This article was downloaded by: [218.89.242.

93] On: 20 September 2024, At: 05:13

Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA

Organization Science
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org

Open Source Collaboration in Digital

Entrepreneurship
Yu-Kai Lin, Likoebe M. Maruping

To cite this article:

Yu-Kai Lin, Likoebe M. Maruping (2022) Open Source Collaboration in Digital Entrepreneurship. Organization
Science 33(1):212-230. https://doi.org/10.1287/orsc.2021.1538

Full terms and conditions of use: https://pubsonline.informs.org/Publications/Librarians-Portal/PubsOnLine-

Terms-and-Conditions

This article may be used only for the purposes of research, teaching, and/or private study. Commercial use or
systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher approval,
unless otherwise noted. For more information, contact [email protected].

The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or support
of claims made of that product, publication, or service.

Copyright © 2021, INFORMS

Please scroll down for article—it is on subsequent pages

With 12,500 members from nearly 90 countries, INFORMS is the largest international association of operations
research (O.R.) and analytics professionals and students. INFORMS provides unique networking and learning
opportunities for individual professionals, and organizations of all types and sizes, to better understand and use
O.R. and analytics tools and methods to transform strategic visions and achieve better outcomes.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
 ORGANIZATION SCIENCE
Vol. 33, No. 1, January–February 2022, pp. 212–230
http://pubsonline.informs.org/journal/orsc ISSN 1047-7039 (print), ISSN 1526-5455 (online)

Open Source Collaboration in Digital Entrepreneurship

Yu-Kai Lin,a Likoebe M. Marupinga
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

a
Robinson College of Business, Georgia State University, Atlanta, Georgia 30303
Contact: [email protected], https://orcid.org/0000-0002-3928-9847 (Y-KL); [email protected], https://orcid.org/0000-0001-5105-6635
(LMM)

Received: January 10, 2020 Abstract. Emerging digital technologies give rise to digital entrepreneurship and the
Revised: December 12, 2020; June 10, 2021; widespread phenomenon of open source collaboration (OSC) on GitHub for entrepreneur-
August 11, 2021 ial pursuits. Although openness is a common theme in digital entrepreneurship, it is un-
Accepted: October 2, 2021 clear how digital startups—that is, startups that that have digital artifacts at the core of
Published Online in Articles in Advance: their business model for value creation and capture—actually realize value from their OSC
November 22, 2021
engagement. We develop a theoretical framework to explain how the engagement in OSC
https://doi.org/10.1287/orsc.2021.1538 may affect the value of digital startups and how the effect is contingent on the stage of ven-
ture maturity (conception, commercialization, or growth) and the mode of OSC engage-
Copyright: © 2021 INFORMS ment (inbound or outbound). In analyses that pool 17,552 matched digital startups with
monthly panel observations between 2008 and 2017, we find digital startups in the concep-
tion and commercialization stages benefit more from inbound OSC whereas the ones in
the growth stage benefit more from outbound OSC. As digital startups increasingly use
OSC for ideation, experimentation, and scaling, our contribution is to show whether,
when, and how knowledge flows through OSC might affect the value of digital startups.
We discuss implications for research on organizing for digital entrepreneurship as well as
open innovation.

History: This paper has been accepted for the Special Issue on Emerging Technologies and Organizing.

Keywords: open innovation • entrepreneurship • organizing for innovation in the digitized world • digital innovation •
digital entrepreneurship • open source collaboration • paradox of openness

“Without a doubt, the most exciting technological hand, the affordances and generativity of digital tech-
innovations occurring today are taking shape in open nologies and platforms drastically reduce the costs of
source communities.” —Jim Whitehurst, President open collaboration and at the same time, increase the
of IBM benefits of sharing and cocreation (Zittrain 2006, Yoo
et al. 2012).
Although openness has been studied extensively
Introduction in the open innovation literature (Chesbrough 2003,
There is growing research interest in digital entrepre- Bogers et al. 2017), we are only beginning to under-
neurship among business scholars (Huang et al. 2017, stand how digital startups—which we define as start-
Nambisan 2017, von Briel et al. 2018). This research ups that have digital artifacts at the core of their
examines how and why emerging digital technologies business model for value creation and capture—can
change the ways in which entrepreneurs identify, pur- organize to use openness to garner value.1 However,
sue, and develop new business opportunities. Studies the consequences of openness, and the management
have found that a salient theme from the digitalization thereof, in digital entrepreneurship are not obvious.
of entrepreneurial processes and outcomes is the Indeed, extant theories and evidence for open innova-
emphasis on openness (Nambisan et al. 2018). As tion largely come from researching large and mature
Nambisan (2017) explained, for two broad categories firms, and until recently, scholars paid relatively less
of reasons, digital technologies have rendered entre- attention to the roles and context of digitalization for
preneurial boundaries to be more fluid and entrepre- open innovation (Bogers et al. 2017, Dahlander et al.
neurial agency to be more distributed. On one hand, 2021). Despite the promises of openness, digital start-
because digital artifacts are editable and expansible, ups may confront a different calculus on whether and
the scope, features, and value of digital products and how to engage in openness on the one hand, and the
services can rapidly evolve with inputs from custom- value that they can extract from their open innovation
ers and contributions from third-party complementors engagement on the other (Stam 2009, Greul et al.
(Yoo et al. 2010, Kallinikos et al. 2013). On the other 2018).

212
 Lin and Maruping: Open Source Collaboration in Digital Startups
Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS 213

In examining openness in digital entrepreneurship, venture maturity (conception, commercialization, and

we focus on digital startups’ open source collabora- growth; cf., Kazanjian 1988, Fisher et al. 2016). We the-
tion (OSC) on GitHub, that is, the purposeful yet orize how digital startups at early versus late stages
loosely connected interaction and exchange between of maturity have different needs vis-à-vis their inno-
digital startups and open source communities on Gi- vation process in a way that we believe can affect
tHub regarding the use and development of open the value they can gain from different modes of
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

source software (OSS). Although there are many ways OSC engagement. In particular, we develop and test
to approach open innovation (e.g., alliances, licensing, hypotheses that digital startups in the conception
and user communities), OSC on GitHub is arguably and commercialization stages benefit more from in-
the most important one for digital startups. Given the bound OSC whereas those in the growth stage benefit
central role of digital technologies in digital entrepre- more from outbound OSC. Our longitudinal analysis
neurship, software undoubtedly is critical in the de- of digital startups’ OSC engagement on GitHub be-
velopment of these technologies. Meanwhile, open tween 2008 and 2017 provides strong support for our
source is a key driving force in the advance of modern framework.
digital innovations, such as artificial intelligence (AI), This research seeks to make three main contribu-
big data, and cloud services. The vast majority of OSS tions related to emerging technologies and organizing.
projects are hosted on GitHub because the platform First, we provide a nuanced understanding of open-
drastically reduces the barriers and increases the val- ness in digital entrepreneurship. Although emerging
ue of OSC. As a result, a growing number of digital digital technologies, such as platforms like GitHub,
startups are leveraging GitHub for ideation, experi- enable entrepreneurs to acquire resources, create val-
mentation, and scaling. Indicative of the promise of ue, and scale business from open collaboration, open-
GitHub to startups’ open innovation initiatives, 15 out ness also gives rise to tension and conflict between
of the top 20 global unicorns (i.e., startups with at least value creation and value capture in entrepreneurial
a $1 billion valuation) have public code repositories pursuits (Nambisan et al. 2018). We elucidate the
on GitHub, and most of them have participated in paradoxical nature of openness in digital entrepre-
OSC on GitHub since their early venture stage before neurship and develop a novel conceptual framework
they achieved unicorn status.2 to explain how the value of openness manifests in dig-
Theoretically and empirically, it remains an open
ital entrepreneurship. Second, this research is closely
question whether and how digital startups realize val-
related to the paradox of openness literature. As we
ue from their OSC. Indeed, Felin and Zenger (2014,
will show in the next section, a key limitation in the
2020) argue that openness is not a free and obvious
extant approaches to the paradox of openness is
resource and should be conceived as a purposeful
the lack of a temporal/dynamic lens. Adding to this
activity. Central to the question is how digital startups
stream of work, we provide an alternative conceptual-
navigate “the paradox of openness” (Laursen and
ization that takes time into account when approaching
Salter 2014, Wadhwa et al. 2017). The paradox is root-
the paradox (Smith et al. 2017). Third, we contribute
ed in two competing but interdependent innovation
to the open innovation literature. Our theorizing and
objectives: sharing is critical for value creation but
evidence of OSC by digital startups addresses salient
protecting is vital for value capture. As West and
Gallagher (2006) observed, sharing knowledge and research gaps regarding open innovation among start-
software through OSC could create value for firms be- ups (Laursen and Salter 2014), through digital plat-
cause it attracts improvements and complements from forms (Dahlander et al. 2021), differentiating inbound
OSS communities, but whatever is open sourced can and outbound openness and considering both at the
be easily imitated by current or potential rivals, making same time (Greul et al. 2018), or with a longitudinal
it more difficult for the firms to extract rents from their research design for unveiling the temporal dimension
investments. This is clearly a challenge for digital start- (Wadhwa et al. 2017).
ups—one that raises fundamental questions of interest
to organization scholars about organizing for digital Conceptual Background
entrepreneurship and organizing for open innovation. Digital Entrepreneurship
To advance our understanding of OSC in digital en- Digital entrepreneurship is an important global phe-
trepreneurship, we follow the strategies proposed by nomenon shaped by digital technologies. As digital
Poole and Van de Ven (1989) for theory building from enablers, von Briel et al. (2018) explain that digital tech-
paradox. We attempt a theoretical framework to dis- nologies involve six enabling mechanisms in new
aggregate and resolve the paradox of openness by venture creation: compress time, conserve resources,
spatially separating two modes of OSC engagement expand availability, substitute resources, combine re-
(inbound and outbound; cf., Chesbrough 2003, Greu- sources to create new bundles, and generate new arti-
l et al. 2018) and temporally separating three stages of facts. As digital outcomes, studies have shown how
 Lin and Maruping: Open Source Collaboration in Digital Startups
214 Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS

digitalization can be used to deliver new value propo- as the broadest concept of the three, has been used to
sitions through digital products and services (Yoo et al. describe any purposeful knowledge flows across or-
2010). As digital contexts, digital technologies create or ganizational boundaries for innovation (Chesbrough
reshape the environment in which entrepreneurial ac- 2003). Over the past two decades, researchers have
tivities take place, for example, crowdfunding and proposed different conceptualizations of open inno-
crowdsourcing platforms for startups to acquire capi- vation. For instance, Dahlander and Gann (2010) pro-
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

tal or ideas from the crowd (Foege et al. 2019). pose four categories of open innovation based on the
With openness being a common feature in digital direction of knowledge flows and the type of re-
entrepreneurship (Nambisan 2017, Nambisan et al. wards: acquiring (inbound and pecuniary), selling
2018), researchers have explored opportunities and is- (outbound and pecuniary), sourcing (inbound and
sues of openness in digital startups. For instance, nonpecuniary), and revealing (outbound and nonpe-
Gruber and Henkel (2006) examine new ventures in cuniary). In a similar vein, Zobel and Hagedoorn
the OSS industry and show that OSS startups draw (2020) propose a framework to distinguish different
from and contribute to OSS projects to mitigate mar- types of organizational boundaries (competence, effi-
ket entry barriers and the liabilities of smallness and ciency, identity, and power) and discuss the implica-
newness. More recently, Greul et al. (2018) examine tions of their permeability for the governance of open
the strategic openness choices confronted by 3D print- innovation. Taxonomies like these showcase the di-
ing startups in which they weigh the benefits and versity of approaches under the umbrella of open
risks of engaging in an online innovation community. innovation. Some studies go beyond categorizing
Echoing this point, West and Kuk (2016) argue that “a openness and examine how firms should choose be-
firm can only be open if it has something that is closed tween different innovation approaches. For example,
which allows it to generate revenues and profits to Lakhani et al. (2013) argue that the locus of innova-
support the business” (p. 169). tion should increasingly shift outward when tasks are
To date, however, openness in digital entrepreneur- more decomposable and knowledge is more broadly
ship has lacked a comprehensive theoretical frame- distributed. Similarly, Felin and Zenger (2014) sug-
work. Although prior studies have illuminated rich gest that the optimal solution search method should
insights regarding the opportunities from openness be guided by the degrees of hidden knowledge and
and the tension between closed and open innovation problem complexity.
in digital startups, how the opportunities and tensions Open collaboration is a type of open innovation
evolve over the venture creation journey has received (Baldwin and von Hippel 2011). The hallmarks of
little attention. Openness in digital entrepreneurship open collaboration are community-oriented activities
is typically not a static, one-time activity. Instead, it among purposeful yet loosely coordinated partici-
should be viewed as an ever-evolving strategy that pants for the creation of public goods that can then be
continues to be adapted and recalibrated depending accessed and modified by anyone (Levine and Prietu-
on dominant problems in different venture stages. la 2014, Kane and Ransbotham 2016). Baldwin and
Further, the close/open dichotomy is too coarse a dis- von Hippel (2011) are among the first to point out the
tinction to adequately examine the effects of openness. potential of open collaboration for innovation. They
To this point, Felin and Zenger (2020) argue that we analytically show that (a) open collaborative innova-
need “more discriminating and nuanced research on tion is economically viable as long as communication
firm-level strategy and open innovation” (p. 231). costs are low, and (b) it could displace producer inno-
Whereas prior research tends to show a compounded vation when design costs are high. Similarly, Levine
effect of openness (e.g., Spaeth et al. 2015), we seek to and Prietula (2014) use computational models to
provide a more refined view of openness. Clearly, study the dynamics and principles in open collabora-
opening to inbound knowledge flows is different from tion. They show that open collaborative innovation
opening to outbound knowledge flows, and the two could thrive even with few contributors, free riding,
should yield vastly different opportunities and risks in rival goods, and low participant diversity.
the process of entrepreneurial endeavors. In all, we are Our conceptualization of OSC draws upon open
motivated to develop a more comprehensive theoreti- collaboration. Indeed, OSS development has been con-
cal framework in understanding openness in digital sidered as an early and canonical instance of open
entrepreneurship. We do so by contextualizing our collaboration. As such, we conceptualize OSC as a
theorizing in a salient avenue for openness among dig- specific type of open collaboration in which the inno-
ital startups, OSC, which we elaborate next. vation or production activities center on the develop-
ment or use of OSS. Contemporary research in OSS
Open Innovation, Open Collaboration, and OSC has examined various facets of organizing for OSC by
There is a hierarchical relationship among open inno- firms. Although these works covered a range of phe-
vation, open collaboration, and OSC. Open innovation, nomena, they can be broadly categorized by their
 Lin and Maruping: Open Source Collaboration in Digital Startups
Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS 215

emphasis on firm contributions to OSS communities, users in one place. As a result, it draws an ever-
OSS community contributions to firm projects, and growing number of new projects and users to the eco-
firm-OSS community arrangements. For firm contri- system, which in turn benefits the existing projects
butions to OSS, research has examined how firms di- and users on the platform because of the added devel-
rect their employees to make OSS contributions to opers and code (Parker et al. 2017). Meanwhile, the
align with their own proprietary offerings to custom- large and diverse pool of projects and users on Gi-
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

ers (Dahlander and Wallin 2006, Capra et al. 2011) tHub has brought unprecedented levels of generativ-
and when it is beneficial to devote employee resources ity (Zittrain 2006). As Yoo et al. (2010, p. 730) point
to OSS communities (Colombo et al. 2013). For OSS out, “the greater the heterogeneity, the more genera-
community contributions to firms, research has exam- tive the platform becomes.” OSS innovations on Gi-
ined how the for-profit versus nonprofit nature of the tHub are more likely to improve and grow through
sponsoring firm affects community contributions the generative process with contributions or comple-
(Stewart et al. 2006) and how sponsoring firm attrib- ments from other members in the ecosystem.
utes such as credibility and openness attract communi-
ty contributions (Spaeth et al. 2015). Finally, research The Paradox of Openness
on managing firm-OSS community collaboration has The paradoxical nature of openness is at the core of
examined how task execution is organized via open OSC in digital entrepreneurship. The term paradox
superposition and its effects on success for firm- came from ancient Greek para (against; contrary to) and
sponsored OSS projects (Medappa and Srivastava doxa (to appear; to think; to accept) and is often illustrat-
2019) and identified the governance forms that firms ed by the Chinese yin–yang symbol, representing the
enact to manage OSC (Schaarschmidt et al. 2015) as dualism of two contradictory but interdependent ele-
well as the obligations, roles, and the responsive de- ments (Smith and Lewis 2011). Drawing on these philo-
sign practices involved when firms collaborate with sophical foundations, Schad et al. (2016) define paradox
OSS communities (Agerfalk and Fitzgerald 2008, as “persistent contradiction between interdependent el-
Germonprez et al. 2017). Taken together, research in ements” (p. 10). In digital entrepreneurship specifically,
OSS has engaged with various aspects of firm organiz- OSC is paradoxical because value creation and value
ing for OSC. Although OSC has been considered as a capture from this form of openness are interdependent
valuable means for innovation and scaling entrepre- but persistently contradictory. OSC often enables digi-
neurial ventures (Parker et al. 2017, Shaikh and Levina tal startups to create valuable innovations through net-
2019), the OSS literature has not theorized this with work effects and generativity as described earlier. But
much depth, especially for startups outside the OSS OSC also makes it difficult for digital startups to cap-
industry. ture value from innovations because rivals can easily
The emergence of GitHub since 2008 represents a imitate or copy any software that is open sourced.3
shift of paradigm in the OSC practice in two signifi- To explore the extant literature on the paradox of
cant ways that reduce the barriers and increase the openness, we conducted a systematic literature search
value of OSC participation. First, unlike the conven- and classified the identified articles based on the para-
tional open source practice in which each project had dox taxonomies proposed by Schad et al. (2016) and
its own protocols, processes, and infrastructures (e.g., Poole and Van de Ven (1989).4 In their review of
kernel.org for Linux), GitHub enables a standardized paradox research in management, Schad et al. (2016)
glossary, interfaces, and practices in the creation, shar- identified six themes and organized them into three cat-
ing, and maintenance of any open source packages egories based on their nature (types and relationships),
(Faraj et al. 2016). Clearly, this system-wide standardi- impact (outcomes and dynamics), and approaches (in-
zation reduces the costs and barriers in OSC participa- dividual or collective). Poole and Van de Ven (1989)
tion. This effect of standardization is in contrast to further showed that the approaches to paradox can be
(and in addition to) project-level approaches for ad- categorized into four modes: acceptance (to live with
dressing challenges in OSS development, such as spe- the paradox without trying to resolve it), spatial separa-
cialization (division of labor in a project; von Krogh tion (to resolve the paradox by structurally or spatially
et al. 2003), modularization (decomposition of a com- separating the competing demands), temporal separa-
plex project into more manageable tasks; Baldwin and tion (to resolve the paradox by allocating the competing
Clark 2006), and superposition (sequential layering of demands to different time periods), and synthesis (to
individual tasks; Howison and Crowston 2014). resolve the paradox by devising a solution that takes all
Second, GitHub cultivates an ecosystem for OSC elements in a paradox into account). It is important to
that creates value for OSC participation through two note that resolving a paradox is not to eliminate the ten-
reinforcing and multiplicative mechanisms: network sion but, rather, to find “a means of meeting competing
effects and generativity. In terms of network effects, demands or considering divergent ideas simulta-
GitHub hosts a large and diverse pool of projects and neously” (Smith and Lewis 2011, p. 386).
 Lin and Maruping: Open Source Collaboration in Digital Startups
216 Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS

Overall, prior research has provided rich discus- Theory and Hypotheses
sions on the nature of the paradox of openness. For This section explicates our proposed conceptual frame-
example, many studies have shown a concave rela- work for OSC in digital entrepreneurship and develops
tionship between openness and appropriability testable hypotheses related to our conceptualization.
(Laursen and Salter 2014, Wadhwa et al. 2017). The lit- We begin with brief discussions of the theoretical basis
erature has also suggested that the tensions are mod- of two contextual factors in our framework: (a) modes
erated by firms’ technological capabilities (Arora et al.
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

of OSC and (b) stages of venture maturity. We then ex-

2016), the characteristics of the focal firm (service or plore how their interplay affects the feasibility and
manufacturing) or the external partner (domestic or benefits of OSC to digital startups and theorize how
international) (Stefan and Bengtsson 2016, Yacoub digital startups may approach the paradoxical tensions
et al. 2020), and the complementarity and substitut- to realize value.
ability of resources (Alexy et al. 2018). To resolve the
paradox, scholars have developed novel concepts to Modes of OSC
reconcile or reduce the tensions. For instance, Alexy The open innovation literature has established that
et al. (2018) view firms as resource bundles and dis- openness is multidirectional and can involve inbound as
cuss the concept of value migration, in which firms well as outbound knowledge flows across organization-
open up parts of the resources to increase profits for al boundaries (Chesbrough 2003, Bogers et al. 2017).
their still-proprietary complements. Similarly, Zobel
These have also been described as the receiving and the
and Hagedoorn (2020) propose the use of informal,
giving sides of the innovation process (Sojer and Henkel
open-ended relational contract design as an alterna-
2010). The distinction between inbound and outbound
tive appropriation regime to govern interorganiza-
openness is important to improve conceptual clarity, al-
tional knowledge flows.
lowing more precision in theorizing (Dahlander and
In spite of this richness of research on the paradox
Gann 2010, Foege et al. 2019). Otherwise, there can be
of openness, there are still several notable gaps, espe-
substantial theoretical ambiguity, even inconsistency,
cially for understanding OSC in digital entrepreneur-
because inbound and outbound openness present dis-
ship. First, we found that none of the articles in our
tinct opportunities and risks to the firm.
review examine paradox in the process of new ven-
Accordingly, we differentiate inbound and out-
ture creation. As such, the extant theories and findings
bound OSC based on the primary direction of knowl-
regarding the paradox of openness may need to be ex-
tended to predict or explain the use of openness in edge flows. In outbound OSC, knowledge flows from
digital entrepreneurship. Second, most research has inside-out in which participants contribute their own
focused inward on the relationships between the software innovations to the open source community.
contradictory elements without exploring how the In inbound OSC, knowledge flows from outside-in in
tensions and their management affect organizational which participants absorb external software innova-
outcomes. Consequently, the literature does not pro- tions from the open source community.
vide sufficient insight in explaining whether and how The two modes of OSC enable different mechanisms
engaging in OSC might benefit the outcomes of digital for value creation. In outbound OSC, knowledge shar-
startups. Third, extant research has not differentiated ing is the value creation mechanism. The literature has
and addressed inbound and outbound knowledge collectively shown that knowledge sharing enables
flows in their theorizing in depth. We need to consider firms to create value by eliciting collaboration (Dah-
both directions of knowledge flows to be able to un- lander and Gann 2010), attracting talent (Lerner and
derstand the paradox of openness more precisely in Tirole 2002), establishing standards (Harhoff et al.
digital startups’ OSC. Finally, there is an absence of 2003), growing the market (Parker et al. 2017), stimu-
temporal separation in the extant approaches to the lating third-party complements (Alexy et al. 2018), and
paradox of openness. This suggests that extant re- improving organizational learning (Nagle 2018). In in-
search tended to assume that the poles of the paradox bound OSC, the mechanism for value creation is
remain stable over time (cf., Smith et al. 2017). Since knowledge reuse. Firms’ reuse of external knowledge
new venture creation consists of several distinct can have beneficial effects on the costs, speed, and
phases and each comes with a unique set of chal- novelty of their innovations. Reusing external knowl-
lenges and needs (Kazanjian 1988, Fisher et al. 2016), edge can lower the costs and increase the speed of in-
we expect digital startups’ resolution of the paradox- novations as it negates the need to reinvent solutions
ical demands in OSC to shift over time. In all, our that have already been generated (Haefliger et al.
study complements extant paradox of openness re- 2008). Further, reuse of external knowledge can inspire
search by providing greater nuance in the use of ideas that are otherwise hard to imagine and has been
openness in general and OSC in particular for digi- shown to improve the output and the degree of novel-
tal entrepreneurship. ty of innovation (Majchrzak et al. 2004).
 Lin and Maruping: Open Source Collaboration in Digital Startups
Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS 217

Despite these value creation mechanisms, research- commercialization, growth, and stability. Since our
ers find that value capture is also critical to realizing primary interest is in new ventures as opposed to es-
gains from innovation (Teece 1986, Laursen and Salter tablished firms, we follow Fisher et al. (2016) and focus
2014, Alexy et al. 2018). To capture value from innova- on the first three stages of Kazanjian’s model in our
tion, firms’ appropriability strategies play a significant theoretical development.
role. The literature has suggested a menu of formal As Kazanjian (1988) explains, each stage of venture
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

and informal appropriability strategies to capture val- maturity has a unique set of dominant problems. Dur-
ue. In terms of formal strategies, firms may use pat- ing the conception stage, startups’ dominant problems
ents, trademarks, or other regulatory means to protect are about ideation—the generation, selection, and pro-
their intellectual assets. For informal strategies, firms totyping of ideas for the creation of a new venture.
may protect innovation returns by keeping some ele- Startups at this stage are concerned with learning how
ments of the innovation secret or exploiting lead time to invent creative and feasible solutions and overcome
or complementary assets of the innovation to stay key technical hurdles. As a startup receives funding
ahead of competitors. from venture capital (VC) firms or angel investors, it
In the context of OSC by digital startups, informal progresses into the commercialization stage. Having
appropriability strategies are expected to be more sa- the business ideas validated by experienced financial
lient for three reasons. First, studies have found that backers, startups at this stage shift their focus to en-
new and small firms generally have fewer resources sure the technology works well and goes beyond a
for formal intellectual property (IP) protection barely functional prototype. A startup reaches the
schemes (Leiponen and Byma 2009). Second, formal growth stage when it establishes at least a proven
appropriability strategies are less relevant when it product or service with substantial market size. At
comes to OSS since the source code is made available this stage, technology risks and market uncertainties
to everyone for free (Harhoff et al. 2003). Although are significantly reduced, and the dominant problems
most OSS comes with a license agreement, such open shift toward growing sales and market share. Taken
source licenses do not guard against imitation and ex- together, the discrepancies in dominant problems
propriation. Third, complementary knowledge and across stages of venture maturity imply that, depend-
assets have been considered as the most critical value- ing on its current stage, a digital startup may be more
capturing mechanism in the digital economy as plat- or less prone to risks of OSC or realizing its benefits.
forms and ecosystems become pervasive (Teece 2018). Digitalization has implications for venture stages,
To be clear, outbound OSC and IP protection are not an but it does not invalidate the stage perspective of en-
either/or choice, and we do expect digital startups to trepreneurship. According to Nambisan (2017), digiti-
take protection measures when engaging in outbound zation of entrepreneurial processes and outcomes
OSC. For example, research on selective openness sug- helps break down the boundary between experimen-
gests that firms protect their knowledge outflows by tation and implementation, leading to greater levels of
controlling complementary and still-proprietary resour- overlap in different stages and nonlinearity in how a
ces (Henkel et al. 2014, West and Kuk 2016, Alexy et al. startup advances from one to the next. Indeed, digital
2018). innovations are often incomplete by design such that
digital startups may launch a less mature product to
Stages of Venture Maturity the market and update the product “on the go” based
There is a growing consensus among entrepreneurship on market feedback (Yoo et al. 2012). Nevertheless,
scholars that entrepreneurship, and new venture crea- digital startups still need to attend to the stage-
tion in particular, should be viewed as an unfolding specific dominant problems that are unfolding across
process (Shane and Venkataraman 2000, Davidsson the organizational life cycle. After all, and despite the
and Gruenhagen 2021). Whereas there are multiple inherent process flexibility in the digital world, it re-
ways to characterize process patterns (e.g., speed, mains necessary for digital startups to conceptualize
cycles, and trajectories), Davidsson and Gruenhagen and identify opportunities, develop a minimum viable
(2021) find that differentiating stages is most common product, commercialize the product, and grow their
in the entrepreneurship literature. Indeed, the process market, in that order.
of new venture development takes anywhere from the
formation of an idea in an entrepreneur’s mind to a The Proposed Conceptual Framework
mature product that has amassed a large customer We propose that value creation and capture in a digi-
base. The notion of venture stages helps surface quali- tal startup’s OSC are contingent on the mode of OSC
tative changes and milestones in the journey of entre- engagement and the stage of venture maturity. As il-
preneurial pursuits. To characterize stages of venture lustrated in Table 1, although digital startups enact
maturity, Kazanjian (1988) proposes a stage-of-growth both modes of OSC, we theorize that the relative effect
model with four distinctive stages: conception, of each in enabling value creation and capture will
 Lin and Maruping: Open Source Collaboration in Digital Startups
218 Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS

Table 1. Theoretical Framework

Stage of venture maturity

Mode of OSC Stage 1: Conception Stage 2: Commercialization Stage 3: Growth
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

vary for digital startups in an earlier stage of maturity inbound OSC may enable nascent digital startups to
compared with those in a later stage of maturity. This generate new ideas by exploring, customizing, and re-
relative effect is the net of multiple factors, including combining existing open source innovations (Majchr-
the feasibility to create and capture value from in- zak et al. 2004, Kyriakou et al. 2017). This promotes
bound/outbound OSC and their benefits in address- digital startups’ product ideation and prototyping in
ing the dominant problems of different venture stages. three ways that are highly valuable. First, with scant
Consistent with Poole and Van de Ven (1989), we ap- resources at this stage, engaging in inbound OSC
proach the tensions in digital startups’ OSC through could be a cost-effective means to acquire intangible
temporal separation and spatial separation to identify assets (e.g., existing solutions and best practices) need-
insights on organizing. In what follows, we arrange ed in kickstarting a new venture. Second and consis-
our theoretical arguments based on venture stages. tent with research findings on OSS code reuse (Hae-
fliger et al. 2008, Sojer and Henkel 2010), inbound
Stage 1: Conception. As indicated earlier, the domi- OSC speeds up the time to market as the startup is
nant problems for startups at the conception stage are able to build upon and adapt existing tools as op-
related to ideation of the product or service. Startups fo- posed to developing everything from the ground up.
cus on the search, learning, and experimentation of Finally, drawing from external knowledge on GitHub
their product concepts during this exploration phase. could improve the quality and novelty of ideation be-
Given their emphasis on ideation and their lack of com- cause such knowledge reuse may enable digital start-
plementary assets, we infer that inbound OSC would ups to invent solutions in nontraditional ways
be more effective relative to outbound OSC in enabling (Majchrzak et al. 2004).
stage 1 digital startups to create and capture value. Although the search and integration of external
With new venture concepts still in flux, digital start- knowledge could incur substantial costs and efforts
ups at this early stage may find it difficult to engage in (Majchrzak et al. 2004), digital platforms such as Gi-
outbound OSC to create value toward new product tHub have drastically reduced these burdens in in-
ideation. Since digital startups at this stage are virtually bound OSC. In terms of search, GitHub offers conve-
unknown to the market, it is difficult for them to favor- nient and powerful search functions on public OSS
ably attract volunteer developers. As Spaeth et al. repositories using keywords and filters (cf, Kane and
(2015) explain, the perceived credibility of a sponsoring Ransbotham 2016). As for integration, Git—which is
firm is critical in attracting volunteer contributors the code management system used by GitHub—
because it offers more valuable learning opportuni- makes it easier to customize, experiment, and inte-
ties and a more trustworthy environment for partici- grate existing code. To the extent that stage 1 digital
pation. Further, in the absence of a marketable 0.01w?>startups possess adequate prior technical
product and financial backing, stage 1 digital start- and domain knowledge to absorb this inbound OSC
ups are unlikely to own any complementary assets,
(Shane 2000, Zahra and George 2002), they should be
which would in turn limit their ability to capture
able to garner greater value from inbound OSC than
value from their knowledge outflows (Teece 1986,
outbound OSC.
Laursen and Salter 2014). Therefore, exposing prod-
uct concepts in platform environments at such an Hypothesis 1. Inbound OSC is associated with a more
early stage of maturity can be risky as it enables ri- positive impact than outbound OSC to the value of digital
vals to engage in imitation before the startup has a startups in the conception stage.
product in market (Huang et al. 2020).
In contrast, inbound OSC could be highly valuable Stage 2: Commercialization. With the infusion of capi-
for stage 1 digital startups. Knowledge inflows from tal from a VC firm or an angel investor, the focus of
 Lin and Maruping: Open Source Collaboration in Digital Startups
Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS 219

startups at the commercialization stage is on develop- thousands of users and developers and at the same
ing a refined product that works well beyond a proof- time cut costs” (p. 277).
of-concept prototype and meets certain market needs Second, inbound OSC enables digital startups to ap-
(Kazanjian 1988). In other words, the dominant prob- propriate value by capitalizing upon opportunities for
lems for startups at this stage are centered around complementarities (Fosfuri et al. 2008). That is, in-
product development. The two modes of OSC engage- bound OSC offers digital startups privileged observ-
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

ment may affect product development in stage 2 digi- ability into the underlying architecture of existing
tal startups, as we discuss next. OSS, enabling them to shape their own product or ser-
Engaging in outbound OSC may create value in vice offering to be complementary. Bonaccorsi et al.
stage 2 digital startups’ product development through (2006) suggest that it enables firms to collect product
two notable ways. From a platform ecosystem per- and service information that can reveal market niches
spective, outbound OSC may allow stage 2 digital that are untapped and may eventually open up. Stam
startups to reach out to their potential consumers and (2009) argues that such privileged access to informa-
partners and engage in cocreation (Hoyer et al. 2010). tion on market opportunities enables firms to organize
With input from external partners, such cocreation and reorient their internal research and development
activities could help digital startups better identify (R&D) activities toward promising avenues. A multi-
market needs and increase the value of the product. ple case study by Dahlander (2005) reveals that one
Another avenue through which stage 2 digital start- strategy by which firms appropriate rents from open
ups may create value through outbound OSC is by source is by bundling such source code with their
leveraging volunteer developers to improve the reli- own code and selling the installations. He also finds
ability of the product functions or components. This that one firm developed and sold a proprietary con-
could involve testing and debugging the software in tent management system that integrated with a popu-
diverse settings and improving its usability to a lar open source web server in a manner that was easy
broader population. to configure and use.
There are, however, several impediments that in- In sum, although both modes of OSC could enable
hibit value creation and capture from outbound OSC stage 2 digital startups to create and capture value in
in stage 2 digital startups. Since their product is not areas related to new product development, significant
yet market proven and their reputation remains large- impediments in outbound OSC may limit the utility
ly unknown to the general public, stage 2 digital start- and feasibility of such engagement. Therefore, we
ups face similar risks of expropriation and difficulties posit that inbound OSC should better contribute to
of attracting volunteer developers when engaging in the value of stage 2 digital startups.
outbound OSC as stage 1 digital startups do. More-
over, investors in a stage 2 digital startup may have Hypothesis 2. Inbound OSC is associated with a more
concerns about the consequences of opening up IP positive impact than outbound OSC to the value of digital
through outbound OSC (Mann and Sager 2007). In the startups in the commercialization stage.
event of startup failure, sale of IP can enable investors
to recoup some of their losses (Serrano and Ziedonis Stage 3: Growth. In contrast to earlier stage digital
2019). Therefore, outbound OSC when there is still startups, those at the growth stage have a reputable
technological or market uncertainties could erode the product or service in market and have demonstrated
value investors ascribe to the startup. traction with customers (Kazanjian 1988). The main
Inbound OSC can also benefit product development emphasis for growth stage startups is on scaling to in-
in stage 2 digital startups through two main mecha- crease market share and sales. With the emphasis on
nisms. First, GitHub offers access to low-cost, high- scaling, we theorize that for stage 3 digital startups,
quality source code. When digital startups achieve outbound OSC is likely to be more salient than in-
a stage of maturity where they are focused on refining bound OSC in value creation and capture.
their product functioning, they can be deliberate With demonstrated market traction and credibility,
in their decision making about the software and em- outbound OSC in stage 3 digital startups is expected
bedded technological know-how that is needed to to garner greater attention and interest from volunteer
enhance their offering. Such inbound OSC enables OSS developers (von Krogh et al. 2012, Spaeth et al.
digital startups to appropriate value by saving costs 2015). Increased third-party developer attention can
of reproducing source code and functionality using create value to stage 3 digital startups in three ways
their own resources (Germonprez et al. 2017). Further, that are beneficial to scaling: (1) fixing problems, (2)
it enables digital startups to leverage established OSS stimulating complements, and (3) signaling confi-
to improve product performance. As Dahlander dence. First, outbound OSC enables developers on
(2005) observed, “use of open components allow GitHub to discover, investigate, and address potential
[firms] to use components which are tested by bugs and usability issues in the startup’s software.
 Lin and Maruping: Open Source Collaboration in Digital Startups
220 Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS

Such outbound OSC improves the quality of the offer- activity logs whereas the latter provides startup-
ings, which could help expand sales and market related information. Sampling is a salient issue in our
share. Second, outbound OSC encourages the supply empirical setting because digital entrepreneurship is a
of third-party complements. Outbound OSC enables relatively new phenomenon, involving more than just
other participants on the platform to fork the software software or information technology (IT) startups. To
for their own pursuits. Such openness of source code our knowledge, there is not a list or database of digital
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

encourages third parties to develop complementary startups. As such, we identify a set of industry groups
products and services that enhance the value proposi- from Crunchbase that have been commonly suggested
tion of the digital startup’s offering (Dahlander 2005, by the digital entrepreneurship literature, including
Nambisan et al. 2018). Meanwhile, increased comple- (a) AI, (b) apps, (c) data and analytics, (d) Internet/
ments can have cross-side network effects by attract- cloud services, (e) IT, (f) platforms, and (g) software.
ing additional customers (Parker et al. 2017). Third, We construct our research data based on the follow-
the visibility of problem fixes and third-party integra- ing procedure. First, from the Crunchbase data set we
tions help signal an increase in the quality and the in- identify all for-profit, U.S.-based private firms
dispensable nature of the startup’s offering. This founded between 2008 and 2017 within the aforemen-
could engender a perception of lower risk and greater tioned industry groups. This is the universe of digital
confidence in the startup’s business among consum- startups that we consider for our sample construction.
ers, which in turn contributes to its legitimacy and We limit our sample to U.S. startups because the Unit-
market growth. ed States has more new venture activities and more
From a value capture perspective, growth stage dig- GitHub activities than other countries. The window of
ital startups have amassed substantial capital invest- startup founding years was chosen to ensure that the
ments and product revenues, which allow them to startups were indeed new ventures and GitHub was
accumulate and secure complementary assets. The available when they were founded.
complementary assets may come from a variety of Second, we identify organization accounts on
sources, such as technology standards, marketing ca- GitHub and match these accounts with the digital
pabilities, manufacturing know-how, user/developer startup records from Crunchbase. We are able to es-
communities, platforms, and ecosystems, among tablish record linkages for 4,388 digital startups. Con-
others (Teece 1986, 2018). These complementary assets sistent with the econometric literature (e.g., Austin
help guard against the threat of expropriation in and Stuart 2015), we label these digital startups as the
knowledge outflows when digital startups engage in treatment group. We define a treatment as demon-
outbound OSC (Laursen and Salter 2014). strated participation in OSC, and digital startups are
In contrast to outbound OSC, inbound OSC is pri- considered treated at the point when they have a pub-
marily directed toward improving internal R&D per- lic repository on GitHub.
formance (e.g., inspiring new products or enhancing Third, we identify a set of untreated startups that
existing products). This provides added value for ex- are similar to the treated startups as the control group.
isting customers and may also attract new customers Specifically, for each treated startup, we randomly
(Huang et al. 2017). However, for stage 3 digital start- sample three untreated startups from the universe of
ups, such enhancements may not scale the startup’s digital startups that were (a) founded within three
market presence to the same degree as would an in- months of the treated startup’s founding date and (b)
crease in the number of complementors. From a plat- operated in the same industry. This approach is akin
form ecosystem perspective, the network externality to blocking in experimental design in that startups in
of complementarities is more likely to cause positive the treatment and control groups are blocked at the
feedback and exponential growth (Parker et al. 2017, start in two key drivers of startup performance: tim-
Nambisan et al. 2018), and the network effects are an ing of entry and type of business. Since the treated
essential force in the diffusion of technological inno- and the untreated startups are comparable along these
vation (Abrahamson and Rosenkopf 1997). Conse- two dimensions, this sample construction strategy
quently, we propose the following hypothesis. allows us to implicitly control for environmental con-
ditions, such as market and competition, that may
Hypothesis 3. Outbound OSC is associated with a more otherwise affect our inference.
positive impact than inbound OSC to the value of digital Finally, we derive a panel data set to characterize
startups in the growth stage. the longitudinal observations of the digital startups in
our sample. For each startup, we create a monthly
Data and Methods panel from its month of founding to the month before
Sample and Research Design its exit/closure, or December 2017 (the last time peri-
To test our hypotheses, we collect data from GitHub od of our panel data) if the startup is still operating
and Crunchbase. The former provides detailed OSC and privately held. This is to ensure that the firms in
 Lin and Maruping: Open Source Collaboration in Digital Startups
Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS 221

our analyses are active and private new ventures, in VC firm. We categorize startups as stage 3 once they
accordance with the scope of this study. In all, this raise a series B or later round from VC firms.
process yields a total of 17,552 startups in our sample
(of which 4,388 were treated) and 1,196,706 startup- Control Variables. We control for an array of time-
month observations. varying factors that may affect relationships of inter-
est in this study, including company age, the overall
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

Measures quality of the board (board centrality), investor experi-

Dependent Variable (DV). The outcome of interest ence, prior professional experience of the top manage-
in this study is firm value of a digital startup. Unlike ment team (TMT job experience), and news coverage and
public firms, market-based valuation measures are Internet search volume on the focal startup.
not applicable to privately held startups. Following Taken together, Table 2 summarizes the descriptive
prior entrepreneurship research (e.g., Gompers et al. statistics of these variables. As evident from Table 2,
2010), we use the premoney valuation from the most the distributions of these measures are very skewed.
recent VC-backed funding round to represent a digital As such, we log transform all nonbinary measures in
startup’s value. The premoney valuation of a startup all our regression analyses to address the skewness in
represents a professional and comprehensive evalua- their distributions.
tion of the “perceived net present value of the
company” (Gompers et al. 2010, p. 31) before the latest Empirical Models and Identification Strategies. We
round of funding.5 model a startup’s value as a function of OSC activities,
time-varying control variables, firm fixed effects (FEs),
OSC Variables. We measure and quantify OSC in and time FEs, and other unobserved idiosyncratic er-
three different ways: overall, outbound, and inbound. rors. To avoid the problem of reverse causation, we
We use the total number of public repositories the dig- use the startup’s value 12 months ahead of the current
ital startup has on GitHub to represent overall OSC. To period (t + 12) as the DV. The empirical model is
distinguish the direction of knowledge flows, we use estimated using the weighted least squares (WLS)
the number of public repositories created by the approach, and we will discuss how to calculate the
startup to represent outbound OSC and the number of weights later in this section. The weights in WLS ad-
public repositories forked by the startup to repre- just each startup’s propensity for being treated (i.e.,
sent inbound OSC. OSC) when estimating the treatment effects. Since our
DV and numeric independent variables are log trans-
Moderator. Venture stage is our moderator for digital formed, the coefficients from our regression model
startup maturity. We consider the first three stages of can be interpreted as the percentage change in the val-
Kazanjian’s (1988) stage-of-growth model as outlined ue of a digital startup contributed by the correspond-
earlier. Stage 1 indicates that the startups have not yet ing regression terms.
received any funding from angel investors or VC To understand how OSC affects digital startups’ val-
firms. Startups progress to stage 2 when they obtain ue, we must explicitly address two endogeneity concerns
angel investment or seed or series A funding from a that are common in observational studies: selection

Table 2. Descriptive Statistics

Correlation

Variable Mean SD Max (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

1. Firm value 4.64 312.47 59,000

2. Overall OSC 1.80 22.88 2,882 0.01
3. Outbound OSC 0.98 11.60 1,898 0.01 0.57
4. Inbound OSC 0.83 18.82 2,836 0.01 0.86 0.08
5. Board centrality 183,467.9 753,610.3 29,019,561 0.02 0.06 0.06 0.03
6. Investor experience 94.58 380.25 9,440 0.18 0.11 0.11 0.06 0.40
7. TMT job experience 55.57 271.18 11,835 0.14 0.04 0.05 0.02 0.21 0.38
8. Company age 3.69 2.54 71 0.02 0.06 0.06 0.04 0.03 0.13 0.07
9. News coverage 0.004 0.25 125 0.74 0.01 0.01 0.00 0.02 0.13 0.10 0.01
10. Search volume 0.28 2.79 100 0.06 0.00 0.01 0.00 0.02 0.05 0.02 −0.01 0.04
11. Venture stage 1 0.74 0.44 1 −0.03 −0.05 −0.04 −0.03 −0.26 −0.38 −0.20 −0.11 −0.02 −0.03
12. Venture stage 2 0.22 0.41 1 −0.01 0.01 0.01 0.01 0.12 0.13 0.03 0.04 0.00 0.02 −0.89
13. Venture stage 3 0.04 0.20 1 0.07 0.08 0.08 0.05 0.32 0.56 0.37 0.17 0.05 0.03 −0.35 −0.11
Notes. SD, standard deviation. The minimum value is 0 in each of these variables. Venture stages 1, 2, and 3 represent the conception,
commercialization, and growth stages in Kazanjian (1988).
 Lin and Maruping: Open Source Collaboration in Digital Startups
222 Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS

bias and omitted variable bias (Wooldridge 2002). First, omitted variables. In this approach, we first predict
the treated and control startups may differ in systematic the degree of OSC activities in each startup using IVs,
ways such that the startups that self-selected into the and then use the predicted values in the main regres-
treatment group may attain a better outcome even with- sion to estimate the impact of OSC on startups’ value.
out the treatment. Second, there may be omitted factors The intuition of our instrumentation is motivated by
that correlate with OSC activities such that the observed Fan (2013) and Karanam et al. (2020), in which we
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

effects are actually caused by the omitted factors rather construct a set of IVs based on the OSC behaviors
than OSC. We mitigate selection bias using the inverse among competitors’ competitors in nonoverlapping
probability of treatment weighting (IPTW) strategy and business categories of the focal startup.7 The identify-
address omitted variable bias through instrumental vari- ing assumption in the IV estimation concerns the exo-
ables (IVs), as we elaborate next. geneity of the instruments, which is not empirically
To mitigate selection bias, we employ the IPTW verifiable in our context. However, since IPTW and
strategy in our empirical estimation (Austin and IVs rely on different assumptions, the IV estimates
Stuart 2015). The intuition for IPTW is to adjust the can validate the results from the IPTW-based WLS
weight of each startup in the model fitting procedure estimation.
so as to accommodate their heterogeneous propensity
toward receiving the treatment. If a startup is ex ante Results
likely (unlikely) to participate in OSC, it should have Table 4 shows our results from the WLS and IV esti-
a lower (higher) weight in our estimation. In doing so, mations. In each set of analyses, we present the rela-
we reduce potential endogeneity concerns in our caus- tionship between OSC and the value of digital start-
al inference from observational data.6 To ensure that ups in four different ways: (1) the effect of overall
the IPTW strategy is valid, we conduct a balance OSC, (2) the effect of overall OSC moderated by ven-
check on observed covariates between treatment and ture stages, (3) the effects of outbound and inbound
control startups, following Austin and Stuart (2015). OSC, and (4) the effects of outbound and inbound
Table 3 shows that there were significant imbalances OSC moderated by venture stages. Although our con-
for most variables before IPTW and that all these im- ceptual framework and hypotheses concern mainly
balances become negligible after IPTW. This suggests the last (fourth) condition, we consider all these four
that IPTW effectively addresses the systematic differ- analytical setups for two reasons. First, they allow us
ences in observed characteristics between the treated to detect the roles of distinguishing modes of OSC
and untreated digital startups in our sample. and stages of venture maturity. Second, it is the best
To address the potential omitted variable bias, we practice in regression analysis to model and present
implement an IV estimation. Omitted variables are estimates without interaction terms along with the
plausible in our context since we do not have data on ones with those terms.
many other value-relevant factors that vary over time
such as R&D spending, market share, competition, WLS Estimation. Columns (1) to (4) of Table 4 present
and so on. The IV estimator allows us to isolate and the results of the WLS estimation.8 From column (1), we
identify the effects of OSC even when there exist find that the estimate of overall OSC is positive and

Table 3. IPTW Validation: Covariates Balance Check

Before IPTW After IPTW

Weighted mean Weighted mean Absolute standardized Weighted mean Weighted mean Absolute standardized
Covariate (SD) in treatment (SD) in control difference (SD) in treatment (SD) in control difference

Firm value 0.019 (0.255) 0.014 (0.163) 0.023 0.014 (0.194) 0.016 (0.185) 0.012
Board centrality 3.708 (5.952) 1.351 (3.958) 0.466 1.961 (4.674) 1.954 (4.662) 0.002
Investor experience 1.008 (2.076) 0.405 (1.343) 0.345 0.570 (1.594) 0.565 (1.601) 0.003
TMT job experience 1.621 (2.069) 0.919 (1.651) 0.375 1.128 (1.761) 1.106 (1.826) 0.012
Company age 0.540 (0.547) 0.538 (0.548) 0.004 0.545 (0.556) 0.540 (0.548) 0.010
News coverage 0.001 (0.029) 0.001 (0.022) 0.018 0.001 (0.023) 0.001 (0.024) 0.003
Search volume 0.108 (0.425) 0.057 (0.326) 0.134 0.074 (0.340) 0.072 (0.373) 0.007
Venture stage 1 0.202 (0.402) 0.136 (0.343) 0.177 0.154 (0.361) 0.153 (0.360) 0.003
Venture stage 2 0.019 (0.135) 0.005 (0.073) 0.123 0.009 (0.096) 0.009 (0.094) 0.005
Venture stage 3 0.779 (0.415) 0.859 (0.349) 0.207 0.836 (0.370) 0.838 (0.369) 0.004
Notes. All nonbinary covariates were nature-log transformed before calculating the weighted mean, SD, and absolute standardized difference.
Venture stages 1, 2, and 3 represent the conception, commercialization, and growth stages in Kazanjian (1988). When the absolute standardized
difference is below 0.1, the degree of imbalance in covariate x between treatment and control subjects is considered negligible.
 Lin and Maruping: Open Source Collaboration in Digital Startups
Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS 223

Table 4. Impact of OSC

(A) WLS estimation (B) IV estimation

Dependent variable: Firm valuei,t+12 (1) (2) (3) (4) (5) (6) (7) (8)

Overall OSC 0.018*** −0.012*** 0.150*** −0.051***

(0.005) (0.002) (0.044) (0.023)
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

Overall OSC × Venture stage 2 0.018*** 0.084***

(0.005) (0.029)
Overall OSC × Venture stage 3 0.131*** 0.331**
(0.030) (0.145)
Outbound OSC 0.005 −0.011*** 0.137 −0.518***
(0.007) (0.002) (0.284) (0.056)
Inbound OSC 0.021*** −0.005*** 1.083*** −0.249***
(0.008) (0.002) (0.356) (0.051)
Outbound OSC × Venture stage 2 0.015*** 0.262***
(0.006) (0.066)
Inbound OSC × Venture stage 2 0.042*** 0.479***
(0.007) (0.088)
Outbound OSC × Venture stage 3 0.176*** 0.577***
(0.018) (0.079)
Inbound OSC × Venture stage 3 0.056*** 0.140***
(0.012) (0.028)
Board centrality −0.000 −0.001 −0.000 −0.001 −0.001 −0.001 −0.001 −0.002
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) (0.002)
Investor experience 0.014*** 0.013*** 0.014*** 0.013*** 0.015*** 0.014*** 0.013*** 0.012***
(0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
TMT job experience 0.018*** 0.014** 0.017** 0.015** 0.014** 0.013** 0.013* 0.009
(0.007) (0.007) (0.007) (0.007) (0.007) (0.007) (0.007) (0.006)
Company age −0.020* −0.011 −0.020* −0.012 −0.019** −0.019** −0.018** −0.007
(0.011) (0.011) (0.011) (0.011) (0.009) (0.009) (0.009) (0.008)
News coverage 0.508*** 0.497*** 0.508*** 0.496*** 0.489*** 0.488*** 0.487*** 0.465***
(0.083) (0.083) (0.083) (0.082) (0.072) (0.072) (0.072) (0.070)
Search volume 0.646*** 0.639*** 0.645*** 0.640*** 0.769*** 0.765*** 0.763*** 0.746***
(0.150) (0.151) (0.150) (0.151) (0.175) (0.175) (0.175) (0.176)
Venture stage 2 0.025** 0.022* 0.025** 0.022* 0.017 0.018* 0.018* 0.019*
(0.012) (0.013) (0.012) (0.012) (0.011) (0.011) (0.011) (0.011)
Venture stage 3 0.375*** 0.240*** 0.375*** 0.256*** 0.406*** 0.403*** 0.402*** 0.232***
(0.034) (0.041) (0.034) (0.039) (0.036) (0.036) (0.036) (0.039)
Adjusted R2 0.651 0.653 0.651 0.653 0.651 0.652 0.648 0.629
Cragg-Donald Wald F — — — — 167.9 257.3 196.5 281.2
Number of startups 17,552
Number of observations 1,196,706
Notes. Firm and time FEs in all models. Robust standard errors, clustered on firm, in parentheses. We apply ln(X + 1) to transform all nonbinary
variables. Venture stages 1 (baseline), 2, and 3 represent the conception, commercialization, and growth stages in Kazanjian (1988).
*p < 0.10; **p < 0.05; ***p < 0.01.

significant to the value of digital startups. The coefficient in understanding the roles of OSC to digital entrepre-
of overall OSC is 0.018. To interpret this effect, we notice neurship. In column (3), we distinguish outbound and
that the mean and standard deviation of overall OSC are inbound OSC. We find that the estimate of inbound
1.80 and 22.88, respectively. Therefore, a one standard OSC is positive and significant but the estimate of out-
deviation increase in overall OSC from the mean trans- bound OSC is not. In other words, inbound OSC, when
lates to a 22.88 ( 22.88/1.80 * 0.018) percent increase in in aggregate, has a larger impact on digital startups’
the value of digital startups. By interacting overall OSC value than outbound OSC. The sharp difference be-
with venture stages, the estimates from column (2) tween inbound and outbound OSC points to the impor-
show that the value-adding benefit of OSC increases tance of specifying the directions of knowledge flows
with the venture stage. However, the negative coeffi- when studying OSC in digital entrepreneurship.
cient of overall OSC, which represents the effect of over- Column (4) of Table 4 presents the main results of
all OSC to stage 1 digital startup, suggests that OSC our study regarding the discrepancies in how digital
might not be so beneficial, at least in the short term, in startups can best extract value from different modes
the conception stage. The patterns observed in column of OSC across different stages of venture maturity.
(2) confirm that the venture stage can be a salient factor For ease of comparison, we visualize the contingency
 Lin and Maruping: Open Source Collaboration in Digital Startups
224 Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS

Figure 1. (Color online) Contingency Effects of OSC in Digital Entrepreneurship

Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

Notes. The bars represent the estimates, along with their standard errors, of the respective mode of OSC engagement in a specific
stage of venture maturity. Stages of venture maturity 1, 2, and 3 represent the conception, commercialization, and growth stages in
Kazanjian (1988).

effects in Figure 1 based on the estimates and stan- the value of digital startups as more causal instead of
dard errors from column (4). Consistent with our just correlational.
hypotheses, the estimates of inbound OSC are higher
than those for outbound OSC in stages 1 and 2 digital Discussion and Conclusion
startups, but for stage 3 digital startups, outbound Emerging large-scale collaboration platforms such as
OSC becomes relatively more valuable than inbound GitHub bring questions about organizing innovation
OSC. The Wald test reveals that the difference in in- to the forefront in digital entrepreneurship for three
bound and outbound OSC estimates at each stage of main reasons. First, digitalization creates more fluid
venture maturity is statistically significant. Consistent entrepreneurial boundaries and distributed entrepre-
with the patterns observed in column (2), we also find neurial agency (Nambisan 2017), and as a result, digi-
that regardless of the direction of knowledge flows, tal startups are more likely to include external actors
the estimates of OSC are increasing with the venture or resources in the process of new venture develop-
stage and the effect of OSC in stage 1 is negative. Im- ment. Second, the processes and outcomes of digital
portantly, we detect a significant shift in the relative entrepreneurship are enabled by or embodied in digi-
importance in the mode of OSC from inbound to out- tal technologies (such as AI and cloud computing),
bound as digital startups advanced from stage 2 to and open source is a powerful setting for sourcing
stage 3. In particular, for stage 2 digital startups, the and developing such technologies (Agerfalk and
coefficient of inbound OSC is almost three times larg- Fitzgerald 2008, Germonprez et al. 2017). Third,
er than that of outbound OSC, but for stage 3 digital GitHub, an emerging digital technology in itself, dras-
startups, the coefficient of inbound OSC is only about tically reduces the barriers and expands the benefits
one third as large as the coefficient of outbound OSC. for startups to participate in OSC through standardi-
zation, network effects, and generativity (Yoo et al.
IV Estimation. To verify findings from the WLS esti- 2012, Parker et al. 2017), which in turn engenders nov-
mation, we use IVs as an alternative empirical strate- el entrepreneurial opportunities.
gy to estimate the effects of OSC. Columns (5) to (8) of Against this backdrop, this paper aims to better
Table 4 present the results from this IV estimation. We conceptualize and understand OSC in digital entre-
begin by noting that the Cragg-Donald Wald F statis- preneurship. Despite the prevailing OSC in digital en-
tic in each model is well above the critical value for trepreneurship, there is an inherent tension in the use
the weak instrument test (Stock and Yogo 2005), of openness in entrepreneurial initiatives (Nambisan
which rejects the null hypothesis that the instruments et al. 2018). Research on the paradox of openness sug-
are weak. From the estimates, we find that they have gests that openness tends to promote value creation,
consistent signs as the ones from the WLS estimation on the one hand, and undercut value capture, on the
across the four models, suggesting that our findings other (West and Gallagher 2006, Laursen and Salter
are robust to the potential threats of sample selection 2014, Wadhwa et al. 2017). As such, a theoretically im-
bias and omitted variable bias. Although true causal portant and practically relevant puzzle in organizing
claims are always difficult with observational data, for digital entrepreneurship is whether and how digi-
the consistent results from different methods help tal startups actually realize value from OSC. Our re-
move a step further in interpreting the role of OSC in search examines this puzzle and presents a novel
 Lin and Maruping: Open Source Collaboration in Digital Startups
Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS 225

conceptual framework to explicate how the value of openness in digital entrepreneurship (Nambisan 2017,
OSC to digital startups unfolds across different stages Nambisan et al. 2018), but offers little guidance on
of venture maturity and modes of OSC engagement. how digital startups should organize for value creation
Our theorizing and evidence suggest that digital start- and value capture in community-based and platform-
ups in the conception and commercialization stages oriented environments. We suggest that the OSC-
realize higher value from their inbound OSC, whereas value nexus rests on not only mode of OSC and stage
of venture maturity, but also the fit between the two.
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

those in the growth stage realize higher value from

their outbound OSC. By juxtaposing the potential for value creation and val-
Moreover, our auxiliary results foreshadowed inter- ue capture as well as the dominant problems at differ-
esting patterns when modes of OSC and stages of ven- ent venture stages, we theorize and demonstrate the
ture maturity are considered separately. Between the relative importance of inbound OSC in the conception
two modes of OSC, we found that inbound OSC, over- and commercialization stages and outbound OSC in
all, has a larger impact on digital startups’ value than the growth stage for digital startups. We thus shed
outbound OSC. This finding is in sharp contrast with light on a key question in digital entrepreneurship re-
Marullo et al. (2018), who examined how open inno- garding how the impact of inbound and outbound
vation influences startup success and found outbound knowledge flows changes with the phase of the venture
openness to be more beneficial than inbound open- life cycle (Nambisan et al. 2018). More broadly, our
ness. It is noteworthy that Marullo et al. (2018) studied findings on the evolving roles of inbound and out-
formal open innovation and quantified the breadth of bound OSC across the development stages of a new
inbound/outbound knowledge flows in terms of the venture have implications for the use of openness in en-
types of open innovation used by the startup (e.g., IP trepreneurship in ways that elucidate “why, when, and
licensing, R&D alliance, etc.). In contrast, we focus on how different modes of action are used to exploit entre-
only one type of informal open innovation (i.e., OSC) preneurial opportunities” (Shane and Venkataraman
and emphasize the depth of openness in terms of the 2000, p. 218).
number of repositories being created or forked. This Second, in examining the tension of OSC in digital
implies that the formality of openness initiatives (for- startups, we also contribute to work on the paradox
mal versus informal) and the characteristics of en- of openness. Our review of the extant literature re-
gagement (breadth versus depth) could affect how vealed that whereas novel conceptualizations have
startups realize benefits from inbound/outbound been advanced to address the tension in the paradox
openness. As for the stages of venture maturity, prior of openness (Alexy et al. 2018, Zobel and Hagedoorn
research has suggested that internal resources and ca- 2020), the temporal dimension had been largely
pabilities are key contingent factors in realizing the overlooked. Indeed, Smith et al. (2017) point out that
benefits of open innovation (West and Gallagher 2006, “scholars often assume that the poles of paradox re-
Dahlander and Gann 2010, Bogers et al. 2017). These main stable; and that their dynamics depend on how
factors naturally grow as digital startups move to later individuals experience the poles and/or the relation-
venture maturity stages, and may explain the in- ship between the poles” (p. 312). Our theorizing
creased effect of OSC across stages of venture maturi- sheds light on a temporal perspective of the paradox
ty in our results. One interesting and unexpected find- of openness by explicitly recognizing the dynamic
ing from our empirical analysis is that conception poles in the paradox and approaching the paradox
stage digital startups do not realize value from OSC. through temporally separating venture stages. Giv-
This implies that a certain level of internal capabilities en that new venture creation is an unfolding process
or resources is needed in order to begin to capitalize (Shane and Venkataraman 2000, Davidsson and
the value of OSC as predicted in the absorptive capac- Gruenhagen 2021), we believe our theoretical model
ity literature (Zahra and George 2002). It reinforces and arguments serve as a sensitizing device that is
the importance of temporal dynamics and processes especially useful in explaining and managing the
in understanding how new ventures organize for tension of OSC in digital entrepreneurship. By over-
open innovation in emerging large-scale collaboration laying the temporal dimension (venture stages) with
environments. the spatial dimension (OSC modes) to better make
sense of OSC in digital entrepreneurship, we also
Contributions and Implications draw attention to the need for future scholarship to
On the theme of emerging technologies and organizing, incorporate multiple dimensions of paradox when
this study makes several noteworthy contributions. examining the paradox of openness.
First, we contribute to the digital entrepreneurship Third, this study contributes to the open innovation
literature by developing a theoretical framework on literature. Although open innovation has attracted
organizing innovation on GitHub—a large-scale open substantial research interest, a significant gap remains
collaboration platform. The literature has recognized in our understanding of how startups should organize
 Lin and Maruping: Open Source Collaboration in Digital Startups
226 Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS

for such activity (Greul et al. 2018). Indeed, Laursen unbounded and distributed rather than selective and
and Salter (2014, p. 877) argue that “we need more targeted. This is important because in platform envi-
knowledge about how choices about appropriability ronments, partners fluidly come and go—making it in-
and openness shape the firm’s performance, perhaps efficient to establish relational contracts on a partner-
especially in the case of young, small firms.” Our con- by-partner basis. Meanwhile, the scope, functionality,
textual focus on digital startups surfaces consequen- and value of digital innovations are dynamic and con-
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

tial tensions that they must navigate if they are to real- tinuously evolving—making it difficult to employ a
ize value from open innovation. In addition, scholars “targeted flashlight approach to creating and identify-
have pointed out that extant open innovation research ing value” (Felin and Zenger 2020, p. 230). As emerg-
draws heavily on case studies or self-reported innova- ing digital technologies continue to be a key driver of
tion surveys. There is need for “better measurement” new entrepreneurial opportunities and the nature of
and “greater generalizability” (West et al. 2014, p. innovation becomes increasingly open and collabora-
807), especially in ways that expose the temporal di- tive, this research adds to our understanding of open
mension of open innovation (Wadhwa et al. 2017) and innovation with a platform logic and a digital perspec-
decouple inbound and outbound knowledge flows tive. In particular, our findings imply that the roles of
(Greul et al. 2018). Our study thus offers a useful com- firm boundaries may become less salient in manage-
plement to prior open innovation research. ment and organization of digital innovation in plat-
Fourth, as discussed in our overview of the litera- form contexts (Nambisan et al. 2017, Parker et al.
ture, although OSS research has engaged with various 2017). Once new ventures move past the initial concep-
facets of firm organizing for OSC, there are notable tion phase, entrepreneurs and managers should in-
gaps that warrant attention. This research contributes creasingly take a community- and ecosystem-oriented
to the extant OSS literature in the following notable perspective, rather than a firm-specific view, toward
regards. Although prior work identifies how firms or- openness in order to commercialize and grow their
ganize for OSC, they do not explicitly differentiate the digital innovations more effectively.
form of OSC or theorize the implications for value cre-
ation and capture. We contribute by drawing atten- Limitations and Future Directions
tion to the meaningful distinction between inbound This study has limitations that must be acknowledged.
and outbound OSC for making sense of how firms One main limitation is that our research data and de-
derive value from such arrangements. Further, empir- sign rely on publicly observable and quantifiable
ically, the OSS literature on firm OSC has overwhelm- measures of digital startups and their OSC behaviors
ingly focused on mature firms or OSS startups (e.g., on GitHub. As such, unobservable factors related to
Dahlander 2005, Germonprez et al. 2017, Nagle 2018). the technology, startup, or industry could affect our re-
Thus, it offers little guidance on the implications of sults. One salient unobservable is the venture’s busi-
such arrangements for digital ventures, which are ness model. Definitions and interpretations of business
mostly outside the OSS industry. We contribute by models vary significantly in the literature (Massa et al.
contextualizing our theorizing to digital startups and 2017), but in general, the concept cannot be reliably
articulating how different stages of venture maturity quantified with public data. Similarly, factors such as
shape the benefits derived from different modes of the founders’ capabilities or attitudes toward open
OSC. Given distinct differences in the challenges faced source may affect the motivation and performance of
by digital startups relative to established firms or OSC. That being said, our empirical methods were de-
OSS startups, such contextualized theorizing is impor- signed to mitigate endogeneity concerns from such
tant for the OSS literature. Our framework offers the omitted variables in our analysis. Future studies can
conceptual foundation for OSS literature to advance build on our findings and employ case studies, inter-
its understanding of how young firms are affected views, or surveys to tease out how such omitted varia-
by OSC. bles affect the use and benefits of OSC in digital
Finally, and more broadly, our results have implica- startups.
tions for organization theory and design. Contempo- Another limitation of this study is that we only con-
rary frameworks for open innovation such as those sidered OSC on GitHub. For well-known open source
proposed by Zobel and Hagedoorn (2020) and Felin projects, there may exist dedicated online communi-
and Zenger (2020) advocate for more relational con- ties outside GitHub, for example, kernel.org for Linux.
tracts and targeted mechanisms when approaching Similarly, although GitHub is by far the leading and
external partnerships and ideas. Our focus on OSC en- most popular OSC platform, there are many other
abled by emerging technology uncovers an alternative similar platforms, such as SourceForge, Bitbucket, and
possibility—in platform environments and for digital GitLab, where startups can engage in OSC. We did
innovations, firms may not need to establish relational not observe OSC outside of GitHub. Thus, our find-
contracts, and the nature of knowledge flows is more ings should be interpreted in light of this limitation.
 Lin and Maruping: Open Source Collaboration in Digital Startups
Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS 227

An extension of our work is to explore startups’ plat- average, each has 62 public repositories on GitHub. All of these 15
form choices for OSC and to consider their OSC activi- unicorns are digital startups and none of their business models are
about selling open source products or services. Therefore, their pres-
ties on different platforms. ence on GitHub and engagement with open source communities
suggests that the value of OSC in digital entrepreneurship goes be-
Conclusion yond monetizing open source per se.
Emerging digital technologies give rise to digital en- 3
To be clear, digital startups can, and often do, pursue open and
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

trepreneurship and the use of OSC to identify and ex- closed innovations at the same time, and their products may have
ploit entrepreneurial opportunities. This study offers both open- and closed-source components. However, at a finer level,
such as a software package, component, or module, it remains a bi-
theoretical insights and empirical evidence into the in-
nary choice whether to share or to withhold the innovation in OSC.
terplay between digital startups’ OSC and firm value. 4
We searched the top 28 journals related to innovation manage-
We highlight the tension in openness and appropri- ment and identified 14 articles on the paradox of openness. Details
ability and present a theoretical framework to unpack and results of our literature search and coding are available at re-
the effect of OSC to digital ventures. We show that quest from the corresponding author.
whereas digital startups can indeed create and cap- 5
We acknowledge that not all startups receive VC funding and if
ture value from engaging in OSC, the effect is regulat- they do, information about firm valuation may not be released to the
ed by the stage of venture maturity and the mode of public. This induces measurement errors into our DV. Such mea-
surement errors would result in an attenuation bias in the estimated
OSC engagement. In doing so, this study advances ex-
coefficients (Wooldridge 2002). In other words, in the presence of
tant digital entrepreneurship and open innovation re- the measurement errors our estimates would be more conservative
search and provides a nuanced account of utilizing than the true effects.
OSC to realize value in the process of digital venture 6
Formally, let zk denote treatment assignment (zk  1 if startup k is
creation. in the treatment group; zk  0 if in the control group) and X k denote
a vector of attributes about the startup before treatment. The pro-
Acknowledgments pensity score of startup k is defined as ek  P(zk  1 | X k ), which
needs to be estimated from data. We denote the estimated propensi-
The authors express their sincerest gratitude to the special
ty score of startup k by e
k : We use a logistic regression and observa-
issue editors for guiding them through the review process.
tions from a pretreatment period to estimate the propensity score.
The authors are grateful to the anonymous reviewers, Once we obtain e
k , the startup-specific weight for digital startup k
whose feedback helped to improve the quality of the man- in WLS is: wk  zekk + 1−e
1−zk
:
uscript. The authors benefited from the many insights

k
7
For each startup i, its OSC behaviors can be influenced by its com-
of participants in the 2020 Association for Information Sys- petitors (denoted as C1), and the latter group can be further influ-
tems (AIS) Special Interest Group on Digital Innovation, enced by its own competitors (denoted as C2) which are not in i’s
Transformation, and Entrepreneurship (SIG DITE) paper business categories. Since there can be many startups in C2, we use
development workshop, organized by Jan Recker, Sanja textual descriptions of companies from Crunchbase to calculate the
Tumbas, and Julian Lehmann. The authors also acknowl- cosine similarity between each startup in C2 and startup i. We de-
edge feedback from presentations of this research at Uni- note startups in C2 that have an above average cosine similarity
versity of Mannheim, Western University, University of with startup i as C3. Taken together, the degree of OSC among start-
Georgia, University of Miami, HEC Montreal, Hong Kong ups in C3 can indirectly influence startup i’s OSC behaviors through
Baptist University, and the University of Queensland’s startups in C1 (the relevance condition of an IV). Meanwhile, since
startups in C3 are not in the same business categories as startup i,
Business Information Systems (BIS) Research Conference.
the degree of OSC in C3 should have no influence on the value of
startup i (the exclusion restriction of an IV). Given that our models
Endnotes have three endogenous variables (i.e., overall, outbound, and inbound
1
Our definition of digital startups draws on Steininger (2019), who OSC), the instruments for each startup are constructed as the mean
emphasizes digitalization at the core of a new venture’s business of the respective endogenous measure in their corresponding C3.
model. This is more restrictive than some of the earlier conceptuali- For interaction terms that involve the moderator and an endoge-
zations in the literature. For instance, Davidson and Vaast (2010, p. 2) nous variable, we create additional instruments based on the inter-
referred to digital entrepreneurship as “the pursuit of opportunities action of the moderator and the base IV.
based on the use of digital media and other information and commu- 8
In an unreported analysis, we implement alternative measures for
nication technologies.” Huang et al. (2017, p. 301) considered digital startup value and OSC. For startup value, we use the total amount of
startups as “start-ups growing by drawing on and adding to digital capital raised to represent startup’s value. As for OSC activities, we
infrastructures.” von Briel et al. (2021, p. 33) suggested that digital quantify them using GitHub commits made by the startup. Results
entrepreneurship “focused on creating new economic activities em- based on these alternative measures yield qualitatively similar conclu-
bodied in or enabled by digital technologies.” In a modern digital sions and are available on request from the corresponding author.
economy, these earlier definitions may be too broad and cover any
startup that sells products on the Internet. For the purpose of our
study, we opt for a more restrictive definition that is more congruent References
with our theorizing and sampling. Abrahamson E, Rosenkopf L (1997) Social network effects on the ex-
2
See https://www.cbinsights.com/research-unicorn-companies (ac- tent of innovation diffusion: A computer simulation. Organ. Sci.
cessed on April 22, 2020). Based on our analysis, 15 of the top 20 uni- 8(3):289–309.
corns have public GitHub repositories: Bytedance, Stripe, Didi Agerfalk P, Fitzgerald B (2008) Outsourcing to an unknown work-
Chuxing, Instacart, Klarna, Epic Games, Databricks, Nubank, DJI force: Exploring opensourcing as a global sourcing strategy.
Innovations, SHEIN, Checkout, Canva, Grab, Plaid, BYJU’s. On Management Inform. Systems Quart. 32(2):385–409.
 Lin and Maruping: Open Source Collaboration in Digital Startups
228 Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS

Alexy O, West J, Klapper H, Reitzig M (2018) Surrendering control Fosfuri A, Giarratana MS, Luzzi A (2008) The penguin has entered
to gain advantage: Reconciling openness and the resource- the building: The commercialization of open source software
based view of the firm. Strategic Management J. 39(6):1704–1727. products. Organ. Sci. 19(2):292–305.
Arora A, Athreye S, Huang C (2016) The paradox of openness revis- Germonprez M, Kendall JE, Kendall KE, Mathiassen L, Young B,
ited: Collaborative innovation and patenting by UK innovators. Warner B (2017) A theory of responsive design: A field study
Res. Policy 45(7):1352–1361. of corporate engagement with open source communities.
Austin PC, Stuart EA (2015) Moving toward best practice when us- Inform. Systems Res. 28(1):64–83.
ing inverse probability of treatment weighting (IPTW) using Gompers P, Kovner A, Lerner J, Scharfstein D (2010) Performance
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

the propensity score to estimate causal treatment effects in ob- persistence in entrepreneurship. J. Financial Econom. 96(1):18–32.
servational studies. Statist. Medicine 34(28):3661–3679. Greul A, West J, Bock S (2018) Open at birth? Why new firms
Baldwin CY, Clark KB (2006) The architecture of participation: Does do (or don’t) use open innovation. Strategic Entrepreneurship J.
code architecture mitigate free riding in the open source devel- 12(3):392–420.
opment model? Management Sci. 52(7):1116–1127. Gruber M, Henkel J (2006) New ventures based on open innova-
Baldwin C, von Hippel E (2011) Modeling a paradigm shift: From tion—An empirical analysis of start-up firms in embedded
producer innovation to user and open collaborative innovation. Linux. Internat. J. Tech. Management 33(4):356–372.
Organ. Sci. 22(6):1399–1417. Haefliger S, von Krogh G, Spaeth S (2008) Code reuse in open
Bogers M, Zobel AK, Afuah A, Almirall E, Brunswicker S, source software. Management Sci. 54(1):180–193.
Dahlander L, Frederiksen L, et al. (2017) The open innovation Harhoff D, Henkel J, von Hippel E (2003) Profiting from voluntary
research landscape: Established perspectives and emerging information spillovers: How users benefit by freely revealing
themes across different levels of analysis. Indust. Innovation their innovations. Res. Policy 32(10):1753–1769.
24(1):8–40. Henkel J, Schöberl S, Alexy O (2014) The emergence of openness:
Bonaccorsi A, Giannangeli S, Rossi C (2006) Entry strategies under How and why firms adopt selective revealing in open innova-
competing standards: Hybrid business models in the open tion. Res. Policy 43(5):879–890.
source software industry. Management Sci. 52(7):1085–1098. Howison J, Crowston K (2014) Collaboration through open superpo-
Capra E, Francalanci C, Merlo F, Rossi-Lamastra C (2011) Firms’ in- sition: A theory of the open source way. Management Inform.
volvement in open source projects: A trade-off between soft- Systems Quart. 38(1):29–50.
ware structural quality and popularity. J. Systems Software 84(1): Hoyer WD, Chandy R, Dorotic M, Krafft M, Singh SS (2010) Con-
144–161. sumer cocreation in new product development. J. Service Res.
Chesbrough HW (2003) Open Innovation: The New Imperative for Cre- 13(3):283–296.
ating and Profiting from Technology (Harvard Business Review Huang J, Henfridsson O, Liu M, Newell S (2017) Growing on
Press, Boston, MA). steroids: Rapidly scaling the user base of digital ventures
Colombo MG, Piva E, Rossi-Lamastra C (2013) Authorising employ- through digital innovation. Management Inform. Systems Quart.
ees to collaborate with communities during working hours: 41(1):301–314.
When is it valuable for firms? Long Range Planning 46(3):236–257. Huang H, Parker G, Tan Y, Xu H (2020) Altruism or shrewd busi-
Dahlander L (2005) Appropriation and appropriability in open source ness? Implications of technology openness on innovations and
software. Internat. J. Innovation Management 09(03):259–285. competition. Management Inform. Systems Quart. 44(3):1049–1071.
Dahlander L, Gann DM (2010) How open is innovation? Res. Policy Kallinikos J, Aaltonen A, Marton A (2013) The ambivalent ontology of
39(6):699–709. digital artifacts. Management Inform. Systems Quart. 37(2):357–370.
Dahlander L, Wallin MW (2006) A man on the inside: Unlocking Kane GC, Ransbotham S (2016) Content as community regulator: The
communities as complementary assets. Res. Policy 35(8):1243– recursive relationship between consumption and contribution in
1259. open collaboration communities. Organ. Sci. 27(5):1258–1274.
Dahlander L, Gann DM, Wallin MW (2021) How open is innova- Karanam A, Agarwal A, Barua A (2020) Designing for Visibility: The
tion? A retrospective and ideas forward. Res. Policy 50(4): Case of Mobile Apps (Social Science Research Network, Roches-
104218. ter, NY).
Davidson E, Vaast E (2010) Digital entrepreneurship and its socio- Kazanjian RK (1988) Relation of dominant problems to stages of
material enactment. 2010 43rd Hawaii Internat. Conf. Systems Sci. growth in technology-based new ventures. Acad. Management J.
(IEEE, Honolulu, HI), 1–10. 31(2):257–279.
Davidsson P, Gruenhagen JH (2021) Fulfilling the process promise: Kyriakou H, Nickerson J, Sabnis G (2017) Knowledge reuse for cus-
A review and agenda for new venture creation process re- tomization: Metamodels in an open design community for 3D
search. Entrepreneurship Theory Practice 45(5):1083–1118. printing. Management Inform. Systems Quart. 41(1):315–332.
Fan Y (2013) Ownership consolidation and product characteristics: Lakhani KR, Lifshitz-Assaf H, Tushman M (2013) Open innova-
A study of the US daily newspaper market. Amer. Econom. Rev. tion and organizational boundaries: task decomposition,
103(5):1598–1628. knowledge distribution and the locus of innovation. Grand-
Faraj S, von Krogh G, Monteiro E, Lakhani KR (2016) Online ori A, ed. Handbook of Economic Organization: Integrating Eco-
community as space for knowledge flows. Inform. Systems. Res. nomic and Organization Theory (Edward Elgar, Cheltenham,
27(4):668–684. UK), 355–382.
Felin T, Zenger TR (2014) Closed or open innovation? Problem solv- Laursen K, Salter AJ (2014) The paradox of openness: Appropriabil-
ing and the governance choice. Res. Policy 43(5):914–925. ity, external search and collaboration. Res. Policy 43(5):867–878.
Felin T, Zenger TR (2020) Open innovation: A theory-based view. Leiponen A, Byma J (2009) If you cannot block, you better run:
Strategic Management Rev. 1(2):223–232. Small firms, cooperative innovation, and appropriation strate-
Fisher G, Kotha S, Lahiri A (2016) Changing with the times: An inte- gies. Res. Policy 38(9):1478–1488.
grated view of identity, legitimacy, and new venture life cycles. Lerner J, Tirole J (2002) Some simple economics of open source. J.
Acad. Management Rev. 41(3):383–409. Indust. Econom. 50(2):197–234.
Foege JN, Lauritzen GD, Tietze F, Salge TO (2019) Reconceptual- Levine SS, Prietula MJ (2014) Open collaboration for innovation:
izing the paradox of openness: How solvers navigate Principles and performance. Organ. Sci. 25(5):1414–1433.
sharing-protecting tensions in crowdsourcing. Res. Policy 48(6): Majchrzak A, Cooper LP, Neece OE (2004) Knowledge reuse for in-
1323–1339. novation. Management Sci. 50(2):174–188.
 Lin and Maruping: Open Source Collaboration in Digital Startups
Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS 229

Mann RJ, Sager TW (2007) Patents, venture capital, and software Steininger DM (2019) Linking information systems and entrepre-
start-ups. Res. Policy 36(2):193–208. neurship: A review and agenda for IT-associated and digital
Marullo C, Casprini E, Di Minin A, Piccaluga A (2018) “Ready for entrepreneurship research. Inform. Systems J. 29(2):363–407.
take-off”: How open innovation influences startup success. Cre- Stewart KJ, Ammeter AP, Maruping LM (2006) Impacts of license
ativity Innovation Management 27(4):476–488. choice and organizational sponsorship on user interest and de-
Massa L, Tucci CL, Afuah A (2017) A critical assessment of business velopment activity in open source software projects. Inform.
model research. Acad. Management Ann. 11(1):73–104. Systems Res. 17(2):126–144.
Medappa PK, Srivastava SC (2019) Does superposition influence the Stock JH, Yogo M (2005) Testing for weak instruments in linear IV
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.

success of FLOSS projects? An examination of open-source soft- regression. Andrews DWK, Stock JH, eds. Identification and In-
ware development by organizations and individuals. Inform. ference for Econometric Models (Cambridge University Press,
Systems Res. 30(3):764–786. Cambridge, UK), 80–108.
Nagle F (2018) Learning by contributing: Gaining competitive ad- Teece DJ (1986) Profiting from technological innovation: Implica-
vantage through contribution to crowdsourced public goods. tions for integration, collaboration, licensing and public policy.
Organ. Sci. 29(4):569–587. Res. Policy 15(6):285–305.
Nambisan S (2017) Digital entrepreneurship: Toward a digital tech- Teece DJ (2018) Profiting from innovation in the digital economy:
nology perspective of entrepreneurship. Entrepreneurship Theory Enabling technologies, standards, and licensing models in the
Practice 41(6):1029–1055. wireless world. Res. Policy 47(8):1367–1387.
Nambisan S, Siegel D, Kenney M (2018) On open innovation, plat- von Briel F, Davidsson P, Recker J (2018) Digital technologies as ex-
forms, and entrepreneurship. Strategic Entrepreneurship J. 12(3): ternal enablers of new venture creation in the IT hardware sec-
354–368. tor. Entrepreneurship Theory Practice 42(1):47–69.
Nambisan S, Lyytinen K, Majchrzak A, Song M (2017) Digital inno- von Briel F, Recker J, Selander L, Jarvenpaa S, Hukal P, Yoo Y, Leh-
vation management: Reinventing innovation management mann J, et al. (2021) Researching digital entrepreneurship: Cur-
research in a digital world. Management Inform. Systems Quart. rent issues and suggestions for future directions. Comm. Assoc.
41(1):223–238. Inform. Systems 48(33):284–304.
Parker G, Alstyne M, Jiang X (2017) Platform ecosystems: How von Krogh G, Spaeth S, Lakhani KR (2003) Community, joining,
developers invert the firm. Management Inform. Systems Quart. and specialization in open source software innovation: a case
41(1):255–266. study. Res. Policy 32(7):1217–1241.
Poole MS, Van de Ven AH (1989) Using paradox to build manage- von Krogh G, Haefliger S, Spaeth S, Wallin MW (2012) Carrots and
ment and organization theories. Acad. Management Rev. 14(4): rainbows: Motivation and social practice in open source soft-
562–578. ware development. Management Inform. Systems Quart. 36(2):
Schaarschmidt M, Walsh G, von Kortzfleisch HFO (2015) How do 649–676.
firms influence open source software communities? A frame- Wadhwa A, Bodas Freitas IM, Sarkar MB (2017) The paradox of
work and empirical analysis of different governance modes. In- openness and value protection strategies: Effect of extramural
form. Organ. 25(2):99–114. R&D on innovative performance. Organ. Sci. 28(5):873–893.
Schad J, Lewis MW, Raisch S, Smith WK (2016) Paradox research in West J, Gallagher S (2006) Challenges of open innovation: The para-
management science: Looking back to move forward. Acad. dox of firm investment in open-source software. R & D Manage-
Management Ann. 10(1):5–64. ment 36(3):319–331.
Serrano CJ, Ziedonis R (2019) How redeployable are patent assets? West J, Kuk G (2016) The complementarity of openness: How Mak-
Evidence from failed startups. Acad. Management Proc. (Acade- erBot leveraged Thingiverse in 3D printing. Tech. Forecasting
my of Management, Briarcliff Manor, NY), 15631. Soc. Change 102:169–181.
Shaikh M, Levina N (2019) Selecting an open innovation community West J, Salter A, Vanhaverbeke W, Chesbrough H (2014) Open inno-
as an alliance partner: Looking for healthy communities and vation: The next decade. Res. Policy 43(5):805–811.
ecosystems. Res. Policy 48(8):103766. Wooldridge JM (2002) Econometric Analysis of Cross Section and Panel
Shane S (2000) Prior knowledge and the discovery of entrepreneur- Data (MIT Press, Cambridge, MA).
ial opportunities. Organ. Sci. 11(4):448–469. Yacoub G, Storey C, Haefliger S (2020) Appropriability mechanisms
Shane S, Venkataraman S (2000) The promise of entrepreneur- for manufacturing and service firms: The contingencies of
ship as a field of research. Acad. Management Rev. 25(1): openness and knowledge intensity. R & D Management 50(5):
217–226. 551–572.
Smith WK, Lewis MW (2011) Toward a theory of paradox: A dy-
Yoo Y, Henfridsson O, Lyytinen K (2010) The new organizing logic
namic equilibrium model of organizing. Acad. Management Rev.
of digital innovation: An agenda for information systems re-
36(2):381–403.
search. Inform. Systems Res. 21(4):724–735.
Smith WK, Erez M, Jarvenpaa S, Lewis MW, Tracey P (2017) Add-
ing complexity to theories of paradox, tensions, and dualities of Yoo Y, Boland RJ, Lyytinen K, Majchrzak A (2012) Organizing for
innovation and change. Organ. Stud. 38(3–4):303–317. innovation in the digitized world. Organ. Sci. 23(5):1398–1408.
Sojer M, Henkel J (2010) Code reuse in open source software devel- Zahra SA, George G (2002) Absorptive capacity: A review, reconcep-
opment: Quantitative evidence, drivers, and impediments. J. tualization, and extension. Acad. Management Rev. 27(2):185–203.
Assoc. Inform. Systems 11(12):868–901 Zittrain J (2006) The generative internet. Harvard Law Rev. 119:
Spaeth S, von Krogh G, He F (2015) Perceived firm attributes and 1974–2040.
intrinsic motivation in sponsored open source software projects. Zobel AK, Hagedoorn J (2020) Implications of open innovation for
Inform. Systems Res. 26(1):224–237. organizational boundaries and the governance of contractual
Stam W (2009) When does community participation enhance relations. Acad. Management Perspect. 34(3):400–423.
the performance of open source software companies? Res. Policy
38(8):1288–1299. Yu-Kai Lin is an assistant professor of computer informa-
Stefan I, Bengtsson L (2016) Appropriability: A key to opening tion systems at the Robinson College of Business, Georgia
innovation internationally? Internat. J. Tech. Management 71(3–4): State University. He received his PhD from the Eller College
232–252. of Management at the University of Arizona. He studies a
 Lin and Maruping: Open Source Collaboration in Digital Startups
230 Organization Science, 2022, vol. 33, no. 1, pp. 212–230, © 2021 INFORMS

broad range of topics including digitalization, open innova- innovation in small- and large-scale collectives such as
tion, intellectual property, health information technology, teams, communities, and crowds. His interests in this area
and business analytics. include the enabling role of digital collaboration platforms,
Likoebe M. Maruping is professor of computer informa- the mechanisms underlying the collaboration process, and
tion systems in the Robinson College of Business, Georgia the leadership and governance of collaborative efforts in or-
State University. His research focuses on collaboration and ganizational and open environments.
Downloaded from informs.org by [218.89.242.93] on 20 September 2024, at 05:13 . For personal use only, all rights reserved.