Research Policy 43 (2014) 632–644
Contents lists available at ScienceDirect
Research Policy journal homepage: www.elsevier.com/locate/respol
Acceptance of monetary rewards in open source software development夽 Sandeep Krishnamurthy a,1 , Shaosong Ou b,2 , Arvind K. Tripathi c,∗ a b c
School of Business, University of Washington, Bothell, 18115 Campus Way NE, Room UW1-381B, Bothell, WA 98011-8246, United States Information Systems and Operations Management, Michael G. Foster School of Business, University of Washington, Seattle, WA 98195-3200, United States Information Systems and Operations Management, University of Auckland Business School, Auckland 1142, New Zealand
a r t i c l e
i n f o
Article history: Received 8 April 2012 Received in revised form 27 June 2013 Accepted 20 October 2013 Available online 27 December 2013 Keywords: Open source software Monetary rewards Crowd in Sustainability of open source software
a b s t r a c t The open source software (OSS) movement thrives on innovation and volunteer effort of developers. Scholars have expressed widespread concern about the sustainability of the OSS movement due to high levels of volunteerism. In this paper, we address a central challenge to the sustainability of OSSdevelopers’ acceptance of monetary rewards. We strive to explain why some OSS developers accept monetary rewards and others do not. Viewed through the theoretical lens of the private-collective innovation model (Von Hippel and Von Krogh, 2003, 2006), this allows us to describe when developers will accept private financial rewards. Our main research objective is to clearly map the web of relationships between causal antecedents, and developers’ acceptance behavior. Using a unique dataset that combines survey and behavioral measures, we find that – (a) intention to accept monetary rewards mediates the impact of motivational elements on developers’ acceptance of monetary rewards; (b) intrinsic and extrinsic motivations positively affect their intention to accept monetary rewards, community motivation negatively impacts intention and ideological motivation does not affect the intention to accept rewards and (c) these effects are obtained even after inclusion of several control variables. The theoretical and managerial implications of our work are described. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The central idea behind open source software is that greater market innovation results when developers are provided the necessary freedom to create new products (Raymond, 1998; Ghosh, 1998). Developers work in loosely organized communities built on the private-collective innovation model, i.e., a “best of both worlds” scenario where programmers make private gains while contributing to the collective good (Von Hippel and Von Krogh, 2003, 2006) through “freely revealing” the source code for the product (Von Krogh and Von Hippel, 2006). Open source software “democratizes” the innovation process (Von Hippel, 2005) by making code freely available through diverse licensing arrangements (Krishnamurthy, 2003). The extensive and free distribution of the code allows a large group of individuals to examine and use software leading to a more reliable and robust product (Bonaccorsi and Rossi, 2003;
夽 Order of authors is alphabetical – all the authors contributed equally. ∗ Corresponding author. Tel.: +64 09 923 4922; fax: +64 09 373 7430. E-mail addresses:
[email protected] (S. Krishnamurthy),
[email protected] (S. Ou),
[email protected] (A.K. Tripathi). 1 Tel.: +1 425 352 5229; fax: +1 425 352 5277. 2 Tel.: +1 206 543 0153; fax: +1 206 543 3968. 0048-7333/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.respol.2013.10.007
Rossi, 2004). Unlike proprietary software, the open source software movement is characterized by a “diversity of project structures, diverse employment arrangements, co-existence of corporations and communities and co-existence of the creative and commercial elements” (Krishnamurthy, 2006). In contrast to the proprietary software regime where an organization makes a limited set of strategic choices, open source promotes multiple innovation trajectories. The open source literature has provided different perspectives on the relative place of private vs. collective rewards to the developer. The early characterizations described an altruistic, community-minded developer interested in purely collective gains through volunteerism (Raymond, 1998; Ghosh, 1998). Of late, the focus has turned to the sustainability of the open source software movement, i.e., whether open source innovation can sustain if developers are not making private gains. Scholars have expressed widespread concern about the sustainability of the OSS movement due to high levels of volunteerism (Von Hippel and Von Krogh, 2003). It has been argued that while open source developers may derive joy from coding and might be ideologically motivated, many may “simply wish to earn a reasonable livelihood from their efforts” (Fitzgerald, 2006). In the current system where open source and proprietary software often co-exist in the same marketplace (Bonaccorsi and Rossi, 2003), corporations may be “harvesting the
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
altruism” of volunteer developers (Haruvy et al., 2003) leading to a system with “altruistic individuals and selfish firms” (Rossi and Bonaccorsi, 2005; Bonaccorsi and Rossi, 2004). There is a traditional free-rider problem in this situation where corporations often benefit from the source code of open source software without providing any compensation to the developers (Haruvy et al., 2008). Financial incentives in the open source landscape (i.e., the pathways to private financial gain) are of many types. They can be categorized based on the distribution pattern (all or few), type of provider (corporation, individual), contingent or fixed and conditional or not (Krishnamurthy, 2006). Open source developers might be paid a fixed salary by a non-profit organization or forprofit corporation (Roberts et al., 2006), offered a bounty to solve a particular problem (Krishnamurthy and Tripathi, 2006) or be paid through voluntary contributions (Krishnamurthy and Tripathi, 2009). Theoretically, our view is that these differences matter in our understanding of when developers will accept private financial rewards. The implicit assumption made by the literature thus far (e.g., Von Hippel and Von Krogh, 2003) is that when financial rewards are made available to open source developers, they will simply accept them. Put otherwise, the literature has examined the provision of and not the acceptance of these incentives. Interestingly, empirical data show that not all developers accept financial incentives when they are provided. Consider the case of Mozilla which offers a regular bounty to identify bugs in its code. A news story reports that 10–15% of participants turn down the financial rewards.1 This is a poorly understood phenomenon and ties directly to the sustainability of the OSS movement. The implications for the private-collective innovation model (Von Hippel and Von Krogh, 2003) when many developers do not accept financial incentives are considerable. In effect, it switches from the privatecollective innovation model to simply the collective innovation model (Osterloh and Rota, 2006) when this happens. Therefore, our work helps deepen our understanding of the relative place of private and collective rewards in the open source ecosystem. This research is also managerially relevant. Understanding why OSS developers do not accept incentives will afford companies, non-governmental organizations (NGOs) and policy makers an opportunity to rethink their strategies. For one, companies are investing considerable amounts in this arena with the hope of attracting developers. These companies assume that developers will be attracted to these financial rewards. However, if there is systematic self-selection due to acceptance behavior, results might be different from what was intended or predicted. If acceptance behavior effects are massive, the resulting self-selection might jeopardize the network externality that companies might be counting on for rapid product development and diffusion (Bonaccorsi and Rossi, 2003; Dahlander, 2007; Bitzer et al., 2007). In order to deepen our understanding, we begin with a short qualitative study. We find there is considerable diversity of opinion on how developers should be compensated. Not surprisingly, the fiercely independent open source developers themselves do not agree on one optimal arrangement. Table 1 summarizes openended responses to the question – “How should open source developers be paid?” This is based on survey data we gathered from OSS developers. Interestingly, one respondent explicitly argues that open source developers should not be paid. This confirms our intuition that not all developers are likely to accept all financial incentives. Based on our analysis, we identified ten themes from the qualitative data – no compensation, voluntary monetary rewards, plurality
1 http://www.wincert.net/news/software/2209-more-than-1-in-10-mozillabug-finders-turn-down-cash-reward.
633
of methods, rewards in kind, subscription, salary/bounty, payfor-service, contextual support, consensus-based decision and software sales. It is beyond the scope of this paper to analyze the developers’ relationship with these nine different types of financial incentives. We focus on one of the identified sources of payment – voluntary monetary rewards. Our research interest in this paper is to identify developers’ acceptance behavior as central to the discourse on OSS sustainability and to clearly map the web of relationships between causal antecedents, and acceptance behavior. We examine how motivations drive developers to accept voluntary monetary rewards for their work on OSS projects. Through this paper, we make the following contributions – (1) we identify acceptance behavior as requiring special attention within the open source literature, (2) we propose a scale to measure the intention to accept financial incentives, (3) we identify the impact of various established motivational components on acceptance behavior, and (4) we provide an empirical study using a dataset that combines attitudinal and behavioral variables from multiple sources to answer our research questions.
2. Literature review OSS developer motivation is an issue that is of great interest to researchers. The current literature has identified several motivational components that drive open source software developers – intrinsic, i.e., originating from the act of participation (Lakhani and Von Hippel, 2003; Lakhani and Wolf, 2005), extrinsic, i.e., originating from external rewards (Lerner and Tirole, 2002, 2005), ideological, i.e., stemming from a strong belief structure in the values and norms underpinning OSS development methodology (Stewart and Gosain, 2006) and community, i.e., deriving from a strong sense of identification with the open source community (Hertel et al., 2003; Jannsen and Huang, 2008). These diverse motivational components are not necessarily mutually exclusive and may co-exist within a developer (Franck and Jungwirth, 1999; Krishnamurthy, 2006; Roberts et al., 2006). The main focus of this stream of literature is to understand how developers’ motivations drive their participation/effort on OSS project which in turn affects overall performance and effectiveness of developers and projects. A summary of main findings from this stream of literature is provided in Table 2. Generally speaking, the link between motivation and effort is well established in various literatures. While the literature is advanced in the identification of motivational components, the focus has been on the relationship between these components and effort/performance at either the individual or team level. While we know about how these motivations affect software-related performance, we do not know much about the link between these components and behaviors related to financial incentives, e.g., it is not clear what motivates some developers to accept some types of financial rewards in some situations. We aim to address that gap in this research. The fundamental tenets of open source software are rooted in the principles of sharing freely with a virtual community of individuals (Raymond, 1998; Ghosh, 1998). These early descriptions of open source viewed financial incentives as applicable only to proprietary software regimes and considered open source as a volunteer-based system driven by ideology and freedom. Reciprocity was identified as one of the motivating factors in participation. Specifically, there was great interest in the generalized exchange model where individuals reciprocate contributions by strangers (Ekeh, 1974; Kollock, 1999; Lakhani and Von Hippel, 2003). Kollock (1999) argues that the General Public License (GPL) enables generalized exchange because it “creates an incentive
634
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
Table 1 Indicative responses to “how should open source developers be paid?”. #
Comment
Theme
1.
“This presumes that developers should be compensated, per se, for their work. I’m not at all certain that I agree with this premise. I have no problem with nominal charges for software (a.k.a. ‘shareware’) and such charges being disbursed to developers. However, making such development an actual livelihood is likely to result in attracting people whose primary motivation is making money, and as far as I am aware, there would seem to be an inverse correlation to such motivation and actual ability as a developer.” “Donations are the most satisfying compensations for work spent on Open Source projects. If people donate for a project, they show that they care about it’s future. Them spending money for something they can also get for free shows that the developer’s work is important for them.” “Fee for service. writing custom code that is outside the main development stream. Donations. Lecture fees. Book Deals. Donation to charities of developers choice (ala sourceforge)” “Donations of any type that will further enhance the project on some way. Including server hardware, bandwidth for servers, hardware for developer computers, hardware for servers, upgrades for servers or developer computer and so on.” “If their only source of income seems to be open source software, then a subscription charge could be considered, or a donate to the programmer arrangement, or helping place him in an environment that would allow him to earn income from other programming tasks as well but not require him to abide by what most view as the (predatory) corporate culture.” “Either by professional position (such as Novell, Red-hat, etc.), or bounty type arrangement.” “Direct payment to accelerate bug fixes and feature adds that are outside of their current plans. Paid for end-user training, programmer training, and documentation.” “I doubt there is a one-size-fits-all answer to this question. Some projects are frivolous; or so easy that the satisfaction of a job well done is more than adequate compensation. Others are so useful or important that governments should divert tax revenue to provide resources and personnel to insure completion and support.” “Short answer: however they want, whatever is agreed upon. Preferably by consensus of the stake holders.” “Corporate sponsorship might be one way to compensate open source developers. However, as an owner of a 1-person software company I find the best way for me to be compensated is by selling my software. Having customers pay for only support and not the software itself would be nice, I personally have not found a way to make this work for my own company. I still must rely on customers buying a license to use certain software products I produce.”
No Payment
2.
3. 4. 5.
6. 7. 8.
9. 10.
structure in which programmers are encouraged to contribute modifications to the program because they are assured that everyone will have access to their contributions and that they will have access to any modifications other people have made, either currently or in the future.” Lakhani and Von Hippel (2003) is one of the few papers in the literature that has explored generalized reciprocity in the open source context. Interestingly, they find that infrequent providers of user-to-user customer service are more likely to be motivated by generalized reciprocity vis-à-vis frequent providers (Lakhani and Von Hippel, 2003, table 14, item #1). Related to the thread on reciprocity, is the literature that views open source from the lens of gift giving (Bergquist and Ljungberg, 2001). Developers freely donate their software programs, source code and their time to volunteer on various tasks such as bug fixing and providing user-to-user assistance (Raymond, 1998; Ghosh, 1998). These are gifts not made to a specific individual, but rather to a large group, i.e., users and developers of open source software (Kollock, 1999). The free nature of the software and code enable product acceptance leading to network externality advantages for the authors (Boncarossi and Rossi, 2003). Gift economies are characterized by norms, beliefs and values (Stewart and Gosain, 2006; Bergquist and Ljungberg, 2001; Mauss, 1955). While the larger focus of this literature has been on nonfinancial rewards, acceptance of rewards is an integral part of this gift economy. As noted by a previous study: “One of the norms of gift giving is the rule that gifts be accepted. Therefore, the receiver becomes subordinate to the giver. In the open source community, this can be the fact. But the opposite is also important. Refusing to accept a gift can, in some situations, be a way of showing superiority.” (Bergquist and Ljungberg, 2001, p. 310) This argument, rooted in the anthropology literature on gift giving, argues that gifting is strategic action. Refusing financial rewards might be a way of strategically positioning oneself to gain reputation within the community. This is, perhaps, one of the earliest arguments for acceptance behavior as a separate area of study. The open source literature has long argued that developers participate in the creation of such software for the sake of rewards.
Voluntary Monetary Rewards as compensation Plurality of Methods Donation in Kind Subscription
Salary/Bounty Pay-for-Service Contextual Support
Consensus-based Decision Software Sales
For instance, Lerner and Tirole (2002) argued that participation in open source might lead to “future job offers” or “future access to the venture capital market” (p. 213). Hars and Ou (2002) also specifically argue that open source participation brings with it four future rewards: revenues from related products and services, human capital, self-marketing and peer recognition. However, volitional acceptance of these rewards has not received much attention. The open source literature on crowding out of motivation as a result of financial rewards (Roberts et al., 2006) assumes that developers do not have a choice in accepting these rewards. The psychology and economics literature on crowding out is based on the idea that externally imposed rewards are controlling in nature. However, “intrinsic motivation may be crowded out or crowded in (if the external intervention is perceived to be supportive instead of controlling)” (Frey and Goette, 1999). But, this literature assumes that financial incentives, once provided, will always be accepted. If developers are actually provided a choice of accepting or rejecting financial rewards, that is likely to be perceived as supportive rather than controlling intervention and might, therefore, not diminish their intrinsic motivation.
3. Hypotheses development 3.1. Motivations and intention to accept monetary rewards Drawing from well-established theories on individual motivations and behavior, we propose several hypotheses linking developers’ motivations to their intention to accept monetary rewards. By definition, extrinsic motivation refers to motivation when driving forces such as rewards, are external to an individual or an open source software developer in this context. Extrinsically motivated individuals do not derive satisfaction from performing the tasks, but from external sources such as rewards which may come before or after performing the task. The need and presence of external rewards has been documented clearly in the OSS literature (Lerner and Tirole, 2002, 2004; Okoli and Oh, 2007). Extrinsically motivated open source developers are motivated by
Table 2 Summary of open source literature on developers’ motivations. Motivational components included
Methodology
Dependent variable
Mediating variables
Sample description
Findings
Hars and Ou (2002)
Internal (self-determination, altruism, community identification), future rewards (selling products, human capital, self-marketing, peer recognition), personal need Enjoyment-based intrinsic motivation, economic-based extrinsic motivation, obligation/community-based intrinsic motivation Intrinsic, reciprocity, reputation (extrinsic), community-oriented, job responsibility (extrinsic) Identification, pragmatic (extrinsic), norm-oriented, social and political, hedonistic (intrinsic) and reward (extrinsic) motives Intrinsic, extrinsic, status (extrinsic), use value (extrinsic) motives
Regression
Effort
None
79 individuals from various open source projects
Selling products, self-marketing and personal need were most strongly correlated with effort
Regression/clustering
Effort (project hours per week)
None
684 software developers in 287 F/OSS projects
Strong support for intrinsic motivation
Survey and Behavior
Related to the provision and quality of customer service Effort (hours per week), Willingness to be involved in Linux development in future Performance (number of changes in ASF rank)
None
336 individuals working for Apache help system
None
141 contributors in a large OSS project (Linux Kernel)
Motivational process in OSS projects is similar to other social communities
Participation (number of source code contributions)
288 software developers in Apache projects
Affective trust, cognitive trust Helping, career advancement, enhancing human capital, satisfying personal needs Restrictiveness of open source license
67 OSS projects
Extrinsic motivations do not crowd-out intrinsic motivations. All motivations have equal effect on OSS participation Ideological motivation affects team effectiveness Satisfaction affects continuance intention, Mixed findings on motivation
Lakhani and Wolf (2005)
Lakhani and Von Hippel (2003) Hertel et al. (2003)
Roberts et al. (2006)
Stewart and Gosain (2006) Wu et al. (2007)
Fershtman and Gandal (2007)
Regression/factor analysis
Structural equation model
Ideological motivation (OSS values, norms, beliefs) Helping, career advancement, enhancing human capital, satisfying personal needs
Structural equation model Structural equation model
Team effectiveness
None included. Findings consistent with intrinsic motivation, status and signaling arguments
Regression analysis
Output per contributor
OSS continuance intention
148 OSS participants
71 open source projects over 18 month period
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
Paper reference
Output per contributor is lower in projects with restrictive licenses
635
636
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
the possibility of receiving monetary and non-monetary rewards (Ryan and Deci, 2000). Such individuals are likely to be influenced by a utilitarian cost-benefit calculus that is the sum of “the immediate payoff (current benefit minus current cost) plus the delayed payoff (delayed benefit minus delayed cost)” (Lerner and Tirole, 2002, p. 213). Acceptance offers the possibility of immediate and delayed payoffs as a result of immediate and delayed benefits with minimal delayed costs. Monetary rewards to developers can be considered as reciprocation from users to Developers. Reciprocity is one of the important underlying forces in Open Source Community exchanges where individuals try to reciprocate at a level higher than they received. Hence we argue that extrinsically motivated individuals are likely to consider monetary rewards as reciprocity from the consumers of their code/contribution. Therefore, it follows that H1. A greater level of extrinsic motivation leads to a greater propensity of an OSS developer’s intention to accept monetary rewards. Intrinsic motivation has been characterized as a motive to participate or engage in activities that are challenging, playful, and novel. An individual’s intrinsic motivations are driven by need for competence, autonomy and creativeness (Deci and Ryan, 1985). Open source software research has shown that the act of participation in open source software development is inherently enjoyable to developers (Lakhani and Von Hippel, 2003; Lakhani and Wolf, 2005). Indeed, intrinsically motivated open source developers derive their satisfaction from the properties of the task (Calder and Staw, 1974; Deci, 1975). They are inherently curious (Amabile et al., 1994) and inspired by the creative process of open innovation and hence, find the freedom offered by open source to be empowering. From a theoretical standpoint, the literature on intrinsic motivation in open source software has presupposed that financial rewards automatically lead to crowding out of intrinsic motivation. Hence, early research found the absence of a relationship between financial rewards and intrinsic motivation to be novel (Lakhani and Von Hippel, 2003; Lakhani and Wolf, 2005), although early work in management has found that crowding out does not always take place (Dermer, 1975). Recent work continues this trend as evidenced by Alexy and Leitner (2011, p. 166): “Yet, surprisingly, extent research in the field (of) OSS development, a prime example of distributed innovation (Lee and Cole, 2003), has found that financial rewards do not crowd out intrinsic motivation in OSS (e.g., Hars and Ou, 2002; Lakhani and Wolf, 2005; Roberts et al., 2006) – which contradicts outright the very theory on which these results are built.” Alexy and Leitner (2011) propose the importance of norms to explain this phenomenon (Stewart and Gosain, 2006). In an experiment, they propose that crowding out occurs only for those individuals who find accepting financial rewards to be “socially unacceptable.” In particular, financial norms moderate the impact of financial rewards in intrinsic motivation – i.e., there are no main effects due to financial rewards or norms, but their interaction impacts the level of intrinsic motivation. They also find the importance of self-determination on total motivation. Lost in this discussion is the lucid explanation of the impact of external rewards on motivation that is found in Frey and Jegen (2001). They identify two possible effects (p. 594): “(1) External interventions crowd-out intrinsic motivation if the individuals affected perceive them to be controlling. In that case, both self-determination and self-esteem suffer, and the individuals react by reducing their intrinsic motivation in the activity controlled.
(2) External interventions crowd-in intrinsic motivation if the individuals concerned perceive it as supportive. In that case, self-esteem is fostered, and individuals feel that they are given more freedom to act, thus enlarging self-determination.” To date, the open source literature has not explored crowding in effects as it pertains to intrinsic motivation. We contribute to the literature by highlighting this possibility. The context of our investigation is dissimilar from the vast literature on motivation. The nature of rewards and their relationship to motivation has been studied extensively in various literatures (Cameron and Pierce (1994) provide a meta-analysis focused on intrinsic motivation, Frey and Jegen, 2001). Generally, in this literature, the provision of rewards is equated with the acceptance of them. The reward may be unexpected, contingent on performance and tangible or otherwise (Cameron and Pierce, 1994). However, in all cases, the principal provides the reward and the agent accepts it. This is true with the open source software literature as well (Alexy and Leitner, 2011; Lakhani and Von Hippel, 2003; Lakhani and Wolf, 2005). An agent provides the rewards (e.g., teacher, manager, and corporation) and the participant simply accepts them. In direct contrast, the developer decides on the provision of rewards in contrast to an external authority figure. Conceptually, we regard this as a three-stage process: Stage 1: Developers accepts possibility of rewards. (or NOT) Stage 2: Any interested party or parties provides rewards. (or NOT) Stage 3: The open source platform transfers these rewards to the developers. Acceptance of rewards is an exercise of freedom – a value that is central to open source software itself (e.g., Bonaccorsi and Rossi, 2003). Since the developer determines whether financial rewards are to be accepted or not, s/he is able to ensure that these rewards do not control the agenda of the project. The spirit of free innovation is maintained while negotiating the ability to receive rewards. Consistent with Frey and Jegen (2001), we would expect that this would crowd in intrinsic motivation. The provider of the reward matters here, as well. In the case of online open source platforms, the giver may be a community member and is likely to be anonymous or pseudonymous. In direct contrast to a salary or a bonus, receiving rewards from an appreciative user removes the controlling relationship between a supervisor and an employee. These monetary rewards do not result from any contract between developers and consumers of the code, rather they are unsolicited and voluntary in nature; hence they are considered unbiased certification of the quality of the code/contribution. Therefore, open source developers are likely to find the receipt of voluntary donations to be “supportive” (Frey and Jegen, 2001) and empowering. This is also consistent with the discussion in Benabou and Tirole (2003, p. 504): “By contrast, rewards that are discretionary (not contracted for) may well boost the agent’s self-esteem or intrinsic motivation (emphasis added)” Therefore, we hypothesize that: H2. A greater level of intrinsic motivation leads to a higher propensity of an OSS developer’s intention to accept monetary rewards. Literature in psychology defines ideology as common values that bonds individuals of a community and helps them achieve collective goals. OSS ideologies have been shaped and developed by community stalwarts such as Richard Stallman and Eric Raymond over time. Stallman argued that OSS development contributes to
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
faster and secure software development. Individuals motivated by OSS ideology consider their contribution to OSS projects more purposeful and meaningful. Ideologically motivated OSS developers closely identify with and derive their identity from the open source movement. They adhere closely to the norms, beliefs and values inherent to open source (Stewart and Gosain, 2006; Bergquist and Ljungberg, 2001) and are more likely to be principle-driven (Lakhani and Wolf, 2005). The central elements of OSS ideology include knowledge sharing, software freedom and voluntary cooperation (Stewart and Gosain, 2006). Indeed, individuals who hold these values understand the need for volunteer effort for OSS development and do not indulge in opportunistic behaviors and are motivated in way to be productive for their project team (Stewart and Gosain, 2006). Those driven by ideological positions are likely to believe that the place of money in software development should be limited (Stallman, 1992). Adherence to OSS ideology reinforces egalitarian and utopian motivations and thus, ideologically motivated individuals are less likely to accept monetary rewards for their effort on OSS projects. Formally, we hypothesize that: H3. A greater level of ideological motivation leads to lower propensity of an OSS developer’s intention to accept monetary rewards. The community orientation of open source software has been pointed out in the earliest descriptions of the phenomenon with the use of evocative phraseology such as “bazaar” (Raymond, 1998) and “cooking pot markets” (Ghosh, 1998). Linus Torvalds echoed the importance of community contribution in an interview2 – “Making Linux freely available is the single best decision I’ve ever made” and “It was a natural decision within the community that I felt I wanted to be a part of”. Hertel et al. (2003) liken developers’ motivations to work on OSS projects, to motivations of volunteers contributing to social communities. Communities are assemblages of volunteers who share a common goal and are united in the successful attainment of a goal. In open source, the evolution of a project fosters the growth of developer community associated with the project, such as Linux Kernel community. Individuals engaged with OSS projects identify themselves as a part of the community (Raymond, 1998) and follow the norms of the community (Lakhani and Wolf, 2005). Members of the project community earn respect and trust through their effort and contribution to the OSS project. These project communities enable “social integration” (Von Hippel and von Krogh, 2003) thereby motivating developers to contribute for the project (Hars and Ou, 2002; Bagozzi and Dholakia, 2006). Developer motivations to identify with these project communities drive their contributions to the project (Hertel et al., 2003). Hars and Ou (2002) also find that developers are community oriented and driven by altruistic motives to join OSS projects. Community motivated developers see themselves as integral part of the community and thus consider benefits of the community above and beyond their personal benefits/gains. Since these developers themselves benefit from others’ work on OSS, they feel obliged to give back to the OSS community by putting their effort on OSS development (Wu et al., 2007; Shah, 2006). Interest in benefiting the community over personal gain is common among altruistic developers (Hars and Ou, 2002; Haruvy et al., 2003) who are motivated by the “perceived benefit to society” (Chakravarty et al., 2007). An altruistic orientation causes developers to have preferences over others’ financial rewards (Haruvy et al., 2003). A community where developers’ status is defined by giving, accepting monetary rewards may seem to signal a desire to work only for paid tasks and reduce the status of such members in the community.
637
Consequently, those who accept monetary rewards risk losing status and respect among their peers and are less likely to seek such rewards. As a result, we hypothesize that: H4. A greater level of community motivation leads to lower propensity of an OSS developer’s intention to accept monetary rewards. 3.2. Intention to accept monetary rewards and actual acceptance behavior The importance of cognition in guiding human behavior has been well documented in social psychology and behavior science literature (Sorrentino and Higgins, 1986). The Theory of Reasoned Action (Ajzen and Fishbein, 1974) is a well-accepted theory in psychology which posits the importance of intention as a mediating construct in the relationship between attitude and performance. A meta-analysis of this literature conducted by Sheppard et al. (1988) documents the wide applicability of this model. The Theory of Reasoned Action has been applied and has influenced work in marketing (e.g., Bagozzi et al., 2000; Shimp and Kavas, 1984), business ethics (e.g., Randall, 1989; Chang, 1998), management information systems (e.g., Karahanna et al., 1999; Venkatesh et al., 2003). Following this extensive literature, we hypothesize that behavioral intention is the immediate antecedent of OSS programmers’ decision to accept monetary rewards. Our hypothesis can then be stated as: H5. OSS developers who report high intention to accept monetary rewards are more likely to accept rewards. These hypotheses are illustrated in Fig. 1. 4. Research method 4.1. Context of investigation We tested our research model with data from open source developers who work on OSS projects hosted on Sourceforge.net. Sourceforge (www.sourceforge.net) was chosen because it is the largest repository of OSS projects, hosting more than 200,000 projects. This site is a free online open source software development platform (Krishnamurthy and Tripathi, 2009) that enables the OSS ecosystem. For this study, the unit of analysis is the individual developer. We use a unique dataset that combines observed behavioral data from Sourceforge with responses to an online survey administered to randomly selected OSS developers on Sourceforge.net. Many previous studies have exclusively used behavioral data available on Sourceforge.net (Lerner and Tirole, 2005; Gonzalez-Barahona et al., 2008). However, at least one paper has suggested that analyzing behavioral data alone may be dangerous for various reasons (Howison and Crowston, 2004). In addition, the limitations of exclusively using survey-based are documented in the literature (Lerner and Tirole, 2005, p. 105). To address these concerns, this study combines data from two different sources – OSS developers’ behavioral data from Sourceforge and survey data to capture developers’ motivations for monetary rewards. To the best of our knowledge, this is the first study to take this approach. Combining observed behavioral data from Sourceforge with survey-based data allows us to perform a robust test of our model while sidestepping the concerns of exclusively using either survey-based or behavioral data. 4.2. Control variables
2 http://firstmonday.org/htbin/cgiwrap/bin/ojs/index.php/fm/article/view/1475/ 1390.
Our research focuses on the relationships between OSS developers’ motivation and their monetary reward acceptance behavior.
638
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
+
Intrinsic Motivation
-
Tenure +
Ideological Motivation
-
+ Intention to Accept Monetary Rewards
Extrinsic Motivation
Subscribe
+
Acceptance of Monetary Rewards
+ +
Community Motivation
Sponsor
-
Income Non-NA-EUR Control Variables
-
Independent Variables Fig. 1. Hypothesized relationships between motivations and acceptance.
A number of other factors could also affect these relationships. In order to control for the possible confounding effects, we included the following five factors as control variables: Tenure, Subscription, OSS Income, Sponsor, and Non-NA-EUR. We discuss these control variables next. Undoubtedly, the OSS platform (in this case, Sourceforge.net) plays an important role in developments and dissemination of OSS projects, which has been acknowledged by the OSS community. Tenure measures the length of the developers’ association with the OSS platform (specifically, the number of days a developer has been registered on Sourceforge). Tenure is a surrogate to measure developers’ association with open source software development teams. Individual with longer tenure have a stronger sense of identification with OSS development teams and are more affiliated with OSS ideology (Krishnamurthy and Tripathi, 2009). These individuals are more likely to contribute to OSS development and platform such as Sourceforge, to support their belief in OSS values and ideology. Therefore, they will be less likely to accept monetary rewards for their effort on OSS projects. We control for this variable in our research model. OSS developers on Sourceforge.net can subscribe to the news updates of their projects and signal their subscription status. We believe that subscription status of OSS developers signals relational commitment (Morgan and Hunt, 1994) with OSS development and the open source platform, in this case. The Subscription variable (measured as 1 if the developer is a subscriber and 0 otherwise) is included to control for its potential influence on relational commitment. We control for differences in the economic status of open source developers by using three variables – Sponsor, OSS Income and NonNA-EUR. First, we control for developers who work on employersponsored projects. Several corporations have adapted their business model to actively sponsor OSS projects. In these cases, salaried employees are assigned to work on OSS projects. These developers are likely to have stable and high incomes and therefore, less likely to voluntarily accept monetary rewards. We control for these professionally employed developers by including the variable
Sponsor which takes the value of one for a developer if s/he works on employer-sponsored projects. Second, open source developers could potentially earn money through their open source activities through multiple means. Some might earn a salary. Others might participate in bounties offered by corporations (Krishnamurthy and Tripathi, 2006), provide hosting services for OSS, develop customized OSS solutions, earn speaker and trainer fees and offer other software services. Those who earn more through open source development are less likely to accept monetary rewards voluntarily. Therefore, we control for this through the OSS Income variable which measures the earning through open source activities over the past five years. We believe that the combination of these three control variables (Sponsor, OSS Income and Non-NA-EUR) sufficiently control for developers’ economic status in this context. Finally, OSS programmers come from a wide variety of geographical regions (Gonzalez-Barahona et al., 2008) around the world with different economic and cultural backgrounds. Developers from economically disadvantaged nations are more likely to accept monetary rewards when made available. In addition, this variable controls for cultural differences. Prior studies have suggested that cultural factors can affect OSS development at both the project-level (Nakakoji et al., 2002) and the individuallevel (Dafoulas and Macaulay, 2001). This might be through programmers with different cultural backgrounds have different commitment levels to the projects (Padmanabha, 2007), which may further affect their decision to accept monetary rewards. Therefore, to capture geographical and cultural differences, we define a dummy Non-NA-EUR. Non-NA-EUR is 0 if the developer is from North America/Europe and 1 otherwise. 4.3. Data collection We focus on OSS developers associated with the top 15percentile active Sourceforge projects for this study. Sourceforge classifies projects based on their activity. The project activity index is computed based on a complex formula that considers number of bugs, patches, commits tasks, etc. related to the project (Healy
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
639
Table 3 Research variables. Variables
Factor loading
Survey items InM: intrinsic motivation InM1 InM2 InM3 InM4
Writing open source software programs is fun (Hars and Ou, 2002) Working on an open source project is fun Code for project is intellectually stimulating to write (Lakhani and Wolf, 2005) It gives me a sense of personal achievement (Roberts et al., 2006)
0.868 0.857 0.651 0.688
IdM: ideological motivation IdM1 IdM2 IdM3 IdM4 IdM5 IdM6 IdM7
I believe that source code should be open (Lakhani and Wolf, 2005) I dislike proprietary software and want to defeat them (Lakhani and Wolf, 2005) Owners of software should not restrict its use Corporations should make the source code of their software open Users of all software should be free to copy it at no charge Users of all software should be free to modify it at no charge Not providing access to source code for software is unethical
0.834 0.675 0.607 0.725 0.747 0.661 0.751
Working on an open source project gives me the chance to attain a recognized qualification (Roberts et al., 2006 modified) Working on an open source project gives me the chance to attain a recognized skill (Roberts et al., 2006 modified) Working on an open source project gives me status at work (Roberts et al., 2006) Working on an open source project enhances my professional status (Lakhani and Wolf, 2005) Working on an open source project increases my opportunities for a better job (Roberts et al., 2006) The programming skills I develop on an open source project will help me get a better job (Lakhani and Wolf, 2005 modified)
0.722
EM: extrinsic motivation EM1 EM2 EM3 EM4 EM5 EM6
CM: community motivation The work culture of open source projects is more fun compared to working on proprietary software CM1 CM2 I feel a greater affinity with other developers on open source projects in comparison to proprietary software projects CM3 Unlike proprietary software, there is a greater sense of cohesiveness in open source project teams
0.833 0.720 0.802 0.893 0.783
0.762 0.778 0.675
INT: intention to accept monetary rewards INT1 I would accept monetary rewards from for-profit corporations INT2 I would accept monetary rewards from non-profit corporations INT3 I would accept monetary rewards from individuals
0.816 0.820 0.909
Moderating variables Sponsor OSS Income
n/a n/a
Is your work on an open source project sponsored by your employer? (Yes/No) In the last five years, what is the total amount of money you have earned as a result of your open source activities? (Logarithm transformed)
Observation data (from Sourceforge.net) Acceptance behavior of monetary rewards Accept Individual’s decision to accept monetary rewards on Sourceforge.net (Yes/no)
1.000
Tenure Tenure
Number of days since a developer registered on Sourceforge
n/a
Subscription Subscribe
Indicates if an individual is a subscriber to Sourceforge.net (Yes/no)
n/a
Region Non-NA-EUR
Measures the region of the developer (0 if in North America/Europe, 1 otherwise)
n/a
and Schussman, 2004). It has been documented that open source projects are asymmetric in their size, prominence and activity levels (Krishnamurthy, 2002) and many projects report low or no activity over an extended period of time (Healy and Schussman, 2004). Since the inclusion of developers associated with low activity project may skew the results, some prior studies have focused on top active projects (Krishnamurthy, 2002; Zhou and Davis, 2005; Krishnamurthy and Tripathi, 2009; Krishnaraj and Srinivasa, 2011). Focusing on active projects allows us to focus on the population of interest, i.e., active OSS developers. The survey instrument was first sent to a few expert OSS developers for their feedback on the structure and readability of the survey and on the inclusion of specific questions. The survey was modified in light of this feedback. The final instrument was sent via email to 4000 active OSS developers, randomly chosen from a pool of active developers who were associated with the top 15 percentile active OSS projects. We received a total of 320 usable responses with a response rate of 8%, which compares favorably with other
surveys of this population. For the subjects who responded to the survey, additional information was obtained by querying their personal profile pages on Sourceforge.net. The descriptive statistics of our dataset is reported in Table 6. On average, our respondent is an OSS developer who has been associated with Sourceforge.net for more than 5 years, is affiliated with 2.66 projects, and serves as the admin of 1.74 projects. Nearly half of our sample comes from the North America/Europe region (45%). The demographic profile of our sample is consistent with that of Sourceforge. 4.4. Construct operationalization We followed recent studies in OSS to develop our survey instrument and adopted items from the literature when available – see Table 3 for details. All items were measured on seven-point Likert Scales. Intrinsic motivation (InM) was a four-item scale and included items from Hars and Ou (2002), Lakhani and Wolf (2005), and Roberts et al. (2006). Community motivation (CM) was a
640
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
Table 4 Reliabilities, AVEs, and inter-construct correlations. Construct
Cronbach ˛
Cons. reliab
AVE
Intrinsic motivation
Ideological motivation
Extrinsic motivation
Intrinsic motivation Ideological motivation Extrinsic motivation Community motivation Intention to accept Acceptance behaviora
0.843 0.877 0.906 0.781 0.885 1.000
0.853 0.880 0.911 0.783 0.886 1.000
0.596 0.515 0.631 0.547 0.722 1.000
0.772 0.159* 0.456*** 0.510*** 0.334*** 0.127*
0.718 0.185** 0.402*** 0.005 −0.034
0.794 0.429*** 0.295*** 0.133*
a * ** ***
Community Intention motivato accept tion
0.740 −0.018 0.027
0.850 0.182**
Acceptance behavior
1.000
One-item constructs. p ≤ 0.05. p ≤ 0.01. p ≤ 0.001.
three-item scale. Ideological motivation (IdM) was a seven-item scale that included items from Lakhani and Wolf (2005). Extrinsic motivation (EM) was a six-item scale that included items from Roberts et al. (2006) and Lakhani and Wolf (2005). We introduce a three-item scale, intention to accept monetary rewards (INT). This is a contribution to the open source literature. INT was operationalized by three survey items, each asking for the subjects’ volitional intent to accept monetary rewards from various sources. These items were adapted from similar measures for the behavioral intention construct in the Theory of Reasoned Action (TRA) literature (Ajzen and Fishbein, 1980). OSS developers on Sourceforge can change their account settings that allow them to accept monetary rewards from the consumers of their code/contributions. We measure developers’ decision to accept by a variable Acceptance Behavior (ACC). ACC is a one-item construct, measured by the subject’s acceptance status on Sourceforge.net. This is a binary variable (yes = 1 and no = 0) that was directly observed from the Sourceforge.net website. 5. Results Before testing the hypotheses, the validity of the measurement model needs to be established first (Kline, 1998). Our study then tested for the overall model fit as well as the individual hypothesized path coefficients. 5.1. Measurement model The measurement model is specified by allowing unmeasured covariance between each pair of latent constructs, which includes intrinsic motivation, community motivation, ideological motivation, extrinsic motivation, intention to accept monetary rewards and acceptance behavior. For the single-item construct, ACC, the error term of the only measurement item is fixed to have a mean of 0 and variance of 0 following the guidelines in Hair et al. (2006). The model was then assessed its overall goodness-of-fit using AMOS 16.0. The standardized factor loading scores are listed in Tables 3 and 4 which report the construct reliability, Cronbach’s ˛, and the average variance extracted (AVE) for each construct. The construct reliability (CR) and AVE are calculated by the following formulas respectively, where i is the loading score for the ith factor: CR =
(˙i )
2
2
(˙i ) + ˙var(εi )
AVE =
˙2i ˙2i + ˙var(εi )
=
=
(˙i )
2
2
(˙i ) + ˙(1 − 2i ) ˙2i ˙2i + ˙(1 − 2i )
We validated our model on convergent validity, discriminant validity and nomological validity. First the convergent validity of
Table 5 Hypotheses summary. Hypothesis
Result
H1: A greater level of extrinsic motivation leads to a greater propensity of an OSS developer’s intention to accept monetary rewards H2: A greater level of intrinsic motivation leads to a higher propensity of an OSS developer’s intention to accept monetary rewards H3: A greater level of ideological motivation leads to lower propensity of an OSS developer’s intention to accept monetary rewards H4: A greater level of community motivation leads to lower propensity of an OSS developer’s intention to accept monetary rewards H5: OSS developers who report high intention to accept monetary rewards are more likely to accept rewards
Supported
Supported
Not Supported
Supported
Supported
the measurement model is well supported: all factor loading scores are greater than 0.60 and statistically significant at p = .001 and all Constructs’ reliability scores and Cronbach’s ˛s are higher than 0.70. Further, the AVE of all the constructs is greater than 0.50, which confirms that all of our constructs capture more constructrelated variance than the error variance. According to Chin (1998), to establish the discriminant validity of the constructs, the square root of a construct’s AVE must be greater than all its inter-construct correlations to ensure the variance shared by the construct and its items are greater than that of another construct. Table 4 lists all the correlation coefficients between the constructs, with the square root of the AVEs on the diagonal. It can be seen that all inter-construct correlations are lower than the numbers on the diagonal. This suggests acceptable discriminant validity between the constructs. We test the nomological validity of the constructs by examining the correlation matrix, which is also listed in Table 4. As can be seen from the table, all the motivation constructs are positively correlated with each other. This is not surprising. OSS motivation factors are highly concomitant with each other (Roberts et al., 2006) and OSS programmers are usually driven by quite a few motivations simultaneously instead of one. Of the four motivations, intrinsic motivations (InM) and extrinsic motivations (EM) correlate with intention to accept monetary rewards (INT), and acceptance behavior (ACC) significantly, while the other two motivations (ideological and community oriented motivations) do not. Finally, intention to accept (INT) and acceptance (ACC) are positively correlated with each other. 5.2. Structural model Having established the validity of the measurement model, we move on to test the structural model. The structural model is specified by replacing the unmeasured covariance between the
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644 Table 6 Descriptive statistics. Variables Subscriber (%) Average tenure (months) Developers accepting monetary rewards (%) Avg. number of active projects developers are associated with Avg. number of active projects developers work as admin Percentage of developers from North America Avg. income of a developer from open source services (over 5 years)
7% 61 18% 2.66 1.74 45% $20,929
independent and dependent variables in the measurement model with hypothesized causal relationships. We then used AMOS to estimate both the overall goodness-of-fit of the model as well as the individual path coefficients. We use multiple fit indices to test the overall fit of our model (Hair et al., 2006). In addition to the absolute fit index 2 and the associated d.f., we also report fit indices from each of the following categories: one incremental fit index (CFI), one parsimony-adjusted fit index (PCFI), two goodness-of-fit indices (NNFI and CFI), and one badness-of-fit index (RMSEA). The results are summarized in Fig. 2. We evaluate these fit indices against the benchmark values in Hair et al. (2006). Our model has 27 observed variables with a sample size of 320. For a model of this complexity, most of the indices suggest a very good overall model fit. In particular, CFI (0.929) is higher than the 0.90 benchmark and the associated RMSEA (0.052) is lower than the accepted 0.07 threshold. Together they suggest very good model fit. This is confirmed by the TLI (or NNFI) value (0.908), which is also above the 0.90 benchmark. However the parsimony fit index, PCFI, suggests only moderate model fit with a value of 0.720. This may be attributed to the relatively large sample size of our study (N = 320) and the complexity of our model (27 observed variables and 5 latent constructs). Noting that there is no general accepted threshold for parsimony-adjusted indices like PCFI (Singh, 2009). Fig. 2 reports the estimates of the hypothesized path coefficients and a summary of our results is shown in Table 5. First, we find that the intention to accept monetary rewards has a positive and significant impact on actual acceptance behavior. Next, we investigate the relationships between various types of motivation and acceptance behavior. Four out of five hypotheses are supported. We find that extrinsic and intrinsic motivations have significant positive effects on the intention to accept monetary rewards. Community motivation (CM) has a significant and negative effect on intention to accept monetary rewards. Ideological motivation (IdM) does not have a significant relationship with the intention to accept monetary rewards. We also find that the intention to accept monetary rewards leads to actual acceptance behavior. Tenure, Subscribe, Sponsor, OSS Income and Non-NA-EUR are the control variables in our study. They are modeled as directly affecting acceptance behavior. We find that Non-NA-EUR and Subscribe are positive and strongly significant. A SourceForge subscriber is more likely to accept monetary rewards than a non-subscriber and someone outside of North America/Europe is more likely to accept monetary rewards than otherwise. OSS Income is negative – i.e., those with high OSS Income are less likely to accept monetary rewards. Tenure and Sponsor are insignificant. 5.3. Discussion of results Our paper contributes to our understanding of an underresearched area – acceptance of financial incentives in OSS. We believe that this is essential to understand how innovation through open source software is to be sustained. In particular, our larger contribution is to surface the importance of acceptance behavior.
641
Our study has begun to uncover the reasons for this behavior. The theoretical implications of this are that the private-collective innovation model might revert to a collective innovation model if acceptance effects are strong. Our first finding is that extrinsic motivation has a significant positive effect on developers’ intention to accept rewards. This should not be a surprising result to open source scholars – although it has never been formally established in this form previously. Extrinsically motivated developers might “simply want to earn a living” (Fitzgerald, 2006) through their open source work. Second, we find that intrinsic motivation has a significant positive effect on the developers’ intention to accept rewards. This will be a counterintuitive result for many who have accepted “crowding out” as the only possible scenario vis-à-vis financial rewards. In particular, as argued earlier, our position is that in some circumstances “crowding in” occurs. This is tied to the particular nature of the financial incentives. Invoking Frey and Jegen (2001) and Benabou and Tirole (2003), we argue that since developers voluntarily chose to accept rewards, they will perceive such incentives to be supportive rather than controlling. Hence, “crowding in” is likely to occur. Besides the theoretical explanation for this empirical finding, our work provides new evidence on the nature of intrinsic motivation in open source software. Thus far, the literature has evidence of no crowding out (Lakhani and Wolf, 2005; Roberts et al., 2006) and the importance of norms in intrinsic motivation (Alexy and Leitner, 2011). Our work adds to this by suggesting that when studying the reaction to financial incentives, “crowding in” can occur with a particular subset of rewards. The implications of this and the boundaries of this finding will continue to be explored as we gain a deeper understanding of intrinsic motivation. Community motivation (CM) has a significant and negative effect on intention to accept monetary rewards. This might be because accepting monetary rewards might be viewed as individual gain rather than benefiting the community. This finding underscores some of the core concerns about the sustainability of the open source movement. If developers regard financial rewards as selfish and decline them, that might lead to an unsustainable financial position in the long run. Future research must explore under what circumstances the open source community will accept financial rewards. They might be open to certain types of rewards, for instance. Ideological motivation (IdM) does not appear to have any direct relationship with intention to accept monetary rewards. The reasons for this are not immediately obvious. It might be that developers are able to psychologically accommodate the seeming contradiction of open source ideology and voluntary rewards. There might also be different segments of developers who approach ideology in different ways. We do not have the data to disentangle these explanations and leave it to future research. Finally, intention, and acceptance are positively correlated with each other. This is consistent with our theoretical argument that OSS developers’ intention to accept monetary rewards affects their actual acceptance behavior. 6. Limitations Our paper contributes to the central discourse in open source around the motivation of developers. We add to previous studies (e.g., Hars and Ou, 2002; Lakhani and von Hippel, 2003) by introducing scales for community and ideological motivations as shown in Table 3. However, every empirical research paper necessarily suffers from some limitations. We describe these limitations and their potential impact on our findings in this section. First, our study is constrained by the source of our data, i.e., Sourceforge.net. With a different source, different behaviors may
642
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
Intrinsic Motivation
0.389***
0.000 0.238***
Ideological Motivation
0.035
0.267***
Sponsor Acceptance of Monetary Rewards OSS_Income -0.021~ 0.191***
Community Motivation
-0.350***
Subscribe
0.076
Intention to 0.170** Accept Monetary Rewards Extrinsic Motivation
Tenure
Non-NA-EUR Control Variables
Independent Variables Fig. 2. Model results: standardized regression weights and significance levels. Model fit indices: N = 320, d.f. = 337, 2 = 627.881, CFI = 0.929, PCFI = 0.720, TLI (NNFI) = 0.908, RMSEA = 0.052. ∼p ≤ 0.1, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.
have been assumed. Also, the observed behavior is a direct result of the specific site design of Sourceforge.net. We have only included developers who were registered Sourceforge users. This may have systematically omitted certain types of open source developers. Second, our study design was based on the top 15% of projects. A few prior studies have pointed out the uneven distribution of activity of Sourceforge projects (Howison and Crowston, 2004), our study was designed to rule out the explanation that non-acceptance of rewards was due to the inactivity of a project. However, focusing on the most active projects may have led to certain biases. For one, this might have led to floor effects with respect to motivation – we might have studied only the more engaged developers. This self-selection might have biased our findings and this must be taken into consideration. Other sampling strategies (e.g., focusing on one category of projects) might have led to a broader set of developers including those who are not as engaged. The problem is that engagement and survey response are also related, making it difficult to ascertain the net impact of this study design. Future empirical studies must consider other sampling strategies to explore the generalizability of our findings. Third, the robustness of our results is enhanced by the inclusion of several control variables – Tenure, Subscribe, Sponsor, OSS Income and Non-NA-EUR. While three of these variables pertain to the economic and cultural aspect of a developer (Sponsor, OSS Income and Non-NA-EUR), we have not explicitly included the income of the developer. Including the income might have distorted the results considerably. The direction of this distortion is a matter of speculation – although, one might reasonably argue that income might have had a dampening effect on all motivational variables. Our dataset is also lacking in several demographic variables – notably, age, gender and educational levels. The impact of such demographic variables on developer behavior has received scant attention in the literature thus far and this must be explored in future research. 7. Conclusion The private-collective innovation model (Von Hippel and Von Krogh, 2003, 2006) argues for a mix of private and collective gains in the open source innovation model. Our work indicates that the
type of incentive matters immensely in this tradeoff, i.e., it is not just the size of the incentive, but how it is structured. Since voluntary monetary rewards are provided by users, such rewards can be “supportive” of the open source innovation system (Frey and Jegen, 2001; Benabou and Tirole, 2003). Those who provide these rewards do not set the innovation agenda for the project, nor do they control changes to the code. Future research must examine developers’ relationship with other types of rewards. The open source literature has suggested that by not providing developers adequate financial rewards, corporations might be exploiting altruistic individuals (Bonaccorsi and Rossi, 2003, 2005; Haruvy et al., 2003). Our findings highlight a greater degree of complexity in this phenomenon than was initially understood. By demonstrating that some developers choose to accept rewards and others do not, we introduce the volition of the individual programmer as an important variable. What appears as “harvesting altruism” (Haruvy et al., 2003), might be the result of a developer voluntarily choosing to not accept the reward for various reasons. The freedom inherent in open source innovation is well documented (e.g., Von Hippel and Von Krogh, 2003, 2006; Lerner and Tirole, 2002; Osterloh and Rota, 2006). Usually, this is discussed in the context of software development, innovation or knowledge sharing. Our work demonstrates that these developers are also fiercely independent about their views of financial rewards. Not accepting rewards could be interpreted as developers exercising their freedom. It must be clarified that our findings do not invalidate the private-collective innovation model. Rather, we see a deepening of our understanding of the place of private gains vs. collective rewards in open source communities. Our paper should motivate a broader dialog about the antecedent factors that drive the acceptance of private vs. collective rewards. From a policy perspective, governments and NGOs seeking to expand open source development must understand that simply providing grant money is not going to solve all problems. Key developers might simply not accept this grant money – especially if they fear that this will lead to some sort of control on the innovation process. Governments and NGOs must engage deeply with open source communities to encourage further innovation.
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
Significant research questions about the role of financial incentives in the OSS development model remain. Are the four motivational components (intrinsic, extrinsic, ideological and community) stable over time? Do developers behave differently with respect to various financial incentives? For instance, are those who accept monetary rewards also likely to participate in a bounty? How do OSS developers perceive those who accept rewards? Do such individuals loose or gain status? Does the magnitude of the financial incentives affect acceptance behavior? These and other research questions should be investigated in future research. References Ajzen, I., Fishbein, M., 1974. Factors influencing intentions and the intentionbehavior relation. Human Relations 27 (1), 1–15. Ajzen, I., Fishbein, M., 1980. Understanding Attitudes and Predicting Social Behavior. Prentice-Hall, Inc., Englewood Cliffs, NJ. Alexy, O., Leitner, M., 2011. A fistful of dollars: are financial rewards a suitable management practice for distributed models of innovation? European Management Review 8 (3), 165–185. Amabile, T.M., Hill, K.G., Hennessey, B.A., Tighe, E.M., 1994. The work preference inventory – assessing intrinsic and extrinsic motivational orientations. Journal of Personality and Social Psychology 66 (5), 950–967. Bagozzi, R.P., Dholakia, U.M., 2006. Open source software user communities: a study of participation in Linux user groups. Management Science 52 (7), 1099–1115. Bagozzi, R.P., Wong, N., Abe, S., Bergami, M., 2000. Cultural and situational contingencies and the theory of reasoned action: application to fast food restaurant consumption. Journal of Consumer Psychology 9 (2), 97–106. Benabou, R., Tirole, J., 2003. Intrinsic and extrinsic motivation. Review of Economic Studies 70 (3), 489–520. Bergquist, M., Ljungberg, J., 2001. The power of gifts: organizing social relationships in open source communities. Information Systems Journal 11 (4), 305–320. Bitzer, J., Schrettl, W., Schroder, P.J.H., 2007. Intrinsic motivation in open source software development. Journal of Comparative Economics 35 (1), 160–169. Bonaccorsi, A., Rossi, C., 2003. Why open source software can succeed. Research Policy 32 (7), 1243–1258. Bonaccorsi, A., Rossi, C., 2004. Altruistic individuals, selfish firms? The structure of motivation in open source software. First Monday 9 (1.). Calder, B.J., Staw, B., 1974. The interaction of intrinsic and extrinsic motivations: Some methodological notes. Journal of Personality and Social Psychology 31, 76–80. Cameron, J., Pierce, W.D., 1994. Reinforcement, reward, and intrinsic motivation: A meta-analysis. Review of Educational Research 64, 363–423. Chakravarty, S., Haruvy, E., Wu, F., 2007. The link between incentives and product performance in open source development: an empirical investigation. Global Business and Economics Review 9 (2/3), 151–169. Chang, M.K., 1998. Predicting unethical behavior: a comparison of the theory of reasoned action and the theory of planned behavior. Journal of Business Ethics 17 (16), 1825–1834. Chin, W.C., 1998. The partial least squares approach for structural equation modeling. In: Marcoulides, G.A. (Ed.), Modern Methods for Business Research. Lawrence Erlbaum Associates, pp. 295–336. Dafoulas, G., Macaulay, L., 2001. Investigating cultural differences in virtual software teams. Electronic Journal on Information Systems in Developing Countries (74), 1–14. Dahlander, L., 2007. In the club: human and social capital of leaders in free and open source software communities. Academy of Management Proceedings (August), 1–6. Deci, E.L., 1975. Intrinsic Motivation. Plenum, New York. Deci, E., Ryan, R., 1985. Intrinsic Motivation and Self-Determination in Human Behavior. Plenum Press, New York. Dermer, J., 1975. The interrelationship of intrinsic and extrinsic motivation. Academy of Management Journal 18 (1), 125–129. Ekeh, P., 1974. Social Exchange Theory: The Two Traditions. Harvard University Press. Fershtman, C., Gandal, N., 2007. Open source software: motivation and restrictive licensing. International Economics and Economic Policy 4, 209–225. Fitzgerald, B., 2006. The transformation of open source software. MIS Quarterly 30 (3), 587–598. Franck, E., Jungwirth, C., 1999. Reconciling rent-seekers and donators – the governance structure of open source. Journal of Management & Governance 7 (4), 401–421. Frey, B.S., Goette, L., 1999. Does Pay Motivate Volunteers. University of Zurich, Institute of Empirical Research in Economics (Working Paper). Frey, B.S., Jegen, R., 2001. Motivation crowding theory. Journal of Economic Surveys 15 (5), 589–611. Ghosh, R.A., 1998. Cooking pot markets: an economic model for the trade in free goods and services on the internet. Brazilian Electronic Journal of Economics 1 (1). Gonzalez-Barahona, J.M., Robles, G., Andradas-Izquierdo, R., Ghosh, R.A., 2008. Geographic origin of Libre software developers. Information Economics and Policy 20 (4), 356–363.
643
Hair, J., Black, B., Babin, B., Anderson, R., Tatham, R., 2006. Multivariate Data Analysis, sixth ed. Prentice-Hall, Upper Saddle River, NJ. Hars, A., Ou, S., 2002. Working for free? Motivations for participating in open-source projects. International Journal of Electronic Commerce 6 (3), 25–39. Haruvy, E., Prasad, A., Sethi, S., 2003. Harvesting altruism in open-source software development. Journal of Optimization Theory and Applications 118 (2), 381–416. Haruvy, E., Prasad, A., Sethi, S., Zhang, R., 2008. competition with open source as a public good. Journal of Industrial and Management Optimization 4 (1), 199–211. Healy, K., Schussman, A., (Working Paper) 2004. The Ecology of Open Source Software Development. http://www.kieranhealy.org/files/drafts/oss-activity.pdf Hertel, G., Niedner, S., Herrmann, S., 2003. Motivation of software developers in Open Source projects: an internet-based survey of contributors to the Linux kernel. Research Policy 32 (7), 1159–1177. Howison, J., Crowston, K., 2004. The perils and pitfalls of mining Sourceforge. In: Proceedings of the International Workshop on Mining Software Repositories, http://msr.uwaterloo.ca/papers/Howison.pdf Jannsen, O., Huang, X., 2008. Us and me: team identification and individual differentiation as complementary drivers of team members’ citizenship and creative behaviors. Journal of Management 34, 69–88. Karahanna, E., Straub, D.W., Chervany, N.L., 1999. Information technology adoption across time: a cross-sectional comparison of pre-adoption and post-adoption beliefs. MIS Quarterly 23 (2), 183–213. Kline, Rex B., 1998. Principles and Practice of Structural Equation Modeling. Guilford Press, New York. Kollock, P., 1999. The economies of online cooperation: gifts and public goods in cyberspace. In: Smith, M., Kollock, P. (Eds.), Communities in Cyberspace. Routledge, London, pp. 220–239. Krishnamurthy, S., 2002. Cave or community? An empirical examination of 100 mature open source projects. First Monday 7 (6). Krishnamurthy, S., 2003. A Managerial Overview of Open Source Software. Business Horizons 46 (5), 47–56. Krishnamurthy, S., 2006. On the intrinsic and extrinsic motivation of open source developers. Knowledge, Technology & Policy 18 (4), 17–39. Krishnamurthy, S., Tripathi, A., 2006. Bounty programs in free/Libre/open source software (floss): an economic analysis. In: Bitzer, J., Schroder, P.J.H. (Eds.), The Economics of Open Source Software Development. Elsevier Publications. Krishnamurthy, S., Tripathi, A., 2009. Monetary donations to an open source software platform. Research Policy 38 (2), 404–414. Krishnaraj, P.M., Srinivasa, K.G., 2011. Analysis of projects and volunteer participation in large scale free and open source software ecosystem. ACM SIGSOFT Software Engineering Notes 36, 1–5. Lakhani, K., von Hippel, E., 2003. How open source software works: ‘Free’ user-touser assistance. Research Policy 32 (7), 923–943. Lakhani, K., Wolf, B., 2005. Why hackers do what they do: understanding motivation and effort in free/open source software projects. In: Feller, J., Fitzgerald, B., Hissam, S., Lakhani, K. (Eds.), Perspectives on Free and Open Source Software. MIT Press. Lee, G.K., Cole, R.E., 2003. From a firm-based to a community-based model of knowledge creation: The case of the Linux Kernel development. Organization Science 14 (6), 633–649. Lerner, J., Tirole, J., 2002. Some simple economics of open source. Journal of Industrial Economics 50 (2), 197–234. Lerner, J., Tirole, J., 2005. The economics of technology sharing: open source and beyond. Journal of Economic Perspectives 19 (2), 99–120. Mauss, M., 1955. The gift. Cohen and West Publishing, London. Morgan, R.M., Hunt, S.D., 1994. The commitment–trust theory of relationship marketing. Journal of Marketing 58 (3), 20–38. Nakakoji, K., Yamamoto, Y., Nishinaka, Y., Kishida, K., Ye, Y., 2002. Evolution patterns of open-source software systems and communities. In: Proceedings of the international Workshop on Principles of Software Evolution, Orlando, FL, May 19–20, IWPSE ‘02. ACM, New York, NY, pp. 76–85. Okoli, C., Wonseok, O., 2007. Investigating recognition-based performance in an open content community: a social capital perspective. Information & Management 44 (3), 240–252. Osterloh, M., Rota, S., 2006. Open source software development—just another case of collective invention? Research Policy 36 (2), 157–171. Padmanabha, R., (Working Paper) 2007. FLOSS (Free/Libre open source software): A Theme for Studying Cultural Differences. http://mpra.ub.uni-muenchen.de/ 4253/1/MPRA paper 4253.pdf Raymond, E., 1998. The cathedral and the bazaar. First Monday 3 (2). Roberts, J., Hann, I.H., Slaughter, S., 2006. Understanding the motivations, participation, and performance of open source software developers: a longitudinal study of the apache projects. Management Science 52 (7), 984–999. Randall, D.M., 1989. Taking stock: can the theory of reasoned action explain unethical conduct? Journal of Business Ethics 8 (11), 873–882. Rossi, M.A., (Working Paper) 2004. Decoding the Free/Open Source (F/OSS) Software Puzzle: A Survey of Theoretical and Empirical Contributions. http://citeseerx.ist. psu.edu/viewdoc/download?doi=10.1.1.112.4800&rep=rep1&type=pdf Rossi, C., Bonaccorsi, A., 2005. Intrinsic vs. extrinsic incentives in profit-oriented firms supplying Open Source products and services. First Monday 10 (5). Ryan, R.M., Deci, E.L., 2000. Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist 55 (1), 68–78.
644
S. Krishnamurthy et al. / Research Policy 43 (2014) 632–644
Shah, S., 2006. Motivation, governance, and the viability of hybrid forms in open source software development. Management Science 52 (7), 1000–1014. Sheppard, B.H., Hartwick, J., Warshaw, P.R., 1988. The theory of reasoned action: a meta-analysis of past research with recommendations for modifications and future research. Journal of Consumer Research 15 (3), 325–343. Shimp, T.A., Kavas, A., 1984. The theory of reasoned action applied to coupon usage. Journal of Consumer Research 11 (December (3)), 795–809. Singh, R., 2009. Does my structural model represent the real phenomenon?: a review of the appropriate use of Structural Equation Modelling (SEM) model fit indices. The Marketing Review 9 (3), pp. 199-212(14). Sorrentino, R.M., Higgins, E.T., 1986. Handbook of Motivation and Cognition: Vol. 1: Foundations of Social Behaviour. John Wiley & Sons Ltd. Stallman, R., 1992. Why Software Should be Free. GNU Operating System http://www.gnu.org/philosophy/shouldbefree.html Stewart, K.J., Gosain, S., 2006. The impact of ideology on effectiveness in open source software development teams. MIS Quarterly 30 (2), 291–314.
Venkatesh, V., Morris, M.G., Davis, G.B., Davis, F.D., 2003. User acceptance of information technology: toward a unified view. MIS Quarterly 27 (September (3)), 425–478. Von Hippel, E., 2005. Democratizing Innovation. The MIT Press, Boston, MA. Von Hippel, E., Von Krogh, G., 2003. Open source software and the “privatecollective” model: issues for organization science. Organization Science 14 (2), 209–223. Von Hippel, E., Von Krogh, G., 2006. Free revealing and the private-collection model for innovation incentives. R&D Management 36 (3), 295–306. Von Krogh, G., Von Hippel, E., 2006. The promise of research on open source software. Management Science 52 (7), 975–983. Wu, C.-G., Gerlach, J.H., Young, C.E., 2007. An empirical analysis of open source software developers’ motivations and continuance intentions. Information and Management 44 (3), 253–262. Zhou, Y., Davis, J., 2005. Open source software reliability model: an empirical approach. In: Fifth Workshop of Open Source Software Engineering (5-WOSSE), St. Louis, MO, USA.