ArticlePDF Available
Electronic copy available at: https://ssrn.com/abstract=2897264
The Effect of Acquisitions on Product Innovativeness, Quality, and
Sales Performance: Evidence from the Console Video Game
Industry (2002-2010)
Masakazu Ishihara
Stern School of Business
New York University
Joost Rietveld
Rotterdam School of Management
Erasmus University
January 10, 2017
We have benefited from discussions with Ron Borkovsky, Junhong Chu, J.P. Eggers, T¨ulin Erdem, Clarence Lee,
Robin Lee, Vishal Narayan, Sherif Nasser, Vithala Rao, Melissa Schilling, and Vishal Singh. We also thank seminar
and conference participants at 2014 Marketing Science Conference, NUS, 2015 IIOC, NYU, BAM, and Cornell for
their helpful comments. All remaining errors are ours. Authors contributed equally and are listed in alphabetical
order.
Masakazu Ishihara: Assistant Professor of Marketing, Stern School of Business, New York University, Email:
mishihar@stern.nyu.edu.
Joost Rietveld: Assistant Professor of Strategic Management, Rotterdam School of Management, Erasmus Uni-
versity, Email: rietveld@rsm.nl.
Electronic copy available at: https://ssrn.com/abstract=2897264
Abstract
Previous work has mostly looked at how acquisitions affect firm-level outcomes. This paper in-
vestigates how acquisitions affect product-level outcomes in the context of the console video game
industry. We model the effect of acquisitions on video games’ innovativeness, quality, and sales
performance in a Structural Equation Modeling framework. We control for endogenous partner se-
lection by implementing a two-sided matching framework. We jointly estimate these models using a
Bayesian MCMC algorithm on a dataset of 5,916 video games released in the UK from 2002 to 2010.
We find that video game publishers acquire developers for their product development capabilities
and to secure access to hit properties. Prior alliances and geographic proximity between firms also
predict acquisitions. We further find that publishers fail to leverage developers’ capabilities for on-
going innovation as product innovativeness decreased post-acquisition. Lastly, acquisitions create
value through enhanced inter-firm coordination resulting in higher quality products and increased
sales performance.
Keywords: Acquisition, Innovation, Product Quality, Performance, Two-Sided Matching Models,
Bayesian Estimation, Video Games
1 Introduction
Mergers and acquisitions continue to be an economically and scholarly relevant subject.1Growth
through acquisitions is an oft-deployed strategy for firms to remain competitive in rapidly changing
environments as well as increasing pressures for globalization. The worldwide M&A deal value
totaled US$ 4.7 trillion in 2015, the strongest reported since records began.2A wealth of research
in strategy, finance, economics, and recently marketing has been dedicated to understanding the
effects of acquisitions and whether they are, in fact, successful. Many papers have looked into the
drivers of financial outcomes (Higgins and Rodriguez 2006; King, Slotegraaf and Kesner 2008).
Some have looked at how acquisitions affect inter-firm coordination and efficiency (Cassiman et al.
2005; Liu, Lu and Qiu 2016). And another body of work studied the effects of acquisitions on firms’
innovation capability (Ahuja and Katila 2001; Prabhu, Chandy and Ellis 2005; Zhao 2009). While
these papers make important contributions to our understanding of how acquisitions create value,
they largely look at firm level outcomes. These findings offer very little insight into how mergers
and acquisitions affect the product characteristics of firms involved in a deal and how acquisitions
impact the performance of firms’ products.
Reconfigurations in firms’ product portfolios including diversification, improved quality, and
innovation are prime motives for firms to engage in acquisitive strategies (Krishnan, Joshi and
Krishnan 2004; Puranam, Singh and Zollo 2006). Products are the nexus of competition, and
innovation trajectories often result in the introduction of new and improved products to the market
(Brown and Eisenhardt 1995; Chao, Kavidas and Gaimon 2009; Hauser, Tellis and Griffin 2006).
The consequences and outcomes of acquisitions therefore are ultimately reflected in changes to
firms’ product characteristics and their market performance. Why, then, are there so few papers
studying the effects of acquisitions at the product level? Coarse measures such as abnormal stock
returns and asset growth are generally used because post-acquisition data at the product level is
unavailable or difficult to obtain. Furthermore, target firms are typically integrated into acquirers’
corporate structures which complicates identification of post-acquisition changes at the target level.
1The terms mergers and acquisitions have often been used interchangeably in extant literature. Generally speaking,
acquisitions involve takeover deals between two firms where one firm can be clearly identified as ‘buyer’ and the other
as ‘target.’ Mergers can be seen more as a marriage of equals, involving deals between two firms of somewhat similar
financial status. The focus of this study is on acquisitions of game developers by publishers of video games.
2Thompson Reuters (2015) Mergers & Acquisition Review: Financial Advisors.http://share.thomsonreuters.
com/general/PR/MA-4Q15-(E).pdf – Last accessed December, 2016.
1
Additionally, the granularity of financial reporting on measures such as product-level outcomes may
not be down to the level of individual subunits within a firm. And even if it were, pre-acquisition
benchmark data may not be readily available to determine any post-acquisition treatment effects.
Addressing this gap in the literature, in this paper we examine how acquisitions affect targets’
product-level innovativeness, quality, and sales performance. We study these effects in the context
of the console video game industry. The console video game industry is an ideal setting for our
objectives because the industry is replete with inter-firm collaborations as downstream publishers
either contract for the development of video games with independent upstream developers, produce
games internally, or acquire independent developers. We exploit a key feature of this industry for
identification: acquired developers and their products remain identifiable post-acquisition. This
allows for a granular approach towards estimating the effect of acquisitions on product-level out-
comes including how game innovativeness and game quality change, as well as if there is any impact
on games’ post-acquisition market performance. We estimate results using a unique dataset of 85
acquisitions of developers by downstream video game publishers. Our dataset includes various
product-level measures for 5,916 video games released in the UK from 2002 to 2010. We combine
firm level data from multiple secondary sources including the Securities Data Corporation (SDC)
Platinum database with a proprietary product-level dataset containing information on product
characteristics and sales performance measures. We estimate the effects of acquisitions through
Structural Equation Modeling (SEM), and we control for the acquisition process by implement-
ing a two-sided matching game (Ni and Srinivasan 2015; Park 2013; Rao, Yu and Umashankar
2016; Sørensen 2007). We jointly estimate these models by a Bayesian Markov chain Monte Carlo
(MCMC) algorithm.
Our study makes several contributions to the literature on mergers and acquisitions. First,
our study is among a select group of papers to study the impact of acquisitions on product-
level outcomes. Acquisitions affect product characteristics in a number of important ways, but
these changes are hardly reflected in short-term stock market reactions or financial accounting
measures. This is particularly the case when acquisitions involve privately held firms or when
target firms are acquired for technology-related reasons that financial markets may struggle to
evaluate appropriately. Analysis at the product level can thus provide greater insight into how
mergers and acquisitions affect the firms involved in a deal. We find that acquisitions affect target
2
firms’ product mix strategies in a number of ways: Following an acquisition, targets’ product quality
improves as video games by acquired developers are more likely to be rated favorably by expert
critics. We also find, however, that acquisitions lower targets’ product innovativeness as acquired
developers are less likely to release video games based on original intellectual property.
Second, we contribute by studying the effect of acquisitions on multiple outcome measures in a
single framework. Although review studies typically find that acquisitions create value for the firms
involved in a deal, individual evidence has been mixed with some studies suggesting that acquisi-
tions boost targets’ performance (e.g., Liu et al. 2016) and others suggesting that acquisitions are
detrimental to targets’ performance (e.g., Ravenscraft and Scherer 1989). Acquisitions are complex
features with multifarious consequences, some of which are positive (e.g., improved efficiency, com-
plementarities) and others that are negative (e.g., loss of autonomy, employee turnover). By jointly
studying the effects of acquisitions in a Structural Equation Modeling framework we can enhance
our understanding of the mechanisms behind how acquisitions create value. While we find that the
quality enhancing effect of acquisitions has a positive mediation effect on developers’ video game
sales, the reduced innovation effect has a negative mediation effect on sales performance. The over-
all effect of acquisitions on video games’ sales performance is positive suggesting that acquisitions
do indeed create value in our empirical context.
Lastly, we study acquisitions in a novel empirical setting, the console video game industry.
Acquisitions are not only popular in industries such as pharmaceutics and technology, but are
increasingly prevalent in the consolidating media and entertainment industries. It is important
to know how existing theories of acquisitions hold in creative and innovation-driven industries
such as video games. In this light, one contribution that we make is through our measure of
innovation. Studies measuring the effect of acquisitions on innovation typically look at inventor or
scientist turnover, R&D intensity, and patent related measures such as patent counts and citations.
While these are valid measures of invention, none of these outcomes can be directly linked to the
exploitation of novel ideas. Furthermore, patent applications are much less salient in the media
and entertainment industries where novelty emerges from (team-based) creative processes which are
then transformed into new products. In this study we therefore measure innovation at the product
level by determining if a video game is based on a novel intellectual property (IP) rather than based
on an existing video game IP (e.g., sequel, prequel, or spin-offs), or based on an external media
3
tie-in such as a movie or book adaptation. We find that the dominant theory, that acquisitions
stifle target-level innovation, holds even when using this unexplored measure of innovation.
The rest of the paper is organized as follows. Section 2 provides a literature review and forms
conjectures. Section 3 follows with a description of the video game industry and the data used in
our study. In Section 4 describes the two-sided matching model and a Structural Equation Model
for product-level outcomes, and explains our estimation strategy. In Section 5, subsequently, we
discuss the estimation results. In Section 6 we discuss our findings and implications for the literature
on acquisitions. And, finally, in Section 7 we offer limitations and suggestions for future research.
2 Related literature and conjectures
In this paper we study the effect of acquisitions on three product-level outcomes: innovativeness,
quality, and sales performance. We further look into firms’ strategic selection of partner firms for ac-
quisition. Taken together, these objectives are captured in the conceptual model in Figure 1, where
the arrows in the model reflect the product development process. In order to arrive at conjectures,
in this section we first review the literature on acquisitions regarding these relationships.3
2.1 The impact of acquisitions on innovation
The effect of acquisitions on innovation received ample attention in the literature. Early work by
Hitt and colleagues provided compelling arguments for why acquisitive growth reduces innovation.
At the acquirer level, Hitt, Hoskisson and Ireland (1990) argue that acquisitions can serve as a
substitution for internal innovation and that R&D investments can stifle as the acquirer focuses on
short-term profits to pay for the costs related to acquisition. Furthermore, acquisitions increase the
level of diversification in the organization which pushes managers to rely more on financial controls
(in lieu of strategic controls) and increase the level of bureaucracy. In subsequent work, Hitt et
al. (1991) empirically validated these arguments in a longitudinal sample of 191 US acquisitions as
acquisitions were found to negatively impact R&D intensity and patent intensity.
In order to maintain competitive parity, acquiring firms often turn to innovative target firms
that have successfully developed new products that can be incorporated into and exploited by the
3Since the main focus of our study is how acquisitions affect innovativeness, quality and performance (and not about
how innovation affects quality, or how innovation and quality affect performance), we do not offer any conjectures
about the effect of innovation on quality or the effects of innovation and quality on performance.
4
acquiring firm (Hitt et al. 1996; Schweizer 2005). Acquiring firms, however, often fail to leverage
target firms’ ongoing innovation capabilities. Post-acquisition, targets become part of a common
control system that allows the acquirer to leverage the existing knowledge that resides within the
target firm (Gulati, Lawrence and Puranam 2005). Nevertheless, the acquirer typically struggles
to leverage the target’s capabilities for ongoing innovation as the acquisition ends the autonomous
existence of the acquired firm (Puranam and Srikanth 2007). Following an acquisition, the target
firm’s routines may change to improve coordination with the acquirer, which can undermine the
target’s innovative capabilities (Benner and Tushman 2003; Leonard-Barton 1992). Furthermore,
loss of autonomy can also lead to a decline of target employees’ intrinsic motivation and may trigger
a potential increase in key employee turnover, further hampering the target’s innovative capabilities
(Ernst and Vitt 2000; Osterloh and Frey 2000).
While there have not been many studies to investigate the effect of acquisitions at the target
level, the ones that have largely support this view. In their study of 207 acquisitions in the
information technology hardware industries, Puranam et al. (2006) found a strong and negative
main effect of integrative acquisitions on targets’ likelihood of introducing new products. In their
study of 304 acquisitions of small- and medium-sized German firms by foreign acquirers, Stiebale
and Reize (2011) similarly found that acquisitions had a strong and negative impact on targets’
post-acquisition R&D activities. In a study of 3,933 inventors in the pharmaceutical industry,
Paruchuri, Nerkar and Hambrick (2006) found that the patent application count of scientists in
acquired firms declined following an acquisition. This finding was echoed in a study by Kapoor and
Lim (2007) who found that the number of successful patent applications by acquired inventors in
the semiconductor industry was significantly lower in the five year period following an acquisition
compared to a matched sample of non-acquired inventors.
2.2 The impact of acquisitions on quality
While clearly an important topic, the effect of acquisitions on quality has received considerably less
attention in the literature. Existing research has argued for two competing effects of how acquisi-
tions affect quality. On one hand, acquisitions can improve quality through enhanced coordination
in production processes, increased efficiencies, and the pooling of knowledge bases between the
target firm and the acquirer (Cassiman et al. 2005), as well as through the elimination of a hold-up
problem (Klein, Crawford and Alchian 1978; Williamson 1979). On the other hand, acquisitions
5
can incentivize firms to reduce their quality provisions when the merging of a target and an acquirer
firm reduces competitive pressures in the marketplace (e.g., Masta 2011). Previous research in the
US airline industry found that average product quality dropped marginally following a merger of
two competing airlines, but reverted back to its initial levels and sometimes even improved in the
long run (Prince and Simon 2015). A related paper found that the magnitude of the change in
post-acquisition quality depended on the intensity of pre-acquisition competition between merging
airline companies (Chen and Gayle 2013). Fan’s (2013) study in the newspaper industry found
more conclusive evidence that newspaper content quality dropped following a decrease in compet-
itive intensity as a result of consolidation. In line with the coordination argument, Sheen (2014)
found that the average quality of merging firms in consumer product markets did not change, but
became more similar post-acquisition. These sparse and mixed results suggest that more research
is needed on the effect of acquisitions on (product) quality.
2.3 The impact of acquisitions on performance
A vast and multidisciplinary body of work has addressed the effects of acquisitions on performance
(see reviews by Datta, Pinches and Narayanan 1992; King et al. 2004; and Lubatkin 1983). These
studies conclude that the average overall returns to firms involved in acquisitions are positive, but
that the value created from these deals is disproportionately skewed in favor of target firms. Early
empirical work by Chatterjee (1986) and Singh and Montgomery (1987) echo these conclusions as
both studies find that the value of target firms’ stock prices increased significantly in the period
following the announcement.
The dominant logic for why acquisitions create value hearkens back to synergy from pooling
complementary resources between the target and the acquiring firm (Barney 1988; Harrison et al.
1991; 2001). Resources are complementary when the value created by bundling them exceeds the
value created from deploying those resources in isolation. For example, when a target firm with
strong product development capabilities is acquired by a firm that is highly capable in complemen-
tary domains such as marketing, access to distribution channels, or ties with external actors, the
appropriability conditions and thus the value capturing potential of the target firm’s innovations will
increase post-acquisition. Subsequent research provided empirical support for the complementar-
ity hypothesis, showing that abnormal returns originated from combining complementary resource
profiles in the form of marketing and technology resources (King et al. 2008; Yu, Umashankar and
6
Rao 2015).
An important limitation of this body of work, however, is that in almost all cases performance
is operationalized as returns to stock prices or other financial book measures measured at the level
of the acquiring firm. While certainly informative, these measures are rather coarse and apprehend
a nuanced interpretation of the results. A few notable exceptions exist of studies that quantify the
performance impact of acquisitions at the target level. A study by Ravenscraft and Scherer (1989)
on 251 US acquirees found that targets’ post-merger profitability declined. More attuned with
existing theory, a study by Krishnan et al. (2004) found that merging hospitals’ profits increased
following a merger. Similarly, a study by Liu et al. (2016) found that the performance of 775
Chinese target firms improved following acquisition by a non-Chinese acquirer.
2.4 Strategic factors in target selection
From the discussion above it follows that, in order for firms to create value from acquisitions, there
must exist heterogeneities in the resource allocations between firms within the pool of potential
targets as well as within the pool of potential acquirers. As a result, there will exist variation in the
‘match value’ of pairs of potential targets and acquirers. In a given market, certain target-acquirer
pairs can achieve greater synergy than others.
This notion led some to study (but rarely control for) firms’ strategic selection of partners for
acquisition. Firm characteristics such as the number of past acquisitions (Park 2013), reputation in
the product market (Saxton and Dollinger 2004), firm innovativeness (Higgins and Rodriguez 2006;
Zhao 2009), and financial status (Park 2013; Saxton and Dollinger 2004), are important factors for
predicting acquisition activity. Wang and Zajac (2007) suggest taking a dyadic perspective and look
at the factors that motivate a pair of firms to engage in acquisition. Synergy arising from resource
complementarities can only be evaluated by looking at both firms’ resource allocations (Rao et al.
2016). Some studies found that complementarities in the product market increase the likelihood of
acquisition (Makri, Hitt and Lane 2010; Yu et al. 2015). One important factor in gauging potential
sources of complementarities is acquirers’ access to (private) information. Several studies have
found that previous alliances between the acquirer and target increases the likelihood of acquisition
(Higgins and Rodriguez 2006; Porrini 2004; Yu et al. 2015). Conversely, other research found
that geographic distance between target and acquirer firms is inversely correlated with acquisition
incidence (Chakrabarti and Mitchell 2013).
7
2.5 Synthesis
Past literature studied the effect of acquisitions on a number of important outcomes including inno-
vation, quality, and performance. This yielded a number of theoretical predictions: (1) acquisitions
have a negative effect on innovation; (2) acquisitions have mixed effects on post-acquisition quality;
and (3) acquisitions have a positive effect on target-level performance. Since the dominant logic
for why acquisitions create value resides in synergy from bundling complementary resources, past
studies also pointed to the importance of strategic partner selection. In this light, firms’ access to
private information (e.g., in the form of prior alliances, geographical proximity) plays an important
role in predicting why firms engage in acquisitions.
3 Data and methods
3.1 Empirical setting
We study the effect of acquisitions in the context of the console video game industry. In the
video game industry, publishers and developers work together to develop and commercialize video
games. Publishers are responsible for IP management, financing the development of video games,
as well as executing on marketing activities including advertising, distribution, and pricing. Game
developers are specialized in game development and undertake game design, art, programming, and
testing. Developers can be either independent, in which case they develop games potentially for any
publisher, or internal, in which case they develop games only for their parent publishers. Internal
developers can further be classified into internally founded developers and acquired developers. Our
focus in this study is on the difference between independent developers and acquired developers.4
The video game industry offers an ideal setting for studying the impact of acquisitions on
product characteristics. First, there are a large number of firms in the industry, and acquisitions
are commonly observed. Since the mid-nineties, the industry has witnessed a trend of consolidation
with publishers acquiring independent developers to increase scale and secure access to valuable IP
(Johns 2006). During our sample period we observe 85 acquisitions of independent developers by
publishers (see Appendix A for an overview of deals). Figure 2 shows the number of acquisitions over
4In this study we do not look at the effect of vertical integration on platform-level outcomes, which is the main
focus of Lee’s (2013) paper. We acknowledge that the market for console video games is a two-sided market, and as
such we control for platform level factors that are likely to affect games’ characteristics and performance.
8
time, as well as the number of active publishers and the number of active independent developers.5
The numbers of publishers and independent developers are stable over time, fluctuating around 80
publishers and 550 independent developers. The number of acquisitions shows an increasing pattern
in 2004 and 2005. We conjecture that this spike in acquisition activity is due to the introduction
of next generation consoles (i.e. Nintendo Wii, Microsoft Xbox 360, and SONY PlayStation 3) in
2005 and 2006. While this is an interesting pattern, we do not further investigate it in this study,
but we control for it by including year dummies.
A second major advantage of the video game industry is that acquired developers typically
maintain their pre-acquisition identities after they are acquired.6For example, after publisher
Activision acquired developer Treyarch for nearly US$ 20 million in 2001, the studio kept its name
and continues to be credited for all the games (e.g., Call of Duty 2: Big Red One) it develops.
This unique feature of our data enables us to conduct our analysis at the target-level, instead of
the acquiring firm.
Finally, the video game industry offers an opportunity to study the impact of acquisitions at
the product level. Our data allow us to quantify not only the direct impact of acquisitions, but also
the indirect impact on sales performance through changes in product innovativeness and quality.
3.2 Data
We compile a unique dataset from five different sources. First, similar to many other papers we
collected data on acquisitions from the SDC Platinum database. From this dataset we collected data
on 64 acquisitions. While the SDC Platinum database is known to be among the most complete
databases for inter-firm collaborations, such databases tend to under-report less visible deals, such
as those involving private companies (Schilling 2009). Second, since the majority of firms in the
video game industry are private companies, we complemented the SDC data with searches on
secondary sources such as Mobygames.com, firm websites and other online sources. This additional
search yielded information on another 21 acquisitions. Third, data on video game titles, sales,
distribution, developer and publisher, genre, age rating, and release date come from a proprietary
dataset provided by one of the platform owners in the UK. The sales data account for 90% of all
5For all publishers and developers in our sample, we collected data on their founding and closing years. We then
use this information to determine whether they were active in any given year.
6In some rare cases, the developer is rebranded post-acquisition and adopts a name that corresponds with the
acquiring entity. For example, when the now defunct publisher THQ acquired developer Rainbow Studios the firm
was renamed into THQ Digital Phoenix. We control for these instances by identifying such rebrands.
9
UK retail transactions (online retailers included) and range from 2000 to 2011. We excluded games
released in 2000 and 2001 to facilitate firm level measures based on two year rolling windows. We
also excluded games released in 2011 to allow for product lifecycles of at least one year. Fourth,
data on video game innovativeness and other firm and product characteristics were hand-collected
by three research assistants, one of which with working experience in the industry.7Fifth, data on
video game quality come from Metacritic.com and were collected using a web-scraper. Metacritic
collects and aggregates expert scores from over 300 online and offline publications which it then
transforms into a weighted “Metascore” ranging from 0-100. Our final sample for analysis includes
5,916 video games of which 972 are developed by acquired developers.
3.3 Variable definitions
Our first set of analyses (predicting who acquires whom) is at the firm level, while the second set
of analyses (the effect of acquisition on game outcomes) is at the product level. We use different
predictor variables in these sets of analyses that we discuss in detail below. Table 1 summarizes
the variables used in our analyses.
3.3.1 Outcome variables
We measure acquisition as a dummy variable that takes the value of 1 if a publisher acquired a
developer. The event we code for is the year when a deal is reported as completed. At the game level,
we measure product innovativeness by whether a game introduces a new IP to the market. Video
game publishers, like other media and entertainment firms, see creating new IPs as the fundamental
innovation activity in their industry: Like other entertainment companies, our business is based
on the creation, acquisition, exploitation, and protection of intellectual property (IP).”8Following
Tschang (2007) and Rietveld and Eggers (2016), we therefore operationalize product innovation as
games that are based on original intellectual property rather than existing video game properties
such as sequels, sports licenses, or non-video game media tie-ins (movie adaptations or based on
books etc.). We measure video game quality in terms of games’ expert review scores as reported
on Metacritic. We choose Metascores as a proxy for game quality mainly for three reasons. First,
7We included some overlap between RA’s to assess the validity of the data and conclude that inter-rater reliability
measures are at sufficient levels (κ= 0.64) (Fleiss 1971; Landis and Koch 1977).
8Electronic Arts (2006) Proxy Statement and Annual Report.http://files.shareholder.com/downloads/ERTS/
3538257882x0x203378/25B063C8-94BC-4857-B821- 40E2403C9829/2006%20Annual%20Report%20and%20Proxy%
20Statement.pdf – Last accessed January, 2017.
10
unlike user reviews, expert reviews are less subject to selection bias.9Second, it is well known
that publishers regard Metascores as a proxy for game quality and even use them for compensation
schemes with independent developers.10 Third, the majority of expert reviews are written and
published prior to the launch of a focal game. Since the correlation between Metascores and sales
performance is likely non-monotonic and characterized by strong cut-offs, we operationalize high
quality games using a dummy variable that takes the value of 1 if a game has a Metascore of
75 or higher. Metacritic uses the 75 threshold score to indicate (using colors) that game reviews
are “generally favorable” and of “universal acclaim.”11 Since not all games have Metascores, and
because critics’ decisions to cover games are likely non-random and potentially correlated with sales
(Hsu 2006), we introduce an additional outcome variable: review selection. Review selection is a
dummy variable that indicates whether games have a Metascore registered on Metacritic. In our
system of equations, we estimate the probability of a game receiving a Metascore jointly with the
effect of quality on sales performance and other outcome equations, so as to control for the potential
selection bias. We operationalize games’ sales performance as their cumulative revenues (in GBP)
measured at the end of our data collection period. Console video games have very short product
lifecycles with most games generating the bulk of their revenues within the first few months (Nair
2007; Ishihara and Ching 2016). We therefore are not concerned about potential biases due to data
truncation or right-censoring.
3.3.2 Partner selection (match valuation) predictors
We rely on the existing literature on partner selection for the identification of relevant predictor
variables. Since all developers in our dataset are private firms and since access to information is
important for gauging the existence of potential complementarities, particularly for private firms,
we control for the number of times a developer-publisher pair has worked together in a two year
rolling window period. Furthermore, and for similar reasons, we measure the geographical distance
between the locations of the headquarters of a publisher-developer pair (in km) expecting an inverse
9Expert reviews are not entirely free from potential selection bias. We attempt to account for this in our proposed
model. Moreover, experts may also not be completely unbiased in their reviews. However, since we use aggregated
review scores from over 300 publications, we increase the accuracy of this measure through averaging individual
expert errors and biases (Liu and Ishihara 2016; Rousseau 1985).
10For example, see Activision’s four-game contract with developer Bungie, section 10.3:
http://documents.latimes.com/bungie-activision-contract – last accessed April 15, 2015.
11For more on Metacritic’s colored grading scheme, see: http://www.metacritic.com/about-metascores.
11
relationship with acquisition match value. We measure complementarities in the product market
as the additional genre experience a publisher would gain by acquiring a focal developer. We first
identify the set of game genres released by the publisher and the developer over a two year rolling
window. We then count the number of game genres in the developer’s set, but not in the publisher’s
set. If the developer’s set is completely included in the publisher’s set, then acquiring the developer
does not add any new genre experience to the publisher. This variable takes a larger value if the
developer has more genre experience that the publisher does not.
We also include a number of firm characteristics in our matching equation. At the publisher
level, we control for the number of past acquisitions a publisher has made over a period of a two year
rolling window expecting that experienced acquirers are more likely to make future acquisitions. We
include publishers’ financial status as measured in total sales (in GBP) of all games released over a
two year rolling window. We further include measures of acquirer diversification (genre experience)
as measured by the number of unique genres a publisher released games in over a two year rolling
window. Similarly, we document the number of unique developers a publisher has worked with
over a two year rolling window. Lastly, we include measures of market competition at the acquirer
level by counting the unique number of competing publishers releasing games across all platforms
on which the focal publisher released at least one game over a two year rolling window, as well
as the average platform-level Herfindahl Index (HHI) across all platforms on which the publisher
released at least one game over a two year rolling window. At the developer level we include similar
measures of financial status, diversification, and competition. We also include three measures of
product reputation at the developer level: the number of game awards received by the developer,
the percentage share of video games developed which are based on new IP, and the percentage
share of video games developed which are of high quality (i.e. Metascore 75).12
3.3.3 Outcome equation predictors
Given the peculiar nature of the media and entertainment industries, we take a more contextualized
approach in choosing predictor variables for our outcome equations. In all our outcome equations
we include the following predictor variables. First, we control for whether a game is developed by
an internally founded developer. By including this dummy the base category for the acquisition
12We collected the award data from six prominent video game awards: Academy of Interactive Arts & Sciences,
British Academy of Film and Television Arts, Game Developers Choice, Golden Joystick, IGN, and Spike.
12
dummy are games developed by independent developers. Second, we control for the number of
platforms a game is launched on. We expect platform exclusive games to be more likely to be
innovative, have higher quality due to better platform specific technical fit, but have lower sales
given their restricted market potential. Third, we control for the age of the focal platform as well
as the age of the next generation platform to control for demand heterogeneity in the platform
consumer pool as well as for overall market size (Rietveld and Eggers 2016). Fourth, we control for
the financial status of the publisher as a proxy for marketing budget (publishers typically spend a
fixed proportion of their allocated development budgets on marketing activities). Fifth, we control
for the number of awards the focal developer won over a two year rolling window as a proxy for
their technical and creative capabilities. Lastly, with the exception of the new IP equation, we
control for whether games feature any famous persons (e.g., Tony Hawk, Michael Jackson) either
on the box cover or in the game title. We expect stars to have a positive effect on the likelihood
of receiving a Metascore, as well as on games’ quality and sales. All outcome equations further
include publisher fixed effects, year of release fixed effects, platform fixed effects, calendar month
of release fixed effects, genre fixed effects, and age restriction fixed effects.
We also include some specific control variables in each of our outcome equations. As a way of
exclusion restriction, and building off work by Hsu (2006) on the attention of expert critics in the
motion picture industry, we include the following controls in the review selection outcome: First,
we control for how coherent experts have rated games within a focal genre by including the variance
of Metascores over a two year rolling window. Second, we control for the overall crowdedness of the
genre by including the number of games that have been released in the focal genre over a two year
rolling window. Third, we control for publishers’ relationships with expert critics by including a
ratio measure of the number of games by the focal publisher that have received Metascores relative
to all games launched by that publisher, again over a two year rolling window. We expect that
lower genre variance, less crowded product categories, and better relationships with critics improves
the changes of a game having a Metascore registered on Metacritic.
In the quality equation we control for firms experiences within the focal genre as well as for their
dyadic, developer-publisher pair specific working experience. We expect that the number of games
developers and publishers have released within the focal genre will result in valuable experience.
We also expect that the number of times a developer-publisher pair has worked together in the past
13
(over a two-year rolling window) will likely increase the quality of their products through improved
efficiencies.
In the final outcome equation estimating games’ sales performance we include a measure of
competition at the genre-platform level. Following Rietveld and Eggers (2016) we count the number
of games released on the platform with the focal genre a month prior to the release of the focal
game. We expect more crowded product categories to negatively correlate with sales. Finally, we
include a dummy indicating whether a game has a Metascore registered on Metacritic so that the
base category against which the high quality score is estimated are games with low or medium
quality (i.e. Metascore <75). As we discussed earlier, this variable is subject to the selection issue
and we control for it by the inclusion of the review selection model.13
4 Model
4.1 A Two-Sided Matching Model
We control for non-random partner selection by estimating a two-sided matching model of pub-
lishers’ acquisitions of independent developers. The acquisition process is not a one-way decision
of which developer(s) a publisher acquires. Developers choose whether to get acquired (and by
whom) or remain independent. Unlike a typical discrete choice model such as Probit, a two-sided
matching model allows to capture such a mutual decision making process. Consequently, it has
been advocated in recent studies on M&A (e.g., Park 2013, Rao et al. 2016). The type of match-
ing model we employ in this paper is a one-to-many matching model. This is because, given the
acquisition market, which we will define below, a publisher may acquire multiple developers, but a
developer can be acquired by at most one publisher.
Our empirical two-sided matching framework is built on Sørensen (2007). Because we observe
acquisitions at the yearly level, we assume that each of the years in our sample period (2002-2010)
defines an acquisition market.14 We treat each acquisition market as a separate market, so that
13Note that we do not control for games selling prices in the sales specific outcome equation for two main reasons.
First, console video games have platform recommended retail prices (MSRPs) which reduces the variance in first
period selling prices (Nair 2007). Second, we observe only cumulative sales revenues for games, and publishers
adjustments to games’ selling prices in subsequent periods is a reflection of their market performance (i.e. games that
sell poorly will receive greater price drops than games that sell well). Fluctuations in price are therefore determined
by factors endogenous to our outcome variable. When we include games average selling prices (in GBP) as a control
variable, our main results remain unchanged.
14There is only one geographic market in our empirical context, as most of the video game publishers are multina-
tional companies releasing games in all major geographical markets.
14
acquisition decisions in one year are independent of those in another year. The acquisition market
in year tconsists of two disjoint sets of players: active publishers (potential acquirers) and active
independent developers (potential acquirees). Let It, Jtbe the sets of publishers and developers,
respectively. At each time t, publisher iItcan acquire multiple developers or not acquire at all,
whereas developer jJtcan be acquired by one publisher, or remain independent. A matching
at time t,µt, is a collection of acquisitions. Let µit Jt∪ {0}be the set of developers publisher
iacquires at time t, and µjt It∪ {0}be the publisher acquiring developer j. In both cases,
µit ={0}and µjt = 0 represent a situation where publisher idoes not acquire any developer, and
developer jremains independent, respectively.
4.1.1 Preferences
To represent publishers’ preferences over developers and developers’ preferences over publishers, we
introduce a joint valuation for each potential match. Let Vij t be the valuation of a match between
publisher iand developer jin year t. We can view Vijt as an approximation for the present
discounted value of current and future joint profits if publisher iacquires developer jat time t. In
order to derive preference ordering and the unique stable match, we follow the previous literature
and make the following two assumptions on Vijt and the way Vijt is split between publisher iand
developer j.
Assumption 4.1. Valuations are distinct at each time t.
This assumption states that Vijt 6=Vij 0tfor all j06=jand Vijt 6=Vi0j t for all i06=i. In other
words, preferences are strict: there is no tie in publishers’ preference over developers and developers’
preference over publishers. Given Vijt, we assume the following valuation sharing rule between the
publisher and the developer:
Assumption 4.2. The valuation, Vij t, is divided between publisher iand developer jaccording to
a fixed sharing rule determined by λt(0,1).
Let πit and πjt be the payoff from a match µt. The fixed sharing rule implies
πit =λtX
jµit
Vijt ,and πjt = (1 λt)Vµjtjt .
Publishers receive the λtproportion of the joint valuation from each of the matched developers,
and developers receive the (1 λt) proportion from the acquiring publishers. The proportion
15
λtis assumed to be uniform across all possible matches within year t. It follows from the two
assumptions that publisher iprefers developer jto developer j0when Vijt > Vij 0t, and developer j
prefers publisher ito publisher i0if Vijt > Vi0j t.
Before characterizing stable matching, we make a few remarks on the assumptions. First, the
above approach does not consider complementary preferences. For example, we do not allow for
preferences where publisher 1 prefers to acquire developer 1 only when developer 2 is acquired
together (i.e., acquiring both developers 1 and 2 is preferred to not acquiring, which is preferred
to acquiring developer 1 only). We do not allow for complementary preferences mainly due to a
technical reason: we cannot guarantee the existence of a stable matching (Roth and Sotomayor
1990). However, substitutable preferences are not far from reality in our empirical setting. Pub-
lishers rarely acquire two developers in anticipation of synergies between the two. This is mainly
because independent developers tend to have different cultures, and even after they are acquired
by a publisher, they operate rather autonomously. Indeed, in our data, we did not find instances
where two developers who were acquired by the same publisher were re-branded as one developer.
This claim does not exclude the possibility that informal interaction among employees of acquired
developers generates potential synergies. However, we believe that this would not be the main
reason for acquiring multiple developers jointly.
Second, we acknowledge that although the fixed sharing rule assumption makes the model
empirically tractable, it might impose a strong assumption. In general, a publisher may offer a
higher proportion of the joint valuation to an attractive developer in order to win the acquisition
competition with other interested publishers. Given that most of the publishers and developers in
the video game industry are private companies, we are unable to obtain transfer data to verify our
assumption. However, we note that since publishers’ characteristics enter Vijt, different publishers
pay a different amount to acquire the same developer (Park 2013). That is, if Vijt > Vi0j t, publisher
ipays to acquire developer jmore than publisher i0does. The same logic applies to developers:
developer jreceives more from publisher ithan developer j0does if Vijt > Vij0t. Although a model
with endogenous transfer could better accommodate the variation in πit and πjt across different
matches via an endogenous sharing rule, estimating such a model with our data set and for our
empirical context is technically challenging and beyond the scope of this paper.
16
4.1.2 Stable matching
The equilibrium concept in our two-sided matching model is group stability. However, as shown in
Roth and Sotomayor (1990), a matching is group stable if and only if it is pairwise stable. Thus,
we can focus only on pairwise stability. Assumptions 4.1 and 4.2 imply that a stable matching
exists and it is unique, and that the unique matching can be characterized by a set of inequalities
described below (Park 2013; Sørensen 2007):
Vijt > V ijt (i, j)µt,and Vij t <¯
Vijt (i, j )/µt,
where Vijt max nmaxi0Sj t Vi0jt,maxj0Sit Vij0towith Sjt niIt∪ {0}|Vijt >minj0µit Vij0to
and Sit njJt∪ {0}|Vijt > Vµj tjto, and ¯
Vijt max nVµj tjt,minj0µit Vij 0to.
First, the set of inequalities Vijt > V ijt for (i, j)µtguarantees that for a matched pair (i, j),
the current valuation is larger than any valuation that could be obtained by deviating to a partner
who is willing to form a new match with ior j. The first term in the max function, maxi0Sjt Vi0j t,
represents the maximum valuation developer jcan obtain from deviating to a publisher who is
willing to match with developer j.Sjt is the set of publishers who are willing to match with
developer j, i.e., the valuation with developer jis larger than the lowest valuation from the currently
matched developers. The second term, maxj0Sit Vij0t, represents the maximum valuation publisher
ican obtain from deviating to a developer who is wiling to match with publisher i.Sit is the set of
developers who are willing to match with publisher i, i.e., the valuation with publisher iis larger
than the valuation from its current match.
Second, the set of inequalities Vijt <¯
Vijt for (i, j )/µtguarantees that for an unmatched pair
(i, j), the valuation is lower than that under the current match µt. The terms in the max function
represent the current match value for developer j, and the minimum value publisher ireceives from
the current match.
These inequalities will be exploited in the estimation as a set of constraints that parameters
need to satisfy. We will elaborate on this in the estimation strategy section (Section 4.4).
4.2 The empirical specification of the match valuation
In our empirical analysis, we model the valuation of a match between publisher iand developer j
in year t,Vijt, as
Vijt =Wij tα+ηij t,
17
where Wijt is a vector of observed variables that might affect the valuation, and ηijt is an error that
is assumed to be normally distributed with zero mean and variance normalized to one. Further,
we normalize the mean valuation of a match (i, j) with i= 0 or j= 0 (i.e., not being acquired and
not acquiring, respectively) to zero.15
4.3 The empirical structure of the outcome equations
Our goal is to examine the effect of acquisitions on (1) product innovativeness, (2) product quality,
and (3) sales performance for each of the games produced by the matched publisher and developer
pair. As in Figure 1, we expect that the decision on new IP is made early in the game development,
and whether a game is a new IP might affect experts’ choice on reviewing, actual expert score upon
being reviewed, and sales performance. The review decision then plays a role of the selection equa-
tion for the quality equation (Heckman 1976), and also helps control for the endogenous treatment
effect of a game being reviewed on sales performance. Finally, we expect that quality affects games’
sales performance.
Since new IP, review selection, and high quality are all indicator variables, we model them in a
Probit framework. Sales performance is modeled in a log-linear specification, and we assume that
the error is normally distributed. Let I
ijkt , R
ijkt , Q
ijkt , and Y
ijkt be the error terms for new IP, review
selection, high quality, and sales performance models, respectively, for game kby publisher iand
developer jreleased in year t. We allow these errors to be correlated and estimate the correlation
parameters.
4.4 Estimation Strategy
We estimate the four outcome equations together with the two-sided matching model we presented.
Our major econometric issue is endogenous sorting: developers who are acquired are systematically
different from those who are not. The source of unobservables that generate such a systematic
difference may come from developers’ potential skills in developing a high quality game and/or a
successful new IP, synergies between a publisher and a developer, etc. We expect these unobserv-
ables to affect both the joint valuation, Vijt, as well as the key outcome variables for games developed
by an acquired developer and released in year τt. Thus, in order to control for the endogeneity
15The same assumption was made in Park (2013). We acknowledge that this assumption could be restrictive, as
different publishers and developers might have different outside options. In general, it is difficult to assess the value
of those options firm by firm.
18
of acquisitions, we allow for correlations between ηijt and ij (I
ijkτ , R
ijkτ , Q
ijkτ , Y
ijkτ ).
We further note that since our outcome analysis is conducted at the game-level, it is possible
that a developer may release multiple games post-acquisition. In this case, we take ηijt at the
time of of acquisition and let it be correlated with the errors of the outcome equations for all
post-acquisition games. We provide a detailed description of the error distributions in Appendix
B.
4.4.1 Estimation Algorithm
We estimate the proposed model by a Bayesian Markov chain Monte Carlo (MCMC) algorithm.
Parameters are drawn by Gibbs sampling, based on the conditional error distribution obtained from
the distribution described in Appendix B. A Bayesian approach is attractive for several reasons.
First, we can apply the data augmentation approach to draw the match valuation, which avoids
costly multi-dimensional integration (Sørensen 2007). Furthermore, the outcome models for new
IP, review selection, and high quality are Probit. Similar to the match valuation, we can apply the
data augmentation approach to reduce the computational burden (Albert and Chib 1993).
The Gibbs sampling mainly consists of five blocks: (1) augment Vijt from the set of inequalities
that characterize stable matching for all matched and unmatched pairs (i, j); (2) draw αin the
match valuation equation conditional on augmented Vijt using the Bayesian linear regression; (3)
augment the latent variables of the Probit models for new IP, review selection, and high quality,
(say, UI
ijkt , U R
ijkt , U Q
ijkt ), from the inequalities for their corresponding Probit models; (4) draw the
parameters of the outcome equations conditional on augmented UI
ijkt , U R
ijkt , U Q
ijkt , and (logged)
observed sales (Yijkt) using the Bayesian linear regression; (5) draw the parameters of the covariance
matrix of the errors conditional on the residuals from (1)-(4). In all five steps, we use appropriate
conditional distributions of the error terms. Lastly, we use an uninformative prior for all parameters.
5 Results
5.1 Summary Statistics
Prior to our full investigation on the data, we compare key outcome variables (product innova-
tiveness, quality, and sales performance) by categorizing games into three groups: games by inde-
pendent developers, games by acquired developers, and games by internally founded developers.
19
Out of 5,916 games in our sample, 3,746 games were developed by independent developers, 972 by
acquired developers, and 1,198 by internally founded developers.
Table 2 shows the results. We first compare the key variables of games by independent de-
velopers (labeled as “Independent All”) vs. acquired developers (“Acquired”). Our findings here
are summarized as follows: (1) there is a considerable difference in the likelihood of introducing a
new IP by the two types: about 29% of games by independent developers are new IPs while only
17% by acquired developers are new IPs. This initial examination confirms our conjecture that one
important reason for acquiring a developer is to secure future access to proven hit-properties; (2)
not all games in our sample received expert scores. Here we also see a difference between acquired
and independent developers. Games by acquired developers are more likely to be reviewed than
those by independent developers, by about 16 percentage points. This difference is important in
our analysis, as unobserved factors that affect the experts’ decision on reviewing could bias the
effect of acquisitions on quality; (3) the percentage of high quality games by acquired developers is
higher than that by independent developers: about 26% of games by independent developers are
of high quality while about 48% by acquired developers are of high quality; (4) games by acquired
developers earn twice as much revenues as those by independent developers on average (GBP 1.06
million vs. 2.21 million). The bottom panel of Table 2 shows the mean differences and significance
levels from a two-tailed t-test with equal variances. All the differences are significant at 1% level.
The above differences between independent vs. acquired developers may be due to acquisition.
Alternatively, they may be explained if acquired developers are systematically different from other
independent developers who were not acquired during our sample period. If the latter is the case,
the above mean differences are simply due to sample selection. To disentangle these effects further,
we separate games into those by independent developers who were always independent (“Always
Independent”) and those by independent developers who were later acquired by a publisher (“Later
Acquired”). We report the mean differences at the bottom of Table 2 under “Later Acquired - Al-
ways Independent.” We find that (1) although later acquired developers develop less games based
on new IP by 3.5 percentage points than always independent developers do, the difference is not sta-
tistically significant; (2) games by later acquired developers are more likely to be reviewed (by 17.5
percentage points) than those by always independent developers; (3) later acquired developers are
more likely to develop a high quality game (by 9.8 percentage points); (4) games by later acquired
20
developers earn more revenues (by about GBP 290,000), but the difference is not significant. These
results indicate that “Later Acquired” developers are indeed different from “Always Independent”
developers on some aspects.
Lastly, to see the acquisition treatment effect, we provide the comparison of the key variables
between “Later Acquired” independent developers and “Acquired” developers. We find that al-
though the mean differences are smaller as compared to the differences between “Independent All”
and “Acquired,” they are still statistically significant at 1% level: compared to their pre-acquisition
games, post-acquisition games are (1) less likely to be based on new IP (by 9.3 percentage points),
(2) more likely to be reviewed by experts (by 6.9 percentage points), (3) more likely to be of high
quality (by 12.6 percentage points), and (4) earn more revenues (by about GBP 880,000).
These descriptive data patterns are useful for gaining understanding about the impacts of
acquisitions on key outcome variables. At first glance, these results suggest that there are both
selection effects and acquisition treatment effects at play. However, it does not allow us to learn
about the mechanism behind why some key variables by acquired developers are higher or lower
than those by independent developers. Moreover, these patterns may be subject to unobservables
that are correlated with acquisition decisions. The next section discusses the results from our
proposed model that deals with these issues.
5.2 Results from the proposed model
We now report the results from our proposed framework that estimates the two-sided matching
model and the outcome equations jointly to control for potential endogenous sorting. The matching
model does not only allow us to control for this potential bias, but also provides insights about the
determinants of partner selection. Thus, we first discuss the parameter estimates for the match
valuation, and then move on to the parameter estimates for the outcome equations. We draw 30,000
parameters from the posterior distribution, and throw away the first 10,000 draws as burn-in. To
reduce the serial correlation, we use every 10th draw of the remaining 20,000 draws to compute
the posterior mean and standard deviation of the parameters. Figures 4 and 5 plot the parameter
draws for key variables in the matching and outcome models, respectively.
21
5.2.1 Results for the match valuation
Recall that we included in Wijt pair-specific variables as well as publisher- and developer-specific
variables. Table 3 reports descriptive statistics for the variables included in the match valuation,
and Table 4 reports estimation results. The first two columns of Table 4 report the posterior
mean and standard deviation of the parameters, respectively. Since the match valuation is a latent
variable, the parameter estimates are not readily interpretable. To make the estimates interpretable
and comparable across parameters, the third column (dP /dW ) presents a measure similar to the
marginal effect, proposed by Sørensen (2007). It measures the change in the likelihood of one match
being preferred to another, due to a change in the focal variable, assuming that the two matches
have the same valuation except for the change in the focal variable.16
First, we discuss pair-specific variables. We find that the number of times a developer-publisher
pair has worked together has a positive and significant effect, and the corresponding “marginal
effect” is 0.139. This indicates that an increase in the (logged) number of times a developer-publisher
pair has worked together by one unit increases the likelihood of a match being preferred by 13.9
percentage points. The physical distance variable has a negative and marginally significant effect
(dP/dW =0.004). These findings are consistent with the importance of access to information
and the facilitating role of geographic proximity herein (Chakrabarti and Mitchell 2013). We find
that complementarities (additional genre experience a publisher would gain by acquiring a focal
developer) are not significant.
Second, among publisher-level variables, we find the number of past acquisitions by a publisher
and a publisher’s financial status have a positive and significant effect on the match value (dP/dW =
0.065 and 0.007, respectively), confirming that experienced and financially healthy publishers are
more likely to make further acquisitions (Park 2013). We also find a negative impact of the number
of competitors for a publisher (dP/dW =0.021), suggesting that we observe more acquisitions
when the focal publisher faces less competition. This result is in line with recent studies on vertical
integration (e.g., Loertscher and Reisinger 2014). A market with lower competitiveness provides
more incentives for firms to vertically integrate, because the potential increase in profit due to the
elimination of double marginalization is larger when the market is less competitive.
Most of the developer-level variables are not significant, except for the percentage share of
16Mathematically, it is computed as φ(0)α/2, where φ(·) is the standard normal probability density function.
22
games developed that are of high quality (dP/dW = 0.052). This result is consistent with our
preliminary finding in Table 2 that independent developers who were later acquired tended to
develop more high quality games than those who were never acquired. The non-significant effect
of the percentage share of games developed that are based on novel intellectual properties may
suggest that innovativeness by developers itself is not necessarily attractive to publishers. It may
also mean that innovative developers prefer to remain independent. This is further evidenced by
the non-significant effect of the interaction between innovativeness and quality: even if developers
are both innovative and capable of developing high quality games, they are not likely to be acquired
by publishers.
5.2.2 Results for the outcome equations
We now move on to discuss the results for the outcome equations. Table 5 lists descriptive statistics
and the estimation results are reported in Table 6. For each of the four outcome equations in Table
6, the first and second columns report the posterior mean and standard deviation, respectively. The
first three outcomes are modeled in a Probit framework, so we add the third column and report
the marginal effect (dF/dX) at the average value of the independent variables. For continuous
independent variables, the marginal effect measures the percentage point change due to a marginal
change in those independent variables. For dummy variables, the marginal effect measures the
percentage point change when the dummy variables change from zero to one. Sales performance
is modeled in a log-linear specification. Thus, the posterior mean corresponds to the percentage
change in sales (in GBP) due to a marginal change in independent variables.
The effect of acquisitions on product-level outcomes
We first discuss the direct effect of acquisitions on key outcome variables. The direct effects are
largely consistent with our preliminary analysis in Table 2. The effect on new IP is negative and
significant. The marginal effect suggests that acquisition reduces the probability of releasing games
based on novel intellectual properties by 6.0 percentage points. The effect of acquisitions on the
chance of receiving an expert review is positive and significant: acquisition increases the probability
of a game being reviewed by experts by 8.2 percentage points. The chance of producing a high
quality game also increases by 21.6 percentage points as a result of acquisition.
23
Lastly, contrary to our finding in Table 2, we find no direct effect of acquisition on sales per-
formance. Thus, the large mean difference in sales performance between games by acquired vs.
independent developers in Table 2 may be either due to control variables, or mediated by the effect
of other key outcome variables on sales performance. For example, we find a positive impact of
high quality games on sales performance, and we have discussed improved quality post-acquisition.
As a result, high quality may mediate the effect of acquisition on sales performance. We uncover
such relationships by examining both indirect and total effects of acquisitions.
Based on the parameter estimates, we calculate the indirect and total effects of acquisitions. As
in the conceptual framework in Figure 1, acquisitions could affect quality indirectly via its effect on
new IP and review selection, and it could affect sales performance indirectly via its effects on new
IP, review selection, and quality. We first examine the indirect effects. The bottom panel in Table
6 shows the indirect effects and the total effects in terms of the marginal effects. For example, the
indirect effect of acquisitions on quality via new IP (AIin the “High Quality” column) is
0.023, suggesting that the chance of a game being high quality increases by 2.3 percentage points
due to this indirect effect: Acquisition leads to less new IP, which in turn improves product quality.
We find that some of the indirect effects are significantly different from zero: (1) the indirect effect
on review selection via new IP is 0.027, or 2.7 percentage points, (2) the indirect effect on sales
performance via new IP is -0.030, and (3) the indirect effect on sales performance via quality is
0.644. The latter two suggest important roles that innovativeness and quality play in understanding
the mechanism behind the effect of acquisitions on sales performance. On one hand, acquisitions
lead to less innovative games, which decreases sales performance. On the other hand, acquisitions
lead to higher quality games, which increases sales performance.
To see the total effects of acquisitions, we present them below the indirect effects in the same
panel of Table 6. We find that acquisitions overall reduce innovativeness (same as the direct effect),
but increases the change of being reviewed and game quality by 9.7 percentage points and 28.7
percentage points, respectively. The total effect on sales performance is positive (0.484). Thus,
the positive indirect effect via quality improvement dominates the negative indirect effect via less
new IP. Since the sales performance equation is a log-linear specification, acquisition leads to an
increase in sales performance by 48.4%. Given that the average sales (in GBP) in our sample is
about GBP 1.55 million, a 48.4% increase is roughly equivalent to GBP 0.775 million per game.
24
Control variables
We end our discussion on parameter estimates by briefly discussing the estimates for control vari-
ables. First, most of the control variables have expected signs. In the new IP equation, we find a
negative effect for most of the control variables, except for platform exclusive games (8.9 percentage
points). Games by internally founded developers have a similar negative effect as acquired devel-
opers. In the review selection equation, we find that new IP games have a lower chance of being
reviewed (by 60.2 percentage points). Platform exclusive games are also less likely to be reviewed
(by 7.4 percentage points), but games with star endorsement are more likely to be reviewed (by
11.2 percentage points). All the exclusion restrictions are significant and have the expected signs:
variance of reviews (disagreement) and the number of reviews have a negative effect, and the per-
centage of the publisher’s games reviewed in the past has a positive effect. In the quality equation,
we find a negative effect of new IP (by 17.3 percentage points) and a positive effect of awards
received by developer (1.7 percentage points), platform exclusive games (17.4 percentage points),
and games with star endorsement (17.8 percentage points). We also find that the number of times
a publisher-developer pair has worked has a positive impact (7.5 percentage points). Lastly, in the
sales performance equation, we find a positive effect of high quality games (49.5%), new IP games
(223.8%), games by internally founded developers (34.9%), publisher’s financial status (80.8%), and
games with star endorsement (54.7%). We also find platform exclusive games have a lower sales
performance (by 51.6%). Games that are not reviewed by experts have a higher sales performance
than games that are reviewed but not of high quality (i.e. Metascore <75).
5.3 Robustness checks
We checked the robustness of our main results for the outcome equations in several ways, and
report the main results of the outcome equations in Table 7. Throughout the robustness checks,
we used the same specification of the matching model as our main specification (except for the first
robustness check where we dropped the matching model). The estimates of the matching model
are very similar to those of the main specification. Thus we focus on the results of the outcome
equations.
First, we estimated the outcome model without the matching model. This robustness check
allows us to see the effect of potential endogenous sorting on the impact of acquisitions. We report
25
the correlations between the error of the matching equation and those of the outcome equations
under Correlation. Our results suggest that the correlation is significant for the sales performance
(0.054). This positive correlation indicates that there is an unobserved factor that affects the match
valuation and the sales performance positively (e.g., any unobserved factor that affects sales can
be valued as part of the match valuation during the acquisition process because the match value
is an approximation of future profits). If we ignore such an unobservable, we may over-estimate
the effect of acquisitions on sales performance. This is indeed what we find. The total effect of
acquisitions on sales performance in the absence of the matching model turned out to be 0.790,
which is much higher than that of our main specification (0.484). This finding reverberates with
recent studies (e.g., Park 2013; Rao et al. 2016; Yu et al. 2015) emphasizing the importance of
controlling for non-random partner selection when measuring the effect of acquisitions on outcome
measures.
Second, our sample of games include those by internally founded developers. These games do
not help identify the effect of acquisitions on key outcome variables. We thus re-estimated the entire
model by restricting our sample of games to those by independent and acquired developers only.
Our main results remain unchanged. Third, in our main analysis, we do not model any role of prices.
As we explained earlier, this is partly because of little variation in first-period prices across games
(Nair 2007), and also because our data does not include price trajectories across games but only the
average selling price based on the cumulative dollar and unit sales. Nevertheless, we re-estimated
the entire model by including the average selling price as a control variable in the sales performance
equation. We find a positive and significant effect of the average selling price. Partly due to this
effect, the effect of high quality on sales performance decreases slightly. Overall, our main results
remain unchanged. Fourth, we included the measure of market competition in the quality equation
and re-estimated the entire model. Previous studies suggest that product quality may decrease as
market competitiveness decreases (Fan 2013; Masta 2011). We find that the measure of market
competition is not significant, and the main results on the effect of acquisitions on key outcome
variables remain unchanged.
Lastly, we examined whether complementarities between a publisher and its acquired developer
have any effect on key outcome variables as suggested by previous studies (King et al. 2008;
Makri et al. 2010; Zhao 2009). In the main specification, we have this variable as a predictor
26
for the match valuation, but did not include it in the outcome equations. To examine the role of
complementarities, in all outcome equations, we include an interaction term between the acquisition
dummy and the measure of complementarities (defined as additional genre experience the publisher
gained by acquiring the developer). We find that for the new IP and review selection equations,
the interaction term is not significant. However, we find complementarities help improve quality
but have a negative direct effect on sales performance. Overall, this robustness check adds a
more granular interpretation of the effect of acquisitions on key outcome variables, but the main
conclusion remains unchanged.
6 Discussion
This paper contributes to the literature by looking at how vertical acquisitions affect targets’
product-level outcomes. Compared to past work that has mostly focused on the firm level, a
focus on products allows for a more fine-grained analysis of how mergers and acquisitions create
value. We also contribute by jointly estimating the antecedents and consequences of acquisitions
in a Structural Equation Modeling framework. While the importance of controlling for partner
selection has been emphasized in past work, we are unaware of any studies that jointly estimate
the effect of acquisitions on multiple related outcome variables. The main results of our analyses
are summarized in Figure 3.
The findings from our match valuation analysis are generally consistent with existing theories
of partner selection in the acquisitions literature. Experienced and financially healthy video game
publishers acquire upstream game developers with a reputation for high quality products mainly to
secure access to existing hit-properties and valuable product development capabilities. These results
correspond with Schweizer’s (2005) qualitative study of biotech acquisitions by pharmaceutical
companies. Furthermore, the selection of targets and the timing of acquisitions appear to be driven
by uncertainty. First, we found that one of the strongest predictors of a publisher-developer pair
engaging in acquisition is the number of times they had collaborated in the past. This finding
re-emphasizes the importance of prior alliances and access to private information in gauging the
presence of potential synergies (Higgins and Rodriguez 2006; Porrini 2004; Yu et al. 2015). Second,
we found that greater geographical distance is inversely correlated with partner selection. This
finding is in line with the notion that the search for valuable resources becomes more difficult
27
as distance increases (Chakrabarti and Mitchell 2013). This particularly applies to video game
developers whose product development capabilities are difficult to assess.
The findings from our outcome equations yield new insights in how acquisitions affect firms’
products and in the mechanism behind how acquisitions create value. First, we found that acquired
video game developers were less likely to release innovative video games than independent game
developers. While publishers acquire developers for their product development capabilities, the
integrated organizational form hinders developers’ subsequent innovation output. This finding sug-
gests that by acquiring upstream video game developers, publishers can effectively secure access to
successful intellectual properties, but fail to leverage developers’ capabilities for ongoing innovation
output. Previous research reporting similar results has linked these findings to a drop in autonomy
and increased financial controls for employees in the target firm (Hitt et al. 1996; Kapoor and
Lim 2007; Puranam and Srikanth 2007). This explanation corresponds with our understanding
of the institutional context where video game designers dread working for large and profit-driven
publishers and where only games based on new intellectual properties are seen as truly creative
projects.
Second, we found that video games by acquired developers are more likely to be of high quality
than video games released by independent developers. Past research suggested two competing
hypotheses for how acquisitions affect (product) quality. On the one hand, some studies have
argued for a coordination argument where integration between a target and a buyer firm enhances
efficiency resulting in improvements in quality (Cassiman et al. 2005; Klein et al. 1978; Prince
and Simon 2015; Williamson 1979). On the other hand, when acquisitions reduce competition in
the marketplace this may disincentivize the acquired firm resulting in a drop in post-acquisition
quality (Fan 2013; Masta 2011). Our findings are consistent with the coordination hypothesis. Even
when we controlled for platform-level competition in our quality equation as a robustness check,
the positive effect of acquisition persisted. Since quality in our study is externally determined by
expert critics, our findings suggest that by merging, video game publishers are able to exert greater
direction and coordination over the product development process to improve video games’ appeal
to external critics.
Third, while we found no direct effect of acquisitions on games’ sales performance, the overall
effect is mediated by the positive effect of quality on sales. Our findings suggest that value in the
28
video game industry is created by bundling complementary resources (Barney 1988; Harrison et
al. 1991; 2001). Video game developers specialize in the ideation and creation of game content,
while publishers specialize in commercializing and marketing video games. Merging these activities
provides publishers with greater control and direction, which allows them to optimize the product
development process and achieve quality enhancing efficiencies. Our finding that games by acquired
developers have a higher likelihood of being reviewed by expert critics further underscores the
complementarity hypothesis. Marketing capabilities including relationships with expert critics are
strong predictors of having products reviewed, and video game publishers are more experienced in
these activities than developers are. The sum of our findings suggest that acquisitions create value
in the video game industry through improved inter-firm coordination resulting in enhanced product
quality and higher sales performance. Overall, we find that there exists an acquisition premium of
GBP 775,000 per video game.
Our outcome equations yield a number of additional findings of interest. We found that video
games based on novel intellectual properties have a lower probability of being reviewed by expert
critics. This contributes to Hsu’s (2006) work on critics’ evaluative schemas and attention. Our
results show that critics’ attention is not only affected by uncertainty at the product-category
level but equally so by uncertainty at the product level. When expert critics do decide to review
innovative video games, we found that they are generally less likely to be evaluated favorably
or receive critical acclaim. While the direct effect of innovativeness on sales performance was
positive, the combined negative effects on review selection and high quality render the total effect
of innovativeness on video games’ sales performance non-significant. These results align with the
general notion that innovative products exhibit greater risk for their stakeholders (e.g., producers,
evaluating experts, and consumers).
7 Limitations and Future Research
Our study has a number of important limitations that provide avenues for future research. First,
our measure of product performance is essentially a short term measure as console video games
generate the bulk of their revenues within the first six months after launch. The long term effects of
acquisitions on performance may be different. Looking at additional data we collected, for example,
we found that of the 85 firms that had been acquired in our study timeframe, 37 had been dissolved
29
or divested by 2012. This unusually high dissolution rate of almost 44% could potentially be linked
to the negative effect of acquisitions on innovation. It is well established that (product) innovation
is positively correlated with firm survival (Cefis and Marsili 2005; 2006; 2012). Therefore, in order
to more thoroughly understand the effects of acquisition on performance, future work needs to
address the long-term effects of acquisition on product outcomes and firm survival.
A second limitation of our study is that it looks at the effects of acquisitions on the target
firm but not on existing business units within the broader acquiring organization. Firm resources,
including managers’ attention and marketing capabilities, are finite. Acquisitions of target firms
and their product launches may thus distract managers or siphon resources away from existing
business units. On the other hand, acquired innovation capabilities and knowledge may spill over
into existing business units, depending on their proximity and degree of relatedness. It seems likely,
therefore, that acquisitions do not solely affect the products of the acquired firm, but also have
important effects on existing business units within the acquiring organization (Karim 2009; Mingo
2013). We encourage future research to address the effect of acquisitions on product-level outcomes
at the broader organizational level.
A third suggestion for future research is to study the timing of merger and acquisitions in the
context of technology platforms. We observed that acquisition activity spiked in 2005, one year
prior to the launch of seventh generation video game consoles. As new video game platforms are
announced well ahead of launch (and publishers are typicaally involved in the development process),
we conjecture that publishers strategically timed their acquisitions in order to prepare for the
impending next generation console launch. This observation is consistent with theories on merger
waves that show that technology shocks drive acquisitive behavior (Harford 2005), and with theories
on how inter-firm collaborations can help firms deal with uncertainty in relation to technology
shocks (Schilling 2015). We hope future research will explore these dynamics further by looking at
how platform lifecycle dynamics affect participating firms’ structural form and whether innovation
outcomes and product performance are contingent on organizational form as the ecosystem evolves.
There are also important limitations to our modeling approach. First, we assume that the
acquisition market is independent across years. In reality, firms could be forward-looking and
acquisition decisions in a given year are made by accounting for potential future acquisitions. If
such behavior is severe, our assumption that the errors of the match valuation are independent
30
across years may not be valid. Furthermore, another type of developers, those who were founded
internally, are treated as exogenous. In our example, this assumption may not be as restrictive as
it may seem, because most of those internally founded studios were formed in eighties and nineties,
when the video game market was relatively small, and many entering publishers were developers.
However, it is possible that publishers in these years might choose to found a studio by themselves
and such a decision could be endogenous and linked to the acquisition market condition (e.g., there
are not many attractive independent developers to acquire). Our current approach does not address
such an endogenous decision on founding a development studio internally. We leave these for future
research.
31
References
Ahuja, Gautam, Riitta Katila. 2001. Technological Acquisitions and the Innovation Performance
of Acquiring Firms: A Longitudinal Study. Strategic Management Journal 22(3) 197–220.
Albert, James H., Siddhartha Chib. 1993. Bayesian Analysis of Binary and Polychotomous Re-
sponse Data. Journal of the American Statistical Association 88(422) 669–679.
Barney, Jay B. 1988. Returns to Bidding Firms in Mergers and Acquisitions: Reconsidering the
Relatedness Hypothesis. Strategic Management Journal 9(S1) 71–78.
Benner, Mary J., Michael L. Tushman. 2003. Exploitation, Exploration, and Process Management:
The Productivity Dilemma Revisited. Academy of Management Review 28(2) 238–256.
Brown, Shona L., Kathleen M. Eisenhardt. 1995. Product Development: Past Research, Present
Findings, and Future Directions. Academy of Management Review 20(2) 343–378.
Cassiman, Bruno, Massimo G. Colombo, Paola Garrone, Reinhilde Veugelers. 2005. The Impact
of M&A on the R&D Process: An Empirical Analysis of the Role of Technological- and Market-
Relatedness. Research Policy 34(2) 195–220.
Cefis, Elena, Orietta Marsili. 2005. A Matter of Life and Death: Innovation and Firm Survival.
Industrial and Corporate Change 14(6) 1167–1192.
Cefis, Elena, Orietta Marsili. 2006. Survivor: The Role of Innovation in Firm’s Survival. Research
Policy 35(5) 626–641.
Cefis, Elena, Orietta Marsili. 2012. Going, Going, Gone. Exit Forms and the Innovation Capabilities
of Firms. Research Policy 41(5) 795–807.
Chakrabarti, Abhirup, Will Mitchell. 2013. The Persistent Effect of Geographic Distance in Ac-
quisition Target Selection. Organization Science 24(6) 1805–1826.
Chao, Raul O., Stylianos Kavadias, Cheryl Gaimon. 2009. Revenue Driven Resource Allocation:
Funding Authority, Incentives, and New Product Development Portfolio Management. Manage-
ment Science 55(9) 1556–1569.
Chatterjee, Sayan. 1986. Types of Synergy and Economic Value: The Impact of Acquisitions on
Merging and Rival Firms. Strategic Management Journal 7(2) 119–139.
Chen, Yongmin, Philip Gayle. 2013. Mergers and Product Quality: Evidence from the Airline
Industry. Working Paper, Department of Economics, University of Colorado at Boulder.
32
Datta, Deepak K., George E. Pinches, V. K. Narayanan. 1992. Factors Influencing Wealth Creation
from Mergers and Acquisitions: A Meta-Analysis. Strategic Management Journal 13(1) 67–84.
Ernst, Holger, Jan Vitt. 2000. The Influence of Corporate Acquisitions on the Behaviour of Key
Investors. R&D Management 30(2) 105–120.
Fan, Ying. 2013. Ownership Consolidation and Product Characteristics: A Study of the US Daily
Newspaper Market. American Economic Review 103(5) 1598–1628.
Fleiss, Joseph L. 1971. Measuring Nominal Scale Agreement Among Many Raters. Psychological
Bulletin 76(5) 378–382.
Gulati, Ranjay, Paul R. Lawrence, Phanish Puranam. 2005. Adaptation in Vertical Relationships:
Beyond Incentive Conflict. Strategic Management Journal 26(5) 415–440.
Harford, Jarrad. 2005. What Drives Merger Waves? Journal of Financial Economics 77(3) 529–
560.
Harrison, Jeffrey S., Michael A. Hitt, Robert E. Hoskisson, R. Duane Ireland. 1991. Synergies and
Post-Acquisition Performance: Differences versus Similarities in Resource Allocations. Journal
of Management 17(1) 173–190.
Harrison, Jeffrey S., Michael A. Hitt, Robert E. Hoskisson, R. Duane Ireland. 2001. Resource
Complementarity in Business Combinations: Extending the Logic to Organizational Alliances.
Journal of Management 27(6) 679–690.
Hauser, John, Gerard J. Tellis, Abbie Griffin. 2006. Research on Innovation: A Review and Agenda
for Marketing Science.Marketing Science 25(6) 687–717.
Heckman, James J. 1976. The Common Structure of Statistical Models of Truncation, Sample
Selection and Limited Dependent Variables and a Simple Estimator for Such Models. Annuals
of Economic and Social Measurement 5(4) 475–492.
Higgins, Matthew J., Daniel Rodriguez. 2006. The Outsourcing of R&D Through Acquisitions in
the Pharmaceutical Industry. Journal of Financial Economics 80(2) 351–383.
Hitt, Michael A., Robert E. Hoskisson, R. Duane Ireland. 1990. Mergers and Acquisitions and Man-
agerial Commitment to Innovation in M-Form Firms. Strategic Management Journal 11(Special
Issue) 29–48.
Hitt, Michael A., Robert E. Hoskisson, R. Duane Ireland, Jeffrey S. Harrison. 1991. Effects of
Acquisitions on R&D Inputs and Outputs. Academy of Management Journal 34(3) 693–706.
33
Hitt, Michael A., Robert E. Hoskisson, Richard A. Johnson, Douglas D. Moesel. 1996. The Market
for Corporate Control and Firm Innovation. Academy of Management Journal 39(5) 1084–1119.
Hsu, Greta. 2006. Evaluative Schemas and the Attention of Critics in the US Film Industry.
Industrial and Corporate Change 15(3) 467–4960.
Ishihara, Masakazu, Andrew T. Ching. 2016. Dynamic Demand for New and Used Durable Goods
without Physical Depreciation: The Case of Japanese Video Games. Working Paper, Stern
School of Business, New York University.
Johns, Jennifer. 2006. Video Games Production Networks: Value Capture, Power Relations and
Embeddedness. Journal of Economic Geography 6(2) 151–180.
Kapoor, Rahul, Kwanghui Lim. 2007. The Impact of Acquisitions on the Productivity of Inventors
at Semiconductor Firms: A Synthesis of Knowledge-Based and Incentive-Based Perspectives.
Academy of Management Journal 50(5) 1133–1155.
Karim, Samina. 2009. Business Unit Reorganization and Innovation in New Product Markets.
Management Science 55(7) 1237–1254.
King, David R., Rebecca J. Slotegraaf, Idalene Kesner. 2008. Performance Implications of Firm
Resource Interactions in the Acquisition of R&D-Intensive Firms. Organization Science 19(2)
327–340.
King, Major David R., Dan R. Dalton, Catherine M. Daily, Jeffrey G. Covin. 2004. Meta-Analyses
of Post-Acquisition Performance: Indications of Unidentified Moderators. Strategic Management
Journal 25(2) 187–200.
Klein, Benjamin, Robert G. Crowford, Armen A. Alchian. 1978. Vertical Integration, Appropriable
Rents, and the Competitive Contracting Process. Journal of Law & Economics 21(2) 297–326.
Krishnan, Ranjani A., Satish Joshi, Hema Krishnan. 2004. The Influence of Mergers on Firms’
ProductMix Strategies. Strategic Management Journal 25(6) 587–611.
Landis, J. Richard, Gary G. Koch. 1977. The Measurement of Observer Agreement for Categorical
Data. Biometrics 33(1) 159–174.
Lee, Robin S. 2013. Vertical Integration and Exclusivity in Platform and Two-Sided Markets.
American Economic Review 103(7) 2960–3000.
Leonard-Barton, Dororthy. 1992. Core Capabilities and Core Rigidities: A Paradox in Managing
New Product Development. Strategic Management Journal 13(Special Issue) 111–125.
34
Liu, Qing, Ruosi Lu, Larry D. Qiu. 2016. Foreign Acquisitions and Targe Firms’ Performance in
China. The World Economy Forthcoming.
Liu, Yuzhou, Masakazu Ishihara. 2016. A Dynamic Structural Model of Endogenous Consumer
Reviews in Durable Goods Markets Working Paper, Stern School of Business, New York Univer-
sity.
Loertscher, Simon, Markus Reisinger. 2014. Market Structure and the Competitive Effects of
Vertical Integration. RAND Journal of Economics 45(3) 471–494.
Lubatkin, Michael. 1983. Mergers and the Performance of the Acquiring Firm. Academy of Man-
agement Review 8(2) 218–225.
Makri, Marianna, Michael A. Hitt, Peter J. Lane. 2010. Complementary Technologies, Knowledge
Relatedness, and Invention Outcomes in High Technology Mergers and Acquisitions. Strategic
Management Journal 31(6) 602–628.
Masta, David A. 2011. Competition and Product Quality in the Supermarket Industry. Quarterly
Journal of Economics 126(3) 1539–1591.
Mingo, S. 2013. The Impact of Acquisitions on the Performance of Existing Organizational Units
in the Acquiring Firm: The Case of an Agribusiness Company. Management Science 59(12)
2687–2701.
Nair, Harikesh. 2007. Intertemporal Price Discrimination with Forward-Looking Consumers: An
Application to the US Market for Console Video-Games. Quantitative Marketing and Economics
5(3) 239–292.
Ni, Jian, Kannan Srinivasan. 2015. Matching in the Sourcing Market: A Structural Analysis of the
Upstream Channel. Marketing Science 34(5) 722–738.
Osterloh, Margit, Bruno S. Frey. 2000. Motivation, Knowledge Transfer, and Organizational Forms.
Organization Science 11(5) 538–550.
Park, Minjung. 2013. Understanding Merger Incentives and Outcomes in the US Mutual Fund
Industry. Journal of Banking and Finance 37 4368–4380.
Paruchuri, Srikanth, Atul Nerkar, Donald C. Hambrick. 2006. Acquisition Integration and Produc-
tivity Losses in the Technical Core: Disruption of Inventors in Acquired Companies. Organization
Science 17(5) 545–562.
35
Porrini, Patrizia. 2004. Can a Previous Alliance Between an Acquirer and a Target Affect Acqui-
sition Performance? Journal of Management 30(4) 545–562.
Prabhu, Jaideep C., Rajesh K. Chandy, Mark E. Ellis. 2005. The Impact of Acquisitions on
Innovation: Poison Pill, Placebo, or Tonic? Journal of Marketing 69(1) 114–130.
Prince, Jeffrey, Daniel H. Simon. 2015. The Impact of Mergers on Quality Provision: Evidence
from the Airline Industry. Working Paper, Indiana University.
Puranam, Phanish, Harbir Singh, Maurizio Zollo. 2006. Organizing for Innovation: Managing the
Coordination-Autonomy Dilemma in Technology Acquisitions. Academy of Management Journal
49(2) 263–280.
Puranam, Phanish, Kannan Srikanth. 2007. What They Know vs. What They Do: How Acquirers
Leverage Technology Acquisitions. Strategic Management Journal 28(8) 805–825.
Rao, Vithala R., Yu Yu, Nita Umashankar. 2016. Anticipated vs. Actual Synergy in Merger Partner
Selection and Post-Merger Innovation. Marketing Science 35(6) 934–952.
Ravenscraft, David J., F.M. Scherer. 1989. The Profitability of Mergers. International Journal of
Industrial Organization 7(1) 101–116.
Rietveld, Joost, J.P. Eggers. 2016. Demand Heterogeneity and the Adoption of Platform Comple-
ments. Working Paper, Rotterdam School of Management, Erasmus University.
Roth, Alvin, Marilda Sotomayor. 1990. Two-sided Matching: A Study in Game-theoretic Modeling
and Analysis. Cambridge University Press, Cambridge.
Rousseau, Denise M. 1985. Issues of Level in Organizational Research: Multi-Level and Cross-Level
Perspectives. Research in Organizational Behavior 71–37.
Saxton, Todd, Marc Dollinger. 2004. Target Reputation and Appropriability: Picking and Deploy-
ing Resources in Acquisitions. Journal of Management 30(1) 123–147.
Schilling, Melissa A. 2009. Understanding the Alliance Data. Strategic Management Journal 30(3)
233–260.
Schilling, Melissa A. 2015. Technology Shocks, Technological Collaboration, and Innovation Out-
comes. Organization Science 26(3) 668–686.
Schweizer, Lars. 2005. Organizational Integration of Acquired Biotechnology Companies into Phar-
maceutical Companies: The Need for a Hybrid Approach. Academy of Management Journal
48(6) 1051–1074.
36
Sheen, Albert. 2014. The Real Product Market Impact of Mergers. Journal of Finance 69(6)
2651–2688.
Singh, Harbir, Cynthia A. Montgomery. 1987. Corporate Acquisition Strategies and Economic
Performance. Strategic Management Journal 8(4) 377–386.
Sørensen, Morten. 2007. How Smart Is Smart Money? A Two-Sided Matching Model of Venture
Capital. Journal of Finance 62(6) 2725–2762.
Stiebale, Joel, Frank Reize. 2011. The Impact of FDI Through Mergers and Acquisitions on
Innovation in Target Firms. International Journal of Industrial Organization 29(2) 155–167.
Tschang, F. Ted. 2007. Balancing the Tensions Between Rationalization and Creativity in the
Video Games Industry. Organization Science 18(6) 989–1005.
Wang, Lihua, Edward J. Zajac. 2007. Alliance or Acquisition? A Dyadic Perspective on Interfirm
Resource Combinations. Strategic Management Journal 28(13) 1291–1317.
Williamson, Oliver E. 1979. Transaction-Cost Economies: The Governance of Contractual Rela-
tions. Journal of Law & Economics 22(2) 233–261.
Yu, Yu, Nita Umashankar, Vithala R. Rao. 2015. Choosing the Right Target: Relative Preferences
for Resource Similarity and Complementarity in Acquisition Choice. Strategic Management Jour-
nal 37(8) 1808–1825.
Zhao, Xinlei. 2009. Technological Innovation and Acquisitions. Management Science 55(7) 1170–
1183.
37
Table 1: Definition of variables
Variable Definition Source
Dependent variables
Acquisition
Dummy = 1 if the game is developed by an acquired developer
S; O
New IP
Dummy = 1 if the game is a new Intellectual Property (IP)
O
High Quality
Dummy = 1 if Metascore 75
M
ln Sales in GBP
Total sales of the game in GBP, [logged]
P
Review Selection
Dummy = 1 if the game is reviewed by experts
M
Outcome equations
Founded Internally
Dummy = 1 if the game is developed by an internally founded developer
O
Age focal pf in month
Focal platform age (in months) at the time of the game's release
P
Age next gen pf in month
Age of next generation platform (in months) at the time of the game's release, applies only to Xbox,
GameCube, and PlayStation 2 games that are released late in the platform cycle
P
ln Fin status pub
Total sales (in GBP) of all games the publisher released (2-year window), [logged]
P
Awards dev
Number of game awards received by the developer (2-year window). Included game awards by
Academy of Interactive Arts & Sciences, British Academy of Film and Television Arts (BAFTA),
Game Developers Choice (GDC), Golden Joystick, IGN, and Spike.
O
Platform exclusive
Dummy = 1 if the game is released exclusively for one platform
P
Star
Dummy = 1 if the game features a real-life celebrity on box cover or title
O
Var critic within genre
Variance of expert scores of all games within the focal game's genre (2-year window)
M; P
ln Num games within genre
Number of games released within the focal game's genre (2-year window), [logged]
P
% games reviewed within genre pub
Percentage of the publisher's games reviewed by experts within the focal game's genre (2-year window) M; P
ln Genre exp pub
Number of games with the focal genre released by the publisher (2-year window), [logged]
P
ln Genre exp dev
Number of games with the focal genre developed by the developer (2-year window), [logged]
P
ln Num cowork
Number of games released jointly by the publisher & the developer (2-year window), [logged]
P
Game intro within genre-pf
Total number of games released one month prior to the focal game's release, on the focal game's
platform and within the focal game's genre
P
Missing quality
Dummy = 1 if the game has no score registered on Metacritic
M
Publisher dummies
Dummies for the publisher
P
Year dummies
Dummies for the release year of the game: 2002 - 2010.
P
Platform dummies
Dummies for the platform of the game, where platform is one of the eight platforms: Microsoft Xbox,
Microsoft Xbox 360, Nintendo DS, Nintendo GameCube, Nintendo Wii, Sony PlayStation 2, Sony
PlayStation 3, Sony PlayStation Portable
P
Release month dummies
Dummies for release month of the game: Jan. - Dec.
P
Genre dummies
Dummies for twelve genres: Action, Fighting, Music, Platform, Puzzle, Racing, RPG, Shooting,
Simulation, Sports, War, Misc.
P
Age restriction dummies
Dummies for five age ratings: Age 3, Age 7, Age 12, Age 16, Age 18. Sourced by British Board of Film
Classification (BBFC), Entertainment and Leisure Software Publishers Association (ELSPA), and Pan
European Game Information (PEGI).
P
Matching equation
ln Num cowork
Number of games released jointly by the publisher & the developer (2-year window), [logged]
P
ln Physical Dist
Distance (in km) between the HQ locations of the publisher & the developer, [logged]
O
ln Add genre exp pub dev
Additional genre count the publisher would gain if it acquired the developer (2-year window), [logged]
P
ln Num past acquisitions pub
Number of past developer acquisitions by the publisher (2-year window), [logged]
S; O
ln Fin status pub
Total sales (in GBP) of all games the publisher released (2-year window), [logged]
P
ln Genre count pub
Number of unique genres of games released by the publisher (2-year window), [logged]
P
ln Tie pub
Number of unique developers the publisher worked with (2-year window), [logged]
P
ln PF HHI pub
Average platform-level Herfindahl Index across all platforms on which the publisher released a game (2-
year window), [logged]
P
ln Num pf comp pub
Unique number of competing publishers across all platforms on which the publisher released a game (2-
year window), [logged]
P
ln Fin status dev
Total sales (in GBP) of all games the developer worked for (2-year window), [logged]
P
ln Genre count dev
Number of unique genres of games developed by the developer (2-year window), [logged]
P
ln Tie dev
Number of unique publishers the developer worked with (2-year window), [logged]
P
ln PF HHI dev
Average platform-level Herfindahl Index across all platforms on which the developer released a game
(2-year window), [logged]
P
ln Num pf comp dev
Unique number of competing developers across all platforms on which the developer released a game
(2-year window), [logged]
P
Awards dev
Number of game awards received by the developer (2-year window). Same game awards as above.
O; P
% New IP dev
Percentage of new IP games by the developer (2-year window)
O; P
% High quality dev
Percentage of high quality games by the developer (2-year window)
M; P
Legend: M = data comes from Metacritic; O = data comes from secondary online sources; P = data comes from proprietary platform owner dataset;
S = data comes from SDC Platinum
38
Table 2: Sample average of key outcome variables by developer type
Notes: ** p < 0.01, * p < 0.05, +
p < 0.10. Mean differences and significance levels reported from a two-tailed t-test with equal variances.
-0.125**
0.228**
0.212**
1,144,992.1**
291,870.1
0.098**
0.175**
-0.035
-0.093**
0.069**
0.126**
879,924.8**
Developer Type
Acquired
- Independent All
Later Acquired
- Always Independent
All
Independent
Acquired
Internally
founded
Independent
All
Always
Independent
Later
Acquired
Acquired
- Later Acquired
Mean Difference
Table 3: Summary statistics: match valuation
Variable Mean SD Min Max
Acquisition 2.08e-4 0.014 0 1
Publisher-Developer
ln Num cowork 0.007 0.089 0 3.43
ln Physical dist 7.59 1.70 0 9.39
ln Add genre exp pub dev 0.246 0.413 0 2.40
Publisher
ln Num past acquisitions pub 0.125 0.337 0 1.95
ln Fin status pub 8.20 7.96 0 20.1
ln Genre count pub 0.889 0.943 0 2.56
ln Tie pub 1.12 1.27 0 4.36
ln PF HHI pub 0.188 0.244 0 1.03
ln Num pf comp pub 2.06 1.94 0 4.25
Developer
ln Fin status dev 5.34 6.67 0 19.6
ln Genre count dev 0.347 0.467 0 2.40
ln Tie dev 0.350 0.467 0 2.48
ln PF HHI dev 0.033 0.064 0 0.719
ln Num pf comp dev 2.17 2.68 0 6.32
Awards dev 0.303 2.65 0 93
% New IP dev 0.145 0.322 0 1
% High quality dev 0.071 0.233 0 1
# observations
409,352
39
Table 4: The proposed model estimates: match valuation
Variable Mean SD dP/dW
Publisher-Developer
ln Num cowork
0.491**
0.071 0.139
ln Physical dist
-0.013+
0.008 -0.004
ln Add genre count pub dev
-0.042
0.065 -0.012
Publisher
ln Num past acquisitions pub
0.229**
0.048 0.065
ln Fin status pub
0.025**
0.010 0.007
ln Genre count pub
-0.065
0.076 -0.018
ln Tie pub
0.044
0.051 0.012
ln PF HHI pub
-0.041
0.101 -0.012
ln Num pf comp pub
-0.076*
0.037 -0.021
Developer
ln Fin status dev
0.015
0.010 0.005
ln Genre count dev
-0.08
0.105 -0.022
ln Tie dev
0.09
0.099 0.025
ln PF HHI dev
-0.214
0.314 -0.060
ln Num pf comp dev
-0.032
0.022 -0.009
Awards dev
0.001
0.005 2.20e-4
% New IP dev
0.018
0.062 0.005
% High quality dev
0.183*
0.086 0.052
% New IP dev X % High quality dev
-0.071 0.139 -0.020
Year dummies
# observations
409,352
Included
Match Valuation
Notes: **, *, +
: zero not contained in 1%, 5%, 10% credible interval, respectively. Year
fixed effects are included. We draw 30,000 parameters from the posterior distribution,
and throw the first 10,000 draws as burn-in. To reduce the serial correlation, we use
every 10th draw of the remaining 20,000 draws to compute the posterior mean and s.d.
Table 5: Summary statistics: outcome equations
Variable # observations Mean SD Min Max
Acquisition 5,916 0.164 0.371 0 1
New IP 5,916 0.239 0.427 0 1
High Quality 4,148 0.348 0.476 0 1
ln Sales in GBP 5,916 12.6 2.09 1.39 18.5
Reviewed by Experts 5,916 0.701 0.458 0 1
Founded Internally 5,916 0.203 0.402 0 1
Age focal pf in month 5,916 38.3 22.1 0 120
Age next gen pf in month 5,916 0.975 4.39 0 44
ln Fin status pub 5,916 13.8 1.21 2.82 16.1
Awards dev 5,916 3.73 9.42 0 93
Platform exclusive 5,916 0.316 0.465 0 1
Star 5,916 0.093 0.291 0 1
Var critic within genre 5,916 175.7 39.7 0 282.4
ln Num games within genre 5,916 4.56 1.00 0.693 5.92
% games reviewed within genre pub 5,916 0.637 0.390 0 1
ln Genre exp pub 5,916 3.90 1.03 0.693 5.38
ln Genre exp dev 5,916 1.91 0.824 0.693 4.01
ln Num cowork 5,916 1.20 1.19 0 4.25
Game intro within genre-pf 5,916 2.56 2.93 0 17
40
Table 6: The proposed model estimates: outcome equations
Variable Mean SD dF/dX Mean SD dF/dX Mean SD dF/dX Mean SD
Acquisition
-0.233**
0.078 -0.060 0.345** 0.132 0.082 1.101** 0.407 0.216 -0.127 0.093
New IP
-1.840**
0.223 -0.602 -2.978*1.172 -0.173 0.495*0.177
High Quality
2.238**
0.114
Founded Internally
-0.282**
0.059 -0.073 0.034 0.093 0.009 0.624+0.344 0.100 0.349** 0.071
Age focal pf in month
-0.008**
0.001 -0.002 -0.013** 0.003 -0.003 0.009 0.009 0.001 -0.021** 0.002
Age next gen pf in month
-0.013*
0.007 -0.004 -0.043** 0.008 -0.012 -0.034 0.043 -0.003 -0.047** 0.007
ln Fin status pub
-0.148**
0.036 -0.042 0.018 0.056 0.005 -0.393 0.255 -0.040 0.808** 0.045
Awards dev
-0.007**
0.003 -0.002 0.005 0.004 0.001 0.157** 0.021 0.017 3.65e-04 0.003
Platform exclusive
0.305**
0.047 0.089 -0.215** 0.074 -0.060 1.132** 0.306 0.174 -0.516** 0.059
Star
0.469**
0.112 0.104 0.920*0.402 0.178 0.547** 0.084
Var critic within genre
-0.002*
0.001 -0.001
ln Num games within genre
-0.195*
0.081 -0.053
% games reviewed within genre pub
0.558**
0.086 0.150
ln Genre exp pub
-0.590*
0.258 -0.061
ln Genre exp dev
-1.145**
0.226 -0.124
ln Num cowork
0.688**
0.168 0.075 0.015 0.024
Game intro within genre-pf -0.002 0.009
Dummy for missing quality
0.811**
0.168
Correlation
Match Valuation 0.034 0.030 -0.014 0.028 0.027 0.029
0.054**
0.025
New IP
0.731**
0.037 0.334*0.119 -0.315** 0.059
Review
-0.378**
0.107 0.154** 0.059
High Quality
-0.632**
0.035
Indirect effects of Acquisition (dF/dX)
A I
0.027**
0.008 0.023*0.012 -0.030*0.015
A R -0.003 0.015
A Q
0.644**
0.299
Total effect of Acquisition (dF/dX)
A
-0.060**
0.018 0.097** 0.021 0.287** 0.131 0.484*0.281
# observations
5,936
New IP
(Probit)
Review Selection
(Probit)
High Quality
(Probit)
ln Sales in GBP
(Linear)
Notes: **, *, +
: zero not contained in 1%, 5%, 10% credible interval, respectively. All four equations include fixed effects for publisher, year, platform, release month,
genre, and age restriction. We draw 30,000 parameters from the posterior distribution, and throw the first 10,000 draws as burn-in. To reduce the serial correlation, we
use every 10th draw of the remaining 20,000 draws to compute the posterior mean and s.d.
41
Table 7: Robustness checks: outcome equations
Variable Mean SD dF/dX Mean SD dF/dX Mean SD dF/dX Mean SD
1. Without matching model
Acquisition
-0.178**
0.057 -0.048 0.298** 0.080 0.075 1.265** 0.335 0.274 0.029 0.067
New IP
-1.676**
0.305 -0.555 -3.219** 1.157 -0.226 0.610** 0.179
High Quality
2.254**
0.117
2. Drop games by internally founded developers (N=4,718)
Acquisition
-0.207**
0.070 -0.580 0.375** 0.134 0.071 1.611** 0.499 0.076 -0.159 0.098
New IP
-2.919**
0.316 -0.829 -3.693** 1.019 -0.026 0.234 0.215
High Quality
2.359**
0.115
3. Average selling price in sales performance
Acquisition
-0.186*
0.092 -0.049 0.353*0.142 0.085 1.395** 0.532 0.195 -0.048 0.085
New IP
-1.795**
0.179 -0.592 -4.445** 1.083 -0.159 0.550** 0.185
High Quality
1.841**
0.101
Average Selling Price
0.101**
0.003
4. Competition measure in quality equation
Acquisition
-0.220*
0.092 -0.058 0.325** 0.130 0.074 1.167** 0.462 0.120 -0.132 0.095
New IP
-2.106**
0.224 -0.671 -4.520** 0.942 -0.134 0.511** 0.182
Game intro within genre-pf 0.041 0.039 0.002
High Quality
2.269**
0.114
5. Complementarities
Acquisition
-0.291**
0.082 -0.074 0.341*0.142 0.084 0.584 0.464 0.106 -0.047 0.103
Acquisition X ln Add genre count pub dev
0.140 0.118 0.043 -0.039 0.165 -0.014 1.568** 0.623 0.404 -0.247+0.128
New IP
-1.676**
0.262 -0.556 -3.143** 1.255 -0.202 0.589** 0.169
High Quality
2.216**
0.122
Notes: **, *, +
: zero not contained in 1%, 5%, 10% credible interval, respectively. All four equations include the remaining variables of the main specification. We draw
30,000 parameters from the posterior distribution, and throw the first 10,000 draws as burn-in. To reduce the serial correlation, we use every 10th draw of the remaining
20,000 draws to compute the posterior mean and s.d.
New IP
(Probit)
Review Selection
(Probit)
High Quality
(Probit)
ln Sales in GBP
(Linear)
42
Figure 1: Conceptual framework
Target
selection
Acquisition
Product
innovativeness
Product quality
Sales
performance
Main effect
Control effect
Figure 2: Industry evolution and the number of acquisitions
0
100
200
300
400
500
600
0
5
10
15
20
2002 2003 2004 2005 2006 2007 2008 2009 2010
Active firms
Acquisitions
Independent developers Publishers Acquisitions
43
Figure 3: Conceptual framework with results
!
Partner
selection
Acquisition
Product
innovativeness
Product quality
Sales
performance
Acquirer-target interactions:*
Prior alliances (.491)
Acquirer:
Prior acquisitions (.229)
Financial status (.025)
Platform competition (-.076)
Target:
Quality products (.183)
-.127
-2.798
ª Effects in bold indica te zero is not included in 10% credible interval.
* Only match valuation coefficients listed where zero is not included in
10% credible interval.
Figure 4: MCMC plot for key variables: match valuation
.2
.4
.6
.8
0 10000 20000 30000
ln Num cowork
.04
.02
0
.02
0 10000 20000 30000
ln Physical dist
.3
.2
.1
0
.1
.2
0 10000 20000 30000
ln Add genre count pub dev
0
.1
.2
.3
.4
0 10000 20000 30000
ln Num past acquisitions pub
.02
0
.02
.04
.06
0 10000 20000 30000
ln Fin status pub
.4
.2
0
.2
0 10000 20000 30000
ln Genre count pub
.1
0
.1
.2
.3
0 10000 20000 30000
ln Tie pub
.6
.4
.2
0
.2
.4
0 10000 20000 30000
ln PF HHI pub
.2
.15
.1
.05
0
.05
0 10000 20000 30000
ln Num pf comp pub
.02
0
.02
.04
0 10000 20000 30000
ln Fin status dev
.6
.4
.2
0
.2
0 10000 20000 30000
ln Genre count dev
.4
.2
0
.2
.4
0 10000 20000 30000
ln Tie dev
1.5
1
.5
0
.5
1
0 10000 20000 30000
ln PF HHI dev
.1
.05
0
.05
0 10000 20000 30000
ln Num pf comp dev
.02
.01
0
.01
.02
0 10000 20000 30000
Awards dev
.2
.1
0
.1
.2
.3
0 10000 20000 30000
% New IP dev
.2
0
.2
.4
0 10000 20000 30000
% High quality dev
.6
.4
.2
0
.2
.4
0 10000 20000 30000
% New IP dev X % High quality dev
Parameter Value
MCMC Plots
MCMC Iterations
44
Figure 5: MCMC plot for key variables: outcome equations
.04
.03
.02
.01
0
.01
0 10000 20000 30000
[I] Acquisition
.5
.4
.3
.2
.1
0
0 10000 20000 30000
[I] Founded Internally
.015
.01
.005
0
0 10000 20000 30000
[I] Age focal pf in month
.012
.01
.008
.006
.004
0 10000 20000 30000
[I] Age next gen pf in month
.1
.2
.3
.4
.5
0 10000 20000 30000
[I] ln Fin status pub
.3
.2
.1
0
0 10000 20000 30000
[I] Awards dev
.4
.6
.8
1
0 10000 20000 30000
[I] Platform exclusive
.2
.4
.6
.8
1
0 10000 20000 30000
[R] Acquisition
.4
.2
0
.2
.4
0 10000 20000 30000
[R] New IP
0
.2
.4
.6
.8
0 10000 20000 30000
[R] Founded Internally
.01
0
.01
.02
0 10000 20000 30000
[R] Age focal pf in month
.025
.02
.015
.01
.005
0 10000 20000 30000
[R] Age next gen pf in month
0
.2
.4
.6
.8
1
0 10000 20000 30000
[R] ln Fin status pub
.2
.1
0
.1
.2
0 10000 20000 30000
[R] Awards dev
0
.5
1
1.5
2
0 10000 20000 30000
[R] Platform exclusive
.6
.4
.2
0
.2
0 10000 20000 30000
[R] Star
.08
.06
.04
.02
0
0 10000 20000 30000
[R] Var critic within genre
.006
.004
.002
0
.002
0 10000 20000 30000
[R] ln Num games within genre
.6
.4
.2
0
.2
0 10000 20000 30000
[R] % games reviewed within genre pub
.2
.1
0
.1
.2
0 10000 20000 30000
[Q] Acquisition
1
0
1
2
0 10000 20000 30000
[Q] New IP
0
.5
1
1.5
2
2.5
0 10000 20000 30000
[Q] Founded Internally
.05
.1
.15
.2
.25
0 10000 20000 30000
[Q] Age focal pf in month
.02
0
.02
.04
0 10000 20000 30000
[Q] Age next gen pf in month
2
1.5
1
.5
0
0 10000 20000 30000
[Q] ln Fin status pub
1.5
1
.5
0
.5
0 10000 20000 30000
[Q] Awards dev
0
1
2
3
4
5
0 10000 20000 30000
[Q] Platform exclusive
0
.5
1
1.5
2
2.5
0 10000 20000 30000
[Q] Star
0
.5
1
1.5
0 10000 20000 30000
[Q] ln Genre exp pub
1.5
1
.5
0
.5
0 10000 20000 30000
[Q] ln Genre exp dev
1
0
1
2
0 10000 20000 30000
[Q] ln Num cowork
.07
.06
.05
.04
.03
.02
0 10000 20000 30000
[Y] Acquisition
.1
.2
.3
.4
.5
.6
0 10000 20000 30000
[Y] New IP
.5
0
.5
1
0 10000 20000 30000
[Y] High Quality
.6
.4
.2
0
.2
0 10000 20000 30000
[Y] Founded Internally
.01
.005
0
.005
.01
.015
0 10000 20000 30000
[Y] Age focal pf in month
.03
.025
.02
.015
0 10000 20000 30000
[Y] Age next gen pf in month
.1
.05
0
.05
.1
0 10000 20000 30000
[Y] ln Fin status pub
.6
.7
.8
.9
1
0 10000 20000 30000
[Y] Awards dev
.04
.02
0
.02
.04
0 10000 20000 30000
[Y] Platform exclusive
.7
.6
.5
.4
.3
0 10000 20000 30000
[Y] Star
.2
.4
.6
.8
0 10000 20000 30000
[Y] ln Num cowork
.2
.4
.6
.8
1
1.2
0 10000 20000 30000
[Y] Game intro within genrepf
.5
1
1.5
2
2.5
0 10000 20000 30000
[Y] Dummmy for missing quality
Parameter Value
MCMC Plots
MCMC Iterations
45
Appendix A List of acquisition deals: 2002-2010
Table A1 shows the details of all 85 acquisition deals observed during our sample period.
Appendix B Error distributions
Without loss of generality, we assume the following error structure. If game kis released by
publisher iand developer jwho has been acquired by publisher iin year t, then we need to control
for the correlation between ηijt and ijkτ (I
ijkτ , R
ijkτ , Q
ijkτ , Y
ijkτ ) for all games kreleased in year
τt. Let
ηijt N(0,1)
I
ijkτ =ρIηij t +νI
ijkτ , νI
ijkτ N(0,1)
R
ijkτ =ρRηij t +σIRνI
ijkτ +νR
ijkτ , νR
ijkτ N(0,1)
Q
ijkτ =ρQηij t +σIQνI
ijkτ +σRQ νR
ijkτ +νQ
ijkτ , νQ
ijkτ N(0,1)
Y
ijkτ =ρYηij t +σIY νI
ijkτ +σRY νR
ijkτ +σQY νQ
ijkτ +νY
ijkτ , ν Y
ijkτ N(0, σ2
Y),
where we assume ηijt,νI
ijkτ ,νI
ijkτ ,νI
ijkτ , and νY
ijkτ are i.i.d. across i, j, k, t, and τ. The distribution
of (ηijt , I
ijkτ , R
ijkτ , Q
ijkτ , Y
ijkτ ) is given by
ηijt
I
ijkτ
R
ijkτ
Q
ijkτ
Y
ijkτ
N0,1,
where
1=
1ρIρRρQρY
·ρ2
I+ 1 ρIρR+σI R ρIρQ+σI Q ρIρY+σIY
· · ρ2
R+σ2
IR + 1 ρRρQ+σI RσI Q +σRQ ρRρY+σI RσI Y +σRY
· · · ρ2
Q+σ2
IQ +σ2
RQ + 1 ρQρY+σI QσI Y +σRQσRY +σQY
· · · · ρ2
Y+σ2
IY +σ2
RY +σ2
QY +σ2
Y
.
In our data set, some of the acquired developers develop multiple games after acquisition.
Let Gijt be the set of post-acquisition games developed by developer jwho has been acquired
by publisher iat time t, and let M≡ |Gijt |. Our above construction implies that conditional
on ηijt ,ijkτ are i.i.d. across kGij t. For notational convenience, we re-index kGijt as
46
k1, . . . , kMand let tmbe the release year of game kmfor m= 1, . . . , M . The joint distribution of
(ηijt , ijk1t1, . . . , ij kMtM) is normal with mean zero and the covariance matrix ΩMis given by
M=
1 Σ01 . . . Σ0m. . . Σ0M
Σ10 Σ11 . . . Σ1m. . . Σ1M
. . . . . . . . . . . . . . . . . .
Σm0Σm1. . . Σmm . . . ΣmM
. . . . . . . . . . . . . . . . . .
ΣM0ΣM1. . . ΣM m . . . ΣM M
,
where (1) Σ0m= (ρI, ρR, ρQ, ρY)m; (2) Σmm mis given by a 4×4 matrix obtained by deleting
the first row and column of Ω1; and (3) Σmm0m, m0with m6=m0is given by
Σmm0=
ρ2
IρIρRρIρQρIρY
·ρ2
RρRρQρRρY
· · ρ2
QρQρY
· · · ρ2
Y
.
We note that some games are not reviewed by experts and thus there is no quality equation.
For those games, we make an adjustment to the covariance matrix by removing the error of the
quality equation.
In the Gibbs sampling procedure, we take ΩMas the unconditional covariance matrix, and
compute appropriate mean and covariance matrix of the conditional distributions given the param-
eters and augmented latent variables for the matching and Probit models. Finally, for games by
independent developers, we assume no correlation between ηijt and ijkt . However, we still assume
correlations among (I
ijkt , R
ijkt , Q
ijkt , Y
ijkt ) and obtain the corresponding covariance matrix using a
similar approach as above.
47
Table A1: List of acquisition deals: 2002-2010
Target Buyer Data source Year Target location Games in sample %New IP %High quality Avg. revenues
1 7 STUDIOS ACTIVISION BLIZZARD SDC Platinum 2009 Germany 14 21.43% 0.00% 990,744.86£
2 BEENOX ACTIVISION BLIZZARD SDC Platinum 2005 Italy 14 0.00% 14.29% 630,211.50£
3 BIZARRE CREATIONS ACTIVISION BLIZZARD SDC Platinum 2007 France 11 45.45% 63.64% 3,854,825.36£
4 BUDCAT CREATIONS ACTIVISION BLIZZARD SDC Platinum 2008 United States 16 25.00% 37.50% 1,029,841.69£
5 FREESTYLEGAMES ACTIVISION BLIZZARD SDC Platinum 2008 United Kingdom 10 20.00% 70.00% 1,049,111.50£
6 HIGH MOON STUDIOS ACTIVISION BLIZZARD SDC Platinum 2007 United States 6 33.33% 50.00% 987,275.83£
7 INFINITY WARD ACTIVISION BLIZZARD SDC Platinum 2003 United States 6 0.00% 83.33% 36,161,494.17£
8LUXOFLUX ACTIVISION BLIZZARD SDC Platinum 2002 United States 13 23.08% 38.46% 3,330,964.08£
9 SHABA ACTIVISION BLIZZARD SDC Platinum 2002 United States 8 0.00% 37.50% 1,000,327.25£
10 SWINGIN' APE STUDIOS ACTIVISION BLIZZARD SDC Platinum 2005 United States 3 100.00% 100.00% 169,731.00£
11 TOYS FOR BOB ACTIVISION BLIZZARD SDC Platinum 2005 United States 10 0.00% 30.00% 988,538.40£
12 UNDERGROUND DEVELOPMENT ACTIVISION BLIZZARD SDC Platinum 2002 United States 16 37.50% 31.25% 447,083.56£
13 VICARIOUS VISIONS ACTIVISION BLIZZARD SDC Platinum 2005 U nited States 45 6.67% 28.89% 1,811,123.42£
14 EDEN STUDIOS ATARI SDC Platinum 2002 France 7 14.29% 42.86% 1,554,067.86£
15 KROME STUDIOS ATARI Other 2006 Australia 31 19.35% 3.23% 890,058.71£
16 ARKANE STUDIOS BETHESDA SOFTWORKS SDC Platinum 2010 Germany 1 100.00% 0.00% 65,218.00£
17 ID SOFTWARE BETHESDA SOFTWORKS SDC Platinum 2009 United States 1 0.00% 100.00% 1,900,772.00£
18 BLUE CASTLE GAMES CAPCOM SDC Platinum 2010 Canada 5 20.00% 60.00% 2,570,127.60£
19 K2 CAPCOM SDC Platinum 2008 Japan 9 11.11% 11.11% 646,933.33£
20 SEGA RACING STUDIO CODEMASTERS Other 2008 United Kingdom 2 0.00% 100.00% 1,991,044.00£
21 VICIOUS CYCLE SOFTWARE D3P Other 2007 United States 13 38.46% 15.38% 74,544.38£
22 AVALANCHE SOFTWARE DISNEY INTERACTIVE STUDIOS SDC Platinum 2005 France 28 10.71% 21.43% 637,564.00£
23 BLACK ROCK STUDIOS DISNEY INTERACTIVE STUDIOS SDC Platinum 2006 United Kingdom 55 20.00% 21.82% 682,910.91£
24 JUNCTION POINT STUDIOS DISNEY INTERACTIVE STUDIOS SDC Platinum 2007 United States 1 0.00% 0.00% 3,646,322.00£
25 WIDELOAD GAMES DISNEY INTERACTIVE STUDIOS Other 2009 United States 1 100.00% 100.00% 431,666.00£
26 CORE DESIGN EIDOS Other 2006 United Kingdom 2 50.00% 0.00% 5,239,682.50£
27 BIOWARE ELECTRONIC ARTS SDC Platinum 2008 United States 10 40.00% 70.00% 3,349,314.50£
28 CRITERION GAMES ELECTRONIC ARTS SDC Platinum 2004 United Kingdom 22 18.18% 90.91% 4,152,851.64£
29 DIGITAL ILLUSIONS ELECTRONIC ARTS SDC Platinum 2004 Sweden 14 28.57% 64.29% 3,478,305.43£
30 HYPNOTIX ELECTRONIC ARTS Other 2005 United States 10 40.00% 10.00% 113,990.70£
31 NUFX INC ELECTRONIC ARTS SDC Platinum 2004 United States 3 0.00% 100.00% 398,899.00£
32 PANDEMIC STUDIOS ELECTRONIC ARTS SDC Platinum 2007 United States 22 31.82% 36.36% 3,064,525.55£
33 STUDIO 33 ELECTRONIC ARTS SDC Platinum 2003 United Kingdom 1 0.00% 0.00% 332,920.00£
34 VR1 JALECO SDC Platinum 2002 United States 1 100.00% 0.00% 177,494.00£
35 HUDSON KONAMI SDC Platinum 2005 Japan 40 30.00% 12.50% 552,038.43£
36 CRYO INTERACTIVE MC2-MICROIDS Other 2008 France 1 0.00% 0.00% 139,685.00£
37 BIGPARK MICROSOFT Other 2009 United Kingdom 1 100.00% 0.00% 4,754,946.00£
38 LIONHEAD STUDIOS MICROSOFT SDC Platinum 2006 United Kingdom 4 25.00% 75.00% 9,170,226.75£
39 RARE MICROSOFT SDC Platinum 2002 United Kingdom 11 36.36% 72.73% 4,083,027.45£
40 PITBULL SYNDICATE MIDWAY SDC Platinum 2005 United Kingdom 7 28.57% 0.00% 1,065,504.00£
41 COKTEL VISION MINDSCAPE UK SDC Platinum 2005 France 1 100.00% 0.00% 42,553.00£
42 KAOLINK MINDSCAPE UK Other 2008 France 6 50.00% 0.00% 451,275.00£
43 D3 PUBLISHER NAMCO BANDAI GAMES SDC Platinum 2009 Japan 25 84.00% 0.00% 80,395.20£
44 NAMCO NAMCO BANDAI GAMES SDC Platinum 2006 Japan 79 20.25% 24.05% 1,181,184.15£
45 RETRO STUDIOS NINTENDO SDC Platinum 2002 United States 5 0.00% 100.00% 4,179,223.20£
46 PIVOTAL GAMES SCI SDC Platinum 2003 United Kingdom 14 21.43% 0.00% 2,256,635.00£
47 CREATIVE ASSEMBLY SEGA SDC Platinum 2005 United Kingdom 7 57.14% 0.00% 1,124,261.00£
48 SECRET LEVEL SEGA Other 2006 United States 9 0.00% 0.00% 476,229.00£
49 SMILEBIT SEGA Other 2004 Japan 4 50.00% 25.00% 497,790.25£
50 SPORTS INTERACTIVE SEGA SDC Platinum 2006 United Kingdom 11 0.00% 9.09% 1,501,806.36£
51 BIGBIG STUDIOS SONY COMPUTER ENT. SDC Platinum 2007 Japan 4 25.00% 50.00% 1,228,125.00£
52 EVOLUTION STUDIOS SONY COMPUTER ENT. SDC Platinum 2007 United Kingdom 6 16.67% 33.33% 3,868,331.33£
53 GUERRILLA GAMES SONY COMPUTER ENT. SDC Platinum 2005 Netherlands 5 60.00% 40.00% 4,053,495.20£
54 INCOGNITO SONY COMPUTER ENT. SDC Platinum 2002 United States 5 40.00% 100.00% 765,546.20£
55 MEDIA MOLECULE SONY COMPUTER ENT. SDC Platinum 2010 United Kingdom 2 50.00% 50.00% 9,044,240.00£
56 ZIPPER INTERACTIVE SONY COMPUTER ENT. SDC Platinum 2006 United States 7 14.29% 85.71% 2,356,331.43£
57 IO INTERACTIVE SQUARE ENIX EUROPE SDC Platinum 2004 Denmark 19 42.11% 63.16% 2,366,934.16£
58 TAITO SQUARE ENIX EUROPE Other 2005 Japan 25 40.00% 12.00% 290,638.40£
59 CAT DADDY TAKE 2 Other 2003 United States 4 75.00% 0.00% 779,956.00£
60 FIRAXIS TAKE 2 SDC Platinum 2005 United States 4 0.00% 100.00% 1,178,874.75£
61 INDIE BUILT TAKE 2 Other 2004 United States 2 0.00% 50.00% 1,020,724.00£
62 IRRATIONAL GAMES TAKE 2 SDC Platinum 2006 United States 4 50.00% 100.00% 5,172,525.50£
63 KUSH GAMES TAKE 2 SDC Platinum 2005 United States 7 0.00% 42.86% 178,489.86£
64 PAM DEVELOPMENT TAKE 2 Other 2006 France 10 40.00% 40.00% 1,097,711.30£
65 POPTOP SOFTWARE TAKE 2 SDC Platinum 2005 United States 1 100.00% 0.00% 112,716.00£
66 VENOM GAMES TAKE 2 Other 2004 United Kingdom 4 25.00% 25.00% 1,565,077.75£
67 VISUAL CONCEPTS TAKE 2 SDC Platinum 2005 United States 58 5.17% 70.69% 174,347.38£
68 TECMO TECMO KOEI EUROPE SDC Platinum 2009 Japan 13 46.15% 7.69% 250,408.62£
69 BLUE TONGUE THQ SDC Platinum 2004 Australia 18 5.56% 11.11% 566,206.39£
70 HEAVY IRON STUDIOS THQ Other 2009 United States 29 0.00% 3.45% 1,270,053.41£
71 JUICE GAMES THQ SDC Platinum 2006 United Kingdom 8 25.00% 0.00% 2,165,078.75£
72 OUTRAGE THQ SDC Platinum 2002 United States 3 66.67% 0.00% 114,364.00£
73 RAINBOW STUDIOS THQ SDC Platinum 2004 United States 24 25.00% 41.67% 667,609.75£
74 RELIC THQ SDC Platinum 2004 Canada 1 100.00% 0.00% 1,704,536.00£
75 UNIVERSOMO THQ SDC Platinum 2007 Finland 1 100.00% 0.00% 23,899.00£
76 VIGIL GAMES THQ SDC Platinum 2006 United States 2 100.00% 100.00% 1,991,570.50£
77 GAMELOFT UBISOFT SDC Platinum 2008 France 13 53.85% 0.00% 549,037.54£
78 SOUTHLOGIC STUDIOS UBISOFT SDC Platinum 2009 Brazil 4 0.00% 0.00% 921,515.50£
79 RADICAL ENTERTAINMENT VIVENDI GAMES SDC Platinum 2005 Canada 37 13.51% 18.92% 1,974,570.51£
80 SWORDFISH STUDIOS VIVENDI GAMES SDC Platinum 2005 United Kingdom 3 33.33% 0.00% 879,823.00£
81 MONOLITH WARNER BROS. INTERACTIVE SDC Platinum 2004 United States 7 14.29% 71.43% 1,513,484.00£
82 ROCKSTEADY STUDIOS WARNER BROS. INTERACTIVE Other 2010 United Kingdom 6 33.33% 33.33% 3,184,147.50£
83 SNOWBLIND STUDIOS WARNER BROS. INTERACTIVE Other 2009 United States 7 0.00% 57.14% 523,803.29£
84 TT GAMES PUBLISHING WARNER BROS. INTERACTIVE Other 2007 United Kingdom 68 1.47% 36.76% 2,530,135.35£
85 HOTHOUSE CREATIONS ZUSHI GAMES Other 2004 United Kingdom 3 66.67% 0.00% 1,486,465.00£
48
... However, as we saw, since the optimal δ is independent of p h at least for an interior solution, we expect that the impact of γ on δ will still remain unchanged. 13 We note that it is possible that in-house software is developed by an independent software developer and published by a platform provider (see Gil and Warzynski 2015;Ishihara and Rietveld 2017). In this study, we focus on publisher identity because the decision to release a game is made by publishers. ...
Article
Full-text available
This paper examines the role of software piracy in digital platforms where a platform provider makes a decision of how much software to produce in-house and how much to outsource from a third-party software provider. Using a vertical differentiation model, we theoretically investigate how piracy influences the software outsourcing decision. We find that when piracy is intermediate, the loss in in-house software profits due to piracy outweighs the loss in licensing fee profits. As a result, an increase in piracy leads to more outsourcing. However, when piracy is high, it becomes too expensive for the platform provider to subsidize the software provider, resulting in a decrease in outsourcing. Moreover, when software variety is also endogenously chosen by firms, the platform provider’s incentive to develop software variety in-house depends not only on the return from software profits but also on the return from hardware profits. Under such a situation, an increase in piracy always leads to less outsourcing and less total software variety. To provide additional insights on the outsourcing decision, we conduct empirical analyses using data from the U.S. handheld video game market between 2004 and 2012. This market is a classical two-sided market, dominated by two handheld platforms (Nintendo DS and Sony PlayStation Portable) and is known to have suffered from software piracy significantly. Our regression results show that in this market, piracy increases outsourcing but has no effect on the total software variety.
Article
Full-text available
Past research has primarily focused on what happens after a merger. This research attempts to determine whether anticipated benefits from the merger actually accrue. We characterize the effects of observed variables on whether pairs of firms merge, vis-à-vis roommate matching, and then link these factors to post-merger innovation (i.e., number of patents). We jointly estimate the two models using Markov Chain Monte Carlo methods with a unique panel data set of 1,979 mergers between 4,444 firms across industries and countries from 1992 to 2008. We find that similarity in national culture and technical knowledge has a positive effect on partner selection and post-merger innovation. Anticipated synergy from subindustry similarity, however, is not realized in post-merger innovation. Furthermore, some key synergy sources are unanticipated when selecting a merger partner. For example, financial synergy from higher total assets and complementarity in total assets and debt leverage as well as knowledge synergy from breadth and depth of knowledge positively influence innovation but not partner selection. Furthermore, factors that dilute synergy (e.g., higher debt levels) are unanticipated, and firms merge with firms that detract from their innovation potential. Overall, the results reveal some incongruity between anticipated and realized synergy. Data, as supplemental material, are available at https://doi.org/10.1287/mksc.2016.0978 .
Article
Full-text available
Building on the structural two-sided matching model, we develop a framework to study the sourcing market in the context of marketing firms matching with manufacturers. Both sides prefer partners that could generate significant values with better sourcing process abilities. Moreover, experienced manufacturers are preferred by the branded marketing firms who may even be willing to compensate the matching intermediary more for facilitating that preference. Empirical research, measuring the values of such matching and the intermediary’s pricing (through commission) on observed marketing firms’ characteristics and the low-cost manufacturers and the deals that result, is problematic, when some of the characteristics are only partially observed and the matching is endogenous. With the matching model, we can control for endogenous matching. We find evidence of positive assortative matching of pairs’ size on both sides of the market. We also find that manufacturers’ location and tenure, and whether the marketing firms are listed, are important factors in identifying the preferred matching partners and the related ranking. Without controlling for endogenous matching, the intermediary’s pricing equation estimates are biased, especially for the marketing firms that specialize in luxury products.
Article
Full-text available
sResearch SummaryCorporate acquisition is a popular strategic option for firms seeking new resources. However, little research exists on the question of why one firm is chosen over another. We develop a model relating characteristics of similarity and complementarity between acquirers’ and target firms’ key resources, including products and R&D pipelines, to the likelihood of choosing a particular firm. We construct measures of similarity and complementarity between and across products and R&D pipelines, and test their effects using a novel application of the choice model. Findings reveal that acquirers view similarity and complementarity differently, based on the resource they are comparing. When making comparisons to their own R&D pipelines, acquirers prefer similarity over complementarity whereas when making comparisons to their product portfolios, they prefer complementarity over similarity.Managerial SummaryCorporate acquisition is a popular way for firms to grow and obtain innovative resources. However, we know little about why acquirers choose one firm over another. We capture the influence of similarity and complementarity between acquirers’ and target firms’ products (current innovative value) and R&D pipelines (future innovative value) on whether a particular target firm is acquired. Insights from the Pharmaceutical industry reveal that acquirers value similarity and complementarity in target firms differently, based on whether the comparison being made is with respect to their products or their R&D pipelines. Regarding their R&D pipelines, acquirers prefer that the target firm has similar, rather than complementarity, resources. However, the opposite is true concerning their own products: acquirers prefer that the target firm has complementarity, versus similar, resources.
Article
Retrospective studies of horizontal mergers have focused on their price effects, leaving the important question of how mergers affect product quality largely unanswered. This paper empirically investigates this issue for two recent airline mergers. Consistent with the theory that mergers facilitate coordination but diminish competitive pressure for quality improvement, we find that each merger is associated with a quality decrease (increase) in markets where the merging firms had (had no) pre-merger competition with each other, and the quality change can have a U-shaped relationship with pre-merger competition intensity. Consumer gains/losses associated with quality changes, which we monetize, are substantial.
Article
This paper empirically investigates the effects of foreign acquisitions on several performance measures of Chinese target firms. Using a self-constructed database that includes information on foreign acquisitions in China and the accounting information of Chinese manufacturing firms from 1998 to 2007, we find that foreign acquisitions significantly improve the productivity, sales and fixed asset investment of the target firms. We address the potential endogeneity issue in the OLS estimation using the difference-in-differences technique, with two control groups, namely the would-be targets and the propensity-score-matched targets. We also find that the performance-enhancing effect of foreign acquisitions becomes stronger when larger technology gaps are observed between the acquirers and the targets. An inverted-U relationship is observed between the post-acquisition performance of the target firms and the target firm equity that is held by the foreign acquirers. The performance effects are qualitatively different among vertical, horizontal and conglomerate acquisitions.
Article
We show how knowledge-based and incentive-based perspectives complement each other to explain the effects of acquisitions on the productivity of inventors from acquired firms. Incentive-based theories account for their lower productivity relative to that of inventors at nonacquired firms, and both perspectives jointly explain why their productivity converges with that of inventors from acquiring firms. Higher productivity is achieved when there is greater overlap in routines and moderate overlap in skills, and when the acquired firm is large relative to its acquirer. This study clarifies the subtle manner in which incentives and the knowledge-based view are intertwined.
Article
This paper examines the nature of the core capabilities of a firm, focusing in particular on their interaction with new product and process development projects. Two new concepts about core capabilities are explored here. First, while core capabilities are traditionally treated as clusters of distinct technical systems, skills, and managerial systems, these dimensions of capabilities are deeply rooted in values, which constitute an often overlooked but critical fourth dimension. Second, traditional core capabilities have a down side that inhibits innovation, here called core rigidities. Managers of new product and process development projects thus face a paradox: how to take advantage of core capabilities without being hampered by their dysfunctional flip side. Such projects play an important role in emerging strategies by highlighting the need for change and leading the way. Twenty case studies of new product and process development projects in five firms provide illustrative data.
Article
Do acquisitions increase, decrease, or have no effect on innovation? The empirical research on this question suggests that acquisitions may hurt innovation; that is, they may be a "poison pill" for innovation. The authors present an alternative view. For firms that first engage in internal knowledge development, the knowledge-based view the authors present suggests that acquisitions can help innovation; that is, they can be a tonic for innovation. Analysis of cross-sectional, time-series data on a sample of pharmaceutical firms during 1988-97 provides evidence to support the thesis.