corporate strategy, conformism, and the stock market"
Post on 04-Feb-2017
212 Views
Preview:
TRANSCRIPT
Corporate Strategy, Conformism, and the StockMarket�
Thierry FoucaultHEC Paris
Laurent FrésardUniversity of Maryland
April 2015
Abstract
In choosing their business strategy, managers must account for the e¤ectof their choice on the value of the real options associated with each strategy.When managers rely on stock market information to exercise or not these op-tions, their value is higher when stock prices are more informative. We showtheoretically that this induces conformity in strategic choices because the stockmarket is more informative about the value of common strategies. This con-formity e¤ect is stronger for private �rms because investors have no incentiveto produce information about their strategies. Consistent with this prediction,we show empirically that �rms choose more di¤erentiated products after goingpublic and more so when their managers are better privately informed or whentheir peers�stock prices are less informative.
�We are grateful to Maria Cecilia Bustamante, Julien Cujean, Claudia Custodio, Miguel Ferreira,and seminar participants at the Nova School of Business for useful comments. We thank Jerry Hobergand Gordon Phillips for sharing their TNIC data, and Jay Ritter for sharing its IPO data. All errorsare ours.
1
1 Introduction
A central tenet of strategic management is that �rms should choose corporate strate-
gies �the business in which they operate �that give them the strongest competitive
advantage (e.g. Porter (1985)). This competitive advantage can be obtained by
unique combinations of resources (e.g. through acquisition of new assets or restruc-
turing), di¤erentiation of product o¤ering, or lower costs (e.g. through innovations or
the adoption of new technologies). According to this view, managers create value by
choosing unique strategies, i.e., strategies that are di¤erent from those of other �rms,
and that others cannot easily replicate (e.g. Barney (1986) or Hoberg and Phillips
(2014)).
In choosing their strategy, managers often face considerable uncertainty about
their payo¤s, however. One way for managers to reduce this uncertainty is to rely on
stock market prices as a source of information. In this paper, we show theoretically
that the reliance on the stock market can signi�cantly reduce managers�incentives to
choose unique strategies and we provide evidence thereof. That is, the possibility of
learning information from stock prices is a source of conformism in strategic choices,
even when managers choose strategies that maximize the value of their �rm.
We formally derive this result in a model in which stock price informativeness
and managers�strategic decisions are jointly determined. In our model, a manager
can choose to follow either a common strategy, already followed by several established
public �rms, or a unique strategy. The net present value of each strategy is uncertain.
Their payo¤depends on a future state speci�c to each strategy that is unknown when
the manager chooses his strategy (e.g., the strength of consumers�demand). Ex post,
once the state is known, the type of the chosen strategy can be either good (if it yields
a payo¤higher than expected) or bad (if it yields a smaller payo¤than expected). The
net present value of a good unique strategy is higher than that of a good common
strategy because uniqueness generates a higher payo¤. However, a bad unique or
common strategy has a negative net present value. After announcing his strategy,
the manager receives additional information about the type of his strategy and has
2
the option to abandon it. Thus, when the manager chooses his strategy, he must
incorporate the impact of his choice on the information he will obtain about its type
since this information in�uences the value of his abandonment option.
As a benchmark, we �rst consider the case in which the manager only relies on his
own private information in exercising the option to pursue or to abandon his chosen
strategy. In this case, he always chooses the unique strategy because the expected
net present value of this strategy (including the value of the option to abandon the
strategy) is higher. The reason is simply that di¤erentiation yields a higher payo¤
on average, in particular when the unique strategy is good. Then, we consider the
possibility for the manager to learn information from its own stock price (if his �rm
is publicly listed) and established �rms�stock prices, in addition to his own private
information. In this case, the manager sometimes �nds optimal to choose the common
strategy.
The reason is that the value of the abandonment option is higher for the common
strategy because the stock market is more informative. This superior informativeness
stems from two complementary economic forces. First, speculators who are informed
about the type of the common strategy can exploit this information by trading the
stocks of all �rms following this strategy. For instance they buy the stocks of all
these �rms if they learn that the common strategy is good. As a result, intermediaries
setting prices (i.e., market makers) are more likely to discover the type of the common
strategy than that of the unique strategy. Therefore, prices contain more information
about whether the strategy is good or bad. Second, there are economies of scale
in trading on information about the common strategy: The �xed cost of obtaining
information about the common strategy can be amortized by trading all stocks of �rms
following it. Thus, the information cost per stock is naturally lower for the common
strategy. So, when the demand for information about a strategy is endogenous,
speculators trading the common strategy must obtain a smaller expected pro�t per
stock in equilibrium. This implies that stock prices are more information about the
type of the common strategy.
3
When the manager relies in part on stock market prices as a source of information,
he faces a trade-o¤ when choosing his strategy. Ignoring the abandonment option,
the unique strategy has a higher expected net present value. The value of the aban-
donment option is however smaller for the unique strategy because the stock market
is (endogenously) less informative about its payo¤. Given this trade-o¤, there is a
range of values for the parameters such that the manager is better o¤ choosing the
common strategy while he would not in the absence of stock market information. The
reliance on the stock market can thus push the manager�s decision toward strategic
conformity. This �conformity e¤ect�becomes stronger when the bene�t of obtaining
information from the stock market is higher for the manager. This happens when
(a) his private information is less precise, (a) the number of investors producing in-
formation about the common strategy increases, or (c) the number of investors who
produce information about the unique strategy decreases.
While the model o¤ers several new insights, one unique testable implication of
our theory is that the conformity e¤ect is stronger for a private �rm than for a public
�rm, other things equal. By construction, a private �rm cannot rely on its own
stock price as a source of information. This lowers the value of the abandonment
option associated with the unique strategy compared to the case where the �rm is
public. Thus, the model predicts that �rms that go public should be relatively more
di¤erentiated after their Initial Public O¤ering (IPO) compared to prior their IPO.1
We test this novel prediction using a sample of 1,231 U.S. �rms that go public
between 1996 and 2011. We measure the uniqueness of the strategy chosen by a
�rm by the degree of di¤erentiation of its product o¤ering relative to that of related
(peer) �rms. We identify the set of peers for each newly public �rm at the time of its
1Some papers (e.g., Maksimovic and Pichler (2001), Spiegel and Tookes (2009), or Chod andLyandres (2011)) analyze the e¤ects of IPOs on competitive interactions in the product market.However, this literature has not considered a possible direct e¤ect of the going-public decision ondi¤erentiation choices, as we propose in this paper. For instance, Chod and Lyandres (2011) showsthat newly-public �rms compete more aggressively with their rivals after going public because theirowners can better diversify idiosyncratic risks in capital markets. However, their analysis and testsassume that industry de�nition �and the extent of di¤erentiation among �rms �is �xed before andafter the IPOs.
4
IPO using Hoberg and Philipps (2015)�s Text-Based network Classi�cation (TNIC).
This classi�cation is based on textual analysis of the product description sections in
�rms�10-Ks. For every pair of �rms, Hoberg and Phillips (2015) de�ne an index
of product similarity based on the relative number of words that �rm-pairs share in
their product description. We use (one minus) this index to measure the extent of
product di¤erentiation between each �rm-pairs.2
For each going-public �rm, we then measure the change in its di¤erentiation vis-à-
vis each of its established peers, measured at the time of the IPO and de�ned as �rms
that have been listed for more than �ve years. We track the change in di¤erentiation
within each pair over the �ve years following the IPO. To better isolate e¤ects that are
due to the IPO and not re�ecting general trend in di¤erentiation or peers�decisions
to di¤erentiate, we construct counterfactual �rm-pairs that are made of established
peers of peers of the IPO �rm i that are not peers of �rm i.
Consistent with the model�s prediction, we �nd that going-public �rms become
signi�cantly more di¤erentiated in the years that follow their initial public listing.
In particular, the average degree of product di¤erentiation between a newly-public
�rm and an established peer increases signi�cantly more over time compared to that
observed for counterfactual pairs. Notably, this result is obtained using regressions
that include �rm-pair �xed-e¤ects that capture any time-invariant di¤erences within
pairs (e.g. age di¤erences or geographical location) and time-varying control variables
that could a¤ect the evolution of �rms�di¤erentiation choices over time (e.g. size,
growth opportunities, or access to capital). In a similar vein, we �nd a signi�cant
decrease in the return co-movement between IPO �rms and their established peers
throughout the post-IPO period, which con�rms that the fundamentals of IPO �rms
become less similar than that of their peers after they become publicly traded.
Our model further suggests that the weakening of the conformity e¤ect following
an IPO should be larger for newly-public �rms for which the informational cost of
2For instance the peers of �rm i at a given point in time are �rms for which the index of productsimilarity exceeds a pre-de�ned threshold. A decrease in this index of similarity for a �rm i relativeto one of its peers j indicates that the degree of di¤erentiation increases between these two �rms.
5
di¤erentiation is smaller, that is, when the managers of going public �rms are better
informed, or when the stock prices of established peers are less informative. Our
empirical analysis con�rms these predictions. We �nd that the increase in product
di¤erentiation of IPO �rms is larger for �rms whose managers are better informed, as
measured by proxies for the intensity and pro�tability of insider trading. In addition,
IPO �rms for which established peers�stock prices are less informative (as proxied by
the PIN measure or the size of price reactions to earnings surprises) appear to increase
their degree of di¤erentiation signi�cantly more over time. In addition, the increase
in di¤erentiation is larger for �rms whose peers receive less coverage by professional
�nancial analysts.
There is a vast theoretical and empirical literature on conformism in managerial
decisions (see Lieberman and Asaba (2006) for a review). The economic literature on
this topic emphasizes that conformism in strategic choices can originate in reputation
concerns (e.g. Scharfstein and Stein (1990)), information cascades and herding (e.g.
Bikhchandani, Hirshleifer, and Welch (1998), or simply as a response to correlated
signals among related �rms. The mechanism and the consequences analyzed in our
paper are distinct. In our theory, conformism arises in equilibrium because it allows
managers to receive more precise information from the stock market. In addition,
imitation is perfectly rational and value enhancing. To our knowledge, our paper is
the �rst to analyze the connections between conformism and informational feedbacks
from the stock market.
Our paper builds upon the growing literature that studies corporate decision mak-
ing when managers learn information from the stock market (see Bond, Edmans, and
Goldstein (2012) for a recent survey). In general, this literature has focused, both
empirically or theoretically, on the e¤ects of stock price information on real invest-
ment decisions by �rms (see, for instance, empirical analyses in Chen, Goldstein, and
Jiang (2007), Bakke and Whited (2010), Edmans, Goldstein, and Jiang (2012), or
Foucault and Frésard (2012)). This literature has however not considered how the
prospect of learning information from stock prices a¤ects strategic decisions by �rms,
6
such has product di¤erentiation and product market positioning.3
Our paper also adds to the literature that examines the connections between �nan-
cial and product market decisions. Models analyzing the interplay between product
market competition and �rms��nancing choises do not consider the information pro-
duced by the stock market, nor its e¤ect on �rms�product market strategies (e.g.,
Titman (1984), Brander and Lewis (1986), Maksimovic (1988), or Bolton and Scharf-
stein (1990)). Similary, existing research that links product market characteristics to
stock prices typically take the intensity of competition in product markets as given
and analyze how (various dimension of) competition in�uences stock returns (e.g.
Hou and Robinson (2006) or Bustamante (2015)) or informed investors�trading deci-
sions (e.g. Peress (2010) or Tookes (2008)). Our paper focuses on the reverse e¤ect:
How information produced in the stock market can in�uence �rms�di¤erentiation
decisions, and ultimately shapes product market structures.
The rest of the paper is organized as follows. In the next section, we describe
the model and show how the reliance on stock market information can lead to con-
formism in strategic choices. We also show that the going public decision should
weaken this bias, which leads to our main prediction: going public �rms should, on
average, increase product di¤erentiation after IPOs. We present the data used to test
this prediction in Section 3. Section 4 reports the empirical �ndings and Section 5
concludes.
2 Model
Choosing a strategy. At date 1, �rm A chooses a �strategy�, denoted SA. At date
2, the stock market opens, investors observe �rms�strategy, and trade (see below).
At date 3, the manager of �rm A decides to implement or not the strategy chosen
3In our model, a �rm manager learns information from his own stock price. In equilibrium, whenthe �rm chooses the common strategy, its stock price is a su¢ cient statistic for the information inits peer stock price. Thus, implications of the model are identical if �rms learn only from their ownstock price or more generally from the stock price of all �rms following the same strategy. Foucaultand Frésard (2014) provide evidence that �rms rely on their peers�stock prices for their investmentdecisions.
7
at date 1, after observing stock prices at date 2 and receiving private information on
the payo¤ of his strategy (see below). At date 4, the payo¤s of all �rms are realized.
Figure 1 describes the timing of the model.4
[Insert Figure 1 about here]
Firm A can choose one of two possible strategies, denoted Su or Sc. Strategy Su is
a unique strategy whereas strategy Sc is a common strategy, already chosen by n other
public �rms. We interpret a strategy as a di¤erentiation choice. The unique strategy
allows �rm A to signi�cantly di¤erentiate its strategy from its competitors�strategy
while the common strategy, Sc, does not. We denote by n(S) the number of �rms
following strategy S at the end of date 3. As strategy Su is unique, we have n(Su) = 1
if A chooses it and implements it. Otherwise n(Su) = 0. Similarly, n(Sc) = n + 1 if
the manager of �rm A chooses strategy Sc and implements it. Otherwise n(Sc) = n.
If the manager of �rm A abandons his strategy at date 3, he bears no cost but he
cannot switch to a new strategy. Firm A�s payo¤ is then zero. If instead the manager
of �rm A chooses to implement his strategy, �rm A must invest an indivisible amount,
normalized to one. For public �rms following strategy Sc, the implementation cost
is sunk. These �rms represent established �rms who have already decided to follow
strategy Sc and incurred corresponding investments.
We denote by r(S; n(S)) the expected return of strategy S per dollar invested.
The actual return of the strategy is uncertain, however. For instance, a �rm with
a di¤erentiated product might �nd no consumers with a taste for his product. To
capture this uncertainty, we assume that both the common and the unique strategy
can be Good (G) or Bad (B) with equal probabilities and we denote by tS 2 fG;Bg,the type of strategy S 2 fSu; Scg. This type becomes known at date 4 and determinesthe realized return on a strategy at this date. The realized return on a good strategy
is r(S; n(S); G) = r(S; n(S)) + �S while the realized return on a bad strategy is
4While we focus on product di¤erentiation strategies, the timing of the model resemble theevidence provided by Luo (2005) who document that �rms annouce an acquisition strategy, andthen decide to implement or not the strategy (i.e. pursue the acquisition) based on the stock marketreaction.
8
r(S; n(S); B) = r(S; n(S)) � �S. Thus, the expected net present value (NPV) ofimplementing strategy S for �rm A is
E(NPV(S; n(S))) = r(S; n(S))� 1 for S 2 fSu; Scg. (1)
We assume that the payo¤ of a good (bad) unique strategy is higher than the
payo¤ of a good (bad) common strategy:
A.1: �(n) � r(Su; 1; tS)� 1r(Sc; n+ 1; tS)� 1
> 1, (2)
This assumption captures the notion that di¤erentiation is a way to increase revenues,
provided the common and the di¤erentiated strategy have the same type.5 It is
natural (but not critical for our �ndings) to assume that �(n) increases with n: as
more �rms follow the common strategy, competition among these �rms intensi�es
and the return of the common strategy decreases.6 For the problem to be interesting,
arrival of information about the type of a strategy must a¤ect the manager�s decision
to pursue or not his strategy. Thus, we assume that the net present value of a good
strategy is positive (r(S; n(S); G) > 1) but that the net present value of a bad strategy
is negative (r(S; n(S); B) < 1) for �rm A. Thus, �rm A will implement his strategy
at date 3 if its manager learns that the strategy is good and abandons it otherwise.
We also assume that the expected net present value of both strategies for �rm A is
negative:
A.2 : r(S; n(S)) � 1 for S 2 fSu; Scg. (3)
Thus, absent new information at date 3, the manager of �rm A always abandons
his strategy, whether he chose the unique or the common strategy at date 1. This
assumption can be relaxed without a¤ecting qualitatively the main �ndings of the
5In line with this assumption, Hoberg and Phillips (2015) show empiricall that unique �rms(de�ned as �rms whose product o¤ering is di¢ cult to replicate using a combination of other �rms)display higher valuation compared to less unique �rms.
6The alternative assumption is that �(n) < 1 and �(n) decreases with n. In this case, �rms�strategies are complements rather than substitutes in the sense that a �rm�s payo¤ is higher whenmore �rms choose the same strategy. In this situation, �rms would choose not to be di¤erentiatedeven when they do not learn information from the stock market. We focus on the other, morenatural case, precisely to highlight the fact learning from the stock market can induce conformity.
9
paper. Finally, we assume that :
A.3: r(Sc; n+ 1; G) > r(Sc; n; B);
which is equivalent to r(Sc; n) � r(Sc; n + 1) < 2�c. This means that the payo¤ of
a good common strategy chosen by n + 1 �rms is higher than the payo¤ of a bad
strategy chosen by n �rms. This assumption implies that observing that the strategy
of a �rm is good is a more positive news than observing that it is bad even if, in
the former case, one extra �rm (�rm A) decides to adop the strategy. It guarantees
that, in equilibrium, informed investors prefer to buy stocks of established �rms than
selling them when they learn that the common strategy is good.
In the baseline version of the model, we assume that �rm A is public. Hence,
the manager of �rm A has three sources of information when he decides or not to
implement his strategy at date 3. First, he privately observes a signal sm 2 fG;B;?gabout the type of his strategy. Speci�cally, sm = tS with probability or sm = ?
with probability (1� ), where ? is the null signal corresponding to no signal. Thus, measures the likelihood that the manager has full information about the type of
his strategy. We refer to sm as �direct managerial information� and to as the
quality of this information. Second, the manager of �rm A observes the stock prices
of established �rms, denoted by pj2 for j 2 f1; :::; ng. Finally, the manager of �rm A
observes his own �rm�s stock price.
Let I be the manager�s decision at date 3, with I = 1 if the manager of �rm A
implements his strategy and zero otherwise. At date 3, for a given decision, I, the
expected value of �rm A is
VA3(I; SA) = I � E(NPV(SA; n(SA)) j3 ); (4)
where 3 = fp12; ::; pn2; pA2; smg is the information set of the manager when he makeshis decision at date 3. Firm A faces no �nancing constraints and, at date 3, its
manager makes the decision I that maximizes VA3(I; SA). We denote by I�(3; SA)
the optimal decision of the manager at date 3 given his information at this date.
10
At date 1, the manager has no information and the value of �rm A is
VA1(SA) = E(I�(3; SA)�NPV(SA; n(SA))): (5)
The manager chooses the strategy, S�A, at date 1 that maximizes VA1(SA).
The Stock Market. There are three types of investors in the stock market:
(i) a continuum of risk-neutral speculators, (ii) liquidity traders with an aggregate
demand zj, uniformly and independently distributed over [�1; 1], for �rm j, and (iii)risk neutral dealers.
Each speculator assesses strategies chosen by publicly listed �rms and obtains a
signal bsi(S) 2 fG;B;?g about the type of strategy S. We assume that a fraction �Sof speculators receives a perfect signal (i.e., bsi(S) = G) about strategy S. Remainingspeculators observe no signal about this strategy (bsi(S) = ? for these speculators).
After receiving her signal on strategy S, a speculator can choose to buy or sell one
share in all stocks of �rms following this strategy or he can decide not to trade at
all. We denote by xi(bsi(S(j))) 2 f�1; 0;+1g the demand of speculator i for sharesof �rm j following strategy S(j) given her signal about this strategy.
Let fj(S(j)) be the order �ow �the sum of speculators and liquidity traders�net
demand �for the stock of �rm j when it follows strategy S(j):
fj = zj + xj(S(j)); (6)
where xj =R 10xi(bsi(S(j)))di is speculators�aggregate demand of stock j. As in Kyle
(1985), order �ow in each stock is absorbed by dealers at a price such that they
just break even given the information contained in the order �ow. We assume that
market makers observe the realizations of order �ows in each market when they set
their prices.7 Thus,
pA2(fA(SA)) = E(VA3(I�(3; SA); SA) j 2), (7)7Thus, the stock price for each stock re�ects all available information at the end of date 1, not
just the information contained in the order �ow of the stock. It is natural to focus on this case sincemanagers make their decisions at low frequency relative to the frequency at which market makerscan observe order �ows in all stocks and adjust their prices accordingly. Thus, by the time at whichmanagers make their decisions, stock prices are likely to re�ect all order �ow information. In anycase, the main implications of the model are identical if market makers can condition their priceson the order �ow in their stock only.
11
and,
pj2(fj(Sc)) = E(r(Sc; n(Sc); tSc) j 2) for j 2 f1; ::; ng, (8)
where 2 = ff1; :::; fn; fAg. Hence, using the Law of Iterated Expectations, the
stock prices of �rms A and j 2 f1; ::; ng at date 1 are pA1(I�) = VA1(SA) and
pj1(fj(Sc)) =E(r(Sj; n(Sc); tSc)), respectively.
Equilibrium. The stock price of �rm A clearly depends on the manager�s opti-
mal decision, I�(3; SA). The stock prices of established �rms also depend on this
decision when �rm A chooses the common strategy. Indeed, the manager�s decision
to pursue the strategy at date 3 determines the number of �rms following the es-
tablished strategy, which in turn determines established �rms�payo¤s. In turn, the
manager�s optimal decision at date 3 itself depends on stock prices. Thus, in equi-
librium, the manager�s optimal decision, I�(3; SA), and the stock prices of all �rms
are jointly determined. Formally, a stock market equilibrium for is a set of trading
decisions x�i (bsi(Sj)), stock prices p�j2(2) and a decision rule I�(3; SA) such that (i)the trading strategy x�i (�) maximizes the expected pro�t for speculator i given othertraders�actions and �rm A�s decision rule, (ii) the decision rule I�(�) maximizes theexpected value of �rm A, VA3(I; S), at date 3, and (iii) p�j2(�) solves (7) given thatagents behave according to x�i (�), and I�(�).
2.1 The stock market and strategic conformity
As a benchmark, we �rst consider the case in which the manager does not have access
to stock market information (or ignores it). In this case, the manager only relies on
his private information. Therefore, he implements his strategy at date 2 if he learns
that the strategy is good and does nothing otherwise. Thus, in this case, the expected
value of �rm A at date 1 is:
V benchmarkA1 (SA) =
2(r(SA; n(SA); G))� 1):
We deduce that V benchmarkA1 (Su)=VbenchmarkA1 (Sc) = �(n) > 1. Hence, the manager
optimally chooses the unique strategy in the benchmark case.
12
Proposition 1 (benchmark) The manager of �rm A optimally chooses the unique
strategy when he does not use information from the stock market.
We now analyze how stock market information a¤ects the choices of the manager
of �rm A. For this we must �rst derive the equilibrium of the stock market when �rm
A chooses the common strategy. Let de�ne pHA (SA) = r(SA; n(SA); G)�1, pLA(SA) = 0,and pMA (SA) = V
benchmarkA1 (SA). Observe that pHA (SA) > p
MA (SA) > p
LA(Sc).
Lemma 1 When �rm A chooses the common strategy, the equilibrium of the stock
market at date 2 is as follows:
1. Speculator i buys one share of �rm j if bsi(Sc) = G, sells one share of �rm j ifbsi(Sc) = B, and does not trade otherwise.2. The stock price of an established �rm is (i) pj = if the order �ow of one stock
(including stock A) is larger than (1��c), (ii) pj = (1� =2)r(Sc; n)+ r(Sc; n+1)=2 if the order �ow of all stocks (including stock A) belongs to [�(1��c); (1��c)], (iii) pj = r(Sc; n; B) if the order �ow of one stock (including stock A) is
less than (1� �c):
3. The stock price of �rm A is (i) pHA (Sc) if the order �ow of one stock (including
stock A) is larger than (1 � �c), (ii) pMA (Sc) if the order �ow of all stocks
(including stock A) belongs to [�(1��c); (1��c)], and (iii) pLA(Sc) if the order�ow of one stock (including stock A) is less than (1� �c):
4. If the manager receives a private signal then he implements the strategy is his
signal indicates that the strategy is good and abandons it if his signal indicates
that the strategy is bad. Else, the manager follows his stock price: he implements
his strategy at date t = 2 if his stock price is pHA (Sc) and abandons it otherwise.
If the manager receives a private signal about the type of his strategy, he only
relies on this signal for his decision at date 2 since this signal is perfect (Part 4 of
13
Lemma 1). Otherwise, the manager relies on the information conveyed by his own
stock price to make his decision.
The information obtained from the stock market depends on the realization of
the total demand (fj) for each stock. If speculators have negative information, they
optimally sell stocks and, therefore, the largest possible realization of the order �ow
in this case is less than (1 � �c). Thus, when the demand for one stock is higherthan (1� �c), market makers infer that speculators know that the common strategyis good. Thus, the stock price of all �rms, including �rm A, adjusts to its highest
possible level (Part 3 of Lemma 1), which signals to the manager of �rm A that the
common strategy is good.8 Hence, in this case, the manager of �rm A decides to go on
with the common strategy and he implements it (Part 4 of Lemma 1), even if he has
no private information. Symmetrically, market makers infer from a relatively weak
demand in one stock (i.e., a realization of the order �ow less than �(1 � �c)) thatspeculators have negative information about the type of the common strategy. Thus,
the stock price of all �rms, including �rm A, adjusts to its lowest possible level,
which signals to the manager of �rm A that the common strategy should not be
pursued further. Last, market makers cannot infer whether speculators have received
a positive or a negative signal when the demand for each stock is neither relatively
strong or relatively weak, that is, when fj belongs to [�(1 � �c); (1 � �c)] for eachstock. Indeed, realizations of the order �ow in this range are equally likely when the
strategy is good or when the strategy is bad. Thus, in this case, stock prices do not
contain information and the manager of �rm A optimally chooses not to pursue the
strategy since its unconditional net present value is negative (Assumption A.1).
8The stock price of established �rm is higher when market makers learn from order �ow that thestrategy is good than when they learn no information because r(Sc; n+ 1; G) > (1� =2)r(Sc; n) + r(Sc; n+1)=2; under Assumption A.3. When Assumption A.3 is not satis�ed, this is not necessarilythe case. Intuitively, discovery by the stock market that the common strategy is good has anambiguous e¤ect on the value of established �rms. On the one hand, it raises their value because,other things equal, a good strategy yields a larger cash-�ow than a bad strategy. On the other hand,it lowers the value of established �rm because it implies that �rm A will implement his strategy(which means that established �rms willl face one additional competitor). The former e¤ect alwaysdominates i¤ A.3 is satis�ed.
14
Using eq.(5), the value of �rm A at date 1 is:
VA1(Sc) = (Pr(I� = 1 jtSc = G) (r(Sc; n+1; G)�1)+Pr(I� = 1 jtSc = B) (r(Sc; n+1; B)�1))=2:
(9)
From the last part of Lemma 1, we deduce that:
Pr(I� = 1 jtSc = G) = + (1� ) Pr(pA1 = pHA (Sc) jtSc = G) . (10)
Thus, stock market information increases the likelihood that the manager of �rm A
will implement the strategy chosen at date 1 when it is good. Indeed, conditional
on the strategy being good, the manager will implement the strategy either when
(i) he receives managerial information (as in the benchmark case) or (ii) if its stock
price at date 1 is high (i.e., equal to pHA (Sc)) when his private signal is uninformative.
Using the fact that the demand from liquidity traders is uniformly and independently
distributed across stocks, we deduce that9
Pr(pA1 = pHA (Sc) jtSc = G) = �(n; �c)
def= 1� (1� �c)n+1
As the number of �rms following the common strategy increases, the likelihood that
stock prices reveal the type of the common strategy increases. This explains why
�(n; �c) increases with n.
When the common strategy is bad, speculators sell all stocks. Accordingly, the
order �ow in each stock is at most (1 � �c) and therefore the stock price of �rmA has a zero probability of being high. Moreover, if the manager receives private
information, this information indicates that the strategy is bad and, therefore, the
manager does not implement the strategy. Thus, the likelihood that the manager of
�rm A implements a bad strategy is zero: Pr(I� = 1 jSc = B) = 0. We deduce fromeq.(9) that the expected value of �rm A at date 1 is:
VA1(Sc) =( + (1� )�(n; �c))
2(r(Sc; n+ 1; G)� 1): (11)
9To see this, observe that, Pr(pA1 = pHA (Sc) jtSc = G) = Pr([j=n+1j=1 (fj) � (1 � �c) jSc = G) =1 � Pr(\j=nj=1 (fj) < (1 � �c) jSc = G) according to Lemma 1. As fj = zj + �c when tSc = G, we
deduce that Pr(pA1 = pHA (Sc) jtSc = G) = 1�Pr(\j=n+1j=1 fzjg < (1� 2�c)) = 1� (1� �c)n+1, where
the last equality from the fact that the zjs are uniformly and independently distributed over [�1; 1].
15
The next lemma describes the equilibrium of the stock market when �rmA chooses
the unique strategy.
Lemma 2 :When �rm A chooses the unique strategy, the equilibrium of the stock
market at date 2 is as follows:
1. Speculator i buys one share of �rm j if bsi(Sj) = G, sells one share of �rm j ifbsi(Sj) = B, and does not trade otherwise.2. The stock price of an established �rm is (i) pj = r(Sc; n;G) if the order �ow in
the stock of one established �rm is larger than (1� �c), (ii) pj = r(Sc; n) if theorder �ow of all stocks of established �rms belongs to [�(1� �c); (1� �c)], (iii)pj = r(Sc; n; B) if the order �ow in the stock of one established �rm is less than
(1� �c):
3. The stock price of �rm A is (i) pHA (Su) if the order �ow in stock A is larger than
(1��u), (ii) pMA (Su) if the order �ow in stock A belongs to [�(1��u); (1��u)],and (iii) pLA(Su) if the order �ow in stock A is less than (1� �u):
4. If the manager receives a prive signal then he implements the strategy is his
signal indicates that the strategy is good and abandons it if his signal indicates
that the strategy is bad. Else, the manager follows his stock price: he implements
his strategy at date t = 2 if his stock price is pHA (Su) and abandons it otherwise.
The equilibrium is very similar to that obtained when �rm A chooses the common
strategy. The only important di¤erence is that when �rm A chooses the unique
strategy, market makers in �rm A cannot learn information from trades in stocks of
established �rms. Thus, the stock price of �rm A only depends on the order �ow in
its stock. Accordingly the probability that the manager of �rm A will implement its
strategy when it is good depends only on the fraction of speculators informed about
the unique strategy, �u. This probability is
Pr(I� = 1 jtSu = G) = + (1� ) Pr(pA1 = pHA (Su) jtSu = G) = + (1� )�u:
16
Then proceeding as in the case in which �rm A chooses the common strategy, we
deduce that the expected value of the �rm at date 1 when it chooses the unique
strategy is
VA1(Su) =( + (1� )�u)
2(r(Su; 1; G)� 1): (12)
For a given strategy, the expected value of �rm A when it uses stock market
information is higher than when it ignores it: VA1(SA) � V benchmarkA1 (SA) for SA 2fSc; Sug, with a strict inequality if < 1. The reason is that stock market informationcomplements managerial information and therefore enhances the manager�s ability to
make value enhancing decisions. Moreover, as explained previously, the stock price of
�rm A at date 1 is p1A = VA1(SA). Hence, whether the manager of �rm A announces
the common or the unique strategy, we have pLA2 < pMA2 < pA1 < p
HA2. Thus, Lemma 1
and 2 imply that the manager of �rm A goes on with the strategy announced at date
1 if, in reaction to this announcement, his stock price goes up (�pA = pA2� pA1 > 0)and abandons it if his stock price goes down. This explains why the stock price of �rm
A drops even if market makers cannot infer from demands of various stocks whether
speculators have a good or a bad signal about �rm A�s strategy (i.e., why pMA2 < pA1).
Indeed, in this case, market makers factor in their price the fact that the �rm is less
likely to pursue his strategy, which tends to lower its value.
In either case, the value of �rm A is higher when the fraction of speculators who
are informed about the payo¤ of its strategy (�u or �c) is higher. The reason is that
the informativeness of the stock market about a strategy increases with the fraction
of speculators informed about this strategy. For instance, consider the case in which
�rm A chooses the common strategy. It follows from the third part of Lemma 1 that
the stock price of �rm A reveals the type of its strategy with probability �(n; �c),
which increases with �c. Thus, we refer to �(n; �c) (or �u depending on the strategic
choice of �rm A) as the informativeness of the stock market for the manager of �rm
A.
The next proposition states our main result.
17
Proposition 2 (Conformity e¤ect): If �u < �(n; �c) then, at date 1, �rm A op-
timally chooses the common strategy if �(n) < b�( ; �u; �c; n) and it chooses theunique strategy if �(n) > b�( ; �u; �c; n), where b�( ; �u; �c; n) = ( +(1� )�(n;�c)
( +(1� )�u) > 1. If
�u > �(n; �c) then �rm A always chooses the unique strategy.
Hence, there is a set of values for the parameters (�u < �(n; �c) and �(n) <b�( ; �u; �c; n)) such that the manager of �rm A chooses the common strategy while
he always chooses the unique strategy when he ignores stock market information
(see Proposition 1). This shows that when managers rely on the stock market as a
source of information, their incentive to di¤erentiate is weakened. We label this the
conformity e¤ect.
The intuition for this �nding is as follows. When �rm A chooses the unique
strategy, its stock price reveals the type of its strategy with probability �u while when
�rm A chooses the common strategy, its stock price reveals its type with probability
�(n; �c). Thus, if �u < �(n; �c), the stock market is more informative about the
value of �rm A�s strategy if it does not di¤erentiate. In this case, the manager of
�rm A faces a trade-o¤ in choosing his strategy: di¤erentiation yields a larger payo¤
if the unique strategy is good but the manager receives a less informative signal from
the stock market about whether the strategy is good or bad. He is therefore less
likely to implement a good (unique) strategy when it should indeed be implemented.
If �(n) < b�( ; �u; �c; n)), the latter e¤ect dominates the former and the manager isbetter o¤not di¤erentiating. If �u > �(n; �c), there is no trade-o¤since di¤erentiation
brings both a larger payo¤ if the manager�s strategy is good and is associated with a
more informative stock market.
The case in which �u < �(n; �c) is more likely to occur in reality for two reasons.
First, �(n; �c) quickly increases with the number of established �rms. Thus, even if
�c is smaller than �u, �(n; �c) can be larger if n is large enough (e.g., for n = 5,
we have �(n; �c) = 40% if �c = 10%). The reason is that intermediaries (market
makers) can extract more information from speculators�trades since trades in one
18
stock are informative about the payo¤s of other stocks.10 Second, speculators can
use information about the common strategy to speculate in all stocks of �rms fol-
lowing this strategy. Thus, for a �xed cost of producing information, they bene�t
from economies of scale when producing information about the common strategy.
If information acquisition is costly and endogenous, this scale e¤ect should tend to
make the informativeness of the stock market about the common strategy (�(n; �c))
larger than its informativeness (�u) about unique strategies in equilibrium, as the
next corollary shows.11
Corollary 1 Suppose that speculators must pay a cost C to become informed about
a strategy. In equilibrium, the informativeness of the stock market about the common
strategy when n + 1 �rms follow this strategy is higher than the informativeness of
the stock market about the unique strategy when one �rm follows this strategy, i.e.,
�(n; ��c) > ��u if n �
(r(Su;1)+�u��c)(2�c� (r(Sc;n)�r(Sc;n+1)) :
To understand the corollary, observe that the standard deviation of possible pay-
o¤s for an established �rm is �c while it is (r(Su; 1) + �u)=2 for �rm A when it
follows the unique strategy.12 Thus, one attractive feature of the unique strategy
for speculators is that its speculative value is higher if 2�c < (r(Su; 1) + �u). If
n � (r(Su;1)+�u��c)(2�c� (r(Sc;n)�r(Sc;n+1)) , this e¤ect however is not strong enough to o¤set the econ-
omy of scale e¤ect and, in equilibrium, the fraction of informed speculators about
each type of strategy is such that �(n; ��c) > ��u.13 Importantly, one can obtain
10See Pasquariello and Vega (2014) and Boulatov et al.(2011) for evidence of such cross assetlearning by market makers.11In unreported tests, we con�rm this intuition using the proxies for uniqueness and price infor-
mativeness de�ned in the empirical section below.12Indeed, if �rm A follows the unique strategy, it will keep the strategy if it learns that it is good
and abandons it if it is bad or if the manager learns no information about the type of the strategy.Thus, in equilibrium (i.e., accounting for the optimal exercise policy of the abandonment option),the payo¤s for a �rm adopting the unique strategy are r(Su; 1) + �u or zero.13The corollary compares the informativeness of the stock market about the common strategy
when �rm A chooses this strategy (so that n+1 �rms choose this strategy) with the informativenessof the stock market about the unique strategy when �rm A chooses this strategy. Of course, if �rmA chooses the common strategy, the informativeness of the stock market about the unique strategyis then nil since no �rms choose this strategy.
19
�(n; ��c) > ��u even if the equilibrium demand of information about the common
strategy ��c is less than that about the unique strategy, ��u.
It is easily seen that b�( ; �u; �c; n) decreases with and goes to one when goesto one. Indeed, when the manager has more precise private information, he needs
to rely less on stock market information. Hence, the informational gain of adopting
the common strategy �the conformity e¤ect �is smaller. This informational gain is
also smaller (larger) when �u (�(n; �c)) is higher so that b�( ; �u; �c; n) decreases with�u and increases with �(n; �c). Moreover when goes to zero and �u goes to zero,b�( ; �u; �c; n) become in�nitely large. In these cases, �rm A chooses the common
strategy even if the increase in payo¤ with a successful unique strategy is very large.
2.2 A unique testable implication
2.2.1 Strategic choices of public status
One way to test whether the conformity e¤ect plays a role in di¤erentiation decisions
of �rms is to analyze empirically whether and how a variation in �u a¤ects the
choice of its strategy by a �rm. For instance, an exogenous increase of �u increases
the likelihood that a given �rm shifts to a more unique strategy (since it lowersb�( ; �u; �(n; �c))). One immediate problem with this approach is that �u is not
observed unless the �rm decides to di¤erentiate and choose the unique strategy in
the �rst place. When a �rm is private, however, �u is by de�nition zero since the
�rm is not publicly traded such that it cannot attract information gathering by any
speculator. In contrast, �c is di¤erent from zero irrespective of whether �rm A is
private or public as long as established �rms are public. Now suppose that �rm A
goes public. If it shifts to the unique strategy then �u > 0 as its stock will attract
some trading from informed investors (maybe a small fraction but at least a strictly
positive fraction). Hence, going public is a positive shock on �u. All else equal,
going-publci should therefore increase �rm A�s incentive to di¤erentiate its strategy.
To formalize this intuition, suppose �rst that �rm A is private. In this case, if
�rm A chooses the unique strategy then it cannot obtain information from the stock
20
market. Thus, its expected value at date 1 is identical to that in the benchmark case
(or to the case in which �u = 0):
V privateA1 (Su) = VbenchmarkA1 (Su): (13)
If instead, �rm A chooses the common strategy then its manager can learn infor-
mation about his strategy from the stock price of the n established public �rms. In
particular, if the stock price of these �rms is high then the manager of �rm A can
infer that the common strategy is good and therefore he will choose to implement
the common strategy in this case, even if he does not receive managerial information.
Thus, proceeding as in the case in which �rm A is public, we deduce that the expected
value of �rm A when it is private and when it chooses the common strategy is14
V privateA1 (Sc) = VA1(Sc) =( + (1� )�c(n� 1))
2(r(Sc; n+ 1; G)� 1): (14)
The only di¤erence with the expression obtained when �rm A is public (eq.(7)) is that
�(n � 1; �c) replaces �(n; �c). The reason is that the manager of �rm A learns that
the common strategy is good if the stock price of the n established public �rms is high
but it cannot learn information from its own stock price. For reasons explained in
the previous section, this event happens with probability �(n�1; �c) = 1� (1��c)n.
Proposition 3 When �rm A is private, it optimally chooses the common strategy at
date 1 if �(n) < b�private( ; �c; n) and it chooses the unique strategy if b�private( ; �c; n),where b�private( ; �c; n) = ( +(1� )�c(n�1)
.
In choosing its strategy, �rm A faces the same trade-o¤ when it is private and
when it is public. Thus, its behavior is identical in each case: it chooses to di¤eren-
tiate only if the gain from di¤erentiation o¤sets losses due to less informed decision,
that is if �(n) exceeds a threshold (b�private( ; �(n; �c))) that is stricly larger thanone. However, this threshold is di¤erent than that obtained when the �rm is pub-
lic (b�private( ; �c; n) 6= b�( ; �u; �c; n))). In particular, if �u > (�(n;�c)��(n�1;�c)) +(1� )�(n�1;�c) then
14For brevity, we formally derive eq.(14) in the on-line appendix for the paper. When �rm A isprivate and chooses the common strategy, its decision to implement or not its strategy is determinedby established �rms�stock prices rather than its own stock price, as in Lemma 1).
21
b�private( ; �c; n) > b�( ; �u; �c; n). That is, the manager of �rm A chooses the commonstrategy for a larger set of parameters when it is private than when it is public. In
this sense, the conformity e¤ect is stronger when a �rm is private than when it is
public. The reason is as follows. When �rm A is public, the informativeness of the
stock market about the strategy chosen by �rm A is enhanced but for two di¤erent
reasons depending on whether �rm A chooses the unique strategy or the common
strategy. If it chooses the common strategy, stock price informativeness is larger by
�(n; �c)� �(n� 1; �c) because market makers can now learn from trades in the mar-ket of stock A as well. This increases the likelihood that speculators�trades reveal
their information about the common strategy. If instead �rm A chooses the unique
strategy, the stock market is more informative for �rm A simply because �rm A can
learn from its own stock price when public while it cannot if private. This second
e¤ect dominates the former when �u > (�(n;�c)��(n�1;�c)) +(1� )�(n�1;�c) . We therefore obtain the
following implication.
Corollary 2 (the conformity e¤ect is weaker for public �rms). Suppose that �u > (�(n;�c)��(n�1;�c)) +(1� )�(n�1;�c) . If �(n) 2 [b�( ; �u; �c; n); b�private( ; �c; n)] then �rm A chooses the
common strategy if it is private and the unique strategy if it is public. Otherwise �rm
A chooses the same strategy whether public or private. Thus, other things equal, a
public �rm is more likely to choose a di¤erentiated strategy than a private �rm.
Figure 2 illustrates Corollary 2. It shows the optimal strategic choice of �rm A
when it is public or private for speci�c parameter values. When �u < (�(n;�c)��c(n�1)) +(1� )�c(n�1)
then b�private( ; �c; n) < b�( ; �u; �c; n). In this case, if �(n) 2 [b�private( ; �c; n); b�( ; �u; �c; n)]then �rm A chooses the unique strategy when private and the common strategy when
public.15 Thus, the previous result is reversed: the conformity e¤ect is now stronger
for public �rms. The reason is that stock price informativeness about the common
strategy increases by �(n; �c) � �(n � 1; �c) = �nc (1 � �c) when �rm A is public
rather than private while the informativeness of price about the unique strategy in-
creases by �u if �rm A chooses this strategy when public. If �u is small relative to15For other values of �(n), �rm A chooses the same strategy whether public or private.
22
�(n; �c)��c(n�1) then the common strategy is relatively more attractive when �rmA is public. This case however is unlikely because �(n; �c)��(n�1; �c) = �nc (1��c)decreases quickly with n and in fact the condition �u <
(�(n;�c)��(n�1;�c)) +(1� )�(n�1;�c) can never
be satis�ed if �u > 14. Thus, Corollary 2 describes the most plausible scenario and is
our main testable implication.
2.3 Methodological issues
A natural way to test Corollary 2 is to study empirically whether a given �rm chooses
to be more di¤erentiated when it is public than when it is private. In principle, our
model applies to any strategic choices. In our tests, we measure strategic di¤erentia-
tion with an index of product di¤erentiation for each �rm relative to a set of publicly
listed peers, J ( "established �rms" in the model). Let B be a �rm from the set JA
of peers for �rm A and let �A;B(kA; �A) be the degree of product di¤erentiation of
�rm A vis-à-vis B when �rm A has ownership status kA 2 fprivate; publicg and type�A represents the gain of being di¤erentiated for A (i.e., �(n) in the model). Finally,
let �A(�A) be the average di¤erence in the degree of product di¤erentiation of �rm
A and is peers when A is private and when A is public:
�A(�A) =1
n
XB2JA
(�A;B(public; �A)��A;B(private; �A)); (15)
Corollary 2 implies that �A(�A) is either zero or positive depending on �A. One
di¢ culty is that we can empirically measure the degree of di¤erentiation of a �rm
and its peers only when a �rm is public as we do not have proxies for di¤erention
before a �rm goes public (see Section 3). To overcome this problem, we look at the
evolution of �A;B(public; �A) over time, starting from the year in which a �rm goes
public.16 Speci�cally, let de�ne the event time variable � = 0; 1; :::; T as the public
age of �rm A since its IPO, with � = 0 being the year of its IPO. We assume that
�A;B(private; �A) can be proxied using the degree of product di¤erentiation between
A and B measured at the time of �rm A�s IPO, or �A;B;�=0(public; �A). We then
16This is similar to the approach in Spiegel and Tookes (2014) or Chod and Lyandres (2011).
23
measure �A(�A) by considering the evolution of �A;B;� (public; �A) over time. We do
so by estimating a linear regression of the form:
�A;B;� (public; �A) = �A;B + �A � � + �A;B for all B 2 JA; (16)
where the coe¢ cient �A measures the average change of product di¤erentiation be-
tween �rms A and the set of �rms B over time. Thus, �A is the empirical counterpart
of �A(�A) and therefore our prediction is �A � 0. Empirically, we estimate eq.(16) fora panel dataset that stacks �A;B;� (public; �A) for a large number of IPO �rms equiv-
alent to A, and public �rms B, and create a set of counterfactual pairs.17 The gain of
di¤erentiation is unobserved for each �rm but on average it should be strictly posi-
tive as, for some �rms, �A should fall in the interval [b�( ; �u; �c; n); b�private( ; �c; n)].Thus, we expect �A (the cross-sectional average evolution of di¤erentiation after an
IPO) to be strictly positive. We describe in the next section the sample construction
and detail the econometric implementation of eq.(16).
3 Data and Methodology
3.1 Measuring strategic choices
To measure the degree of product di¤erentiation between two �rms (�i;j), we rely
on the Text-Based Network Classi�cation (TNIC) developed by Hoberg and Phillips
(2015). This classi�cation is based on textual analysis of the product description
sections of �rms�10-K (Item 1 or Item 1A) �led every year with the Securities and
Exchange Commission (SEC). The classi�cation covers the period 1996-2011.18 For
each year, Hoberg and Phillips (2015) compute a measure of product similarity (�i;j)
for every pair of �rms by parsing the product descriptions from their 10-Ks. This
measure is based on the relative number of product words that two �rm share in
their product description, and ranges between 0% and 100%. Intuitively, the more
17Note that by capturing �A with a �rm �xed e¤ect we implicitly assume that the gain of beingdi¤erentiatied is �xed for a given �rm around its IPO.18This limitation arises because TNIC industries require the availability of 10-K in electronically
readable format.
24
common words two �rms use in describing their products, the closer they are in the
product market space, or equivalently the less di¤erentiated they are.
Hoberg and Phillips (2015) then de�ne, for each year, each �rm i�s set of peers to
include all �rms j with pairwise similarity scores relative to i above a pre-determined
threshold (equal to 21.32%).19 This represents the TNIC network of �rm pairwise
similarity. Unlike standard industry de�nitions, the TNIC network does not require
relations between �rms to be transitive. Each �rm has its own distinct set of peers,
that can change over time as �rms modi�es their product ranges, innovate, and enter
new markets. Following Hoberg and Phillips (2015), we use the similarity score (�i;j)
for each pair in the TNIC network as the basis to measure the intensity of product
di¤erentiation of IPO and established �rms. We simply de�ne the degree of product
di¤erentiation between any two �rms i and j (in the TNIC network) in year t as
�i;j;t = 1� �i;j;t.As as alternative way to measure strategic di¤erentiation, we rely on stock return
comovement between two �rms (�i;j). The idea is that stock returns of �rms that
follow less di¤erentiated strategies should co-move more.20 On this ground, we rely
on stock return co-movements between �rms in a pair to capture the extent to which
their strategies are related. To obtain this measure, we estimate for each �rm-pair-
year the following speci�cation:
ri;w;t = �0 + �m;trm;w;t + �i;j;trj;w;t + �i;w;t , (17)
where ri;w;t is the (CRSP) return of �rm i in week w of year t, rm;w;t is the return
of the market (CRSP value-weighted index), and rj;w;t is the return of �rm j. The
estimate of �i;j;t thus measures the return co-movement between �rms i and j in year
t. We conjecture that increased di¤erentiation leads to lower return co-movement.21
19This threshold is chosen to generate set of product market peers with the same fraction of pairsas 3-digit SIC industries.20This is the case in the model. The stock returns of �rms that follow the common strategy
are perfectly positively correlated. In contrast, the stock return of �rm A and the stock return ofestablished �rms are uncorrelated if �rm A follows the unique strategy.21Consistent with this claim, the correlation between�i;j and �i;j is -0.29 across all �rm-pair-years
of our sample.
25
3.2 Initial public o¤erings
We obtain the name, CRSP identi�er, and �ling date of �rms going public from the
IPO database assembled by Jay Ritter.22 We restrict our attention to the IPOs during
the 1996-2011 period. The sample includes IPOs with an o¤er price of at least $5.00,
and excludes American Depositary Receipts (ADRs), unit o¤ers, closed-end funds,
Real Estate Investment Trusts (REITs), partnerships, small best e¤orts o¤ers, and
stocks not listed on CRSP (CRSP include Amex, NYSE and NASDAQ stocks). We
further restrict the sample to exclude non-�nancial �rms (SIC codes between 6000
and 6999) and utilities (SIC codes between 4000 and 4999), �rms that are not present
in the TNIC network, �rms without any TNIC peers on their IPO year, �rms that
are listed for less than one year, and �rms with missing information on total assets
in COMPUSTAT. The �nal IPO sample comprises 1,214 going public �rms.
3.3 Econometric speci�cation
To implement Equation (16) and empirically measure the evolution of di¤erentiation
for IPO �rms, we �rst need to identify the set of established �rms for each newly
public �rm. We select, for each IPO �rm A, the set of TNIC peers on the year of
�rm A�s IPO (� = 0). We label this set, whose size varies by IPO �rm, as the initial
peers. To best map the model�s structure, we consider only established peers, de�ned
as peers that have been publicly listed for at least �ve years on �rm A�s IPO year.23
Then, we track the product di¤erentiation between the IPO �rm A and each of its
initial peer B over the �ve year that follows the IPO year (�A;B;� with � = 0; :::; 5). If
a peer B leaves the set of initial peers (i.e. is no longer in �rm A�s TNIC network) in a
given year � (where � > 0) , we set �A;B;� equals to one (i.e., perfectly di¤erentiated).
Arguably, the degree of di¤erentiation between any two �rms A and B re�ects
their joint product market strategies. Hence an increase of �A;B;� following �rm
A�s IPO might indicate that A di¤erentiates from B, but also that B di¤erentiates
22We thank Jay Ritter for sharing this data with us.23All our results are robust i¤ we de�ne established �rms as �rms that have been listed for more
than 3 years.
26
from A, or that both �rms (independently) increase their product di¤erentiation. In
addition, a change in the degree of di¤erentiation post IPO could also be observed if
di¤erentiation naturally change for every �rm over their life-time.
To better measure the situation where the IPO �rm A di¤erentiates from the
established �rm B and to capture general di¤erentiation patterns, we construct the
following counterfactual sample. For each initial peer �rm B (of the IPO �rm A), we
select its set of peers �rms on the year of the �rm A�s IPO (� = 0) that are not in the
set of initial peers of the �rm A, and that have been publicly listed for at least �ve
years on the IPO year. These are the initial established peers of the peer of A that
are not themselves peers of A. We label such peers of peers as B0. Among this set, we
select the three peers of peers B0 that exhibit similar levels of product di¤erentiation
with B, than B with the IPO �rm A, such that E(�B;B0) � �A;B for any pair A;B.
We then track the product di¤erentiation between the IPO �rm A and B0 over �ve
years (�B;B0;� with � = 0; :::; 5).24
We combine together the pairs made of an IPO �rms and their initial peers (A,B),
with all the counterfactual pairs (B,B0). For every actual or counterfactual pair and
event-time year, we compute di¤erentiation � and label the former set of pairs as
treated pairs, and the latter set as counterfactual pairs. To estimate the extent
to which IPO �rms change their product di¤erentiation after they become publicly
listed, we consider the following baseline linear speci�cation:
�i;j;� ;t = �i;j + �0� + �1(� � Treatedi;j;� ;t) + �t + �Xi;j;� ;t + "i;j;� ;t; (18)
where the subscripts i and j represent respectively a pair of �rms, t represents calendar
time, and � represent event time (� = 0; :::; 5). The unit of observation is at the �rm-
pair-time level. The variable Treated is an indicator variable that equals one if a pair
includes an IPO �rm and a peer (i.e., a pair (A,B)), and zero otherwise (if a pair
includes a peer of an IPO �rm and of of its peer �a pair (B,B0)). The �rm-pair �xed
e¤ects (�i;j) capture any time-invariant �rm-pair characteristics (e.g., �rms�intrinsic
24As we did for the pairs of �rms (A,B), if a peer of peer B0leaves the set of initial peers (i.e. is
no longer in �rm B�s TNIC network) in a given year � (where � > 0) , we set �B;B0;� equals to one.
27
gain from di¤erentiation �) and the (calendar) time �xed e¤ects (�t) absorbs any
common time-speci�c factor such as IPO booms or waves of product di¤erentitation.25
The vectorX includes several time-varying �rm-pair characteristics (e.g. di¤erence in
total assets between �rms i and j). We allow the error term ("i;j;� ;t) to be correlated
within pairs and we correct the standard errors as in Petersen (2009).
In estimating equation (18), we are interested in the coe¢ cient �1. Indeed, �0
measures the average within-pair change of di¤erentiation for all pairs over time, and
�0 + �1 measures the average within-pair change of di¤erentiation for treated pairs.
In the spirit of a di¤erence-in-di¤erences estimation, �1 measures the average relative
change of di¤erentiation between an IPO �rm and an established peer, compared
to a change occuring in similar counterfactual pair. As explained in Section 2.3, it
should be interpreted as a measure of the di¤erence in the extent to which a �rm is
di¤erentiated when it is public and when it is private (�A(�A)). Corollary 2 implies
that this di¤erence should be positive on average, that is, �1 > 0.
[Insert Table 1 about Here]
Table 1 presents descriptive statistics for the sample we use for the estimations.
Panel A indicates that for � = 0 the sample comprises 1,231 distinct IPO �rms (A),
2,678 distinct established peers (B), and 2,961 distinct peers of peers (B0). The
average (public) age of peers and peers of peers is 13.535 and 14.304, respectively.
Unsurprisingly, IPO �rms are smaller than their established peers, and have higher
market-to-book ratio. Peers and peers of peers are overall similar in term of age, size,
and market-to-book ratio. The average degree of product di¤erentiation (�i;j) and
return co-movement (�i;j) are roughly similar across the three sets of �rms at � = 0
(by construction). Panels B and C report pair level information for � = 0 and pair-
year level information across � = 0; :::; 5. There are 122,195 pairs (633,745 pair-year
observations) in the sample, separated into 31,427 treated pairs (139,101 pair-years)
and 90,768 counterfactual pairs (494,644 pair-years). On average, we observe a given
25Note the the inclusion of �rm pair �xed e¤ects instead of �rm i �xed e¤ects deliver (mechanically)the same results.
28
(treated or counterfactual) pair for 4.55 years post IPO.
4 Empirical Findings
4.1 The conformity-decreasing e¤ect of IPOs
Table 2 presents estimates for various speci�cations of Equation (18). The �rst
colum reports a baseline speci�cation that includes only the event time variable �
as explanatory variable, together with �rm-pair and calendar year �xed e¤ects. The
coe¢ cient on � is positive and signi�cant, indicating that the average degree of prod-
uct di¤erentiation in a given �rm pair (�i;j) increases over time for all (treated and
counterfactual) pairs. The economic magnitude of our estimate is non-trivial: The
average within-pair level of di¤erentiation increases by 0.147% per year, equivalent
to a 0.735% increase over our �ve years window.26
[Insert Table 2 about Here]
More important for our purpose, we observe in column (2) that the coe¢ cient
on the interaction term � � Treated is also positive and signi�cant (0.026 with at-statistic of 4.732). IPO �rms appear to di¤erentiate signi�cantly more from their
initial product market peers over time compared to counterfactual (established) �rms.
The point estimate indicates that the increase of di¤erentiation for newly public
�rms with their initial peers is about 20% larger than that oberved for counterfactual
pairs. This �nding is new, and consistent with the model�s main prediction that
(newly public) �rms have higher likelihood to reduce conformity and select the unique
strategy after going public as they can bene�t from an additional informative signal
coming from their own stock price. Of course, as predicted by the model, not all
�rms should shift their strategy when becoming public. Nevertherless, our estimates
suggest an average positive shift towards less conformity in our sample of 1,231 IPO
�rms.26Because we evaluate changes in di¤erentiation among TNIC pairs that are by construction the
closest pairs in terms of product o¤erings, our estimates likely represents a lower bound.
29
In the remaining columns of Table 2, we check the robutness of our �ndings to
changes in the baseline speci�cation. In column (3), we control for di¤erences in
size, age, and market-to-book ratios in each �rm-pair and observe similar results. In
columns (4) we constrain the sample to include only �rm-pairs for which we have
non-missing observations for at least three years in the post-IPO period. In columns
(5) and (6) we alter the construction of counterfactual pairs by taking �ve matches
instead of three in column (5), and by matching on size di¤erence instead of product
di¤erentiation in column (6). Our conclusion remains unchanged.
[Insert Figure 2 about Here]
Figure 2 displays the pattern of di¤erentiation for treated and counterfactual �rm-
pairs in event time. We construct this �gure by replacing the event-time variable �
in Equation (18) by a set of event-time dummy variables (D� ) and their interaction
with Treated.27 In line with the results presented in Table 2, Figure 1 con�rms the
larger increase in product di¤erentiation among treated pairs. For each � , we observe
that �1;� > 0. Moreover, the e¤ect of IPO on di¤erentiation appears to be increasing
over time, as the gap between treated and counterfactual pairs is widening over time
(�1;� ).
[Insert Table 3 about Here]
Table 3 presents the results when we measure di¤erentiation using return comove-
ment among �rm-pairs (�i;j) instead of product di¤erentiation (�i;j). The results are
largely similar. Column (1) reveals a negative coe¢ cient on � indicating an overall
decrease in return co-movement between �rm-pairs over time. In line with a move
towards the unique strategy post IPO, we observe in column (2) a negative and sig-
ni�cant coe¢ cient on � � Treated. The stock return of newly public �rms becomesless correlated with that of their initial peers over time after they become publicly
listed. The remaining columns of Table 3 con�rm the robustness of this result.
27Speci�cally, we estimte �i;j;�;t = �i;j+5P
�=0�0;�Di;j;�+
5P�=0
�1;� (Di;j;��Treatedi;j;� )+�t+"i;j;�;t
30
4.2 Cross-sectional contrasts
Our model shows that �rms face a trade-o¤ between gaining a competitive advantage
through di¤erentiation and learning information about their strategic choices from
the stock market. This trade-o¤ reduces �rms�incentive to di¤erentiate. This logic
implies that �rms should choose to di¤erentiate more when they rely less on stock
market information because their managers are better privately informed ( high) or
because stock prices are not very informative about the common strategy ( �(n; �c)
low). This suggests that going public �rms should di¤erentiate their product relatively
more after their IPOs when (i) their manager are better informed or (ii) their peer
stock prices are less informative.
To examine how the observed changes in product di¤erention post-IPO varies
with proxies for (1) managerial information, and (2) informed trading, we augment
the baseline speci�cation as follows:
�i;j;� ;t = �0� + �1(� � Treatedi;j;� ;t) + �2(� � Treatedi;j;� ;t � �i;j) + ::: (19)
where �i;j represents a proxy for one of the model�s parameters ( or �(n; �c)). Equa-
tion (19) enables us to decompose the change in di¤erentiation for IPO �rms into an
unconditional component and a component that depends linearly on the interacted
variable of interest (�i;j).
4.2.1 Private information of managers ( )
We �rst consider proxies the private information of the managers. Following Chen,
Goldstein, and Jiang (2007) and Foucault and Frésard (2014) we use the trading
activity of �rms�insiders and the pro�tability of their trades. We posit that managers
should be more likely to trade their own stock and make pro�t on these trades if they
possess more private information, if is larger.28
We measure the trading activity (Insideri) of IPO �rm i�s insiders in year t
as the number of shares traded (buys and sells) by its insiders during that year
28For instance, Fahlenbrach and Stulz (2009) report that large increases in insider ownership isassociated with increase in �rm value.
31
divided by the total number of shares traded for stock i in year t. The pro�tability
of insiders�trades in �rm i in year t (InsiderARi) is measured by the average one
month market-adjusted returns of holding the same position as insiders for each
insider�s transaction.29 We then aggregate each variable by taking its average value
over the 5-year period following the IPO. We obtain corporate insiders�trades from
the Thomson Financial Insider Trading database.30 As in other studies (e.g., Beneish
and Vargus (2002), or Peress (2010)), we restrict our attention to open market stock
transactions initiated by the top �ve executives (CEO, CFO, COO, President, and
Chairman of the Board).31 Finally, we use CRSP to compute the total number
of shares traded (turnover) in each stock and market-adjusted returns on insiders�
positions.
[Insert Table 4 about Here]
Table 4 presents estimates for equation (19) when �i = Insideri and �i =
InsiderARi. In columns (1) and (2), we �nd that the coe¢ cient (�2) on the triple
interaction between � , Treated, and the proxies for is positive and signi�cant at
the 5%-level, while the coe¢ cient on � � Treated remains positive and signi�cant.The increase in product di¤erentiation (�i;j) following the IPO is thus statistically
larger when managers are more informed. We �nd a similar pattern when we focus
on return co-movement (�i;j) in columns (3) and (4). The increase in di¤erentiation
post-IPO (i.e., the decrease of �i;j) increases signi�cantly with the intensity of insider
trading (but not with the pro�tability of insiders�trades). The �nding reported in
Table 4 are in line with the model, because informed managers are less sensitive to
29The average value of InsiderAR is 0:75% in our sample of IPO �rms and is signi�cantly di¤erentfrom zero at the 1%-level. This �nding supports the notion that insiders have private informationand is in line with �ndings on the pro�tability of insiders�trades in the literature (for recent evidence,see Seyhun (1998), or Ravina and Sapienza (2010)).30This database contains all insider trades reported the the SEC. Corporate insiders include those
who have "access to non-public, material, insider information" and required to �le SEC forms 3, 4,and 5 when they trade in their �rms stock.31Arguably owners typically sells large stakes upon public listing (e.g. Helwege, Pirinsky and Stulz
(2007)). This could potentially jeopardize the value of using insider trade to measure managers�information. To adress this issue we separately consider the number shares bought or sold dividedby the total number of shares traded instead of all transaction. Interestingly our results only holdfor buys, but not for sells.
32
the information contained in the stock prices of peers. As a result they are more likely
to switch to the unique strategy after the IPO when they receive some information
from their own stock price (�u > 0).32
4.2.2 Price informativeness of established �rms (�(n; �c))
Next, we examine how the change in product di¤erentiation of newly-public �rms
varies with measures of the informativeness of established peers� stock prices. We
use three variable as proxies for �(n; �c). First, we follow Chen, Goldstein and Jiang
(2007) and consider the probability of informed trading, PIN , measure developed by
Easley, Kiefer, and O�Hara (1996, 1997) and Easley, Hvidkjaer, and O�hara (2002).
The PIN measure is based on a structural market microstructure model in which
trades can come from noise traders or from informed traders. It measures the prob-
abilty that trading in a given stock comes from informed traders. Similar to Chen,
Goldstein, and Jiang (2007), because PIN provides a direct estimate of the probabil-
ity of informed trading, we posit that it is positively linked to the amount of private
information re�ected in stock prices. We use data on the (annual) PIN measure
computed using the Venter and De Jongh�s (2004) method. The data is provided and
discussed by Brown and Hillegeist (2007) and covers the period 1996-2010.33
Our second proxy for the informativeness of prices relies on the sensitivity of stock
prices on earnings news using the �earnings reponse coe¢ cients�or ERC (e.g. Ball
and Brown (1968)). Following the accounting literature, we conjecture that infor-
mative stock prices should better and more timely incorporate relevant information
about future earnings. Hence, earnings news of �rms with informative prices should
trigger a lower price response (lower ERC). We compute ERC for each �rm-year as
the average of the three-day window absolute market-adjusted stock returns over the
four quarterly announ cement periods.
Our third proxy focuses on the coverage from professional �nancial analysts re-
32As noted earlier, if managers are perfecly informed or not not rely on the stock market ( = 1),they should always choose the unique strategy. In this case, going public should have no e¤ect ontheir strategic choices.33See http://scholar.rhsmith.umd.edu/sbrown/pin-data
33
ceived by peer �rms. More coverage is typically associated with improved informa-
tional environment. Indeed, analysts gather information about the value of business
strategies, and provide value-adding information (e.g. Healy and Palepu (2001)) that
has a signi�cant e¤ect on stock prices (e.g. Womack (1996) and Barber, Lehavy, Mc-
Nichols, and Trueman (2006)). Hence, we argue that �rms covered by more analysts
bene�t from more information about the value of their strategy.
[Insert Table 5 about Here]
Table 5 presents estimates for equation (19) when �i;j = PINj, �i;j = ERCj, and
�i;j = Coveragej. We aggregate each proxy by taking its average value over the 5-year
period following the IPO (of �rm i). Columns (2) to (3) present results for our measure
of product di¤erentiation (�i;j). Across the three proxies �i;j, we �nd that the increase
in product di¤erentiation for IPO (relative to counterfactual pairs) is signi�cantly
smaller when the stock price of peer �rms is more informative. The same results
emerge when we look at stock return co-movement. With the exception of analyst
coverage, we observe a smaller decrease in �i;j when the established of newly-public
�rms have more informative prices. These cross-sectional results con�rm the model�s
prediction, according to which the bene�t of choosing the unique strategy is lower
when the ability to obtain information about the common strategy is higher. Our
results indicate that higher price informativeness of peer �rms accentuate conformism
in choosing strategies.
5 Conclusion
34
Appendix
Proof of Proposition 1: Directly follows from the argument in the text.
Proof of Lemma 1: We show that the strategies described in Lemma 1 form an
equilibrium.
Speculators�strategy. Let �j(xij; bsi(Sj)) be the expected pro�t of a speculatorwho trades xij 2 f�1; 0;+1g shares of �rm j when his signal about the strategy of
�rm j is bsi(Sj). First, consider an informed speculator who observes that the typeof the common strategy is good, i.e., tSc = G. If he buys the stock of an established
�rm j, the speculator�s expected pro�t is:
�j(+1; G) = E(r(Sc; en;G)� epj2 jbsi(Sc) = G) ; for j 2 f1; :::; ng:where en = n if �rm A abandons the common strategy and en = n + 1 if �rm A
implements it. The speculator expects all other informed speculators to buy one
share of all established �rms and �rm A in equilibrium. Thus, since each speculator
is in�nitesimal, he expects the total demand for each stock to be �c. This implies that
the likelihood that the demand for one stock is less than �1 + �c is zero conditionalon bsi(Sc) = G (Pr([j=n+1j=1 ffjg < �1 + �c jbsi(Sc) = G) = 0). This also implies thatPr(�1 + �c < \j=n+1j=1 ffjg < 1� �c jbsi(Sc) = G) = 1� �(n; �c).Thus, given that bsi(Sc) = G, the speculator expects the price of stock j to be
either r(Sc; n + 1; G) if the demand for one stock (including �rm A) exceeds 1 � �cor (1� =2)r(Sc; n)+ r(Sc; n+1; G)=2 if the demand for all stocks is in the interval[1 + �c; 1� �c]. In the �rst case, the manager of �rm A will implement his strategy
and the payo¤ of �rm j is r(Sc; n+1; G). Hence, the speculator earns a zero expected
pro�t since he trades at a price that is just equal to the asset payo¤. In the second
case, the manager of �rm A implements his strategy i¤ only if he learns privately that
the strategy is good, which happens with probability since tSc = G. Hence, in this
case, the speculator earns an expected pro�t of r(Sc; n+1; G)+ (1� )r(Sc; n;G)�
35
(1� =2)r(Sc; n)� r(Sc; n+ 1)=2 = �c� 2(r(Sc; n)� r(Sc; n+ 1)). We deduce that
�j(+1; G) = E(r(Sc; en;G)�epj1 jbsi(Sc) = G) = (1� �(n; �c))2
�(2�c� (r(Sc; n)�r(Sc; n+1))) > 0;(20)
where the last inequality follows from Assumption A.3. If instead, the speculator sells
the asset, his expected pro�t is �j(�1; G) =E(�(r(Sc; en;G) � epj1) jbsi(Sc) = G) =��j(+1; G) < 0, again using the fact that the speculator�s order is in�nitesimal andtherefore cannot a¤ect the total demand for the stock. Thus, the speculator opti-
mally buys stock j when he knows that the common strategy is good. A symmetric
reasoning shows that the speculator optimally sells stock j when he knows that the
common strategy is bad. Similarly, we can proceed in the same way to show that a
speculator optimally buys stock A if he knows that the common strategy is good and
sells it if he knows that the common strategy is bad.
The manager�s decision at date 3. Now consider the investment policy for the
manager of �rm A given equilibrium prices. If the manager receives managerial in-
formation, he just follows his signal since this signal is perfect. Hence, in this case,
he implements the common strategy if sm = G and he does not if sm = B. If he
receives no managerial information (sm = ?), the manager relies on stock prices. If
he observes that pA2 = pHA , the manager deduces that fj > 1��c for at least one �rmand infers that tSc = G. The reason is that fj > 1� �c if and only if speculators buystock j, i.e., if tSc = G. In this case, the manager optimally implements the common
strategy. If instead the manager observes that pA2 = pLA(Sc) then the manager de-
duces that fj < �1+�c for at least one �rm and he infers that tSc = B. In this case,the manager optimally abandons his strategy.
Finally if he observes pA2 = pMA (Sc) then the manager deduces that �1 + �c �fj � 1 � �c for all �rms (including �rm A). In this case, the stock market does not
a¤ect the manager�s priors since trades are uninformative. Hence, ifthe manager has
no private information, he does not invest since he expects the NPV of the common
36
strategy to be negative (Assumption A.2). In sum:
I�(3; Sc) =
8<:1 if sm = G;
1 if sm = ? and p�A1 = pHA ;0 otherwise,
; (21)
as claimed in the last part of Lemma 1.
Equilibrium prices. Now consider equilibrium stock prices. We must check that
equilibrium conditions (7) and (8) are satis�ed by the equilibrium prices given in the
lemma, i.e., these prices satisfy:
pA2(fA(Sc)) = E(VA3(I�(3; Sc); Sc) j 2); (22)
and
pj2(fj(Sc)) = E(r(Sc; n(Sc); tSc) j 2) for j 2 f1; ::; ng, (23)
where I�(3; Sc) is given by (21) and n(Sc) = n+1 if I� = 1 and n(Sc) = n if I� = 0.
We check that the prices given in the second and third parts of Lemma 1 satisfy these
conditions.
Suppose �rst that fj � (1� �c) for at least one �rm j. In this case, investors�netdemand in stock j reveals that the common strategy is good, i.e., tSc = G. Moreover,
according to the conjectured equilibrium, the stock price of �rm A is pHA . Hence,
I� = 1. We deduce that:
E(VA3(I�(3; Sc); Sc) j 2) = r(Sc; n+1; G)�1 if 9j 2 f1; :::; n; Ag s.t. fj � (1��c);
which is equal to pHA . Hence, if there is one �rm j such that fj � (1� �c), Condition(22) is satis�ed. Moreover, we deduce that:
E(r(Sc; n(Sc); tSc) j 2) = r(Sc; n+ 1; G) if 9j 2 f1; :::; n; Ag s.t. fj � (1� �c);
which is equal to the stock price of �rm j 6= A when 9j 2 f1; :::; n; Ag s.t. fj �(1� �c). Thus, in this case, Condition (23) is satis�ed as well.Now suppose that 9j 2 f1; :::; n; Ag s.t. fj � �(1��c). In this case, investors�net
demand in stock j reveals that the common strategy is bad, i.e., Sc = B. Moreover,
37
according to the conjectured equilibrium, the stock price of �rm A is pLA. Hence, the
manager never implements his strategy in this case (if he receives private managerial
information, he observes that Sc = B and otherwise he infers it from the stock price
of �rm A). Hence, we deduce that:
E(VA3(I�(3; Sc); Sc) j 2) = 0 if 9j 2 f1; :::; n; Ag s.t. fj � �(1� �c);
which is equal to pLA. Hence, if 9j 2 f1; :::; n; Ag s.t. fj � �(1� �c)., Condition (22)is satis�ed. Moreover, we deduce that:
E(r(Sc; n(Sc); tSc) j 2) = r(Sc; n; B) if 9j 2 f1; :::; n; Ag s.t. fj � �(1� �c);
which is equal to the stock price of �rm j 6= A when 9j 2 f1; :::; n; Ag s.t. fj �(1� �c). Thus, in this case, Condition (23) is satis�ed as well.Last, consider the case in which�1+�c � fj � 1��c for all �rms (including A). In
this case, investors�demand in each stock is uninformative about the common strategy
because Pr(Sc = G���1 + �c � \j=n+1j=1 fj � 1� �c
�= 1
2. Moreover, according to the
conjectured equilibrium, the stock price of �rm A is pMA . Hence, the manager of �rm
A will implement the common strategy (I� = 1) if and only if he receives a signal
that the common strategy is good, just as in the benchmark case. We deduce that,
E(VA3(I�(3; Sc); Sc) j 2) = V BenchmarkA1 ; if 8j; � 1 + �c � fj � 1� �c,
which is equal to pMA . Hence, if �1 + �c � fj � 1� �c for all stocks, Condition (22)is satis�ed. Moreover, we deduce that:
E(r(Sc; n(Sc); tSc) j 2) = r(Sc; n+1; G)=2+ r(Sc; n; B)=2+(1� )r(Sc; n) if 8j; �1+�c � fj � 1��c;
which is equal to (1� =2)r(Sc; n) + r(Sc; n+1)=2 As this is the stock price of �rmj 6= A when �1 + �c � fj � 1 � �c for all stocks, we obtain that Condition (23) issatis�ed as well.
Proof of Lemma 2: The proof of Lemma 2 follows the same steps as the proof of
Lemma 1 and is therefore omitted for brevity.
38
Proof of Proposition 2: Firm A will choose the unique strategy i¤ VA1(Su) >
VA1(Sc): The proposition follows by replacing VA1(Su) and VA1(Sc) by their expressions
given in (11) and (12).
Proof of Corollary 1: First consider the case in which �rm A chooses the common
strategy. Consider a speculator who buys information about the common strategy.
The expected pro�t of the speculator in �rm j 2 f1; :::; ng if he learns that thecommon strategy is good is �j(+1; G) given in eq.(20). By symmetry, the expected
pro�t of the speculator in �rm j 2 f1; :::; ng if he learns that the common strategy isgood is �j(�1; B) = �j(+1; G). Thus, the ex-ante expected gross pro�t of receivinginformation about the common strategy and trading on this information in established
�rm j is:
�(�c) =1
2�j(+1; G)+
1
2�j(�1; B) = �j(+1; G) =
(1� �(n; �c))2
(2�c� (r(Sc; n)�r(Sc; n+1));(24)
where the last equality follows from eq.(20). A speculator who is informed about
the type of the common strategy can also use his information to speculate in the
stock market of �rm A. If he learns that the common strategy is good the speculator
buys one share of stock A. The speculator can make a pro�t only if in this case the
order �ow of all �rms (including A) does not reveal that the strategy is good, that
is, if the stock price of �rm A is pMA (Sc) = r(Sc; n + 1; G)=2. This happens with
probability (1� �(n; �c). In this case, if the manager privately learns the type of thestrategy then he will implement the strategy since it is good. Otherwise the manager
abandons the strategy. Thus, in this case, the speculator expects to receive a cash
�ow of r(Sc; n+1; G) on his share of �rm A. Hence, the speculator�s expected pro�t
on buying one share of �rm A when (i) �rm A chooses the common strategy and (ii)
this strategy is good is
�A(+1; G) = (1� �(n; �c)) r(Sc; n+ 1; G)=2: (25)
A similar reasoning yields �A(�1; B) = �A(+1; G). Thus, the ex-ante expected
gross pro�t of receiving information about the common strategy and trading on this
39
information in the stock market of �rm A is:
�A(�c) =(1� �(n; �c))
2(r(Sc; n+ 1) + �c � 1).
A speculator who receives information about the common strategy can pro�tably
trade in all stocks of �rms following this strategy. Thus, his total expected pro�t is:
�(n; �c) = n�(�c) + �A(�c). (26)
In equilibrium, the demand for information about the common strategy, ��c , adjusts
such that the expected pro�t from trading on information on the common strategy
is just equal to the cost of obtaining information on this strategy, i.e., �(n; ��c) = C.
Using eq.(24), (25), and (26), this implies that:
(1� �(n; ��c)) =2C
n(2�c � (r(Sc; n)� r(Sc; n+ 1)) + (r(Sc; n+ 1) + �c � 1), (27)
when �rm A chooses the common strategy at date 1.
Now suppose that �rm A chooses the unique strategy and consider a speculator
who buys information on the type of this strategy. The ex-ante expected pro�t of this
speculator can be computed as for a speculator who buys information on the common
strategy. The only di¤erence is that the speculator can only use his information to
speculate in the stock market of �rm A (the �rm choosing the unique strategy). As we
denote the fraction of speculators who buy information about the type of the unique
strategy by �u, we deduce that the total expected pro�t of a speculator informed
about the type of the unique strategy when �rm A chooses this strategy at date 1 is:
�A(�u) =(1� �u)
2(r(Su; 1) + �u � 1): (28)
In equilibrium, the demand for information about the unique strategy, ��u, adjusts
such that the expected pro�t from trading on information on the unique strategy
is just equal to the cost of obtaining information on this strategy, i.e., ��u solves
�A(��u) = C when �rm A chooses the unique strategy. Using eq.(28),
(1� ��u) =2C
(r(Su; 1) + �u � 1). (29)
40
We have ��u � �(n; ��c) i¤ (1���u) � (1��(n; ��c)). Using eq.(27) and (29), we deducethat this is the case i¤:
(r(Su; 1)+�u�1) � n(2�c� (r(Sc; n)�r(Sc; n+1))+ (r(Sc; n+1)+�c�1): (30)
It is then immediately checked that the condition given in Corollary 1 is su¢ cient for
Condition (30) to hold true.
Proof of Proposition 3: When �rm A is private, it will choose the unique strategy
i¤ V privateA1 (Su) > VprivateA1 (Sc): The proposition follows by replacing V
privateA1 (Su) and
V privateA1 (Sc) by their expressions given in (13) and (14).
Proof of Corollary 2: Using the expressions for b�( ; �u; �c; n) and b�private( ; �c; n),we obtain that b�( ; �u; �c; n) < b�private( ; �c; n) i¤�u > (�(n;�c)��c(n�1))
+(1� )�c(n�1) . Thus under
this condition, a �rm for which b�( ; �u; �c; n) < �(n) < b�private( ; �c; n) chooses thecommon strategy if it is private but the unique strategy if it is public.
41
References
[1] Ball, R, and P. Brown, 1968. An empirical evaluation of accounting income numbers.
Journal of Accounting Research 6, 159-178.
[2] Bakke, T., Whited, T. M., 2010. Which �rms follow the market? An analysis of
corporate investment decisions. Review of Financial Studies 23, 1941�1980.
[3] Barber, B., Lehavy, R., McNichols, M., Trueman, B., 2006. Can investors pro�t from
the prophets? security analyst recommendations and stock returns?. Journal of Fi-
nance 56, 531-564.
[4] Barney, J. (1986). Strategic factor markets: expectations, luck, and business strategy.
Management Science 32, 1231�1241.
[5] Beneish, M. D., Vargus, M. E., 2002. Insider trading, earnings quality, and accrual
mispricing. The Accounting Review 77, 755�791.
[6] Bond, P., Edmans, A., Goldstein, I., 2012. The real e¤ects of �nancial markets. Annual
Review of Financial Economics 4, 339�360.
[7] Bolton, P., and D., Scharfstein, 1990. A theory of predation based on agency problems
in �nancial contracting. American Economic Review 80, 93-106.
[8] Brandenburger, A., Polak, B., 1996. Whan managers cover their posteriors:making the
decisions the market wants to see. RAND Journal of Economics 27, 523-541.
[9] Brander, J., Lewis, T., 1988. Oligopoly and �nancial structure: the limited liability
e¤ect. American Economic Review 76, 956-970.
[10] Brown, S., Hillegeist, S., 2007. How disclosure quality a¤ects the level of information
asymmetry. Review of Accounting Studies 12, 443-477.
[11] Bustamante, C., 2015. Strategic investment and industry risk dynamics. Review of
Financial Studies 28, 279-341.
42
[12] Chen, Q., Goldstein, I., Jiang, W., 2007. Price informativeness and investment sensi-
tivity to stock price. Review of Financial Studies 20, 115�147.
[13] Chod, J. and Lyandres, E., 2011. Strategic IPOs and product market competition.
Journal of Financial Economics 100, 45�67.
[14] Easley, D., Kiefer, N., O�Hara, M., Paperman, J., 1996. Liquidity, information, and
infrequently traded stocks. Journal of Finance 51, 1405�1436.
[15] Easley, D., Hvidkjaer, S., O�Hara, M., 2002. Is information risk a determinant of asset
returns?. Journal of Finance 57, 2185-2221.
[16] Edmans, A., Goldstein, I., Jiang, W., 2012. The real e¤ects of �nancial markets: The
impact of prices on takeovers. Journal of Finance 67, 933�971.
[17] Fahlenbrach, r., Stulz, R., 2009. Mangerial ownership dynamics and �rm value. Journal
of Financial Economics 97, 12-32.
[18] Foucault, T., Frésard, L., 2012. Cross-listing, investment sensitivity to stock price and
the learning hypothesis. Review of Financial Studies 25, 3305�3350.
[19] Foucault, T., Frésard, L., 2014. Learning from peers�stock prices and corporate in-
vestment. Journal of Financial Economics 111, 554�577.
[20] Healy. P., Palepu, K., 2001. Information asymmetry, corporate disclosure, and the
capital markets: a review of the empirical disclosure literature. Journal of Accounting
and Economics 31, 405-440.
[21] Helwege, J., Pirinsky, C., Stulz, R., 2007. Why do �rms become widely held? an
analysis of the dynamics of corporate ownership. Journal of Finance 62, 995-1028.
[22] Hoberg, G., Phillips, G., 2014. Product market uniqueness, organizational form and
stock market valuations. Working Paper.
[23] Hoberg, G., Phillips, G., 2015. Text-based network industries and endogenous product
di¤erentiation. Journal of Political Economy, forthcoming.
43
[24] Hou, K., Robinson, D., 2006. Industry concentration and average stock returns. Jour-
nal of Finance 61, 1927-1956.
[25] Kyle, A., 1985. Continuous auctions and insider trading. Econometrica 53, 1315�1335.
[26] Luo, Y., 2005. Do insiders learn from outsiders? Evidence from mergers and acquisi-
tions. Journal of Finance 60, 1951�1982.
[27] Maksimovic,M., 1988. Capital structure in a repeated oligopoly. Rand Journal of Eco-
nomics, 389�407.
[28] Maksimovic, M., Pichler, P., 2001. Technological innovation and initial public o¤erings,
Review of Financial Studies 14, 459-494.
[29] Porter M., 1985. Competitive Advantage. Creating and Sustaining Superior Perfor-
mance, Free Press.
[30] Petersen, M., 2009. Estimating standard errors in �nance panel data sets: comparing
approaches. Review of Financial Studies 22, 435-480.
[31] Peress, J., 2010. Product market competition, insider trading, and stock market e¢ -
ciency. Journal of Finance 65, 1�44.
[32] Ravina, E., Sapienza, P., 2010. What do independent directors know? Evidence from
their trading. Review of Financial Studies 23, 962�1003.
[33] Roll, R., 1988. R2. Journal of Finance 43, 541�566.
[34] Seyhun, H. N., 1998., Investment Intelligence from Insider Trading. MIT Press, Cam-
bridge, MA.
[35] Spiegel, M., Tookes, H., 2014. An IPO�s impact of rival �rms. Working Paper, Yale
University.
[36] Scharfstein, D. and Stein, J., 2010. Herd behavior and investment, American Economic
Review 80, 465�479.
44
[37] Titman, S., 1984. The e¤ect of capital structure on a �rm�s liquidation decisions.
Journal of Fianncial Economics 13, 137-151.
[38] Tookes, H., 2008. Information, trading, and product market interactions: Cross-
sectional implications of informed trading. Journal of Finance 63, 379�413.
[39] Venter, J., De Jongh, D., 2004. Extending the EKOP model to estimate the probability
of informed trading. Working Paper. North-West University.
[40] Womack, K., 1996. Do brokerage analysts�recommendations have investment value?.
Journal of Finance 54, 137-157.
45
Table 1: Descriptive Statistics
This table reports the summary statistics of the main variables used in the analysis. We present the number of observation and mean. All variables are defined in the text. In Panel A, we present the statistics for firm‐level observations (IPO firms, established peers of IPO firms, and established peers of peers of IPO firms). In Panel B, we present statistics for average firm‐pair level observations. Pairs that include an IPO firm are treated pairs (Treat=1) and pairs without an IPO firm are counterfactual pairs (Treat=0). In Panel C, we present statistics for average firm‐pair‐year level observations. Pairs that include an IPO firm are treated pairs (Treat=1) and pairs without an IPO firm are counterfactual pairs (Treat=0). Peers and Peers of Peers are defined using the TNIC industries developed by Hoberg and Phillips (2014). We define established peers as public firms that have been listed for more than 5 years. We track pairs over five years following each IPO, so that the event time variable τ = 0, 1, 2, 3, 4, or 5. The sample period is from 1996 to 2011.
Panel A: Firm‐level
τ=0 IPO firms Established Peers Peers of Peers
N 1,231 2,678 2,961
Age 0.000 13.535 14.304
Δi,j 0.974 0.970 0.980
βi,j 0.085 0.088 0.073
# of Peers 86.128 72.126 58.985
log(A) 4.987 5.665 5.617
MB 3.480 2.283 2.144
Panel B: Pair‐level
τ=0 Treat=1 Treat=0 all
N 31,427 90,768 122,195
Δi,j 0.967 0.976 0.973
βi,j 0.109 0.100 0.102
Agei‐Agej ‐13.164 ‐2.129 ‐4.967
log(A)i‐log(A)j ‐0.969 ‐0.079 ‐0.308
MBi‐MBj 1.089 0.214 0.439
Panel C: Pair‐year‐level
All τ Treat=1 Treat=0 all
N 139,101 494,644 633,745
Δi,j 0.972 0.981 0.979
βi,j 0.121 0.113 0.115
Agei‐Agej ‐13.265 ‐2.082 ‐4.536
log(A)i‐log(A)j ‐0.847 ‐0.047 ‐0.222
MBi‐MBj 0.426 0.109 0.178
τ 4.037 4.699 4.554
Table 2: Differentiation post‐IPO: Product Differentiation (Δi,j)
This table presents the results from estimations of equation (17). The dependent variable is the degree of product differentiation between firm i and j. The unit of observation is a firm‐pair‐year. The sample includes pairs where one firm (firm i) is an IPO firm and the other firm (firm j) is an established peer, and pairs where one firm (firm i) is a peer of an IPO firm and the other firm (firm j) is a peer of firm i than is not a peer of the IPO firm. We identify peers using the TNIC network developed by Hoberg and Phillips (2014) and defined established peers as public firms that have been listed for more than 5 years. We select peers of peers using a matching procedure as defined in Section 3. Pairs that include an IPO firm are treated pairs (Treat=1) and pairs without an IPO firm are counterfactual pairs (Treat=0). We track pairs over five years following each IPO, so that the event time variable τ = 0, 1, 2, 3, 4, or 5. The control variables include the difference in size and market‐to‐book ratio between firms in each pair. The sample period is from 1996 to 2011. The control variables are divided by their sample standard deviation to facilitate economic interpretation. Columns (1) to (3) present baseline estimations. In column (4) we constrain the sample to include only firm‐pairs for which we have non‐missing observations for at least three years. In column (5) we consider 5 matches to construct counterfactual pairs instead of 3. In column (6) we consider differences in size matches to construct counterfactual pairs instead of product differentiation. The standard errors used to compute the t‐statistics (in squared brackets) are adjusted for heteroskedasticity and within‐firm‐pair clustering. All specifications include firm‐pair and calendar year fixed effects. Symbols ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.
Dependent Variable: Product Differentiation (Δi,j)
(1) (2) (3) (4) (5) (6)
controls >3yrs M‐5 M‐size
τ 0.147*** 0.142*** 0.142*** 0.130*** 0.135*** 0.146***
[62.907] [57.389] [57.331] [50.524] [70.914] [55.948]
τ x Treat 0.027*** 0.026*** 0.014** 0.035*** 0.021***
[4.824] [4.736] [2.501] [6.461] [3.763]
log(A)i‐log(A)j ‐0.024*** ‐0.014** ‐0.020*** ‐0.070***
[‐3.718] [‐2.068] [‐3.771] [‐10.108]
MBi‐MBj ‐0.007*** ‐0.004** ‐0.007*** ‐0.002
[‐3.350] [‐2.044] [‐4.350] [‐0.825]
Pair Fixed Effects (αi,j) Yes Yes Yes Yes Yes Yes
Year Fixed Effects (δt) Yes Yes Yes Yes Yes Yes
Obs. 633,745 633,745 633,745 558,680 943,223 638,986
Adj. R2 0.729 0.729 0.729 0.712 0.714 0.770
Table 3: Differentiation post‐IPO: Return Co‐movement (βi,j)
This table presents the results from estimations of equation (17). The dependent variable is return co‐movement between firm i and j. The unit of observation is a firm‐pair‐year. The sample includes pairs where one firm (firm i) is an IPO firm and the other firm (firm j) is an established peer, and pairs where one firm (firm i) is a peer of an IPO firm and the other firm (firm j) is a peer of firm i than is not a peer of the IPO firm. We identify peers using the TNIC network developed by Hoberg and Phillips (2014) and defined established peers as public firms that have been listed for more than 5 years. We select peers of peers using a matching procedure as defined in Section 3. Pairs that include an IPO firm are treated pairs (Treat=1) and pairs without an IPO firm are counterfactual pairs (Treat=0). We track pairs over five years following each IPO, so that the event time variable τ = 0, 1, 2, 3, 4, or 5. The control variables include the difference in size and market‐to‐book ratio between firms in each pair. The sample period is from 1996 to 2011. The control variables are divided by their sample standard deviation to facilitate economic interpretation. Columns (1) to (3) present baseline estimations. In column (4) we constrain the sample to include only firm‐pairs for which we have non‐missing observations for at least three years. In column (5) we consider 5 matches to construct counterfactual pairs instead of 3. In column (6) we consider differences in size matches to construct counterfactual pairs instead of product differentiation. The standard errors used to compute the t‐statistics (in squared brackets) are adjusted for heteroskedasticity and within‐firm‐pair clustering. All specifications include firm‐pair and calendar year fixed effects. Symbols ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.
Dependent Variable: Return Comovement ()
(1) (2) (3) (4) (5) (6)
controls >3yrs M‐5 M‐size
τ ‐0.003*** ‐0.002*** ‐0.002*** ‐0.003*** ‐0.002*** ‐0.000
[‐7.223] [‐5.828] [‐5.810] [‐6.647] [‐6.609] [‐0.892]
τ x Treat ‐0.002** ‐0.001* ‐0.001** ‐0.001* ‐0.004***
[‐2.393] [‐1.654] [‐1.990] [‐1.883] [‐5.716]
log(A)i‐log(A)j ‐0.016*** ‐0.017*** ‐0.016*** ‐0.016***
[‐16.095] [‐16.553] [‐18.995] [‐15.259]
MBi‐MBj ‐0.000 ‐0.000 ‐0.000 ‐0.002***
[‐1.284] [‐1.250] [‐0.700] [‐6.656]
Pair Fixed Effects (αi,j) Yes Yes Yes Yes Yes Yes
Year Fixed Effects (δt) Yes Yes Yes Yes Yes Yes
Obs. 618,811 618,811 618,811 545,005 921,058 624,325
Adj. R2 0.261 0.261 0.261 0.246 0.259 0.262
Table 4: Differentiation post‐IPO: Managerial Information (λ)
This table presents the results from estimations of equation (18). The dependent variable is the degree of product differentiation (columns (1) and (2)) or return co‐movement (columns (3) and (4)) between firm i and j. The unit of observation is a firm‐pair‐year. The sample includes pairs where one firm (firm i) is an IPO firm and the other firm (firm j) is an established peer, and pairs where one firm (firm i) is a peer of an IPO firm and the other firm (firm j) is a peer of firm i than is not a peer of the IPO firm. We identify peers using the TNIC network developed by Hoberg and Phillips (2014) and defined established peers as public firms that have been listed for more than 5 years. We select peers of peers using a matching procedure as defined in Section 3. Pairs that include an IPO firm are treated pairs (Treat=1) and pairs without an IPO firm are counterfactual pairs (Treat=0). We track pairs over five years following each IPO, so that the event time variable τ = 0, 1, 2, 3, 4, or 5. The control variables (unreported) include the difference in size and market‐to‐book ratio between firms in each pair. The sample period is from 1996 to 2011. φ represents proxies for managerial information of the IPO firm i (Insider and InsiderAR). The proxies φ are divided by their sample standard deviation to facilitate economic interpretation. The standard errors used to compute the t‐statistics (in squared brackets) are adjusted for heteroskedasticity and within‐firm‐pair clustering. All specifications include firm‐pair and calendar year fixed effects. Symbols ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.
Dep. Variable: Produt Differentation (Δi,j) Return Co‐movement (βi,j)
φ: Insideri InsiderARi Insideri InsiderARi
(1) (2) (3) (4)
τ 0.142*** 0.142*** ‐0.002*** ‐0.002***
[57.440] [57.315] [‐5.862] [‐5.804]
τ x Treat 0.020*** 0.018*** ‐0.001 ‐0.001
[3.574] [2.799] [‐1.028] [‐0.976]
τ x Treat x φ 0.022*** 0.009** ‐0.002*** 0.000
[5.407] [2.166] [‐3.061] [‐0.803]
Control Variables Yes Yes Yes Yes
Pair Fixed Effects Yes Yes Yes Yes
Year Fixed Effects Yes Yes Yes Yes
Obs. 633,745 633,745 618,811 618,811
Adj. R2 0.729 0.729 0.261 0.261
Table 5: Differentiation post‐IPO: Peers’ Price Informativeness (πc)
This table presents the results from estimations of equation (18). The dependent variable is the degree of product differentiation (columns (1) to (3)) or return co‐movement (columns (4) to (6)) between firm i and j. The unit of observation is a firm‐pair‐year. The sample includes pairs where one firm (firm i) is an IPO firm and the other firm (firm j) is an established peer, and pairs where one firm (firm i) is a peer of an IPO firm and the other firm (firm j) is a peer of firm i than is not a peer of the IPO firm. We identify peers using the TNIC network developed by Hoberg and Phillips (2014) and defined established peers as public firms that have been listed for more than 5 years. We select peers of peers using a matching procedure as defined in Section 3. Pairs that include an IPO firm are treated pairs (Treat=1) and pairs without an IPO firm are counterfactual pairs (Treat=0). We track pairs over five years following each IPO, so that the event time variable τ = 0, 1, 2, 3, 4, or 5. The control variables (unreported) include the difference in size and market‐to‐book ratio between firms in each pair. The sample period is from 1996 to 2011. φ represents proxies for the (average) price informativeness of peer firm j (PIN, ERC, and Coverage). The proxies φ are divided by their sample standard deviation to facilitate economic interpretation. The standard errors used to compute the t‐statistics (in squared brackets) are adjusted for heteroskedasticity and within‐firm‐pair clustering. All specifications include firm‐pair and calendar year fixed effects. Symbols ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels, respectively.
Dep. Variable: Produt Differentation (Δi,j) Return Co‐movement (βi,j)
φ PINj ERCj Coveragej PINj ERCj Coveragej
(1) (2) (3) (4) (5) (6)
τ 0.146*** 0.136*** 0.139*** ‐0.004*** ‐0.003*** ‐0.003***
[56.430] [50.981] [55.625] [‐9.366] [‐6.329] [‐6.361]
τ x Treat 0.093*** ‐0.266*** 0.143*** ‐0.008** 0.006 0.004
[3.308] [‐8.322] [6.847] [‐2.240] [1.637] [1.595]
τ x Treat x φ ‐0.022** 0.073*** ‐0.061*** 0.002** ‐0.002** ‐0.003*
[‐2.306] [9.657] [‐5.394] [2.020] [‐2.086] [‐1.892]
Control Variables Yes Yes Yes Yes Yes Yes
Pair Fixed Effects Yes Yes Yes Yes Yes Yes
Year Fixed Effects Yes Yes Yes Yes Yes Yes
Obs. 577,179 569,191 633,745 573,554 567,106 618,811
Adj. R2 0.736 0.735 0.729 0.273 0.266 0.261
Figure 1: Timing of the Model
The manager of firm A
announces his
strategy:
SA{ Su ,Sc}
Su is “unique”
Sc is “common” and
followed by n firms.
Stock market
opens: trades and
stock prices are
realized.
Manager of firm A receives
managerial information and
observes stock prices. She
decides to implement his
strategy or abandons it.
Firms’ payoffs are
realized.
t=1 t=2 t=3 t=4
Figure 2: Differentiation post‐IPO: Product Differentiation (Δi,j)
This figure displays the pattern of differentiation for treated and counterfactual firm‐pairs in event time. We construct this figure by replacing the event‐time variable τ in Equation (16) by a set of event‐time dummy variables and their interaction with Treated. The solid line plots the estimated coefficients for the treated pairs (Treat=1), while the dotted line plots the estimated coefficients for the counterfactual pairs (Treat=0). The sample period is from 1996 to 2011. The specification includes firm‐pair and calendar year fixed effects.
0.2
.4.6
.8
0 1 2 3 4 5Event-time (tau)
Treat=1 Treat=0
top related