The effect of entry on R&D networks
| Published date | 01 September 2018 |
| Date | 01 September 2018 |
| DOI | http://doi.org/10.1111/1756-2171.12250 |
| Author | Emmanuel Petrakis,Nikolas Tsakas |
RAND Journal of Economics
Vol.49, No. 3, Fall 2018
pp. 706–750
The effect of entry on R&D networks
Emmanuel Petrakis∗
and
Nikolas Tsakas∗∗
We investigate the effect of potential entry on the formation and stability of R&D networks
considering farsighted firms. The presence of a potential entrant often alters the incentives of
incumbents to collaborate. Incumbent firms may form an otherwise undesirable collaboration to
deter entry of a new firm. Moreover, an incumbent may refrain from establishing an otherwise
desirablecollaboration, expecting to form a more profitablelink with the entrant. Finally,potential
entry may lead an inefficient incumbent to exit the market. Welfare analysis shows market and
social incentives to be often misaligned. We propose a subsidy scheme that encourages welfare-
improving entry.
1. Introduction
Motivation. R&D partnerships are the different modes of interfirm collaborations in which
several independent economic agents agree to share part of their R&D activities and outcomes
(Hagedoorn, 2002). The collection of R&D partnerships between firms that operate in the same or
in related industries forms an R&D network. R&D networks haveattracted attention recently due
to the substantial increase in the number of interfirm collaborations. Such collaborations arise in
most industries, but are more common in industries with rapid technological development (K¨
onig,
2013), where the importance of R&D is vital, as, for example, in pharmaceutical (Roijakkers and
Hagedoorn, 2006), computer, and military industries.
In addition, there is strong evidence of substantial entry and exit rates in the R&D networks
of several manufacturing sectors (Gulati, Sytch, and Tatarynowicz, 2012; K¨
onig and Rogers,
2016; Tomasello et al., 2016).1Finally, despite the fact that in general it is difficult to identify
∗University of Crete; petrakis@uoc.gr.
∗∗University of Cyprus; tsakas.nikolaos@ucy.ac.cy.
We are indebted to the Editor and two anonymous referees, as well as Sanjeev Goyal, David Minarsch, and Jos´
eLuis
Moraga-Gonzalez, Hans Frankort, and the seminar participants at University of Crete, Luxembourg University, ASSET
2015 in Granada, Oligo Workshop 2016 in Paris, CRETE 2016 in Tinos,GAMES 2016 in Maastricht, and 2016 GAEL
Conference in Grenoble for useful comments and suggestions. All remaining errors are our sole responsibility. Part of
this project was carried out while Tsakas was at SUTD–MIT International Design, Center at Singapore University of
Technologyand Design supported by grant no. IDG31300110.
1Tomasello et al. (2016), using a large database of publicly announced R&D alliances, investigate empirically
the evolution of R&D networks and the process of alliance formation in several manufacturing sectors overthe period
1986–2009, and identify a rise-and-fall dynamics property of R&D networks, with a peak in the mid-1990s. K¨
onig and
706 C2018,The RAND Cor poration.
PETRAKIS AND TSAKAS / 707
potential entrants in a given industry, the literature reports a few cases in which R&D links have
explicitly been formed in order to deter entry of potential competitors. For instance, “Intel and TI
(TexasInstruments) agreed to swap designs and manufacturing technologies and to act as alternate
chip suppliers for each other” in order to “make it hard for small custom producers to survive in
this rapidly growing market” (Lewis, 2002). An even more sound case is the Symbian platform,
a software company founded in 1998 by Ericsson, Nokia, Motorola, Psion, Inc. and Siemens,
whose primary focus was to design and license the operating system (OS) and user interface to
the world’s leading handset markets to produce smartphones.2The main goal of Symbian was
to complicate and even block Microsoft’s entry in the smartphone and wireless markets (Smith,
2002; West and Wood, 2013; West, 2014).3
This article investigates the effect of potential entry on the formation and stability of R&D
networks, as wellas on market outcomes and social welfare. A main question addressed is whether
and how the presence of potential entrants alters the incentives of incumbent firms to form R&D
collaborations. In this way, we will get a better understanding of why certain network structures
are more likely to be sustainable than others in the long run. Moreover, we will be able to identify
novel strategies used either by incumbent firms that seek to deter entry, or by potential entrants
that seek to enforce accommodation. For incumbent firms, these strategies may be particularly
crucial in industries with a small number of firms, where entry deterrence can secure for them
larger market shares and profits. For entrant firms, these strategies may be particularly beneficial
in highly asymmetric industries, in which efficient incumbent firms may be willing to collaborate
with entrants in order to marginalize or even to induce exit of their inefficient rivals.
It is apparent that the study of the effects of potential entry on the formation of R&D networks
is of vast importance. It is a decisive stepping stone in the process of identifying those features
which guide the investmentand collaboration decisions of the fir ms. To the best of our knowledge,
this is the first article that considers how potential entry can shape the incentives for R&D collab-
orations between firms, leading them to establish links they would otherwise find unprofitable,as
well as to prevent the establishment of collaborations that would otherwise be profitable.4
Setting and results. We consider the simplest setup with potential entry that draws on
D’Aspremont and Jacquemin (1988) and is in line with K¨
onig, Liu, and Zenou (2014). There are
two incumbent firms with (possibly) different initial marginal costs and a potential entrant. Each
firm produces a brand of a horizontally differentiated good. An R&D link between any twofir ms
leads each to share a given part of its R&D outcome with its collaborator. The incumbent firms
first decide whether to establish an R&D link and then the entrant decides whether to enter or
not. If the entrant stays out, the incumbent firms choose their R&D efforts and outputs. If entry
occurs, all the firms meet sequentially in a random order in pairs and decide whether to form
R&D links or not. An R&D link is established only if both firms agree on it. This generates a
dynamic phase of discussions that ends either if a complete network has been formed, or if after
a complete round of discussions, no new R&D link has been agreed upon. After the end of these
discussions, all firms decide their R&D efforts and outputs.
Rogers (2016), considering a longer time series of data, have shown that R&D networksactually exhibit an oscillatory
pattern. In an earlier article, Gulati, Sytch, and Tatarynowicz (2012), using data from the global computer industry from
1996 to 2005, have shownthat the evolutionary dynamics of the network structure exhibit a rise-and-fall pattern.
2Although Symbian resembles a research joint venture, there is evidence that, by no means, was maximizing its
founders’ overall profits. In fact, Symbian’s eventual failure was mainly due to the conflicting interests of its founders,
each of which had several other R&D collaborations with third parties (see Cloodt, Hagedoorn, and Roijakkers, 2010).
Therefore, Symbian can be reasonably considered as a set of links or evena distinct node in a complicated R&D network.
3There are also other instances of R&D alliances established to develop newproducts and technologies that would
indirectly complicate entry by others. Moreover,airline code-sharing alliances are often used to deter entry of competitors.
These, however,are not R&D endeavors but a cooperative marketing strategy (see Goetz and Shapiro, 2012).
4K¨
onig and Rogers (2016) also allow for entry and exit, but their focus is on the study of the coevolutionary
dynamics of knowledge creation, diffusion, and the formation of R&D collaboration networks. In their article, it is high
enough subsidies to incumbent firms’ R&D linking costs that deter the entry of new firms.
C
The RAND Corporation 2018.
708 / THE RAND JOURNAL OF ECONOMICS
Evidently,a static concept such as pairwise stability is unable to capture the dynamic nature
of entry process. Wethus postulate that fir ms are farsightedwhen deciding whether to establish an
R&D link.5In particular, and in contrast with what is implied by the notion of pairwise stability,
under farsighted stability, firms are able to anticipate networks that may arise in the future as a
consequence of their current decisions. Therefore, a firm will refrain from forming a link yielding
a higher payoffunder the cur rent network,as long as it realizes that this will lead to an equilibrium
network in which it will be worse off than in the network the firms will end up if the link under
discussion is not formed. We thus introduce a dynamic process of network formation in which
each pair of firms is allowed to reconsider forming an R&D collaboration as long as a new R&D
link has been established between another pair of firms. The only restriction weimpose is that any
two firms that have established an R&D link cannot break it afterward.6Nevertheless, one should
acknowledge both that farsightedness requires firms to be highly sophisticated when making their
long-run plans and that in large markets, it may require significant amounts of information on the
side of the firms, which might be a potential burden in empirical applications.7
An important feature of farsightedness is that it can act as a coordination device for some
firms, helping them to achieve their maximum payoffs. For instance, in our setting with three
firms, if two of them obtain their maximum payoffs in the unconnected network in which they
are connected, then this is the unique equilibrium network,despite not being necessarily pairwise
stable.
In the symmetric benchmark case, all firms have initially equal marginal costs, thus, their
incentives are identical. Under no entry threat, incumbents always establish an R&D link. The
same happens under potential entry, where by establishing a link, the incumbents can deter
the entry of the new firm. This occurs when competition is intense and R&D spillovers take
intermediate values, in which case the potential entrant remains isolated in an unconnected
network and decides not to enter, even if the entry cost is nil. This result is in line with Goyal
and Moraga (2001). For lower R&D spillovers, entry is accommodated with the new firm being
isolated in the unconnected network. Finally, in the rest of the cases, entry is again accommodated
and the complete network arises.
Under asymmetric marginal costs, there are often different incentives across firms. Under
no entry threat, the inefficient incumbent always wantsto establish an R&D link, but the efficient
incumbent will agree on its establishment only if its partner-rival is not too inefficient. When the
potential entrant and the efficient incumbent are equally efficient, we identify conditions under
which the latter will opt for an R&D link with a quite inefficient incumbent—that wouldotherwise
be ignored—to deter the entry of the new firm. This occurs when the competition is fierce relative
to the benefits of R&D spillovers, and an R&D link between any two firms would force the third
one to exit the market. To avoid such a forced exit, the efficient incumbent establishes an R&D
link with its inefficient counterpart at the outset of the game. In this way, it also enjoys a larger
market share and profits in the ensuing duopoly. As a consequence, the entry of an efficient new
firm is deterred. In addition, the incentives of the efficient incumbent are altered due to potential
entry.
Further, we identify conditions under which entry is accommodated and may also force
the exit of the inefficient incumbent. When competition is fairly intense, the efficient incumbent,
anticipating entry, refrains from establishingan otherwise profitable R&D link with the inefficient
incumbent and forms instead a link with the more efficient new firm the first time they meet to
discuss. This is because a farsighted efficient incumbent correctly anticipates entry and also
5The notion of farsightedness is an adaptation of the Dutta, Ghosal, and Ray (2005)’s model, which has been
studied extensively and possesses severalinteresting proper ties (see Herings, Mauleon, and Vannetelbosch, 2009, 2010;
Mauleon, Sempere-Monerris, and Vannetelbosch, 2014).
6This is reasonable, as partnerships between firms are usually official and protected by binding contracts, hence,
breaking a link may induce large costs. Webriefly discuss the robustness of our results if this assumption is dropped.
7Such limitations could be overcomeby considering an extension of the model with farsighted firms and incomplete
information, to which the current article would act as a benchmark.
C
The RAND Corporation 2018.
To continue reading
Request your trial