discussion

RESEARCH ARTICLE

DEVELOPER CENTRALITY AND THE IMPACT OF VALUE CONGRUENCE AND INCONGRUENCE ON COMMITMENT

AND CODE CONTRIBUTION ACTIVITY IN OPEN SOURCE SOFTWARE COMMUNITIES1

Likoebe M. Maruping J. Mack Robinson College of Business, Georgia State University,

Atlanta, GA 30303 U.S.A. {lmaruping@gmail.com}

Sherae L. Daniel Carl H. Linder College of Business, University of Cincinnati,

Cincinnati, OH 45221 U.S.A. {sherae.daniel@uc.edu}

Marcelo Cataldo Google, New York, NY 10011 U.S.A. {chelo.cataldo@gmail.com}

Open source software (OSS) communities are dependent on the code contributions of developers who, in many cases, never meet face-to-face and collaborate primarily through technology-enabled means. With their fluid membership, such communities often rely on engaging the commitment of developers to their cause. Given the changing nature of OSS communities, developers face barriers in appreciating appropriate ways of con- tributing to the collaborative effort. Such uncertainty about how to contribute results in OSS communities losing developers as they devote their attention to other, more welcoming, communities. In this research, we draw upon uncertainty reduction theory to argue that developers have two alternative avenues at their disposal to gain certainty about how to contribute: passive and interactive. Leveraging the person–environment fit perspective, we argue that congruence and incongruence in the OSS values of a developer and an OSS com- munity serve as an avenue for passive approaches to gaining certainty, to the degree that appropriate ways of contributing are encoded in these values. Further, leveraging social network theory, we argue that centrality within a community’s communication network constitutes an avenue for interactive approaches for gaining certainty about how to contribute. Using polynomial regression analysis, we analyze survey and archival data from 410 developers in an OSS community. Results suggest that developer centrality moderates the impact of congruence and incongruence in OSS values on commitment. Moreover, commitment fully mediates the impact of OSS value congruence and incongruence on developer contribution activity. We discuss the implications of our findings for research and practice.

1

Keywords: Open source software, P-E fit, uncertainty reduction theory, social networks, centrality, value congruence, value incongruence, commitment, code contribution, polynomial regression analysis

1Sue Brown was the accepting senior editor for this paper. Robert Fuller served as the associate editor.

The appendices for this paper are located in the “Online Supplements” section of MIS Quarterly’s website (https://misq.org).

DOI: 10.25300/MISQ/2019/13928 MIS Quarterly Vol. 43 No. 3, pp. 951-976/September 2019 951

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Introduction

The open source software (OSS) development model has grown in popularity from a fledgling movement to an attrac- tive alternative to closed source development. The attractive- ness of the model has increased with OSS project success stories such as Linux, Eclipse, mySQL, and Open Office (Dahlander and O’Mahony 2011; Kogut and Metiu 2001; Shah 2006). The model entices developers and corporations because it leverages volunteer developer contributions and can produce software that is at least as robust and creative as closed source (Kuan 2002; Paulson et al. 2004). In many ways, the OSS approach has revolutionized the way we manage innovation (Singh et al. 2011; Von Krogh et al. 2012). It is no wonder, then, that there has been a change in the nature of OSS communities due to major software firm and venture capitalist investments (Daniel et al. 2018; Dwoskin 2016). For example, Oracle Corporation partici- pates in more than 700 OSS projects (Aksulu and Wade 2010).

As successful as the OSS approach has been, many OSS projects fail because they fall short in drawing developer con- tributions (Fang and Neufeld 2009; Robles and Gonzalez- Barahona 2006). Developer code contributions take the form of commits—implemented changes or additions to the soft- ware (Singh et al. 2011). Developers use commits to create features, fix bugs, and improve application robustness. With- out developer contributions, OSS applications remain stagnant, fail to keep up with user demands, become costly to maintain, and disappoint developers looking to learn innova- tive software development techniques (Dahlander and O’Mahony 2011). OSS projects face challenges in attracting the commitment of developer contributions, given the wide range of projects from which developers can choose and the lack of contractual obligation tying them to any one project (Robles and Gonzalez-Barahona 2006; Seidel and Stewart 2011).

Developing software is demanding on time and cognitive effort. It behooves OSS community leaders to make the col- laboration process as frictionless as possible, lest developers take their time and talents elsewhere. Because of the dis- tributed and digitally mediated nature of OSS collaboration (Agerfalk and Fitzgerald 2008; Howison and Crowston 2014; Singh et al. 2011; Stewart and Gosain 2006), developers rely on a common understanding of the project contribution pro- cesses. Developers care deeply about how OSS communities orchestrate contributions. Yet, researchers and practitioners understand little about the ramifications of inadequate atten- tion to such matters. OSS community leaders, then, must grapple to understand how to direct developer resources in a

way that can make or break their projects. In particular, how can OSS community leaders maintain developer commitment when there is uncertainty around how developers should make contributions?

To elaborate the theoretical mechanisms linking developer uncertainty to individual attitudes and behaviors, we leverage uncertainty reduction theory (URT) (Berger and Calabrese 1975, Hogg et al. 2010). Our core thesis is that as certainty around the contribution process increases, developer commit- ment rises and, consequently, their contribution to the OSS community grows. Leveraging URT, we posit that this rela- tionship is built on a person’s aversion to unpredictability in the behavior of those around them and the uncertainty created about how they themselves should behave (Berger and Calabrese 1975; Hogg 2000). Given the collaborative nature of OSS communities, we identify two relational mechanisms that can minimize developer uncertainty. We integrate the person–environment fit perspective and social network theory as alternative relational mechanisms to minimize developer uncertainty by providing clarity about how to engage within the community. We draw upon these two theoretical perspec- tives because both are relational in nature and offer comple- mentary perspectives in their treatment of individuals within their environment. The person–environment fit perspective explicitly recognizes an individual’s attitudes and behavior as being shaped by characteristics of the individual relative to the environment in which they engage (Kristof-Brown and Guay 2011; Kristof-Brown et al. 2005). It represents a pas- sive uncertainty reduction mechanism. However, the person– environment fit perspective says little about how embedded individuals are—by way of social connections—within their environments.

Social network theory enables us to further elaborate this influence by considering the individual’s embeddedness within their environment via their social connectedness within the community (Pollack et al. 2012; Vardaman et al. 2015). A developer’s position within the community’s communica- tion network can provide an alternative avenue for managing uncertainty by serving as a pipeline for the flow of, and access to, information about how to engage within the community (Podolny 2001). Compared to a person–environment fit per- spective, an individual’s pattern of accessing information is more interactive. As such, it enables us to conceptualize how different individuals within the same environment might react to similar person–environment relations (i.e., who is more or less likely to anchor their commitment to the community on congruence or incongruence in values). The potential impact of such social connectedness on developer reaction remains unclear. On the one hand, a highly connected developer might experience visceral concerns about OSS value congru-

952 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

ence and incongruence, whereas those with limited connec- tions might not care much given their relative isolation from community members. On the other hand, by virtue of the social connections to the community, a highly connected developer may have sufficient anchoring in the community to render congruence and incongruence less relevant to their experience of participation. Isolated developers might care more because congruence and incongruence affect their ability to make code contributions. Hence, we seek to under- stand whether a developer’s connectedness within a com- munity’s communication network shapes their reaction to congruence and incongruence in OSS values. To this end, we aim to address the following research questions: How do OSS value congruence and incongruence between developers and community affect developers’ commitment and code contri- bution to the community? How salient is the influence of this congruence and incongruence among developers with dif- ferent degrees of connectedness within the community?

Theoretical Background

Commitment in OSS Communities

With well over 100,000 OSS projects in existence and digital repositories (e.g., Github, Sourceforge) providing easy access to projects, developers have a plethora of choices of where to devote their time and effort. Faced with numerous choices and such ease of movement, developers often make one-off OSS community contributions (Pham et al. 2013; Pinto et al. 2016; Zhou and Mockus 2012). However, many devel- opers—who are not project owners—make sustained contri- butions to a particular OSS community (Fang and Neufeld 2009; Qureshi and Fang 2010). Why do developers do this and what can OSS community leaders do to attract and retain them? Research provides insight into why developers volun- tarily contribute to OSS communities in general (e.g., Ke and Zhang 2010; Lakhani and Wolf 2005; Roberts et al. 2006). To gain better understanding of why developers devote them- selves to a particular OSS community, we draw upon the concept of organizational commitment.

Organizational commitment represents a psychological bond that individuals form with an organization that stabilizes their behaviors toward the organization (Mathieu and Zajac 1990; Meyer and Allen 1991). It has been used as a basis for understanding a variety of individual attitudes and behavior toward organizations including job performance (e.g., Meyer et al. 1989), satisfaction (e.g., Dishon-Berkovits and Koslow- ski 2002), and prosocial behavior (e.g., Meyer and Hersco- vitch 2001). We focus on the affective form of organizational

commitment, which emphasizes the emotional attachment that individuals feel toward an organization (Meyer and Allen 1991). In an environment where developers can contribute to numerous OSS communities, emotional attachment holds the greatest potential to retain participant interest. Emotional attachment is particularly important in digital environments where participants generally do not interact with each other face-to-face (Seidel and Stewart 2011; Stewart and Gosain 2006). Bateman et al. (2011) found affective commitment to be the only form of commitment that prompts online com- munity participants to post replies—a behavior that requires more effort than reading posts. Commitment likely plays a similar role in influencing developers to make code com- mits—a particularly effortful form of OSS community contribution.

Three primary reasons point to why organizational commit- ment is well-suited for our research. First, although it has received a significant amount of empirical and theoretical attention within employment contexts, organizational commit- ment traces its roots to efforts to understand why volunteers in nonprofits exhibit differences in their level of dedication (Becker 1960). This speaks directly to our phenomenon of interest with regard to developers volunteering their time and effort to create OSS. Second, a focus on organizational com- mitment provides a locally situated explanation for individual behavior within a particular organizational context, by recognizing both the individual and the organization (Allen and Meyer 1997; Bateman et al. 2011). This is relevant because it can provide insight into the code contribution behaviors (e.g., commits) of developers in OSS communities. Third, as Bateman et al. (2011) point out in the context of online communities, a focus on commitment helps managers explain why such individuals “engage in particular activities in their community” (p. 842), not in online communities in general. This is germane given our theoretical interest in understanding why developers choose to commit to a specific OSS community.

A person does not develop emotional attachments at first sight—or at first encounter, in the OSS case. Before they form an emotional attachment to an OSS community, a developer must experience contributing to that community. In attempting to contribute code, developers can experience frustration with issues that make contributing more difficult. Indeed, recent research suggests that a major reason that developers stop contributing to OSS communities is not because software coding is difficult, but because the process of working in that environment is cumbersome (Steinmacher et al. 2015). Next, we elaborate this barrier to contributing to OSS communities and explain why OSS values and commu- nication networks can serve as mechanisms for reducing uncertainty around the process of working together.

MIS Quarterly Vol. 43 No. 3/September 2019 953

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Uncertainty about Participation Processes in OSS Communities

Steinmacher et al. (2015) identify several social barriers to OSS community participation. They include a lack of com- munity responsiveness to developer communications and cultural differences. With regard to a lack of community responsiveness, they found that many developers who turned away did so after not receiving answers to their posts to the community or after continuously receiving delayed responses that made it difficult to complete tasks in a timely manner. A lack of response can result in duplicated code and effort which can frustrate developers. Other studies found that com- munity responsiveness drives developers’ continued participa- tion (Zhang et al. 2013; Zhou and Mockus 2015). Some developers further reported receiving rude responses as a major factor. In terms of cultural differences, developers described having difficulty dealing with differences in community members’ communication styles. A Brazilian developer interpreted a German developer’s directness as rude (Steinmacher et al. 2015) and Daniel et al. (2013) found that cultural diversity decreased community engagement.

From a technical standpoint, the main barriers identified by Steinmacher et al. (2015) include trouble with orienting them- selves within the community and inadequate documentation. Developers noted difficulty knowing on which tasks to begin work, finding the appropriate artifacts to use in fixing an issue, knowing how to contribute, setting up the local environment, and understanding the contribution flow. In addition, inadequate documentation made it difficult for developers to understand the architecture of the software, how to set up the workspace, and how to write code documen- tation. Notably, one developer lamented “I think I will have to look at all the documentation. I have no idea of where the code that I need to change is” (Steinmacher et al. 2016, p. 6). These barriers made it difficult for developers to engage in the process of contributing code (Steinmacher et al. 2016).

As this discussion suggests, developers face challenges in the process of contributing to OSS communities. These barriers create developer uncertainty about how to use their time and skills to interact with community members and to make code contributions. Uncertainty can erode developer attachment to an OSS community. Next, we discuss URT as a lens for understanding why developers are less likely to commit to OSS communities lacking mechanisms that offer guidance about how to contribute. We then discuss the importance of OSS values and communication network position as just such mechanisms.

Uncertainty Reduction Theory

Uncertainty emerges when a person cannot predict the behaviors of others or the outcomes of his own actions (Berger and Calabrese 1975; Hogg et al. 2010; Hogg and Terry 2000). It also emerges when experiences do not match expectations and when relationships change (Kramer 1993). Uncertainty often leads to anxiety and stress (Hogg 2000; Miller and Monge 1985; Schweiger and Denisi 1991) and within organizational settings, uncertain situations—such as secrecy during mergers and job uncertainty—are associated with lower job satisfaction and commitment (Ashford et al. 1989; Bordia et al. 2004). Consequently, people hold a core motivation to reduce uncertainty (Hogg and Terry 2000; Schweiger and Denisi 1991).

People generally employ three approaches to reduce uncer- tainty: passive, active, and interactive (Berger and Calabrese 1975; Ramirez et al. 2002). Passive approaches involve information acquisition about how others act and what they expect through inconspicuous means such as behavior obser- vation (Ramirez et al. 2002). Behavior observation is one of the main approaches that people use to gather information (Berger and Calabrese 1975; Knobloch 2015) and is the primary form of information seeking on discussion lists (Non- necke and Preece 2000) and online communities (Smith 1993). Digital traces of contribution activity enable OSS developers to passively infer OSS community member values through observation. Active approaches involve searching for information about the target from third parties.

Finally, interactive strategies encompass information acquisi- tion through direct interaction with the target. This can be accomplished through posing questions to the target. Berger and Calabrese (1975) observed that people communicated as a way to reduce uncertainty about strangers and subsequent research found that communication patterns often mirror the degree of uncertainty in the task environment in technical work (Tushman 1979, 1979). OSS developers can communi- cate as a way to reduce uncertainty, reducing the friction involved in engaging with the OSS community, and enabling them to focus on the more enjoyable work of writing code to solve technical problems (Weinberg 1998). Next, we high- light OSS values and communication networks as two mechanisms for managing OSS developer uncertainty.

OSS Values and Communication Networks as Relational Mechanisms

During the early stages of the OSS movement, the OSS ideology emerged as a way to codify an approach to devel- oping software that departed from the predominant closed

954 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

source approach (Ljungberg 2000; Raymond 2001; Stewart and Gosain 2006). The OSS ideology included a set of values. Broadly speaking, values represent a preference for a set of behaviors or outcomes (Trice and Beyer 1993). Values are a vital component of one’s self-concept (Lydon 1996) and include a relatively enduring conviction that a specific mode of conduct or end state is preferable to its opposite (Chatman 1991). OSS values emphasize sharing information, helping each other in development efforts, and voluntarily cooperating with others as preferable modes of conduct and underscore the pursuit of technical knowledge, continuously learning new skills, and the development of reputation through participation as desirable end states (Raymond 2001; Stallman 1992; Stewart and Gosain 2006). OSS values codify the expectations around how developers conduct themselves when contributing to an OSS community. In this way, OSS values facilitate a passive mechanism through which developers can minimize uncertainty.

Although the emphasis has not been relational in nature, prior research has focused on the impact of OSS values. One body of work focuses exclusively on the effect of OSS developer values on attitudes and contributions (Benbya and Belbaly 2010; Henkel 2008; Xu et al. 2009), while another solely considers the influence of community OSS values on devel- opers’ attitudes and contributions (Chou and He 2011; Stewart and Gosain 2006). In one of the earliest studies to consider both developer and OSS communities, Ke and Zhang (2009) conceptualize ideological conviction as the extent to which a developer identifies with the OSS values. They argue that when developers feel that their values align with those of the OSS movement, they are likely to feel energized in their community contributions. Similarly, Ke and Zhang (2010) conceptualize integrated motivation as a developer’s convic- tion with the OSS movement ideological values. They argue that the greater the conviction developers feel, the greater their task effort. Both of these studies primarily focus on the motivational role of OSS values. A summary of the research on OSS values and its impact is in Appendix A.

Research on a variety of phenomena, including values, has shown the theoretical insights that we can gain by isolating the effects of person characteristics (i.e., developer), envi- ronment characteristics (i.e., OSS community), the similarities between them, and the differences when one exceeds the other (Kristof-Brown et al. 2005). We follow this work and lever- age the person–environment fit perspective because it provides an explicit treatment of individuals within their environments (Kristof-Brown et al. 2005). Person–environ- ment fit represents the compatibility between a person and their environment on the basis of some characteristic (Kristof- Brown and Guay 2011; Kristof-Brown et al. 2005). Con- sistent with this perspective, we focus on the extent to which

a developer’s OSS values match his perception of those of the OSS community. Congruence is the amount of similarity between the characteristics of an individual and the charac- teristics of the environment in which that individual operates (Hoffman and Woehr 2006). We also examine incongruence, which is the extent to which a developer’s OSS values are discrepant from the perception of those of the OSS community in which he is embedded.

As noted earlier, in some communities, the opportunity to leverage passive mechanisms may not necessarily be obvious or present. Yet, in such conditions, we still see developers willing to commit to contributing to the community. Com- munication networks offer an opportunity to enact interactive approaches for attaining understanding of how to engage within the community. By considering the communication network, we address a limitation of the person–environment fit perspective. While person–environment fit explicitly accounts for person–environment relations with regard to characteristics, it does not account for the individual’s connectedness within that environment. This is theoretically significant because social connections represent a vital part of organizational and community life (Vardaman et al. 2015), particularly in digitally enabled settings (Zhang et al. 2013). Reviews of person–environment fit have recognized this limitation and called for research to incorporate considera- tions of individuals’ local context within their environment (Kristof-Brown and Guay 2011; Kristof-Brown et al. 2005). Consequently, we leverage social network theory to under- stand the local context of developers in relation to other community members by examining the degree to which a developer’s centrality in the communication network shapes the effects of value congruence and incongruence on commitment.

Consistent with social network theory (Freeman 1978), we believe that a developer’s position within the communication network holds important implications for their ability to access information about how to contribute. Broadly, social network theory posits that one’s position within a network of relationships facilitates access to resources (Freeman 1978). Accessing these resources can facilitate performance or improve other career outcomes (Ahuja et al. 2003). Some of the most commonly studied structural properties of a person’s position within a social network include centrality (Ahuja et al. 2003; Freeman 1978), closure (Coleman 1988), brokerage (Freeman 1978), and structural holes (Burt 2004). In this research, we focus on degree centrality. Degree centrality is the number of people to whom a person is connected. In a communication network, that is the number of people with whom they have directly communicated. We focus on degree centrality for several reasons. First, communication network centrality reflects a developer’s linkages with other OSS com-

MIS Quarterly Vol. 43 No. 3/September 2019 955

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

munity members (Sarker et al. 2011) and echoes the notion of communication networks as pipes through which information flows (Podolny 2001). This fits well with our interest in com- munication networks as an avenue for developers to enact interactive information seeking approaches to understand how to contribute within an OSS community. Further, degree centrality, compared to betweenness or closeness centrality, is appropriate when the context is related to communication (Freeman and Spyridakis 2004). In contrast, other social net- work structural features such as brokerage and bridging of structural holes speak to power and advantage attained by having privileged access to unevenly distributed information or connecting people who might otherwise not be connected (Freeman 1978).

Second, centrality reflects the extent to which a person sits at the nexus of information flow within the communication net- work (Freeman 1978). As such, it provides an ideal lens for characterizing a developer’s ability to understand contribution behavior through the communication that occurs between many participants. Centrality within a communication net- work can enable a developer to learn about rapidly changing technical factors including awareness of other developers’ work, their style of working, as well as their skills (Dabbish et al. 2012). Finally, centrality is effectively used when com- munication structure is less strongly dictated by formal structures compared to traditional organizations (Rice and Aydin 1991).

Taken together, we believe that OSS value congruence and incongruence provide opportunities to enact passive informa- tion seeking, and that communication network centrality provides opportunities for interactive information seeking. These serve as essential relational strategies OSS developers can use to gain certainty about how to make code contribu- tions and such information is likely to engender greater com- mitment to the community. Next, we introduce and develop the logic behind our research model. We begin by arguing for the importance of considering the developer’s OSS values within the context of the OSS community within which they are embedded. We then consider the implications of OSS values for commitment among developers who are not cen- trally located within the communication network versus those who hold more central positions. We ultimately link the influence of these factors on commitment to the developer’s code contributions.

Hypothesis Development

Figure 1 shows the proposed model. Prior empirical research supports the relationships indicated by the dotted lines con- necting centrality to code contribution activity (Ehrlich and

Cataldo 2012; Hinds and McGrath 2006; Sparrowe et al. 2001) and to commitment (Lee and Kim 2011; Siciliano and Thompson 2015) and are beyond the scope of this research. We focus on centrality as a moderator in the model for several reasons. First, psychological and network variables jointly impact outcomes (Vardaman et al. 2015) and centrality, in particular, can alter the way negative attitudes affect out- comes (Pollack et al. 2012; Vardaman et al. 2015). For example, Vardaman et al. (2015) showed that an employee’s centrality in the workplace social network weakened the relationship between their turnover intentions and actual turnover and Pollack et al. (2012) provided evidence that centrality in a social network attenuated the positive relation- ship between economic stress and depressed affect.2 Consis- tent with these and other prior works that treat social network position as a moderator, our interest rests in advancing understanding of how developer centrality moderates the influence of congruence and incongruence in person and environment OSS values. Second, person–environment fit and social network theory represent complementary views about how individuals relate to their environment. As such, we believe that an individual’s social connectedness to their environment acts as an important boundary condition that shapes their reactions to congruence and incongruence in OSS values between themselves and the environment. Next we discuss how OSS value congruence impacts commitment and then we describe how centrality in the communication net- work shapes the relationship.

Influence of OSS Value Congruence

As noted earlier, the OSS movement is defined by a sense of community and common purpose (Ljungberg 2000). In fact, Bagozzi and Dholakia (2006) suggest that OSS communities tend to be defined by a sense of “we-”ness. This sense of community, fueled by congruence between the developer and community values (i.e., seeing helping, sharing, cooperation, and learning as important), evokes a sense of certainty about how to collaborate and interact with others in the community. When a developer’s preferred way of working in the OSS environment matches that of the community, the developer feels he can predict how others will behave. This allows the

2Although we have not come across any empirical studies of this, we acknowledge the possibility that commitment can influence developer cen- trality (i.e., more committed developers may take actions that increase their centrality within the community). However, there are many circumstances in the OSS context in which the two may be unrelated in this way. For instance, some developers may connect with a particular developer because she possesses a valued technical skill or because she is employed by a popular company (e.g., Google). This developer may not necessarily have formed an affective commitment to the OSS community. In any case, we empirically account for potential endogeneity between commitment and centrality in our robustness analysis.

956 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Figure 1. Research Model

developer to focus his energy on the joy of making technical contributions (Ke and Zhang 2010).

Although an overarching set of OSS values that define the open source movement exists, not all OSS communities adhere to those values. Indeed, Ljungberg (2000) and others (e.g., Stewart and Gosain 2006) note that some communities include zealots who see these OSS values as a way of life while other communities view the OSS model as simply a means to an end—a way of developing software. Some of the differences in the extent to which communities adhere to OSS values may be attributable to the participation of organiza- tions with commercial interests (Daniel et al. 2018; Spaeth et al. 2015; Stewart et al. 2006). In much the same way, indi- vidual developers differ in the extent to which they embrace the OSS values. Some developers participate in OSS commu- nities for purely instrumental reasons (e.g., they need the software that is under construction), some are paid for their contributions, and others participate because they believe in the OSS movement (Fang and Neufeld 2009). These moti- vations for joining may correlate with certain values in that a developer who participates for ideological reasons might value cooperation, sharing, and learning (Ke and Zhang 2009, 2010).

Value congruence reflects a passive information seeking opportunity in that contribution actions taken by a developer may be reinforced by the community, giving the developer some sense that the behavior is acceptable. Observation of such positive feedback (and even just a lack of negative feed- back) is consistent with passive information seeking approaches when there is uncertainty (Gibbs et al. 2015, Ramirez et al. 2002). By reinforcing the developer’s pre- ferred approach to doing things (i.e., the developer’s preferred approach to contributing is also the community’s preferred

approach), the community provides clarity to the developer about how to engage. Developers are more likely to form an emotional attachment to an OSS community that shares a similar value system to their own (i.e., greater commitment). Ke and Zhang (2010) argue that developers who identify with the values of an OSS project are likely to be more energized in their work on the project. Ke and Zhang (2009) note that OSS developers who feel their values are embraced in the community find the work more rewarding and worthwhile, suggesting commitment should be highest among developers who embrace OSS values and feel that the OSS community adheres to the same values.

OSS value congruence also suggests that developers and their community share aspects of cognitive processing, fostering comparable methods of classifying and interpreting events. Successful collaborative activities require these qualities because they reduce uncertainty (Fisher and Gitelson 1983; Schein 1985). OSS value congruence assists developers in knowing what to expect and in accurately predicting others’ behavior (Kluckhohn 1951). Greater certainty about the contribution process enables developers to enjoy community participation, increasing commitment.

OSS Value Congruence and Developer Centrality in the Communication Network

As the preceding discussion suggests, we anticipate that OSS value congruence positively influences developer commit- ment to the community. We expect this relationship to be more salient among developers who are less central in the community’s communication network. Connections in the communication network represent the pipelines through which information about ongoing community work flows

MIS Quarterly Vol. 43 No. 3/September 2019 957

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

(Zhang et al. 2013). Developers with limited connections in the communication network have few channels for assessing the appropriate methods and procedures for contributing to the community (Steinmacher et al. 2015; Steinmacher et al. 2016). Their opportunities for enacting interactive ap- proaches is limited. OSS value congruence offers a passive approach to attain validation and reassurance for their pre- ferred approach. Lower centrality in the communication network provides less access to information that might pro- vide validation, making OSS value congruence particularly salient in shaping less central developers’ commitment to the community. It serves as the main channel to make them feel that they fit in.

In contrast, developers who hold central positions in the com- munication network rely less on OSS value congruence in shaping their commitment to the community. Two main reasons underlie this belief. First, connections in the com- munity’s communication network offer an awareness of the community’s practices (Ahuja et al. 2003, Zhang et al. 2013). Central developers in the communication network sit at the nexus of the community’s information flow. Through mailing lists, developers communicate about topics including design and implementation decisions, developer roles on the project, decision-making processes, and coordination (Jensen et al. 2011; Kidane and Gloor 2007). Through such communication networks, developers typically receive contextualized infor- mation that they can apply to their specific needs. This provides an alternative, interactive channel for them to under- stand the community contribution practices. Interactive information seeking through direct communication is more efficient and yields more accurate feedback in reducing uncertainty compared to passive approaches (Antheunis et al. 2010). Second, centrality in social networks leaves indi- viduals more embedded in their social contexts and less likely to voluntarily leave for fear of forfeiting their social capital (Borgatti and Halgin 2011; Lee and Kim 2011; Lee et al. 2004). These social connections endear central developers to a community. Consequently, OSS value congruence, while important in influencing developers’ commitment to the com- munity, is likely to play a lesser role among central developers in the communication network.

H1: OSS value congruence will have a stronger positive relationship with commitment among less central developers than among highly central developers.

Influence of OSS Value Incongruence

In the same way that we expect low centrality to enhance the impact of OSS value congruence, we anticipate low centrality

will heighten the influence of two forms of OSS value incon- gruence. In one form of OSS value incongruence, a developer perceives that the OSS community adheres to the OSS values more than he does. That is, the developer may not be in the community because of the values of knowledge sharing, learning, helping, cooperation, and reputation; but never- theless, the community values these things. Incongruence also emerges when a developer embraces the OSS values more than the OSS community does. For instance, the devel- oper may participate in a community in which knowledge sharing, learning, helping, cooperation, and reputation are not valued as highly as he values them. While incongruence typically negatively impacts attitudinal outcomes, recent empirical research suggests that the two scenarios can lead to distinct impacts on such outcomes (Edwards 1996; Edwards and Rothbard 1999; Jansen and Kristof-Brown 2005). We expect the two forms of incongruence to yield distinct impacts in the OSS context. We first consider incongruence when the OSS community embraces OSS values more than the OSS developer does and then we consider incongruence when the developer embraces the OSS values more than the community does. In both cases we theorize how the influence of incon- gruence on commitment is more potent for developers who are less central in the communication network.

Developer Centrality When the Community Embraces OSS Values to a Greater Degree than the Developer

Many developers expect that OSS communities embrace OSS values, even if they themselves do not. When the community embraces OSS values, it affords developers predictability in terms of what to expect from other community participants and in terms of what the community expects of them. Ac- cording to URT, such predictability makes for a desirable environment as it reduces anxiety and stress (Hogg et al. 2010; Hogg and Terry 2000). Given that the environment is desirable in its predictability, we expect developers who perceive the OSS community to adhere to OSS values more than they themselves do to be committed to the community. Edwards and Rothbard’s (1999) idea of carryover reinforces this expectation. They argue that an individual is likely to exhibit prosocial behaviors toward the organization when he feels the environment provides more of something (in this case, adherence to the OSS values) than he needs. Consistent with this idea, such environments promote greater commit- ment (Ashford et al. 1989).

The positive impact of incongruence is consequential for a developer with low centrality in the communication network because he may have few other avenues for more actively ascertaining acceptable behavior (Steinmacher et al. 2015).

958 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

For these developers, the practices that underlie OSS values can substitute for the collaboration information to which they are not privy in the communication network. In digitally enabled environments, such as OSS communities, developers can observe the behaviors dictated by community OSS values. Observation represents one of the few avenues for gathering information about appropriate ways of engaging for those who are not otherwise well-connected within the community (Lindberg et al. 2016). The OSS values manifest through digital traces of community member activities as they go about interacting with other community members and making code contributions. Digital repositories provide the requisite degree of permanence to these behaviors that allow devel- opers to observe engagements over time. Indeed, many developers spend a month or so observing how people con- tribute and behave in a community before making their first contribution (von Krogh et al. 2003). For such developers, commitment forms after they observe and understand the behaviors embedded in the values of the community. This process corresponds with a passive information seeking process to reduce uncertainty. Communities that embrace OSS values enact the underlying practices (i.e., they help, cooperate, and share knowledge), making them responsive to developer needs (Zhang et al. 2013). Such welcoming ges- tures should instill greater commitment on the developer’s part, even if he does not embrace the same values.

In contrast to developers with low centrality in the com- munication network, developers with high centrality can use their connections to attain understanding and predictability with regard to the community’s way of collaborating. Since they can use communication to access information and there- by lower uncertainty (Brashers 2007), a developer with high centrality in the communication network can get personalized responses to her queries about the process. Communication patterns help determine a person’s ability to process informa- tion. As task complexity increases, the communication patterns needed to execute the task also change (Tushman 1979). Because of the complex nature of software develop- ment and the distribution of knowledge across developers, access to multiple developers helps a developer feel certain about how to do her work. For instance, numerous analytic solutions to programming problems exist, requiring the developer to select the best one. A developer could find information about the skills of co-developers to be useful in this decision-making process. This makes highly central developers less reliant on the OSS values in shaping their understanding of the contribution process and associated commitment to the community. Therefore, the impact of incongruence, where the OSS community embraces OSS values more than the developer does, is likely to be quite muted at best. The preceding arguments imply that when the OSS community adheres to OSS values more than the

developer does, the positive effect on commitment will be stronger among less central developers than among highly central developers.

H2a: When OSS community OSS values exceed developer OSS values, the effect of value incongruence on commitment will be more positive among less central developers than among highly central developers.

Developer Centrality When the Developer Embraces OSS Values to a Greater Degree than the Community

Next we consider the second case of value incongruence when a developer embraces OSS values more than the com- munity does. We expect the impact of such incongruence on commitment to be stronger among developers who are less central in the communication network. In such conditions, a developer is likely to experience uncertainty. Instead of having his approach to developing software reinforced—as would be the case if the values were congruent—the OSS community’s differing approach may cause him to question his approach or feel tension related to the discrepancy. Further, unlike the case where the OSS community strongly embraces the OSS values, in this case the developer cannot determine the community’s processes, placing the developer in an uncertain situation. Edwards and Rothbard (1999) would describe this scenario as one of the environment not providing enough of a desired characteristic (in this case, OSS values) for the developer and would expect antisocial attitudes to result. In line with this logic, Elliott and Scacchi (2003) document two cases where an OSS community participant experiences discomfort because he senses the rest of the com- munity does not sufficiently embrace OSS values. In the first case, the developer speaks out against the method (non-free Adobe Photoshop software) used to produce a graphical representation of the system architecture on the GNU website. Over the next couple of days, the argument about this practice continued and tension between the developer and the com- munity rose. In the second case, a developer expressed his disagreement with the use of non-free tools to develop GNU documentation. The discussion lasted three days. This kind of tension could lead to frustration and lower commitment.

As the community battles about what is, or is not “OSS enough,” developers lose time writing code. In these ex- amples, the OSS community is not operating in a way that is consistent with the OSS values, leaving developers unsure about what to expect and what is acceptable. Steers (1977) found that individuals who perceived their organizations to be undependable in carrying out their commitments to employees

MIS Quarterly Vol. 43 No. 3/September 2019 959

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

were, in turn, less committed to their organizations. URT suggests that people communicate when they lack certainty (Berger and Calabrese 1975). Unfortunately, limited social ties mean that less central developers in the communication network may not have interactive information seeking as an option, creating further discomfort. Such a situation can make the community appear less welcoming to developers with low centrality, especially because they already have limited social ties to the community—making them feel isolated (Vardaman et al. 2015; Williams 2007). Their low embeddedness in the community, coupled with unpredictability regarding accept- able practices, erodes commitment.

In contrast, for highly central developers in the community’s communication network, we expect this form of value incon- gruence to minimally impact commitment. First, through their communication network connections, such developers can resolve uncertainty that arises from incongruence. Reponses from social connections can offer support for any ramifications and clarify acceptable practices, enabling them to easily make contributions (Ehrlich and Cataldo 2012; Steinmacher et al. 2015; Wagstrom 2009). Second, highly central developers are embedded within the community, affording them significant social capital within the com- munity. Vardaman et al. (2015) argue that highly central individuals are generally reluctant to forgo the social capital they have built even when the environment pushes them to consider leaving. In much the same way, we expect that highly central developers in the communication network are unlikely to relinquish their social capital on account of discrepancies in preferred community practices. Conse- quently, OSS value incongruence is less likely to impact the commitment of highly central developers. Taken together, the preceding arguments point to the fact that, in situations where the developer embraces OSS values to a greater degree than the OSS community does, OSS value incongruence should have a stronger negative influence on commitment among less central developers than among highly central developers.

H2b: When developer OSS values exceed OSS community OSS values, the effect of value incongruence on commitment will be more negative among less central developers than among highly central developers.

Mediating Role of Developer Commitment to the OSS Community

A developer’s commitment to the community is expected to mediate the relationship between OSS value congruence and incongruence and code contributions. URT indicates that in relational situations, reduction in uncertainty fosters greater

psychological commitment to the relationship in a way that guides future behavior (Knobloch 2015; Knobloch et al. 2010). This suggests that commitment plays an important role in connecting uncertainty reduction activities, such as infor- mation seeking (Knobloch 2015), to subsequent relational behaviors, such as conforming to behavioral norms. Per URT, people are naturally drawn to predictability in behavior as it provides greater certainty about how to act within the relationship (Berger and Calabrese 1975). As elaborated in H1 and H2, we expect OSS value congruence and incon- gruence to influence affective commitment (i.e., the emotional bond that developers feel toward the community). We expect commitment, in turn, to promote greater code contribution. That is, commitment facilitates developers’ actualization of the affinity they form for the community based on value congruence or incongruence. Recent research on online com- munities identifies commitment as a chief antecedent of ongoing contribution behavior (Bateman et al. 2011). Similarly, Ke and Zhang (2009) find that goal commitment in OSS communities is positively associated with developers’ self-reported task performance. Drawing on these findings, we expect that value congruence and incongruence influence code contributions through the commitment that developers form because they care about the community’s viability and want to ensure its success.

H3: Commitment will mediate the relationship between OSS value congruence and incongru- ence and developer code contribution activity.

Building on the mediating logic outlined in H3, we expect the mediating role of commitment to be stronger among devel- opers who are less central in the community’s communication network and weaker among highly central developers in the community’s communication network. Two main considera- tions provide a foundation for the rationale explaining the differences in the strength of the mediating role of com- mitment. First, as discussed earlier, interactive and passive relational uncertainty reduction mechanisms are two alter- natives that can be leveraged in digitally enabled environ- ments (Ramirez et al. 2002). Observation of community members’ activities reflects a passive approach and devel- opers can leverage the interactive approach to information seeking through their communication network connections. Highly central developers can use both mechanisms. Less central developers primarily have passive approaches at their disposal, making OSS value congruence and incongruence a main driver of commitment and concomitant code contribution.

Second, given the choice between the two uncertainty reduction mechanisms, interactive approaches—such as that represented in the communication network—are likely to be

960 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

preferred by developers who are highly central in the communication network. Developers prefer an interactive approach, through the communication network, because they can receive clear, direct, and immediate answers to their questions rather than having to observe and infer acceptable forms of engaging within the community (Ramirez et al. 2002). Their high centrality within the community’s com- munication network affords them such access to information. Indeed, Antheunis et al. (2010) find that interactive ap- proaches are the most effective in reducing uncertainty. Consequently, as noted in H1 and H2, highly central developers rely less on OSS values to guide their formation of commitment to the community and their code contributions are driven less by the mediated influence of value congruence and incongruence. As such, commitment should play less of a mediating role in linking OSS value congruence and incon- gruence to code contributions among such developers.

H4: Commitment will have a stronger mediating effect in the relationship between OSS value congruence and incongruence and developer code contribution activity among less central developers than among highly central developers.

Methodology

Research Setting

We tested the hypotheses using data collected from the GNOME OSS community. Formed in 1997, the GNOME community creates and supports a graphical user interface platform for Unix operating systems such as Linux. Over the decades that have followed, volunteers and paid contributors from around the world have contributed in order to create a freely available desktop platform and many supporting applications. Red Hat, Cisco, Novell, and numerous other companies make contributions to GNOME. GNOME project members utilize two tools to enable their work. Mailing lists represent the main communication mechanism among project members (e.g., developers) as well as between project members and community members (e.g., developers in other projects). Source code repositories enable developers to make changes to the system and also allow them access to the history of the development activities.

Sample and Procedure

We used a combination of archival and survey data to test our model. The GNOME source code repositories allowed us to identify contributors across the various projects in GNOME

as well as to extract developer contribution activity, developer degree centrality and control variables used in our analyses. At the time of data collection, the GNOME community had 734 different projects. Since the requirements to create a project in the GNOME source code repository system remain relatively easy to meet, projects tend to differ significantly in their development activity, size, and participation rate. Building on criteria used in past research (Crowston et al. 2003), we only considered active projects that satisfied the following criteria: (1) continuity of development activity (at least one year), (2) amount of development activity (at least 100 commits),3 (3) attractiveness of project for developers (at least 10 committers), and (4) user interest to participate (at least one community hosted mailing list). Use of projects that satisfied these criteria provided more complete information about email communication networks and developer con- tribution activity over time. A total of 91 projects satisfied these criteria. These 91 projects are part of the official GNOME distribution.

A total of 2,576 developers contributed at least one commit to the development effort from 1998 until December of 2010. In all, 2,341 of them made contributions to the 91 projects in our sampling frame. However, 553 of these developers did not make any contributions beyond 2005 and were not considered in our analyses. The remaining 1,788 developers contributed 91.6% of the commits (595,327 out of 649,526) to those 91 projects throughout the period covered by our data. Further- more, those 1,788 developers contributed 90.2% (596,320 out of 661,109) of all development activity across all 734 pro- jects. Therefore, we consider the 1,788 developers to be a representative subgroup of the entire population of developers in the GNOME community.

We invited the 1,788 contributors to complete our survey by sending them an email. One week later a follow-up reminder went out and three weeks after that a final reminder was sent. As an incentive, each participant who completed the survey earned entry into a drawing for a monetary prize. Two parti- cipants, randomly selected through the drawing, received $200.00 each. A total of 562 individuals responded to the survey (31% response rate) and 410 provided usable responses (filled in all survey questions used in analysis). The 410 respondents represented 55 countries distributed across North America (69), Europe (248), South America (20), Asia (45), Africa (4), and Oceania (13). Eleven parti- cipants did not indicate their location. The sample of respon- dents included 93.7% males. Mean age for the respondents was 29.7 years, with a standard deviation of 8.3 years. Our

3Source code repositories manage the software development process by keeping track of changes: what was changed, when it was changed, who made the change. A commit occurs when a developer uploads altered code.

MIS Quarterly Vol. 43 No. 3/September 2019 961

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

analysis showed that our sample of 410 participants did not differ significantly in their patterns of contribution activity from the other 152 individuals who did not provide complete responses to the survey in terms of the number of projects in which the individuals participated (z = -0.51, p = 0.61), the number of commits made (z = -0.40, p = 0.69), the number of lines of software code contributed (z = 1.70, p = 0.09), the number of messages contributed to the mailing lists (z = -1.41, p = 0.16) and the number of messages posted on the defect tracking system of the GNOME community (z = -1.30, p = 0.19). Additionally, analysis showed that our sample of 410 participants did not differ significantly from the 1,226 developers who did not participate in the survey in terms of the number of projects in which the individuals participated (z = -0.05, p = 0.96), number of commits made (z = -1.10, p = 0.27), number of lines of software code contributed (z = -0.95, p = 0.36), and number of messages contributed to the mailing lists (z = -0.78, p = 0.44). The two groups differed in terms of the number of messages posted on the defect tracking system of the GNOME community (z = -4.46, p < 0.00).4

Measurement

Dependent Variable

Developer contribution activity (number of commits). We counted the number of commits on all GNOME projects over the 9 months following the survey administration (between January 2011 and September 2011) to measure each devel- oper’s contribution activity. Since the main task of OSS development communities involves creating and extending software, we use the total number of concurrent versions system (CVS) commits. A CVS commit occurs when a con- tributor uploads some new or modified software code to a project. CVS commits have been used to measure OSS productivity because CVS commits are similar to completed modification requests (MRs) in commercial development environments (Grewal et al. 2006; Midha and Bhattacherjee 2012; Mockus et al. 2002; Singh et al. 2011).

Mediator Variable

Developer commitment. We modified the organizational commitment scale items from Ahuja et al. (2006) to make

GNOME the referent by substituting the word GNOME for organization. Sample questions include “I show by my actions that I really care about the fate of GNOME” and “I am extremely glad to have chosen GNOME to work for over other projects.” The Cronbach’s alpha reliability score for the scale is .79.

Independent Variables

Developer OSS values. We modified the values scale used by Stewart and Gosain (2006) to measure developer OSS values. Sample items include “I believe in helping others” and “I value the reputation I gain by participating in open source projects.” Reliability for the scale is .72.

Perceived OSS community values. In keeping with the person–environment fit literature’s use of commensurate measures, the items from the OSS values scale were modified to reference the OSS community. Sample questions include: “Members of this OSS community believe in helping others” and “Members of this OSS community value the reputation gained by participating in open source projects.” The reliability of the scale is .87.

Moderator Variable

Developer degree centrality. Our measure of degree centra- lity represents the potential number of paths through which information—in particular project-related information (e.g., norms of conduct, technical aspects of the system, etc.)— spreads. The degree centrality measure introduced by Free- man (1978) guided our operationalization of degree centrality. It is measured as the number of developer i’s (direct) ties, normalized by the size of the network. This is given by Ni/N where Ni is the number of ties that developer i has and N is the total number of developers in the network.

We created dyads based on email senders and receivers. Mailing lists embody the main forum for posting questions as well as comments and responding to them. The topics range from specific technical discussions to broad discussions about processes, norms, or strategy in the community. An indi- vidual was uniquely identified by determining the one or more email addresses that he used for his contributions in the mailing lists as well as that listed in the commits (if they had any). The links represent an email interaction in a discussion. For every email, we extracted from the email header the message identifier, the sender, the sent time, and the identifier of the message (if any) to which this message was a reply. The origin of an email message can be determined by the email address of the sender. The responder is more difficult to identify because when a person sends an email it can be

4Since we had pre- and post-survey data on network position, communication activity, and coding activity for all 1,788 developers, we conducted a Heck- man two-step procedure to ensure that there was no selection bias in the sample that participated in the survey (Heckman 1979). The results of the Heckman selection model indicate that selection bias was not a concern in our model estimation.

962 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

read by anyone who subscribes to the list. But deciphering if a person reads the email remains difficult. Consequently, we only create an interaction dyad based on the people who explicitly replied to an email. We assume that the responder read the message. When a “reply-to” header was found, the responder found the initial message to be of interest, and the responder was marked as a recipient of the original message. Any person who responds to the email is connected to the sender of the email (whether the person is the 1st or 31st

responder). The sample includes emails sent during the 12- month period (December 2009 and December 2010). The GNOME community product release cycle, which occurred once every 6 months, informed the selection of the 12-month period used. The releases typically take place in March and September. Considering CVS commit and communication data over 12 months allows us to capture the typical nature of the work around two releases.

Control Variables

Using data collected through the survey, we controlled for gender, age, education, and volunteer status (i.e., whether the developer was paid to contribute or was volunteering) because these factors may impact commitment (Brierley and Cowton 2000; Daniel et al. 2018; Hrebiniak and Alutto 1972). These control variables are included in all model estimations.

Results

In order to assess the validity of the measurement scales, we conducted a confirmatory factor analysis using varimax rota- tion (Fornell and Larcker 1981). Results of the factor analysis are shown in Appendix B. With the exception of two items, all items had loadings above .66 on their expected constructs and cross-loadings lower than .31, thus demonstrating ade- quate convergent validity and discriminant validity. The two items with low loadings were removed from further analysis. In addition, the square root of the average variance extracted for each construct was higher than the correlation between that construct and all other constructs, further demonstrating the discriminant validity of the measures (Fornell and Larcker 1981). Table 1 displays descriptive statistics and correlations for all measures.

Polynomial Regression Analysis

The hypotheses in the research model emphasize the effects of congruence and incongruence in values between devel- opers and their perceptions of the environment (the OSS community). Prior research on congruence often relies on

single-index measures such as difference scores, direct mea- surement, or profile similarity indices (Edwards 1993; Klein et al. 2009). However, due to limitations related to concep- tual ambiguity, discarded information, and unrealistically restrictive constraints, these approaches have received criti- cism (Edwards and Parry 1993). In contrast, polynomial regression analysis incorporates separate person and envi- ronment ratings on commensurate measures and permits a high degree of precision in specifying and testing relation- ships about congruence and incongruence (Brown et al. 2012, 2014; Edwards 2002; Klein et al. 2009). For these reasons, polynomial regression is well-suited for our examination of the effects of OSS value congruence and incongruence on commitment.

Polynomial regression analysis involves the estimation of coefficients for higher-order terms (e.g., quadratic terms) and such terms are especially susceptible to outliers in the data (Edwards 2002). To determine the extent to which this might raise concern in our sample, we screened the data for outliers by examining Cook’s D and the standardized residuals from the polynomial regression equations (Bollen and Jackman 1990). Although three cases were flagged as potential out- liers, no major outliers emerged from the analysis. Estimating the models with and without these observations did not mean- ingfully affect the results. It is rare for two or three cases to severely influence observed patterns of relationships in samples of this size (Bollen and Jackman 1990). We mean- centered the variables prior to computing the interaction and quadratic terms in order to reduce the potential for nonessen- tial multicollinearity and facilitate better interpretation of the response surface plots (Aiken and West 1991; Dalal and Zickar 2012).

The regression coefficients from a polynomial regression equation enable the generation of three-dimensional response surface graphs to depict the relationship between two com- mensurate measures (e.g., developer values and perceived OSS community values) and their effect on an outcome (e.g., developer commitment). These graphs facilitate better inter- pretation of the precise nature of congruence and incon- gruence relationships (e.g., Brown et al. 2012, 2014). Given our focus on the influence of value congruence and incon- gruence on developer commitment among developers with different levels of degree centrality, we used moderated polynomial regression to test the hypotheses and the response surface methodology to interpret the results. In polynomial regression analysis, value congruence is operationalized by all values at which developer OSS values (X) are equal to per- ceived OSS community values (Y) (i.e., all cases where X = Y). Tests of the relationship between OSS value congruence and commitment involve the use of nonparametric techniques to determine if the slope of the relationship is statistically sig-

MIS Quarterly Vol. 43 No. 3/September 2019 963

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Table 1. Correlations and Descriptive Statistics

Variable Mean SD 1 2 3 4 5 6 7 8

1. Developer contribution activity

30.21 57.59

2. Developer commitment 4.77 1.15 .23*** (.89)

3. Developer values 6.34 0.51 .04 .27*** (.71)

4. Perceived OSS community values

5.78 0.86 .10* .51*** .33*** (.86)

5. Developer centrality 0.01 0.05 .29*** .10* .02 .01

6. Age 29.7 8.26 -.02 -.06 -.01 -.01 .02

7. Gender NA NA .06 .07 .09† .08† .05 -.04

8. Education NA NA .03 -.09† .07 .01 .05 .20*** .10*

9. Volunteer status NA NA -.14** -.06 .11* -.03 .06 .01 -.05 .04

Notes: n = 410. 1. For the purpose of interpretability, the non-transformed means and standard deviations for number of commits are shown. 2. Gender = dummy variable (0 = women, 1 = men), volunteer status (0 = not paid, 1 = paid). †p < .10, *p < .05, **p < .01, ***p < .001.

nificantly different from 0. Value incongruence is operation- alized by all values at which developer OSS values are differ- ent from perceived OSS community values. This involves cases where developer OSS values exceed perceived OSS community values (X > Y) as well as cases where perceived OSS community values exceed developer OSS values (X < Y). Tests of the relationship between value incongruence and commitment employ nonparametric tests to determine if their effects are statistically significantly different from 0.

The polynomial regression model was estimated using Stata with project-level fixed effects. In polynomial regression analysis, simple regression models are tested and then pro- gressively higher-order terms are entered into the equation. The simpler model is rejected when a higher-order model explains statistically significantly greater additional variance and when the constraints imposed by the simpler model are rejected (Edwards 2002). Following convention in polyno- mial regression analysis, we estimated the following equa- tions to represent the theoretical model (Edwards and Parry 1993):

Z = b0 + b1X + b2Y + e (1)

Z = b0 + b1X + b2Y + b3X 2 + b4XY + b5Y

2 + e (2)

Z = b0 + b1X + b2Y + b3X 2 + b4XY + b5Y

2 + b6W + b7WX + b8WY b9WX

2 + b10WXY + b11WY 2 + e

(3)

where Z = developer commitment, X = the developer’s values, Y = perceptions about the OSS community’s values, and W = developer’s centrality within the OSS community communication network.

Confirmatory Polynomial Regression Analysis

Polynomial regression analysis also makes it possible to empirically test specific constraints that might be imposed on the specification of the regression model by a theoretical viewpoint. Polynomial regression analysis can empirically test whether only the developer OSS values matter or only the perceived OSS community OSS values matter in predicting developer commitment (Brown et al. 2014; Edwards 2002). If developer OSS values alone matter in predicting developer commitment, then, as Brown et al. (2014) suggest, the fol- lowing constraints would be imposed on Equation (1):

Constraint 1: b1 > 0 (i.e., the coefficient on developer OSS values should be positive)

Constraint 2: b2 = 0 (i.e., the coefficient on OSS community values should be nonsignificant)

Conversely, if OSS community OSS values alone matter in predicting developer commitment, then the following con- straints would be placed on Equation (1):

Constraint 1: b1 = 0 (i.e., the coefficient on developer OSS values should be nonsignificant)

Constraint 2: b2 > 0 (i.e., the coefficient on OSS community values should be positive)

Table 2 shows the results of the polynomial regression pre- dicting OSS commitment. In model 1, we tested Equation (1). As the results for model 1 show, the linear terms for devel- opers’ OSS values and perceptions of OSS community values

964 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Table 2. Results of Moderated Polynomial Regression Analysis Predicting Developer Commitment

Variable Coefficient 1 2 3

Age -.03 -.04 -.04

Gender .03 .03 .03

Education -.10* -.10* -.10*

Volunteer status -.05 -.05 -.05

Developer values b1 .16*** .16* .15*

OSS community values b2 .45*** .50*** .56***

Developer values-squared b3 .13* .13*

Developer values x OSS community values b4 -.14* -.09 †

OSS community values-squared b5 .15* .21**

Developer centrality b6 .15**

Developer values × developer centrality b7 .07

OSS community values × developer centrality b8 -.16**

Developer values-squared × developer centrality b9 .00

Developer values × OSS community values × developer centrality b10 .07

OSS community values-squared × developer centrality b11 .16**

Project fixed effects Yes Yes Yes

Lowest VIF 1.023 1.025 1.030

Highest VIF 1.098 3.303 4.435

Adjusted R² .27*** .30*** .33***

∆R² .03*** .03***

Notes: n = 410. *p < .05, **p < .01, ***p < .001.

explained 27% of the variance in developer commitment. The coefficients on developer OSS values and perceived OSS community values are positive and significant. Thus, the constraints implicitly imposed in prior research are rejected. This suggests that there is value in considering the joint effects of developer OSS values and perceived OSS community values on developer commitment. In light of the constraints being rejected, we proceeded to perform an exploratory polynomial regression analysis in which the constraints were relaxed.

Exploratory Polynomial Regression Analysis

In model 2, we tested Equation (2). The results show that the quadratic model (model 2) explained statistically signi- ficantly greater variance in developer commitment than did the linear model (model 1) (R² = .30, ΔR² = .03, p < .05) resulting in a medium effect size of .43 (Cohen et al. (2003). Therefore, per the guidelines of Edwards (2002), we reject the linear model in favor of the quadratic model. In model 3, we tested for the moderating effect of developer degree centrality. This model explained statistically significantly greater variance in developer commitment than did the

quadratic equation with no moderation (i.e., model 2) (R² = .33, ΔR² = .03, p < .05) resulting a medium effect size of .49.5

Response Surface Methodology

The polynomial regression equations represent an important first step in modeling the joint effects of commensurate measures on the outcome of interest. To meaningfully interpret the results, we graphically plot these relationships and examine their key features (Edwards 2002; Edwards and Parry 1993). The response surface methodology pro- vides the necessary analytical tools to conduct hypothesis tests of various response surface features. Given our focus on the OSS value congruence and incongruence effects on developer commitment, we tested the slope of the surface along these specific lines of interest. In light of the signi- ficant moderating effect of developer social network cen-

5We also estimated a cubic model to see if the moderated quadratic model would be rejected. The cubic model did not explain any additional vari- ance in OSS commitment. Therefore, model 3 was retained for the pur- poses of the response surface analysis.

MIS Quarterly Vol. 43 No. 3/September 2019 965

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

trality, we graphed the response surface at high (i.e., one standard deviation above the mean) and low (i.e., one standard deviation below the mean) levels of developer centrality.

Testing the significance of the slopes along the line of con- gruence (i.e., the line along which developer OSS values equal perceived OSS community values) and incongruence (i.e., the line along which developer OSS values and per- ceived OSS community values are unequal) at high versus low levels of developer centrality requires nonparametric techniques. Following the recommendation of Edwards (2002), we used a bootstrapping procedure to test the significance of the slope of the response surface along the line of congruence and incongruence (Efron and Tibshirani 1993). Bootstrapping is generally preferred over jack- knifing for small sample sizes such as that used in this study (Efron and Tibshirani 1993). Using the bootstrapping ap- proach, we constructed bias-corrected confidence intervals around the estimates of the slopes of the response surface based on equation (4) below (Edwards 2002).

Z = (b0 + b6W) + (b1 + b7W)X + (b2 + b8W)Y + (b3 + b9W)X

2 + (b4 + b10W)XY + (b5 + b11W)Y 2 (4)

+ e

where values of W at one standard deviation above versus below the mean for developer centrality were used to com- pute the simple coefficients for the response surface.6

Tests of Hypotheses Regarding Effects of OSS Value Congruence and Incongruence

Table 3 presents the results of the tests of slopes along the line of congruence and incongruence and Figure 2 shows the graphical plots of the response surfaces. In H1, we posited that OSS value congruence would have a stronger positive effect on commitment among less central devel- opers than among highly central developers. As the results in Table 3 show, when developer centrality is high, the slope of the surface is positive (ax = .69, p < .001). When developer centrality is low, the slope of the surface along the line of congruence is positive (ax = .73, p < .001). The linear slope of the response surface along the line of congruence is stronger when developer centrality is low than when developer centrality is high. This supports H1.

H2a predicted that, as OSS community values exceed devel- oper OSS values, the effects on commitment would be more positive among less central developers than among highly central developers. H2b posited that, as developer OSS values exceed OSS community values, the effects on com- mitment would be more negative among less central developers than among highly central developers. For these hypotheses to be supported, the linear slope of the response surface along the line of incongruence should be more negative among less central developers than among highly central developers. As the results in Table 3 show, the linear slope of the response surface along the line of incongruence is negative and significant among less central developers (ay = -.47, p < .05) and negative but nonsigni- ficant among highly central developers (ay = -.35, p > .10). This provides support for H2a and H2b. This indicates that, among highly central developers, the effects on developer commitment are the same whether X > Y or X < Y. The significant negative linear slope of the response surface for less central developers also suggests that commitment increases as OSS community values exceed developer OSS values; commitment decreases as developer OSS values exceed OSS community values.

Test of Mediation Hypothesis

In H3 we posited that the relationship between value con- gruence and incongruence on developer code contributions would be mediated by commitment. To test this hypothesis, we conducted mediation tests using bootstrapping (Edwards and Lambert 2007; Preacher et al. 2007), which makes no assumptions about the shape of the sampling distribution for the mediation effects and, thus, is robust to non-normal distributions (MacKinnon et al. 2004; Preacher and Hayes 2004).

Before commencing with the mediation tests, we computed a block variable for value congruence and incongruence using equation (2) to account for the interactive, linear and nonlinear effects of the two variables (developers’ values and perceived OSS community values). Klein et al. (2009) recommend this approach for testing the effects of poly- nomial regression coefficients in path models such as mediation and moderated-mediation models. Following the procedure outlined by Cable and Edwards (2004), we regressed the dependent variable (number of commits) on the five value congruence terms using negative binomial regression analysis.7 The coefficients were then used to compute the block variable. This block variable was used

6The simplified polynomial regression equations at low versus high values of developer centrality are shown in Appendix C.

7Negative binomial regression is robust to the non-normal distribution reflected in count data such as number of commits.

966 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Table 3. Results of Tests of Slopes Along Lines of Congruence and Incongruence

Surface along line of congruence H1 (X = Y)

Surface along line of incongruence H2 (X = -Y)

Level of centrality (W) Linear slope

(ax) Curvature

(ax2) Linear slope

(ay) Curvature

(ay2)

High centrality .69*** .31 -.35 .45

Low centrality .73*** .19 -.47* .41

Notes: 1. The linear slope of the response surface along the line of congruence is given by ax = (b1 + b2 + [b7 + b8]W) and the curvature of the slope

along this line is given by ax2 = (b3 + b4 + b5 + [b9 + b10 + b11]W). 2. The linear slope of the response surface along the line of incongruence is given by ay = (b1 - b2 + [b7 - b8]W) and the curvature of the slope

along this line is given by ay2 = (b3 - b4 + b5 + [b9 - b10 + b11]W). 3. Significance levels of estimated coefficients are based on bias-corrected confidence intervals constructed from 10,000 bootstraps at low and

high centrality. 4. *p < .05, **p < .01, ***p < .001.

(a) High Developer Centrality (b) Low Developer Centrality

Figure 3. Graphical Plots of the Response Surfaces for High Versus Low Developer Centrality

Table 4. Results of Mediation (H3) and Moderated-Mediation (H4) Tests on Number of Commits

Hypothesis Level of Developer

Centrality Direct Effect Indirect Effect

95% Bias-Corrected CI

R²Lower Bound Upper Bound

H3 .55 (.31) .32 (.14) .08 .63 .21***

H4 High .41 (.33) .39 (.19) .02 .97 .30***

Low .58 (.19) .28 .99

Notes: For H4, value congruence and incongruence block variable is the independent variable, developer commitment is the mediator, developer centrality is the moderator; control variables were included as covariates in the analysis; standard errors are in parentheses; dependent variable was log-transformed prior to analysis ***p < .001

MIS Quarterly Vol. 43 No. 3/September 2019 967

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

as the independent variable in the mediation tests. To esti- mate the mediation (indirect) effects and the confidence intervals for the differences between estimates at high versus low levels of the moderator (developer centrality), we drew 1,000 bootstrap samples. To account for any potential difference between the product of coefficients from the full sample and the median estimate from the bootstrap samples, we computed bias-corrected confidence intervals (Edwards and Lambert 2007; Efron and Tibshirani 1993; Stine 1989). Table 4 shows the results of the analysis (see Appendix D for coefficient estimates for the moderated regression models).

The results in Table 4 show that value congruence and incon- gruence does not have a significant direct effect on number of commits (.55, p > .10); instead the mediated effect through commitment is significant as indicated by the bias-corrected confidence intervals (indirect effect: .32, CI: .08, .63). This provides support for H3.

In H4, we posited that the mediation effect of commitment in the relationship between value congruence and incongruence and developer code contributions would be stronger among less central developers than among highly central developers in the communication network. Thus, we tested the extent to which the mediation effect in H3 was moderated by developer centrality. We drew 1,000 bootstrap samples to estimate the confidence intervals for the differences between mediation effects at high versus low levels of the moderator (developer centrality). The mediated effect of value congruence and incongruence on number of commits was significant at high (indirect effect = .39, CI: .02, .97) and low (indirect effect = .58, CI: .28, .99) levels of developer centrality. Tests of differences in the mediated effects at high versus low developer centrality were nonsignificant (.39 – .58 = -.19, p > .10), suggesting that there was no difference in the indirect effect of value congruence and incongruence on developer code contributions. H4 is not supported, suggesting that the moderating effect of developer centrality is mediated by developer commitment. It is worth noting that we found support for H4 in our robustness analysis when using more granular measures of code contribution—namely, number of lines of code added or deleted and number of files changed (see Appendix G). Number of commits does not reflect this level of granularity in the amount of work contributed by each developer.

Robustness Analysis

We conducted additional analyses to assess the robustness of our results including using different time frames for com- puting degree centrality in the communication network (past

24 months), changing the basis for computing degree cen- trality to collaboration on prior projects (past 12 months and 24 months), performing a two-stage least squares analysis to account for potential endogeneity between developer commit- ment and centrality, and performing a Heckman selection procedure to rule out the influence of self-selection. These robustness tests are detailed in Appendices E, F, and H, respectively .

Discussion

The OSS approach to developing software—which has traditionally been enshrined in the OSS values—has attained success over the years, in large part, by attracting developer participation. Given the increasing corporate presence in OSS communities, practices have changed to varying degrees (Daniel et al. 2018; Spaeth et al. 2015, Stewart et al. 2006, von Krogh et al. 2012) and researchers and practitioners need to understand what happens when a developer grapples with issues of gaining certainty about how to participate in the community. A mismatch in values can fuel uncertainty and result in negative outcomes. However, OSS communities with corporate influence desire to retain the ongoing contribu- tions of developers with different values. After all, corporate entities participate to reap the benefits of volunteer developer efforts. To advance our theoretical understanding of how OSS leaders can foster developer commitment and ongoing contributions, we focused on the role of OSS value con- gruence, incongruence, and centrality in the communication network as opportunities for passive and interactive informa- tion seeking to reduce uncertainty about how to contribute code. That is, we integrated the person–environment fit perspective with social network theory. Our results indicate that the impact of OSS value congruence and incongruence is more salient among developers who are less central in the community’s communication network. Next, we discuss the theoretical implications of our findings.

Theoretical Contributions

Implications for OSS Community Research

This research makes several contributions to the literature. From a theoretical standpoint, this research leverages URT to advance understanding of the roles of OSS values and com- munication networks as uncertainty reduction mechanisms. Research in this domain has not considered OSS values and communication networks in this light. By leveraging URT, we explain that these constitute alternative relational infor-

968 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

mation seeking mechanisms that facilitate developers’ ability to meaningfully participate in making code contributions by clarifying appropriate collaboration practices. Using this theoretical lens, we gain insight into the role of OSS values as a relational mechanism that facilitates developers’ passive information seeking approaches to uncertainty reduction, and communication networks as a relational mechanism that facilitates interactive information seeking approaches to uncertainty reduction. These avenues represent an essential means for enabling developers to contend with common barriers to participation in digitally enabled communities such as OSS (Steinmacher et al. 2015; Steinmacher et al. 2016). Research has not considered the nature of the relationship between these two relational aspects of developers within their OSS communities. This is theoretically significant because it points to a degree of equifinality in fostering OSS developer commitment. By using this theoretical perspective, we add to the current body of work that has used theories including self-determination theory (Daniel et al. 2018; Ke and Zhang 2009) and legitimate peripheral learning to understand code contributions in OSS development (Fang and Neufeld 2009).

We contribute to theoretical understanding of OSS values and their impact on OSS development. As our review suggests, this topic has received considerable attention in the IS litera- ture since the seminal work by Stewart and Gosain (2006). Much of the research on this topic theorized the impact of OSS values in isolation, focusing only on developers or only on OSS communities. Consequently, researchers reached mixed, and often contradictory, conclusions about the impact of OSS developer values on attitudes and behavior based on an incomplete view of developers and their environment. Our research shifts the conversation in this research stream by demonstrating the necessity of understanding developers within their community environment. By outlining this developer-within-community-environment (i.e., OSS value congruence and incongruence, and social connectedness) ÷ attitudes (i.e., commitment) ÷ contribution behavior (i.e., code commits) chain, we show that the emotional attachments reflected in commitment represent an important mediating mechanism that informs developers’ code contributions to an OSS community. This repesents a reorientation in the focus of research on OSS values because it adopts a more holistic and relational view of developers and the communities in which they participate vis-à-vis congruence and incon- gruence. The results of our research demonstrate that developer OSS values—relative to their perceptions of the community OSS values—shape developers’ commitment. This finding enables future research to explicitly account for developers within the context of their community environment.

From an empirical standpoint, we contribute to the literature through our moderated polynomial regression model. To the best of our knowledge, this is the first empirical study to examine moderation effects on polynomial regression and response surface features within the information systems literature. With the growing number of studies employing this analytical method, this research provides a useful blueprint for testing boundary conditions that might affect the topology of response surfaces.

Implications for Uncertainty Reduction Theory

This research contributes to URT by elaborating the theo- retical consideration of passive approaches to uncertainty reduction in digitally enabled environments. Prior theoretical views on uncertainty reduction in digitally enabled environ- ments, such as online communities, have primarily focused on explicitly communicated information (in the form of text) as representing the passive approach. In this view, someone could passively reduce uncertainty by reading what others post. This is not surprising given that text communication represents the main purpose and form of exchange in such environments. Our context, software development, enables us to extend theoretical treatment of passive approaches in digitally enabled environments beyond text communication. We focused on productive technical work, where explicit communication represents only one aspect of ongoing exchange. In addition to text communication, the behaviors that we observe manifest as encoded actions (e.g., helping and sharing code). As exchange between participants in digitally enabled environments grows to encompass more than just text communication, our conceptualization of passive approaches to uncertainty reduction—as encompassing this broader source of information—will enable more robust application of URT to digital phenomena.

This research contextualizes URT to digitally enabled forms of organizing. Although URT has been applied to a variety of digital domains including online dating (e.g.,Gibbs et al. 2015), social networking (e.g., Anthuenis et al. 2010), and access to e-government services (e.g., Venkatesh et al. 2016), much of this prior work has focused on the independent relationships of different uncertainty reduction approaches on outcomes of interest. We contribute to this existing corpus of work by theoretically elaborating how these approaches interact. We show that approaches to uncertainty reduction in digitally enabled environments are complementary in nature and the determination of which approach has primacy in enabling individuals to feel emotionally committed to such environments is informed, in part, by one’s connectedness within the community.

MIS Quarterly Vol. 43 No. 3/September 2019 969

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Implications for the Person–Environment Fit Perspective

This research makes a theoretical contribution to the person–environment fit perspective. By incorporating the moderating role of communication network centrality, we advance the person–environment fit perspective by explicitly recognizing that the relationship between individuals and their environment is characterized by two layers. The first layer comprises the foundation upon which the person–environment fit perspective is built—namely, the relation between an individual and their environment with regard to a specific characteristic (e.g., OSS values). The second relational layer which we incorporate into the person–environment fit perspective concerns the individual’s connectedness to the environment. This second relational layer gives meaning to the salience of congruence or incongruence in the first relational layer. This is the first study to theoretically com- bine these two relational views to provide a holistic appre- ciation for individuals within their environments. It answers a call by prior reviews and meta-analyses of person–environ- ment fit research to develop a more theoretically nuanced understanding of the relationship between people and the environments in which they work (Kristof-Brown and Guay 2011; Kristof-Brown et al. 2005). This research suggests that, in order to develop a better understanding of phenomena, future person–environment fit research needs to account for the connectedness of the “person” aspect of person–environ- ment fit within the environment being studied.

Our examination of the impact of value congruence and incongruence in the OSS community context contributes to the broader person–environment fit literature by considering an understudied context. Johns (2006) emphasized the value of theorizing how well-studied phenomena unfold in different contexts. Open collaboration environments, such as OSS communities, depend on a different set of governance mech- anisms than the traditional organization settings in which person–environment fit has typically been studied (O’Mahony and Bechky 2008; O’Mahony and Ferraro 2007; Seidel and Stewart 2011). As a form of virtual organization, these com- munities rely on the collaborative effort of a voluntary and paid workforce. This context brings two key elements to the foreground that are secondary in typical person–environment fit study settings. First, in OSS communities, the values come to embody preferred approaches to collaborating. This is significant because OSS community membership is fluid in virtual organizations (Faraj et al. 2011). Leaders must enshrine acceptable practices in the community values because newcomers enter frequently and are likely unfamiliar with acceptable behaviors (Steinmacher et al. 2015). One implication of this contextual difference is that, counter to meta-analytic findings in traditional organizational settings,

excess values (i.e., conditions where the OSS community embraces the values more than the developer does) results in greater commitment among developers. Kristof-Brown et al.’s (2005) meta-analytic review states that “excess E condi- tions have little negative effect on attitudes, whereas excess P conditions accompany dramatic decreases in attitudes” (p. 313). Drawing on URT, we argue that the nature of OSS communities makes clarity around practices especially important and desirable even when they are at odds with an individual’s own preferences. This is particularly salient for socially isolated developers within the community’s commu- nication network. In virtual organizations, such as OSS communities, digital communications serve as the primary channel for understanding how to contribute within the community.

Implications for Practice

We stated at the outset that a major challenge for OSS communities rests in attracting the commitment of developer contributions. This challenge has its roots at least partially in the shifting make-up of OSS communities as commercial firms increasingly devote their own resources (including their developers) to such communities (Aksulu and Wade 2010; Daniel et al. 2018; Dwoskin 2016; Shah 2006; Spaeth et al. 2015; Stewart et al. 2006). While OSS community leaders hold little control over who contributes and what values they embrace, the research findings suggest that leaders should err in favor of greater clarity regarding the acceptable community practices. This clarity holds unique importance for gaining the commitment of less well-connected developers. Prior empirical evidence suggests that such developers stand out in their ability to promote the popularity and quality of OSS projects (Setia et al. 2012). Developers face real barriers (Steinmacher et al. 2015) and OSS communities can ill-afford to lose such a vital resource. Embracing the OSS values to emphasize cooperation, helping, and the pursuit of technical knowledge clearly increases the commitment of such developers, even if they do not embrace those values.

As hybrid forms of software development become increasingly common (Aksulu and Wade 2010), OSS com- munity managers must look to alternative ways of endearing themselves to developers in a manner that is robust to the fact that OSS values surely differ. In this regard, the research findings suggest that having more social connections within the communication network is one such mechanism for main- taining developer commitment. Importantly, a high degree of connectedness within the community’s communication network mitigates the negative influence of OSS value incongruence. OSS community managers may benefit from efforts to ensure that developers on the periphery of the

970 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

community can access a pathway to becoming more con- nected to other members of the community. For instance, community managers can embed gamification incentives (e.g., badges, leaderboards) that reward establishment of social connections to less well-connected developers. Such efforts might encourage developers to reach out to less well- connected members of the community and explicitly bring them into the fold.

Limitations and Future Research

As with any field research, our study has several limitations that should be taken into account when interpreting the results. First, observing GNOME developers allowed us to naturally control for unobserved between-community dif- ferences that might impact the results. At the same time it limits the generalizability of the conclusions. GNOME remains much larger than many other OSS communities and probably receives more contributions from corporate sponsors. Generalizing to smaller communities must be done with caution. It is worth noting that GNOME represents the way in which an increasing number of OSS communities are evolving as corporate sponsors increase their engagement (Aksulu and Wade 2010; Dwoskin 2016).

An associated limitation related to the sample of GNOME developers is that it may not represent all developers in the GNOME community. To minimize concern related to this issue we verified that on key variables (e.g., the number of concurrent versions system commits, the number of emails) the developers in our sample do not differ significantly from the larger set of developers in the GNOME community. Further, the results of our Heckman (1979) selection model indicated that selection bias was not a concern. The results of this analysis suggest that the sample likely represents the broader community.

One other limitation related to the sample is that the devel- opers were predominantly men. Although men make most code contributions to OSS development (Ghosh et al. 2002; Hars and Ou 2002; Jensen et al. 2011; Ke and Zhang 2009; Kuechler et al. 2012; Nafus 2012; Robles et al. 2016), one must question the degree to which gender shapes our results. Prior work suggests that men tend to be more instrumental in their interactions with technology while women tend to be more relational (Venkatesh and Morris 2000; Venkatesh et al. 2000). This distinction raises an important question about how the findings might differ in the event of a greater repre- sentation of women in such environments. Of particular interest is whether and how the combination of instrumental (e.g., executing coding tasks) and relational (e.g., sharing knowledge, communicating with others) factors might impact

commitment to a particular OSS community. If prior research provides a guide, it is possible that, among women, the relational mechanisms studied here are likely leveraged to establish social connections to community as opposed to being driven by purely instrumental ends. Future research may want to consider methods to reduce uncertainty that are less instrumental and more relationship focused.

There are several research opportunities to extend what we learned here to the study of other virtual communities. Given that the mechanisms used to govern contributor efforts differ across various OSS communities (O’Mahony and Bechky 2008), exploring the model developed here in OSS communities that work using different governing bodies represents one such opportunity. Future research should explore this model in communities that create products other than software.

Each OSS developer can take part in several networks that can be defined by participating in OSS development efforts, conference attendance, telephone calls, emails, or sequential contributions of source code. Each network connection could represent an alternative route through which a developer could learn about OSS values and what matters in an OSS community. This research considers both email communica- tion and contributions to the same project. Future research should consider other potentially relevant networks in this context. In particular, observing the spillover of unique values from one OSS development effort to another may represent a useful avenue for future research. Finally, we explore value congruence and incongruence, and commitment at one point in time. Commitment often varies over time (Solinger et al. 2013). Thus exploring how commitment to an OSS community changes over time may offer deeper insight into activity patterns. For instance, Solinger et al. (2013) report a scenario they name “Honeymoon Hangover,” where commitment starts high and quickly drops off. In addition, observing value congruence and incongruence, and commit- ment at the same time limits the ability to infer causality. Now that we empirically demonstrated relationships between them, future research is necessary to identify the nature of causation.

Conclusion

OSS communities leverage resources from a globally dis- tributed network of skilled volunteers and paid contributors. To understand the factors that motivate and enable developers to contribute to these communities requires an analysis of both developer and OSS community values, as well as the developers’ structural network position within the community.

MIS Quarterly Vol. 43 No. 3/September 2019 971

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

The research presented here leverages URT to provide nuanced insights into how value congruence and incon- gruence impact commitment and contribution activity among OSS developers who are well-connected within the com- munication network versus those who are less so.

Acknowledgments

The authors express their gratitude for the tremendous guidance of the senior editor, Sue Brown, the exceptional direction from the associate editor, and the constructive feedback from the reviewers. All errors or omissions are those of the authors. We thank seminar series participants at University of Notre Dame, Bocconi University, and Georgia State University’s Center for Process Innovation for helpful feedback and engaging discussions during earlier presenta- tions of the work. We also thank Jan DeGross for her exceptional effort with copyediting and typesetting the article.

References

Agerfalk, P. and Fitzgerald, B. 2008. “Outsourcing to an Unknown Workforce: Exploring Opensourcing as a Global Sourcing Strategy,” MIS Quarterly (32:2), pp. 385-409.

Ahuja, M. K., Galletta, D. F., and Carley, K.M. 2003. “Individual Centrality and Performance in Virtual R&D Groups: An Empirical Study,” Management Science (49:1), pp. 21-38.

Aiken, L. S., and West, S. G. 1991. Multiple Regression: Testing and Interpreting Interactions, Newbury Park, CA: Sage Publications.

Aksulu, A., and Wade, M. 2010. “A Comprehensive Review and Synthesis of Open Source Research,” Journal of the Association for Information Systems (11:11), pp. 576-656.

Allen, N. J., and Meyer, J. P. 1997. Commitment in the Workplace: Theory, Research and Application, Thousand Oaks, CA: Sage Publications.

Antheunis, M. L., Valkenburg, P. M., and Peter, J. 2010. “Getting Acquainted Through Social Network Sites: Testing a Model of Online Uncertainty Reduction and Social Attraction,” Computers in Human Behavior (26:1), pp. 100-109.

Ashford, S. J., Lee, C., and Bobko, P. 1989. “Content, Cause, and Consequences of Job Insecurity: A Theory-Based Measure and Substantive Test,” Academy of Management Journal (32:4), pp. 803-829.

Bagozzi, R. P., and Dholakia, U. M. 2006. “Open Source Software User Communities: A Study of Participation in Linux User Groups,” Management Science (52:7), pp. 1099-1115.

Bateman, P., Gray, P., and Butler, B. 2011. “The Impact of Community Commitment on Participation in Online Commu- nities,” Information Systems Research (22:4), pp. 841-854.

Becker, H. S. 1960. “Notes on the Concept of Commitment,” American Journal of Sociology (66:1), pp. 32-40.

Benbya, H., and Belbaly, N. 2010. “Understanding Developers’ Motives in Open Source Projects: A Multi-Theoretical Frame- work,” Communications of the AIS (27:1), pp. 589-610.

Berger, C. R., and Calabrese, R. J. 1975. “Some Explorations in Initial Interaction and Beyond: Toward a Developmental Theory of Interpersonal Communication,” Human Communication Research (1:2), pp. 99-112.

Bollen, K. A., and Jackman, R. W. 1990. Regression Diagnostics: An Expository Treatment of Outliers and Influential Cases, Newbury Park, CA: Sage Publications.

Bordia, P., Hunt, E., Paulsen, N., Tourish, D., and DiFonzo, N. 2004. “Uncertainty During Organizational Change: Is it All about Control?,” European Journal of Work and Organizational Psychology (13:3), pp. 345-365.

Borgatti, S. P., and Halgin, D. S. 2011. “On Network Theory,” Organization Science (22:5), pp. 1168-1181.

Brashers, D. E. 2007. Communication and Uncertainty Manage- ment, Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Brierley, J., and Cowton, C. 2000. “Putting Meta-Analysis to Work: Accountants’ Organizational-Professional Conflict,” Journal of Business Ethics (24:4), pp. 343-353.

Brown, S. A., Venkatesh, V., and Goyal, S. 2012. “Expectation Confirmation in Technology Use,” Information Systems Research (23:2), pp. 474-487.

Brown, S. A., Venkatesh, V., and Goyal, S. 2014. “Expectation Confirmation in IS Research: A Test of Six Competing Models,” MIS Quarterly (38:3), pp. 729-756.

Burt, R. 2004. “Structural Holes and Good Ideas,” American Journal of Sociology (110:2), pp. 349-399.

Cable, D. M., and Edwards, J. R. 2004. “Complementary and Supplementary Fit: A Theoretical and Empirical Integration,” Journal of Applied Psychology (89:5), pp. 822-834.

Chatman, J. 1991. “Matching People and Organizations: Selection and Socialization in Public Accounting Firms,” Administrative Science Quarterly (36:3), pp. 459-484.

Chou, S.-W., and He, M.-Y. 2011. “Understanding OSS Develop- ment in Communities: The Perspectives of Ideology and Knowl- edge Sharing,” Behaviour & Information Technology (30:3), pp. 325-337.

Cohen, J., Cohen, P., West, S. G., and Aiken, L.S. 2003. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Coleman, J. 1988. “Social Capital in the Creation of Human Capital,” American Journal of Sociology (94:Supplement), pp. 95-120.

Crowston, K., Annabi, H., and Howison, J. 2003. “Defining Open Source Software Project Success,” in Proceedings of the 24th

International Conference on Information Systems, Seattle, WA, pp. 327-340.

Dabbish, L., Stuart, C., Tsay, J., and Herbsleb, J. 2012. “Social Coding in Github: Transparency and Collaboration in an Open Software Repository,” in Proceedings of the Conference on Computer Supported Cooperative Work, Seattle, Washington, pp. 1277-1286.

Dahlander, L., and O’Mahony, S. 2011. “Progressing to the Center: Coordinating Project Work,” Organization Science (22:4), pp. 961-979.

Dalal, D. K., and Zickar, M. J. 2012. “Some Common Myths about Centering Predictor Variables in Moderated Multiple Regression and Polynomial Regression,” Organizational Research Methods (15:3), pp. 339-362.

972 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Daniel, S., Agarwal, R., and Stewart, K. J. 2013. “The Effects of Diversity in Global, Distributed Collectives: A Study of Open Source Project Success,” Information Systems Research (24:2), pp. 312-333.

Daniel, S., Maruping, L. M., Cataldo, M., and Herbsleb, J. 2018. “The Impact of Ideology Misfit on Open Source Software Com- munities and Companies,” MIS Quarterly (42:4), pp. 1069-1096.

Dishon-Berkovits, M., and Koslowski, M. 2002. “Determinants of Employee Punctuality,” The Journal of Social Psychology (142:6), pp. 723-739.

Dwoskin, E. 2016. “Tech Firms Grapple with How to Make Open Source Pay,” The Wall Street Journal, (available at http://www.wsj.com/articles/tech-firms-grapple-with-how-to- make-open-source-pay-1457031731; accessed May 5, 2016).

Edwards, J. 1993. “Problems with the Use of Profile Similarity Indicies in the Study of Congruence in Organizational Research,” Personnel Psychology (46:3), pp. 641-665.

Edwards, J. R. 1996. “An Examination of Competing Versions of the Person–Environment Fit Approach to Stress,” Academy of Management Journal (39:2), pp. 292-339.

Edwards, J. R. 2002. “Alternatives to Difference Scores: Poly- nomial Regression and Response Surface Methodology,” in Advances in Measurement and Data Analysis, F. Drasgow and N. W. Schmitt (Eds.), San Francisco: Jossey-Bass, pp. 350-400.

Edwards, J. R., and Lambert, L. S. 2007. “Methods for Integrating Moderation and Mediation: A General Analytical Framework Using Moderated Path Analysis,” Psychological Methods (12:1), pp. 1-22.

Edwards, J. R., and Parry, M. E. 1993. “On the Use of Polynomial Regression Equations as an Alternative to Difference Scores in Organizational Research,” Academy of Management Journal (36:6), pp. 1577-1613.

Edwards, J. R., and Rothbard, N. P. 1999. “Work and Family Stress and Well-Being: an Examination of Person–Environment fit in the Work and Family Domains,” Organizational Behavior and Human Decision Processes (77:2), pp. 85-129.

Efron, B., and Tibshirani, R. J. 1993. An Introduction to the Bootstrap, New York: Chapman & Hall.

Ehrlich, K., and Cataldo, M. 2012. “All-for-One and One-for-All? Multilevel Analysis of Communication Patterns and Individual Performance in Geographically Distributed Software Develop- ment,” in Proceedings of the Conference on Computer Supported Cooperative Work, Seattle, WA, pp. 945-954.

Elliott, M. S., and Scacchi, W. 2003. “Free Software Developers as an Occupational Community: Resolving Conflicts and Fostering Collaboration,” in Proceedings of the International ACM SIGGROUP Conference on Supporting Group Work, Sanibel Island, FL, pp. 21-30.

Fang, Y., and Neufeld, D. 2009. “Understanding Sustained Participation in Open Source Software Projects,” Journal of Management Information Systems (25:4), pp. 9-50.

Faraj, S., Jarvenpaa, S. L., and Majchrzak, A. 2011. “Knowledge Collaboration in Online Communities,” Organization Science (22:5), pp. 1224-1239.

Fisher, C. D., and Gitelson, R. 1983. “A Meta-Analysis of the Correlates of Role Conflict and Ambiguity,” Journal of Applied Psychology (68:2), pp. 320.

Fornell, C., and Larcker, D. 1981. “Evaluating Structural Equation Models with Unobservable Variables and Measurement Error,” Journal of Marketing Research (18:1), pp. 39-50.

Freeman, K. S., and Spyridakis, J. H. 2004. “An Examination of Factors That Affect the Credibility of Online Health Informa- tion,” Technical Communication (51:2), pp. 239-263.

Freeman, L. C. 1978. “Centrality in Social Networks Conceptual Clarification,” Social Networks (1:3), pp. 215-239.

Ghosh, R. A., Glott, R., Krieger, B., and Roles, G. 2002. “Free/ Libre Open Source Software: Survey and Study,” FLOSS Deliverable D18: Final Report, Part 4A: Survey of Developers– Annexure on Validation and Methodology, International Institute of Infonomics, University of Maastricht, The Netherlands.

Gibbs, J. L., Ellison, N. B., and Lai, C.-H. 2015. “First Comes Love, Then Comes Google: An Investigation of Uncertainty Reduction Strategies and Self-Disclosure in Online Dating,” Communication Research (38:1), pp. 70-100.

Grewal, R., Lilien, G. L., and Mallapragada, G. 2006. “Location, Location, Location: How Network Embeddedness Affects Project Success in Open Source Systems,” Management Science (52:7), pp. 1043-1056.

Hars, A., and Ou, S. 2002. “Working for Free? Motivations for Participating in Open Source Projects,” International Journal of Electronic Commerce (6:3), pp. 25-39.

Heckman, J. 1979. “Sample Selection as a Specification Error,” Econometrica (47:1), pp. 153-161.

Henkel, J. 2008. “Champions of Revealing—The Role of Open Source Developers in Commercial Firms,” Industrial and Corporate Change (18:3), pp. 435-471.

Hinds, P., and McGrath, C. 2006. “Structures that Work: Social Structure, Work Structure and Coordination Ease in Geographi- cally Distributed Teams,” in Proceedings of the Conference on Computer Supported Cooperative Work, Alberta, Canada, pp. 343-352.

Hoffman, B. J., and Woehr, D. J. 2006. “A Quantitative Review of the Relationship between Person–Organization Fit and Behav- ioral Outcomes,” Journal of Vocational Behavior (68:3), pp. 389-399.

Hogg, M. A. 2000. “Subjective Uncertainty Reduction Through Self-Categorization: A Motivational Theory of Social Identity Processes,” European Review of Social Psychology (11:1), pp. 223-255.

Hogg, M. A., Meehan, C., and Farquharson, J. 2010. “The Solace of Radicalism: Self-Uncertainty and Group Identification in the Face of Threat,” Journal of Experimental Social Psychology (46:6), pp. 1061-1066.

Hogg, M. A., and Terry, D. J. 2000. “Social Identity and Self- Categorization Processes in Organizational Contexts,” Academy of Management Review (25:1), pp. 121-140.

Howison, J., and Crowston, K. 2014. “Collaboration Through Open Superposition: A Theory of the Open Source Way,” MIS Quarterly (38:1), pp. 29-50.

Hrebiniak, L., and Alutto, J. 1972. “Personal and Role-Related Factors in the Development of Organizational Commitment “ Administrative Science Quarterly (17:4), pp. 555-573.

Jansen, K. J., and Kristof-Brown, A. L. 2005. “Marching to the Beat of a Different Drummer: Examining the Impact of Pacing

MIS Quarterly Vol. 43 No. 3/September 2019 973

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Congruence,” Organizational Behavior and Human Decision Processes (97:2), pp. 93-105.

Jensen, C., King, S., and Kuechler, V. 2011. “Joining Free/Open Source Software Communities: An Analysis of Newbies’ First Interactions on Project Mailing Lists,” in Proceedings of the 44th

Hawaii International Conference on System Sciences, New York: IEEE Computer Society.

Johns, G. 2006. “The Essential Impact of Context on Organiza- tional Behavior,” Academy of Management Review (31:2), pp. 396-408.

Ke, W., and Zhang, P. 2009. “Motivations in Open Source Soft- ware Communities: The Mediating Role of Effort Intensity and Goal Commitment,” International Journal of Electronic Com- merce (13:4), pp. 39-66.

Ke, W., and Zhang, P. 2010. “The Effects of Extrinsic Motivations and Satisfaction in Open Source Software Development,” Jour- nal of the Association for Information Systems (11:12), pp. 784-808.

Kidane, Y. H., and Gloor, P. A. 2007. “Correlating Temporal Com- munication Patterns of the Eclipse Open Source Community with Performance and Creativity,” Computational and Mathematical Organization Theory (13:1), pp. 17-27.

Klein, G., Jiang, J. J., and Cheney, P. 2009. “Resolving Difference Score Issues in Information Systems Research,” MIS Quarterly (33:4), pp. 811-826.

Kluckhohn, C. 1951. Values and Value-Orientations in the Theory of Action: An Exploration in Definition and Classification, Cambridge, MA: Harvard University Press.

Knobloch, L. K. 2015. “Uncertainty Reduction Theory,” in International Encyclopedia of Interpersonal Communication, New York: Wiley.

Knobloch, L. K., Satterlee, K. L., and DiDomenico, S. M. 2010. “Relational Uncertainty Predicting Appraisals of Face Threat in Courtship: Integrating Uncertainty Reduction Theory and Polite- ness Theory,” Communication Research (37:3), pp. 303-334.

Kogut, B., and Metiu, A. 2001. “Open Source Software Develop- ment and Distributed Innovation,” Oxford Review of Economic Policy (17:2), pp. 248-264.

Kramer, M. W. 1993. “Communication and Uncertainty Reduction During Job Transfers: Leaving and Joining Processes,” Commu- nications Monographs (60:2), pp. 178-198.

Kristof-Brown, A. L., and Guay, R. P. 2011. Person–Environment Fit, Washington DC: American Psychological Association.

Kristof-Brown, A. L., Zimmerman, R., and Johnson, E. 2005. “Consequences of Individuals’ Fit at Work: A Meta-Analysis of Person–Job, Person–Organization, Person–Group and Person– Supervisor Fit,” Personnel Psychology (58:2), pp. 281-342.

Kuan, J. 2002. “OSS as Lead User’s Make or By Decision: A Study of Open and Closed Source Quality,” paper presented at Open Source Software: Economics, Law and Policy, Toulouse, France, June 20-21.

Kuechler, V., Gilbertson, C., and Jensen, C. 2012. “Gender Differ- ences in Early Free and Open Source Software Joining Process,” IFIP International Conference on Open Source Systems: Long- Term Sustainability, I. Hammouda, B. Lundell, T. Mikkonen, and W. Scacchi (eds.), Berlin: Springer, pp. 78-93.

Lakhani, K., and Wolf, R. 2005. Why Hackers Do What They Do: Understanding Effort and Motivation in Open Source Projects, Cambridge, MA: MIT Press.

Lee, J., and Kim, S. 2011. “Exploring the Role of Social Networks in Affective Organizational Commitment: Network Centrality, Strength of Ties, and Structural Holes,” The American Review of Public Administration (41:2), pp. 205-223.

Lee, T. W., Mitchell, T. R., Sablynski, C. J., Burton, J. P., and Holtom, B. C. 2004. “The Effects of Job Embeddedness on Organizational Citizenship, Job Performance, Volitional Ab- sences, and Voluntary Turnover,” Academy of Management Journal (47:5), pp. 711-722.

Lindberg, A., Berente, N., Gaskin, J., and Lyytinen, K. 2016. “Coordinating Interdependencies in Online Communities: A Study of an Open Source Software Project,” Information Systems Research (27:4), pp. 751-772.

Ljungberg, J. 2000. “Open Source Movements as a Model for Organizing,” European Journal of Information Systems (9:4), pp. 208-216.

Lydon, J. 1996. Toward a Theory of Commitment, Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

MacKinnon, D. P., Lockwood, C. M., and Williams, J. 2004. “Confidence Limits for the Indirect Effect: Distribution of the Product and Resampling Methods,” Multivariate Behavioral Research (39:1), pp. 99-128.

Mathieu, J. E., and Zajac, D. M. 1990. “A Review and Meta- Analysis of the Antecedents, Correlates, and Consequences of Organizational Commitment,” Psychological Bulletin (108:2), pp. 171-194.

Meyer, J. P., and Allen, N. J. 1991. “A Three-Component Concep- tualization of Organizational Commitment,” Human Resource Management Review (1:1), pp. 61-89.

Meyer, J. P., and Herscovitch, L. 2001. “Commitment in the Work- place: Toward a General Model,” Human Resource Management Review (11:3), pp. 299-326.

Meyer, J. P., Paunonen, S. V., Gellatly, I. R., Goffin, R. D., and Jackson, D. N. 1989. “Organizational Commitment and Job Performance: It’s the Nature of the Commitment that Counts,” Journal of Applied Psychology (74:1), pp. 152-156.

Midha, V., and Bhattacherjee, A. 2012. “Governance Practices and Software Maintenance: A Study of Open Source Projects,” Decision Support Systems (54:1), pp. 23-32.

Miller, K. I., and Monge, P. R. 1985. “Social Information and Employee Anxiety about Organizational Change,” Human Communication Research (11:3), pp. 365-386.

Mockus, A., Fielding, R. T., and Herbsleb, J. D. 2002. “Two Case Studies of Open Source Software Development: Apache and Mozilla,” ACM Transactions on Software Engineering and Methodology (11:3), pp. 309-346.

Nafus, D. 2012. “‘Patches Don’t Have Gender’: What Is Not Open in Open Source Software,” New Media & Society (14:4), pp.= 669-683.

Nonnecke, B., and Preece, J. 2000. “Lurker Demographics: Counting the Silent,” in Proceedings of CHI 2000 Conference, The Hague, Netherlands, pp. 73-80.

974 MIS Quarterly Vol. 43 No. 3/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

O’Mahony, S., and Bechky, B. A. 2008. “Boundary Organizations: Enabling Collaboration Among Unexpected Allies,” Administra- tive Science Quarterly (53:2), pp. 422-459.

O’Mahony, S., and Ferraro, F. 2007. “The Emergence of Gover- nance in an Open Source Community,” Academy of Management Journal (50:5), pp. 1079-1106.

Paulson, J., Succi, G., and Eberlein, A. 2004. “An Empirical Study of Open-Source and Closed-Source Software Products,” IEEE Transactions on Software Engineering (30:4), pp. 246-256.

Pham, R., Singer, L., and Schneider, K. 2013. “Building Test Suites in Social Coding Sites by Leveraging Drive-By Commits,” in Proceedings of the International Conference on Software Engineering, San Francisco, CA, pp. 1209-1212.

Pinto, G., Steinmacher, I., and Aur, M. 2016. “More Common Than You Think: An In-Depth Study of Casual Contributors,” IEEE 23rd International Conference on Software Analysis, Evolution, and Reengineering, Osaka, Japan, pp. 112-123.

Podolny, J. M. 2001. “Networks as the Pipes and Prisms of the Market,” American Journal of Sociology (107:1), pp. 33-60.

Pollack, J. M., Vanepps, E. M., and Hayes, A. F. 2012. “The Moderating Role of Social Ties on Entrepreneurs’ Depressed Affect and Withdrawal Intentions in Response to Economic Stress,” Journal of Organizational Behavior (33:6), pp. 789-810.

Preacher, K. J., and Hayes, A .F. 2004. “SPSS and SAS Procedures for Estimating Indirect Effects in Simple Mediation Models,” Behavior Research Methods (36:4), pp. 717-731.

Preacher, K. J., Rucker, D. D., and Hayes, A. F. 2007. “Addressing Moderated Mediation Hypotheses: Theory, Methods, and Prescriptions,” Multivariate Behavioral Research (42:1), pp. 185-227.

Qureshi, I., and Fang, Y. 2010. “Socialization in Open Source Software Projects: A Growth Mixture Modeling Approach,” Organizational Research Methods (14:1), pp. 208-238.

Ramirez, A., Walther, J. B., Burgoon, J. K., and Sunnafrank, M. 2002. “Information-Seeking Strategies, Uncertainty, and Computer-Mediated Communication: Toward a Conceptual Model,” Human Communication Research (28: pp. 213-228.

Raymond, E. S. 2001. The cathedral and the Bazaar: Musing on Linux and Open Source by an Accidental Revolutionary, Sabastopol, CA: O’Reilly.

Rice, R. E., and Aydin, C. 1991. “Attitudes Toward New Organi- zational Technology: Network Proximity as a Mechanism for Social Information Processing,” Administrative Science Quar- terly (36:2), pp. 219-244.

Roberts, J. A., Hann, I.-H., and Slaughter, S. A. 2006. “Under- standing the Motivations, Participation, and Performance of Open Source Software Developers: A Longitudinal Study of the Apache Projects,” Management Science (52:7), pp. 984-1000.

Robles, G., and Gonzalez-Barahona, J. 2006. Contributor Turnover in Libre Software Projects, Boston: Springer.

Robles, G., Reina, L. A., González-Barahona, J. M., and Domínguez, S. D. 2016. “Women in Free/Libre/Open Source Software: The Situation in the 2010s,” IFIP International Conference on Open Source Systems: Integrating Communities, K. Crowston, I. Hammouda, B. Lundell, and G. Robles (eds.), New York: Springer, pp. 163-173.

Sarker, S., Ahuja, M., Sarker, S., and Kirkeby, S. 2011. “The Role of Communication and Trust in Global Virtual Teams: A Social Network Perspective,” Journal of Management Information Systems (28:1), pp. 273-310.

Schein, E. H. 1985. Organisational Culture and Leadership: A Dynamic View, San Francisco: Jossey-Bass.

Schweiger, D. M., and Denisi, A. S. 1991. “Communication with Employees Following a Merger: A Longitudinal Field Experi- ment,” Academy of Management Journal (34:1), pp. 110-135.

Seidel, M.-D., and Stewart, K. J. 2011. “An Initial Description of the C-Form,” Research in the Sociology of Organizations (33:November), pp. 37-72.

Setia, P., Rajagopalan, B., Sambamurthy, V., and Calantone, R. 2012. “How Peripheral Developers Contribute to Open-Source Software Development,” Information Systems Research (23:1), pp. 144-163.

Shah, S. 2006. “Motivation, Governance and the Viability of Hybrid Forms in Open Source Software Development,” Manage- ment Science (52:7), pp. 1000-1014.

Siciliano, M. D., and Thompson, J. R. 2015. “If You Are Com- mitted, Then So Am I: The Role of Social Networks and Social Influence on Organizational Commitment,” Administration & Society (50:7), pp. 916-946.

Singh, P., Tan, Y., and Mookerjee, V. 2011. “Network Effects: The Influence of Structural Social Capital on Open Source Project Success,” MIS Quarterly (35:4), pp. 813-829.

Smith, M. A. 1993. “Voices from the WELL: The Logic of the Virtual Commons,” unpublished paper, University of California, Los Angeles.

Solinger, O. N., Olffen, W. V., Roe, R. A., and Hofmans, J. 2013. “On Becoming (Un)Committed: A Taxonomy and Test of Newcomer Onboarding Scenarios,” Organization Science (24:6), pp. 1640-1661.

Spaeth, S., von Krogh, G., and He, F. 2015. “Research Note— Perceived Firm Attributes and Intrinsic Motivation in Sponsored Open Source Software Projects,” Information Systems Research (26:1), pp. 224-237.

Sparrowe, R., Liden, R. C., Wayne, S. J., and Kraimer, M. L. 2001. “Social Networks and the Performance of Individuals and Groups,” Academy of Management Journal (44:2), pp. 316-325.

Stallman. 1992. “Why Software Should Be Free,” GNU Operating System (http://www.gnu.org/philosophy/shouldbefree.html; accessed March 1, 2003)

Steers, R. M. 1977. “Antecedents and Outcomes of Organizational Commitment,” Administrative Science Quarterly (22:1), pp. 46-56.

Steinmacher, I., Conte, T., Gerosa, M. A., and Redmiles, D. 2015. “Social Barriers Faced by Newcomers Placing Their First Contribution in Open Source Software Projects,” in Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social Computing, Vancouver, BC, Canada, pp. 1379-1392.

Steinmacher, I., Conte, T. U., Treude, C., and Gerosa, M. A. 2016. “Overcoming Open Source Project Entry Barriers with a Portal for Newcomers,” in Proceedings of the 38th International Conference on Software Engineering, Austin, TX.

MIS Quarterly Vol. 43 No. 3/September 2019 975

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Stewart, K., Ammeter, A. P., and Maruping, L. M. 2006. “Impacts of License Choice and Organizational Sponsorship on User Interest and Development Activity in Open Source Software Projects,” Information Systems Research (17:2), pp. 126-145.

Stewart, K., and Gosain, S. 2006. “The Impact of Ideology on Effectiveness in Open Source Software Development Teams,” MIS Quarterly (30:2), pp. 291-314.

Stine, R. 1989. “An Introduction to Bootstrap Methods Examples and Ideas,” Sociological Methods and Research (18:2), pp. 243-291.

Trice, H. M., and Beyer, J. M. 1993. The Cultures of Work Organizations, Englewood Cliffs, NJ: Prentice Hall.

Tushman, M. L. 1979. “Impacts of Perceived Environmental Variability on Patterns of Work Related Communication,” Academy of Management Journal (22:3), pp. 482-500.

Tushman, M. L. 1979. “Work Characteristics and Subunit Com- munication Structure: A Contingency Analysis,” Administrative Science Quarterly pp. 82-98.

Vardaman, J. M., Taylor, S. G., Allen, D. G., Gondo, M. B., and Amis, J. M. 2015. “Translating Intentions to Behavior: The Interaction of Network Structure and Behavioral Intentions in Understanding Employee Turnover,” Organization Science (26:4), pp. 1177-1191.

Venkatesh, V., and Morris, M. G. 2002. “Why Don’t Men Ever Stop to Ask for Directions? Gender, Social Influence, and Their Role in Technology Acceptance and Usage Behavior,” MIS Quarterly (24:1), pp. 115-139.

Venkatesh, V., Morris, M. G., and Ackerman, P. L. 2000. “A Longitudinal Field Investigation of Gender Differences in Individual Technology Adoption Decision-Making Processes,” Organizational Behavior and Human Decision Processes (83:1), pp. 33-60.

Venkatesh, V., Thong, Y. Y. L., Chan, F., and Hu, P. J. 2016. “Managing Citizens’ Uncertainty in E-Government Services: The Mediating and Moderating Roles of Transparency and Trust,” Information Systems Research (27:1), pp. 87-111.

Von Krogh, G., Haefliger, S., Spaeth, S., and Wallin, M. W. 2012. “Carrots and Rainbows: Motivation and Social Practice in Open Source Software Development,” MIS Quarterly (36:2), pp. 649-676.

Wagstrom, P. 2009. “Vertical Communication in Open Software Engineering Communities,” CMU-ISR-09-103, Carnegie Insti- tute of Technology, Carnegie Mellon University, Pittsburgh, PA.

Weinberg, G. 1998. The Psychology of Computer Programming (Silver Anniversary Edition), New York: Van Nostrand Reinhold.

Williams, K. 2007. “Ostracism,” Annual Review of Psychology (58:1), pp. 425-452.

Xu, B., Jones, D. R., and Shao, B. 2009. “Volunteers’ Involvement in Online Community Based Software Development,” Informa- tion and Management (46:3), pp. 151-158.

Zhang, C., Hahn, J., and De, P. 2013. “Continued Participation in Online Innovation Communities: Does Community Response Matter Equally for Everyone?,” Information Systems Research (24:4), pp. 1112-1130.

Zhou, M., and Mockus, A. 2012. “What Makes Long Term Contributors: Willingness and Opportunity in OSS Community,” in Proceedings of the International Conference on Software Engineering, Zurich, Switzerland, pp. 518-528.

Zhou, M., and Mockus, A. 2015. “Who Will Stay in the FLOSS Community? Modeling Participant’s Initial Behavior,” IEEE Transactions on Software Engineering (41:1), pp. 82-99.

About the Authors

Likoebe M. Maruping is an associate professor of Computer Infor- mation Systems and a member of the Center for Process Innovation (CEPRIN) in the J. Mack Robinson College of Business at Georgia State University. His research is primarily focused on collaboration and innovation in small- and large-scale collectives such as teams, communities, and crowds. His interests in this area pertain to the enabling role of digital collaboration platforms, the mechanisms underlying the collaboration process, and the leadership and gover- nance of collaborative efforts in organizational and open environ- ments. Likoebe is currently an associate editor for Information Systems Research and a senior editor for Journal of the Association for Information Systems. He has previously served as an associate editor for MIS Quarterly. Likoebe is a recipient of MIS Quarterly’s “Reviewer of the Year” and “Outstanding Associate Editor of the Year” awards.

Sherae L. Daniel is an associate professor of Operations, Business Analytics and Information Systems in the Carl H. Lindner College of Business at the University of Cincinnati. She earned her Ph.D. in Information Systems from the Robert H. Smith School of Business at the University of Maryland. Sherae’s research seeks to reveal how to best manage collaboration challenges in nontraditional work environments. In particular, she seeks to uncover the keys that will unlock doors to future success for OSS collaborators. Sherae’s research has been published or is forthcoming in premier outlets such as Information Systems Research, MIS Quarterly, and Journal of Association for Information Systems, among others. She is a member of the Association for Information Systems.

Marcelo Cataldo is currently a software engineering manager at Google. He received a BS degree in information systems from the Universidad Tecnologica Nacional, Argentina in 1996 and MS and Ph.D. degrees in societal computing from Carnegie Mellon Univer- sity in 2007. His research interests include geographically distri- buted software development with special focus on the relationship between the software architecture and the organizational structure in large-scale software development projects. Marcelo’s research has been published in premier software engineering and HCI venues and has been recognized with several awards including Alan Newell Award for Research Excellence (2014), ACM Distinguished Paper Award (ICSE 2011), ACM Distinguished Paper Award (ESEM 2008), and Best Paper Award (CSCW 2006).

976 MIS Quarterly Vol. 43 No. 3/September 2019

RESEARCH ARTICLE

DEVELOPER CENTRALITY AND THE IMPACT OF VALUE CONGRUENCE AND INCONGRUENCE ON COMMITMENT

AND CODE CONTRIBUTION ACTIVITY IN OPEN SOURCE SOFTWARE COMMUNITIES1

Likoebe M. Maruping J. Mack Robinson College of Business, Georgia State University,

Atlanta, GA 30303 U.S.A. {lmaruping@gmail.com}

Sherae L. Daniel Carl H. Linder College of Business, University of Cincinnati,

Cincinnati, OH 45221 U.S.A. {sherae.daniel@uc.edu}

Marcelo Cataldo Google, New York, NY 10011 U.S.A. {chelo.cataldo@gmail.com}

Appendix A Summary of Empirical Research on Impact of OSS Values on Developers

A growing body of empirical research has examined the role of OSS values in affecting developers (Benbya and Belbaly 2010; Chou and He 2011; Stewart and Gosain 2006). With few exceptions, the research to date has typically focused on either the community (or OSS team) OSS values or the developer OSS values. In their seminal study, Stewart and Gosain (2006) find that when OSS teams embrace OSS values, there is a positive influence on communication quality and affective trust in OSS teams. Surprisingly, they also find that embracing OSS values negatively influences task completion. They explain their findings by suggesting that teams who embrace OSS values that are geared toward collaboration tend to prioritize consensus over completing tasks. Chou and He (2011) find that a team embracing OSS values positively impacts its collaborative elaboration, and communication decoding and encoding competence. Taken together, this research suggests that the community (or team) OSS values can influence team attitudes and activity level. A looming open issue in this research is that researchers do not consider the role of the individual developer OSS values. As such, it is possible that a developer may not necessarily support the OSS values in the same way as the community.

Other researchers focus on the individual developer OSS values as a determinant. For instance, researchers find that a developer who embraces such values is likely to expend a greater amount of time and effort contributing to OSS initiatives (Benbya and Belbaly 2010) and is likely to report being more involved in OSS communities (Xu et al. 2009). In contrast to these findings, Henkel (2008) finds that OSS values play no role in affecting developers’ contribution to OSS communities. In sum, this stream of research suggests that the developer OSS values can influence developer contribution to OSS communities. However, this research overlooks the role of the OSS community’s values in affecting developer attitudes and contribution behavior. As such, it provides an incomplete picture of the role of OSS values in shaping developers’ attitudes and behavior in OSS communities.

MIS Quarterly Vol. 43 No. 3–Appendices/September 2019 A1

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Table A1. Summary of Empirical Research Findings on OSS Values

Referent Article

Focus of OSS Values Key FindingDeveloper OSS Community

Xu et al. 2009 % + impact on developer involvement Henkel 2008 % n.s. developer contribution Benbya and Belbaly 2010 % + impact on developer participation type

+ impact on developer effort

Ke and Zhang 2009 % % – developer goal commitment n.s. developer effort intensity

Ke and Zhang 2010 % % – developer task effort Chow and He 2011

% + collaborative elaboration + communication decoding competence + communication encoding competence

Stewart and Gosain 2006 %

+ impact on communication quality + impact on affective trust – task completion

+ denotes values antecedent has positive impact on listed outcome, – denotes values antecedent has negative impact on listed outcome, n.s. denotes values antecedent has nonsignificant impact on listed outcome.

Appendix B Results of Confirmatory Factor Analysis

Factors

Items 1 2 3

As a software developer… Developer values1: I value sharing knowledge. .32 .01 .24

Developer values2: I believe in helping others. .76 .17 .06

Developer values3: I place great value on technical knowledge. .70 .23 .13

Developer values4: I am driven by a desire to learn new things. .70 .03 .05

Developer values5: I think cooperation is important. .75 .08 .11

Developer values6: I value the reputation I gain by participating in open source. projects. .67 .17 .09

In my view, members of this OSS community… Perception of OSS values1: value sharing knowledge. .13 .24 .28

Perception of OSS values2: believe in helping others. .20 .82 .22

Perception of OSS values3: place great value on technical knowledge. .13 .83 .17

Perception of OSS values4: are driven by a desire to learn new things. .17 .69 .22

Perception of OSS values5: think cooperation is important. .18 .73 .18

Perception of OSS values6: value the reputation gained by participating in open source projects. .08 .78 .23

Commitment1: I am willing to put in effort beyond the norm for the success of GNOME. .15 .18 .79

Commitment2: For me, this is the best of all possible OSS projects for which to work. .10 .21 .85

Commitment3: I am extremely glad to have chosen GNOME to work for over other projects. .11 .23 .86

Commitment4: GNOME inspires me to be my best technical work. .09 .30 .79

Commitment5: I show by my actions that I really care about the fate of GNOME. .10 .30 .77

Note: Items with italicized loadings were dropped.

A2 MIS Quarterly Vol. 43 No. 3–Appendices/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Appendix C Coefficients for Response Surface Analysis Predicting OSS Commitment at Low Versus High Developer Centrality

Variable Coefficient Low Developer

Centrality High Developer

Centrality

(Intercept) b0 4.98*** 5.06***

Developer values b1 .13 .17*

OSS values b2 .60*** .52***

Developer values-squared b3 .13 .13

Developer values × OSS values b4 -.11* -.07

OSS values-squared b5 .17 .25**

Notes: Significance levels are based on bias-corrected confidence intervals generated from bootstrap estimates. Analysis includes project fixed effects. *p< .05, **p < .01, ***p < .001.

Appendix D Results of Moderation–Mediation Analysis

Developer Contribution Activity

OSS Commitment

Number of Commits

Lines of Code Added/deleted

Number of Files Changed

(Intercept) 5.42*** -4.99*** -4.28*** -4.35***

Age -.01 -.02 -.03 -.02

Gender .35 2.42* 3.76* 2.33*

Education -.14* -.24 -.57 -.36

Volunteer status -.08 -1.04* -1.42* -.94*

OSS Value (in)congruence .69*** .41 .67 .41

Developer centrality .46*

OSS commitment .71*** .90** .58**

Value congruence × developer centrality -.52*

Project fixed effects Yes Yes Yes Yes

R² .46*** .30*** .29*** .28***

Notes: n = 410 1. Value congruence is computed from block variable using the predicted value from the polynomial regression equations. 2. Pattern of results is similar when using different measures of developer centrality. *p < .05, **p < .01, ***p < .001.

MIS Quarterly Vol. 43 No. 3–Appendices/September 2019 A3

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Appendix E

Robustness Analysis Regarding Degree Centrality in Communication Network

In order to ensure that our results were not an artifact of the timeframe upon which our degree centrality measure was based, we conducted the moderated polynomial regression analysis using degree centrality based on the past 24 months of email communication. The pattern of results was consistent across this alternative operationalization as shown in Table E1. Further, in order to ensure that the results were not an artifact of the use of email communication as the basis for computing degree centrality, we computed degree centrality based on developers’ collaboration on the same projects. We examined degree centrality based on collaboration on the same projects in the past 12 months and the past 24 months. As the results show, the pattern of results was similar to that of the analysis using degree centrality based on email communication.

Table E1. Results of Polynomial Regression Analysis Using Alternative Measures of Developer Centrality

Variable Coefficient 1

2 Centrality: 12-month

email

3 Centrality: 24-month

email

4 Centrality: 12-month

project

5 Centrality: 24-month

project

Age -.04 -.04 -.03 -.02 .00

Gender .03 .03 .04 .03 .02

Education -.10* -.10* -.10* -.10* -.10*

Volunteer status -.05 -.05 -.06 -.03 -.06

Developer values b1 .16* .15* .18* .19** .20**

OSS values b2 .50*** .56*** .53*** .41*** .43***

Developer values-squared b3 .13* .13* .14* .15* .15*

Developer values × OSS values b4 -.14* -.09 † -.10* -.14* -.15*

OSS values-squared b5 .15* .21** .11* .10* .13*

Developer centrality b6 .15** .18* .30*** .31***

Developer values × developer centrality b7 .07 .08 .06 .07

OSS values × developer centrality b8 -.16** -.21** -.12* -.19*

Developer values-squared × developer centrality

b9 .00 .03 .06 .03

Developer values × OSS values × developer centrality

b10 .07 .13 .01 -.01

OSS values-squared × developer centrality b11 .16** .14* .10* .16*

Project fixed effects Yes Yes Yes Yes Yes

Adjusted R² .30*** .33*** .33*** .34*** .36***

∆R² .03** .03** .04*** .06***

Notes: n = 410. 1. Developer centrality in model 2 is measured on email communication activity over prior 12 months, developer centrality in

model 3 is measured on email communication activity over prior 24 months, developer centrality in model 4 is measured on project collaboration activity over prior 12 months, developer centrality in model 5 is measured on project collaboration activity over prior 24 months.

1. ∆R² for models 2 through 5 represent change in variance explained over and above model 1 (i.e., the polynomial regression model without moderation).

†p < .10, *p < .05, **p < .01, ***p < .001.

A4 MIS Quarterly Vol. 43 No. 3–Appendices/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Appendix F

Robustness Tests for Endogeneity

An argument could be made that developers who have greater commitment to the OSS community are more likely to be highly central in the communication network.1 Empirically, our study design does not lend itself to testing the influence of commitment on centrality since our data were time lagged. That is, the developer centrality variable is based on email communication before commitment was measured. However, we acknowledge that this does not preclude the possibility of there being endogeneity between the two variables. Consequently, we estimated the polynomial regression models using an instrumental variables two-stage least squares (2SLS) analysis. We identified an instrumental variable that was correlated with commitment but uncorrelated with the error term. As shown in Table F1, the pattern of results from the 2SLS are similar to those of our main analysis. This provides additional confidence in the robustness of the model specification and results.

Table F1. Results of Instrumental Variables Two-Stage Least Squares Analysis with Robust Standard Errors

OSS Commitment

Variable 1 2 3

Age .00 (.00) .00 (.01) .00 (.00)

Gender .06 (.18) .03 (.17) .12 (.19)

Education -.10* (.05) -.10† (.06) -.14* (.06)

Volunteer status .09 (.12) .07 (.12) .07 (.13)

Developer values b1 1.71* (.73) 2.15* (1.00) 2.40* (1.05)

OSS community values b2 2.36*** (.25) 2.62*** (.45) 2.06*** (.51)

Developer values-squared b3 8.91** (2.66) 5.94* (2.93)

Developer values × OSS community values b4 -9.09* (3.83) -7.60 † (4.02)

OSS community values-squared b5 .86** (.28) 1.21* (.60)

Developer centrality b6 2.59** (.91)

Developer values × developer centrality b7 .06 (.08)

OSS community values × developer centrality b8 -1.41** (.52)

Developer values-squared × developer centrality b9 .00 (.05)

Developer values × OSS community values × developer centrality

b10 4.49 (3.66)

OSS community values-squared × developer centrality b11 2.49** (1.05)

Wald χ² 150.20*** 212.67*** 220.292*** Adjusted R² .28 .30 .32

Notes: n = 410; standard errors are in parentheses; country, number of commits made (pre-survey), total number of messages posted to listserv (pre-survey), total number of projects (pre-survey), and number of replies posted to listserv (pre-survey) were used as instruments. †p < .10, *p < .05, **p < .01, ***p < .001.

1We thank an anonymous reviewer for drawing our attention to this possibility.

MIS Quarterly Vol. 43 No. 3–Appendices/September 2019 A5

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Appendix G

Robustness Analysis Regarding Operationalization of Dependent Variable

We wanted to ensure that the results of the mediation analysis were not an artifact of the specific operationalization of developer contribution activity as number of commits. Therefore, we repeated the analysis using two alternative operationalizations: number of lines of code added/ deleted and number of files changed. As the results in Table G1 show, the mediating role of commitment is stronger among less central developers than among highly central developers in predicting number of lines of code added and deleted (test of differences: .75 – .50 = .25, p < .05) and number of files changed (test of differences: .48 – .32 = .16, p < .05). Specifically, we find that the mediating effect of commitment is nonsignificant among highly central developers and is significant among less central developers, providing support for H4. Number of commits may be a less granular measure of the volume of work compared to the actual lines of code and files changed.

Table G1. Results of Mediation (H3) and Moderated-Mediation (H4) Tests on Number of Lines of Code Added and Deleted and Number of Files Changed

Level of Developer Centrality Direct Effect Indirect Effect

95% Bias-Corrected CI

R² Dependent Variable Lower Bound

Upper Bound

.68 (.46) .39 (.20) .04 .86 .23*** Number of lines of code added/deletedHigh .68 (.49) .50 (.31) -.01 .97 .29***

Low .75 (.27) .29 .99

.46 (.30) .33 (.14) .07 .64 .19*** Number of files changed

High .41 (.32) .32 (.20) -.01 .62 .28***

Low .48 (.18) .18 .89

Notes: 1. Value congruence and incongruence block variable is the independent variable, developer commitment is the mediator,

developer centrality is the moderator; control variables were included as covariates in the analysis. 2. Standard errors are in parentheses. 3. Dependent variables were log-transformed prior to analysis. 4. The pattern of results is the same when using degree centrality based on email communication in past 24 months,

collaboration activity on projects in past 12 months and past 24 months. ***p < .001.

A6 MIS Quarterly Vol. 43 No. 3–Appendices/September 2019

Maruping et al./Commitment & Code Contribution Activity in OSS Communities

Appendix H Results of Model Estimation Using Heckman Two-Step Selection Procedure

Variable Commitment

Age .00 (.01)

Gender .16 (.26)

Education -.14* (.06)

Volunteer status .02 (.10)

Developer values b1 2.29* (1.01)

OSS community values b2 2.60*** (.40)

Developer values-squared b3 6.99* (3.41)

Developer values × OSS community values b4 -7.45 † (3.93)

OSS community values-squared b5 1.62* (.60)

Developer centrality b6 .28*** (.06)

Developer values × developer centrality b7 .93 (.66)

OSS community values × developer centrality b8 -.99** (.45)

Developer values-squared × developer centrality b9 .01 (.05)

Developer values × OSS community values × developer centrality b10 3.97 (3.61)

OSS community values-squared × developer centrality b11 .78* (.35)

Inverse Mills ratio .10 (.52)

Wald χ² 234.98***

Notes: n = 410; standard errors are in parentheses; developer degree centrality in the communication network (pre-survey), number of commits (pre-survey), number of different projects (pre-survey), and number of messages and replies posted to the listserv (pre-survey) were used as determinants in the estimation of the first stage probit model. †p < .10, *p < .05, **p < .01, ***p < .001

References

Benbya, H., and Belbaly, N. 2010. “Understanding Developers’ Motives in Open Source Projects: A Multi-Theoretical Framework,” Communications of the AIS (27:1), pp. 589-610.

Chou, S.-W., and He, M.-Y. 2011. “Understanding OSS Development in Communities: The Perspectives of Ideology and Knowledge Sharing,” Behaviour & Information Technology (30:3), pp. 325-337.

Henkel, J. 2008. “Champions of Revealing—The Role of Open Source Developers in Commercial Firms,” Industrial and Corporate Change (18:3), pp. 435-471.

Ke, W., and Zhang, P. 2009. “Motivations in Open Source Software Communities: The Mediating Role of Effort Intensity and Goal Commitment,” International Journal of Electronic Commerce (13:4), pp. 39-66.

Ke, W., and Zhang, P. 2010. “The Effects of Extrinsic Motivations and Satisfaction in Open Source Software Development,” Journal of the Association for Information Systems (11:12), pp. 784-808.

Stewart, K., and Gosain, S. 2006. “The Impact of Ideology on Effectiveness in Open Source Software Development Teams,” MIS Quarterly (30:2), pp. 291-314.

Xu, B., Jones, D. R., and Shao, B. 2009. “Volunteers’ Involvement in Online Community Based Software Development,” Information and Management (46:3), pp. 151-158.

MIS Quarterly Vol. 43 No. 3–Appendices/September 2019 A7

Copyright of MIS Quarterly is the property of MIS Quarterly and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.