and time 2, i.e., after the first round of sales of the firm’s

product but before the second round, a signal of product quality becomes publicly available. We can think of

this signal as a (probabilistic) breakdown of the firm’s product. We assume that the probability of breakdown is

α for a poor quality product and 0 for a superior quality product, with 0 < α < 1.14 The production capacity

of the firm is Q units, which is not variable. The production costs for the superior and poor quality products

are assumed to be the same, and normalized to zero for analytical simplicity. We assume that quality is not a

choice variable for the firm: i.e., each firm is endowed with technology which allows it to manufacture products

only of a given quality.

2.3 The Firm’s Projects and the Financial Market

For simplicity, we assume that the investment technology for firms with both good and bad projects is

piecewise linear, as depicted in Figure 2. Thus, for good projects, the productivity of investment is g when the

investment 0 < I ≤ i1; it is b, 1 < b < g, when i1 < I ≤ i2, and 1 for I > i2. For bad projects, we assume the14 Thus, in the case of pooling in the product market at time 1, customers update their prior at time 2 based on the signal of

product quality occurring between time 1 and time 2. In particular, at time 2, a product is believed to be of superior quality with

a probabilityγ1

γ1+(1−γ1)(1−α) and poor quality with a probability(1−γ1)(1−α)

γ1+(1−γ1)(1−α) in the absence of a breakdown of the product.

9

NPV of marginal investment: b-1

NPV of marginal investment: b-1

NPV of marginal investment: g-1

Amount of Investment

NPV

i1 i2

NPV of marginal investment: 0

NPV of marginal investment: 0

Good Projects

Bad Projects

Figure 3: Investment Technologies for Good (G) and Bad (B) Projects

productivity is b if 0 < I ≤ i2, and 1 for I > i2. We can think of i1 as the high productivity threshold level of

investment; i2 as the full-investment level (in the sense that for either kind of project, if the investment made

is below i2, then NPV is wasted). We assume that these investment technologies are common knowledge. We

also assume that gb >

RSRP, implying that the ratio of the intrinsic values of good and bad projects is greater

than the ratio of full-information prices of superior and poor quality products.

As discussed before, the firm has two projects: an existing project initiated at time 0 when the firm is private

(and funded with only internal capital), and a new project which becomes available at time 1 (which may be

funded with money raised in the financial market if the firm goes public at time 1). Both projects of any given

firm are of the same type.15 We assume that the internal capital owned by each firm before going public,

W ≤ i2, so that any firm can finance at most only one project (either the existing or the new project) to its

full investment level by using internal capital. Thus, if the firm wishes to fund both its existing and its new

project to the full investment level, it has to go public at time 1. In this case, the firm will fund its new project

partially or fully using the external capital raised from its IPO, depending on whether there are any funds

from its internal capital W left over after investing in the existing project. Thus, the magnitude of the internal

capital W is a measure of the degree of the financial constraint faced by the firm prior to its IPO. Throughout

this paper, we assume that the productivities of both types of projects, i.e., g and b, are large enough that all

firms will choose to go public to raise capital for investment in their new projects even after accounting for the

15 This assumption is made only for convenience. Our results are driven only by asymmetric information about the new project,and will therefore be qualitatively unchanged in the absence of asymmetric information about the firm’s existing project.

10

cost associated with doing so (i.e., the dilution of firm insiders’ equity).

2.4 The Firm’s Objective

The firm’s objective is to maximize the expected long-term (time 3) value of equity held by the entrepreneur

(or equivalently, the present value of expected total cash flows to the entrepreneur). The firm’s revenue arises

from two sources: the two-rounds of sales of its product, and the revenue stream from investment in its existing

project (at time 0) and in its new project (at time 1). Thus, the entrepreneur strategically chooses the firm’s

expenditure on advertising, the amount of external capital raised by going public, the amount of investment in

its projects, and the price of equity in the firm’s IPO, to maximize this objective.

3 Equilibrium

Equilibrium strategies and beliefs in our model are defined as those constituting a Perfect Bayesian Equilib-

rium (PBE) satisfying the Cho-Kreps intuitive criterion, and which minimizes the dissipative costs of separation

incurred.16 In other words, the equilibrium concept we use is that of a Pareto dominant or efficient Perfect

Bayesian Equilibrium which survives the Cho-Kreps intuitive criterion.17 We will focus only on separating

equilibria, where the three types of firms fully reveal their types to both the financial and the product markets

in equilibrium.18 Before going on to characterize the equilibria of our model, we analyze the problem faced by

each type of firm.

3.1 Analysis of the Firm’s Problem

We now analyze the trade-offs faced by the three types of firms in arriving at their equilibrium strategies.

In particular, we analyze how each firm arrives at its equilibrium choice of signals. In our discussion below, we

will focus primarily on the type H and type M firms (and not on the type L firm), since, given the equilibrium

choices made by the type H and M firms, the type L is always worse offmimicking the above firm types compared

to its payoff if it follows its full information equilibrium strategy.

16 See Fudenberg and Tirole (1991) for a formal definition of a PBE and Cho and Kreps (1987) for a definition of the Cho-Krepsintuitive criterion.

17 See Milgrom and Roberts (1986) or Engers (1987) for a detailed discussion of why the notion of a Pareto dominant or efficientPBE is the appropriate equilibrium concept here.

18 Given that two dimensions of asymmetric information exist in our model, one can also think of a variety of pooling equilibriasatisfying PBE, involving either pooling in the financial market, or in the product market, or in both markets. However, most ofthese equilibria do not satisfy the Cho-Kreps intuitive criterion, and thus do not satisfy our equilibrium definition. The focus ofthe analysis in this paper will therefore only be on separating equilibria.

11

3.1.1 The Type H Firm’s Problem

The objective of the type H firm is to maximize the expected long-term (time 3) value of the equity held by

the entrepreneur. This is accomplished by maximizing the type H’s value while minimizing the dilution in the

entrepreneur’s equity holdings in the firm. This, in turn, is achieved by ensuring that the firm is able to raise

the optimal amount of external capital, while separating itself from lower type (type M and type L) firms. The

type H can use three alternative signals to accomplish this: (1) raising less capital externally than the amount

that would be raised in a full-information setting (underfinancing), with the resulting underinvestment in its

new project;19 (2) expending resources on advertising; and (3) underpricing the firm’s equity in its IPO. Denote

by Πk, k ∈ {H,M,L} the entrepreneur’s objective function given that his firm is of type k. Further, let Vk

represent the firm value at time 3; Ck the external capital raised by the firm in the IPO market, and I0k and I1k

the amounts invested by the firm in the existing project (at time 0) and the new project (at time 1), respectively.

Further, denote Πk(j) and Vk(j) the type k firm’s objective function and total firm value, respectively, if it is

perceived as a type j, where k, j ∈ {H,M,L}; k 6= j. Finally, let Fk, represent the value of the entire equity of

the firm at the IPO price. Then, the type H firm’s problem is given by:

MaxAH , FH , IH , CH

ΠH = VH

µ1− CH

FH

¶, (1)

subject to the incentive compatibility (IC) constraints which ensure that both the type M and type L firms do

not mimic it:

ΠM = VM

µ1− CM

FM

¶≥ ΠM(H) = VM(H)

µ1− CH

FH

¶, (2)

ΠL = VL

µ1− CL

FL

¶≥ ΠL(H) = VL(H)

µ1− CH

FH

¶, (3)

and the feasibility constraints which ensure that the sum of the firm’s investments in its existing and new

projects do not exceed the total capital available at each point in time:

I0H + I1H ≤ W + CH −AH , (4)

I0H ≤ W −AH . (5)

19We assume that the amount raised by the firm in its IPO is publicly observable, as is the case in practice. Given the observabilityof the amount raised, outsiders can also infer the investment level chosen by the firm. Therefore, throughout this paper, we willuse the terms underfinancing and underinvestment interchangeably.

12

In any separating equilibrium, the firm cannot overprice equity in its IPO, since, otherwise, there would be no

demand for its equity. This market rationality constraint is given by:

Fk ≤ Vk, (6)

where Fk < Vk implies that a firm underprices equity in its IPO. In the following analysis, we define an

“underpricing factor” fk =FkVk, which measures the extent of underpricing in the IPO.20

The solution to the above problem involves the type H firm choosing that combination of signals which

minimizes firm insiders’ aggregate signaling cost. In order to understand the trade-off driving the type H’s

equilibrium choice of signals, it is useful to study the cost of each signal separately. Consider first the case

where the type H firm attempts to use underfinancing in the IPO market alone as a signal. In this case, it will

have to cut back on investment in its new project, thus losing part of its value. If the type M or type L firm

chooses to mimic the type H, they will also have to incur a similar loss in value as the type H firm. However,

this cost in firm value due to underfinancing is partially offset by the reduced dilution in the entrepreneur’s

equity that results from the firm raising a smaller amount of external financing. The benefit from this reduction

in dilution is greater for the type H firm (since its intrinsic value is greater) than for the type M or the type L

firm, thus allowing the type H to use underfinancing as a signal.

Consider now the case where the type H firm attempts to use advertising alone as a signal. In order to

advertise, the type H firm needs to cut back on investment in its existing project, thereby reducing firm value.

This loss in value also results in an increase in the dilution of the entrepreneur’s equity holdings in the firm. If

the type M or type L firm attempts to mimic the type H, they will also have to incur a loss in firm value of

similar magnitude to the type H. However, given that the intrinsic value of the type H is greater than that of

the type M or type L, this dilution in insiders’ equity holdings will always impose a greater cost on the type H

firm compared to that on the type M or type L firm. Thus, advertising alone will be a costlier signal compared

to underfinancing.

Finally, consider the case where the type H attempts to signal using IPO underpricing alone. Given that the

underpricing of a firm is not directly observable by outsiders (recall that only the IPO share price is observable),

20 Thus, optimally, the type H would like to choose the signal that reduces ΠL(H) and ΠM(H) the most while reducing ΠH theleast (i.e., to achieve the maximal effect on the type L and the type M while incurring the smallest signaling cost). Mathematically,

if we define s as the signal chosen, then the type H chooses s such that

¯̄̄∂ΠH / ∂s

∂ΠL(H) / ∂s

¯̄̄and

¯̄̄∂ΠH / ∂s

∂ΠM(H) / ∂s

¯̄̄are the lowest.

13

the type H can use underpricing as a signal only by pricing its equity below the intrinsic value of a type M firm

(if it wishes to prevent the type M from mimicking) or below the intrinsic value of a type L firm (if it wishes

to prevent both the type M and the type L from mimicking). Further, given the type H firm’s higher intrinsic

value, the cost arising from the dilution in insiders’ equity holdings resulting from underpricing is greater for the

type H firm than the corresponding cost to a type M or type L firm if they attempt to mimic the type H. This

means that underpricing alone will be a costlier signal to the type H than either underfinancing or advertising.

Given the relative cost of the three possible signals discussed above, the equilibrium (least-cost) combination

of these signals is determined as follows. If the extent of underfinancing required to deter mimicking by the

type M or type L results in the type H firm cutting back on investment in its new project only in the low

productivity range of its investment opportunity set, then it will use underfinancing alone as a signal. However,

if the extent of underfinancing required to deter mimicking by lower firm types is so large as to require the

firm to cut back on investment in the high productivity (as well as low productivity) range of investment in its

new project, the type H firm can minimize its aggregate cost of signaling by adding either advertising alone or

both advertising and underpricing to the equilibrium signaling mix, depending on the amount of internal capital

available to it. If the internal capital available is large enough that the firm can fund the required amount of

advertising by cutting back only on the low-productivity range of investment in its existing project, then it will

add only advertising to the signaling mix. If, however, the internal capital available is small, so that it has to

cut back also on the high productivity range of investment in its existing project to fund the required amount

of advertising, then the type H firm will add both underpricing and advertising to the signaling mix.21

In the rest of the paper, we wish to focus on the more interesting cases where advertising or underpricing

(or both) enters the equilibrium mix of signals (in other words, the type H and type M do not find it optimal

to signal using underfinancing alone). This will be the case when the investment opportunity set available to

the type H and type M firms is such that they will always have to cut back on the high productivity range

21 The benefit to the type H firm from separating itself from the type M and type L firms is that it will be able to obtain itstrue value in the equity market, thus reducing the dilution in the entrepreneur’s equity holdings at the time of the firm’s IPO. Anequilibrium where the type H firm separates itself from the type M and type L firms will exist here, since the above benefit tothe type H firm will be greater than the benefit to the type M and type L firms from mimicking it. This is due to two reasons.First, given the type H firm’s higher intrinsic value, the signaling benefit of the reduction in dilution in the entrepreneur’s equityholdings is greater for the type H compared to the type M or the type L. Second, recall that the products of the type M and typeL firms (being of poor quality) may break down between the two rounds of sales with a certain probability, while the type H firm’sproduct (of superior quality) has zero breakdown probability. Thus, if a type M or type L firm attempts to mimic the type H byusing the same signal, they may not be able to reap the full benefits of doing so, since the breakdown of their product between thetwo rounds of sales may reveal their true quality.

14

of investment in their new project if they are to signal using underfinancing alone. This is guaranteed by the

following parameter constraint, which will be assumed to hold throughout the paper:

i2 − i1 < i1(g − b)Min(2i1,W + i1)

(b− 1)[2RSQ+ bMax(0,W − i1) + gMin(2i1,W + i1)]. (7)

3.1.2 The Type M Firm’s Problem

Clearly, the equity value of the type H firm will be higher than that of the type M in a full information

setting, while the equity value of the type L will be lower than that of the type M. This implies that the type

H has no incentive to mimic the type M, while the type L has such an incentive. Thus, the type M chooses

the least-cost combination of advertising, underpricing, and underfinancing to distinguish itself from the type

L, thus signaling its true value to the financial market. The type M therefore maximizes:

ΠM = VM

µ1− CM

FM

¶, (8)

subject to feasibility constraints similar to (4) and (5), the IC constraint that the type L firm does not mimic:

ΠL = VL

µ1− CL

FL

¶≥ ΠL(M) = VL(M)

µ1− CM

FM

¶, (9)

and the IC constraint ensuring that the type H does not mimic, which is satisfied trivially. The trade-offs facing

the type M in choosing the optimal combination of signals to deter mimicking by the type L is similar to the

trade-offs discussed earlier in the context of the type H firm. However, while the type H firm is concerned about

the type L mimicking it in terms of both product quality and project value, the type M is concerned about the

type L mimicking it only in terms of project value (since the product qualities of the type M and type L firms

are the same).

3.2 Benchmark Equilibria

Before we begin our analysis, it is worth keeping two benchmark cases in mind.22 Consider first the case

where there is no asymmetric information in either the product or the financial market. In this case, no firm

type advertises, since the only reason for advertising in our setting is to signal firm type. Consider now the case

where the product market is characterized by asymmetric information, but the firm does not raise capital in

the financial market. In this case, firms with a superior product use advertising to signal their quality to the

22 Due to space limitations, we will provide only brief and intuitive discussion of these benchmark equilibria here. Explicitcharacterizations of these equilibria are given in the working paper version of this article.

15

product market, though (in general) to a lesser extent than in a situation where they intend to raise capital in

the financial market (as we discuss in more detail in later sections).

3.3 Equilibria with Asymmetric Information in Both the Product and FinancialMarkets

We now study the situation where, in addition to selling its products in the product market, the firm

proposes to raise new equity, with both product and financial markets characterized by asymmetric information.

3.3.1 Equilibrium with Advertising But No Underpricing

When the type H and type M firms have relatively large amounts of internal capital available, the least-

cost combination of signals involves the firm advertising in the product market and underfinancing (i.e., raising

an amount less than the full-information optimal amount) in the IPO market, but does not involve any IPO

underpricing. We characterize this equilibrium below.

Proposition 1 (Equilibrium with Advertising and Underfinancing) When the internal capital availableW ≥ a2 + i1, the equilibrium in the financial and product markets involves the following:The type H firm: It spends an amount a2 on advertising, raises an amount i1 (less than i2, the full-

information optimal level) from its IPO, and invests amounts of (W − a2) and i1 in its existing and its newprojects, respectively. It prices its equity in the IPO at its intrinsic value.The type M firm: It spends a smaller amount a1 on advertising, raises the same amount i1 from its IPO,

and invests amounts of (W − a1) and i1 in its existing and its new projects, respectively. It prices its equity inthe IPO at its intrinsic value (a1 and a2 are defined in the appendix).The type L firm: It does not advertise, raises an amount i2 from its IPO, and invests amounts of W and

i2 in its existing and its new projects, respectively. It prices its equity in the IPO at its intrinsic value.All three types of firms price their products at their true (intrinsic) value to consumers.23

When the type H and type M firms have enough internal capital available, they can fund their advertising

expenditures by reducing their investment only in the low productivity range of their existing project. This

means that the opportunity cost of using advertising as a signal is lower than that of using underpricing. Further,

the type H and type M firms do not find it optimal to deter the type L from mimicking by using underfinancing

alone, since this would be too costly for them. Thus, in equilibrium, both types of firms use a combination of

product market advertising and underfinancing to signal their types.

In equilibrium, the type H advertises more than the type M because information asymmetry exists about its

product quality as well as its project value. This asymmetric information creates an incentive for both the type

M and type L (which have poor quality products) to mimic the type H (which has superior quality products).

23 In equilibrium, the market infers a type H firm with probability 1 if the firm incurs an amount a2 on advertising; infers atype M firm with probability 1 if the firm incurs an amount a1 on advertising; and infers a type L firm with probability 1 if thefirm does not advertise or advertises any amount other than a1 or a2.

16

Thus, it is beneficial for the type H to advertise more than the type M so that it can distinguish itself from the

other two types in terms of both product quality and project value, and avoid mispricing in both the product

and financial markets.

In equilibrium, the type L chooses not to mimic either the type M or the type H, given the large advertising

expenditures it would incur if it were to mimic. Thus, it chooses to raise the full-information (first-best) level

of investment by going public, prices equity in its IPO at the fair market value, and it does not advertise in the

product market.

It is instructive to compare this equilibrium with the benchmark equilibrium where the firm does not raise

capital in the financial market. It can be shown that the extent of product market advertising is greater when

the firm intends to raise external financing relative to the situation when it does not intend to do so. This

is because the information asymmetry about its product now affects the type H firm not only through the

product market, but also through the financial market. This spillover effect of product market asymmetric

information to the financial market, together with the asymmetric information about the value of the firm’s

projects, increases the type L’s incentive to mimic the type H, compared to the case where the firm does not

raise external capital: by mimicking, the type L is not only able to overprice its product, but is also able to

overprice its equity, thus reducing the dilution in its entrepreneur’s ownership. The type M’s incentive to mimic

the type H is also increased when the firm raises external financing, since it now has the opportunity to sell

overpriced equity. Thus, the type H firm is forced to provide stronger signals to prevent mimicry by the lower

type firms when it intends to raise external financing.

3.3.2 Equilibria with Both Advertising and Underpricing

When the type H and type M firms have smaller amounts of internal capital than in the situation char-

acterized in the previous section, these firms will use underpricing as well as advertising and underfinancing

as signals. The extent to which they use underpricing as a signal, however, will depend upon the amount of

internal capital available to these firms prior to the IPO. In the next two propositions, we characterize the

equilibrium as the amount of internal capital available to the firm becomes progressively smaller.

Proposition 2 (Equilibrium with Advertising, Underpricing, and Underfinancing (I)) When the in-ternal capital available is such that a1 + i1 ≤W < a2 + i1, the equilibrium in the financial and product marketsinvolves the following:The type H firm: It spends an amount (W − i1) on advertising, raises an amount i1 (less than i2, the

full-information optimal level) from its IPO, and invests i1 in both its existing and its new projects. It prices

17

its equity at a fraction f2 of its intrinsic value (f2 is defined in the appendix).The type M firm: It spends a smaller amount a1 on advertising, raises the same amount i1 from its IPO,

and invests amounts of (W − a1) and i1 in its existing and its new projects, respectively. It prices its equity inthe IPO at its intrinsic value.The type L firm: It does not advertise, raises an amount i2 from its IPO, and invests amounts of W and

i2 in its existing and its new projects, respectively. It prices its equity in the IPO at its intrinsic value.All three types of firms price their products at their true (intrinsic) value to consumers.24

When the internal capital available to the firm is small enough that it has to sacrifice investment also in the

high productivity range of investment in its existing project in order to fund the required amount of advertising,

the opportunity cost of advertising exceeds the cost of dilution arising from IPO underpricing. In this case,

it is optimal for the type H firm to use underpricing as well as advertising and underfinancing to separate

itself from the type M and the type L firms. Therefore, the type H firm sets the price of its equity below its

full-information value (i.e., underprices its equity in the IPO), thereby reducing the benefit of the type L and

the type M firms from mimicking it. In other words, including underpricing as part of the signaling mix allows

the type H firm to advertise less than it would have to, if it were to use only advertising and underfinancing as

signals. In summary, the equilibrium (least-cost) signaling mix involves the type H making use of advertising,

underfinancing, and underpricing as signals.

In contrast to the type H, the type M is not concerned about the type L mimicking it in terms of product

quality, since its product quality is the same as that of the type L. It therefore does not need to advertise

as much as the type H in equilibrium, so that it is less affected by its limited internal capital compared

to the type H. Consequently, the type M needs to cut investment only in the low-productivity range of its

investment opportunity set in order to advertise to its equilibrium level, so that its cost of using advertising

and underfinancing as signals is smaller than the cost of dilution associated with using underpricing as a signal.

Thus, the type M uses only advertising and underfinancing as signals, and prices equity in its IPO at its full-

information value (i.e., it does not underprice). Finally, since the type L does not find it profitable to mimic

the type H or type M, its equilibrium behavior is similar to that discussed under proposition 1.

We now study the case where the internal capital available to a firm at time 0 is smaller (i.e., its financial

constraint is more severe) than in the situation characterized in proposition 2.

Proposition 3 (Equilibrium with Advertising, Underpricing, and Underfinancing (II)) When the in-ternal capital available W < a1+ i1, the equilibrium in the financial and product markets involves the following:

24 In equilibrium, outsiders infer a type H firm with probability 1 if the firm incurs an amount W − i1 on advertising and infera type M firm with probability 1 if the firm incurs an amount of a1 on advertising. They infer a type L firm with probability 1 ifthe firm does not advertise or incurs any amount other than W − i1 or a1 on advertising.

18

The type H firm: It spends an amount (W − i1) on advertising, raises an amount i1 (less than i2, thefull-information investment level) from its IPO, and invests i1 in both its existing and its new projects. It pricesits equity in the IPO at a fraction f2 of its intrinsic value.The type M firm: It spends the same amount (W − i1) on advertising, raises the same amount i1 from

its IPO, and invests i1 in both its existing and its new projects. It prices its equity in the IPO at a fractionf1(> f2) of its intrinsic value (f1 is defined in the appendix).The type L firm: It does not advertise, raises an amount i2 from its IPO, and invests amounts of W and

i2 in its existing and its new projects, respectively. It prices its equity in the IPO at its intrinsic value.All three types of firms price their products at their true (intrinsic) value to consumers.25

In this equilibrium, the type M (as well as the type H) firm underprices equity in its IPO, unlike in the

equilibrium characterized in proposition 2, where only the type H underprices in equilibrium. When the internal

capital available to the firm is smaller, the type M firm (as well as the type H) has to sacrifice investment in the

high-productivity range of investment in its existing project in order to fund advertising to the extent required

to deter mimicking by the type L. This implies that the type M firm’s aggregate signaling cost would be larger if

it uses only advertising and underfinancing as signals compared to the case where underpricing is also included

in its signaling mix. The equilibrium (least-cost) combination of signals therefore involves the type M firm using

IPO underpricing, as well as advertising and underfinancing, as signals.

The type H also uses a similar combination of signals (driven by a trade-off similar to that faced by the

type M). In equilibrium, the type H firm prices its equity in its IPO below that of the type M firm, due to two

reasons. First, since the levels of advertising of the type H and type M are the same in this equilibrium, the

type H needs to set a lower price for equity in its IPO to deter the type M from mimicking. Second, unlike the

type M, the type H has to prevent mimicry by the type L in terms of both product quality as well as project

value, which is achieved by the type H setting a lower price for its equity than the type M (so that the extent

of underpricing will be greater for the type H than for the type M).

Part of the benefit to the type H firm from underpricing arises from being able to distinguish itself from

the type M and the type L firms in the product market, and thus receiving a higher price for its products.

Conversely, part of the type H firm’s benefit from having a higher level of product market advertising arises

from being able to distinguish itself from the type M and the type L firms in the financial market, thus receiving

a higher price for its equity in its IPO. In summary, in our setting, a firm’s product market advertising has

25 In equilibrium, outsiders infer that the firm is of type H with probability 1 if they observe a firm setting an advertising levelW − i1 with a low share price in its IPO (equal to the equilibrium share price of a type H firm as specified in proposition 3). Theyinfer a type M firm with probability 1 if they observe a firm setting the same advertising level but a higher share price (equal tothe equilibrium share price of a type M firm as specified in proposition 3). They infer that the firm is of low type with probability1 if they observe that the firm making the IPO does not advertise or spends an amount other than W − i1 on advertising.

19

important spillover effects on the market for its financial assets, by serving to reduce the extent of asymmetric

information about the true value of its projects. At the same time, the extent of underpricing and underfinancing

in a firm’s IPO have spillover effects on its product market, since they help to reduce the extent of advertising

required to signal quality to the product market.

4 Implications and Testable Hypotheses

Our model has several testable implications, which we describe below. We will test implications 1 and 2 in

the empirical section of this paper.

1. Product market advertising and new equity issues: Our model predicts that, on average, firms will advertise

more extensively in years when it plans to make a public equity or other information-sensitive security offering,

relative to years when it does not plan to make such an offering. This implication follows from a comparison

between the equilibrium in proposition 1 and the benchmark case (where the firm does not raise capital in the

financial market). When the firm does not plan to issue new equity, higher intrinsic-value firms will advertise

only to the extent required to distinguish themselves from lower intrinsic-value firms in the product market

alone. When they plan to issue new equity, such firms will advertise more, since they need to distinguish

themselves not only in the product market, but also in the financial market. Thus, the first hypothesis (H1)

we test is that the amount of product market advertising of a firm in the year of its IPO will be greater than

the advertising amount in a benchmark year prior to going public.

2. IPO underpricing and product market advertising as substitutes : Our model predicts that firms will use

product market advertising and IPO underpricing as substitutes to distinguish themselves from lower intrinsic-

value firms when they go public. This implication follows directly from propositions 1, 2, and 3. Thus, the

second hypothesis (H2) we test is that the greater the amount of product market advertising of a firm in the

year of its IPO, the smaller the extent of IPO underpricing, and vice versa.

3. Product market advertising in the absence of product market information asymmetry : Our model predicts

that, if a firm plans to make a public offering of new equity or of any other information-sensitive security, it

may choose to undertake some product market advertising even in the absence of asymmetric information in

the product market (or of any product market activity altogether). Here the motivation for product market

advertising is not to convey information regarding quality to the product market, but to convey information

20

regarding the intrinsic value of the firm to the financial market.26

4. Relationship between the extent of asymmetric information in the financial market, IPO underpricing, and

product market advertising : Our model predicts that, when a firm plans to make a public offering of new

equity or of any other information-sensitive security, the extent of product market advertising and (or) IPO

underpricing will be greater if the financial market where it plans to make the security issue is characterized by

a greater extent of asymmetric information.

5. Effects of IPO underpricing on the product market : Our model implies that, in many situations, a combination

of IPO underpricing and advertising (as well as underfinancing) may provide a lower cost signal to the product

market compared to a situation where the firm is constrained (for some reason) to signal using advertising alone.

In particular, our model implies that firms that are financially constrained will signal to the product market (as

well as to the financial market) using a combination of advertising and underpricing (as well as underfinancing).

In other words, when firms are financially constrained, firms may use underpricing as additional “advertising”

for their products. Consistent with this implication, Demers and Lewellen (2003) document that, for a sample of

Internet firms, IPO underpricing is associated with a statistically significant post-IPO increase in their website

traffic.27

5 Empirical Evidence

5.1 Data and Sample Selection

We obtain our initial sample from the Thomson Financial Securities Data Corporation (SDC) new issues

database. We exclude from our initial IPO sample spin-offs, ADRs, unit offerings, LBOs, limited partnerships,

financial firms (SIC codes 6000 through 6999), and utilities (SIC codes 4900 through 4999). Thus, our initial

sample contains 3815 IPOs between 1990 to 2000. We then extract financial statement information for the IPO

firms in our sample from Standard & Poor’s Compustat files, and stock prices from the University of Chicago’s

Center for Research in Securities Prices (CRSP).

26 This implication is in contrast to that of the industrial organization literature, where the objective of advertising is only tosignal quality to the product market. A real-world example of such advertising solely for the purpose of signaling to the financialmarket is provided by the Japanese Telecom company NTT DoComMo running full page advertisements in the U.S. and otherWestern media (e.g., the Economist) for several weeks even though their products were not available (and not expected to beavailable in the immediate future) in these markets. When the company was contacted regarding the motivation for taking outthese advertisements, they mentioned that the company’s objective was to “increase its shareholder base” (the firm had an ADRlisted in a U.S. Exchange). We thank James Weston for this anecdote.

27 We thank an anonymous referee for pointing out this implication of our model.

21

In order to empirically investigate the effect of going public on a firm’s advertising, we need to identify for

each IPO firm a benchmark year when the firm’s advertising policy is unaffected by considerations related to

going public. In the following, we define the year two years prior to the IPO as the benchmark year (year t− 2

hereafter), since the advertising two years prior to the IPO year is unlikely to reach potential investors in the

IPO market.28 29 We decided not to choose the year one year prior to the IPO year (year t − 1) as the

benchmark year in view of the possibility that there may be some lingering effects of advertising in that year

on the firm’s potential investors in the IPO year (year t). Thus, we exclude from our initial sample those firms

with missing data on advertising expenditures (compustat item #45) two years prior to the IPO year. We also

delete the outliers in our sample by winsorization on the advertising expenditures. These processes reduce our

sample significantly to only 178 firms.30 Note that in some of our empirical analysis we are missing up to 3%

of the final sample due to incomplete information on the financial variables other than advertising expenditures.

Panel A of table 1 reports the summary statistics of the change in advertising from two years prior to the

IPO year (year t− 2) to the IPO year (year t), the change in advertising two years subsequent to the IPO year

(year t + 2), the difference between the above two changes, and the first-day initial return following the IPO.

We define RET , the first-day return on the firm’s equity (IPO underpricing), as the percentage change from the

IPO offer price to the first day closing price. We measure the normalized advertising in the IPO year, ADVt, as

AtAssett

, where At is defined as the cost of advertising, media, and promotional expenses in the IPO year t and

Assett is defined as the book value of assets in the IPO year t; the lagged advertising expenditures, ADVt−2 and

ADVt+2, asAt−2

Assett−2and At+2

Assett+2, i.e., the normalized advertising in year t− 2 (i.e., the benchmark year) and

t+2, respectively. We define ∆ADVt−2,t asAt−At−2Assett−2

, i.e., the difference between advertising expenditures in the

28 We need the advertising expenditure in year t − 2 in the tests of both of our hypotheses. In the test of H1, it is used as thebenchmark for advertising in the IPO year t, so that we can estimate the increase in advertising expenditure which results fromthe firm’s going public decision. In the test of H2, we use it as an instrumental variable for the advertising in year t, so that wecan control for any potential endogeneity problem (when we check the robustness of our OLS regression).

29 As we will show later when we discuss table 2, we find no significant pair-wise correlation between advertising expenses twoyears prior to the IPO and the variables associated with the firm’s going public decision such as initial return, IPO proceeds, etc.This suggests indirectly that the firm’s advertising policy in year t− 2 is unaffected by its going public in year t.30 This filtering greatly reduces our sample size due to two reasons: First, some firms choose not to report advertising expenditures

at all for strategic reasons or include these advertising expenses in other accounting variables. Second, Compustat does not havethe financial information (including advertising expenses) for some firms in the years before they go public. In both cases, we areunable to distinguish whether the missing values on advertising expenses are due to zero advertising or due to lack of reporting.Thus, we choose to select into our final sample only those firms which report advertising expenses for two years prior to their IPOyear. Our sample size is reduced from the initial sample size to 1767 firms due to the first reason. Due to the second reason,our sample reduces to 733 firms when we exclude those firms with missing values on advertising expenditures one year prior tothe IPO year, and further reduced to 178 firms when we exclude those firms with missing values and winsorize those firms withextreme values on advertising expenditures two years prior to the IPO year. Our results would be qualitatively unchanged withoutwinsorization, although they would be a little weaker.

22

IPO year t and that in the benchmark year t−2, scaled by the book value of assets in year t−2, and ∆ADVt,t+2

as At+2−AtAssett

, i.e., the difference between advertising expenditures between year t and year t + 2, scaled by the

book value of assets in year t.31 Column (1) of panel A shows that ∆ADVt−2,t is significantly positive, and

that ∆ADVt,t+2 is negative but insignificant. The test on ∆ADVt,t+2 −∆ADVt−2,t shows that it is negative

and significant. These results suggest that, in a five year span surrounding the IPO year, a firm’s advertising

expenditure peaks in the IPO year. As a robustness check, in column (2) of panel A, we present results on

the percentage change of advertising expenditures, i.e., we scale ∆ADVt−2,t and ∆ADVt,t+2 by advertising

expenditures in year t− 2 and t respectively. The results on the percentage change of advertising expenditures

are even stronger than those presented in column (1), suggesting again a firm’s advertising expenditure peaks

in the IPO year. Panel A also shows that RET is significantly positive at the 1% level, which implies that a

firm underprices the equity in its IPO.

Panels B, C, and D of table 1 report the means and medians of RET , ∆ADVt−2,t, and ∆ADVt,t+2 after

several simple sorts: the 1990s (January 1990 - December 1998) versus the internet bubble period (January 1999

- December 2000), tech firms versus non-tech firms, large firms versus small firms (and test for differences in

these sub-samples using the Wilcoxon two-sample test). Here, we scale ∆ADVt−2,t and ∆ADVt,t+2 by Assett−2

and Assett, respectively. According to these three panels, the cross-sectional patterns of the mean and median

first-day return in our sample are very similar to those in Loughran and Ritter (2004), although we have a smaller

sample. In particular, we find that a firm on average underprices the equity in its IPO more in the internet

bubble period than in the 1990s (mean underpricing 64.2% versus 10.2%, median underpricing 34.2% versus

6.9%); tech IPOs are underpriced more than non-tech IPOs (mean underpricing 55.2% versus 14.3%, median

underpricing 34.5% versus 6.3%); and small firm IPOs (defined as firms with book value of assets smaller than

the average asset value of the sample) underpriced more than large firm IPOs (mean underpricing 46.2% versus

19.2%, median 23.6% versus 7.4%). Table 1 also shows that the increase in advertising in the IPO year over

the benchmark year t− 2, as well as the decrease in advertising from year t to year t+2, is significantly greater

31 Empirical studies of advertising in the industrial organization literature sometimes focus on advertising intensity, defined asadvertising expenditures scaled by sales revenue. However, this measure is inappropriate for our purpose. This is because, in oursetting, any given amount of advertising would affect the product and financial markets simultaneously. Therefore, if we normalizeadvertising expenditures using variables measuring the impact of advertising on the product market alone (e.g., sales revenue),we would not be able to correctly measure the impact of advertising on the financial market, thereby biasing our estimate of therelationship between advertising and IPO underpricing. Nevertheless, we still scale advertising expenditures by lagged assets (andlagged advertising expenditures in robustness checks), in order to control for the size effect and also to be consistent with ourmeasure of RET , which is scaled as well.

23

during the internet bubble period (1999-2000) compared to the early 1990s and for a tech firm compared to

a non-tech firm. The increase in advertising expenditures from the benchmark year to the IPO year is also

significantly greater for a small firm compared to a large firm. One implication of our model is that firms facing

a greater extent of asymmetric information advertise more in the product market when they are about to go

public; such firms also underprice equity more in their IPOs. Thus, the above empirical results are consistent

with our model, since, in general, small firms face a greater degree of asymmetric information than large firms,

and tech firms (with less tangible projects than non-tech firms) face a greater degree of asymmetric information

than non-tech firms.32

5.2 Empirical Tests and Results

5.2.1 Change in the Amount of Advertising from the Benchmark Year to the IPO Year

In this part, we test our first hypothesis (H1). The dependent variable in our regression is the change in

advertising expenditures from two years prior to the IPO year to the IPO year, ∆ADVt−2,t. We expect it to

be positive as suggested by H1. However, in order to ascertain whether the firm’s going public decision has an

impact on its advertising expenditures, we need to control for those product market factors that may affect the

firm’s advertising even in the absence of an IPO. For example, many firms choose to advertise heavily in order to

maintain or increase their sales. Thus, we use the sales revenue in the benchmark year, SALEt−2, and the change

of sales revenue from the benchmark year to the IPO year, ∆SALEt−2,t (calculated as SALEt−SALEt−2), as

control variables. Here, we measure SALEt and SALEt−2 as the sales revenue in year t and t− 2 respectively,

divided by the book value of total assets in each year. Also, many firms may have an established advertising

policy which tends to be consistent over time. In order to control for such an established advertising policy,

we also include the lagged advertising expenditures, ADVt−2, as a control variable. In addition, we include the

square of the lagged advertising expenditures, denoted by ADV 2t−2, to control for the nonlinear effects of lagged

advertising on ∆ADVt−2,t. Further, a firm’s advertising may be affected by its financial liquidity constraint as

well. We therefore measure the firm’s liquidity constraint by the lagged operating income, EBITt−2, calculated

as the earnings before interests, taxes, depreciation, and amortization two years prior to the IPO year scaled

32 In the setting of our model, we can measure the extent of asymmetric information facing a firm in the product market bythe difference between the true values of superior and poor quality products to consumers; similarly, the extent of asymmetricinformation facing a firm in the financial market can be measured by the difference in intrinsic values between good and badprojects. Then, it can be shown (as a comparative static result), that the amount of advertising undertaken by a firm and (or) theextent of underpricing are increasing in the extent of asymmetric information facing the firm in either the product market, or thefinancial market, or both.

24

by the book value of assets of the firm in that year. Finally, our model implies that asymmetric information

about quality in the product market will spill over to the financial market and affect the share price in the firm’s

IPO. Thus, a firm would increase its advertising expenditures over the benchmark year to a greater extent

if the extent of asymmetric information facing it is greater. We proxy the degree of asymmetric information

facing the firm using the lagged firm size, SIZEt−2, calculated as the log of the book value of total assets,

since smaller firms would face more severe asymmetric information in the product market, financial market, or

both markets. We also proxy the degree of asymmetric information by a tech dummy TECH, which is equal

to one if the firm going public is classified as a tech firm and zero otherwise.33 On the one hand, a firm

operating in a high technology industry may face more severe asymmetric information and thus has more of an

incentive to advertise; on the other hand, a high tech firm is likely to have less internal capital available and

may thus advertise less. Thus, the manner in which high tech firms would differ from non-high-tech firms in

advertising policy is ambiguous. Table 2 provides summary statistics and pairwise correlations for the above

control variables.

In summary, the linear regression to be estimated in testing H1 is as follows:34

∆ADVt−2,t = α0 + α1∆SALEt−2,t + α2SALEt−2 + α3ADVt−2 + α4ADV 2t−2

+α5SIZEt−2 + α6EBITt−2 + α7TECH + εt.

(10)

Note here that we choose not to control for any factors associated with the financial market alone, since our

objective is to test for the effect of raising capital in the financial market on the amount of the firm’s advertising

expenditures. In other words, we intend to control for those variables which are associated either with the

product market alone, or with both the product and the IPO markets, but not for variables related to the IPO

market alone. Thus, α0 can be viewed as the change of advertising expenditures which is not associated with

the product market factors but with the firm’s going public decision.35 We expect α0 to be positive.

33 The definition of a tech firm is based on the firm’s SIC code and follows that in Loughran and Ritter (2004): tech firms aredefined as those in SIC codes 3571, 3572, 3575, 3577, 3578, 3661, 3663, 3669, 3674, 3812, 3823, 3825 - 3827, 3829, 3841, 3845, 4812,4813, 4899, 7370 - 7375, 7378, and 7379.

34 We have conducted our tests using several alternative specifications. For example, excluding ADVt−2 or ADV 2t−2 (or both)does not change our qualitative results supporting hypothesis H1. Similarly, excluding SIZEt−2 and (or) SALEt−2 does not affectour qualitative results as well.

35 One can also view the first regression model as being derived in the following way. Consider two equations ADVt−2 = γ0+γ1Z,where Z stands for all the product market factors, and ADVt = γ00 + γ01Z. The first equation captures the relationship betweenadverting expenditures and product market factors when the firm has no plan to go public (the benchmark case discussed in section3.2), and the second equation captures a similar relationship but in a scenario where the firm plans to go public (as characterizedin propositions 1, 2, and 3). Thus, α0 = γ00 − γ0, which captures the change in advertising contributed by the firm’s going publicdecision only.

25

Column (1) of table 3 reports the results of the above regression. Overall, the regression has a good fit,

with adjusted R2 of 46.7%. As expected, we find that the intercept α0 is positive (equal to 0.12) and significant

at the 1% level using a two-tailed test. This result suggests that firms on average increase their advertising

expenditures in the IPO year by 12% of the firm’s book value of assets in the benchmark year in anticipation

of their upcoming IPOs.36 37 If one considers that the firm’s advertising in the benchmark year (when the

going public decision has no impact on advertising policy) accounts for 7% of the value of assets on average

(as shown in table 2), this increase in advertising due to the IPO is economically significant as well. Thus, this

result is consistent with hypothesis H1, i.e., a firm increases its advertising expenditures significantly when it is

about to go public. As a robustness check, column (2) of table 3 also presents results from the regression based

on the percentage change in advertising expenditures. Again, the results in column (2) support our hypothesis

H1, suggesting a significant increase in advertising expenditures around the IPO.

In addition, we find that the change in sales revenue has a positive and significant impact (at the 1% level)

on the increase in advertising expenditures, as can be expected from product market considerations. We also

find that smaller firms tend to increase advertising expenditures significantly more than larger firms in the year

of their IPO. This difference is significant at the 1% level. This result is consistent with the predictions of our

model as well, suggesting that the increase in advertising around the IPO is more for firms suffering from more

severe asymmetric information in the product and financial markets.

5.2.2 IPO Underpricing and Advertising As Substitutes

In this subsection, we test our second hypothesis (H2). We will test this hypothesis by running regressions

with advertising and IPO underpricing as dependent variables separately. In the first regression, we regress

IPO underpricing, RET , on the amount of advertising in the IPO year, ADVt, controlling for those variables

36 Note that we normalize the change of advertising by the book value of assets in the benchmark year. Thus, due to the differencebetween the value of assets in the IPO year and that in the benchmark year, 10% of the book value of assets in the benchmarkyear is equivalent to 2.36% of the book value of assets in the IPO year.

37 This increase in product market advertising associated with the IPO may have two components. The first component may beassociated with signaling to the IPO market as discussed in the theoretical section of this paper. In some cases, there may alsobe a second component arising from a regime-shift in the firm’s advertising behavior post-IPO. Such a regime shift can arise, forinstance, from the firm dramatically accelerating its investment in new products (and therefore future sales) much above the rateof annual increase in these variables pre-IPO. While we have controlled for anticipated increases in advertising driven by productmarket considerations from the benchmark year to the IPO year assuming that the rate of annual increase in sales (and thereforeadvertising) after the IPO would be consistent with that pre-IPO, we are unable to rule out the possibility that our results aredriven partially by this second component. However, given that we have demonstrated a decline in the product market advertisinglevels two years after the IPO relative to the IPO year (even when sales are increasing in this two year time period), it is unlikelythat our results are driven by such product market effects.

26

that will directly affect IPO underpricing.38 If advertising and IPO underpricing serve as substitutes for the

purpose of signaling to the IPO market, the coefficient on ADVt is expected to be negative.

In order to find an unbiased estimate, we control for other determinants of underpricing (initial return) in

our regression. Many theories of IPO underpricing predict that the initial return is higher when either the extent

of asymmetric information or the extent of ex-ante uncertainty about an issuer’s value is greater (see, e.g., Allen

and Faulhaber (1989), or Chemmanur (1993)). We first control for this using the tech dummy, TECH, and the

size of the issuer prior to its IPO, SIZEt−1, since smaller firms or those operating in high technology industries

would face more severe asymmetric information. Carter and Manaster (1990) and Megginson and Weiss (1991)

note that more information is available to the market about older firms compared to younger firms, and they

find a negative relationship between initial returns and age. We follow Lowry and Shu (2002) and construct an

age dummy, AGE, which is equal to one if a firm reports earnings data in its prospectus (according to SDC)

five or more years prior to its IPO, and zero otherwise. Chemmanur and Fulghieri (1994) demonstrate that a

higher-ranked underwriter has an incentive to more accurately certify the quality of the issuing firm so as to

protect its reputation, leading investors to demand a smaller IPO discount. We proxy the underwriter rank,

RANK, following the measure used by Loughran and Ritter (2004) (see also Carter, Dark, and Singh (1998)),

with higher ranks representing higher quality. We also control for the syndication dummy, SY ND, which is

equal to one if the IPO underwriting is syndicated and zero otherwise, since IPOs underwritten by a syndicate

may be underpriced less. Similarly, the IPO literature has documented that venture backed firms differ from

non-venture backed firms in terms of underpricing (see, e.g., Megginson and Weiss (1991)). Thus, we include

in the regression a dummy variable, V ENTURE, which is equal to one if the firm making the IPO was backed

by venture capitalists, and zero otherwise.

Further, Loughran and Ritter (2004) find that conditions in the stock market as measured by prior returns

on the relevant market index affect initial IPO returns, though Lowry and Schwert (2002) find that this effect

is not significant. In our regression, we include the compounded return of an equally weighted market index 15

days prior to the IPO, RUNUP , to control for the conditions in the stock market. Moreover, other variables

can affect IPO underpricing as well. For example, the proceeds of an IPO may be related to its initial return in

38 We choose to regress RET on the level of advertising, instead of the change in advertising expenses, since our model predictsthat the levels of advertising and underpricing serve as signalling substitutes. In our model, every dollar of advertising expenses canserve as a signal to both the product and financial markets. Thus, the whole amount of advertising expenses should be consideredwhen we consider the signaling role of advertising.

27

several ways. First, information tends to be more readily available about larger firms which tend to have larger

proceeds. Thus, the information asymmetry hypothesis predicts that IPO proceeds will be negatively related

to initial returns. However, Michaely and Shaw (1994) argue that larger issues can be harder to sell, thereby

motivating underwriters to underprice them by a larger amount. We control for IPO proceeds in our regression

by AMOUNT , calculated as the log of the amount of money raised from the IPO. Also, since we observe higher

first-day returns during the internet bubble period (year 1999 and 2000) compared to other time periods, we

also follow Loughran and Ritter (2004) in controlling for this using a dummy variable Y EAR which is equal

to one if the year of IPO is between 1990 to 1998, and zero if it is between 1999 to 2000. Finally, we control

for lagged advertising expenditures ADVt−1 since the amount of advertising one year prior to the IPO could be

influenced by the firm’s going public decision. Table 2 provides descriptive statistics for all the above control

variables.

In summary, the first regression equation we estimate to test H2 is as follows:

RET = β0 + β1ADVt + β2ADVt−1 + β3TECH + β4SIZEt−1 + β5AGE + β6SY ND

+β7RANK + β8V ENTURE + β9RUNUP + β10AMOUNT + β11Y EAR+ εt.

(11)

Here, we choose not to control for those product market factors that affect IPO underpricing indirectly through

advertising. Including those variables does not add any explanatory power to the estimation, while introducing

the potential for multicollinearity with the amount of advertising.39

In the second regression, we regress advertising expenditures (ADVt) on IPO underpricing (RET ), controlling

for those factors that may affect the firm’s advertising decision but are not associated with the firm’s going public

decision. Thus, the control variables we use here are related to the firm’s product market only, similar to the

control variables we use in the test of hypothesis H1. In particular, we choose the change of sales revenue from

one year prior to IPO, ∆SALEt−1,t (calculated as the change in sales revenue between year t−1 and t, scaled by

the book value of assets in year t−1), lagged sales, SALEt−1, and lagged operating income, EBITt−1, as control

variables. We also include in the regression the lagged advertising expenditures ADVt−1 and ADVt−2 and the

square of one-year lagged advertising expenditures ADV 2t−1, to control for any autocorrelation of advertising

39 Note that including product market factors such as the sales variables as explanatory variables will not change our qualitativeresults. If we include these variables, the p-value of β1 would increase, but still stay in the meaningful range (i.e. less than10%), while the adjusted R2 reduces from 25.4% to 24.6%. Our results are also robust to alternative regression specifications.For example, realizing that the tech dummy TECH is correlated with ADVt (see table 2) and can thus potentially cause multi-collinearity problems, we have tried specifications excluding this variable. The qualitative features of the regression result on β1does not change under this new specification. Similarly, β1 will remain negative and significant even if we exclude the size variableSIZEt−1.

28

expenditures in the time series.40 We also control for firm-specific characteristics, such as SIZEt−1, TECH,

and AGE. Table 2 provides the pair-wise correlation between all the control variables. The following is the

second regression equation we estimate in the test of H2:

ADVt = β00 + β01RET + β02ADVt−1 + β03ADV 2t−1 + β04ADVt−2 + β05SALEt−1

+β06∆SALEt−1,t + β07EBITt−1 + β08SIZEt−1 + β09TECH + β010AGE + ε0t.(12)

Note here that we choose not to control for those factors associated only with the firm’s IPO since their

explanatory power has already been incorporated into RET , and adding them would only cause multicollinearity

with RET . Based on hypothesis H2, we expect β1 in regression (11) and β01 in regression (12) to be negative.

In the estimation of the above two models, we use the ordinary least square (OLS) approach. In the next

section, we will use other approaches to study the robustness of our OLS regression results. The results of

OLS regressions are reported in panels A and B respectively under column (1) of table 4. The overall fit of the

two regressions are quite good, in view of the adjusted R2 of 25.4% for the regression on RET and 51.4% for

the regression on ADVt. As conjectured, we find that the coefficient of ADVt in the regression on RET , i.e.,

β1, is negative (-3.906, as shown in panel A) and significant at the 5% level. Further, the coefficient of RET

in the regression on ADVt, i.e., β01, is negative (-0.005, as shown in panel B) and significant at the 5% level.

Both results are consistent with hypothesis H2, suggesting that a firm uses advertising and IPO underpricing

as substitutes to signal to the product and financial markets.

Note that from the regression on RET (as shown in panel A of table 4), the coefficient of SIZEt−1 is

significantly negative at the 1% level (with coefficient equal to -0.115), suggesting that smaller firms underprice

more than larger firms. This finding is consistent with our model, since smaller firms are likely to face more

severe information asymmetry in both the financial and the product markets. Also in the same regression, the

coefficient of AMOUNT is significantly positive at the 1% level (with the coefficient equal to 0.234). Finally,

our result on the coefficient of Y EAR suggests that, holding other things constant, the initial return of IPOs in

1999 and 2000 was significantly greater than that in the period between 1990 to 1998 by an average of 22.3%.

40 We also ran an alternative regression adding the square of ADVt−2 as an independent variable. The coefficient of this variablein the regression is statistically insignificant.

29

5.2.3 Robustness of Results

In this section, we demonstrate the robustness of our OLS regression results. First, an endogeneity problem

may exist in OLS estimations, since both IPO underpricing and advertising are endogenous variables in our

framework. In particular, if ADVt and εt in regression (11) or RET and ε0t in (12) are correlated, then the

relation between IPO underpricing and advertising estimated using the OLS regression could be biased. Second,

given the correlation between ADVt and εt (and the correlation between RET and ε0t), equations (11) and (12)

are connected. In this case, estimating each equation separately such as in an OLS regression may ignore any

simultaneity between the two equations and waste the information in the other equation, and may be inefficient

compared to a joint estimation. We therefore present results from three simultaneous equation techniques to

demonstrate the robustness of our results.41 First, we use a two-stage least square (2SLS) approach to account

for any endogeneity problems. Second, we use a generalized least square (GLS) approach to account for any

simultaneity problems, ignoring the endogeneity problem. Finally, we use a three-stage least square (3SLS)

approach to account for both endogeneity and simultaneity problems at the same time.42 43

We implement the 2SLS estimation as follows. In the first stage, we regress the endogenous variables, namely,

RET and ADVt, on a set of instrumental variables (IVs). We choose the instrumental variables keeping in mind

that they should be highly correlated with the corresponding endogenous variable, but uncorrelated with the

corresponding disturbances εt or ε0t (or in other words, the corresponding dependent variables). In particular,

we choose Y EAR, AMOUNT , SY ND, RANK, and V ENTURE as the instrumental variables for RET , since

these variables, especially the year dummy Y EAR, are exogenous to the firm’s decision about product market

41 It is worth emphasizing that our objective in presenting results from these simultaneous equation estimation techniques is notto identify any dynamic causal relation between advertising and underpricing: our hypothesis H2 only predicts that the two shouldbe negatively correlated in equilibrium.

42 It is well known that asymptotically, 2SLS dominates OLS, and full-information estimators such as 3SLS and GLS are moreefficient than limited-information estimators such as 2SLS and OLS. However, the small-sample properties of these estimators arenot clear. In a small sample, 2SLS is sensitive to outliers and the choice of instrumental variables, thus the net benefit of using2SLS is ambiguous. Also, the benefit of using systems methods (GLS and 3SLS) rather than single-equation methods is modestin a small sample, since the finite-sample variation of disturbances or specification errors could be transmitted between the twoequations in a joint estimation and thus reduce the efficiency of estimation. Given that each of the above estimation techniqueshas its own individual limitations, we choose to provide results using all three methods to demonstrate the robustness of our OLSresults.

43 To implement the above estimation techniques, we also need to establish that our simultaneous equations are properly identified.This is clearly the case. The endogenous variables in the equations are RET and ADVt. The common independent variables for bothequations are ADVt−1, SIZEt−1, TECH, and AGE, while the independent variables SY ND, RANK, V ENTURE, RUNUP ,AMOUNT , and Y EAR appear in the return equation only and ADV 2t−1, ADVt−2, SALEt−1, ∆SALEt−1,t, and EBITt−1appear in the advertising equation only. Thus, if we convert our model of simultaneous equations to a reduced-form model, thereduced form parameters are more than enough to solve the structural parameters, which means that our model is identified.

30

advertising in the absence of an IPO, but related to the firm’s going public decision.44 45 On the other hand, for

ADVt, we choose as instrumental variables those variables which are most likely to be exogenous to the firm’s

going public decision but related to the firm’s choice of advertising in the absence of an IPO. Thus, lagged

values of those variables related directly to the firm’s product market provide natural candidates to be used

as instrumental variables. We therefore choose ∆SALEt−1,t, SALEt−1, EBITt−1, ADVt−1, ADV 2t−1, and

ADVt−2 as the instrumental variables for ADVt. These instrumental variables are properly constructed, since

advertising two years prior to the IPO year ADVt−2 is unlikely to be affected by the firm’s going public decision

in the IPO year, and thus is exogenous to the initial return.46 47 Then, after regressing on the instrumental

variables, we use the fitted values of the first-stage regressions as the instrumental variables for RET and ADVt

in the second stage. For brevity, we report only the results from the second stage of the estimation in table

4 under column (2). Similar to the results reported in the OLS regressions, both β1 and β01 are negative and

significant at the 10% and 5% levels, respectively. These results are also consistent with our hypothesis H2.

In the GLS estimation, we do not need to construct any instrumental variables. We rely on the covariance

matrix estimated from the OLS results, and estimate (11) and (12) simultaneously. In the 3SLS estimation,

we use the same instrumental variables as those in the 2SLS estimation. Our 3SLS estimator is based on the

covariance matrix estimated from the 2SLS results and we estimate (11) and (12) simultaneously. The results

from the GLS and the 3SLS estimations are presented in columns (3) and (4) of table 4, respectively. As expected,

our results (i.e., significantly negative β1 and β01) support hypothesis H2, suggesting the substitutability between

advertising and underpricing.

6 Conclusion

Practitioners have noted that firms tend to increase their product market advertising prior to an IPO or

a seasoned equity issue. Further, recent empirical evidence by Grullon, Kanatas, and Weston (2004) indicates

that firms with a greater level of product market advertising have lower bid-ask spreads and a larger number

44 According to table 2, Y EAR is highly correlated with RET , but uncorrelated with ADVt.

45 The regression of RET in the first stage is well-fitted with adjusted R2 equal to 24%. Coefficients of Y EAR, AMOUNT , andSIZEt−1 are significantly different from zero at the 1% level.

46 According to table 2, ADVt−2 is highly correlated with ADVt, but uncorrelated with RET .

47 Our regression of ADVt in the first stage is well fitted with adjusted R2 equal to 48.8%. The coefficient of ADVt−2 issignificantly different from zero at the 1% level.

31

of both individual and institutional investors in their equity. We develop a theoretical model of the interaction

between a firm’s product market advertising and its corporate financing decisions in the above context. We

consider a firm which faces asymmetric information in both the product and the financial market (about the

quality of its products and the intrinsic value of its projects) and which needs to raise external financing to

fund its growth opportunity (new project). Any product market advertising undertaken by the firm is visible

in the financial market as well. We show that, in equilibrium, the firm uses a combination of product market

advertising, IPO underpricing, and underfinancing (raising a smaller amount of external capital than the full

information optimum) to convey its true product quality and the intrinsic value of its projects to consumers and

investors. Our model has several implications for IPO underpricing and product market advertising. Two of

these predictions are as follows. First, firms will choose a higher level of product market advertising when they

are planning to issue new equity or other information-sensitive securities, compared to situations where they

have no immediate plans to sell such securities. Second, product market advertising and IPO underpricing are

substitutes for a firm going public. The empirical evidence supports these two predictions: First, firms indeed

increase their product market advertising in their IPO year relative to a benchmark year two years before their

IPO. Further, we find that, in the five year span around the IPO year (i.e., the IPO year, and the two years

before and after the IPO year), the peak advertising level is reached in the IPO year. Second, the extent of

underpricing is smaller as the level of product market advertising is greater.

32

References

Allen, F., and Faulhaber, G., 1989, “Signaling by Underpricing in the IPO Market.” Journal of FinancialEconomics 23, 303-323.

Bhattacharya, S., and Ritter, J. R., 1983, “Innovation and Communication: Signaling with Partial Dis-closure.” Review of Economic Studies 50, 331-346.

Carter, R., Dark, F., and Singh, A., 1998, “Underwriter Reputation, Initial Returns, and the Long-runUnderperformance of IPO Stocks.” Journal of Finance, 53, 285-311.

Carter, R., and Manaster, S., 1990, “Initial Public Offerings and Underwriter Reputation.” Journal ofFinance 45, 1045-1067.

Chemmanur, T., 1993, “The Pricing of Initial Public Offerings: A Dynamic Model with InformationProduction.” Journal of Finance 48, 285-304.

Chemmanur, T., and Fulghieri, P., 1999, “A Theory of the Going-public Decision.” Review of FinancialStudies, 12, 249-279.

Chemmanur, T., and Fulghieri, P., 1994, “Investment Bank Reputation, Information Production, andFinancial Intermediation.” Journal of Finance, 49, 57-79.

Cho, I. K., and Kreps, D., 1987, “Signaling Games and Stable Equilibria.” Quarterly Journal of Economics102, 179-222.

Demers, E., and Lewellen, K., 2003, “The Marketing Role of IPOs: Evidence from Internet Stocks.”Journal of Financial Economics, 68, 413-437.

Engers, M., 1987, “Signaling with Many Signals.” Econometrica 55, 663-674.

Fudenberg, D., and Tirole, J., 1991, “Perfect Bayesian Equilibrium and Sequential Equilibrium,” Journalof Economic Theory 53, 236-60.

Gertner, R., Gibbons, R., and Scharfstein, D., 1988, “Simultaneous Signaling to the Capital and ProductMarkets.” Rand Journal of Economics 19, 173-190.

Grullon, G., Kanatas, G., and Weston, J. P., 2004, “Advertising, Breadth of Ownership, and Liquidity.”Review of Financial Studies 17, 439-461.

Kihlstrom, R. E., and Riordan, M. H., 1984, “Advertising as A Signal.” Journal of Political Economy 92,427-451.

Loughran, T., and Ritter, J. R., 2004, “Why Has IPO Underpricing Changed Over Time?” FinancialManagement 33, 5-37.

Lowery, M., and Schwert, G. W., 2002, “IPO Market Cycles: Bubbles or Sequential Learning.” Journalof Finance, 62, 1171-1200.

Lowry, M., and Shu, S., 2002, “Litigation Risk and IPO Underpricing.” Journal of Financial Economics,65, 309-335 .

Maksimovic, V., and Pichler, P., 2001, “Technological Innovation and Initial Public Offerings.” Review ofFinancial Studies, 14, 459-494.

Michaely, R., and Shaw, W., 1994, “The Pricing of Initial Public Offerings: Tests of Adverse-selection andSignaling Theories.” Review of Financial Studies, 7, 279-319.

Megginson, W., and Weiss, K., 1991, “Venture Capitalist Certification in Initial Public Offerings.” Journalof Finance, 46, 879-904.

Milgrom, P., and Roberts, J., 1986, “Price and Advertising Signals of Product Quality.” Journal of PoliticalEconomy 94, 796-821.

33

Nelson, P., 1974, “Advertising as Information.” Journal of Political Economy 81, 729-754.

Ritter, J. R., and Welch, I., 2002, “A Review of IPO Activity, Pricing, and Allocations.” Journal ofFinance, 57, 1795-1828.

Stoughton, N. M., Wong, K. P. and Zechner, J., 2001, “IPOs and Product Quality.” Journal of Business,74, 375-408.

Welch, I., 1989, ”Seasoned Offerings, Imitation Costs, and the Underpricing of Initial Public Offerings.”The Journal of Finance, 44, 421-449.

34

Appendix

Proof of Proposition 1 to 3. i). For a type L firm, F ∗L = VL; C∗L = i2; A

∗L = 0, and therefore

Π∗L = 2RPQ+ b(W + i2)− i2.

ii). Define dk =1Fk. Thus, the type M’s optimization problem can be rewritten as:

MaxAM , dM , CM

ΠM = VM (1− dMCM ) ,

s.t. Π∗L ≥ ΠL(M) = VL(M) (1− dMCM ) , (A1)

where VM = 2RPQ + G(I0M ) + G(I

1M ), and VL(M) = 2RPQ + B(I

0M ) + B(I

1M ). Given our assumption that b

and g are large enough that the firm is always willing to go public, any type of firm invests all the available

internal wealth W in the existing project and the capital raised from the capital market in the new project.

Thus, I0M =W −AM and I1M = CM .

It is easy to show that in the first best case (the case without asymmetric information), the type M would

choose dk = 1/VM , AM = 0, and CM = i2. However, the first-best choices of the type M do not satisfy the IC

constraint (A1), since, in this case, RHS of (A1) = [2RPQ+ b(W + i2)](1− i22RPQ+G(W )+G(i2)

] > Π∗L. Therefore,

the type M’s problem can be viewed as satisfying constraint (A1) as an equation with a minimal cost incurred

by the type M (defined as the difference between his payoff in the first best and his optimal payoff under the

IC constraint (A1)). In other words, the type M chooses to change the choice variables (i.e., AM , dM , or CM )

from their first-best levels so that, to achieve the same reduction of ΠL(M) from the level in the first-best to

Π∗L, a minimal reduction of ΠM from the first-best level is incurred. Mathematically, the type M chooses to

change the choice variable which corresponds toMin(¯̄̄

∂ΠM/∂AM∂ΠL(M)/∂AM

¯̄̄,¯̄̄

∂ΠM/∂CM∂ΠL(M)/∂CM

¯̄̄,¯̄̄

∂ΠM/∂dM∂ΠL(M)/∂dM

¯̄̄). Solving for

these derivatives, we have:¯̄̄

∂ΠM/∂dM∂ΠL(M)/∂dM

¯̄̄= VM

VL(M);¯̄̄

∂ΠM/∂AM∂ΠL(M)/∂AM

¯̄̄W−AM≥i1

= 1;¯̄̄

∂ΠM/∂AM∂ΠL(M)/∂AM

¯̄̄W−AM<i1

= gb ;¯̄̄

∂ΠM/∂CM∂ΠL(M)/∂CM

¯̄̄CM≥i2

= (1−dMCM )−VMdM(1−dMCM )−VL(M)dM

;¯̄̄

∂ΠM/∂CM∂ΠL(M)/∂CM

¯̄̄i1≤CM<i2

= b(1−dMCM )−VMdMb(1−dMCM )−VL(M)dM

; and¯̄̄

∂ΠM/∂CM∂ΠL(M)/∂CM

¯̄̄CM<i1

=

g(1−dMCM )−VMdMb(1−dMCM )−VL(M)dM

.

We can show¯̄̄

∂ΠM/∂CM∂ΠL(M)/∂CM

¯̄̄i1≤CM<i2

<¯̄̄

∂ΠM/∂AM∂ΠL(M)/∂AM

¯̄̄W−AM≥i1

<¯̄̄

∂ΠM/∂dM∂ΠL(M)/∂dM

¯̄̄<¯̄̄

∂ΠM/∂AM∂ΠL(M)/∂AM

¯̄̄W−AM<i1

<¯̄̄∂ΠM/∂CM

∂ΠL(M)/∂CM

¯̄̄CM<i1

and¯̄̄

∂ΠM/∂dM∂ΠL(M)/∂dM

¯̄̄<¯̄̄

∂ΠM/∂CM∂ΠL(M)/∂CM

¯̄̄CM≥i2

, since 1 < VMVL(M)

=2RPQ+G(I

0M )+G(I

1M )

2RPQ+B(I0M )+B(I1M)< g

b .

Thus, given the objective of the type M to choose the cheapest signal combination, the type M prefers to first

reduce CM from its first-best level (as long as CM ≥ i1), then advertise if necessary (as long as W −AM ≥ i1),

2

and finally underprice its equity. Define a1 = argAM

[Π∗L = (2RPQ+b(W−AM+i1))³1− i1

2RPQ+G(W−AM )+G(i1)´]

and F1 = argFM

{Π∗L = [2RPQ+ bMin(2i1,W + i1)]³1− i1

FM

´}. The equilibrium choices for the type M are:

Case 1: if W ≥ i1 + a1, f∗M = 1; C∗M = i1; and A∗M = a1.

Case 2: if W < i1 + a1, f∗M = f1 ≡ F1/VM , C∗M = i1; and A

∗M =Max (0,W − i1).

iii). Now the type H’s IC constraints become to:

Π∗L ≥ VL(H) (1− dHCH) , and (A2)

Π∗M ≥ VM(H) (1− dHCH) . (A3)

We first compare¯̄̄

∂ΠH/∂AH∂ΠL(H)/∂AH

¯̄̄,¯̄̄

∂ΠH/∂CH∂ΠL(H)/∂CH

¯̄̄, and

¯̄̄∂ΠH/∂dH

∂ΠL(H)/∂dH

¯̄̄. It can be shown that

¯̄̄∂ΠH/∂dH

∂ΠL(H)/∂dH

¯̄̄= VH

VL(H);¯̄̄

∂ΠH/∂AH∂ΠL(H)/∂AH

¯̄̄W−AH<i1

= gb ;¯̄̄

∂ΠH/∂AH∂ΠL(H)/∂AH

¯̄̄W−AH≥i1

= 1;¯̄̄

∂ΠH/∂CH∂ΠL(H)/∂CH

¯̄̄CH≥i2

= (1−dHCH)−VHdH(1−dHCH)−VL(H)dH

;¯̄̄∂ΠH/∂CH

∂ΠL(H)/∂CH

¯̄̄i1≤CH<i2

= b(1−dHCH)−VHdHb(1−dHCH)−VL(H)dH

; and¯̄̄

∂ΠH/∂CH∂ΠL(H)/∂CH

¯̄̄CH≤i1

= g(1−dHCH)−VHdHb(1−dHCH)−VL(H)dH

.

Consider the case when gb >

RSRP. Then g

b >VHVL(H)

= 2RSQ+G(W−AH)+G(CH)[(2−α)RS+αRP ]Q+B(W−AH)+B(CH) . This implies that¯̄̄

∂ΠH/∂CH∂ΠL(H)/∂CH

¯̄̄i1≤CH≤i2

<¯̄̄

∂ΠH/∂AH∂ΠL(H)/∂AH

¯̄̄W−AH≥i1

<¯̄̄

∂ΠH/∂dH∂ΠL(H)/∂dH

¯̄̄<¯̄̄

∂ΠH/∂AH∂ΠL(H)/∂AH

¯̄̄W−AH<i1

<¯̄̄

∂ΠH/∂CH∂ΠL(H)/∂CH

¯̄̄CH≤i1

,

and¯̄̄

∂ΠH/∂dH∂ΠL(H)/∂dH

¯̄̄<¯̄̄

∂ΠH/∂CH∂ΠL(H)/∂CH

¯̄̄CH>i2

. Therefore, the type H prefers to signal its type by first reducing CH

(as long as CH ≥ i1), then advertising if necessary (as long as W − AH ≥ i1), and finally underpricing its

equity. Define a2 = argAH

{Π∗L = [(2− α)RSQ+αRPQ+ b(W −AH + i1)][1− i12RSQ+G(W−AH)+G(i1) ]}, and F2 =

argFH

{Π∗L = [(2− α)RSQ+ αRPQ+ bMin(2i1, i1 +W )]³1− i1

FH

´}. Then, the equilibrium choices for the type

H are:

Case 1: if W ≥ i1 + a2, f∗H = 1; C∗H = i1; and A∗H = a2.

Case 2: if W < i1 + a2, f∗H = f2 ≡ F2/VH , C∗H = i1; and A∗H =Max (0,W − i1).

Now, we want to show that, given the above solutions, the IC constraint (A3) is satisfied as well. In

particular, the RHS of (A3) =VM(H)

VL(H)Π∗L =

[(2−α)RS+αRP ]Q+G(W−A∗H)+gi1[(2−α)RS+αRP ]Q+B(W−A∗H)+bi1Π

∗L, and the LHS of (A3) =

V ∗MVL(M)

Π∗L =

2RPQ+G(W−A∗M )+gi12RPQ+B(W−A∗M )+bi1Π

∗L. Since

VM(H)

VL(H)≤ VM

VL(M), constraint (2) is satisfied.

Thus, when W ≥ i1 + a2, we have proposition 1; when i1 + a1 ≤ W < i1 + a2, we have proposition 2; and

when W < i1 + a1, we have proposition 3.

3